Professional Documents
Culture Documents
Dynamic Analysis
User’s Guide
Corporate Europe Asia Pacific
MSC.Software Corporation MSC.Software GmbH Asia Pacific
2 MacArthur Place Am Moosfeld 13 MSC.Software Japan Ltd.
Santa Ana, CA 92707 81829 Munich Shinjuku First West 8F
Telephone: (800) 345-2078 GERMANY 23-7 Nishi Shinjuku
FAX: (714) 784-4056 Telephone: (49) (89) 43 19 87 0 1-Chome, Shinjuku-Ku
Fax: (49) (89) 43 61 71 6 Tokyo 160-0023, JAPAN
Telephone: 0120-924-832 (toll
free, Japan only)
Mobile phone: 03-6911-1222
Fax: (81) (3)-6911-1201
Worldwide Web
www.mscsoftware.com
Disclaimer
MSC.Software Corporation reserves the right to make changes in specifications and other information contained
in this document without prior notice.
The concepts, methods, and examples presented in this text are for illustrative and educational purposes only,
and are not intended to be exhaustive or to apply to any particular engineering problem or design. MSC.Software
Corporation assumes no liability or responsibility to any person or company for direct or indirect damages resulting
from the use of any information contained herein.
User Documentation: Copyright 2010 MSC.Software Corporation. Printed in U.S.A. All Rights Reserved.
This notice shall be marked on any reproduction of this documentation, in whole or in part. Any reproduction or
distribution of this document, in whole or in part, without the prior written consent of MSC.Software Corporation is
prohibited.
This software may contain certain third-party software that is protected by copyright and licensed from
MSC.Software suppliers.
MSC, MD, Dytran, Marc, MSC Nastran, MD Nastran, MSC Patran, MD Patran, OpenFSI, the MSC.Software
corporate logo, and Simulating Reality are trademarks or registered trademarks of the MSC.Software Corporation
in the United States and/or other countries.
NASTRAN is a registered trademark of NASA. PAMCRASH is a trademark or registered trademark of ESI Group.
SAMCEF is a trademark or registered trademark of Samtech SA. LS-DYNA is a trademark or registered trademark
of Livermore Software Technology Corporation. ANSYS is a registered trademark of SAS IP, Inc., a wholly owned
subsidiary of ANSYS Inc. ABAQUS is a registered trademark of ABAQUS Inc. All other brand names, product
names or trademarks belong to their respective owners. PCGLSS 6.0, Copyright © 1992-2005, Computational
Applications and System Integration Inc. All rights reserved. PCGLSS 6.0 is licensed from Computational
Applications and System Integration Inc. METIS is copyrighted by the regents of the University of Minnesota. A
copy of the METIS product documentation is included with this installation. Please see "A Fast and High Quality
Multilevel Scheme for Partitioning Irregular Graphs". George Karypis and Vipin Kumar. SIAM Journal on Scientific
Computing, Vol. 20, No. 1, pp. 359-392, 1999.
Main Index
Contents
MD Nastran Dynamic Analysis User’s Guide
MD Nastran
Dynamic Analysis
Preface User’s Guide
Acknowledgements xiii
Technical Support xv
Equations of Motion 3
Mass Input 21
Damping Input 35
Main Index
iv MD Nastran Dynamic Analysis User’s Guide
Methods of Computation 53
Examples 70
Main Index
CONTENTS v
Examples 184
Examples 236
6 Damping
Overview 260
Main Index
vi MD Nastran Dynamic Analysis User’s Guide
7 Enforced Motion
Overview 282
Examples 311
8 Random Analysis
Random Analysis with Coupled Excitations 326
Examples 379
Remarks 381
Main Index
CONTENTS vii
Units 292
Mass 293
Damping 294
Main Index
viii MD Nastran Dynamic Analysis User’s Guide
Loads 299
Meshing 300
Groundcheck 317
15 Results Processing
Overview 338
Main Index
CONTENTS ix
16 Special Topics
Direct Matrix Input Dynamics 360
OPENFSI 463
Main Index
x MD Nastran Dynamic Analysis User’s Guide
Bibliography 511
Main Index
MD Nastran Dynamic Analysis User’s Guide
Preface
About this Book
Acknowledgements
List of Nastran Books
Technical Support
Internet Resources
Main Index
xii MD Nastran Dynamic Analysis User’s Guide
About this Book
This guide is an update to the MD Nastran Basic Dynamic Analysis User’s Guide for Version 68, which
borrowed much material from the MSC.Nastran Handbook for Dynamic Analysis. However, not all
topics covered in that handbook are covered here. In addition, dynamic reduction, response spectrum
analysis, random response analysis, complex eigenvalue analysis, nonlinear analysis, fluid-structure
coupling and Enforced Motion and other topics which were part of the MSC.Nastran Advanced
Dynamics Guide for Version 70 and are now included in this guide. In addition, Release Guide material
related to dynamics has also been included in this guide.
This guide contains many highlighted links (in blue) to other MD Nastran documents and all the
documents were delivered together as a collection. If you keep the collection together the links between
documents will work. Two suggestions when working with links are 1) “alt “ returns you back in the
window your mouse is in and 2) you can open the other “linked to” document in a new window from an
Adobe Reader by choosing Preferences Documents Open cross-document links in the same
window; then you would uncheck the and select “OK”.
Main Index
CHAPTER xiii
Preface
Acknowledgements
The 2010 Version of the MD Nastran Dynamic Analysis User’s Guide is part of an ongoing project to
update existing MD Nastran documentation. This guide is a combination of the Basic Dynamics User’s
Guide and Advanced Dynamics User’s Guide. The primary editors for previous versions of Dynamics
Guides include Dr. Richard MacNeal (Version 61), Mr. Michael Gockel (Version 64), Mr. John Caffrey,
Mr. John Lee, and Mr. Grant Sitton (Version 68), Mr. John Lee (Versions 69, 2001 and 2003) and Dave
Herting (Version 70).
This guide incorporates most capabilities related to dynamic analysis into one place. The goal was to
incorporate the two existing dynamcis guides and the material presented in all the Release Guides since
Version 70 into one place and to connect the MD documents together by providing direct links between
manuals. This eliminates the older Release Guides and provides quick and directed access to other
documents like the MD Nastran Quick Reference Guide (QRG).
The editor is grateful to Ms. Lori Lampert for her patience and dedication in updating this user’s guide.
The editor would also like to thank Mr. Mohan Barbela, Mr. Dean Bellinger, Mr. Don Graff, Mr. John
Lee, Mr. Joe Maronick, Mr. Martin McNamee, Mr. Lance Proctor, Mr. Jim Swan, and Mr. Charlie
Wilson for their technical review of this guide.
Donald M. McLean
June 2010
Main Index
xiv MD Nastran Dynamic Analysis User’s Guide
List of Nastran Books
User’s Guides
• Getting Started
• Linear Static Analysis
• Dynamic Analysis
• MD Demonstration Problems
• Thermal Analysis
• Superelements
• Design Sensitivity and Optimization
• Implicit Nonlinear (SOL 600)
• Explicit Nonlinear (SOL 700)
• Aeroelastic Analysis
• User Defined Services
• EFEA User’s Guide
• EFEA Tutorial
• EBEA User’s Guide
Main Index
CHAPTER xv
Preface
Technical Support
For technical support phone numbers and contact information, please visit:
http://www.mscsoftware.com/Contents/Services/Technical-Support/Contact-Technical-Support.aspx
Support Center (www.simcompanion.mscsoftware.com)
Support Online. The Support Center provides technical articles, frequently asked questions and
documentation from a single location.
Main Index
xvi MD Nastran Dynamic Analysis User’s Guide
Internet Resources
Internet Resources
MSC.Software (www.mscsoftware.com)
MSC.Software corporate site with information on the latest events, products and services for the
CAD/CAE/CAM marketplace.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 1: Fundamentals of Dynamic Analysis
Overview
Equations of Motion
Dynamic Analysis Process
Dynamic Analysis Types
Main Index
2 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
In static structural analysis, it is possible to describe the operation of MD Nastran without a detailed
discussion of the fundamental equations. Due to the several types of dynamic analyses and the different
mathematical form of each, some knowledge of both the physics of dynamics and the manner in which
the physics is represented is important to using MD Nastran effectively and efficiently for dynamic
analysis.
You should become familiar with the notation and terminology covered in this chapter. This knowledge
will be valuable to understand the meaning of the symbols and the reasons for the procedures employed
in later chapters. References and Bibliography, 507 provides a list of references for structural dynamic
analysis.
Main Index
CHAPTER 1 3
Fundamentals of Dynamic Analysis
Equations of Motion
The basic types of motion in a dynamic system are displacement u and the first and second derivatives
of displacement with respect to time. These derivatives are velocity and acceleration, respectively, given
below:
du
u· = ------ = v = velocity
dt
(1-1)
2
d u- = dv
u·· = -------- ------ = a = acceleration
dt
2 dt
pt
m = mass (inertia) ut
b = damping (energy dissipation
k = stiffness (restoring force)
m
p = applied force
u = displacement of mass
u· = velocity of mass k b
u·· = acceleration of mass
Main Index
4 MD Nastran Dynamic Analysis User’s Guide
Equations of Motion
The four basic components of a dynamic system are mass, energy dissipation (damper), resistance
(spring), and applied load. As the structure moves in response to an applied load, forces are induced that
are a function of both the applied load and the motion in the individual components. The equilibrium
equation representing the dynamic motion of the system is known as the equation of motion.
Equation of Motion
This equation, which defines the equilibrium condition of the system at each point in time, is represented
as
Inertia Force
An accelerated mass induces a force that is proportional to the mass and the acceleration. This force is
called the inertia force mu·· t .
Viscous Damping
The energy dissipation mechanism induces a force that is a function of a dissipation constant and the
velocity. This force is known as the viscous damping force bu· t . The damping force transforms the
kinetic energy into another form of energy, typically heat, which tends to reduce the vibration.
Elastic Force
The final induced force in the dynamic system is due to the elastic resistance in the system and is a
function of the displacement and stiffness of the system. This force is called the elastic force or
occasionally the spring force ku t .
Applied Load
The applied load p t on the right-hand side of Eq. (1-2) is defined as a function of time. This load is
independent of the structure to which it is applied (e.g., an earthquake is the same earthquake whether it
is applied to a house, office building, or bridge), yet its effect on different structures can be very different.
Main Index
CHAPTER 1 5
Fundamentals of Dynamic Analysis
Dynamic analysis can be divided into two basic classifications: free vibrations and forced vibrations.
Free vibration analysis is used to determine the basic dynamic characteristics of the system with the
right-hand side of Eq. (1-2) set to zero (i.e., no applied load). If damping is neglected, the solution is
known as undamped free vibration analysis.
mu·· t + ku t = 0 (1-3)
Eq. (1-3) has a solution of the form
The quantity u t is the solution for the displacement as a function of time t . As shown in Eq. (1-4),
the response is cyclic in nature.
k-
n = --- (1-5)
m
The circular natural frequency is specified in units of radians per unit time.
Natural Frequency
The natural frequency f n is defined by
f n = ------n (1-6)
2
The natural frequency is often specified in terms of cycles per unit time, commonly cycles per second
(cps), which is more commonly known as Hertz (Hz). This characteristic indicates the number of sine
or cosine response waves that occur in a given time period (typically one second).
The reciprocal of the natural frequency is termed the period of response T n given by
1 2-
T n = ---- = ----- (1-7)
fn n
The period of the response defines the length of time needed to complete one full cycle of response.
Main Index
6 MD Nastran Dynamic Analysis User’s Guide
Equations of Motion
In the solution of Eq. (1-4), A and B are the integration constants. These constants are determined by
considering the initial conditions in the system. Since the initial displacement of the system u t = 0
and the initial velocity of the system u· t = 0 are known, A and B are evaluated by substituting their
values into the solution of the equation for displacement and its first derivative (velocity), resulting in
u· t = 0
B = u t = 0 and A = ---------------------
n
(1-8)
These initial value constants are substituted into the solution, resulting in
u· 0
u t = ----------- sin n t + u 0 cos n t (1-9)
n
Eq. (1-9) is the solution for the free vibration of an undamped SDOF system as a function of its initial
displacement and velocity. Graphically, the response of an undamped SDOF system is a sinusoidal wave
whose position in time is determined by its initial displacement and velocity as shown in Figure 1-2.
Amplitude u t
Time t
If damping is included, the damped free vibration problem is solved. If viscous damping is assumed, the
equation of motion becomes
Damping Types
The solution form in this case is more involved because the amount of damping determines the form of
the solution. The three possible cases for positive values of b are
Main Index
CHAPTER 1 7
Fundamentals of Dynamic Analysis
• Critically damped
• Overdamped
• Underdamped
Critical damping occurs when the value of damping is equal to a term called critical damping b cr . The
critical damping is defined as
b cr = 2 km = 2m n (1-11)
Again, A and B are the constants of integration based on the initial conditions of the system. The new
term d represents the damped circular natural frequency of the system. This term is related to the
undamped circular natural frequency by the following expression:
2
d = n 1 – (1-14)
b
= ------- (1-15)
b cr
The damping ratio is commonly used to specify the amount of damping as a percentage of the critical
damping.
In the underdamped case, the amplitude of the vibration reduces from one cycle to the next following an
exponentially decaying envelope. This behavior is shown in Figure 1-3. The amplitude change from one
cycle to the next is a direct function of the damping. Vibration is more quickly dissipated in systems with
more damping.
Main Index
8 MD Nastran Dynamic Analysis User’s Guide
Equations of Motion
Amplitude u t
Time t
The damping discussion may indicate that all structures with damping require damped free vibration
analysis. In fact, most structures have critical damping values in the 0 to 10% range, with values of 1 to
5% as the typical range. If you assume 10% critical damping, Eq. (1-4) indicates that the damped and
undamped natural frequencies are nearly identical. This result is significant because it avoids the
computation of damped natural frequencies, which can involve a considerable computational effort for
most practical problems. Therefore, solutions for undamped natural frequencies are most commonly
used to determine the dynamic characteristics of the system (see Real Eigenvalue Analysis, 43).
However, this does not imply that damping is neglected in dynamic response analysis. Damping can be
included in other phases of the analysis as presented later for frequency and transient response (see
Frequency Response Analysis, 133 and Transient Response Analysis, 201).
In this equation the circular frequency of the applied loading is denoted by . This loading frequency
is entirely independent of the structural natural frequency n , although similar notation is used.
Main Index
CHAPTER 1 9
Fundamentals of Dynamic Analysis
pk (1-17)
u t = A sin n t + B cos n t + ---------------------------- sin t
2 2
1 – n
B = ut = 0
Again, A and B are the constants of integration based on the initial conditions. The third term in
Eq. (1-17) is the steady-state solution. This portion of the solution is a function of the applied loading
and the ratio of the frequency of the applied loading to the natural frequency of the structure.
The numerator and denominator of the third term demonstrate the importance of the relationship of the
structural characteristics to the response. The numerator p k is the static displacement of the system.
In other words, if the amplitude of the sinusoidal loading is applied as a static load, the resulting static
displacement u is p k . In addition, to obtain the steady state solution, the static displacement is scaled
by the denominator.
The denominator of the steady-state solution contains the ratio between the applied loading frequency
and the natural frequency of the structure.
1
---------------------------- (1-18)
2 2
1 – n
is called the dynamic amplification (load) factor. This term scales the static response to create an
amplitude for the steady state component of response. The response occurs at the same frequency as the
loading and in phase with the load (i.e., the peak displacement occurs at the time of peak loading). As
the applied loading frequency becomes approximately equal to the structural natural frequency, the ratio
n approaches unity and the denominator goes to zero. Numerically, this condition results in an
infinite (or undefined) dynamic amplification factor. Physically, as this condition is reached, the
dynamic response is strongly amplified relative to the static response. This condition is known as
resonance. The resonant buildup of response is shown in Figure 1-4.
Main Index
10 MD Nastran Dynamic Analysis User’s Guide
Equations of Motion
Amplitude u t
Time t
Figure 1-4 Harmonic Forced Response with No Damping
It is important to remember that resonant response is a function of the natural frequency and the loading
frequency. Resonant response can damage and even destroy structures. The dynamic analyst is typically
assigned the responsibility to ensure that a resonance condition is controlled or does not occur.
Solving the same basic harmonically loaded system with damping makes the numerical solution more
complicated but limits resonant behavior. With damping, the equation of motion becomes
Main Index
CHAPTER 1 11
Fundamentals of Dynamic Analysis
In this case, the effect of the initial conditions decays rapidly and may be ignored in the solution. The
solution for the steady-state response is
sin t +
u t = p k ------------------------------------------------------------------------------ (1-20)
2 2 2 2
1 – n + 2 n
The numerator of the above solution contains a term that represents the phasing of the displacement
response with respect to the applied loading. In the presence of damping, the peak loading and peak
response do not occur at the same time. Instead, the loading and response are separated by an interval
of time measured in terms of a phase angle as shown below:
– 1 2 n
= – tan --------------------------------- (1-21)
2 2
1 – n
The phase angle is called the phase lead, which describes the amount that the response leads the
applied force.
Note: Some texts define as the phase lag, or the amount that the response lags the applied force.
To convert from phase lag to phase lead, change the sign of in Eq. (1-20) and Eq. (1-21).
1
-----------------------------------------------------------------------------
- (1-22)
2 2 2 2
1 – n + 2 n
The interrelationship among the natural frequency, the applied load frequency, and the phase angle can
be used to identify important dynamic characteristics. If n is much less than 1, the dynamic
amplification factor approaches 1 and a static solution is represented with the displacement response in
phase with the loading. If n is much greater than 1, the dynamic amplification factor approaches
zero, yielding very little displacement response. In this case, the structure does not respond to the
loading because the loading is changing too fast for the structure to respond. In addition, any measurable
displacement response will be 180 degrees out of phase with the loading (i.e., the displacement response
will have the opposite sign from the force). Finally if n = 1 , resonance occurs. In this case, the
magnification factor is 1 2 , and the phase angle is 270 degrees. The dynamic amplification factor
and phase lead are shown in Figure 1-5 and are plotted as functions of forcing frequency.
Main Index
12 MD Nastran Dynamic Analysis User’s Guide
Equations of Motion
Amplification Factor
360
Phase Lead (Degrees)
180
n
Forcing Frequency
Figure 1-5 Harmonic Forced Response with Damping
In contrast to harmonic loadings, the more general forms of loading (impulses and general transient
loading) require a numerical approach to solving the equations of motion. This technique, known as
numerical integration, is applied to dynamic solutions either with or without damping. Numerical
integration is described in Transient Response Analysis, 201.
Main Index
CHAPTER 1 13
Fundamentals of Dynamic Analysis
Dynamic
Environment
Finite Element
Model
Modal Yes No
Results
Analysis? Satisfactory?
No
Yes
No Results Forced-Response
Satisfactory? Analysis
Yes
End
An overall system design is formulated by considering the dynamic environment. As part of the
evaluation process, a finite element model is created. This model should take into account the
characteristics of the system design; and just as importantly, the nature of the dynamic loading (type and
frequency); and any interacting media (fluids, adjacent structures, etc.). At this point, the first step in
many dynamic analyses is a modal analysis to determine the structure’s natural frequencies and mode
shapes (see Real Eigenvalue Analysis, 43).
Main Index
14 MD Nastran Dynamic Analysis User’s Guide
Dynamic Analysis Process
In many cases the natural frequencies and mode shapes of a structure provide enough information to
make design decisions. For example, in designing the supporting structure for a rotating fan, the design
requirements may require that the natural frequency of the supporting structure have a natural frequency
either less than 85% or greater than 110% of the operating speed of the fan. Specific knowledge of
quantities such as displacements and stresses are not required to evaluate the design.
Forced response is the next step in the dynamic evaluation process. The solution process reflects the
nature of the applied dynamic loading. A structure can be subjected to a number of different dynamic
loads with each dictating a particular solution approach. The results of a forced-response analysis are
evaluated in terms of the system design. Necessary modifications are made to the system design. These
changes are then applied to the model and analysis parameters to perform another iteration on the design.
The process is repeated until an acceptable design is determined, which completes the design process.
The primary steps in performing a dynamic analysis are summarized as follows:
1. Define the dynamic environment (loading).
2. Formulate the proper finite element model.
3. Select and apply the appropriate analysis approach(es) to determine the behavior of the structure.
4. Evaluate the results.
Main Index
CHAPTER 1 15
Fundamentals of Dynamic Analysis
Frequency response analysis is an efficient method for finding the steady-state response to sinusoidal
excitation. In frequency response analysis, the loading is a sine wave for which the frequency, amplitude,
and phase are specified. Frequency response analysis is limited to linear elastic structures. Frequency
response analysis is described in Frequency Response Analysis, 133.
Transient response analysis is the most general method of computing the response to time-varying loads.
The loading in a transient analysis can be of an arbitrary nature, but is explicitly defined (i.e., known) at
every point in time. The time-varying (transient) loading can also include nonlinear effects that are a
function of displacement or velocity. Transient response analysis is most commonly applied to structures
with linear elastic behavior. Transient response analysis is described in Transient Response Analysis,
201.
Additional MD Nastran advanced dynamic analysis capabilities, such as damping, direct enforced
motion, random response analysis, response spectrum analysis and coupled fluid structure analysis can
be used in conjunction with the above analyses. More advanced dynamic analysis capabilities like design
sensitivity, design optimization, aeroelastic, rotor dynamics, control system and nonlinear transient also
build on these capabilities.
In practice, very few engineers use all of the dynamic analysis types in their work. Therefore, it may not
be important for you to become familiar with all of the types. Each type can be considered
independently, although there may be many aspects common to many of the analyses.
Main Index
16 MD Nastran Dynamic Analysis User’s Guide
Dynamic Analysis Types
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 2: Finite Element Input Data
Overview
Strategies for Dynamic Analysis
Plan the Analysis
Mass Input
Coupled Mass Matrix Terms
Mass Effects in Moving Coordinate Systems
Damping Input
Units in Dynamic Analysis
Main Index
18 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
When conducting an MD Nastran dynamic analysis, the formulation of the model depends upon the
desired results and the type of dynamic loading. For example, if stresses are of interest, a finer finite
element mesh is required than if only the system level displacement patterns are needed. Many of the
modeling considerations involved in a static analysis are similarly applied in dynamic analysis. A
dynamic analysis, however, requires additional input data which is used to define the dynamic character
of the structure.
In static analysis the stiffness properties are defined by element and material properties. These same
properties are also required for dynamic analysis along with the addition of mass and damping. Mass
Input, 21 describes mass input and Damping Input, 35 describes damping input.
Correct specification of units is very important for dynamic analysis. Incorrect specification of units is
probably more difficult to diagnose in dynamic analysis then in static analysis. Because MD Nastran has
no built-in set of units, you must ensure their consistency (and accuracy). Units in Dynamic Analysis, 40
describes the common variables and units for dynamic analysis.
In general, the practice of building MD Nastran models for dynamic analysis is similar to building statics
or normal modes models. However, in dynamics the higher costs of running detailed large order models
usually results in a different approach. For instance, a direct frequency response analysis may require
solutions at more than 100 frequencies. Each frequency will require an unsymmetric matrix solution
using complex variables, which will take four to ten times longer to solve then a single static solution. If
memory is sufficient, the running time could be larger by a factor 400 to 1000. If the required memory
(four times more then in a static analyses) is insufficient, the solution will spill, and even longer run times
will result. In addition, the database and scratch file space requirements will be much larger and the size
of the output data becomes very large. The result is that many users who simply take fine mesh static
models directly into a dynamic analysis find the solution time and required resources to be excessive.
Main Index
CHAPTER 2 19
Finite Element Input Data
Main Index
20 MD Nastran Dynamic Analysis User’s Guide
Plan the Analysis
Main Index
CHAPTER 2 21
Finite Element Input Data
Mass Input
F = M u·· (2-1)
where each component of the acceleration vector u·· represents a generalized degree-of-freedom.
The mass matrix is required for nearly all dynamic solution sequences. It is also used for generating
gravity and centrifugal loads for static solutions. Inertia relief solutions require the mass matrix to
balance the unbalanced forces on a free body. All solutions may calculate the total mass and center of
gravity (CG) information for printout.
Note that mass is not required for heat transfer dynamics.
Mass input is one of the major entries in a dynamic analysis. Mass can be represented in a number of
ways in MD Nastran. The mass matrix is automatically computed when mass density or nonstructural
mass is specified for any of the standard finite elements (CBAR, CQUAD4, etc.) in MD Nastran, when
concentrated mass elements are entered, and/or when full or partial mass matrices are entered.
Main Index
22 MD Nastran Dynamic Analysis User’s Guide
Mass Input
The MD Nastran coupled mass formulation is a modified approach to the classical consistent mass
formulation found in most finite element texts. The MD Nastran lumped mass is identical to the classical
lumped mass approach. The various formulations of mass matrices can be compared using the CROD
element. Assume the following properties:
Main Index
CHAPTER 2 23
Finite Element Input Data
1 2 3 4
AE AE-
–---------- 1
------- 0 0
L L
GJ – GJ 2
0 ------- 0 ----------
K = L L (2-2)
AE-
–---------- AE
-------
0 0 3
L L
– GJ GJ
0 ----------
L
0 -------
0
4
The zero entries in the matrix create independent (uncoupled) translational and rotational behavior for
the CROD element, although for most other elements these degrees-of-freedom are coupled.
1
--- 0 0 0
2
M = AL 0 0 0 0 (2-3)
1---
0 0 0
2
0 0 0 0
The lumped mass matrix is formed by distributing one-half of the total rod mass to each of the
translational degrees-of-freedom. These degrees-of-freedom are uncoupled and there are no torsional
mass terms calculated.
Main Index
24 MD Nastran Dynamic Analysis User’s Guide
Mass Input
1 1
--- 0 --- 0
3 6
I I
0 ------
- 0 ------
-
M = AL 3A 6A (2-4)
1 1
--- 0 --- 0
6 3
I I
0 ------
- 0 ------
-
6A 3A
This classical mass matrix is similar in form to the stiffness matrix because it has both translational and
rotational masses. Translational masses may be coupled to other translational masses, and rotational
masses may be coupled to other rotational masses. However, translational masses may not be coupled to
rotational masses.
5 1
------ 0 ------ 0
12 12
M = AL 0 0 0 0 (2-5)
1- 5-
----- 0 ----- 0
12 12
0 0 0 0
The axial terms in the CROD element coupled mass matrix represent the average of lumped mass and
classical consistent mass. This average is found to yield the best results for the CROD element as
described below. The mass matrix terms in the directions transverse to the element axes are lumped mass,
even when the coupled mass option is selected. Note that the torsional inertia is not included in the
CROD element mass matrix.
1.5708 E
-----------
l
Main Index
CHAPTER 2 25
Finite Element Input Data
Using the lumped mass formulation for the CROD element, the first frequency is predicted to be
E
1.414 -----------
l
which underestimates the frequency by 10%. Using a classical consistent mass approach, the predicted
frequency
E
1.732 -----------
l
is overestimated by 10%. Using the coupled mass formulation in MD Nastran, the frequency
E
1.549 -----------
l
is underestimated by 1.4%. The purpose of this example is to demonstrate the possible effects of the
different mass formulations on the results of a simple problem. Remember that not all dynamics
problems have such a dramatic difference. Also, as the model’s mesh becomes finer, the difference in
mass formulations becomes negligible.
ut
1
E E
Theoretical Natural Frequency: n = --- ---------------- = 1.5708 ----------------
2 l l
E E
MD Nastran Lumped Mass: n = 2 ---------------- = 1.414 ----------------
l l
E E
Classical Consistent Mass: n = 3 ---------------- = 1.732 ----------------
l l
E E
MD Nastran Coupled Mass: n = 12 5 ---------------- = 1.549 ----------------
l l
Main Index
26 MD Nastran Dynamic Analysis User’s Guide
Mass Input
Mass Units
Another important aspect of defining mass is the units of measure associated with the mass definition.
MD Nastran assumes that consistent units are used in all contexts. You must be careful to specify
structural dimensions, loads, material properties, and physical properties in a consistent set of units.
All mass entries should be entered in mass consistent units. Weight units may be input instead of mass
units, if this is more convenient. However, you must convert the weight to mass by dividing the weight
by the acceleration of gravity defined in consistent units:
m = 1 g w (2-6)
where:
g = acceleration of gravity
The parameter
PARAM,WTMASS,factor
performs this conversion. The value of the factor should be entered as 1 g . The default value for the
factor is 1.0. Hence, the default value for WTMASS assumes that mass (and mass density) is entered,
instead of weight (and weight density).
When using English units if the weight density of steel is entered as RHO = 0.3 lb/in3 , using
PARAM,WTMASS,0.002588 converts the weight density to mass density for the acceleration of gravity
2
g = 386.4 in/sec . The mass density, therefore, becomes 7.765E-4 lb f -sec 2 /in 4 . If the weight density of steel
Main Index
CHAPTER 2 27
Finite Element Input Data
is entered as RHO = 80000 N/m 3 when using metric units, then using PARAM,WTMASS,0.102 converts
the weight density to mass density for the acceleration of gravity g = 9.8 m/sec 2 . The mass density,
therefore, becomes 8160 kg/m 3 .
PARAM,WTMASS is used once per run, and it multiplies all weight/mass input (including CMASSi,
CONMi, and nonstructural mass input). Therefore, do not mix input type; use all mass (and mass
density) input or all weight (or weight density) input. PARAM,WTMASS does not affect direct input
matrices M2GG or M2PP (see Direct Matrix Input Dynamics, 360). If M2GG is used, then
PARAM,CM1, CM2 can be used to scale all weight/mass input (except for M2GG and M2PP); and
PARAM,CM1, CM2 can be used to scale M2GG; there is no parameter scaling for M2PP.
f1 u··
M –M 1 (2-7)
= – ··
f2 –M M u2
where f 1 and f 2 are the inertia forces acting at degrees-of-freedom 1 and 2, respectively, and
M is the mass coefficient, specified on the CMASSi entry (or on the PMASS entry if i = 2 or 4).
In most applications, the second degree-of-freedom is not specified. In this case, the entry
generates the inertia force f 1 = – Mu·· 1 , and M is added to the mass matrix in the diagonal position
corresponding to u 1 . An important application of the CMASSi entry occurs in the recommended
method for specifying enforced motion at grid points (see Enforced Motion with Loads, 292).
6. Transfer functions defined on the TF Bulk Data entry may contribute terms to the mass matrix.
Main Index
28 MD Nastran Dynamic Analysis User’s Guide
Mass Input
Mass Density
The most common method to enter mass is using the RHO field on the MATi entry. This field is assumed
to be defined in terms of mass density (mass/unit volume). To determine the total mass of the element,
the mass density is multiplied by the element volume (determined from the geometry and physical
properties). For a MAT1 entry, a mass density for steel of 7.76E-4 lb f -sec 2 /in 4 is entered as follows:
1 2 3 4 5 6 7 8 9 10
$MAT1 MID E G NU RHO A TREF GE
MAT1 2 30.0E6 0.3 7.76E-4
Nonstructural Mass
An additional way to input mass is to use nonstructural mass, which is mass not associated with the
geometric cross-sectional properties of an element. Examples of nonstructural mass are insulation,
roofing material, and special coating materials. Nonstructural mass is input as mass/length for line
elements and mass/area for elements with two-dimensional geometry. Nonstructural mass is defined on
the element property entry. Such as the NSM field on the PBAR entry for example.
Main Index
CHAPTER 2 29
Finite Element Input Data
3. PARAM,GRDPNT,V1 – Causes the Grid Point Weight Generator to be executed. The value of
the parameter, V1, is an integer which identifies a grid point at which the rigid body mass
properties of the structure will be computed. See the Grid Point Weight Generator (Ch. 19) in the
MD Nastran Linear Static Analysis User’s Guide for a description of GPWG and an example of
the output format which is provided. This capability can be used as a check on mass and
geometric input data.
Main Index
30 MD Nastran Dynamic Analysis User’s Guide
Coupled Mass Matrix Terms
u1 u2
Structure u
m1 P2
k m2
P1
A spring, k , and two masses, m 1 and m 2 , are attached to a structural model. However, when modeling
fluids or other special connections, the user may wish to use the difference in displacements,
u = u 1 – u 2 , instead of u 2 as the primary degree-of-freedom.
We may find the equivalent coupled mass matrix by the use of energy methods. The
2
V = ku 2
2 2
T = m 1 u·· 1 + m 2 u·· 1 – u··
(2-8)
Main Index
CHAPTER 2 31
Finite Element Input Data
and
m1 + m2 –m2
m = (2-10)
–m2 m2
where u = u 1 u are the active degrees-of-freedom. Two CMASSi elements may be used; one will
be coupled to two points. Note that the spring, k , is now connected to u only.
An alternate method would be to use the MPC constraints to produce the same effect. Simply include
all three DOFs in the model and use conventional lumped masses and a connected spring. Choose
u 2 – u 1 + u = 0 as the MPC equation, and the reduced mass and stiffness matrices will be identical to
the system above.
Main Index
32 MD Nastran Dynamic Analysis User’s Guide
Mass Effects in Moving Coordinate Systems
Spinning Bodies
If the entire structure is spinning at a constant angular velocity, both static centrifugal forces and coupled
dynamic inertial terms may be needed. An example is a high speed turbine wheel that exhibits
gyroscopic stability problems.
For dynamic analysis of rotating bodies, PARAM,CORITAN provided the generation of extra matrix
terms caused by the rotating coordinates. For a body rotation defined by the angular velocity vector, ,
and for a location vector of a point, r , the absolute velocity vector of the point, V , in fixed coordinates is
V = r + u + u·
(2-11)
The first term on the right-hand side of Eq. (2-11) is the static centrifugal force; the second term is the
centripetal stiffness; the third term is the Coriolis force; the last term, u , is the relative acceleration
vector.
Another term that is calculated for spinning bodies is the so-called differential stiffness matrix. It is
proportional to the steady centrifugal preloads in the elements. These are the terms that would stiffen a
string if a weight on the string were swung in a circular motion. These terms are important because they
are approximately the same magnitude as the Coriolis and centripetal stiffness terms defined in
Eq. (2-12). (Centrifugal stiffness and differential stiffness terms are of the same magnitude.)
The basic matrix equation for the forces in the moving system is
c c
B = generates the velocity-dependent Coriolis forces. Note that B is not symmetric!
c
K = the centripetal stiffness matrix
d
K = the differential stiffness matrix
Main Index
CHAPTER 2 33
Finite Element Input Data
The system defined by Eq. (2-13) may be solved with a transient analysis, a frequency response, or a
complex eigenvalue calculation. The complex eigenvalues may be obtained for a series of spin rates to
determine the critical angular speeds. At each angular velocity the complex roots are obtained in the
form p n = n i n . Unstable conditions usually occur when n P , where P is a positive integer. By
definition, the system is unstable if n is a positive number.
Inertia Relief
If a free body is accelerating due to constant unbalanced loads, the inertia relief solution provides the
ability to obtain static deflections relative to a set of reference points attached to the moving coordinate
system. An example is an airplane in a steady turn or accelerating dive. Although this capability is a
static solution, it is obtained from the dynamics theory.
The basic matrix equation for the inertia relief method is
K u = P – M ao (2-14)
where u are displacements relative to the moving system and a o are the steady accelerations to be
determined from the mass and loads. If D is a matrix whose columns define the rigid body motions
of the structure, then for a free body,
T T T
D K u = 0 = D P – D M ao (2-15)
where D is called the rigid body transformation matrix. However, since the full-sized vector, a o , is
a rigid body motion, it may be defined in terms of accelerations at a set of reference coordinates, a r ,
by the equation
ao = D ar (2-16)
–1 T
ao = D m D P (2-17)
Two different methods are used in MD Nastran to calculate the rigid body matrix D in SOL 101 using
PARAM,INREL. See Inertia Relief (Ch. 11) in the MD Nastran Linear Static Analysis User’s Guide for
a description of methods. One uses the SUPORT Bulk Data entry to define the reference
degrees-of-freedom. Any number of u r degrees-of-freedom that provide a nonredundant set of supports
may be used. This option allows partially free bodies and extra mechanisms. The other method
Main Index
34 MD Nastran Dynamic Analysis User’s Guide
Mass Effects in Moving Coordinate Systems
automatically determines the six rigid body displacement vectors. In these cases, only six free motions
are allowed.
Base Excitations
One of several methods to solve problems with enforced motion is to constrain the point of motion and
solve the problem in the accelerating system. This method is related to inertia relief but uses entirely
different inputs. It is easy to use for earthquake analysis of buildings, in which the base is accelerating
uniformly.
If a structure is attached to a semi-rigid base that causes a known stress-free motion, u 0 , the total
structural motion, u A , is
uA = u0 + ug (2-19)
where u g are displacements relative to the base motion. If the structure is not constrained elsewhere,
we may assume that the u 0 base displacements produce no force and
If we move the known base motion u·· 0 to the right hand side it looks almost identical to a gravity load:
Main Index
CHAPTER 2 35
Finite Element Input Data
Damping Input
Damping is a mathematical approximation used to represent the energy dissipation observed in
structures. Damping is difficult to model accurately since it is caused by many mechanisms including
• Viscous effects (dashpot, shock absorber)
• External friction (slippage in structural joints)
• Internal friction (characteristic of the material type)
• Structural nonlinearities (plasticity, gaps)
Because these effects are difficult to quantify, damping values are often computed based on the results
of a dynamic test. Simple approximations are often justified because the damping values are low.
f v = bu· (2-22)
where:
fs = i G k u (2-23)
where:
k = stiffness
u = displacement
For a sinusoidal displacement response of constant amplitude, the structural damping force is constant,
and the viscous damping force is proportional to the forcing frequency. Figure 2-2 depicts this and also
Main Index
36 MD Nastran Dynamic Analysis User’s Guide
Damping Input
shows that for constant amplitude sinusoidal motion the two damping forces are equal at a single
frequency.
At this frequency,
Gk
G k = b or b = ------- (2-24)
where is the frequency at which the structural and viscous damping forces are equal for a constant
amplitude of sinusoidal motion.
Viscous Damping
f v = bu· = i b u
Structural Damping
fs = i G k u
f
Damping Force
Forcing Frequency
Figure 2-2 Structural Damping and Viscous Damping Forces for Constant Amplitude
Sinusoidal Displacement
G k- = G m
b = -------- n (2-25)
n
b cr = 2 km = 2m n (2-26)
Some equalities that are true at resonance ( n ) for constant amplitude sinusoidal displacement are
b G
------- = = ---- (2-27)
bc r 2
Main Index
CHAPTER 2 37
Finite Element Input Data
1 1
and Q = ------ = ---- (2-28)
2 G
where Q is the quality or dynamic magnification factor, which is inversely proportional to the energy
dissipated per cycle of vibration.
1 2 3 4 5 6 7 8 9 10
$MAT1 MID E G NU RHO A TREF GE
MAT1 2 30.0E6 0.3 7.764E-4 0.10
specifies a structural damping coefficient of 0.1.
An alternate method for defining structural damping is through PARAM,G,r where r is the structural
damping coefficient. This parameter multiplies the stiffness matrix to obtain the structural damping
matrix. The default value for PARAM,G is 0.0. The default value causes this source of structural
damping to be ignored. Two additional parameters are used in transient response analysis to convert
structural damping to equivalent viscous damping: PARAM,W3, W4, W3FL, W4FL.
PARAM,G and GE can both be specified in the same analysis.
Main Index
38 MD Nastran Dynamic Analysis User’s Guide
Damping Input
CDAMP1 entry Scalar damper between two degrees-of-freedom (DOFs) with reference to a
PDAMP property entry.
CDAMP2 entry Scalar damper between two DOFs without reference to a property entry.
CDAMP3 entry Scalar damper between two scalar points (SPOINTs) with reference to a
PDAMP property entry.
CDAMP4 entry Scalar damper between two scalar points (SPOINTs) without reference to a
property entry.
CVISC entry Element damper between two grid points with reference to a PVISC property
entry.
CBUSH entry A generalized spring-and-damper structural element that may be nonlinear or
frequency dependent. It references a PBUSH entry.
Viscous damping for modal transient response and modal frequency response is specified with the
TABDMP1 entry.
Note that GE and G by themselves are dimensionless; they are multipliers of the stiffness. The CDAMPi
and CVISC entries, however, have damping units.
Damping is further described in Frequency Response Analysis, 133 and Transient Response Analysis,
201 as it pertains to frequency and transient response analyses. Additional discussions of damping
modeling and effects can be found in Damping (Ch. 6).
Output
Complex displacements, velocities, accelerations, strains, and stresses have been implemented for solid
p-elements. The same, including forces, have been implemented for shell and beam p-elements.
These output quantities can be written into the .f06 and .op2 files in either rectangular or polar format.
For rectangular, which is the default, the keyword REAL or IMAGINARY may be used, and for polar,
the keyword PHASE may be used. Each quantity may be output in a different format.
These output quantities may also be written in either SORT1 or SORT2 order. In SORT1 order, which is
the default for frequency and complex eigenvalue analyses, all the grids/elements are output for each
frequency/time/eigenvalue. In SORT2 order, which is the default for transient analysis, all the
frequencies/times/eigenvalues are output for each grid/element. Only one order may be chosen for the
entire analysis.
Limitations
The p-elements do not adapt in the dynamic solution sequences. Having the elements adapt within the
frequency or time steps would require too many additional resources for the value gained. If an adaptive
analysis is desired, it can be done in linear statics or normal modes, and the resulting p-distribution may
Main Index
CHAPTER 2 39
Finite Element Input Data
be used for the dynamic analysis. This would be especially useful for the modal solution sequences,
where the normal mode vectors could be used in a restart without being recalculated.
The time delay DELAY, phase delay DPHASE, and transient initial condition TIC Bulk Data entries
reference grid points only, which could cause singularities in p-elements. The appropriate functions for
p-elements would reference the edges and faces. However, no problem exists if the specified area is
limited to p=1.
Error estimation is currently not available. Because the p-elements do not adapt, this is unnecessary;
however, it could provide useful feedback. Running a normal modes analysis would provide error
estimation before any dynamic solutions are run.
Main Index
40 MD Nastran Dynamic Analysis User’s Guide
Units in Dynamic Analysis
Main Index
CHAPTER 2 41
Finite Element Input Data
* L Denotes length
M Denotes mass
T Denotes time
-- Denotes dimensionless
Main Index
42 MD Nastran Dynamic Analysis User’s Guide
Units in Dynamic Analysis
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 3: Real Eigenvalue Analysis
Overview
Reasons to Compute Normal Modes
Mathematical Overview of Normal Modes Analysis
Methods of Computation
User Interface for Real Eigenvalue Analysis
Solution Control for Analysis
Massless Mechanism Identification and Control
Prestiffened Normal Mode Analysis
Examples
Rigid Body Modes
SUPORT Entry
Rigid Body Mode Examples
Tools to Help Identify and Validate Normal Modes
Main Index
44 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
The usual first step in performing a dynamic analysis is determining the natural frequencies and mode
shapes of the structure with damping neglected. These results characterize the basic dynamic behavior
of the structure and are an indication of how the structure will respond to dynamic loading.
Natural Frequencies
The natural frequencies of a structure are the frequencies at which the structure naturally tends to vibrate
if it is subjected to a disturbance. For example, the strings of a piano are each tuned to vibrate at a specific
frequency. Some alternate terms for the natural frequency are characteristic frequency, fundamental
frequency, resonance frequency, and normal frequency.
Mode Shapes
The deformed shape of the structure at a specific natural frequency of vibration is termed its normal mode
of vibration. Some other terms used to describe the normal mode are mode shape, characteristic shape,
eigenvector and fundamental shape. Each mode shape is associated with a specific natural frequency.
Natural frequencies and mode shapes are functions of the structural properties and boundary conditions.
A cantilever beam has a set of natural frequencies and associated mode shapes (Figure 3-1). If the
structural properties change, the natural frequencies change, but the mode shapes may not necessarily
change. For example, if the elastic modulus of the cantilever beam is changed, the natural frequencies
change but the mode shapes remain the same. If the boundary conditions change, then the natural
frequencies and mode shapes both change. For example, if the cantilever beam is changed so that it is
pinned at both ends, the natural frequencies and mode shapes change (see Figure 3-2).
z x
1
z x
2
z x
3
z x
4
Main Index
CHAPTER 3 45
Real Eigenvalue Analysis
z x
1
z x
2
z x
3
z x
4
Figure 3-2 The First Four Mode Shapes of a Simply Supported Beam
Computation of the natural frequencies and mode shapes is performed by solving an eigenvalue problem
as described in Mathematical Overview of Normal Modes Analysis, 47. Next, we solve for the eigenvalues
(natural frequencies) and eigenvectors (mode shapes). Because damping is neglected in the analysis, the
eigenvalues are real numbers. (The inclusion of damping makes the eigenvalues complex numbers; see
Complex Eigenvalue Analysis (Ch. 13).) The solution for undamped natural frequencies and mode
shapes is called real eigenvalue analysis or normal modes analysis.
Main Index
46 MD Nastran Dynamic Analysis User’s Guide
Reasons to Compute Normal Modes
The results of the dynamic analyses are sometimes compared to the physical test results. A normal modes
analysis can be used to guide the experiment. In the pretest planning stages, a normal modes analysis can
be used to indicate the best location for the accelerometers. After the test, a normal modes analysis can
be used as a means to correlate the test results to the analysis results.
Design changes can also be evaluated by using natural frequencies and normal modes. Does a particular
design modification cause an increase in dynamic response? Normal modes analysis can often provide
an indication by paying attention to how the frequencies shift and if they now align with frequencies ‘to
be avoided’.
In summary, there are many reasons to compute the natural frequencies and mode shapes of a structure.
All of these reasons are based on the fact that real eigenvalue analysis is the basis for many types of
dynamic response analyses. Therefore, an overall understanding of normal modes analysis as well as
knowledge of the natural frequencies and mode shapes for your particular structure is important for all
types of dynamic analysis.
Main Index
CHAPTER 3 47
Real Eigenvalue Analysis
M u·· + K u = 0 (3-1)
where:
M = mass matrix
K = stiffness matrix
u = displacement
u·· = acceleration
This is the equation of motion for undamped free vibration. To solve Eq. (3-1) assume a harmonic
solution of the form
u = sin t (3-2)
where:
Aside from this harmonic form being the key to the numerical solution of the problem, this form also has
a physical importance. The harmonic form of the solution means that all the degrees-of-freedom of the
vibrating structure move in a synchronous manner. The structural configuration does not change its
basic shape during motion; only its amplitude changes.
If differentiation of the assumed harmonic solution is performed and substituted into the equation of
motion, the following is obtained:
2
– M sin t + K sin t = 0 (3-3)
which after simplifying becomes
2
K – M = 0 (3-4)
This equation is called the eigenequation, which is a set of homogeneous algebraic equations for the
components of the eigenvector and forms the basis for the eigenvalue problem. An eigenvalue problem
is a specific equation form that has many applications in linear matrix algebra. The basic form of an
eigenvalue problem is
A – I x = 0 (3-5)
Main Index
48 MD Nastran Dynamic Analysis User’s Guide
Mathematical Overview of Normal Modes Analysis
where:
A = square matrix
= eigenvalues
I = identity matrix
x = eigenvector
In structural analysis, the representations of stiffness and mass in the eigenequation result in the physical
representations of natural frequencies and mode shapes. Therefore, the eigenequation is written in terms
of K , , and M as shown in Eq. (3-4) with 2 = .
2
K – M = 0
From a structural engineering point of view, the general mathematical eigenvalue problem
reduces to one of solving the equation of the form
2
det K – M = 0 (3-7)
or
det K – M = 0 (3-8)
where = 2
The determinant is zero only at a set of discrete eigenvalues i or 2i . There is an eigenvector i which
satisfies and corresponds to each eigenvalue. Therefore, can be rewritten as
2
K – i M i = 0 i = 1 2 3 (3-9)
Each eigenvalue and eigenvector define a free vibration mode of the structure. The i-th eigenvalue i is
related to the i-th natural frequency as follows:
f i = ------i (3-10)
2
Main Index
CHAPTER 3 49
Real Eigenvalue Analysis
where:
i = i
The number of possible eigenvalues and eigenvectors is equal to the number of degrees-of-freedom that
have mass or the number of dynamic degrees-of-freedom.
There are a number of characteristics of natural frequencies and mode shapes that make them useful in
various dynamic analyses. First, when a linear elastic structure is vibrating in free or forced vibration,
its deflected shape at any given time is a linear combination of all of its normal modes
u = i i (3-11)
i
where:
Second, if K and M are symmetric and real (as is the case for all the common structural finite
elements), the following mathematical properties hold:
T
i M j = 0 if i j (3-12)
T
j M j = m j = j-th generalized mass (3-13)
and
T
j K j = 0 if i j (3-14)
T 2
j K j = k j = j-th generalized stiffness = m j (3-15)
Also, from Eq. (3-13) and Eq. (3-15) Rayleigh’s equation is obtained
T
2 j K j
j = ------------------------------------
- (3-16)
T
j M j
Eq. (3-12) and Eq. (3-14) are known as the orthogonality property of normal modes, which ensures that
each normal mode is distinct from all others. Physically, orthogonality of modes means that each mode
Main Index
50 MD Nastran Dynamic Analysis User’s Guide
Mathematical Overview of Normal Modes Analysis
shape is unique and one mode shape cannot be obtained through a linear combination of any other mode
shapes.
In addition, a natural mode of the structure can be represented by using its generalized mass and
generalized stiffness. This is very useful in formulating equivalent dynamic models and in component
mode synthesis (see Special Topics, 359).
u1 u2
m1 m2
k
u1 1
1 = 0 1 = =
u2 1
When both masses move the same amount (as a rigid body), there is no force induced in the connecting
spring. A detailed discussion of rigid-body modes is presented in Rigid Body Modes, 99.
Scaling of Eigenvectors
An important characteristic of normal modes is that the scaling or magnitude of the eigenvectors is
arbitrary. Mode shapes are fundamental characteristic shapes of the structure and are therefore relative
quantities. In the solution of the equation of motion, the form of the solution is represented as a shape
with a time-varying amplitude. Therefore, the basic mode shape of the structure does not change while
it is vibrating; only its amplitude changes.
For example, three different ways to represent the two modes of a two-DOF structure are shown in
Figure 3-4. The graphical representation of the eigenvectors in the figure shows the modal displacements
rotated by 90 degrees in order to view the deformation better.
Main Index
CHAPTER 3 51
Real Eigenvalue Analysis
u1 u1 u1
m1
k1
u2
u2 u2
m2
k2
Mode 1 Mode 2
u1 3 300 – 0.6 u1 – 1.6 160 1.0
1 = = = = 2 = = = =
u2 1 100 – 0.2 u2 0.8 – 80 – 0.5
A common misconception about mode shapes is that they define the structural response. Again, mode
shapes are relative quantities. They cannot be used alone to evaluate dynamic behavior. As described
earlier, it is the relation between the structural loading and the natural frequencies that determines the
absolute magnitude of dynamic response. The relation of a specific loading to a set of natural
frequencies provides explicit scale factors that are used to determine the extent to which each particular
mode is excited by the loading. After the individual modal responses to a particular loading are
determined, only then can the various engineering design decisions be made with the actual (absolute)
values of stress and/or displacement. Methods that use the modal results to determine forced response
are termed modal methods or modal superposition methods. Modal frequency response analysis and
modal transient response analysis are described in Frequency Response Analysis, 133 and Transient
Response Analysis, 201, respectively.
Numerically this method results in a modal mass matrix that is an identity matrix. This normalization
approach is appropriate for modal dynamic response calculations because it simplifies both
Main Index
52 MD Nastran Dynamic Analysis User’s Guide
Mathematical Overview of Normal Modes Analysis
computational and data storage requirements. When mass normalization is used with a model of a heavy,
massive structure, the magnitude of each of the terms of the eigenvectors is very small.
In MAX normalization, each eigenvector is normalized with respect to the largest a-set component.
(Dynamic Reduction, 371 and The Set Notation System Used in Dynamic Analysis, 503 provide
discussions of the a-set.) This normalization results in the largest a-set displacement value being set to
a unit (1.0) value. This normalization approach can be very useful in the determination of the relative
participation of an individual mode. A small generalized mass obtained using MAX normalization may
indicate such things as local modes or isolated mechanisms.
POINT normalization of eigenvectors allows you to chose a specific displacement component at which
the modal displacement is set to 1 or -1. This method is not recommended because for complex structures
the chosen component in the non-normalized eigenvector may have a very small value of displacement
(especially in higher modes). This small value can cause larger numbers to be normalized by a small
number, resulting in possible numerical roundoff errors in mode shapes.
Although mode shapes are relative quantities, a number of modal quantities can be helpful in predicting
qualitative responses or in isolating troublesome modal frequencies. Since relative strains, internal loads,
and stresses develop when a structure deforms in a mode shape, you may recover these quantities during
a normal modes analyses. Basically, any quantity that you can recover for static analysis is also available
for normal modes analysis.
It is important to remember that these output quantities are based on the relative displacements of a mode
shape. The output quantities can be compared for a given mode, but not necessarily from one mode to
another. However, they can still be effectively used in the analysis/design process.
Modal quantities can be used to identify problem areas by indicating the more highly stressed elements.
Elements that are consistently highly stressed across many or all modes will probably be highly stressed
when dynamic loads are applied.
Modal strain energy is a useful quantity in identifying candidate elements for design changes to eliminate
problem frequencies. Elements with large values of strain energy in a mode indicate the location of large
elastic deformation (energy). These elements are those which most directly affect the deformation in a
mode. Therefore, changing the properties of these elements with large strain energy should have more
effect on the natural frequencies and mode shapes than if elements with low strain energy were changed.
Structures with two or more identical eigenvalues are said to have repeated roots. Repeated roots occur
for structures that have a plane of symmetry or that have multiple, identical pieces (such as appendages).
The eigenvectors for the repeated roots are not unique because many sets of eigenvectors can be found
that are orthogonal to each other. An eigenvector that is a linear combination of the repeated eigenvectors
is also a valid eigenvector. Consequently, small changes in the model can make large changes in the
eigenvectors for the repeated roots. Different computers can also find different eigenvectors for the
repeated roots. Rigid-body modes (see Rigid Body Modes, 99) represent a special case of repeated roots.
Main Index
CHAPTER 3 53
Real Eigenvalue Analysis
Methods of Computation
Seven methods of real eigenvalue extraction are provided in MD Nastran. These methods are numerical
approaches to solving for the natural frequencies and modes shapes. The reason for seven different
numerical techniques is because no one method is the best for all problems. While most of the methods
can be applied to all problems, the choice is often based on the efficiency of the solution process.
The methods of eigenvalue extraction belong to one or both of the following two groups:
• Transformation methods
• Tracking methods
In the transformation method, the eigenvalue equation is first transformed into a special form from which
eigenvalues may easily be extracted. In the tracking method, the eigenvalues are extracted one at a time
using an iterative procedure.
The recommended real eigenvalue extraction method in MD Nastran is the Lanczos method. The
Lanczos method combines the best characteristics of both the tracking and transformation methods. For
most models the Lanczos method is the best method to use.
Four of the real eigenvalue extraction methods available in MD Nastran are transformation methods.
These methods are
• Givens method
• Householder method
• Modified Givens method
• Modified Householder method
Two of the real eigenvalue extraction methods available in MD Nastran are classified as tracking
methods. These methods are
• Inverse power method
• Sturm modified inverse power method
The remainder of this section briefly describes the various methods. The theory and algorithms behind
each method can be found in the MD Nastran Numerical Methods User's Guide.
Lanczos Method
The Lanczos method overcomes the limitations and combines the best features of the other methods. It
requires that the mass matrix be positive semidefinite and the stiffness be symmetric. Like the
transformation methods, it does not miss roots, but has the efficiency of the tracking methods, because
it only makes the calculations necessary to find the roots requested by the user. This method computes
accurate eigenvalues and eigenvectors. Unlike the other methods, its performance has been continually
enhanced since its introduction giving it an advantage. The Lanczos method is the preferred method for
most medium- to large-sized problems, since it has a performance advantage over other methods.
Main Index
54 MD Nastran Dynamic Analysis User’s Guide
Methods of Computation
Also, Lanczos uses Sturm sequence logic to ensure that all modes are found. The Sturm sequence check
determines the number of eigenvalues below a trial eigenvalue, then finds all of the eigenvalues below
this trial eigenvalue until all modes in the designated range are computed. This process helps to ensure
that modes are not missed.
The Givens and Householder methods fail if the mass matrix is not positive definite. To minimize this
problem, degrees-of-freedom with null columns are removed by the automatic application of static
condensation (see Dynamic Reduction, 371) called auto-omit. Applying the auto-omit process is a
precaution and may not remove all possible causes of mass matrix singularity, such as a point mass offset
from a grid point, but it greatly improves the reliability and convenience of the Givens and Householder
methods.
Givens and Householder methods use different transformation schemes to obtain the eigenvalues. For
problems in which no spill occurs (i.e., all of the matrices fit in your computer's main memory), the
Householder method costs about half as much as the Givens method for vector processing computers. In
addition, the Householder method can take advantage of parallel processing computers.
Main Index
CHAPTER 3 55
Real Eigenvalue Analysis
therefore, use the modified methods when necessary for numerical stability but use the standard methods
when the numerical stability is accurate.
Main Index
56 MD Nastran Dynamic Analysis User’s Guide
User Interface for Real Eigenvalue Analysis
1 2 3 4 5 6 7 8 9 10
EIGRL SID V1 V2 ND MSGLVL MAXSET SHFSCL NORM
option_1=value_1, option_2=value_2, etc.
Example:
Field Contents
SID Set identification number. (Unique Integer > 0)
V1, V2 For vibration analysis: frequency range of interest. For buckling analysis: eigenvalue
range of interest. See Remark 4. (Real or blank, – 5 10 16 V1 V2 5. 10 16 )
ND Number of roots desired. See Remark 4. (Integer > 0 or blank)
MSGLVL Diagnostic level. (0 < Integer < 4; Default = 0)
MAXSET Number of vectors in block or set. Default is machine dependent. See Remark 14.
SHFSCL Estimate of the first flexible mode natural frequency. See Remark 10. (Real or blank)
NORM Method for normalizing eigenvectors (Character: “MASS” or “MAX”)
MASS Normalize to unit value of the generalized mass. Not available for buckling
analysis. (Default for normal modes analysis.)
MAX Normalize to unit value of the largest displacement in the analysis set.
Displacements not in the analysis set may be larger than unity. (Default for
buckling analysis.)
Main Index
CHAPTER 3 57
Real Eigenvalue Analysis
Field Contents
ALPH Specifies a constant for the calculation of frequencies (Fi) at the upper boundary segments
for the parallel method based on the following formula. See Remark 13. (Integer > 0.0;
Default = 1.0):
i
1 – ALPH
Fi = V2 – V1 ---------------------------------------
NUMS
-
1. – ALPH
NUMS Number of frequency segments for the parallel method. (Integer > 0; Default = 1)
Fi Frequency at the upper boundary of the i-th segment. See Remark 13. (Real or blank;
V1 F1 F2 F15 V2 )
option_i= Assigns a value to the fields above except for SID. ALPH, NUMS, and Fi must be
value_i specified in this format. V1, V2, ND, MSGLVL, MAXSET, SHFSCL, and NORM may
be specified in this format as long as their corresponding field is blank in the parent entry.
Examples of the results of using explicit or default values for the V1, V2, and ND fields are shown in
Table 3-1. The defaults on the EIGRL entry are designed to provide the minimum number of roots in
cases where the input is ambiguous.
Table 3-1 Number and Type of Roots Found with the EIGRL Entry
Case V1 V2 ND Number and Type of Roots Found
1 V1 V2 ND Lowest ND in range or all in range, whichever is smaller
2 V1 V2 All in range
3 V1 ND Lowest ND in range [V1, ]
4 V1 Lowest root in range [V1, ]
5 ND Lowest ND roots in [-, ]
6 Lowest root
7 V2 ND Lowest ND or all in range [-, V2], whichever is smaller
8 V2 All below V2
The MSGLVL field of the EIGRL entry is used to control the amount of diagnostic output. The value
of 0 produces no diagnostic output. The values 1, 2, or 3 provide more output with the higher values
providing increasingly more output. In some cases, higher diagnostic levels may help to resolve
difficulties with special modeling problems.
The MAXSET field is used to control the block size. The default value of 7 is recommended for most
applications. There may be special cases where a larger value may result in quicker convergence of
many multiple roots or a lower value may result in more efficiency when the structure is lightly coupled.
However, the default value has been chosen after reviewing the results from a wide range of problems
Main Index
58 MD Nastran Dynamic Analysis User’s Guide
User Interface for Real Eigenvalue Analysis
on several different computer types with the goal of minimizing the computer time. A common
occurrence is for the block size to be reset by MD Nastran during the run because there is insufficient
memory for a block size of 7. Computational efficiency tends to degrade as the block size decreases.
Therefore, it is, important to examine the eigenvalue analysis summary output to determine that
MD Nastran has sufficient memory to use an efficient block size. A smaller block size may be more
efficient when only a few roots are requested. The minimum recommended block size is 2.
The SHFSCL field allows a user-designated shift to be used to improve performance especially when
large mass techniques are used in enforced motion analysis (see Enforced Motion, 281). Large mass
techniques can cause a large gap between the rigid body (see Rigid Body Modes, 99) and flexible
frequencies, which can degrade performance of the Lanczos method or cause System Fatal Message
5299. When SHFSCL is used, its value should be set close to the expected first nonzero natural
frequency.
The Lanczos method normalizes (i.e., scales) the computed eigenvectors using the MASS or MAX
method. These methods are specified using the NORM field. The MASS method normalizes to a unit
value of the generalized mass (i.e., m j = 1.0 ). The MAX method normalizes to a unit value of the largest
component in the a-set (see Dynamic Reduction, 371). The default is MASS.
You can use the continuation entry to specify V1, V2, ND, MSGLVL, MAXSET, SHFSCL and NORM
if you have not specified them on the parent entry. To apply the continuation entry use the following
format: 'option_i=value_i', e.g., ND=6. Using the continuation entry is the only way to specify the three
new options, ALPH, NUMS and Fi.
NUMS The number of segments that a frequency range will be broken into for parallel processing.
You must define a value greater than 1 to take advantage of parallel processing. You may
also specify NUMS using the NUMSEG keyword on the NASTRAN statement. If you
specify both, then NUMS takes precedence.
Fi Directly specifies the upper frequencies of each segment, such that
V1 < F1 < F2 < ... F15 < V2.
ALPH Automatically generates the Fi values based on the following formula:
i NUMS
Fi = V2 – V1 1.0 – ALPH 1.0 – ALPH
If you specify both ALPH and Fi, then Fi takes precedence over ALPH as long as they are consistent. If
ALPH is multiplied by 100, it may be specified on the FRQSEQ keyword of the NASTRAN statement.
Main Index
CHAPTER 3 59
Real Eigenvalue Analysis
1 2 3 4 5 6 7 8 9 10
EIGR SID METHOD F1 F2 NE ND
NORM G C
Field Contents
SID Set identification number. (Unique Integer > 0)
METHOD Method of eigenvalue extraction. (Character)
Modern Methods:
LAN Lanczos Method
AHOU Automatic selection of HOU or MHOU method. See Remark 13.
Obsolete Methods:
INV Inverse Power method.
SINV Inverse Power method with enhancements.
GIV Givens method of tridiagonalization.
MGIV Modified Givens method.
HOU Householder method of tridiagonalization.
MHOU Modified Householder method.
AGIV Automatic selection of METHOD = “GIV” or “MGIV”. See Remark
13.
NORM Method for normalizing eigenvectors. (Character: “MASS,” “MAX,” or “POINT”;
Default = “MASS”)
MASS Normalize to unit value of the generalized mass. (Default)
MAX Normalize to unit value of the largest component in the analysis set.
POINT Normalize to a positive or negative unit value of the component
defined in fields 3 and 4. The POINT option is not supported for
METH=LAN. (Defaults to “MASS” if defined component is zero.)
G Grid or scalar point identification number. Required only if NORM = “POINT”.
(Integer > 0)
C Component number. Required only if NORM = “POINT” and G is a geometric grid
point. (1 < Integer < 6)
Main Index
60 MD Nastran Dynamic Analysis User’s Guide
User Interface for Real Eigenvalue Analysis
There is an interrelationship among the F1, F2, and ND fields on the EIGR entry as defined in Table 3-2.
Table 3-2 Relationship Between METHOD Field and Other Fields for Obsolete Methods
METHOD Field
Field INV or SINV GIV, MGIV, HOU, or MHOU
F1, F2 Frequency range of interest. F1 must Frequency range of interest. If ND is not blank, F1
be input. If METHOD = “SINV” and and F2 are ignored. If ND is blank, eigenvectors
ND, is blank, then F2 must be input. are found with natural frequencies that lie in the
See also Remark 21. range between F1 and F2. (Real > 0.0; F1 < F2)
(Real > 0.0)
NE Estimate of number of roots in range Not used.
(Required for METHOD = “INV”).
Not used by “SINV” method.
(Integer > 0)
ND Desired number of roots. If this field Desired number of eigenvectors. If ND is zero, the
is blank and METHOD = “SINV”, number of eigenvectors is determined from F1 and
then all roots between F1 and F2 are F2. If all three are blank, then ND is automatically
searched and the limit is 600 roots. set to one more than the number of degrees-of-
(Integer > 0, Default is 3 NE for freedom listed on SUPORTi entries. (Integer > 0;
METHOD = “INV” only.) Default = 0)
The rules for METHOD = GIV, HOU, MGIV, MHOU, AGIV, and AHOU are identical. If any of these
methods are selected, MD Nastran finds all of the eigenvalues but only computes the eigenvectors
specified by F1 and F2 or those specified by ND (the desired number). F1 and F2 specify the lower and
upper bounds of the frequency range in which eigenvectors are computed, and ND specifies the number
of eigenvectors, beginning with the lowest (or the first rigid-body mode, if present). If F1, F2, and ND
entries are present, ND takes precedence.
If METHOD = SINV, the values of F1, F2, and ND determine the number of eigenvalues and
eigenvectors that are computed. These entries also provide hints to help MD Nastran find the
eigenvalues. F1 and F2 specify the frequency range of interest within which MD Nastran searches for
modes. MD Nastran attempts to find all of the modes in the range between F1 and F2 or the number
specified by ND, whichever is less. If searching stops because ND modes are found, there is no guarantee
that they are the lowest eigenvalues. If ND modes are not found in the range of interest, SINV usually
finds one mode (or possibly more) outside the range F1 and F2 before stopping the search.
The SINV method is particularly efficient when only a small number of eigenvalues and eigenvectors are
to be computed. Often only the lowest mode is of interest. The following example illustrates an EIGR
entry which extracts only the lowest nonzero eigenvalue.
1 2 3 4 5 6 7 8 9 10
EIGR 13 SINV 0.0 0.01 1
Main Index
CHAPTER 3 61
Real Eigenvalue Analysis
It is assumed in the example above that the frequency of the lowest mode is greater than 0.01 cycles per
unit time. MD Nastran finds one eigenvalue outside the range F1, F2, and then stops the search. The
eigenvalue found is the lowest nonzero eigenvalue (or a member of the lowest closely spaced cluster of
eigenvalues in cases with close roots) provided that there are no negative eigenvalues and that the
SUPORT entry has been used to specify the correct number of zero eigenvalues (see Rigid Body Modes,
99).
The following examples demonstrate the use of the EIGR Bulk Data entry.
1 2 3 4 5 6 7 8 9 10
EIGR 1 AHOU 10
In this example, the automatic Householder method is selected, and the lowest 10 modes are requested.
Since the default MASS eigenvector normalization is requested, no continuation entry is needed.
Main Index
62 MD Nastran Dynamic Analysis User’s Guide
Solution Control for Analysis
Main Index
CHAPTER 3 63
Real Eigenvalue Analysis
Main Index
64 MD Nastran Dynamic Analysis User’s Guide
Massless Mechanism Identification and Control
The massless mechanism feature is controlled by several parameters. The following is a list of the
primary parameters along with their functions:
Controls for fixing the massless mechanism problem. The capability is provided
automatically for the default of this parameter, listed above. The capability is executed
only when the eigensolution does not provide answers because of symptoms consistent
with the presence of massless mechanisms. The MMs are removed, and a second
eigensolution is made. If MECHFIX is set to YES, the constraint modes are removed
before attempting an eigensolution. When set to NO, the capability is blocked, and the
eigensolution uses the pre MSC.Nastran 2001 rules, i.e., three failed shifts and a fatal
exit.
MECHPRT Default = NO
For SOL 103 only, if massless mechanisms are found, the constraint modes are printed
with a format similar to eigenvectors when this parameter is set to YES. They are
labeled CONSTRAINT MODES, and are numbered sequentially. Grid points with
only zero values in a mode are not printed. This parameter should be used when
performing initial checkout of a model and a goal is to remove all massless
mechanisms before starting production analysis. The number of each "mode" matches
the corresponding GID,C pair in the high ratio message. If there are many (thousands)
of such modes, the output file will be large. There is no method to plot these shapes at
present.
Main Index
CHAPTER 3 65
Real Eigenvalue Analysis
The following secondary parameters have default values that work well in most circumstances. They are
provided for tuning unusual circumstances.
This is another traditional parameter whose other functions are described in the MD
Nastran Quick Reference Guide. It is also the criterion for identifying massless
mechanisms. A lower value is more conservative in that it is likely to identify and
control more mechanisms. If blocking of this capability is desired, use
PARAM,MECHFIX,NO instead of modifying this parameter. The rationale is that if
the user wants to set it high enough that it never identifies a mechanism, it is better not
to request massless mechanism control.
Criterion for discarding massless mechanism modes with small generalized mass. A
smaller value will result in more marginal constraint modes being retained.
Filter value used to distinguish between MM modes and rigid body modes. A smaller
value may discard rigid body modes. The default value has been effective on all
problems solved to date.
NLMAX Default = 60
The number of suspected MMs is determined from the number of high ratio messages.
If this number exceeds NLMAX, the number of trial MMs is reduced to NLMAX. This
avoids an expensive debug run when there may be thousands of MMs due to systematic
modeling error, such as having CONM2 elements on many grid points for which
structural elements have been left out through oversight. The value of this parameter
may be increased on initial debug runs where many high ratio DOFs may be present,
and the user prefers to see them all at once, rather than on successive runs where only a
part are removed at one time.
NLMIN Default = 10
In the case of only one or a few high ratio DOFs more MMs may be present. More trial
MM vectors are used, and those that do not indicate true problems are discarded. A
smaller value could be considered on a stable model undergoing small modeling
changes.
Main Index
66 MD Nastran Dynamic Analysis User’s Guide
Prestiffened Normal Mode Analysis
The objective of applying a static load to a structure during a normal modes analysis is to use the loaded
structure in obtaining the stiffness matrix for the modal analysis.
The eigenvalue problem equation corresponding to pre-stiffened structures is:
2
– M + K + K D = 0 (3-18)
where K D is the differential stiffness matrix resulting from the applied load.
In general, there are three terms/effects that can influence the stiffness matrix:
• Material nonlinearity, e.g. elastic-plastic behavior
• Geometric stiffness, due to changes in the structural shape... these can be due to small or large
displacement behavior
• Follower-force stiffness, arising as the loading changes its line of action relative to the
displacements... these can be due to small or large displacement behavior
Since the normal modes analysis is linear, only the geometric and follower force stiffness changes due to
small displacements will be considered. For incorporating the full nonlinear effects of the loading, a
nonlinear static solution is required.
The basic approach for incorporating the pre-loaded stiffness in the normal modes run is to run SOL 103
with two subcases, the first subcase being used to obtain the differential stiffness matrix from a linear
static analysis, and the second subcase being used to solve the eigenvalue problem using the differential
stiffness from the prior subcase. The primary trigger for this is the STATSUB Case Control command.
The second subcase will contain the STATSUB command to identify the static subcase from which the
differential stiffness is to be used.
A case study of this method is now shown. The normal modes analysis of a simple thin walled cylinder
with closed ends is considered under two conditions:
• No pre-load
• Internal pressurization with 100 psi
Geometry of the Cylinder
Main Index
CHAPTER 3 67
Real Eigenvalue Analysis
Main Index
68 MD Nastran Dynamic Analysis User’s Guide
Prestiffened Normal Mode Analysis
Listing 3-2 Eigenvalue table from the SOL 103 normal modes analysis, without
prestiffening
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 7.578808E+07 8.705635E+03 1.385545E+03 1.000000E+00 7.578808E+07
2 2 7.578847E+07 8.705657E+03 1.385548E+03 1.000000E+00 7.578847E+07
3 3 8.052795E+07 8.973736E+03 1.428214E+03 1.000000E+00 8.052795E+07
4 4 8.052826E+07 8.973754E+03 1.428217E+03 1.000000E+00 8.052826E+07
5 5 1.871432E+08 1.368003E+04 2.177244E+03 1.000000E+00 1.871432E+08
6 6 1.871440E+08 1.368006E+04 2.177249E+03 1.000000E+00 1.871440E+08
7 7 2.133802E+08 1.460754E+04 2.324862E+03 1.000000E+00 2.133802E+08
8 8 2.133841E+08 1.460767E+04 2.324883E+03 1.000000E+00 2.133841E+08
9 9 2.603917E+08 1.613666E+04 2.568229E+03 1.000000E+00 2.603917E+08
10 10 2.603947E+08 1.613675E+04 2.568243E+03 1.000000E+00 2.603947E+08
Listing 3-3 Eigenvalue table from the SOL 103 normal modes analysis, with prestiffening
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 1.365975E+08 1.168749E+04 1.860122E+03 1.000000E+00 1.365975E+08
2 2 1.365977E+08 1.168750E+04 1.860124E+03 1.000000E+00 1.365977E+08
3 3 2.401212E+08 1.549584E+04 2.466240E+03 1.000000E+00 2.401212E+08
4 4 2.401227E+08 1.549589E+04 2.466248E+03 1.000000E+00 2.401227E+08
5 5 3.519227E+08 1.875960E+04 2.985684E+03 1.000000E+00 3.519227E+08
6 6 3.519234E+08 1.875962E+04 2.985686E+03 1.000000E+00 3.519234E+08
7 7 4.099749E+08 2.024784E+04 3.222543E+03 1.000000E+00 4.099749E+08
8 8 4.099752E+08 2.024784E+04 3.222544E+03 1.000000E+00 4.099752E+08
9 9 4.993418E+08 2.234596E+04 3.556469E+03 1.000000E+00 4.993418E+08
10 10 4.993436E+08 2.234600E+04 3.556476E+03 1.000000E+00 4.993436E+08
Listing 3-4 Eigenvalue table from the nonlinear pre-stiffened normal modes analysis
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 1.364892E+08 1.168286E+04 1.859385E+03 1.000000E+00 1.364892E+08
2 2 1.364894E+08 1.168287E+04 1.859386E+03 1.000000E+00 1.364894E+08
3 3 2.399139E+08 1.548915E+04 2.465175E+03 1.000000E+00 2.399139E+08
4 4 2.399153E+08 1.548920E+04 2.465183E+03 1.000000E+00 2.399153E+08
5 5 3.512732E+08 1.874228E+04 2.982927E+03 1.000000E+00 3.512732E+08
6 6 3.512738E+08 1.874230E+04 2.982929E+03 1.000000E+00 3.512738E+08
7 7 4.097554E+08 2.024242E+04 3.221680E+03 1.000000E+00 4.097554E+08
8 8 4.097556E+08 2.024242E+04 3.221681E+03 1.000000E+00 4.097556E+08
9 9 4.988593E+08 2.233516E+04 3.554751E+03 1.000000E+00 4.988593E+08
10 10 4.988609E+08 2.233519E+04 3.554756E+03 1.000000E+00 4.988609E+08
As long as the nonlinear effects are minimal, these would match the SOL 103 method.
Investigation of the first mode for the two analysis shows that the presence of the pressurization has also
changed the mode shape from 3 lobes to 2 lobes.
Main Index
CHAPTER 3 69
Real Eigenvalue Analysis
First mode for normal modes analysis First mode for normal modes analysis for
pre-stiffened structure
Main Index
70 MD Nastran Dynamic Analysis User’s Guide
Examples
Examples
This section provides several normal-modes analysis examples showing the input and output. These
examples are as follows:
Two-DOF Model
This example is a restrained two-DOF model with two springs and two masses as illustrated in Figure 3-5.
Main Index
CHAPTER 3 71
Real Eigenvalue Analysis
k 1 = 100 N/m
Grid Point 2 m 2 = 10 kg
k 2 = 1.0E4 N/m
z
Figure 3-5 Two-DOF Model
The masses are constrained to deflect in only the y-direction. The example illustrates normal modes
analysis (SOL 103) using automatic selection of the Householder or modified Householder method
(METHOD = AHOU on the EIGR entry). The eigenvectors are normalized to the unit value of the
largest displacement component (NORM = MAX on the EIGR entry). The input file is shown in
Listing 3-5, see Install_dir/mdxxxx/doc/dynamics/bd03two.dat.
Listing 3-5
$ FILE bd03two.dat
$
$ TWO DOF SYSTEM
$ CHAPTER 3, NORMAL MODES
$
TIME 5
SOL 103 $ NORMAL MODES ANALYSIS
CEND
$
TITLE = TWO DOF SYSTEM
SUBTITLE = NORMAL MODES ANALYSIS
$
$ SELECT SPC
SPC = 10
$
$ SELECT EIGR ENTRY
METHOD = 99
$
$ SELECT OUTPUT
SET 1 = 1,2
DISPLACEMENT = 1
SET 2 = 3
SPCFORCE = 2
SET 3 = 11,12
ELFORCE = 3
$
Main Index
72 MD Nastran Dynamic Analysis User’s Guide
Examples
BEGIN BULK
$
$......2.......3.......4.......5.......6.......7.......8.......9.......10..$
$
$EIGR SID METHOD F1 F2 NE ND +EIG
$+EIG NORM G C
EIGR 99 AHOU 0. 200. +EIG1
+EIG1 MAX
$
GRID 1 0. 2. 0.
GRID 2 0. 1. 0.
GRID 3 0. 0. 0.
GRDSET 13456
CONM2 1 1 0.1
CONM2 2 2 10.0
CELAS2 11 100.0 1 2 2 2
CELAS2 12 1.0E4 2 2 3 2
SPC 10 3 2
$
ENDDATA
The printed output is shown in Listing 3-6. The eigenvalue summary lists the eigenvalue 2n , circular
frequency n (radians per second), natural frequency f n (cycles per second), generalized mass (see
Eq. (3-13)), and generalized stiffness (see Eq. (3-15)) for each mode. The eigenvectors, SPC forces, and
spring forces are shown for each mode.
Main Index
CHAPTER 3 73
Real Eigenvalue Analysis
EIGENVALUE = 9.048751E+02
CYCLES = 4.787562E+00 R E A L E I G E N V E C T O R N O . 1
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 1.000000E+00 0.0 0.0 0.0 0.0
2 G 0.0 9.512492E-02 0.0 0.0 0.0 0.0
EIGENVALUE = 1.105125E+03
CYCLES = 5.290854E+00 R E A L E I G E N V E C T O R N O . 2
EIGENVALUE = 9.048751E+02
F O R C E S O F S I N G L E - P O I N T C O N S T R A I N T
EIGENVALUE = 1.105125E+03
F O R C E S O F S I N G L E - P O I N T C O N S T R A I N T
EIGENVALUE = 9.048751E+02
F O R C E S I N S C A L A R S P R I N G S ( C E L A S 2 )
ELEMENT FORCE ELEMENT FORCE ELEMENT FORCE ELEMENT FORCE
ID. ID. ID. ID.
11 9.048751E+01 12 9.512491E+02
EIGENVALUE = 1.105125E+03
F O R C E S I N S C A L A R S P R I N G S ( C E L A S 2 )
ELEMENT FORCE ELEMENT FORCE ELEMENT FORCE ELEMENT FORCE
ID. ID. ID. ID.
11 1.105125E+02 12 -1.051249E+03
x z
L
r
Main Index
74 MD Nastran Dynamic Analysis User’s Guide
Examples
4
L = 3.0 m r = 0.014 m J = 6.0E-8 m
2 4 3
A = 6.158E-4 m I1 = I2 = 3.0E-8 m w = 2.65E4 N/m
2
E = 7.1E10 N/m = 0.33
The w term is the weight density and must be converted to mass density m for consistency of units.
PARAM,WTMASS is used to convert this weight density to mass density.
2
WTMASS = 1 g = 1 9.81 = 0.102 sec m where g is the acceleration of gravity in m/sec 2 . Therefore,
3
. The nonstructural weight of 2.414 N/m is added to
m = w WTMASS = 2.65E4 0.102 = 2703 kg/m
the beam. This nonstructural weight per length is also scaled by PARAM,WTMASS.
The example illustrates normal modes analysis (SOL 103) using the Sturm modified inverse power
method (METHOD = SINV on the EIGR entry). Mass normalization (the default) is chosen for the
eigenvectors. All frequencies between 0 and 50 Hz are requested. Two models are run. In the first
model, manufacturing tolerances make the cross section slightly out of round, making I1 and I2 slightly
different. In the second model, the cross section is perfectly round, making I1 and I2 identical.
Consider the first model. Due to the manufacturing tolerances, I1 = 2.9E-8 m 4 and I2 = 3.1E-8 m 4 . The
input file is shown in Listing 3-7. (See Install_dir/mdxxxx/doc/dynamics/bd03bar1.dat)
Main Index
CHAPTER 3 75
Real Eigenvalue Analysis
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10.$
$
$EIGR SID METHOD F1 F2 NE ND +EIG
$+EIG NORM G C
EIGR 10 SINV 0. 50.
$
$ ALUMINUM PROPERTIES:
$ E = 7.1E10 N/m**2, NU = 0.33, RHO = 2.65E4 N/m**3 (W E I G H T DENSITY)
$MAT1 MID E G NU RHO
MAT1 1 7.1+10 0.33 2.65+4
$
$ CONVERT WEIGHT TO MASS: MASS = (1/G)*WEIGHT
$ G = 9.81 m/sec**2 --> WTMASS = 1/G = 0.102
PARAM WTMASS 0.102
$
$ I1 AND I2 SLIGHTLY DIFFERENT DUE TO MANUFACTURING TOLERANCE
$ ADD NONSTRUCTURAL WEIGHT OF 2.414 N/M
PBAR 1 1 6.158-4 2.9-8 3.1-8 6.-8 2.414
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10.$
$
CBAR 1 1 1 2 0. 1. 0.
CBAR 2 1 2 3 0. 1. 0.
CBAR 3 1 3 4 0. 1. 0.
CBAR 4 1 4 5 0. 1. 0.
CBAR 5 1 5 6 0. 1. 0.
CBAR 6 1 6 7 0. 1. 0.
CBAR 7 1 7 8 0. 1. 0.
CBAR 8 1 8 9 0. 1. 0.
CBAR 9 1 9 10 0. 1. 0.
CBAR 10 1 10 11 0. 1. 0.
GRID 1 0.0 0. 0.
GRID 2 0.3 0. 0.
GRID 3 0.6 0. 0.
GRID 4 0.9 0. 0.
GRID 5 1.2 0. 0.
GRID 6 1.5 0. 0.
GRID 7 1.8 0. 0.
GRID 8 2.1 0. 0.
GRID 9 2.4 0. 0.
GRID 10 2.7 0. 0.
GRID 11 3.0 0. 0.
SPC1 1 123456 1
$
ENDDATA
The first two resulting y-direction modes are illustrated in Figure 3-7. Displacements in the y-direction
displacements are controlled by the I1 term. Because the structure is also free to displace in the z-
direction, similar modes occur in that direction and are controlled by the I2 term.
Main Index
76 MD Nastran Dynamic Analysis User’s Guide
Examples
Printed output is shown in Listing 3-8. Note that modes 1 and 3 are y-direction (T2) modes and modes 2
and 4 are z-direction (T3) modes.
Main Index
CHAPTER 3 77
Real Eigenvalue Analysis
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 1.629657E+02 1.276580E+01 2.031740E+00 1.000000E+00 1.629657E+02
2 2 1.742047E+02 1.319866E+01 2.100632E+00 1.000000E+00 1.742047E+02
3 3 6.258656E+03 7.911166E+01 1.259101E+01 1.000000E+00 6.258656E+03
4 4 6.690287E+03 8.179417E+01 1.301795E+01 1.000000E+00 6.690287E+03
5 6 4.809111E+04 2.192968E+02 3.490218E+01 1.000000E+00 4.809111E+04
6 5 5.140773E+04 2.267327E+02 3.608563E+01 1.000000E+00 5.140773E+04
EIGENVALUE = 1.629657E+02
CYCLES = 2.031740E+00 R E A L E I G E N V E C T O R N O . 1
EIGENVALUE = 1.742047E+02
CYCLES = 2.100632E+00 R E A L E I G E N V E C T O R N O . 2
EIGENVALUE = 6.258656E+03
CYCLES = 1.259101E+01 R E A L E I G E N V E C T O R N O . 3
EIGENVALUE = 6.690287E+03
CYCLES = 1.301795E+01 R E A L E I G E N V E C T O R N O . 4
Now, consider the second model for which I1 and I2 are identical. Printed output is shown in Listing 3-9.
Note that modes 1 and 2; 3 and 4; 5 and 6; etc., have identical frequencies; this is a case of repeated roots.
Main Index
78 MD Nastran Dynamic Analysis User’s Guide
Examples
(See Install_dir/mdxxxx/doc/dynamics/bd03bar2.dat) Also note that the eigenvectors are not pure y- or
pure z-translation (as they were in the first model); the eigenvectors are linear combinations of the y and
z modes since this model has repeated roots.
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 1.685851E+02 1.298403E+01 2.066473E+00 1.000000E+00 1.685851E+02
2 3 1.685851E+02 1.298403E+01 2.066473E+00 1.000000E+00 1.685851E+02
3 2 6.474471E+03 8.046410E+01 1.280626E+01 1.000000E+00 6.474471E+03
4 5 6.474471E+03 8.046410E+01 1.280626E+01 1.000000E+00 6.474471E+03
5 4 4.974941E+04 2.230458E+02 3.549883E+01 1.000000E+00 4.974941E+04
6 6 4.974941E+04 2.230458E+02 3.549883E+01 1.000000E+00 4.974941E+04
7 7 1.870792E+05 4.325266E+02 6.883875E+01 1.000000E+00 1.870792E+05
8 8 1.870792E+05 4.325266E+02 6.883875E+01 1.000000E+00 1.870792E+05
EIGENVALUE = 1.685851E+02
CYCLES = 2.066473E+00 R E A L E I G E N V E C T O R N O . 1
Main Index
CHAPTER 3 79
Real Eigenvalue Analysis
This second model was rerun, changing the mass from lumped (the default) to coupled by adding
PARAM,COUPMASS,1 to the Bulk Data. (See Install_dir/mdxxxx/doc/dynamics/bd03bar3.dat) The
resulting frequencies are shown below and are compared to those of the lumped mass model and the
theoretical results. Note that the frequency difference is greater at higher frequencies. For most
production-type models (i.e., complex three-dimensional structures), this difference is negligible.
Frequencies (Hz)
Theory Lumped Mass Model Coupled Mass Model
2.076 2.066 2.076
13.010 12.806 13.010
36.428 35.499 36.437
71.384 68.838 71.451
Bracket Model
This example is a steel bracket as shown in Figure 3-8.
Main Index
80 MD Nastran Dynamic Analysis User’s Guide
Examples
A concentrated mass is suspended from the center of the hole in bracket. This mass has the following
properties:
2
m = 0.0906 lb-sec /in
2
I11 = 0.35 in-lb-sec
2
I22 = 0.56 in-lb-sec
2
I33 = 0.07 in-lb-sec
The concentrated mass (grid point 999) is connected to the bracket by an RBE2 element connecting 24
grid points, as shown in Figure 3-9.
The bracket is clamped by constraining six degrees-of-freedom for each of 12 grid points near the base.
This example illustrates a normal modes analysis (SOL 103) using the Lanczos method (EIGRL entry).
All frequencies below 100 Hz are requested. The MODES Case Control command is used to specify the
number of times a subcase is repeated and therefore enables different output requests for each mode. The
output requests for this problem are eigenvectors for all modes (DISPLACEMENT = ALL above the
subcase level), corner stresses for the first two modes (STRESS(CORNER) = ALL and MODES = 2 in
Subcase 1), and element strain energies for the third mode (ESE = ALL). An abridged version of the
input file is shown in Listing 3-10. (See Install_dir/mdxxxxmdxxxx/doc/dynamics/bd03bkt.dat)
Main Index
CHAPTER 3 81
Real Eigenvalue Analysis
Main Index
82 MD Nastran Dynamic Analysis User’s Guide
Examples
eigenvectors are printed for each mode, the element corner stresses are printed for the first and second
modes, and the element strain energies are printed for the third mode. (Only the headers are shown in
the figure in order to save space.)
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 3.930304E+03 6.269214E+01 9.977763E+00 1.000000E+00 3.930304E+03
2 2 2.878402E+04 1.696586E+02 2.700200E+01 1.000000E+00 2.878402E+04
3 3 5.498442E+04 2.344876E+02 3.731985E+01 1.000000E+00 5.498442E+04
EIGENVALUE = 3.930304E+03
CYCLES = 9.977763E+00 R E A L E I G E N V E C T O R N O . 1
EIGENVALUE = 2.878402E+04
CYCLES = 2.700200E+01 R E A L E I G E N V E C T O R N O . 2
EIGENVALUE = 5.498442E+04
CYCLES = 3.731985E+01 R E A L E I G E N V E C T O R N O . 3
EIGENVALUE = 3.930304E+03
S T R E S S E S I N Q U A D R I L A T E R A L E L E M E N T S ( Q U A D 4 ) OPTION = CUBIC
EIGENVALUE = 2.878402E+04
S T R E S S E S I N Q U A D R I L A T E R A L E L E M E N T S ( Q U A D 4 ) OPTION = CUBIC
E L E M E N T S T R A I N E N E R G I E S
The deformed shape resulting from the first mode is illustrated in Figure 3-10 and is overlaid on the
undeformed shape. Figure 3-11 illustrates the stress contours plotted on the deformed shape of the
second mode. The element strain energy contour plot for the third mode is shown in Figure 3-12.
Main Index
CHAPTER 3 83
Real Eigenvalue Analysis
Main Index
84 MD Nastran Dynamic Analysis User’s Guide
Examples
Figure 3-12 Element Strain Energy Contours for the Third Mode
Main Index
CHAPTER 3 85
Real Eigenvalue Analysis
Main Index
86 MD Nastran Dynamic Analysis User’s Guide
Examples
Main Index
CHAPTER 3 87
Real Eigenvalue Analysis
Main Index
88 MD Nastran Dynamic Analysis User’s Guide
Examples
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 2.346479E+03 4.844047E+01 7.709540E+00 1.000000E+00 2.346479E+03
2 2 2.654886E+03 5.152559E+01 8.200551E+00 1.000000E+00 2.654886E+03
3 3 3.769821E+03 6.139887E+01 9.771934E+00 1.000000E+00 3.769821E+03
4 4 4.633242E+03 6.806792E+01 1.083335E+01 1.000000E+00 4.633242E+03
5 5 5.078395E+03 7.126286E+01 1.134184E+01 1.000000E+00 5.078395E+03
6 6 8.485758E+03 9.211817E+01 1.466106E+01 1.000000E+00 8.485758E+03
7 7 2.805541E+04 1.674975E+02 2.665805E+01 1.000000E+00 2.805541E+04
8 8 5.350976E+04 2.313218E+02 3.681600E+01 1.000000E+00 5.350976E+04
9 9 5.940912E+04 2.437399E+02 3.879240E+01 1.000000E+00 5.940912E+04
10 10 8.476198E+04 2.911391E+02 4.633622E+01 1.000000E+00 8.476198E+04
11 11 9.134271E+04 3.022296E+02 4.810133E+01 1.000000E+00 9.134271E+04
12 12 9.726959E+04 3.118807E+02 4.963736E+01 1.000000E+00 9.726959E+04
E L E M E N T S T R A I N E N E R G I E S
E L E M E N T S T R A I N E N E R G I E S
Mode shapes for modes 7, 8, 9, and 10 are shown in Figure 3-14. Mode 7 is an overall twisting mode;
mode 8 is a “roof collapse” mode; mode 9 is a local (front) roof mode; and mode 10 is a local rear mode.
Plots such as these, in conjunction with element strain energies, help to illustrate each of the mode shapes.
Main Index
CHAPTER 3 89
Real Eigenvalue Analysis
Main Index
90 MD Nastran Dynamic Analysis User’s Guide
Examples
Main Index
CHAPTER 3 91
Real Eigenvalue Analysis
$
$GRDSET, ,CP, , , , CD, PS
GRDSET , , , , , , , 456
$
$EIGRL, SID, V1, V2, ND, MSGLVL, MAXSET, SHFSCL, NORM
EIGRL , 1, , , 6
$
$MAT1, MID, E, G, NU, RHO
MAT1 , 3, 1.00E7, , 0.334, 2.5383-4
$
... basic model ...
$
ENDDATA
Figure 3-16 shows the first four mode shapes. The first mode is a bending mode, the second and third
modes are twist modes, and the fourth mode is a bending mode.
Main Index
92 MD Nastran Dynamic Analysis User’s Guide
Examples
Lines of Symmetry
ss ss
ss ss
ss ss
ss ss
Full Model
1
z
y ss
2
x
ss
Quarter Model
Figure 3-17 Derivation of Quarter Plate Model
The BC Case Control command identifies multiple boundary conditions. The SPCADD Bulk Data entry
defines a union of SPC sets.
Listing 3-17 Input File (Abridged) for the Quarter Plate Model
$ FILE bd03plt1.dat
$
$ QUARTER PLATE MODEL
$ CHAPTER 3, NORMAL MODES
$
SOL 103 $ NORMAL MODES ANALYSIS
TIME 10
CEND
TITLE = SIMPLY SUPPORTED PLATE USING SYMMETRY
Main Index
CHAPTER 3 93
Real Eigenvalue Analysis
Figure 3-18 shows the quarter plate mode shapes and the corresponding mode shapes for a full model of
the same structure. Note that the quarter plate modes match the full plate modes. (See
Install_dir/mdxxxx/doc/dynamics/bd03plt2.dat)
Main Index
94 MD Nastran Dynamic Analysis User’s Guide
Examples
Figure 3-18 Mode Shapes for the Quarter Model (Left) and Full Model (Right)
DMIG Example
This example illustrates the use of a DMIG entry to input external mass and stiffness. The cantilever
beam model shown in Figure 3-19 is used for this purpose. The model consists of CBAR elements 1
through 4. Element 5 is a model from another subcontractor that is input via DMIG entries. The model
contains two DOFs (R2 and T3) per grid point.
Main Index
CHAPTER 3 95
Real Eigenvalue Analysis
X
1 2 3 4 5 6
1 2 3 4 5
= Grid Point ID
L = Element ID
Figure 3-19 Planar Cantilever Beam
The stiffness and mass matrices from the contractor for element 5 are as follows:
5, 3 5, 5 6, 3 6, 5
5, 3 500038.8
K5 = 5, 5 – 250019.4 166679.6
6, 3 – 500038.8 250019.4 500038.8
6, 5 – 250019.4 83339.8 250019.4 166679.6
5, 3 5, 5 6, 3 6, 5
5, 3 3.5829 0. 0. 0.
M5 = 5, 5 0. 0. 0. 0.
6, 3 0. 0. 3.5829 0.
6, 5 0. 0. 0. 0.
Since the matrices are symmetric, only the lower or upper triangular portion of the matrices need to be
provided via the DMIG entries.
The corresponding input file is shown in Listing 3-19 see
Install_dir/mdxxxx/doc/dynamics/bd03dmi.dat.
Main Index
96 MD Nastran Dynamic Analysis User’s Guide
Examples
SPC = 10
$
$ SPECIFY K2GG AND M2GG
K2GG = EXSTIF
M2GG = EXMASS
$
METHOD = 10
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10..
...$
$
$EIGRL SID V1 V2 ND
EIGRL 10 2
$
CBAR 1 1 1 2 10
CBAR 2 1 2 3 10
CBAR 3 1 3 4 10
CBAR 4 1 4 5 10
$
$ HEADER ENTRY FOR STIFFNESS
DMIG EXSTIF 0 6 1
$
DMIG EXSTIF 5 3 5 3 500039. +000001
++0000015 5 -250019. 6 3 -500039. +000002
++0000026 5 -250019. +000003
$
DMIG EXSTIF 5 5 5 5 166680. +000004
++0000046 3 250019. 6 5 83340. +000005
$
DMIG EXSTIF 6 3 6 3 500039. +000006
++0000066 5 250019. +000007
$
DMIG EXSTIF 6 5 6 5 166680.
$
$ HEADER ENTRY FOR MASS
DMIG EXMASS 0 6 1
$
$ DATA ENTRIES FOR MASS
$
DMIG EXMASS 5 3 5 3 3.5829
DMIG EXMASS 6 3 6 3 3.5829
$
GRID 1 0. 0. 0. 1246
GRID 2 1. 0. 0. 1246
GRID 3 2. 0. 0. 1246
GRID 4 3. 0. 0. 1246
GRID 5 4. 0. 0. 1246
GRID 6 5. 0. 0. 1246
GRID 10 0. 0. 10. 123456
MAT1 1 7.1+10 .33 2700.
Listing 3-19 Input File for the DMIG Example
Main Index
CHAPTER 3 97
Real Eigenvalue Analysis
In this example, EXSTIF is chosen as the name of the input stiffness matrix. Therefore, in order to bring
in this stiffness matrix and add it to the global stiffness matrix, the Case Control command K2GG =
EXSTIF is required. Note that these stiffness terms are additions to any existing terms in the global
stiffness matrix at that location and not replacements of the stiffness terms at that location. In the Bulk
Data Section, five DMIG entries are required—one for the header entry and four for the data column
entries since there are four non-null columns in the above matrix.
For the header entry, the same name EXSTIF must be used to match the name selected in the Case
Control Section. The third field is “0”, which must be the value used for the header entry. The fourth
field (IFO) is set to “6” to denote a symmetric matrix input. The fifth field (TIN) is set to “1” to denote
that the matrix is real, single precision.
The terms in the matrix are referenced in terms of their external grid IDs when using the DMIG entries.
Physically, each term in a particular column of the stiffness matrix ( K ij ) represents the induced reactive
load in the i-th degree-of-freedom due to a unit displacement in the j-th direction with all other
displacement degrees-of-freedom held to zero. Since the matrix is symmetric, only the lower triangular
portion of the matrix is input.
The first DMIG data column entry defines the first column of the above matrix. Field 2 of this DMIG
entry must have the same name EXSTIF as referenced by the Case Control K2GG = EXSTIF command.
Fields 3 and 4 of this entry identify this column in terms of its external grid ID and corresponding degree-
of-freedom, respectively. In this case, it is grid point5, degree-of-freedom 3 (z-translation at grid point
5). Once this column is defined, follow the format description as described in the section for column
data entry format, and then you can input the four terms in this column row-by-row. These four terms
are defined by sets of three fields. They are the external grid ID number, corresponding degree-of-
freedom, and the actual matrix term, respectively. The first term of column one is defined by external
grid ID 5, degree-of-freedom 3 (z-translation at grid point 5) with a stiffness value of 500039. The
second term of column one is defined by external grid ID 5, degree-of-freedom 5 (y-rotation at grid point
5) with a stiffness value of -250019. The third term of column one is defined by external grid ID 6,
degree-of-freedom 3 (z-translation at grid point 6) with a stiffness value of -500039. The fourth term of
column one is defined by external grid ID 6, degree-of-freedom 5 (y-rotation at grid point 6) with a
stiffness value of -250019.
The next DMIG entry defines the second column of the above matrix. Fields 3 and 4 of this entry identify
this column in terms of its external grid ID and corresponding degree-of-freedom, respectively. In this
case, it is grid point 5, degree-of-freedom 5 (y-rotation at grid point 5). The rest of the procedure is
similar to that of column one with the exception that only three terms need to be input due to symmetry.
The next two DMIG entries defines columns three and four of the stiffness matrix, respectively. Note
that due to symmetry, one less row needs to be defined for each additional column.
The mass matrix is input in a similar manner as the stiffness matrix with the following exceptions:
• The command M2GG = EXMASS instead of K2GG = EXSTIF is used in the Case Control
Section. In this case, EXMASS is the name of the mass matrix referenced in field two of the
DMIG Bulk Data entries.
Main Index
98 MD Nastran Dynamic Analysis User’s Guide
Examples
• The matrix defined in the DMIG entries is expressed in the mass matrix terms rather than in
stiffness matrix terms.
• Since there are only two non-null columns for the mass matrix, only two DMIG data entries are
required instead of the four entries needed for the stiffness matrix.
Mass matrices input using DMIG are not scaled by PARAM,WTMASS. PARAM,CM2,xx.xxx can be
used to scale M2GG matrices.
In this example the small-field input format is used, and the maximum number of characters that can be
input are eight (including sign and decimal point). Greater input precision can be achieved by using the
large-field format and by changing the TIN field to 2 for the DMIG entries.
The first two computed natural frequencies for this example are 1.676 Hz and 10.066 Hz.
Main Index
CHAPTER 3 99
Real Eigenvalue Analysis
Rigid-body Modes
Rigid-body modes occur in unconstrained structures, such as satellites and aircraft in flight. For a
general, unconstrained 3-D structure without mechanisms, there are six rigid-body modes often
described as T1, T2, T3, R1, R2, and R3, or combinations thereof. Rigid-body modes can also be
approximated for certain kinds of dynamic or modal tests in which the test specimen is supported by very
flexible supports, such as bungee cords or inflatable bags. In this case the test specimen itself does not
distort for the lowest mode(s) but instead displaces as a rigid body with all of the deformation occurring
in the flexible support. Rigid-body modes can improperly occur if a structure that should be constrained
is not fully constrained (for example, in a building model for which the boundary conditions (SPCs) were
forgotten).
Mechanism Modes
A mechanism mode occurs when a portion of the structure can displace as a rigid body, which can occur
when there is an internal hinge in the structure. An example of a mechanism is a ball and socket joint or
a rudder in an airplane. A mechanism mode can also occur when two parts of a structure are improperly
joined. A common modeling error resulting in a mechanism is when a bar is cantilevered from a solid
element; the bar has rotational stiffness and the solid has no rotational stiffness, resulting in a pinned
connection when the two are joined.
The presence of rigid-body and/or mechanism modes is indicated by zero frequency eigenvalues. Due
to computer roundoff, the zero frequency eigenvalues are numerical zeroes on the order of 1.0E-4 Hz or
less for typical structures. The same unconstrained model may give different values of the rigid-body
frequencies when run on different computer types.
Rigid-body modes generated by MD Nastran are orthogonal with respect to each other and with respect
to the flexible (or elastic) modes. Each rigid-body mode has the following property:
Main Index
100 MD Nastran Dynamic Analysis User’s Guide
Rigid Body Modes
T
rig M rig 0
(3-19)
T
rig K ri g = 0
where rig denotes the rigid-body mode shapes, M denotes the mass matrix, and K denotes the
stiffness matrix. The rigid-body modes contain no strain energy.
Figure 3-20 depicts the rigid-body modes for a two-dimensional model of a straight beam. (See
Install_dir/mdxxxx/doc/dynamics/bd03rmbar1.dat) Because the two-dimensional model has three DOFs
per grid point (x-translation, y-translation, and z-rotation), there are three rigid-body modes. The figure
shows classical rigid-body modes, whereby one mode is purely x-translation, another is purely
y-translation, and another is purely z-rotation about the center of the beam. Because rigid-body modes
are a special case of repeated roots, any linear combination of these displacement shapes also comprises
a valid set of rigid-body modes.
Main Index
CHAPTER 3 101
Real Eigenvalue Analysis
SUPORT Entry
Rigid-body modes are computed in MD Nastran without requiring special user intervention, although the
use of a SUPORT Bulk Data entry makes the mode shapes look cleaner. The SUPORT entry also makes
the rigid-body mode shapes repeatable when the mass or stiffness of the model changes and the model
is reanalyzed. The SUPORT (note spelling!) entry does not constrain the model; it simply provides a
frame of reference for the rigid-body shape calculations by defining the r-set components of motion. The
SUPORT entry is not required for any of the dynamic analysis methods except for response spectrum
analysis (see Response Spectrum Analysis (Ch. 9)).
If the SUPORT is used, one DOF should be specified for each rigid-body mode or mechanism. The
format of the Bulk Data entry SUPORT is shown below:
1 2 3 4 5 6 7 8 9 10
SUPORT ID1 C1 ID2 C2 ID3 C3 ID4 C4
Field Contents
IDi Grid or scalar point identification number.
Ci Component numbers. (0 or blank for scalar points and any unique combination of the integers
1 through 6 for grid points).
Each of the eigenvalue extraction methods selected on the EIGR Bulk Data entry (AGIV, AHOU, GIV,
HOU, INV, MGIV, MHOU, and SINV) treats the SUPORT in the same manner. Eigenvalues are first
computed using the information on the EIGR Bulk Data entry. The frequencies of the first N r modes
(where N r is the number of SUPORT DOFs) are replaced with a value of 0.0 Hz. The first N r
eigenvectors are replaced by modes that are calculated by moving each SUPORT DOF a unit distance
and then mass orthogonalizing them with respect to the other modes. The fact that the eigenvectors are
calculated via kinematics is the reason that the SUPORT entry produces cleaner rigid-body modes (as
opposed to the rigid-body modes computed without the use of the SUPORT entry).
Note that MD Nastran has no built-in checks to warn if any of the N r eigenvalues are not rigid-body
modes. This replacement can hide potential modeling problems. The results from UIM 3035 should be
carefully checked (see Rigid Body Mode Examples, 107). The strain energy term for each SUPORT DOF
should be zero.
A poor choice of DOFs on the SUPORT entry can adversely affect the computation of the flexible modes
for the INV and SINV methods. Flexible modes computed with the other methods (AGIV, AHOU, GIV,
HOU, MGIV, and MHOU) are not adversely affected by a poor choice of SUPORT DOFs. Again, the
Main Index
102 MD Nastran Dynamic Analysis User’s Guide
SUPORT Entry
results of UIM 3035 should be carefully checked to ensure that a proper choice of SUPORT DOFs is
made.
Lanczos Method
In the Lanczos method, the SUPORT entry attempts to give 0.0 Hz modes. The Lanczos routine
compares the values of the calculated modes (without the SUPORT entry) and determines if the
calculated frequencies for the N r modes are near 0.0 Hz. If the computed frequencies are near 0.0 Hz,
the computed values are replaced with 0.0 Hz. If a computed frequency is not near 0.0 Hz, then it is
retained. (Note that this may cause problems in response spectrum analysis where 0.0 Hz is required for
the SUPORT frequencies.) The Lanczos-computed, rigid-body eigenvectors are used, i.e., the rigid-body
eigenvectors computed by the SUPORT entry are not used.
In the above paragraph, “near” means that the eigenvalues are less than 0.01 times the shift scale (the
SHFSCL field on the EIGRL entry). All computed eigenvalues less than this threshold are candidate
rigid-body modes. There is some logic to determine the relationship of these candidate rigid-body modes
to the number of DOFs on the SUPORT entry. Suppose that there are three eigenvalues less than the
threshold of 0.01 times the shift scale. If your SUPORT entry defines two DOFs, then the first two
frequencies become 0.0 Hz, and the third becomes something that is nonzero but small (on the order of
1.0E-6, for example). On the other hand, if your SUPORT entry defines four DOFs, then only the first
three are treated as rigid-body modes, and the fourth is not replaced. Furthermore, the use of a SUPORT
entry forces an extra decomposition, which increases the computer run time.
Because the SUPORT entry is not used to compute the rigid-body eigenvectors, there is probably little
to be gained by using the SUPORT entry with the Lanczos method unless response spectrum analysis is
being performed.
ul
ua = (3-20)
ur
Kl l Kl r ul 0
= P (3-21)
K rl K rr u r r
Main Index
CHAPTER 3 103
Real Eigenvalue Analysis
ul = D ur (3-22)
–1
where: D = – K l l K lr
D
rig = (3-23)
Ir
where rig represents the motion of the a-set for a unit motion of each SUPORT DOF with all other
SUPORT DOFs constrained and I r is an r r identity matrix.
T D T D
M r = rig M a a ri g = Ma a
Ir Ir
To improve the quality of the rigid-body mode shapes, orthogonalization is applied to create a diagonal
mass matrix M o by
T
M o = ro M r = ro (3-24)
This transformation matrix is used to construct the final set of rigid-body mode shape vectors by
D ro
ri g = ri g ro = (3-25)
ro
such that
T
ri g K a a rig 0 (3-26)
T
where: ri g M a a ri g = Mo
Mo is a diagonal matrix.
Care must be taken when selecting SUPORT DOFs. Each SUPORT DOF must be able to displace
independently without developing internal stresses. In other words, the SUPORT DOFs must be
statically determinate. The SUPORT is used only to facilitate the calculation of rigid-body vectors. If
you do not specify the r-set DOFs, the rigid-body modes are calculated directly by the method selected
for the flexible frequency modes. If an insufficient number of r-set DOFs are specified, the calculation
of rigid-body modes is unreliable.
Main Index
104 MD Nastran Dynamic Analysis User’s Guide
SUPORT Entry
As a modeling aid, MD Nastran calculates equivalent internal strain energy (work) for each rigid-body
vector as follows:
K ll K lr D
X = D
T
Ir (3-27)
K rl K r r Ir
When r-set DOFs exist, the printed strain energies are the diagonal elements of X divided by 2 and
should be approximately zero.
Note that X is the transformation of the stiffness matrix K aa to r-set coordinates, which by definition
of rigid-body (i.e., zero frequency) vector properties should be null. If this is not the case, the equilibrium
may be violated by the r-set choice or other modeling errors may exist. The matrix X is also called the
rigid-body check matrix.
MD Nastran also calculates a rigid-body error ratio
T
K r r + K lr D
= ---------------------------------------------------
- (3-29)
K rr
2
where: = Euclidean norm of the matrix = xij
i j
One value of is calculated using Eq. (3-29) based on all SUPORT DOFs. Therefore, in UIM 3035 the
same is printed for every supported DOF.
The rigid-body error ratio and the strain energy should be zero if a set of statically determinate SUPORT
DOFs is chosen. Roundoff error may lead to computational zero values for these quantities.
(“Computational zero” is a small number ( 10 – 5 , for example) that normally is 0.0 except for numerical
roundoff.) The rigid-body error ratio and strain energy may be significantly nonzero for any of the
following reasons:
• Roundoff error accumulation.
• The u r set is overdetermined, leading to redundant supports. The condition gives high strain
energy.
• The u r set is underspecified, leading to a singular reduced-stiffness matrix and a MAXRATIO
error. This condition gives a high rigid-body error ratio.
• The multipoint constraints are statically indeterminate. This condition gives high strain energy
and a high rigid-body error ratio.
Main Index
CHAPTER 3 105
Real Eigenvalue Analysis
• There are too many single-point constraints. This condition gives high strain energy and a high
rigid-body error ratio.
• is null. This condition gives a unit value for the rigid-body error but low strain energy.
K rr
This is an acceptable condition and may occur when generalized dynamic reduction is used (see
Special Topics, 359).
Plane 1
2
4
Plane 3
Plane 2 5 6
1, 2, 3 Normal to Plane 1
4, 5 Normal to Plane 2
6 Normal to Plane 3
Figure 3-21 Statically Determinate r-set
There are special cases where a model need not have six rigid-body modes. A planar model has only
three rigid-body modes, while an airplane with a free rudder has seven, for example. If you use the
SUPORT, it is your responsibility to determine all the modes of rigid-body motion, then provide r-set
DOFs that define these rigid-body modes. Another special case is the application of enforced motion by
the large mass technique (see Enforced Motion, 281). If the input points describe redundant load paths,
Main Index
106 MD Nastran Dynamic Analysis User’s Guide
SUPORT Entry
diagnostics are produced that indicate overconstraint. For this case, these diagnostics may be safely
ignored.
Poorly-constrained rigid-body modes result from either constraining DOFs with relatively little stiffness
or from constraining a set of DOFs that are almost linearly dependent on one another. An example of the
former is a model of a very thin cylindrical shell. The degrees-of-freedom normal to the shell and their
associated bending degrees-of-freedom may all be too soft to avoid numerical conditioning problems. A
modeling cure for this condition is to connect many grid points to a new reference grid point with an
RBE3 element and then to place the reference grid point in the r-set. The RBE3 element does not affect
the flexible modes when applied in this manner.
An example of a structure whose r-set shows poor linear independence is a slightly curved bar modeled
using coordinate systems that follow the curve of the bar such that the x-axis is always tangent to the bar.
The x DOFs at each end of the bar describe linearly independent DOFs in a mathematical sense.
However, numerical truncation produces poor conditioning if the angle between the ends is less than a
few degrees. This condition is detected by the automatic diagnostics discussed earlier. This problem can
be corrected (or better yet, avoided) by making a careful sketch of all r-set DOFs, including their
locations in space and the orientation of their global coordinates. Then apply the three-plane test
described earlier.
Using a physical analogy, a good r-set can be chosen by finding one grid point that sustains all possible
loadings well if it is tied to ground in an actual hardware test. If there is no such grid point, the ties to
ground should be spread over enough grid points to sustain the loads without damaging the structure. An
RBE3 element used for this purpose can provide good rigid-body modes without affecting the flexible
modes.
Main Index
CHAPTER 3 107
Real Eigenvalue Analysis
Z
Figure 3-22 Unconstrained Beam Model
Modes are computed using two methods (Lanczos and SINV), with and without a SUPORT entry. The
SUPORT entry is used in three ways:
• Statically determinate (grid point 1, components 1, 2, and 6)
• Underdetermined (grid point 1, components 1 and 2)
• Overdetermined (grid point 1, components 1, 2, and 6, plus grid point 11, component1)
Main Index
108 MD Nastran Dynamic Analysis User’s Guide
Rigid Body Mode Examples
Listing 3-20 shows a portion of the input file for the statically determinate SUPORT and the Lanczos
method. (See Install_dir/mdxxxx/doc/dynamics/bd03rmbar2.dat)
Table 3-5 lists the computed frequencies. The overdetermined run for the Lanczos method works well;
the same run for the SINV method gives an extra zero frequency mode that does not really exist. The
overdetermined runs have redundant SUPORTs in the x-direction; therefore, two rigid-body modes are
computed in this direction when using the SINV method. In all cases the flexible (greater than 0)
frequencies are correct.
Main Index
CHAPTER 3 109
Real Eigenvalue Analysis
Listing 3-21 shows the output for the rigid-body modes computed for the SINV method when using no
SUPORT. The rigid-body frequencies are denoted by computational zeroes on the order of 10 – 5 Hz or
less. Note that the magnitude may be different when the same problem is run on a different computer
type.
Listing 3-22 shows the output for the rigid-body modes computed for the SINV method when using the
statically determinate SUPORT. The accuracy of the statically determinate SUPORT DOFs is verified
by the computational zeroes for epsilon and the strain energy printed in UIM 3035. Note that the three
rigid-body modes have frequencies of 0.0 Hz. The SUPORT entry provides cleaner mode shapes than
those shown in Listing 3-21 as illustrated by the purely x-translation, y-translation, and z-rotation
eigenvectors shown in Listing 3-22.
Main Index
110 MD Nastran Dynamic Analysis User’s Guide
Rigid Body Mode Examples
EIGENVALUE = 1.866121E-11
CYCLES = 6.875275E-07 R E A L E I G E N V E C T O R N O . 1
EIGENVALUE = 1.819082E-09
CYCLES = 6.788069E-06 R E A L E I G E N V E C T O R N O . 2
EIGENVALUE = 2.000299E-09
CYCLES = 7.118156E-06 R E A L E I G E N V E C T O R N O . 3
Main Index
CHAPTER 3 111
Real Eigenvalue Analysis
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 0.0 0.0 0.0 1.000000E+00 0.0
2 2 0.0 0.0 0.0 1.000000E+00 0.0
3 3 0.0 0.0 0.0 1.000000E+00 0.0
4 4 6.483918E+03 8.052278E+01 1.281560E+01 1.000000E+00 6.483918E+03
5 5 4.732272E+04 2.175379E+02 3.462222E+01 1.000000E+00 4.732272E+04
6 6 1.751285E+05 4.184836E+02 6.660372E+01 1.000000E+00 1.751285E+05
7 7 4.616299E+05 6.794335E+02 1.081352E+02 1.000000E+00 4.616299E+05
EIGENVALUE = 0.000000E+00
CYCLES = 0.000000E+00 R E A L E I G E N V E C T O R N O . 1
EIGENVALUE = 0.000000E+00
CYCLES = 0.000000E+00 R E A L E I G E N V E C T O R N O . 2
EIGENVALUE = 0.000000E+00
CYCLES = 0.000000E+00 R E A L E I G E N V E C T O R N O . 3
Listing 3-23 shows the epsilon and strain energy printed in UIM 3035 for the three SUPORT cases
(statically determinate, overdetermined, and underdetermined).
It can be seen from this example that aside from clean rigid-body vectors there is no advantage to using
a SUPORT entry to compute rigid-body modes.
Main Index
112 MD Nastran Dynamic Analysis User’s Guide
Rigid Body Mode Examples
SUPPORT PT.NO. EPSILON STRAIN ENERGY EPSILONS LARGER THAN 0.001 ARE FLAGGED WITH ASTERISKS
1 7.7496606E-17 5.5879354E-09
2 7.7496606E-17 0.0000000E+00
3 7.7496606E-17 7.1622708E-11
Overdetermined SUPORT:
SUPPORT PT.NO. EPSILON STRAIN ENERGY EPSILONS LARGER THAN 0.001 ARE FLAGGED WITH ASTERISKS
1 1.9913979E-01 7.2869660E+06 ****
2 1.9913979E-01 0.0000000E+00 ****
3 1.9913979E-01 7.1850081E-11 ****
4 1.9913979E-01 7.2869660E+06 ****
Underdetermined SUPORT:
*** USER INFORMATION MESSAGE 4158---STATISTICS FOR SYMMETRIC DECOMPOSITION OF DATA BLOCK KLL FOLLOW
SUPPORT PT.NO. EPSILON STRAIN ENERGY EPSILONS LARGER THAN 0.001 ARE FLAGGED WITH ASTERISKS
1 7.6288287E-17 5.5879354E-09
2 7.6288287E-17 -7.2759576E-12
Main Index
CHAPTER 3 113
Real Eigenvalue Analysis
The six rigid-body modes have computational zero frequencies on the order of 10 –5 Hz. Note that the
magnitudes of the rigid-body modes may be different when the same problem is run on a different
computer type. Also note that the output is sorted by the value of the eigenvalue in ascending order.
Main Index
114 MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
Many of these are also described in Debugging Dynamic Models (Ch. 14).
For the above simple model (L=10in, E=30E6, =.3, =7.4E-4 and cross section as shown), using the
above equation and NASTRAN to compute the natural frequencies result in a pretty good estimation. Of
Main Index
CHAPTER 3 115
Real Eigenvalue Analysis
course, this is for a simple structure, and complex structures won’t be as close, but it helps to know the
primary x-y plane bending mode should be near 100Hz instead of 2000 Hz.
For example, using the beam model in the previous section, the 4th mode, when deformed gives the
following type of motion (undeformed is blue, deformed is black):
Yet, using some of the tools discussed in the following sections, it is known that this mode is really a
torsional mode of the beam rotating about its own axis... obviously very different from the display
above. The remedy to this problem is to add dummy structure to the beam such that when the beam
rotates about its own axis, something else moves, showing motion somewhere other than at a point
(which is what you see when looking down a straight beam). For example, add 2 new nodes (one above
and one below) to the end of the beam and connect them with an RBE2. When this is now shown as a
deformation, you get:
Main Index
116 MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
As can be seen, there is some very small out-of-plane bending, which, lacking any other information to
display, was amplified in the previous deformation. But the deformation here is seen in a more
appropriate context with the majority of motion represented by the beam torsioning.
Always make sure straight lines and point masses have some cross-sectional connectivity to ensure the
post-processor can display the appropriate deformation.
T T
M D R = M
T
M D R = m
where m is the diagonal matrix of generalized masses for the normal modes.
Main Index
CHAPTER 3 117
Real Eigenvalue Analysis
T
= M D R = m i = mi i i
The scaling factor i multiplies the generalized mass m ii to define the participation factor i .
The relationship between the rigid body vector D R and the corresponding rigid body mass, M R , is
T
MR = D R M D R
Using D R =
Then M R = T T M = T m
So the contribution which the i-th mode provides to the rigid body mass M R is 2i m ii . This is known as
the modal effective mass. If you mass normalize the eigenvectors, T M = I , then the
participation factors are i = i , and the modal effective mass is 2i . The modal effective weight is
modal effective mass multiplied by g in the appropriate units.
The MEFFMASS Case Control command can provide this information:
The command has the following form:
SUMMARY, PARTFAC,
MEFFMASS PRINT PUNCH GRID = gid, = YES
MEFFM, MEFFW,
NOPRINT NOPUNCH NO
FRACSUM, ALL
Describer Meaning
PRINT Write output to the print file (Default).
NOPRINT Do not write output to the print file.
PUNCH Write output to the punch file.
NOPUNCH Do not write output to the punch file (Default).
gid Reference a grid point for the calculation of the rigid body mass matrix. The
default is the origin of the basic coordinate system.
SUMMARY Requests calculation of the total effective mass fraction, modal effective
mass matrix, and the rigid body mass matrix (Default).
PARTFAC Requests calculation of modal participation factors.
MEFFM Requests calculation of the modal effective mass in units of mass.
Main Index
118 MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
Describer Meaning
MEFFW Requests calculation of the modal effective mass in units of weight.
FRACSUM Requests calculation of the modal effective mass fraction.
M O D A L E F F E C T I V E M A S S S U M M A R Y
T1 T2 T3 R1 R2 R3
9.654936E-18 9.106090E-01 9.657329E-01 9.544338E-01 9.996986E-01 9.985170E-01
The Effective Mass Matrix is the 6x6 mass represented by the modes, while the A-SET Rigid Body Mass
Matrix is the actual 6x6 rigid body mass of the structure. The Total Effective Mass Fraction printout
shows how much of the total possible rigid body mass is represented by the modes. In this example, you
can see that in the T1 direction, basically no mass is represented. This means there are no modes in which
the mass of the structure is moving in the T1 direction. In this case, it would be impossible to
dynamically excite this structure in the T1 direction, and any applied loading in that direction would
Main Index
CHAPTER 3 119
Real Eigenvalue Analysis
result in no response. In the T2 and T3 directions, 91% and 96.5% of the possible mass is represented
by the first 10 modes.
If the eigensolution solves for 30 modes instead of 10, then 84.9% of the available mass is represented
in the T1 direction, while the T2 and T3 directions improve to 100%. See
Install_dir/mdxxxx/doc/dynamics/beam103m30.dat
M O D A L E F F E C T I V E M A S S S U M M A R Y
T1 T2 T3 R1 R2 R3
8.497244E-01 9.990354E-01 1.000000E+00 1.000000E+00 1.000000E+00 9.999877E-01
In addition to the Modal Effective Mass Summary, the Mass Fraction for each mode is provided. In our
BEAM example, the original visual inspection of the mode shapes indicated mode 4 to be a form of
bending mode with deformation in the T3 direction. Upon evaluating the translational modal effective
mass fractions (in the table below) we see mode 4 as having no motion in the T1 and T2 direction and
only .04% of the mass moving in the T3 direction.
MODE FREQUENCY T1 T2 T3
NO.
FRACTION SUM FRACTION SUM FRACTION SUM
As can be seen from the rotational modal effective mass fractions (in the table below), mode 4 has 66.3%
of all the mass in the R1 direction, which corresponds to the rotation about the axis for the beam. This
helps to identify mode 4 as a torsional mode, not a bending mode as the original deformation suggested.
MODE FREQUENCY R1 R2 R3
NO. FRACTION SUM FRACTION SUM FRACTION SUM
Main Index
120 MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
By reviewing the printout of the full, 30-mode solution, you can see the 23rd mode is required to get T1
mass participation. You can scan the ‘Sum’ column to see how many modes are required to achieve the
desired modal mass participation in a particular direction. If 90% modal effective mass were the desired
threshold, more modes are still needed for the T1 direction, while the T2 direction achieved the goal by
the 9th mode and the T3 direction by the 5th mode.
MODE FREQUENCY T1 T2 T3
NO. FRACTION SUM FRACTION SUM FRACTION SUM
In addition to these tables, Modal Participation Factors, Modal Effective Mass and Modal Effective
Weight are also printed when using the ‘ALL’ setting on the MEFFMASS command.
Example–two springs in series. Which is the best element to stiffen to reduce the tip deflection?
Main Index
CHAPTER 3 121
Real Eigenvalue Analysis
K1 K2
P
1 2
1 2
ESE = --- K
2
where:
K1 = 10
K2 = 1
P = 1
2
for K 1 , ESE 1 = 1--- K 1 21 = 1--- -----
P 1
- = -----
-
2 2 K1 20
P
where 1 = -----
-
K1
1 1 P P P
K 2 ESE 2 = --- K 2 2 – 1 = --- K 2 ------ + ------ – ------
2
2 2 K 1 K 2 K 1
for
1 P 2 1
= --- K 2 ------ = ---
2 K 2 2
P
1
K1
P P
2
K1 K 2
where:
P
1 = ------
K1
P P
2 = ------ + ------
K1 K2
Most of the ESE is in the smaller spring. Therefore, stiffening it is the most efficient way to reduce the
tip deflection.
The Case Control Command to trigger the calculation of element strain energy is shown below. These
results are available in the f06 file from beam103m30.dat, as well as the binary results file for plotting
as fringe results in a post-processor.
Main Index
122 MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
AVERAGE
PRINT, PUNCH
ESE ( AMPLITUDE THRESH = p = ALL
PLOT n
PEAK
Describer Meaning
PRINT Writes energies to the print file (Default).
PUNCH Writes energies to the punch file.
PLOT Do not write energies to either the punch file or the print file.
AVERAGE Requests average energy in frequency response analysis only.
AMPLITUDE Requests amplitude of energy in frequency response analysis only.
PEAK Requests peak energy for frequency response analysis only. PEAK is the sum
of AVERAGE and AMPLITUDE.
THRESH Energies for elements having an energy value of less than p% will be
suppressed in all output files: print, punch, plot, .op2, and .xdb. THRESH
overrides the value of TINY described in Remark 1. (Default = 0.001).
ALL Energy values for all elements will be computed.
n Set identification number. Energy for all elements specified on the SET n
command will be computed. The SET n command must be specified in the
same subcase as the ESE command, or above all subcases (Integer >0).
NONE Element strain energy will not be output.
The strain energies are printed, by mode, then by element type (BAR, QUAD4, etc.), percentage by each
element type, strain energy density (strain energy / volume, which removes some of the biasing of the
element size).
For the sample beam model, mode 1 (as well as the other modes) shows the most modal strain energy in
element 1, so this is where the smallest modification can be made to achieve the biggest effect.
Main Index
CHAPTER 3 123
Real Eigenvalue Analysis
E L E M E N T S T R A I N E N E R G I E S
Where mass
g indicates that the mass-normalized eigenvectors so that the total grid
point kinetic energy is scaled to be unity and the operator indicates term-wise matrix multiplication.
The Grid Point Kinetic Energy output has limited meaning for a coupled mass formulation. Because the
coupled mass formulation produces a coupling of mass across grid points, the sharing of kinetic energy
among grid points can occur. In general, this obscures the meaning of the computation.
Using the previous cantilevered beam example, looking at the f06 output for mode 1 GPKE, you can see
the largest energy is in the T3 direction and pretty well spread among the grids, pointing toward a global
mode.
Main Index
124 MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
EIGENVALUE = 3.061045E+04
CYCLES = 2.784550E+01 G R I D P O I N T K I N E T I C E N E R G Y ( P E R C E N T )
MODE NUMBER = 1
If mode 4 is examined, it can be seen that most of the energy is in the R1 direction, and pretty spread
around, indicating a global rotational mode. This agrees with the previous finding through modal
effective mass and visualization that mode 4 is a torsional mode.
EIGENVALUE = 6.278815E+06
CYCLES = 3.988035E+02 G R I D P O I N T K I N E T I C E N E R G Y ( P E R C E N T )
MODE NUMBER = 4
The Case Control entry for Grid Point Kinetic Energy is:
Requests the output of the kinetic energy at selected grid points in normal modes analysis only.
Format:
ALL
PRINT
GPKE PUNCH, THRESH = e = n
NOPRINT
NONE
Describer Meaning
PRINT The printer will be the output medium.
NOPRINT Generates, but does not print, grid point kinetic energy output.
PUNCH The punch file will be the output medium.
Main Index
CHAPTER 3 125
Real Eigenvalue Analysis
Describer Meaning
e Minimum energy threshold. Only energies above this value will be printed
and/or punched.
ALL Grid point kinetic energy for all grid points will be output.
n Set identification number of a previously appearing SET command. Only grid
points with identification numbers that appear on this SET command will be
included in output (Integer > 0).
NONE Grid point kinetic energy for no points will be output.
Weightcheck
One of the most common errors in normal modes analysis is having the incorrect units for the mass
properties. Since the program assumes consistent units, it is up to the user to provide the correct mass
properties. From a single DOF system, omega is equal the square root of k/m… therefore, large errors
in mass result in large errors in the natural frequency. If the mass density was accidentally entered as
weight density, the mass would be artificially large, resulting in modes that are artificially low.
As an overall aid in checking out the mass properties in your structure, the Case Control Command
WEIGHTCHECK can be used. The output for the WEIGHTCHECK command is similar to that
provided by PARAM, GRDPNT (Grid Point Weight Generator (Ch. 19) in the MD Nastran Linear Static
Analysis User Guide).
At each stage of the mass matrix reduction, compute rigid body mass and compare with the rigid body
mass t the g-set.
Format:
Describer Meaning
PRINT Write output to the print file (Default).
NOPRINT Do not write output to the print file.
SET Selects degree of freedom set(s) (Default SET=G).
Main Index
126 MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
Describer Meaning
gid Reference grid point for the calculation of rigid body motion. The default
is the origin of the basic coordinate system.
CGI For SET G , CGI = YES requests output of center of gravity and mass
moments of inertia (Default: CGI = NO).
WEIGHT/MASS Selects output in units of weight or mass (Default = WEIGHT).
The main difference between PARAM,GRDPNT and WEIGHTCHECK is that the weight check may be
performed at any or all stages of the mass reduction (see the SET keyword).
SET
Keywords DOF Set Description
G g-set before single point, multipoint constraints, and rigid elements are
applied
N n-set after multipoint constraints and rigid elements are applied
N+AUTO n-set with same as the n-set with the rows/columns in the mass matrix
AUTOSPC corresponding to degrees of freedom constrained by the
PARAM,AUTOSPC operation zeroed out
F F-set after single point, multipoint constraints, and rigid elements are
applied
A a-set after static condensation
V v-set Component mode mass which combines the o-set, r-set and c-set
Note: The units of weight or mass output depends on input, it is independent of param,wtmass,x.
Main Index
CHAPTER 3 127
Real Eigenvalue Analysis
Using the sample beam model, here are some different set level outputs.
0 SUBCASE 1
O U T P U T F R O M W E I G H T C H E C K
0 DEGREES OF FREEDOM SET = G
0 REFERENCE POINT = 0
M O
* 4.514000E-05 0.000000E+00 0.000000E+00 0.000000E+00 -5.128499E-12 -1.878838E-06 *
* 0.000000E+00 4.514000E-05 0.000000E+00 5.128499E-12 0.000000E+00 2.257000E-04 *
* 0.000000E+00 0.000000E+00 4.514000E-05 1.878838E-06 -2.257000E-04 0.000000E+00 *
* 0.000000E+00 5.128499E-12 1.878838E-06 3.423766E-07 -9.394190E-06 2.564249E-11 *
* -5.128499E-12 0.000000E+00 -2.257000E-04 -9.394190E-06 1.512190E-03 2.134607E-13 *
* -1.878838E-06 2.257000E-04 0.000000E+00 2.564249E-11 2.134607E-13 1.512268E-03 *
S
* 1.000000E+00 0.000000E+00 0.000000E+00 *
* 0.000000E+00 1.000000E+00 0.000000E+00 *
* 0.000000E+00 0.000000E+00 1.000000E+00 *
DIRECTION
MASS AXIS SYSTEM (S) MASS X-C.G. Y-C.G. Z-C.G.
X 4.514000E-05 0.000000E+00 4.162246E-02 -1.136132E-07
Y 4.514000E-05 5.000000E+00 0.000000E+00 -1.136132E-07
Z 4.514000E-05 5.000000E+00 4.162246E-02 0.000000E+00
I(S)
* 2.641748E-07 0.000000E+00 3.231174E-27 *
* 0.000000E+00 3.836900E-04 2.524355E-29 *
* 3.231174E-27 2.524355E-29 3.836900E-04 *
I(Q)
* 2.641748E-07 *
* 3.836900E-04 *
* 3.836900E-04 *
Q
* 1.000000E+00 0.000000E+00 0.000000E+00 *
* 0.000000E+00 1.000000E+00 0.000000E+00 *
* 0.000000E+00 0.000000E+00 1.000000E+00 *
Below, you see the printout for the N-Set. Since there are no MPC’s in this model, the N-Set is the same
as the G-Set. If it was desired to see the CG and inertia terms, CGI=YES could have been added to the
WEIGHTCHECK command.
0 SUBCASE 1
O U T P U T F R O M W E I G H T C H E C K
0 DEGREES OF FREEDOM SET = N
0 REFERENCE POINT = 0
M O
* 4.514000E-05 0.000000E+00 0.000000E+00 0.000000E+00 -5.128499E-12 -1.878838E-06 *
* 0.000000E+00 4.514000E-05 0.000000E+00 5.128499E-12 0.000000E+00 2.257000E-04 *
* 0.000000E+00 0.000000E+00 4.514000E-05 1.878838E-06 -2.257000E-04 0.000000E+00 *
* 0.000000E+00 5.128499E-12 1.878838E-06 3.423766E-07 -9.394190E-06 2.564249E-11 *
* -5.128499E-12 0.000000E+00 -2.257000E-04 -9.394190E-06 1.512190E-03 2.134607E-13 *
* -1.878838E-06 2.257000E-04 0.000000E+00 2.564249E-11 2.134607E-13 1.512268E-03 *
0 CHANGE WITH RESPECT TO G-SET RIGID BODY MASS/WEIGHT
* 0 % *
* 0 % *
* 0 % *
* 0 % *
* 0 % *
* 0 % *
Now, the F-Set printout is shown. Here you can see a change with respect to the G-Set. As you move
down the set processing, the F-Set is the result after any SPC’s have been applied. Our cantilevered beam
was fixed in all DOF at the left end, so the mass on that grid is removed.
Main Index
128 MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
O U T P U T F R O M W E I G H T C H E C K
0 DEGREES OF FREEDOM SET = F
0 REFERENCE POINT = 0
M O
* 4.288300E-05 0.000000E+00 0.000000E+00 0.000000E+00 -4.872074E-12 -1.784896E-06 *
* 0.000000E+00 4.288300E-05 0.000000E+00 4.872074E-12 0.000000E+00 2.257000E-04 *
* 0.000000E+00 0.000000E+00 4.288300E-05 1.784896E-06 -2.257000E-04 0.000000E+00 *
* 0.000000E+00 4.872074E-12 1.784896E-06 3.252578E-07 -9.394190E-06 2.564249E-11 *
* -4.872074E-12 0.000000E+00 -2.257000E-04 -9.394190E-06 1.512190E-03 2.027877E-13 *
* -1.784896E-06 2.257000E-04 0.000000E+00 2.564249E-11 2.027877E-13 1.512264E-03 *
After the F-Set, we see the A-Set. Since there were no further reduction in the sample cantilevered beam
(no ASET or OSET definitions), the A-Set is no different from the F-Set.
O U T P U T F R O M W E I G H T C H E C K
0 DEGREES OF FREEDOM SET = A
0 REFERENCE POINT = 0
M O
* 4.288300E-05 0.000000E+00 0.000000E+00 0.000000E+00 -4.872074E-12 -1.784896E-06 *
* 0.000000E+00 4.288300E-05 0.000000E+00 4.872074E-12 0.000000E+00 2.257000E-04 *
* 0.000000E+00 0.000000E+00 4.288300E-05 1.784896E-06 -2.257000E-04 0.000000E+00 *
* 0.000000E+00 4.872074E-12 1.784896E-06 3.252578E-07 -9.394190E-06 2.564249E-11 *
* -4.872074E-12 0.000000E+00 -2.257000E-04 -9.394190E-06 1.512190E-03 2.027877E-13 *
* -1.784896E-06 2.257000E-04 0.000000E+00 2.564249E-11 2.027877E-13 1.512264E-03 *
Groundcheck
A terrific model check tool, GROUNDCHECK helps identify grounding problems in the model. One of
the biggest advantages of this capability is to evaluate the model at various stages of processing. This
helps to identify at what stage the grounding is introduced, helping to pinpoint the source of the error.
While not specific to normal modes (it does not require mass to calculate) it is available in the normal
modes solution.
A grounding check is also referred to as a rigid body check because the stiffness matrix is multiplied by
the rigid body transformation matrix. The grounding check identifies unintentional constraints and ill-
conditioning in the stiffness matrix.
The grounding check is requested with the GROUNDCHECK Case Control command.
Main Index
CHAPTER 3 129
Real Eigenvalue Analysis
Perform grounding check analysis on the stiffness matrix to expose unintentional constraints by moving
the model rigidly.
Format:
Describer Meaning
PRINT Write output to the print file. (Default)
NOPRINT Do not write output to the print file.
PUNCH Write output to the punch file.
SET Selects degree-of-freedom set(s) (Default: SET=G).
gid Reference grid point for the calculation of the rigid body motion.
e Maximum strain energy which passes the check. The default value is
computed by dividing the largest term in the stiffness matrix by 1.E10.
DATAREC Requests data recovery of grounding forces (Default: DATAREC=NO).
r Grounding forces which are larger than r percent of the largest grounding
force will be printed if DATAREC=YES (Default = .10; 0. < r < 1.0).
User Information Message 7570 is issued by GROUNDCHECK for each DOF set requested. The strain
energy is computed in each direction of the rigid body motion and by default, if the strain energy exceeds
the tolerance, then “FAIL” is printed out for that direction. The tolerance is set by dividing the largest
stiffness term by 1.E10. The THRESH keyword can be used to further control the amount of output.
Possible reasons for failure are also printed after the strain energies.
From the sample beam model, the check indicates passing for the G-Set. As shown in the informational
message, common reasons for failing at this level are CELAS elements connected to only one grid, or
connected between non-coincident grids or improper DMIG matrices.
Main Index
130 MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
Looking at the F-Set, we see this test was failed. As indicated in the Information Message, constraints
in the model can cause failure, which is the case for the beam model. So, even the fact that this failed is
a good model check… had it not failed all 6 DOf, then there would have been something wrong with the
intended constraints.
If the DATAREC keyword is specified and the rigid body check fails, then data recovery will be
performed to compute and print the grounding forces to make it easier to locate the source of the failure.
By default, only those grounding forces larger than 10 percent of the largest grounding force will be
printed (see RTHRESH keyword).
Since data recovery can only be performed on DOF that are part of the set being evaluated, the grid
showing the forces will often be ‘near’ the real grid. In the beam example, since grid 1 was constrained,
that is the grid that should show up with the grounding forces. But when a grid is constrained, it is
removed for the solution set, so for the F-Set, grid 1 doesn’t exist. So grid 2 shows the grounding force.
Main Index
CHAPTER 3 131
Real Eigenvalue Analysis
G R O U N D C H E C K F O R C E S ( F - S E T )
DIRECTION 2
G R O U N D C H E C K F O R C E S ( F - S E T )
DIRECTION 3
G R O U N D C H E C K F O R C E S ( F - S E T )
DIRECTION 4
G R O U N D C H E C K F O R C E S ( F - S E T )
DIRECTION 5
G R O U N D C H E C K F O R C E S ( F - S E T )
DIRECTION 6
G R O U N D C H E C K F O R C E S ( F - S E T )
Main Index
132 MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 4: Frequency Response Analysis
Main Index
Overview
Frequency response analysis is a method used to compute structural response to steady-state oscillatory
excitation. Examples of oscillatory excitation include rotating machinery, unbalanced tires, and
helicopter blades. In frequency response analysis the excitation is explicitly defined in the frequency
domain. All of the applied forces are known at each forcing frequency. Forces can be in the form of
applied forces and/or enforced motions (displacements, velocities, or accelerations).
Phase Shift
Oscillatory loading is sinusoidal in nature. In its simplest case, this loading is defined as having an
amplitude at a specific frequency. The steady-state oscillatory response occurs at the same frequency as
the loading. The response may be shifted in time due to damping in the system. The shift in response is
called a phase shift because the peak loading and peak response no longer occur at the same time. An
example of phase shift is shown in Figure 4-1.
Complex Numbers
The important results obtained from a frequency response analysis usually include the displacements,
velocities, and accelerations of grid points as well as the forces and stresses of elements. The computed
responses are complex numbers defined as magnitude and phase (with respect to the applied force) or as
Main Index
CHAPTER 4 135
Frequency Response Analysis
real and imaginary components, which are vector components of the response in the real/imaginary
plane. These quantities are graphically presented in Figure 4-2.
i (Imaginary)
u
ui
ur r (Real)
where:
2 2
u = magnitude = u r + u i
–1
= phase angle = tan u i u r
Two different numerical methods can be used in frequency response analysis. The direct method
(SOL 108) solves the coupled equations of motion in terms of forcing frequency. The modal method
(SOL 111) utilizes the mode shapes of the structure to reduce and uncouple the equations of motion
(when modal or no damping is used or no TF, M2PP, B2PP or K2PP DMIG input is present); the solution
for a particular forcing frequency is obtained through the summation of the individual modal responses.
The choice of the method depends on the problem. The two methods are described in Direct Frequency
Response Analysis, 141 and Modal Frequency Response Analysis, 146.
Main Index
136 MD Nastran Dynamic Analysis User’s Guide
Review of Dynamic Excitations
Main Index
CHAPTER 4 137
Frequency Response Analysis
5. Subcases are used in complex eigenvalue analysis (SOLs 107 and 110) for processing multiple
direct input matrices. Typical applications are multiple control system parameters or externally
generated aerodynamic matrices representing a variety of flight conditions.
6. Cyclic symmetry analysis (SOLs 114 through 118) requires subcases to distribute the loads over
different petals (i.e., sections of the structure). The methods are identical to cyclic static analysis.
7. Multiple subcases are not recommended for linear transient analysis.
Loading Methods
An overview of the dynamic load algorithm is shown in Figure 4-3. The input data follow two major
paths: the DAREA method versus the LSEQ method. The DAREA method is the preferred method,
since the use of DAREA without LSEQ will create the equivalent static loads (FORCE, MOMENT,
SLOAD) entries automatically. The DAREA entry is an alternative to using FORCE, MOMENT or
SPCD directly referenced by an RLOADi entry as long as no LOADSET/LSEQ combination is included
in the input file. However, to select a harmonic temperature load, the LOADSET/LSEQ are still required.
Once a LOADSET/LSEQ combination is used all dynamic loading must use the LOADSET/LSEQ
except for the DAREA entry.
Main Index
138 MD Nastran Dynamic Analysis User’s Guide
Review of Dynamic Excitations
Bulk Data
LOADSET = LSEQ
Dynamic
Coefficients Static Loads
TLOADi
DLOAD = RLOADi TABLEDi
DLOAD F(t) or F(f)
Time or Frequency
Functions
Dynamic Loads
P(t) or P(f)
• The LSEQ Bulk Data act much like a set of Case Control commands to generate static loads and
assign them to a specific static load vector. These vectors are assembled, reduced, and combined
for all superelements. A different load vector may be required for each unique function of time
or frequency.
Main Index
CHAPTER 4 139
Frequency Response Analysis
• The DAREA entries used without LSEQ will be converted to the equivalent FORCE,
MOMENT, and/or SLOAD entries. This will work in dynamics and statics.
• The DAREA, DPHASE, and DELAY Bulk Data provide a direct method of distributing the
dynamic loads over the grid points. The DAREA factors correspond to a specific load factor for
a specific degree-of-freedom, much like an area under a pressure load. The DPHASE and
DELAY data are used when the same load versus time function is applied to different points at
different times. An example of a DPHASE application is the frequency response of an
automobile traveling over a highway with a sinusoidal undulation. As the speed increases the
input frequency changes, but the phase difference between the axles is constant.
The TLOADi Bulk Data inputs each define a function of time with coefficients or TABLEDi references.
The RLOADi Bulk Data inputs each define a function of load versus frequency with complex TABLEDi
inputs. The DLOAD entry is used to combine the different load functions, either time or frequency
dependent. These functions may be associated with LSEQ-type vectors or DAREA coefficients, or both.
Recommendations
The following guidelines should be observed when applying dynamic loads.
TABLEDi
1. Remember that the tables are extrapolated at each end from the first or last two points. If the load
actually goes to zero, add two points with values of y = 0.
2. Linear interpolation is used between tabular points. This may lead to accumulated roundoff and
drift errors over a long-period transient analysis. Fix this problem by adding a correcting load
function to the same points on subsequent runs.
3. If a jump occurs (two points with equal values of X), the value of Y at the jump is the average of
the two points. Numerical representation of excitation frequency from a FREQi entry must cause
the excitation frequency to be slightly above or below the jump frequency of the TABLEDi entry.
In this case, different machines and different versions may have somewhat different results at this
specified excitation frequency. To avoid this situation, jumps should not be used on the
TABLEDi entry.
Static Preloads in Frequency Response Analysis
1. A change in the model stiffness due to a static load may be included in a frequency response
analysis by input of suitable case control (see STATSUB usage) and bulk data entries for a static
subcase. In the case control use a static subcase with correct boundary conditions for a linear
static solution and select the static loads, e.g., PLOAD4. The static subcase is followed by the
frequency response subcase where new SPC and DLOAD Case Control commands set the
boundary conditions and specify the applied dynamic load via a RLOADi entry or combination
on the DLOAD Bulk Data entry. The frequency response subcase requires the insertion of the
STATSUB Case Control command to select the static subcase ID.
2. Thermal loads may be applied via the LSEQ method, but note that the dynamic stress recovery
method will not account for the additional strain or stress.
Main Index
140 MD Nastran Dynamic Analysis User’s Guide
Review of Dynamic Excitations
An MD Nastran random analysis requires a preliminary frequency response analysis to generate the
proper transfer functions that define the output/input ratios. The squared magnitudes of the results are
then multiplied by the spectral density functions of the actual loads. Normally, the inputs are unit loads
(e.g., one g constant magnitude base excitation or a unit pressure on a surface).
Fluids and Acoustic Loads
Normally, fluid and acoustic elements are not loaded directly, but are excited by the connected structures.
However, loads on these special grid points may be used to represent sources of fluid motion such as a
small loudspeaker. The actual units of these loads are the second derivative of volume flow with respect
to time. See Axisymmetric Hydroelastic Analysis (Ch. 11) for details.
Main Index
CHAPTER 4 141
Frequency Response Analysis
where u is a complex displacement vector. Taking the first and second derivatives of Eq. (4-2),
the following is obtained:
i t
x· = i u e
(4-3)
2 it
x·· = – u e
When the above expressions are substituted into Eq. (4-1), the following is obtained:
2 i t i t it i t
– M u e + i B u e + K u e = P e (4-4)
2
– M + iB + K u = P (4-5)
The equation of motion is solved by inserting the forcing frequency into the equation of motion. This
expression represents a system of equations with complex coefficients if damping is included or the
applied loads have phase angles. The equations of motion at each input frequency are then solved in a
manner similar to a statics problem using complex arithmetic.
Main Index
142 MD Nastran Dynamic Analysis User’s Guide
Direct Frequency Response Analysis
where:
1
B = damping elements (CVISC, CDAMPi and hybrid damping*) and B2GG
2
B = B2PP direct input matrix and transfer functions
*See Damping (Ch. 6) for more details on on the topic of hybrid damping a unification of vis-
cous, structural and proportional damping to the dynamic equations of motion, see the SDAMP-
ING Case Control command and the Bulk Data entry DAMPING (p. 1579) in the MD Nastran
Quick Reference Guide.
In frequency response, PARAM,G and GE on the MATi entry do not form a damping matrix. Instead,
they form the following complex stiffness matrix:
K = 1 + iG K + i G E K E (4-7)
where:
When the above parameters and/or coefficients are specified, they are automatically incorporated into the
stiffness matrix and therefore into the equation of motion for the solution. All of the forms of damping
can be used in the same analysis, and their effects are added together.
In frequency response analysis, it is not necessary to assume an equivalent viscous form for structural
damping since the solution is complex. Therefore, a complex stiffness matrix is allowed.
Main Index
CHAPTER 4 143
Frequency Response Analysis
FRF Concept
Using the usual symbols, the equation of motion for frequency response analysis is given by:
2
– M + i B + K u = Z u = P (4-8)
–1
u = Z P = HP (4-9)
The H matrix in the above equation is the receptance matrix of the system. It is also commonly
referred to as the FRF (Frequency Response Function) of the system. It should be noted that the FRF of
a system is a function of the forcing frequency . Thus there is a different FRF for the system for each
forcing frequency.
The rows of the FRF matrix represent response DOFs while its columns represent excitation DOFs. Any
term H ij of the FRF matrix represents the response displacement at the i-th DOF due to a unit load
applied to the j-th DOF. Thus, the rows in H j , the j-th column of the FRF, represent the displacements
at all response DOFs due to a unit load applied to the j-th DOF. Similarly, the columns in the i-th row of
the FRF represent the response displacements at the i-th DOF due to unit loads applied to all excitation
DOFs.
In practice, the FRF matrix is computed by solving Eq. (4-9) for a series of unit loads. Thus, consider the
following equation which represents Eq. (4-9) for a series of unit loads on the right hand side:
–1
U = Z I (4-10)
The solution of the above equation for the matrix of displacements U yields the desired FRF matrix
H .
Main Index
144 MD Nastran Dynamic Analysis User’s Guide
Frequency Response Function (FRF) and FRF Based Assembly (FBA)
The matrix of displacements U resulting from the solution of Eq. (4-10), which is equivalent to the
receptance or FRF matrix H of Eq. (4-9), is square because Eq. (4-10) assumes that the responses are
computed at all DOFs of the system and also that unit loads are applied at all DOFs. However, this is
rarely the case in practice. In most practical cases, responses are computed at only a subset of the total
points in a configuration and loads are also applied to only a small subset of the total DOFs in the system.
Therefore, in practice, the actual FRF matrix to be considered for analysis purposes is really a subset (or
smaller partition) of the matrix U given by Eq. (4-10). In general, this FRF matrix, which is what is
actually employed in FRF analysis, is rectangular and can be regarded as an (r x e) matrix where r, the
number of rows, is the number of actual response DOFs and e, the number of columns, is the number of
actual excitation DOFs.
FBA Concept
It is possible to obtain the FRFs of an assembly of components from the FRFs of the individual
components comprising the assembly. This process is referred to as FRF Based Assembly (FBA).
The FRFs of an assembly of components is computed in Nastran by coupling the FRFs of its constituent
components and enforcing conditions of displacement compatibility and force equilibrium at the
connected DOFs. The final result of the FBA process will be a set of FRF matrices for the assembly, with
each such matrix corresponding to a forcing frequency. The number of rows of each such matrix will be
equal to the total number of response DOFs of the assembly and the number of columns will be equal to
the total number of excitation DOFs of the assembly.
Using the FRFs resulting from the FBA process, the program recovers the FRFs for the individual FRF
components comprising the FRF assembly. These results will reveal clearly how excitations on one
component affect the responses at other components of the assembly.
The user should carefully review the the FRF Case Control command and Bulk Data entries before
running the SOL 108 or SOL 111 job. The following additional data item are of particular interest:
a. The DOFs where loads are to be applied must be specified either indirectly via the DLOAD
Case Control command and/or directly via the FRFXIT/FRFXIT1 Bulk Data entries. The
DLOAD Case Control command points to appropriate Bulk Data loading entries. All DOFs
with nonzero load values will have unit loads applied to them. The FRFXIT entry permits
specification of unit load for a single DOF with a label. The FRFXIT1 entry permits
specification of unit loads at multiple DOFs.
Main Index
CHAPTER 4 145
Frequency Response Analysis
b. There is no requirement that unit loading data be defined for every component for which FRFs
are generated, since some components in a configuration may not have any loads applied to
them.
c. Regardless of whether an FRF component has unit loads explicitly specified for it, as in
Remark 12(a) or not, as in Remark 12(b), the program will internally apply unit loads
automatically at all DOFs for all connection points comprising the set referenced by the
CONNPTS keyword. This ensures that correct results are obtained from subsequent FBA
processes.
d. The specific points at which FRFs are computed in an FRF generation run consist of the
following:
• All points specified via DISP, VELO, and ACCE requests
• All points associated with elements for which STRESS/FORCE requests are specified
• All points at which unit loads are applied (as per the scheme indicated in Remark 12(a)
• All points comprising the set referenced by the CONNPTS keyword
• All grid points referenced in PLOTEL Bulk Data entries
e. It is assumed in an FBA process that the FRFs of all of the FRF components have been
generated at the same forcing frequencies, and that these are also the forcing frequencies at
which the FBA process is to be performed. As a result, the FBA process derives these forcing
frequencies from the saved data of the first of the FRF components being assembled, and uses
them in the FBA process.
In order to ensure the validity of the FBA process, the program checks to make sure that all of the FRF
components have been generated using the same number of forcing frequencies and further that all of
these forcing frequencies are the same for all of the FRF components. If both of these conditions are not
met, the program terminates the job with an appropriate fatal message.
Remark 13. on the FRF Case Control command ontains descriptions of examples for various scenarios
of the FRF/FBA capability.
Main Index
146 MD Nastran Dynamic Analysis User’s Guide
Modal Frequency Response Analysis
As a first step in the formulation, transform the variables from physical coordinates u to modal
coordinates by assuming
i t
x = e (4-11)
The mode shapes are used to transform the problem in terms of the behavior of the modes as opposed
to the behavior of the grid points. Equation (4-11) represents an equality if all modes are used; however,
because all modes are rarely used, the equation usually represents an approximation.
To proceed, temporarily ignore all damping, which results in the undamped equation for harmonic
motion
2
– M x + K x = P (4-12)
at forcing frequency .
Substituting the modal coordinates in Eq. (4-11) for the physical coordinates in Eq. (4-12) and dividing
by e it , the following is obtained:
2
– M + K = P (4-13)
Now this is the equation of motion in terms of the modal coordinates. At this point, however, the
equations remain coupled.
2 T T T
– M + K = P (4-14)
where:
T
M = modal (generalized) mass matrix
T
K = modal (generalized) stiffness matrix
T
P = modal force vector
The final step uses the orthogonality property of the mode shapes to formulate the equation of motion in
terms of the generalized mass and stiffness matrices, which are diagonal matrices. These diagonal
matrices do not have the off-diagonal terms that couple the equations of motion. Therefore, in this form
Main Index
CHAPTER 4 147
Frequency Response Analysis
the modal equations of motion are uncoupled. In this uncoupled form, the equations of motion are
written as a set of uncoupled single degree-of-freedom systems as
2
– mi i + ki = pi (4-15)
where:
The modal form of the frequency response equation of motion is much faster to solve than the direct
method because it is a series of uncoupled single degree-of-freedom systems.
Once the individual modal responses i are computed, physical responses are recovered as the
summation of the modal responses using
i t
x = e (4-16)
These responses are in complex form (magnitude/phase or real/imaginary) and are used to recover
additional output quantities requested in the Case Control Section.
T
B diagonal (4-17)
If structural damping is used, the orthogonality property does not, in general, diagonalize the generalized
stiffness matrix
T
K diagonal (4-18)
where K = 1 + iG K + i G E K E
In the presence of a B matrix or a complex stiffness matrix, the modal frequency approach solves the
coupled problem in terms of modal coordinates using the direct frequency approach described in Direct
Frequency Response Analysis, 141
2 T T T T
– M + i B + K = P (4-19)
Equation (4-19) is similar to Eq. (4-5) for the direct frequency response analysis method except that
Eq. (4-19) is expressed in terms of modal coordinates . Since the number of modes used in a solution
Main Index
148 MD Nastran Dynamic Analysis User’s Guide
Modal Frequency Response Analysis
is typically much less than the number of physical variables, using the coupled solution of the modal
equations is less costly than using physical variables.
If damping is applied to each mode separately, the uncoupled equations of motion can be maintained.
When modal damping is used, each mode has damping b i where b i = 2m i i i . The equations of motion
remain uncoupled and have the form
2
– m i i + ib i i + k i i = p i (4-20)
pi
i = ---------------------------------------------
- (4-21)
2
– m i + ib i + k i
The TABDMP1 Bulk Data entry defines the modal damping ratios. A table is created by the
frequency/damping pairs specified on the TABDMP1 entry. The solution refers to this table for the
damping value to be applied at a particular frequency. The TABDMP1 Bulk Data entry has a Table ID.
A particular TABDMP1 table is activated by selecting the Table ID with the SDAMPING Case Control
command.
1 2 3 4 5 6 7 8 9 10
TABDMP1 TID TYPE
f1 g1 f2 g2 f3 g3 -etc.- ENDT
Field Contents
TID Table identification number.
TYPE Type of damping units:
G (default)
CRIT
Q
fi Frequency value (cycles per unit time).
gi Damping value in the units specified.
Main Index
CHAPTER 4 149
Frequency Response Analysis
At resonance, the three types of damping are related by the following equations:
b G
i = ------i- = -----i
b cr 2
b c r = 2m i i (4-22)
1
Q i = ------------ 1
- = -----
2 i Gi
Note that the i subscript is for the i-th mode, and not the i-th excitation frequency.
The values of fi and gi define pairs of frequencies and dampings. Note that gi can be entered as one
of the following: structural damping (default), critical damping, or quality factor. The entered damping
is converted to structural damping internally using Eq. (4-22). Straight-line interpolation is used for
modal frequencies between consecutive fi values. Linear extrapolation is used at the ends of the table.
ENDT ends the table input.
For example, if modal damping is entered using Table 4-1 and modes exist at 1.0, 2.5, 3.6, and 5.5 Hz,
MD Nastran interpolates and extrapolates as shown in Figure 4-4 and in the table. Note that there is no
table entry at 1.0 Hz; MD Nastran uses the first two table entries at f = 2.0 and f = 3.0 to extrapolate
the value for f = 1.0 .
0.20
+
+
+ +
0.10
+ = entered value
= computed value
0.0
2.0 4.0 6.0
f (Hz)
Main Index
150 MD Nastran Dynamic Analysis User’s Guide
Modal Frequency Response Analysis
Entered Computed
f z f z
2.0 0.16 1.0 0.14
3.0 0.18 2.5 0.17
4.0 0.13 3.6 0.15
6.0 0.13 5.5 0.13
1 2 3 4 5 6 7 8 9 10
TABDMP1 10 CRIT +TAB1
+TAB1 2.0 0.16 3.0 0.18 4.0 0.13 6.0 0.13 +TAB2
+TAB2 ENDT
Modal damping is processed as a complex stiffness when PARAM,KDAMP, KDAMPFL is entered as -1.
The uncoupled equation of motion becomes
2
– m i i + 1 + iG k i i = p i (4-23)
The default for PARAM,KDAMP is 1, which processes modal damping as a damping matrix as shown
in Eq. (4-20).
The decoupled solution procedure used in modal frequency response can be used only if either no
damping is present or modal damping alone (via TABDMP1) is used. Otherwise, the modal method uses
the coupled solution method on the smaller modal coordinate matrices if nonmodal damping (i.e.,
CVISC, CDAMPi, GE on the MATi entry, or PARAM,G) is present.
Main Index
CHAPTER 4 151
Frequency Response Analysis
PARAM,LFREQ, LFREQFL gives the lower limit on the frequency range of retained modes, and
PARAM,HFREQ, HFREQFL gives the upper limit on the frequency range of retained modes.
PARAM,LMODES, LMODESFL gives the number of lowest modes to be retained. These parameters can
be used to include the proper set of modes. Note that the default is for all computed modes to be retained.
It should be noted in both of these scenarios that the selected mode shapes represent consecutive mode
numbers. (When dealing with fluid mode shapes, the parameters LMODESFL and
LFREQFL/HFREQFL play a similar role.)
There are cases, particularly when performing modal participation studies, where there may be a need
for a combination of the computed mode shapes whose mode numbers may not all be consecutive. This
may occur when only specific computed mode shapes are to be included in the dynamic analysis or
specific computed mode shapes are to be excluded from the dynamic analysis. Such selections cannot
be accomplished via the LMODES and LFREQ/HFREQ parameters. In order to allow such selections
use the Case Control command called MODESELECT.
The MODESELECT command can also be used to accomplish the same task as the LMODES and
LFREQ/HFREQ parameters. However, its main purpose and usefulness is for making selections of
mode shapes that cannot be accomplished via these parameters. With MODESELECT, specific modes
or modes within frequency ranges can be individually selected for your analysis. See the examples in the
MD Nastran Quick Reference Guide for all the options.
Main Index
152 MD Nastran Dynamic Analysis User’s Guide
Modal Versus Direct Frequency Response
Modal Direct
Small Model X
Large Model X
Few Excitation Frequencies X
Many Excitation Frequencies X
High Frequency Excitation X
Nonmodal Damping X
Higher Accuracy X
In general, larger models may be solved more efficiently in modal frequency response because the
numerical solution is a solution of a smaller system of uncoupled equations. The modal method is
particularly advantageous if the natural frequencies and mode shapes were computed during a previous
stage of the analysis. In that case, you simply perform a restart (see Restarts In Dynamic Analysis, 377).
Using the modal approach to solve the uncoupled equations is very efficient, even for very large numbers
of excitation frequencies. On the other hand, the major portion of the effort in a modal frequency
response analysis is the calculation of the modes. For large systems with a large number of modes, this
operation can be as costly as a direct solution. This result is especially true for high-frequency excitation.
To capture high frequency response in a modal solution, less accurate, high-frequency modes must be
computed. For small models with a few excitation frequencies, the direct method may be the most
efficient because it solves the equations without first computing the modes. The direct method is more
accurate than the modal method because the direct method is not concerned with mode truncation.
Table 4-2 provides an overview of which method to use. Many additional factors may be involved in the
choice of a method, such as contractual obligations or local standards of practice.
Main Index
CHAPTER 4 153
Frequency Response Analysis
Input
Use CBUSH, CELAS1, CELAS3, CDAMP1, and CDAMP3 entries to describe element connectivity
and geometry. CELAS2, CELAS4, CDAMP2, and CDAMP4 cannot be used for frequency-dependent
effects because these element formats do not refer to a property entry. CBUSH is the recommended
format for structural analysis. The CELASi and CDAMPi elements provide the same frequency
dependence as the BUSH element but do not have its capabilities for convenient modeling of complex
geometry.
Use PBUSH, PELAS, and PDAMP entries to specify the nominal properties of the element. Stiffness,
viscous damping (b), and structural damping (ge * Ke) may be specified. The nominal stiffness is used
for real-mode analysis and for cases where no tabular input is used to define frequency-dependent
impedance. The BUSH element has up to six uncoupled stiffness and damping values to define its
nominal properties. These impedances act at a point in space (the elastomer center of elasticity) defined
on the CBUSH entry. The stiffness values at the elastomer center of elasticity are transferred to the
connected grid points by rigid element equations. The orientation of these stiffness values at the
elastomer center of elasticity is defined also on this connection entry, in a manner similar to the Beam
element.
If frequency dependence is needed, the property entries listed above reference a secondary property
entry, listing the table IDs where stiffness and/or damping is defined as a function of excitation
frequency. For example, the PBUSHT entry performs this function for the CBUSH element. Input data
associated with frequency dependence, if present, is ignored in all solution sequences except the
frequency response analysis solutions, SOLs 108, 111, and 200. These elements use only their nominal
values in other types of analysis, even when secondary property entries are specified.
Main Index
154 MD Nastran Dynamic Analysis User’s Guide
Frequency-Dependent Structural Elements
Output
Element force and stress output is available for all of these elements. These elements also appear in grid-
point force balance, element strain energy, and element-oriented force output in static analysis. If present,
the frequency-dependent effects are accounted for exactly in all output quantities, including element
force and stress, and single- and multipoint constraint forces.
Limitation
There are two methods of data recovery in modal dynamic analysis. The default method is called the
Dynamic Data Recovery Matrix Method (DDRMM). The alternate method is called the Mode
Displacement Method. Select the Mode Displacement Method by adding the Bulk Data entry
PARAM,DDRMM,-1.
In transient analysis, if the frequency-dependent elements have damping properties and the matrix
method is used, no element output is provided. In the first example below, Analysis of Spot Welds, a
method is described to use rigid elements as an alternate way to measure the forces in frequency
dependent elements. Output can also be obtained by requesting the Mode Displacement Method.
In frequency response analysis, if frequency-dependent elements are present, MD Nastran automatically
switches to the Mode Displacement Method, and the matrix method cannot be used.
Examples
Analysis of Spot Welds
When running the bushweld.dat example from the MD Nastran Linear Static Analysis User’s Guide
evidence of internal constraints are apparent in static analysis when the resultants are not in balance. The
analysis does not isolate the elements with internal constraints; instead it merely states that some internal
constraints must exist. The evidence in modal analysis appears when expected rigid-body modes have
nonzero natural frequencies or in cases where the mode shapes for the rigid-body modes cause finite
forces to occur in the elements with internal constraints and in other elements adjacent to them.
Frequency Response
Frequency response analysis is demonstrated by a small model, which allows for manual calculation of
results. The model shown in Figure 4-6 is available in the Install_dir/mdxxxx/doc/dynamics/bushver.dat.
The main structure is the grounded grid point 11. It is connected by a vibration control mount to a point
mass at point 12 representing the isolated component. The nominal mount stiffness is input on a PELAS
entry referenced by a CELAS1 entry. The mass is adjusted by a PARAM,WTMASS entry so that unit
stiffness and weight produce a 1.0 Hz. nominal natural frequency. The input file for the verification
problem is shown in Listing 4-2.
Main Index
CHAPTER 4 155
Frequency Response Analysis
Main Index
156 MD Nastran Dynamic Analysis User’s Guide
Frequency-Dependent Structural Elements
Engine-Body Coupling
A series of example problems are based on the planar model shown in Figure 4-6.
25 35 45
44
24
Engine Model
34
23 43
33
Bush Element 15 55
RBAR Element 13 Bush Element
2 12 22 32 42 52 62
Body
Model
1 11 21 31 41 51 61
Example Problem for Bushing Element Frequency Dependent Impedance
.
Main Index
CHAPTER 4 157
Frequency Response Analysis
Main Index
158 MD Nastran Dynamic Analysis User’s Guide
Frequency-Dependent Structural Elements
Forces in rigid elements are requested by MPCFORCE output requests. MPC forces are output in the
displacement coordinates systems of the points to which they are attached. The force at the upper grid
point of the RBAR element is in the same coordinate system as that of the connected grid point, which
was modeled to be the same as that of the CBUSH element. The force transmitted to the body point can
be determined from the MPC force at the other point of the RBAR element. This force is in the basic
coordinate system, the coordinate system of the connected body point because this happens to be the CD
system (blank, or 0) specified for that grid point. MPC (and SPC) forces include effects from all of the
matrices used to compute the dynamic analysis solutions, including all forms of stiffness, viscous
damping, ge and g damping, and mass. This instrumentation technique applies not only to the BUSH
element in transient analysis, but is applicable to all elements in linear analysis.
For direct frequency response analysis, the element force in the BUSH element matches the MPC force to
many digits of accuracy. For the modal frequency response analysis, the two force outputs match within
a few percent, with the differences resulting from modal truncation effects. For the transient response
analysis, the MPC forces provide the only method for output of the force acting on the element.
Performance Analysis
The important parameters in the incremental costs associated with adding frequency-dependent elements
to a model are as follows:
• The ratio of frequency dependent elements to other elements.
• The types of frequency dependent damping requested.
• Whether frequency dependent elements intersect constrained DOFs.
• Whether modal or direct analysis is chosen.
• Whether there are omitted DOFs.
The use of frequency-dependent elements in modal analysis blocks the DDRMM method of data
recovery, and the mode displacement method is used instead. The modal matrices used for solution will
be coupled, not diagonal, as a consequence of the incremental impedances due to frequency-dependent
effects.
In direct analysis, there is no DDRMM method of data recovery and the addition of the incremental
matrices for frequency-dependent elements does not affect the cost of computation of the solution
module itself significantly. There are added costs associated with assembling and reducing the
incremental impedance matrices for the solution module.
Discussion of Results
The modal method costs less than half of the direct method. In fact, if the DDRMM method is not
blocked. If there are no damping effects, a real matrix solution can be used rather than a complex solution,
and even more of the cost can be avoided.
Much of the added cost due to connections of this type can be avoided by not requesting constraint-force
output. GE damping is considerably more expensive than viscous damping. The approximations of the
modal method are more pronounced for a large number of frequency-dependent elements, while
computation costs approach direct analysis costs. These trends indicate that a direct analysis offers
Main Index
CHAPTER 4 159
Frequency Response Analysis
increased accuracy over a modal analysis at small incremental costs when the proportion of frequency-
dependent elements is large. That modal analysis, however, retains its cost advantage over direct analysis
when the proportion of frequency-dependent elements is small.
Main Index
160 MD Nastran Dynamic Analysis User’s Guide
Frequency-Dependent Excitation Definition
The particular entry chosen for defining the dynamic loading is largely a function of user convenience
for concentrated loads. Pressure and distributed loads, however, require a more complicated format if
the LOADSET/LSEQ combination is required.
There are two important aspects of dynamic load definition. First, the location of the loading on the
structure must be defined. Since this characteristic locates the loading in space, it is called the spatial
distribution of the dynamic loading. Secondly, the frequency variation in the loading is the characteristic
that differentiates a dynamic load from a static load. This frequency variation is called the temporal
distribution of the load. A complete dynamic loading is a product of spatial and temporal distributions.
Using Table IDs and Set IDs in MD Nastran makes it possible to apply many complicated and temporally
similar loadings with a minimum of input. Combining simple loadings to create complicated loading
distributions that vary in position as well as frequency is also a straightforward task.
The remainder of this section describes the Bulk Data entries for frequency-dependent excitation. The
description is given in terms of the coefficients that define the dynamic load.
The values of the coefficients are defined in tabular format on a TABLEDi entry. You need not explicitly
define a force at every excitation frequency. Only those values that describe the character of the loading
are required. MD Nastran will interpolate for intermediate values.
Main Index
CHAPTER 4 161
Frequency Response Analysis
1 2 3 4 5 6 7 8 9 10
RLOAD1 SID EXCITEID DELAYI/ DPHASEI/ TC TD TYPE
DELAYR DPHASER
Field Contents
SID Set identification number. (Integer > 0)
EXCITEID Identification number of the DAREA, FBALOAD (in FRF Based Assembly or
FBA process) or SPCD entry set that defines A . See Remarks 5. and 6.
(Integer > 0)
DELAYI Identification number of DELAY or FBADLAY (in FRF Based Assembly or FBA
process) Bulk Data entry that defines time delay . See Remark 2. (Integer > 0 or
blank)
DELAYR Value of time delay that will be used for all degrees-of-freedom that are excited
by this dynamic load entry. See Remark 2. (Real or blank)
DPHASEI Identification number DPHASE or FBAPHAS (in FRF Based Asseembly or FBA
process) Bulk Data entry that defines phase angle . (See Remark 2. (Integer > 0
or blank)
DPHASER Value of phase angle (in degrees) that will be used for all degrees-of-freedom that
are excited by this dynamic load entry. See Remark 2. (Real or blank)
TC Set identification number of the TABLEDi entry that gives C f . See Remark 2.
(Integer > 0)
TD Set identification number of the TABLEDi entry that gives D f . See Remark 2.
(Integer > 0)
TYPE Defines the type of the dynamic excitation. See Remarks 5. and 6. (Integer,
character or blank; Default = 0)
Note that f is the frequency in cycles per unit time and that e iwt = cot wt + i sin wt .
Main Index
162 MD Nastran Dynamic Analysis User’s Guide
Frequency-Dependent Excitation Definition
if
C f + iD f = B f e (4-26)
RLOAD1 RLOAD2
Definition Definition
1 2 3 4 5 6 7 8 9 10
RLOAD2 SID EXCITEID DELAYI/ DPHASEI/ TB TP TYPE
DELAYR DPHASER
Field Contents
SID Set identification number. (Integer > 0)
EXCITEID Identification number of the DAREA, FBALOAD (in FRF Based Assembly or FBA
process) or SPCD entry set that defines A . See Remarks 5. and 6. (Integer > 0)
DELAYI Identification number of DELAY or FBADLAY (in FRF Based Assembly or FBA
process) Bulk Data entry that defines time delay . See Remark 2. (Integer > 0 or
blank)
DELAYR Value of time delay that will be used for all degrees-of-freedom that are excited by
this dynamic load entry. See Remark 2. (Real or blank)
DPHASEI Identification number DPHASE or FBAPHAS (in FRF Based Assembly or FBA
process) Bulk Data entry that defines phase angle . (See Remark 2. (Integer > 0 or
blank)
DPHASER Value of phase angle (in degrees) that will be used for all degrees-of-freedom that
are excited by this dynamic load entry. See Remark 2. (Real or blank)
TB Set identification number of the TABLEDi entry that gives B f . (Integer > 0)
TP Set identification number of the TABLEDi entry that gives f in degrees.
(Integer > 0)
TYPE Defines the type of the dynamic excitation. See Remarks 5. and 6. (Integer, character
or blank; Defaults = 0)
1 2 3 4 5 6 7 8 9 10
DAREA SID P1 C1 A1 P2 C2 A2
Main Index
CHAPTER 4 163
Frequency Response Analysis
Field Contents
SID Set ID specified by RLOADi entires.
Pi Grid, extra, or scalar point ID.
Ci Component number.
Ai Scale factor.
A DAREA entry is selected by RLOAD1 or RLOAD2 entries. Any number of DAREA entries may be
used; all those with the same SID are combined.
1 2 3 4 5 6 7 8 9 10
DELAY SID P1 C1 P2 C2 2
Field Contents
SID Set ID specified by an RLOADi entry.
Pi Grid, extra, or scalar point ID.
Ci Component number.
i Time delay for Pi, Ci. (Default = 0.0)
A DAREA entry must be defined for the same point and component. Any number of DELAY entries
may be used; all those with the same SID are combined.
1 2 3 4 5 6 7 8 9 10
DPHASE SID P1 C1 1 P2 C2 2
Field Contents
SID Set ID specified by an RLOADi entry.
Pi Grid, extra, or scalar point ID.
Main Index
164 MD Nastran Dynamic Analysis User’s Guide
Frequency-Dependent Excitation Definition
Field Contents
Ci Component number.
i Phase lead (in degrees) for Pi, Ci. (Default = 0.0)
A DAREA entry must be defined for the same point and component. Any number of DPHASE entires
may be used; all those with the same SID are combined.
1 2 3 4 5 6 7 8 9 10
TABLED1 TID XAXIS YAXIS
x1 y1 x2 y2 x3 y3 -etc.- ENDT
Field Contents
TID Table identification number.
XAXIS Specifies a linear or logarithmic interpolation for the x-axis. (Character: “LINEAR” or
“LOG”; default = “LINEAR”)
Main Index
CHAPTER 4 165
Frequency Response Analysis
Field Contents
YAXIS Specifies a linear or logarithmic interpolation for the y-axis. (Character: “LINEAR” or
“LOG”; default = “LINEAR”)
xi, yi Tabular values. Values of x are frequency in cycles per unit time.
ENDT Ends the table input.
y = yT x (4-27)
XAXIS YAXIS yT x
xj – x x – xi
---------------- yi + ---------------- yj
LINEAR LINEAR xj – xi xj – xi
ln xj x ln x x i
------------------------- yi + ------------------------- yj
LOG LINEAR ln xj xi ln xj xi
xj – x x – xi
exp ---------------- ln yi + ---------------- ln yj
LINEAR LOG xj – xi xj – xi
ln xj x ln x x i
exp ------------------------- ln y i + ------------------------- yj
LOG LOG ln xj xi ln xj xi
1 2 3 4 5 6 7 8 9 10
TABLED2 TID X1
x1 y1 x2 y2 x3 y3 -etc.- ENDT
Main Index
166 MD Nastran Dynamic Analysis User’s Guide
Frequency-Dependent Excitation Definition
Field Contents
TID Table identification number.
X1 Table parameter.
xi, yi Tabular values. Values of x are frequency in cycles per unit time.
y = y T x – X1 (4-28)
1 2 3 4 5 6 7 8 9 10
TABLED3 TID X1 X2
x1 y1 x2 y2 x3 y3 -etc.- ENDT
Field Contents
TID Table identification number.
X1, X2 Table parameters (X2 0.0).
xi, yi Tabular values. Values of x are frequency in cycles per unit time.
x – X1
y = y T ---------------- (4-29)
X2
1 2 3 4 5 6 7 8 9 10
TABLED4 TID X1 X2 X3 X4
A0 A1 A2 A3 A4 A5 -etc.- ENDT
Field Contents
TID Table identification number.
Main Index
CHAPTER 4 167
Frequency Response Analysis
Field Contents
Xi Table parameters (X2 0.0; X3 < X4).
Ai Coefficients.
N is the degree of the power series. When x < X3, X3 is used for x; when x > X4, X4 is used for x. This
condition has the effect of placing bounds on the table; there is no extrapolation outside of the table
boundaries.
ENDT ends the table input.
DAREA Example
Suppose the following command is in the Case Control Section:
DLOAD = 35
in addition to the following entries in the Bulk Data Section:
1 2 3 4 5 6 7 8 9 10
$RLOAD1 SID DAREA DELAY DPHASE TC TD
RLOAD1 35 29 31 40
$DAREA SID POINT COMPONENT SCALE
DAREA 29 30 1 5.2
$DELAY SID POINT COMPONENT LAG
DELAY 31 30 1 0.2
$TABLED1 ID XAXIS YAXIS
$ x1 y1 x2 y2 x3 y3 x4 y4
TABLED1 40 LINEAR LINEAR
0.0 4.0 2.0 8.0 6.0 8.0 ENDT
The DLOAD Set ID 35 in Case Control selects the RLOAD1 entry in the Bulk Data having a Set ID 35.
On the RLOAD1 entry is a reference to DAREA Set ID 29, DELAY Set ID 31, and TABLED1 Set ID
40. The DAREA entry with Set ID 29 positions the loading on grid point 30 in the 1 direction with a
scale factor of 5.2 applied to the load. The DELAY entry with Set ID 31 delays the loading on grid point
30 in the 1 direction by 0.2 units of time. The TABLED1 entry with Set ID 40 defines the load in tabular
Main Index
168 MD Nastran Dynamic Analysis User’s Guide
Frequency-Dependent Excitation Definition
form. This table is shown graphically in Figure 4-8. The result of these entries is a dynamic load applied
to grid point 30, component T1, scaled by 5.2 and delayed by 0.2 units of time.
10
Amplitude
on 5
TABLED1
0 2 4 6
Frequency (Hz)
Figure 4-8 TABLED1 - Amplitude Versus Frequency
Main Index
CHAPTER 4 169
Frequency Response Analysis
DLOAD
Case Control
Bulk Data
RLOADi
P = S Si Pi (4-31)
i
where:
1 2 3 4 5 6 7 8 9 10
DLOAD SID S S1 L1 S2 L2
Field Contents
SID Load Set ID.
S Overall scale factor.
Si Individual scale factors.
Li Load Set ID numbers for RLOAD1 and RLOAD2 entries.
Main Index
170 MD Nastran Dynamic Analysis User’s Guide
Frequency-Dependent Excitation Definition
Main Index
CHAPTER 4 171
Frequency Response Analysis
Solution Frequencies
A major consideration when you conduct a frequency response analysis is selecting the frequency at
which the solution is to be performed. There are six Bulk Data entries that you can use to select the
solution frequencies. It is important to remember that each specified frequency results in an independent
solution at the specified excitation frequency.
To select the loading frequencies, use the FREQ, FREQ1, FREQ2, FREQ3, FREQ4 and FREQ5 Bulk
Data entries.
The FREQUENCY Case Control command selects FREQi Bulk Data entries. All FREQi entries with the
same selected Set ID are applied in the analysis; therefore, you can use any combination of FREQ,
FREQ1, FREQ2, FREQ3, FREQ4 and FREQ5 entries.
The examples that follow show the formats of the FREQi entries. Notice that the six sets of excitation
frequencies shown in the examples will be combined in a single analysis if the Set IDs are identical.
FREQ
The FREQ Bulk Data entry specifies ten specific (unequally spaced loading frequencies to be analyzed.
1 2 3 4 5 6 7 8 9 10
$FREQ SID F F F F F F F
$ F F F F F F F F
FREQ 3 2.98 3.05 17.9 21.3 25.6 28.8 31.2
29.2 22.4 19.3
Main Index
172 MD Nastran Dynamic Analysis User’s Guide
Solution Frequencies
Field Contents
SID Set ID specified by a FREQUENCY Case Control command.
F Frequency value (cycles per unit time).
The FREQ1 Bulk Data example specifies 14 frequencies between 2.9 Hz and 9.4 Hz in increments of
0.5 Hz.
Field Contents
SID Set ID specified by a FREQUENCY Case Control command.
F start Starting frequency in set (cycles per unit time).
f Frequency increment (cycles per unit time).
NDF Number of frequency increments.
The FREQ2 Bulk Data entry example specifies six logarithmic frequency intervals between 1.0 and 8.0
Hz, resulting in frequencies at 1.0, 1.4142, 2.0, 2.8284, 4.0, 5.6569, and 8.0 Hz being used for the
analysis.
1 2 3 4 5 6 7 8 9 10
$FREQ2 SID F start F end NF
FREQ2 9 1.0 8.0 6
Field Contents
SID Set ID specified by a FREQUENCY Case Control command.
F start Starting frequency (cycles per unit time).
F end Ending frequency (cycles per unit time).
NF Number of logarithmic intervals.
The FREQ3 Bulk Data entry example requests 10 frequencies between each set of modes within the range
20 and 2000, plus ten frequencies between 20 and the lowest mode in the range, plus 10 frequencies
between the highest mode in the range and 2000.
Main Index
CHAPTER 4 173
Frequency Response Analysis
1 2 3 4 5 6 7 8 9 10
$FREQ3 SID F1 F2 TYPE NEF CLUSTER
Field Contents
SID Set ID specified by a FREQUENCY Case Control command.
F1 Lower bound of modal frequency range in cycles per unit time.
(Real > 0.0)
F2 Upper bound of modal frequency range in cycles per unit time.
(Real > 0.0, F2 F1, Default = F1)
TYPE LINEAR or LOG. Specifies linear or logarithmic interpolation between frequencies.
(Character; Default = “LINEAR”)
NEF Number of excitation frequencies within each subrange including the end points.
(Integer > 1, Default = 10)
CLUSTER A CLUSTER value greater than 1 provides closer spacing of excitation frequencies near
the modal frequencies, where greater resolution is needed. (Real > 0.0; Default = 1.0)
The FREQ4 Bulk Data entry example chooses 21 equally spaced frequencies across a frequency band of
0.7 f N to 1.3 f N for each natural frequency between 20 and 2000.
1 2 3 4 5 6 7 8 9 10
$FREQ4 SID F1 F2 FSPD NFM
FREQ4 6 20.0 2000.0 0.30 21
Field Contents
SID Set ID specified by a FREQUENCY Case Control command.
F1 Lower bound of modal frequency range in cycles per unit time. (Real > 0.0)
F2 Upper bound of modal frequency range in cycles per unit time. (Real > 0.0, F2 F1, Default
= F1)
FSPD Frequency spread, +/– the fractional amount specified for each mode which occurs in the
frequency range F1 to F2. (1.0 > Real > 0.0,
Default = 0.10)
NFM Number of evenly spaced frequencies per “spread” mode. (Integer > 0; Default = 3; If NFM
is even, NFM + 1 will be used.)
Main Index
174 MD Nastran Dynamic Analysis User’s Guide
Solution Frequencies
The FREQ5 Bulk Data entry example will compute excitation frequencies which are 0.6, 0.8, 0.9, 0.95,
1.0, 1.05, 1.1, and 1.2 times the natural frequencies for all natural frequencies, but use only the computed
frequencies that fall within the range 20 and 2000.
1 2 3 4 5 6 7 8 9 10
$FREQ5 SID F1 F2 FR1 FR2 FR3 FR4 FR5
FR6 FR7 -etc.-
FREQ5 6 20.0 2000.0 1.0 0.6 0.8 0.9 0.95
1.05 1.1 1.2
Field Contents
SID Set ID specified by a FREQUENCY Case Control command.
F1 Lower bound of modal frequency range in cycles per unit time. (Real > 0.0)
F2 Upper bound of modal frequency range in cycles per unit time. (Real > 0.0, F2 F1;
Default = F1)
FRi Fractions of the natural frequencies in the range F1 to F2. (Real > 0.0)
Main Index
CHAPTER 4 175
Frequency Response Analysis
Use a fine enough frequency step size ( f ) to adequately predict peak response. Use at least five points
across the half-power bandwidth (which is approximately 2f n for an SDOF system) as shown in
Figure 4-10.
Peak Response
Peak
----------- = Half-Power Point
Response 2
Frequency
Half-Power Bandwidth
= Excitation Frequency
For maximum efficiency, an uneven frequency step size should be used. Smaller frequency spacing
should be used in regions near resonant frequencies, and larger frequency step sizes should be used in
regions away from resonant frequencies.
Main Index
176 MD Nastran Dynamic Analysis User’s Guide
Solution Control for Frequency Response Analysis
Table 4-3 Case Control Commands for Frequency Response Solution Control
Case Direct
Control or
Command Modal Description Required/Optional
DLOAD Both Select the dynamic load set from Bulk Data Required
FREQUENCY Both Select FREQi entries from Bulk Data Required
METHOD Modal Select the eigenvalue extraction parameters Required
LOADSET Both Select the LSEQ set from Bulk Data Optional
SDAMPING Modal Select the modal damping table from Bulk Optional
Data
OFREQUENCY Both Select the frequencies for output (Default = Optional
all)
OMODES Both Selects mode for output requests (Default = Optional
all)
The Case Control command, OMODES (for Output MODES), allows for the selection of a set of modes
for data recovery and output purposes. This command is supported for SOLs 103, 105, 107, 110, 111,
112, and 200. It allows selection of modes by mode number rather than by frequency value, as is the case
with the OFREQUENCY command. OMODES is particularly useful because it limits the amount of
output in the case of large, real-life problems without affecting the actual number of modes computed.
For example, assume for a large normal modes model with multiple boundary conditions that the user is
interested in obtaining the output for the first five modes in the first subcase and the next ten modes in
the following subcase. The following OMODES request in each subcase will accomplish this goal.
SUBCASE 10
BC = 1
SPC = 100
METHOD = 101
SET 102 = 1 THRU 5
OMODES = 102
DISP = ALL
$
SUBCASE 20
BC = 2
SPC = 200
METHOD = 201
SET 202 = 6 THRU 15
Main Index
CHAPTER 4 177
Frequency Response Analysis
OMODES = 202
DISP = ALL
The types of results available from a frequency response analysis are similar to those for a typical static
analysis except that the results are a complex function of the applied loading frequency. Additional
quantities (characteristic of dynamic problems) are also available. The output quantities are summarized
in Table 4-4 and Table 4-5.
Frequency response output is in real/imaginary format (the default) or magnitude/phase format (the
phase angle is in degrees). Frequency response output is also in SORT1 or SORT2 format. In SORT1
format, the results are listed by frequency; in SORT2 format, the results are listed by grid point or
element number. SORT1 is the default for direct frequency response analysis (SOL 108), and SORT2
is the default for modal frequency response analysis (SOL 111). PARAM,CURVPLOT,1 and
PARAM,DDRMM,-1 are necessary to obtain SORT1 output in SOL 111. These output formats are
specified with the Case Control commands. The command
DISPLACEMENT(PHASE, SORT2) = ALL
prints displacements in magnitude/phase and SORT2 formats. The output formats are illustrated in the
first example in Examples, 184.
Main Index
178 MD Nastran Dynamic Analysis User’s Guide
Solution Control for Frequency Response Analysis
A number of Bulk Data entries are unique to frequency response analysis. These entries can be combined
with other generic entries in the Bulk Data. Bulk Data entries directly related to frequency response
analysis are summarized in Table 4-6.
Main Index
CHAPTER 4 179
Frequency Response Analysis
Theory
Modal analysis solutions are based on the theory that the effects of a small number of natural modes can
be used to represent the response of the structure. The eigenvectors of these modes are used to transform
the problem from the physical domain into a scalar domain of generalized coordinates. This
transformation results in a much smaller problem size that can be solved rapidly. Once the solution to
the generalized problem is obtained, the solution in the physical domain is obtained by transformation,
again using the eigenvectors of the natural modes. The response at a physical degree of freedom (a grid
point’s translation or rotation component) for a particular forcing frequency or time step can be found by
summing the individual responses of each mode shape. This is accomplished by simple matrix
multiplication of the eigenvectors times the generalized solution vectors. The calculation of modal
contribution fractions can be performed as part of the total response calculation.
In MD Nastran, several different views of the modal contribution fractions are computed. All start with
some basic ingredients. These are the total modal solution response at the required degrees of freedom
and the transformation matrix from modal to physical coordinates. The solution responses are complex
(a real term and an imaginary term) for frequency response and complex eigenvalue solutions. The
responses involve only real terms for transient analysis. The following discussion pertains to complex
response results, but is applicable to transient analysis as well. The only difference is that there are no
imaginary terms and phase angles are all zero. The discussion is limited to operations on a single degree
of freedom for a single solution vector. Extension to multiple degrees of freedom and solution vectors
is a simple matter of substituting matrix operations for vector operations.
The total modal response, U , at a degree of freedom can be found by summing the individual modal
responses.
U = rphi uh (4-32)
In Eq. (4-32), rphi is a real one row by h column physical response data recovery matrix and uh
is a complex (real for transient) h row by one column modal solution vector. The “h” size represents the
number of natural modes that are used in the solution process. The total response, U , is a complex
Main Index
180 MD Nastran Dynamic Analysis User’s Guide
Modal Contribution Fraction (MCFRACTION)
number with real and imaginary terms. It can be represented in polar format (magnitude and phase angle)
with some simple arithmetic manipulation. Now each of the individual modal contributions rphi uh
in the sum can be represented in complex and polar formats also. The natural mode with the largest
individual contribution to the magnitude can be determined by scanning the magnitudes of each of the
individual modal responses. The magnitude of the individual modal response can be projected onto the
magnitude of total response. The relative phase angle of the individual modal response with respect to
the phase angle of the total response can be calculated. Each individual modal response projected
magnitude can be divided by the total response magnitude to obtain the fraction of the total response that
is contributed by each mode. A scaled modal response magnitude can be calculated by dividing each
modal response projected magnitude by the largest modal response magnitude. All of these different
calculated quantities are present in the printed output as shown in Figure 4-11.
Inputs
The MCFRACTION Case Control command gives the user new controls over the amount and ordering of
the produced modal contribution fraction data. The general format of the command is:
FRACTION ALL
FRACTION 0.001
KEY = ITEMS = ALL SOLUTION = m FILTER =
sortitem fratio
(itemlist) NONE
ALL
12
NULL = = n
ipowr
NONE
The SORT and KEY keywords are used to control sorting operations. The printed output can be sorted
in one of several ways using any of the fraction views as a key. The default sorting order is in the order
of the natural frequency (or mode id). The data can also be filtered to remove insignificant terms before
it is printed. The SOLUTION keyword selects a subset of the available solutions for output. Any or all
of the different fraction views can be sent to the punch file.
The SET Case Control command allows the analyst to specify a set of grid point and component code
combinations. An example of the format is “SET 1000 = 917/T3, 85/R2, 127016/T1” which
demonstrates the general input format of grid ID and component code separated by a slash, much like the
manner in which XY-PLOT command grid point entities are defined. No THRU ranges are permitted.
Outputs
The MCFRACTION Case Control command can cause generation of both printed and punched output.
The output data block is OMCFRAC. Items computed and available for output are shown in Table 4-7.
Main Index
CHAPTER 4 181
Frequency Response Analysis
Figure 4-11 is an example of the modal contribution fraction output produced in a modal frequency
response solution. The heading lines identify the grid point and component, total response in both
real/imaginary and magnitude/phase formats, loading condition information, maximum response
contribution for a mode and the mode id, sorting and filtering information. Then, for each natural mode,
a tabular listing of the various views of modal contribution data is presented. The table contains the
response of the mode, in real/imaginary and magnitude phase representations, the projection magnitude,
the relative phase angle, the modal fraction value, and the scaled response magnitude.
M O D A L C O N T R I B U T I O N F R A C T I O N S
GRID POINT = 101/T3, TOTAL RESPONSE (R/I) = 4.73302E-01 / 5.20364E-03, (M/P) = 4.73330E-01 / 0.63
LOAD FREQUENCY = 2.00000E+01, (SUBCASE 1, DLOAD = 15)
MAXIMUM MODAL RESP = 4.67546E-01 FOR MODE ID = 2, SORTKEY = FRACTION, SORT = ABS VALUE DESCENDING, FILTER = 1.00000E-02
MODE NATURAL MODAL RESPONSE MODAL RESPONSE PROJECTION REL. MODAL SCALED RESPONSE
ID FREQ (HZ) REAL IMAGINARY MAGNITUDE PHASE MAGNITUDE PHASE FRACTION MAGNITUDE
2 9.35245E+00 4.67512E-01 5.59608E-03 4.67546E-01 0.69 4.67546E-01 0.06 9.87778E-01 9.99999E-01
3 3.17429E+01 2.36155E-02 -4.93491E-04 2.36207E-02 358.80 2.36087E-02 358.17 4.98778E-02 5.04949E-02
6 7.63429E+01 -1.77298E-02 9.97411E-05 1.77301E-02 179.68 -1.77276E-02 179.05 -3.74530E-02 -3.79164E-02
The headings in Figure 4-11 reflect the output generated in modal frequency response. For modal
transient response, only real (as opposed to complex) numbers are generated and the table is reduced to
that shown in Figure 4-12.
M O D A L C O N T R I B U T I O N F R A C T I O N S
MODE ID NATURAL FREQ. (HZ) MODAL RESPONSE MODAL FRACTION SCALED MAGNITUDE
2 9.35245E+00 -3.30842E-08 8.85432E-01 -1.00000E+00
6 7.63429E+01 -6.92352E-09 1.85294E-01 -2.09270E-01
9 2.54595E+02 -5.27676E-10 1.41222E-02 -1.59495E-02
13 5.30493E+02 4.79519E-10 -1.28334E-02 1.44939E-02
3 3.17429E+01 2.69084E-09 -7.20151E-02 8.13333E-02
Main Index
182 MD Nastran Dynamic Analysis User’s Guide
Modal Contribution Fraction (MCFRACTION)
Example
A simple model demonstrates the usage of the MCFRACTION Case Control command. Three example
problems are available, one for each of the three supported solution sequences. Example mcfr111.dat
(Install_dir/mdxxxx/doc/dynamics/mcfr111.dat) demonstrates usage in a modal frequency response
Main Index
CHAPTER 4 183
Frequency Response Analysis
Main Index
184 MD Nastran Dynamic Analysis User’s Guide
Examples
Examples
This section provides several examples showing the input and output. These examples are:
Two-DOF Model
Consider the two-DOF system shown in Figure 4-13. Modal frequency response (SOL 111) is run with
a 20 N load applied to the primary mass (grid point 2) across a frequency range of 2 to 10 Hz with an
excitation frequency increment of 0.05 Hz. Uniform modal damping of 5% critical damping is used.
Listing 4-4 shows the input file. (See Install_dir/mdxxxx/doc/dynamics/bd04two.dat)
P = 20 N sin t
x
Figure 4-13 Two-DOF Model
Main Index
CHAPTER 4 185
Frequency Response Analysis
Main Index
186 MD Nastran Dynamic Analysis User’s Guide
Examples
Main Index
CHAPTER 4 187
Frequency Response Analysis
Figure 4-14 shows the relationship between the Case Control commands and the Bulk Data entries. Note
that the RLOAD1 entry references the DAREA and TABLED1 entries. The input file also shows an
alternate way to specify the dynamic load, by using a DLOAD Bulk Data entry. Because there is only a
single RLOAD1 entry, the DLOAD Bulk Data entry is not required.
Figure 4-14 Relationship Between the Case Control Commands and Bulk Data Entries for
the Two-DOF Model
Case Control Bulk Data
METHOD EIGRL
FREQUENCY FREQ1
SDAMPING TABDMP1
DLOAD RLOAD1
DAREA
TABLED1
where:
Output can be printed in either real/imaginary or magnitude/phase format and in either SORT1 or
SORT2 format. These formats are illustrated in Listing 4-6, Listing 4-7, and Listing 4-7 showing a portion
of their printed output.
Main Index
188 MD Nastran Dynamic Analysis User’s Guide
Examples
POINT-ID = 1
C O M P L E X D I S P L A C E M E N T V E C T O R
(REAL/IMAGINARY)
FREQUENCY TYPE T1 T2 T3 R1 R2 R3
2.000000E+00 G 0.0 2.813048E-03 0.0 0.0 0.0 0.0
0.0 -2.107981E-04 0.0 0.0 0.0 0.0
2.050000E+00 G 0.0 2.866637E-03 0.0 0.0 0.0 0.0
0.0 -2.229159E-04 0.0 0.0 0.0 0.0
2.100000E+00 G 0.0 2.923136E-03 0.0 0.0 0.0 0.0
0.0 -2.358377E-04 0.0 0.0 0.0 0.0
2.150000E+00 G 0.0 2.982726E-03 0.0 0.0 0.0 0.0
0.0 -2.496357E-04 0.0 0.0 0.0 0.0
2.200000E+00 G 0.0 3.045604E-03 0.0 0.0 0.0 0.0
0.0 -2.643902E-04 0.0 0.0 0.0 0.0
2.250000E+00 G 0.0 3.111983E-03 0.0 0.0 0.0 0.0
0.0 -2.801898E-04 0.0 0.0 0.0 0.0
POINT-ID = 2
C O M P L E X D I S P L A C E M E N T V E C T O R
(REAL/IMAGINARY)
FREQUENCY TYPE T1 T2 T3 R1 R2 R3
2.000000E+00 G 0.0 2.374953E-03 0.0 0.0 0.0 0.0
0.0 -1.129933E-04 0.0 0.0 0.0 0.0
2.050000E+00 G 0.0 2.397706E-03 0.0 0.0 0.0 0.0
0.0 -1.180853E-04 0.0 0.0 0.0 0.0
2.100000E+00 G 0.0 2.421475E-03 0.0 0.0 0.0 0.0
0.0 -1.234172E-04 0.0 0.0 0.0 0.0
2.150000E+00 G 0.0 2.446310E-03 0.0 0.0 0.0 0.0
0.0 -1.290071E-04 0.0 0.0 0.0 0.0
2.200000E+00 G 0.0 2.472262E-03 0.0 0.0 0.0 0.0
0.0 -1.348744E-04 0.0 0.0 0.0 0.0
2.250000E+00 G 0.0 2.499386E-03 0.0 0.0 0.0 0.0
0.0 -1.410402E-04 0.0 0.0 0.0 0.0
POINT-ID = 1
C O M P L E X D I S P L A C E M E N T V E C T O R
(MAGNITUDE/PHASE)
FREQUENCY TYPE T1 T2 T3 R1 R2 R3
2.000000E+00 G 0.0 2.820935E-03 0.0 0.0 0.0 0.0
0.0 355.7145 0.0 0.0 0.0 0.0
2.050000E+00 G 0.0 2.875291E-03 0.0 0.0 0.0 0.0
0.0 355.5535 0.0 0.0 0.0 0.0
2.100000E+00 G 0.0 2.932634E-03 0.0 0.0 0.0 0.0
0.0 355.3874 0.0 0.0 0.0 0.0
2.150000E+00 G 0.0 2.993155E-03 0.0 0.0 0.0 0.0
0.0 355.2159 0.0 0.0 0.0 0.0
2.200000E+00 G 0.0 3.057058E-03 0.0 0.0 0.0 0.0
0.0 355.0386 0.0 0.0 0.0 0.0
2.250000E+00 G 0.0 3.124571E-03 0.0 0.0 0.0 0.0
0.0 354.8552 0.0 0.0 0.0 0.0
POINT-ID = 2
C O M P L E X D I S P L A C E M E N T V E C T O R
(MAGNITUDE/PHASE)
FREQUENCY TYPE T1 T2 T3 R1 R2 R3
2.000000E+00 G 0.0 2.377640E-03 0.0 0.0 0.0 0.0
0.0 357.2761 0.0 0.0 0.0 0.0
2.050000E+00 G 0.0 2.400612E-03 0.0 0.0 0.0 0.0
0.0 357.1805 0.0 0.0 0.0 0.0
2.100000E+00 G 0.0 2.424618E-03 0.0 0.0 0.0 0.0
0.0 357.0823 0.0 0.0 0.0 0.0
2.150000E+00 G 0.0 2.449710E-03 0.0 0.0 0.0 0.0
0.0 356.9813 0.0 0.0 0.0 0.0
2.200000E+00 G 0.0 2.475939E-03 0.0 0.0 0.0 0.0
0.0 356.8773 0.0 0.0 0.0 0.0
2.250000E+00 G 0.0 2.503362E-03 0.0 0.0 0.0 0.0
0.0 356.7702 0.0 0.0 0.0 0.0
Main Index
CHAPTER 4 189
Frequency Response Analysis
FREQUENCY = 2.000000E+00
C O M P L E X D I S P L A C E M E N T V E C T O R
(REAL/IMAGINARY)
FREQUENCY = 2.050000E+00
C O M P L E X D I S P L A C E M E N T V E C T O R
(REAL/IMAGINARY)
FREQUENCY = 2.000000E+00
C O M P L E X D I S P L A C E M E N T V E C T O R
(MAGNITUDE/PHASE)
FREQUENCY = 2.050000E+00
C O M P L E X D I S P L A C E M E N T V E C T O R
(MAGNITUDE/PHASE)
Figure 4-15 shows the plots of the the resulting displacement magnitudes for grid points 1 and 2. Note
that the response for grid point 1 is nearly an order of magnitude larger than that of grid point 2. This
large difference in response magnitudes is characteristic of dynamic absorbers (also called tuned mass
dampers), in which an auxiliary structure (i.e., the small mass and stiffness) is attached to the primary
structure in order to decrease the dynamic response of the primary structure. If this same model is rerun
without the auxiliary structure, the response of the primary structure (grid point 2) at 5.03 Hz is twice
what it was with the auxiliary structure attached, as shown in Figure 4-16. See
Install_dir/mdxxxx/doc/dynamics/bd04twona.dat.
Main Index
190 MD Nastran Dynamic Analysis User’s Guide
Examples
0.12
Displacement 1 (m)
0.014
Displacement 2 (m)
0
2 Frequency (Hz) 10
0.02
Displacement 2 (m)
0
2 Frequency (Hz) 10
Figure 4-16 Displacement Response Magnitude Without the Auxiliary Structure
Main Index
CHAPTER 4 191
Frequency Response Analysis
6 11
x
p(f) 2p(f)
Main Index
192 MD Nastran Dynamic Analysis User’s Guide
Examples
4.0
Load Magnitude 6 (N)
3.0
0
100
Load Phase 6 (deg)
45
0
8.0
Load Magnitude 11 (N)
6.0
0
100
Load Phase 11 (deg)
0
0 Frequency (Hz) 20
Figure 4-18 Applied Loads
Main Index
CHAPTER 4 193
Frequency Response Analysis
Listing 4-8 shows the abridged input file. See Install_dir/mdxxxx/doc/dynamics/bd04bar.dat. The output
quantities, as defined in the Case Control Section, are the applied loads (OLOAD) for grid points 6 and
11, physical displacements (DISPLACEMENT) for grid points 6 and 11, solution set displacements
(SDISPLACEMENT) for modes 1 and 2, and element forces (ELFORCE) for element 6. These output
quantities are plotted rather than printed.
Main Index
194 MD Nastran Dynamic Analysis User’s Guide
Examples
$ DYNAMIC LOADING
$DLOAD SID S S1 L1 S2 L2
DLOAD 22 1.0 1.0 231 1.0 232
$RLOAD2 SID DAREA DELAY DPHASE TB TP
RLOAD2 231 241 261 25
RLOAD2 232 242 25
$DAREA SID P1 C1 A1
DAREA 241 6 2 1.0
DAREA 242 11 2 2.0
$DPHASE SID P1 C1 TH1
DPHASE 261 6 2 45.
$TABLED1 TID +TABL1
$+TABL1 X1 Y1 X2 Y2 ETC.
TABLED1 25 +TABL1
+TABL1 0. 1. 5.0 3. 15.0 3.0 20.0 1. +TABL2
+TABL2 25.0 1. ENDT
$
... basic model ...
$
ENDDATA
Table 4-8 shows the relationship between the Case Control commands and the Bulk Data entries. Note
that the DLOAD Bulk Data entry references two RLOAD2 entries, each of which references a separate
DAREA entry and a common TABLED1 entry. The RLOAD2 entry for grid point 6 also references a
DPHASE entry that defines the 45-degree phase lead.
Table 4-8 Relationship Between Case Control Commands and Bulk Data Entries for the
Beam Model
Case Control Bulk Data
METHOD EIGRL
FREQUENCY FREQ1
SDAMPING TABDMP1
DLOAD DLOAD
RLOAD2 231
DAREA
DPHASE
TABLED1
RLOAD2 232
DAREA
Main Index
CHAPTER 4 195
Frequency Response Analysis
where:
A = 1.0 for grid point 6 and 2.0 for grid point 11 (entered on the DAREA entry)
B = function defined on the TABLED1 entry
= 0.0 (field 7 of the RLOAD2 entry is blank)
= phase lead of 45 degrees for grid point 6 (entered on the DPHASE entry)
= 0.0 (field 4 of the RLOAD2 entry is blank)
Logarithmic plots of the output are shown in the following figures. Figure 4-19 shows the magnitude of
the displacements for grid points 6 and 11. Figure 4-20 shows the magnitude of the modal displacements
for modes 1 and 2. Figure 4-21 shows the magnitude of the bending moment at end A in plane 1 for
element 6. Logarithmic plots are especially useful for displaying frequency response results since there
can be several orders of magnitude between the maximum and minimum response values.
1.0
Displacement 6 (m)
1.0E-5
1.0
Displacement 11 (m)
1.0E-4
0 Frequency (Hz) 20
Figure 4-19 Displacement Magnitude (Log)
Main Index
196 MD Nastran Dynamic Analysis User’s Guide
Examples
1.0
SDisplacement 1
1.0E-4
1.0E-2
SDisplacement 2
1.0E-4
0 Frequency (Hz) 20
Figure 4-20 Modal Displacement Magnitude (Log)
1000
Moment A1 (N-m)
0.1
0 Frequency (Hz) 20
Main Index
CHAPTER 4 197
Frequency Response Analysis
Bracket Model
Consider the bracket model shown in Figure 4-22. An oscillating pressure load of 3 psi is applied to the
elements on the top face in the z-direction. The model is constrained at its base. Modal frequency
response is run from 0 to 100 Hz with a frequency step size of 0.2 Hz. Eigenvalues to 1000 Hz are
computed using the Lanczos method. Modal damping is applied as 2% critical damping for all modes.
Listing 4-9 shows the abridged input file. See Install_dir/mdxxxx/doc/dynamics/bd04bkt.dat. The
RLOAD1 entry is used to apply the pressure loads (PLOAD4 entries). Table 4-9 shows the relationship
between the Case Control commands and the Bulk Data entries.
Main Index
198 MD Nastran Dynamic Analysis User’s Guide
Examples
$
OUTPUT(XYPLOT)
XGRID = YES
YGRID = YES
$
$ PLOT RESULTS
XTITLE = FREQUENCY
$
YLOG = YES
YTITLE = DISPL. MAG. 999
XYPLOT DISP /999(T3RM)
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10..
....
$
$ NORMAL MODES TO 1000 HZ
$EIGRL SID V1 V2
EIGRL 777 -0.1 1000.
$
$ EXCITATION FREQUENCY DEFINITION 0 TO 100 HZ
$FREQ1 SID F1 DF NDF
FREQ1 5 0.0 0.2 500
$
$ MODAL DAMPING OF 2% CRITICAL FOR ALL MODES
$TABDMP1 TID TYPE +TABD1
$+TABD1 F1 G1 F2 G2 ETC.
TABDMP1 4 CRIT +TABD1
+TABD1 0.0 0.02 1000.0 0.02 ENDT
$
$ LOAD DEFINITION
$
$RLOAD1 SID EXCIT DELAY DPHASE TC TD
RLOAD1 2 1 22
$
$
$TABLED1 TID +TABL1
$+TABL1 X1 Y1 X2 Y2 ETC.
TABLED1 22 +TABL1
+TABL1 0.0 1.0 1000.0 1.0 ENDT
$
$ PRESURE LOAD OF 3 PSI PER ELEMENT
$PLOAD4 SID EID P1
PLOAD4 1 171 -3.
PLOAD4 1 172 -3.
PLOAD4 1 160 -3.
....... more pload4's
$
RBE2 999 999 123456 126 127 91 81 80 +RB1
+RB1 90 95 129 128 96 86 85 228 +RB2
+RB2 229 199 189 190 200 195 227 226 +RB3
+RB3 194 184 185
....... grids
GRID 999 3. 3. 3.8
$
CONM2 999 999 .0906 +
+ .35 .56 .07
$
....... element data
Main Index
CHAPTER 4 199
Frequency Response Analysis
ENDDATA
Figure 4-23 shows a logarithmic plot of the z-displacement magnitude of grid point 999, which is the
concentrated mass at the center of the cutout.
10
1.000
0.100
Displacement (m)
0.010
0.001
0.0001
0 Frequency (Hz) 100
Main Index
200 MD Nastran Dynamic Analysis User’s Guide
Examples
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 5: Transient Response Analysis
Overview
Direct Transient Response Analysis
Modal Transient Response Analysis
Initial Condition Usage in Linear Transient Analysis
Dynamic Data Recovery in Modal Transient Response Analysis
Modal Versus Direct Transient Response
Transient Excitation Definition
Integration Time Step
Transient Excitation Considerations
Solution Control for Transient Response Analysis
Restarts in Linear Transient Analysis
Examples
MAX/MIN Output (New Form)
Main Index
202 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
Transient response analysis is the most general method for computing forced dynamic response. The
purpose of a transient response analysis is to compute the behavior of a structure subjected to
time-varying excitation. The transient excitation is explicitly defined in the time domain. All of the
forces applied to the structure are known at each instant in time. Forces can be in the form of applied
forces and/or enforced motions (see Enforced Motion, 281).
The important results obtained from a transient analysis are typically displacements, velocities, and
accelerations of grid points, and forces and stresses in elements.
Depending upon the structure and the nature of the loading, two different numerical methods can be used
for a transient response analysis: direct and modal. The direct method performs a numerical integration
on the complete coupled equations of motion. The modal method utilizes the mode shapes of the
structure to reduce and uncouple the equations of motion (when modal or no damping is used); the
solution is then obtained through the summation of the individual modal responses. The choice of the
approach is problem dependent. The two methods are described in Direct Transient Response Analysis,
203 and Modal Transient Response Analysis, 208.
Main Index
CHAPTER 5 203
Transient Response Analysis
M u·· t + B u· t + K u t = P t (5-1)
The fundamental structural response (displacement) is solved at discrete times, typically with a fixed
integration time step t .
By using a central finite difference representation for the velocity u· t and the acceleration u·· t
at discrete times,
1
u· n = --------- u n + 1 – u n – 1
2t
1
u·· n = -------- u n + 1 – 2u n + u n – 1
2
t (5-2)
and averaging the applied force over three adjacent time points, the equation of motion can be rewritten
as:
M B
-------- u n + 1 – 2u n + u n – 1 + --------- u n + 1 – u n – 1 (5-3)
t
2 2t
K 1
+ ---- u n + 1 + u n + u n – 1 = --- P n + 1 + P n + P n – 1
3 3
A1 un + 1 = A2 + A3 un + A4 un – 1 (5-4)
where:
A1 M + B +K
= -------
- --------- ----
t
2 2t 3
A 2 = 1--- P n + 1 + P n + P n – 1
3
A 3 = 2M K
-------- – ----
2
t 3
M + B –K
A 4 = – -------
- --------- ----
2
t 2t 3
Matrix A 1 is termed the dynamic matrix, and A 2 is the applied force (averaged over three adjacent
time points). This approach is similar to the classical Newmark-Beta direct integration method except
that P t is averaged over three time points and K is modified such that the dynamic equation of
motion reduces to a static solution K u n = P n if no M or B exists.
Main Index
204 MD Nastran Dynamic Analysis User’s Guide
Direct Transient Response Analysis
The transient solution is obtained by decomposing A 1 and applying it to the right-hand side of the above
equation. In this form, the solution behaves like a succession of static solutions with each time step
performing a forward-backward substitution (FBS) on a new load vector. Note that the transient nature
of the solution is carried through by modifying the applied force matrix A 2 with the A 3 and A 4
terms.
In its simplest form, the M , B , and K matrices are assumed to be constant throughout the analysis
and do not change with time. Special solution methods are available in MD Nastran for variations in
these matrices.
A significant benefit presents itself if t remains constant during the analysis. With a constant t , the
A 1 matrix needs to be decomposed only once. Each progressive step in the analysis is only an FBS of
a new load vector. If t is changed, A 1 must be redecomposed, which can be a costly operation in
large problems.
Another efficiency in the direct transient solution is that the output time interval may be greater than the
solution time interval. In many cases it is not necessary to sample output response at each solution time.
For example, if the solution is performed every 0.001 second the results can be output every fifth time
step or every 0.005 second. This efficiency reduces the amount of output.
1 2 G 1
B = B + B + ------- K + ------- G E K E (5-5)
W3 W4
where:
1
B = damping elements (CVISC, CDAMPi) + B2GG
2
B = B2PP direct input matrix + transfer functions
G = overall structural damping coefficient (PARAM,G)
W3 = frequency of interest in radians per unit time (PARAM,W3) for the conversion of overall
structural damping into equivalent viscous damping
K = global stiffness matrix
GE = element structural damping coefficient (GE on the MATi entry)
W4 = frequency of interest in radians per unit time (PARAM,W4) for conversion of element
structural damping into equivalent viscous damping
K E = element stiffness matrix
Main Index
CHAPTER 5 205
Transient Response Analysis
Transient response analysis does not permit the use of complex coefficients. Therefore, structural
damping is included by means of equivalent viscous damping. To appreciate the impact of this on the
solution, a relation between structural damping and equivalent viscous damping must be defined.
The viscous damping force is a damping force that is a function of a damping coefficient b and the
velocity. It is an induced force that is represented in the equation of motion using the [B] matrix and
velocity vector.
M u·· t + B u· t + K u t = P t (5-6)
The structural damping force is a displacement-dependent damping. The structural damping force is a
function of a damping coefficient G and a complex component of the structural stiffness matrix.
M u·· t + 1 + iG K u t = P t (5-7)
Assuming constant amplitude oscillatory response for an SDOF system, the two damping forces are
identical if
Gk = b (5-8)
or
b = Gk
------- (5-9)
Therefore, if structural damping G is to be modeled using equivalent viscous damping b , then the
equality Eq. (5-9) holds at only one frequency (see Figure 5-1).
Two parameters are used to convert structural damping to equivalent viscous damping. An overall
structural damping coefficient can be applied to the entire system stiffness matrix using PARAM,W3,r
where r is the circular frequency at which damping is to be made equivalent. This parameter is used in
conjunction with PARAM,G, GFL. The default value for W3 is 0.0, which causes the damping related
to this source to be ignored in transient analysis.
PARAM,W4,r is an alternate parameter used to convert element structural damping to equivalent
viscous damping. PARAM,W4,r is used where r is the circular frequency at which damping is to be
made equivalent. PARAM,W4 is used in conjunction with the GE field on the MATi entry. The default
value for W4 is 0.0 which causes the related damping terms to be ignored in transient analysis.
Units for PARAM,W3 and PARAM,W4 are radians per unit time. The choice of W3 or W4 is typically
the dominant frequency at which the damping is active. Often, the first natural frequency is chosen, but
isolated individual element damping can occur at different frequencies and can be handled by the
appropriate data entries.
Main Index
206 MD Nastran Dynamic Analysis User’s Guide
Direct Transient Response Analysis
Equivalent
b = Gk 3 (or 4 )
Viscous
Damping
3 (or 4 )
f v = bu· = ibu
u – 1 = u 0 – u· 0 t (5-10)
P – 1 = K u – 1 + B u· 0 (5-11)
P 0 = K u 0 + B u· 0 (5-12)
Regardless of the initial conditions specified, the initial acceleration for all points in the structure is
assumed to be zero (constant initial velocity).
The format for the TIC Bulk Data entry is
1 2 3 4 5 6 7 8 9 10
TIC SID G C U0 V0
Main Index
CHAPTER 5 207
Transient Response Analysis
Field Contents
SID Set ID specified by the IC Case Control command.
G Grid, scalar, or extra point.
C Component number.
U0 Initial displacement.
V0 Initial velocity.
Initial conditions may be specified only in the a-set (see The Set Notation System Used in Dynamic
Analysis (App. B).
Main Index
208 MD Nastran Dynamic Analysis User’s Guide
Modal Transient Response Analysis
As a first step in the formulation, transform the variables from physical coordinates u to modal
coordinates by
The mode shapes are used to transform the problem in terms of the behavior of the modes as opposed
to the behavior of the grid points. Equation (5-13) represents an equality if all modes are used; however,
because all modes are rarely used, the equation usually represents an approximation.
To proceed, temporarily ignore the damping, resulting in the equation of motion
M u·· t + K u t = P t (5-14)
If the physical coordinates in terms of the modal coordinates (Eq. (5-13) is substituted into Eq. (5-14)),
the following equation is obtained:
··
Mt + K t = Pt (5-15)
This is now the equation of motion in terms of the modal coordinates. At this point, however, the
equations remain coupled.
T ·· T T
M + K = Pt (5-16)
where:
T
M = modal (generalized) mass matrix
T
K = modal (generalized) stiffness matrix
T
P = modal force vector
The final step uses the orthogonality property of the mode shapes to formulate the equation of motion in
terms of the generalized mass and stiffness matrices that are diagonal matrices. These matrices do not
have off-diagonal terms that couple the equations of motion. Therefore, in this form, the modal equations
of motion are uncoupled. In this uncoupled form, the equations of motion are written as a set on
uncoupled SDOF systems as
··
mi i t + ki i t = pi t (5-17)
Main Index
CHAPTER 5 209
Transient Response Analysis
where:
Note that there is no damping in the resulting equation. The next subsection describes how to include
damping in modal transient response.
Once the individual modal responses i t are computed, physical responses are recovered as the
summation of the modal responses
Another efficiency option in the modal transient solution is that the output time interval may be greater
than the solution time interval. In many cases, it is not necessary to sample output response at each
solution time. For example, if the solution is performed every 0.001 second, the results can be output
every fifth time step or every 0.005 second. This efficiency reduces the amount of output.
T
B diagonal (5-19)
In the presence of a B matrix, the modal transient approach solves the coupled problem in terms of
modal coordinates using the direct transient numerical integration approach described in Direct Transient
Response Analysis, 203 as follows:
A1 n + 1 = A2 + A3 n + A4 n – 1 (5-20)
where:
T M B K
A 1 = -------2- + --------
2t
- + ----
3
t
1 T
A 2 = --3- P n + 1 + P n + P n – 1
Main Index
210 MD Nastran Dynamic Analysis User’s Guide
Modal Transient Response Analysis
T 2M K
A 3 = -------2- – ---3-
t
T M B K
A 4 = – -------2- + 2t
--------- – ----
3
t
These equations are similar to the direct transient method except that they are in terms of modal
coordinates. Since the number of modes used in a solution is typically much less than the number of
physical variables, the direct integration of the modal equations is not as costly as with physical variables.
If damping is applied to each mode separately, the decoupled equations of motion can be maintained.
When modal damping is used, each mode has damping b i . The equations of motion remain uncoupled
and have the following form for each mode:
·· ·
mi i t + bi i t + ki i t = pi t (5-21)
or
·· · 2 1
i t + 2 i i i t + i i t = ----- p i t (5-22)
mi
where:
The TABDMP1 Bulk Data entry defines the modal damping ratios. A table is created by the
frequency-damping pairs specified on a TABDMP1 entry. The solution refers to this table for the
damping value to be applied at a particular frequency. The TABDMP1 Bulk Data entry has a Set ID. A
particular TABDMP1 table is activated by selecting the Set ID with SDAMPING = Set ID Case Control
command.
1 2 3 4 5 6 7 8 9 10
TABDMP1 ID TYPE
f1 g1 f2 g2 f3 g3 -etc.- ENDT
Field Contents
TID Table identification number.
TYPE Type of damping units:
G (default)
CRIT
Q
Main Index
CHAPTER 5 211
Transient Response Analysis
Field Contents
fi Frequency value (cycles per unit time).
gi Damping value in the units specified.
At resonance, the three types of damping are related by the following equations:
b G
i = ------i- = -----i
b cr 2
b cr = 2m i i
1 1
Q i = -------- = -----
2 i Gi
(5-23)
The values of fi (units = cycles per unit time) and gi define pairs of frequencies and dampings. Note
that gi can be entered as structural damping (default), critical damping, or quality factor. The entered
damping is internally converted to structural damping using Eq. (5-23). Straight-line interpolation is used
for modal frequencies between consecutive fi values. Linear extrapolation is used at the ends of the
table. ENDT ends the table input.
For example, if modal damping is entered using Table 5-1 and if modes exist at 1.0, 2.5, 3.6, and 5.5 Hz,
MD Nastran interpolates and extrapolates as shown in Figure 5-2 and the table. Note that there is no table
entry at 1.0 Hz; MD Nastran uses the first two table entries at f = 2.0 and f = 3.0 to extrapolate the
value for f = 1.0 .
0.20
+
+
+ +
0.10
+ = entered value
= computed value
0.0
2.0 4.0 6.0
f(Hz)
Main Index
212 MD Nastran Dynamic Analysis User’s Guide
Modal Transient Response Analysis
Entered Computed
f z f z
2.0 0.16 1.0 0.14
3.0 0.18 2.5 0.17
4.0 0.13 3.6 0.15
6.0 0.13 5.5 0.13
1 2 3 4 5 6 7 8 9 10
TABDMP1 10 CRIT +TAB1
+TAB1 2.0 0.16 3.0 0.18 4.0 0.13 6.0 0.13 +TAB2
+TAB2 ENDT
With the modal equations in the form of Eq. (5-22), an efficient uncoupled analytical integration
algorithm is used to solve for modal response as decoupled SDOF systems. Each of the modal responses
is computed using
·
–b t 2 m o + b 2m o
t = e o cos d t + --------------------------------------- sin d t (5-24)
d
t
–b t 2 m 1 b 2m
m d
+e ----------- e p sin d t – d
0
In a modal transient analysis, you may add nonmodal damping (CVISC, CDAMPi, GE on the MATi
entry, or PARAM,G). With nonmodal damping, there is a computational penalty due to the coupled B
matrix, causing the coupled solution algorithm to be used. In modal transient response analysis, it is
recommended that you use only modal damping (TABDMP1). If discrete damping is desired, direct
transient response analysis is recommended.
Main Index
CHAPTER 5 213
Transient Response Analysis
The frequency range selected on the eigenvalue entry (EIGRL or EIGR) is one means to control the
frequency range used in the transient response solution. Also, three parameters are available to limit the
number of modes included in the solution. PARAM,LFREQ gives the lower limit on the frequency range
of retained modes, and PARAM,HFREQ gives the upper limit on the frequency range of retained modes.
PARAM,LMODES gives the number of the lowest modes to be retained. These parameters can be used
to include the desired set of modes. Note that the default is for all computed modes to be retained.
The MODESELECT Case Control command permits the user to specify ALL data related to mode
selection without the need for any parameters. The command, which can be employed for selecting either
structure modes or fluid modes, offers five different and distinct options.
1. Mode selection based on arbitrary mode numbers.
2. Mode selection based on the number of lowest modes. This option is similar to the usage of the
LMODES/LMODESFL parameter.
3. Mode selection based on range of mode numbers. This option can be regarded as a variation of
options (1) and (2) above.
4. Mode selection based on frequency range. This option is similar to the usage of the
LFREQ/LFREQFL and HFREQ/HFREQFL parameters. However, this option is more general
since it also allows for the UNCONDITIONAL inclusion or exclusion of selected modes
regardless of their frequencies.
5. Mode selection based on modal effective mass fraction (MEFFMFRA) criteria. This powerful
option allows the user to select modes based on different MEFFMFRA criteria. Further, like
Option (4) above, it also allows for the UNCONDITIONAL inclusion or exclusion of selected
modes regardless of their MEFFMFRA values.
The details and usage are clearly described in the command MODESELECT (Ch. 4) in the MD Nastran
Quick Reference Guide.
It is very important to remember that truncating modes in a particular frequency range may truncate a
significant portion of the behavior in that frequency range. Typically, high-frequency modes are
truncated because they are more costly to compute. So, truncating high-frequency modes truncates high
frequency response. In most cases, high-frequency mode truncation is not of concern. You should
evaluate the truncation in terms of the loading frequency and the important characteristic frequencies of
the structure.
Methods for improving solution accuracy when truncating modes are discussed in Modal Augmentation
Methods (Ch. 12). The mode acceleration method is an ‘after-the-fact’ correction of the results that can
improve accuracy since detailed local stresses and forces are subject to mode truncation and may not be
as accurate as the results computed with the direct method. A newer, preferred method of modal
augmentation is the Residual Vector method, as the ‘correction’ is done by appending additional shapes
to the retained mode set to account for the quasi-static influence of the truncated high-frequency modes.
Main Index
214 MD Nastran Dynamic Analysis User’s Guide
Initial Condition Usage in Linear Transient Analysis
Note: When initial conditions are specified in modal transient analysis (SOL 112), it is important to
employ residual vector processing in order to ensure accurate results. However, residual
vector processing is the default scenario in all modal solution sequences.
IC
The format of the IC Case Control command is as follows:
Format:
PHYSICAL
IC MODAL = n
STATSUB[,DIFFK]
Examples:
IC = 10
IC(PHYSICAL) = 100
IC(MODAL) = 200
IC(STATSUB) = 1000
IC(STATSUB,DIFFK) = 2000
Describer Meaning
PHYSICAL The TIC Bulk Data entries selected by set n define initial conditions for coordinates
involving grid, scalar, and extra points. (Default).
MODAL The TIC Bulk Data entries selected by set n define initial conditions for modal
coordinates and extra points. See Remark 3.
STATSUB Use the solution of the static analysis subcase n as the initial condition. See
Remark 4.
Main Index
CHAPTER 5 215
Transient Response Analysis
Describer Meaning
DIFFK Include the effects of differential stiffness in the solution. See Remarks 4. and 5.
n For the PHYSICAL (the default) and MODAL options, n is the set identification
number of TIC Bulk Data entries for structural analysis (SOL 109, 112 and 129) or
TEMP and TEMPD entries for heat transfer analysis (SOL 159). For the STATSUB
option, n is the ID of a static analysis subcase. (Integer > 0)
Remarks:
1. For structural analysis, TIC entries will not be used (therefore, no initial conditions) unless
selected in the Case Control Section.
2. Only the PHYSICAL option (the default) may be specified in heat transfer analysis (SOL 159).
3. IC(MODAL) may be specified only in modal transient analysis (SOL 112).
4. IC(STATSUB) and IC(STATSUB,DIFFK) may not both be specified in the same execution.
5. The DIFFK keyword is meaningful only when used in conjunction with the STATSUB keyword.
Examples
1. The following example specifies initial conditions in physical coordinates using SOL 109 or
SOL 112.
IC(PHYSICAL) = 100
or
IC = 100
2. The following example specifies initial conditions in modal coordinates using SOL 112.
IC(MODAL) = 200
3. The following example uses the static solution from Subcase 10 as the initial condition in
SOL 109 or SOL 112. The differential stiffness effect is not included.
SUBCASE 10 $ STATIC ANALYSIS
LOAD = 100
$
SUBCASE 20
IC(STATSUB) = 10
4. The following example uses the static solution from Subcase 100 as the initial condition in
SOL 109 or SOL 112. The differential stiffness effect is included.
SUBCASE 100
LOAD = 1000
$
SUBCASE 200
IC(STATSUB,DIFFK) = 100
Main Index
216 MD Nastran Dynamic Analysis User’s Guide
Initial Condition Usage in Linear Transient Analysis
Main Index
CHAPTER 5 217
Transient Response Analysis
Main Index
218 MD Nastran Dynamic Analysis User’s Guide
Modal Versus Direct Transient Response
In general, larger models may be solved more efficiently in modal transient response because the
numerical solution is a solution of a smaller system of uncoupled equations. This result is certainly true
if the natural frequencies and mode shape were computed during a previous stage of the analysis. Using
Duhamel’s integral to solve the uncoupled equations is very efficient even for very long duration
transients. On the other hand, the major portion of the effort in a modal transient response analysis is the
calculation of the modes. For large systems with a large number of modes, this operation can be as costly
as direct integration. This is especially true for high-frequency excitation. To capture high frequency
response in a modal solution, less accurate high-frequency modes must be computed. For small models
with a few time steps, the direct method may be the most efficient because it solves the equations without
first computing the modes. The direct method is more accurate than the modal method because the direct
method is not concerned with mode truncation.
Table 5-2 provides a starting place for evaluating which method to use. Many additional factors may be
involved in the choice of a method, such as contractual obligations or local standards of practice.
Main Index
CHAPTER 5 219
Transient Response Analysis
The particular entry chosen for defining the dynamic loading is largely a function of user convenience
for concentrated loads. Pressure and distributed loads, however, require a more complicated format.
There are two important aspects of dynamic load definition. First, the location of the loading on the
structure must be defined. Since this characteristic locates the loading in space, it is called the spatial
distribution of the dynamic loading. Secondly, the time variation in the loading is the characteristic that
differentiates a dynamic load from a static load. This time variation is called the temporal distribution
of the load. A complete dynamic loading is a product of spatial and temporal distributions.
Using Table IDs and Set IDs in MD Nastran makes it possible to apply many complicated and temporally
similar loadings with a minimum of input. Combining simple loadings to create complicated loading
distributions that vary in position as well as time is also a straightforward task.
The remainder of this section describes the Bulk Data entries for transient excitation. The description is
given in terms of the coefficients that define the dynamic load.
1 2 3 4 5 6 7 8 9 10
TLOAD1 SID EXCITEID DELAYI/ TYPE TID US0 VS0
DELAYR
Main Index
220 MD Nastran Dynamic Analysis User’s Guide
Transient Excitation Definition
Example:
TLOAD1 5 7 15 LOAD 13
Field Contents
SID Set identification number. (Integer > 0)
EXCITEID Identification number of DAREA or SPCD entry set or a thermal load set (in heat
transfer analysis) that defines A . See Remarks 2. and 3. (Integer > 0)
DELAYI Identification number of DELAY Bulk Data entry that defines time delay . See
Remark 9. (Integer > 0 or blank)
DELAYR Value of time delay that will be used for all degrees-of-freedom that are excited by
this dynamic load entry. See Remark 9. (Real or blank)
TYPE Defines the type of the dynamic excitation. See Remarks 2. and 3. (Integer, character
or blank; Default = 0)
TID Identification number of TABLEDi entry that gives F t . (Integer > 0)
US0 Factor for initial displacements of the enforced degrees-of-freedom. See Remarks 10.
and 12. (Real; Default = 0.0)
VS0 Factor for initial velocities of the enforced degrees-of-freedom. See Remarks 11. and
12. (Real; Default = 0.0)
Remarks:
1. Dynamic excitation sets must be selected with the Case Control command DLOAD = SID.
2. The type of the dynamic excitation is specified by TYPE (field 5) according to the following table:
Main Index
CHAPTER 5 221
Transient Response Analysis
The TLOAD2 Bulk Data entry defines dynamic excitation in the form:
0 , t T1 + or t T2 +
Pt = (5-26)
B C t̃
A t̃ e cos 2F t̃ + P , T1 + t T2 +
where: t̃ = t – T1 – .
.
1 2 3 4 5 6 7 8 9 10
TLOAD2 SID EXCITEID DELAYI/ TYPE T1 T2 F P
DELAYR
C B US0 VS0
Example:
Field Contents
SID Set identification number. (Integer > 0)
EXCITEID Identification number of DAREA or SPCD entry set or a thermal load set (in heat
transfer analysis) that defines A . See Remarks 2. and 3. (Integer > 0)
DELAYI Identification number of DELAY Bulk Data entry that defines time delay . See
Remark 5. (Integer > 0 or blank)
DELAYR Value of time delay that will be used for all degrees-of-freedom that are excited by
this dynamic load entry. See Remark 5. (Real or blank)
TYPE Defines the type of the dynamic excitation. See Remarks 2. and 3. (Integer; character
or blank; Default = 0)
T1 Time constant. (Real > 0.0)
T2 Time constant. (Real; T2 > T1)
F Frequency in cycles per unit time. (Real > 0.0; Default = 0.0)
P Phase angle in degrees. (Real; Default = 0.0)
C Exponential coefficient. (Real; Default = 0.0)
B Growth coefficient. (Real; Default = 0.0)
US0 Factor for initial displacements of the enforced degrees-of-freedom. See Remarks 10.
and 12. (Real; Default = 0.0)
VSO Factor for initial velocities of the enforced degrees-of-freedom. See Remarks 11. and
12. (Real; Default = 0.0)
Main Index
222 MD Nastran Dynamic Analysis User’s Guide
Transient Excitation Definition
1 2 3 4 5 6 7 8 9 10
DAREA SID P1 C1 A1 P2 C2 A2
Field Contents
SID Set ID specified by TLOADi entries.
Pi Grid, extra, or scalar point ID.
Ci Component number.
Ai Scale factor.
A DAREA entry is selected by the TLOAD1 or TLOAD2 entry. Any number of DAREA entries may
be used; all those with the same SID are combined.
1 2 3 4 5 6 7 8 9 10
DELAY SID P1 C1 1 P2 C2 2
Field Contents
SID Set ID specified by TLOADi entry.
Pi Grid, extra, or scalar point ID.
Ci Component number.
i Time delay for Pi, Ci.
A DAREA entry must be defined for the same point and component.
Any number of DELAY entries may be used; all those with the same SID are combined.
Main Index
CHAPTER 5 223
Transient Response Analysis
The TABLED1, TABLED2, and TABLED3 Bulk Data entries linearly interpolate between the end
points and linearly extrapolate outside of the endpoints as shown in Figure 5-3. The TABLED1 entry has
the option to perform logrithmic interpolation between points. The TABLED4 Bulk Data entry uses the
endpoint values for values beyond the endpoints.
1 2 3 4 5 6 7 8 9 10
TABLED1 TID XAXIS YAXIS
x1 y1 x2 y2 x3 y3 -etc.- ENDT
Field Contents
TID Table identification number.
XAXIS Specifies a linear or logarithmic interpolation for the x-axis.
(Character: “LINEAR” or “LOG”; Default = “LINEAR”)
YAXIS Specifies a linear or logarithmic interpolation for the y-axis.
(Character: “LINEAR” or “LOG”; Default = “LINEAR”)
xi, yi Tabular values. Values of x are frequency in cycles per unit time.
ENDT Ends the table input.
y = yT x (5-27)
Main Index
224 MD Nastran Dynamic Analysis User’s Guide
Transient Excitation Definition
XAXIS YAXIS yT x
LINEAR LINEAR
xj – x x – xi
---------------- yi + ---------------- yj
xj – xi xj – xi
LOG LINEAR
ln xj x ln x xi
------------------------- yi + ------------------------- yj
ln xj xi ln xj xi
LINEAR LOG
xj – x x – xi
exp ---------------- ln yi + ---------------- ln yj
xj – xi xj – xi
LOG LOG
ln xj x ln x xi
exp ------------------------- ln yi + ------------------------- ln yj
ln xj xi ln xj xi
1 2 3 4 5 6 7 8 9 10
TABLED2 TID X1
x1 y1 x2 y2 x3 y3 -etc.- ENDT
Field Contents
TID Table identification number.
X1 Table parameter.
xi, yi Tabular values.
ENDT Ends the table input.
y = y T x – X1 (5-28)
1 2 3 4 5 6 7 8 9 10
TABLED3 TID X1 X2
x1 y1 x2 y2 x3 y3 -etc.- ENDT
Main Index
CHAPTER 5 225
Transient Response Analysis
Field Contents
TID Table identification number.
X1, X2 Table parameters.
xi, yi Tabular values.
ENDT Ends the table input.
x – X1
y = y T ---------------- (5-29)
X2
1 2 3 4 5 6 7 8 9 10
TABLED4 TID X1 X2 X3 X4
A0 A1 A2 A3 A4 A5 -etc.- ENDT
Field Contents
TID Table identification number.
Xi Table parameters. X2 0.0; X3 X4 .
Ai Coefficients.
N is the degree of the power series. When x X3 , X3 is used for x ; when x X4 , X4 is used for x .
This condition has the effect of placing bounds on the table; there is no extrapolation outside of the table
boundaries.
ENDT ends the table input.
DAREA Example
Suppose the following command is in the Case Control Section:
DLOAD = 35
Main Index
226 MD Nastran Dynamic Analysis User’s Guide
Transient Excitation Definition
1 2 3 4 5 6 7 8 9 10
$TLOAD1 SID DAREA DELAY TYPE TID
TLOAD1 35 29 31 40
$DAREA SID POINT COMPONENT SCALE
DAREA 29 30 1 4.0
$DELAY SID POINT COMPONENT LAG
DELAY 31 30 1 0.2
$TABLED1 ID
$ X1 y1 x2 y2 x3 y3 x4 y4
TABLED1 40 LINEAR LINEAR
0.0 0.0 0.3 1.0 2.0 1.0 ENDT
The DLOAD Set ID 35 in the Case Control selects the TLOAD1 entry in the Bulk Data having a Set ID
35. On the TLOAD1 entry is a reference to DAREA Set ID 29, DELAY Set ID 31, and TABLED1 Set
ID 40. The DAREA entry with Set ID 29 positions the loading on grid point 30 in the 1 direction with a
scale factor of 4.0 applied to the load. The DELAY entry with Set ID 31 delays the loading on grid point
30 in the 1 direction by 0.2 units of time. The TABLED1 entry with Set ID 40 defines the load time
history in tabular form. The result of these entries is a dynamic load applied to grid point 30, component
T1, scaled by 4.0 and delayed by 0.2 units of time.
Figure 5-4 shows the TABLED1 time history and the applied load (scaled by the DAREA entry and time
shifted by the DELAY entry).
Main Index
CHAPTER 5 227
Transient Response Analysis
1.0
Amplitude
on
TABLED1
0
0.5 1.0 1.5 2.0 2.5
Time (sec)
4.0
Force
Amplitude
Time (sec)
Figure 5-4 Time History from the TABLED1 Entry (Top) and Applied Load (Bottom)
Main Index
228 MD Nastran Dynamic Analysis User’s Guide
Transient Excitation Definition
DLOAD
Case Control
Bulk Data
TLOADi
P = S Si Pi (5-31)
i
where:
1 2 3 4 5 6 7 8 9 10
DLOAD SID S S1 L1 S2 L2 -etc.-
Field Contents
SID Load set ID.
S Overall scale factor.
Main Index
CHAPTER 5 229
Transient Response Analysis
Field Contents
Si Individual scale factors.
Li Load set ID number for TLOAD1 and TLOAD2 entries.
1 2 3 4 5 6 7 8 9 10
$DLOAD SID S S1 L1 S2 L2 -etc.-
DLOAD 33 3.25 0.5 14 2.0 27
A dynamic Load Set ID of 33 is created by taking 0.5 times the loads in Load Set ID of 14, adding to it
2.0 times the loads in Load Set ID of 27, and multiplying that sum by an overall scale factor of 3.25.
As with other transient loads, a dynamic load combination defined by the DLOAD Bulk Data entry is
selected by the DLOAD Case Control command.
Main Index
230 MD Nastran Dynamic Analysis User’s Guide
Integration Time Step
You may change t during a run but doing so causes the dynamic matrix to be redecomposed, which
can be costly in direct transient response analysis.
The TSTEP Bulk Data entry has the following format:
1 2 3 4 5 6 7 8 9 10
TSTEP SID N1 t 1 NO1
N2 t 2 NO2
-etc.-
Field Contents
SID Set ID specified by a TSTEP Case Control command.
Ni Number of time steps of value t i .
t i Integration time step.
NOi Output every NOi-th time step.
Main Index
CHAPTER 5 231
Transient Response Analysis
Main Index
232 MD Nastran Dynamic Analysis User’s Guide
Solution Control for Transient Response Analysis
The types of results available from a transient response analysis are similar to those for a static analysis
except that the results vary with time. Additional quantities are also available, which is characteristic of
dynamics problems. The output quantities are summarized in Table 5-5 and Table 5-6.
Main Index
CHAPTER 5 233
Transient Response Analysis
Because the results may be output for many time steps, the volume of output can be very large. Prudent
selection of the output quantities is recommended.
A number of Bulk Data entries are unique to transient response analysis. They can be combined with
other generic entries in the Bulk Data. Bulk Data entries directly related to transient analysis are
summarized in Table 5-7.
Main Index
234 MD Nastran Dynamic Analysis User’s Guide
Solution Control for Transient Response Analysis
Main Index
CHAPTER 5 235
Transient Response Analysis
Main Index
236 MD Nastran Dynamic Analysis User’s Guide
Examples
Examples
This section provides several examples showing the input and output. These examples are
Two-DOF Model
Consider the two-DOF system shown in Figure 5-6. Direct transient response (SOL 109) is run with an
initial displacement of 0.1 meter at grid point 2. The analysis is run for a duration of 10 seconds with a
t of 0.01 second. Damping is neglected in the analysis. Listing 5-1 shows part of the input file. See
Install_dir/mdxxxx/doc/dynamics/db05two.dat.
Grid Point 1 m1
k1
Grid Point 2 m2
k2
x
Figure 5-6 Two-DOF Model
Main Index
CHAPTER 5 237
Transient Response Analysis
Main Index
238 MD Nastran Dynamic Analysis User’s Guide
Examples
Table 5-8 Relationship Between Case Control Commands and Bulk Data Entries for the
Two-DOF Model
Case Control Bulk Data
TSTEP TSTEP
IC TIC
Figure 5-7 shows the plots of the resulting displacements for grid points 1 and 2. Note that there are two
frequencies of response: a higher frequency of about 5 Hz, and a lower frequency of about 0.25 Hz. The
energy (and hence response) appears to be transferred repetitively between grid points 1 and 2 as
represented by the lower frequency response. This energy transfer is called beating. Beating occurs
when there are closely-spaced modes (in this case, 4.79 Hz and 5.29 Hz) in which energy transfer can
readily occur. The response is comprised of two frequencies as given below:
f1 + f2
f h i g h e r = -----------------
2
f2 – f1
f lowe r = -----------------
2 (5-32)
where:
In this example, f higher is 5.04 Hz and f lower is 0.25 Hz. The lower frequency is called the beat frequency
and is the frequency at which energy transfer occurs.
Main Index
CHAPTER 5 239
Transient Response Analysis
Main Index
240 MD Nastran Dynamic Analysis User’s Guide
Examples
6 11
x
p6(t) p11(t)
Force 6 (N)
Time
(sec)
-3
0.1 0.5
Force 11 (N)
Time
(sec)
-6
0.5
Figure 5-9 Applied Loads for the Beam Model
Main Index
CHAPTER 5 241
Transient Response Analysis
Main Index
242 MD Nastran Dynamic Analysis User’s Guide
Examples
$DAREA SID P1 C1 A1
DAREA 241 11 2 6.0
DAREA 242 6 2 3.0
$DELAY SID P1 C1 T1
DELAY 262 6 2 0.1
$
... basic model ...
$
ENDDATA
Table 5-9 shows the relationship between the Case Control commands and the Bulk Data entries. The
DLOAD Bulk Data entry references two TLOAD2 entries, each of which references separate DAREA
entries. A TLOAD2 entry also references a DELAY entry to apply the time delay to the load at grid point
6.
Table 5-9 Relationship Between Case Control Commands and Bulk Data Entries for the
Bar Model
Case Control Bulk Data
METHOD EIGRL
TSTEP TSTEP
SDAMPING TABDMP1
DLOAD DLOAD
TLOAD2 231
DAREA
TLOAD2 232
DAREA
DELAY
Plotted output is shown in the following figures. Figure 5-10 shows the applied loads at grid points 6 and
11. Figure 5-11 shows the plots of the displacements for grid points 6 and 11. Figure 5-12 shows the
accelerations for grid points 6 and 11. Figure 5-13 shows the bending moment at end A in plane 1 for
element 6. Figure 5-14 shows the modal displacements for modes 1 and 2.
Main Index
CHAPTER 5 243
Transient Response Analysis
3.0
Force 6 (N)
0.0
-3.0
6.0
Force 11 (N)
0.0
-6.0
0.0 2.0
Time (sec)
Main Index
244 MD Nastran Dynamic Analysis User’s Guide
Examples
0.03
Displacement 6 (m)
0.0
-0.0
0.08
Displacement 11 (m)
0.0
-0.06
0.0 2.0
Time (sec)
Main Index
CHAPTER 5 245
Transient Response Analysis
4.0
Acceleration 6 m 2 sec
0.0
-6.0
10.0
Acceleration 11 m 2 sec
0.0
-15.0
0.0 2.0
Time (sec)
Main Index
246 MD Nastran Dynamic Analysis User’s Guide
Examples
25.0
Bending Moment A1 (Nm)
0.0
-20.0
0.0 2.0
Time (sec)
Main Index
CHAPTER 5 247
Transient Response Analysis
0.08
SDisplacement 1
0.0
-0.10
0.0015
SDisplacement 2
0.0
-0.0010
0.0 2.0
Time (sec)
Bracket Model
Consider the bracket model shown in Figure 5-15. A pressure load of 3 psi is applied to the elements in
the top face in the z-direction with the time history shown in Figure 5-16. The modal transient analysis
is run for 4 seconds with a time step size of 0.005 second. Modal damping of 2% critical damping is
used for all modes. Modes up to 3000 Hz are computed with the Lanczos method. The model is
constrained near the base.
Main Index
248 MD Nastran Dynamic Analysis User’s Guide
Examples
Pressure
3 psi
0 0.10 0.15 5
Time (sec)
Listing 5-3 shows the abridged input file. See Install_dir/mdxxxx/doc/dynamics/bd05bkt.dat. The LSEQ
entry is used to apply the pressure loads (PLOAD4 entries). Note that the LSEQ and TLOAD1 entries
reference a common DAREA ID (999) and that there is no explicit DAREA entry. Table 5-10 shows the
relationship between the Case Control commands and the Bulk Data entries.
Main Index
CHAPTER 5 249
Transient Response Analysis
Main Index
250 MD Nastran Dynamic Analysis User’s Guide
Examples
$
... basic model ...
$
ENDDATA
Table 5-10 Relationship Between Case Control Commands and Bulk Data Entries for the
Bracket Model
Case Control Bulk Data
METHOD EIGRL
DLOAD TLOAD
SDAMPING TABDMP1
LOADSET LSEQ
Figure 5-17 shows a plot of the z-displacement of grid point 999, which is the concentrated mass at the
center of the cutout.
Main Index
CHAPTER 5 251
Transient Response Analysis
0
=Displacement (m)
-0.04
-0.08
0 Time (sec) 4
Main Index
252 MD Nastran Dynamic Analysis User’s Guide
MAX/MIN Output (New Form)
Main Index
CHAPTER 5 253
Transient Response Analysis
The grid classes are: DISPLACMENT, VELOCITY, ACCELERATION, OLOAD, SPCF (Single
Point Constraint Forces), MPCF (Multi-Point Constraint Forces), and GPKE (Grid Point Kinetic
Energy). The element classes are: STRESS, FORCE and STRAIN. When an element class is selected,
the “element types” area is required. Element types that share the same recovery components can be
entered on a single MAXMIN(DEF) entry.
The component area is always required and can either match NDDL item descriptions or the item codes
presented in Item Codes (p. 961) in the MD Nastran Quick Reference Guide. Note that components can
utilize grouping features, where the grouped components are compared to each other before they are
compared to the other action requested. When an element type has multiple recovery positions, an
example would be stresses for the HEXA element, where data recovery occurs at the center and vertices,
these locations can be limited or grouped by keyword utilization. These operators are /CENTER
(default), /ALL, or /GROUP for the repeating component locations; and /ENDS for the BEAM element.
An example for the selection of the first principal stress component at every data location on the HEXA
would be P1/ALL. This would cause monitoring and performance of data recovery for each of the
monitored nine locations per element.
Action keywords are optional with default values. These keywords consist of MAXALG, MINALG,
ABSOLUTE, and ALL that control the number of items retained during the component evaluations.
When none are specified, the default is ALL with a retained list of five per component. When
MAXALG, MINALG or ABSOLUTE actions are selected, the ones not referenced will default to a
retained set of zero. To override the number of retained values, an integer value of 1 to 25 can be
associated with the action by either an equal delimiter or by encapsulating the value in parenthesis, for
example, MAXA=5 or MINA(2).
The action keyword RMS, causes the root mean squared value to be calculated over the domain, but may
not be meaningful depending upon the interaction with other component actions such as grouping. The
action keywords of BRIEF (default) and FULL controls the amount of presented recovery information.
The BRIEF action only presents the components selected over the retained depth associated with the
selected items in the element or grid set. The FULL action causes the grid or element entry to be retained.
Those in the retained list will be shown in their standard data recovery format. See output section for
illustrations.
The last action keyword concerns the coordinate system in which the comparisons are performed.
Currently this action is only available for grid recovery classes. The action keyword CID={GLOBAL,
BASIC, cid} is used to override the default. The default of the grid comparison is different depending
upon the SDR2 module sort processing. When the matrices are input in SORT1 order, then the
comparison is performed in the BASIC reference system with the “cid” used to select an alternative.
GLOBAL is not supported. When the matrices are input in SORT2 order, the transpose of SORT1, then
the comparisons are performed in GLOBAL, with either BASIC or “cid” as alternative options. Note,
when FULL is requested along with a CID= action, then the entries displayed are in the CID selected
system. Standard data recovery is unaffected and still occurs in the global reference frame.
The MAX/MIN definitions are acted upon only if selected by the additional SUBCASE entries to select
the domain in classical Case Control syntax.
MAXMIN(GRID,[{PRINT/NOPRINT},PUNCH]) = {ALL,SET,NONE)
Main Index
254 MD Nastran Dynamic Analysis User’s Guide
MAX/MIN Output (New Form)
MAXMIN(ELEM,[{PRINT/NOPRINT},PUNCH],[VONMISES/SHEAR],[STRCUR/FIB
ER],[CENTER/BILIN|etc]) = {ALL,SET,NONE)
MAXMIN(BOTH,[{PRINT/NOPRINT},PUNCH],[VONMISES/SHEAR],[STRCUR/FIB
ER],[CENTER/BILIN|etc]) = {ALL,SET,NONE)
Since the MAXMIN(ELEM..) directives act upon all element classes defined within the
MAXMIN(DEF) section, the keywords modifying the data recovery operations for STRESS and
STRAIN are also included since the primary Case Control entries may not be present. The invariant
selection is independent between the MAXMIN and primary data recovery entry. When element strains
recovery is selected, the STRAIN entry when present in the SUBCASE supercedes the MAXMIN
selection concerning fiber or curvature locations for the plate elements. This is also true for the QUAD4
corner selection concerning the method for extrapolation to the vertices.
Outputs
The OMM data block conforms to standard OFP (Output File Processor) format. That is it contains pairs
of records. The first of the pair denoted as the IDENT or HEADER is 146 words in length and describes
the following data record. The data record contains entries that are eleven words in length and are
(perceivably) always in sort1 real format. The OMM data block can be identified as OFP output class
38. Care must be exercised interpreting the data record because item values in the entry change format
classes depending upon settings within the IDENT record. The NDDL contains a full description of the
control values. Another significant difference to the OMM data block when compared to other OFP
tables is that with the FULL action on the MAXMIN(DEF) entry, all other table classes can be contained
as record pairs within the OMM data block.
The output examples presented are intended to assist with column interpretation within analysis approach
and output classes. The two examples were create from a Modal Transient solution sequence. The OMM
table identification line contains the class name of the values presented in the table. The first table
indicates grid class response acceleration, and the second represents an element response, in this case a
BAR element and the component name indicating a stress response. For these two tables, the label of
column one changes from POINT to ELEMENT depending upon the information contained in the table.
The CID or Coordinate System Identification is always present, although currently meaningful for grid
recovery. The value “-1” indicates element 0 basic and positive values, the coordinate system used to
measure response. The OPERATION column contains two pieces of information from the data entry,
first the measure, MAX (maximum algebraic), MIN (minimum algebraic) and ABS (absolute value).
The second value is the retention depth varying from 1 to 25. The next column is labeled COMPONENT
and represents the keyword for the item selected from the NDDL description in the response table.
Additional keywords such as VONMISES that override the NDDL description for this release. The next
column is variable in column label and format. For our current example, the column header is TIME and
the entry item format is real. When Linear Statics is selected as the analysis approach, the column label
would be SUBCASE and the entry item format integer. The next column, VALUE, represents the real
value of the monitored response quantity. These columns form the minimum set of information
presented as basic output.
Main Index
CHAPTER 5 255
Transient Response Analysis
M A X / M I N S U M M A R Y ( ACCELERATION )
M A X / M I N S U M M A R Y ( BAR )
The above table illustrates that two optional columns can appear after the VALUE column. The one with
RMS as the label will always appear as the last column. The other additional column, currently has a
label of either GRID or SD (Station Distance), and appears between VALUE and RMS when the element
has multiple recovery positions, such as the HEXA element. When the GRID columns appears, it
contains two different formats depending upon the component actions selected on the MAXMIN(DEF)
entry. When the OPERATION column has a depth measure of “1”, the identification will appear in the
GRID column. The center of the element has a grid designation of “0”. For higher depth values, a blank
will appear in the GRID column when the /ALL option was selected. When the column item is non-
blank, then the /GROUP option was selected and the position within the element is being identified. The
SD column label indicates that the data entry contains real values for the station distance measure used
for the BEAM element.
Main Index
256 MD Nastran Dynamic Analysis User’s Guide
MAX/MIN Output (New Form)
the BEAM and BEND elements. The MAXMIN operations are performed for SDR2 data recovery
which currently does not include Layered Composites, Element Energies, or Grid Point Forces.
The sort2 Linear Statics solution sequence is currently not supported.
There is no support for complex data recovery for either grids or element for MAXMIN operations.
Examples
Example 1: Modal Transient Model of a Solid
The Case Control Section contains the MAXMIN(DEF) and MAXMIN(ELEM) entries, see
Install_dir/mdxxxx/doc/dynamics/mmhex04.dat. The specific MAXMIN(DEF) entry has been selected
to monitor stresses for the HEXA element and examine the Principal Stresses. Recovery locations are to
be monitored but a single value retained with its associated vertex grid identification. All possible
monitoring actions are to occur to a retained depth value of 10. The parenthesis usage surrounding the
components collapses all of the principals to a single value per element. The MAXMIN(ELEM)
references a set of one element.
MAXMIN(DEF) STRESS HEXA ( P1/group p2/group p3/group ) all=10
TITLE = modal transient with solid elements
SUBCASE 1
method 1
TSTEP = 1
SPC = 1
DLOAD = 2
set 200 = 1
maxmin(elem) = 200
The response recovered information indicated that for the algebraic maximums that the P1 direction over
time was the dominating value although the grid location was not the same for all retained values. The
algebraic minimums switched to the P2 direction, but had a similar pattern concerning vertex location.
The absolute value monitor, however, switched direction and vertex location over the time range. The
coordinate system of element comparison was the basic reference frame because that selection can be
accomplished by setting on the PSOLID entry.
Main Index
CHAPTER 5 257
Transient Response Analysis
SUBCASE 1
M A X / M I N S U M M A R Y ( HEXA )
This example of a Quarter Plate in Linear Statics demonstrates the usage of three MAXMIN(DEF)
entries and different options selected in the MAXMIN(ELEM) entries within the Case Control Section,
see Install_dir/mdxxxx/doc/dynamics/mmq401.dat. The usage of FULL produces limited data recovery
in standard formats encapsulated within the MAXMIN output. Only limited response information is
displayed for illustration purposes.
MAXMIN(DEF) STRAIN QUAD4 EMAX1 EMAX2 MAXA MINA FULL
MAXMIN(DEF) STRESS QUAD4 SMAX1 SMAX2 MAXA MINA FULL
maxmin(def) disp t1 t2 maxa mina full cid=basic
MAXLINES=1000000
echo=none
SPC=1
set 100 = 10 thru 14
SUBCASE 1001
TITLE=Only MAXMIN Shear Invariant, strcur
LOAD=101
maxmin(elem,shear) = all
DISP=100
maxmin(grid) = 100
SUBCASE 1002
TITLE=Only MAXMIN Von Mises Invariant, fiber
LOAD=101
maxmin(elem,fiber) = all
SUBCASE 1003
TITLE=MAXMIN Shear Stress Von Mises, fiber
LOAD=101
maxmin(elem,shear) = all
STRESS=ALL
strain(fiber)=all
BEGIN BULK
Main Index
258 MD Nastran Dynamic Analysis User’s Guide
MAX/MIN Output (New Form)
The recovery for the displacements within the MAXMIN output section was within the basic reference
system. One should observe here that the POINT identification is in retained sequence and that collation
order is by operational depth. Also the identification column has become SUBCASE due to the solution
approach.
M A X / M I N S U M M A R Y ( DISPLACEMENTS )
The standard output recovery for sort1 real is used for the FULL formatting. This output is recovered in
the basic reference system due the request upon the MAXMIN(DEF) entry. The normal POINT collation
order is maintained and the entries correspond to the retained list from the MAXMIN operation section.
SUBCASE 1001
D I S P L A C E M E N T V E C T O R
Once the Grid Point responses have been reported, the element requested are then presented. Because in
Linear Statics, the element reference system may not be consistent, it is suggested that the invariant be
used to determine the ordering for two dimensional elements. The results presented are for the first
SUBCASE which the SHEAR invariant was selected. Again note the element identification ordering is
by component and depth.
Main Index
3MD Nastran Dynamic Analysis User’s Guide
Chapter 6: Damping
6 Damping
Overview
Modeling Damping Effects
Damping Effect on Complex Eigenvalues
Structural Damping for Shell and Bush Elements
Geometric and Material Nonlinear Spring and Damper Element
Main Index
260 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
The physical causes of damping in dynamic analysis are any processes which dissipate energy or reduce
the structural response through internal friction. Furthermore, the internal velocities or displacements
cause reactive damping forces which are irreversible and nonconservative. Examples are mechanical
devices such as shock absorbers, the internal hysteresis that occurs in materials such as rubber, friction
in joints, and other nonlinear effects such as plastic strains in metals.
Main Index
CHAPTER 6 261
Damping
Viscous Damping
The primary method for modeling viscous damping is through the CVISC and CDAMPi Bulk Data
entries. These produce forces which are linearly proportional to the velocities of the connected grid
points. Another source of viscous damping is structural damping that must be converted to viscous
damping in the transient response solutions. Damping in Direct Transient Response, 204 covers this
topic. Viscous damping can be supplied at the basic matrix level using the Direct Matrix Input at
Gridpoints, DMIG, method. This form of viscous damping is selected with the CASE CONTROL
commands B2GG or B2PP.
Structural Damping
Structural damping is intended to simulate the effects of linear material energy loss proportional to the
strains. In other words, this method approximates effects similar to hysteresis. It is specified in the
material definition input, MATi, and on the parameter, G . Note that in the frequency response and
complex eigenvalue solutions, the structural damping produces imaginary numbers in the complex
stiffness matrix. In the transient solutions, the matrix terms are converted to equivalent viscous damping;
i.e., the forces will be proportional to the derivative of strain with respect to time.
The physical basis for the phenomenon of imaginary stiffness is explained by Figure 6-1. Shown is the
force due to a stiffness matrix term, K , plotted against the displacement of the point. As the
displacements cycle sinusoidally, the imaginary stiffness, G , causes a phase lag in the force response
resulting in an elliptical path. The area enclosed by the curve is equal to the dissipated energy.
Main Index
262 MD Nastran Dynamic Analysis User’s Guide
Modeling Damping Effects
Real
Force
Ku 0
= 0
GKu 0
Real
Displacement
u0
= 90
= 180
The basic definition for the steady-state actual displacements, in terms of the complex displacements is
it
u r t = Re u 0 e (6-1)
where u r is the actual real displacement, u 0 is normally a complex variable that we will set to a real
number, is the steady state frequency, and e ix = cos x + i sin x . If a structural damping coefficient,
G , is added to the stiffness matrix, the resulting complex forces are in the following form:
i t
F = 1 + iG Ku 0 e (6-2)
dW = F r du r (6-4)
Main Index
CHAPTER 6 263
Damping
Combining Eq. (6-3) through Eq. (6-5) and integrating over a full cycle, results in the work
2
2
W = – Ku 0 sin t cos t – G sin t dt (6-6)
0
Note that the elastic energy terms average zero over the interval, but the energy dissipated by the
structural damping exists. The area inside the curve in Figure 6-1 is equivalent to the damping energy
loss, W .
This damping method is a reasonable linear approximation to the classical hysteresis effect. It may also
be used to approximate other similar cyclic energy losses such as the effects of loose joints and fasteners.
(Hint: construct a similar trajectory curve and estimate the area.)
The main disadvantage of structural damping is that complex numbers must be converted to real numbers
in transient analysis; i.e., structural damping is converted to linear viscous damping. This is good only
when the response is dominated by a single known frequency.
Modal Damping
For the modal solution formulations, a special damping input is provided in addition to the other damping
terms described above. It is used primarily for efficiency and when test results or contract specifications
provide damping factors. These terms are applied only to the uncoupled modal equations. When
coupling effects are absent, the method avoids expensive matrix calculations.
A second reason for using modal damping is that modal testing may provide accurate damping inputs on
a mode-by-mode basis. These damping factors may be converted to a frequency-dependent table,
TABDMP. If the MD Nastran normal mode frequencies are close and the damping factors are small.
The third reason--when damping factors are specified by a third party--is frequently meant to keep the
analyst more conservative. By restricting the solutions to artificially low damping, this will force the
staff to solve their design problems for the worst case.
Modal damping is available only for the modal solutions method. The matrix terms generated for modal
analysis, and added to any direct matrix inputs, are
·· ·
mi i + bi i + ki i = Pi (6-8)
Main Index
264 MD Nastran Dynamic Analysis User’s Guide
Modeling Damping Effects
In accordance with modal definitions, the modal viscous damping coefficient b i may be expressed as:
bi = gi i mi (6-9)
where i is the undamped vibration frequency (equal to k i m i ), expressed in radians per unit time. The
dimensionless input coefficient, g i , is equal to twice the critical damping ratio of the mode. It is
evaluated by linear interpolation of a user-specified function of frequency, g f .
The physical meaning of modal damping is somewhat clouded. Its effect is viscous damping, but it is
proportional to the stiffness matrix, and varies with modal frequency. The damping effects are
distributed to the structure depending on the energy distribution in each mode shape.
Another aspect of modal damping to remember is that the damping coefficient, b , is constant for each
mode. If the mode is forcibly excited at a different frequency the initial damping factor still applies. In
other words, the damping at any particular frequency is a function of several modal damping factors.
This may cause unexpected results for frequencies where a mode is not dominant.
The user specifies pairs of values f g on a TABDMP1 Bulk Data entry, which is selected by the
SDAMPING Case Control command. There are also provisions for expressing damping as a fraction of
critical damping ( C C c ) or amplification quality factor ( Q ). If accurate test results are available, the user
can specify different damping coefficients, obtained from modal tests, for the different frequency ranges.
The user can ensure that the desired damping has been obtained for a set of modes whose frequencies are
known from a previous run by providing the desired f,g pairs on the TABDMP1 entry.
T v
B = bi + B (6-10)
where b i is a diagonal matrix whose elements are given by Eq. (6-9), is the matrix of eigenvectors,
and B v is the matrix of nonmodal damping terms. Note that the damping effects which enter B v may
well be duplicated by the effects included in b i so that, in general, the user should be careful when using
both forms simultaneously.
Note: In frequency response and complex eigenvalue analysis, the complete damping matrix is
similar, except that structural damping is treated as a complex stiffness matrix.
Also, the matrix T B v in Eq. (6-10), or the equivalent complex stiffness matrix, is generally coupled
so that the efficient uncoupled methods of analysis cannot be used when B v is present.
It is the accepted practice in many industries to express viscous damping as a fraction of critical damping.
Critical damping is defined as the value at which the homogenous solution of Eq. (6-8) transitions from
Main Index
CHAPTER 6 265
Damping
a damped sinusoid to a nonoscillating, decaying exponential. A solution for the modal equation shows
the value of critical damping, C c , to be
Cc = 2 ki mi (6-11)
bi bi
i ------ = -----------------
- (6-12)
Cc 2 ki mi
Another form of damping specification is the amplification quality factor, Q i , with the definition
1 - = ---
1-
Q i = ------- (6-13)
2 i gi
All three forms of damping specification are available for modal damping.
An illustration of the comparison between modal damping and an equivalent function of structural
damping is shown in Figure 6-2, showing the different options for modal damping tables. With unit
modal masses, a constant modal damping table actually corresponds to a viscous damping that increases
with modal frequency. This method falls between the two extremes (constant viscous damping and
equivalent structural damping) and serves as a general purpose compromise.
b i
Constant Viscous Damping
b = c
Structural Damping
2
b G i
Modal Damping
b = mi gi i
i
Modal Frequency i
Main Index
266 MD Nastran Dynamic Analysis User’s Guide
Modeling Damping Effects
Note that structural damping will overdamp the system modes above i and underdamp system modes
below i . The time domain solutions require the specification of i with the PARAMs W3 and W4 and
the units are in radians per secon and not Hertz. The frequency domain solutions do not require PARAMs
W3 and W4.
Nonlinear Damping
This discussion explains the limitations in damping caused by the nonlinear transient solution algorithms
(SOL 129). Frequency response and complex eigenvalue solutions are not available in nonlinear
analysis. For specialized nonlinear dynamics topics, see Nonlinear Transient Response Analysis, 384.
For basic information on the MD Nastran nonlinear solutions, see the MSC.Nastran Handbook for
Nonlinear Analysis or the MD Nastran Reference Manual.
Finite elements may be either linear or nonlinear in the nonlinear formulations, and the damping effects
will be treated differently for each type. The basic rules for nonlinear transient analysis are as follows:
1. Plastic yield in the nonlinear materials automatically absorbs energy when the structure follows a
loading and unloading cycle. This is an actual hysteresis effect that produces an accurate form of
damping. However, note that strain rate effects are not calculated directly. Strain rate effects
must be modeled with structural damping parameters, which are converted internally to viscous
damping.
2. Viscous damping elements are always linear and will participate as constant matrix terms.
3. Structural damping, defined with parameters and material bulk data inputs (the GE field on the
MATi entries), is recognized for both linear and nonlinear elements. The damping matrix terms
are calculated for the current material stiffness moduli and geometry. Note, however, that the
tangent matrices are only updated periodically. The actual damping on nonlinear elements is
unpredictable and can change answers for different runs on the same problem—depending on the
convergence rate and iteration strategy. It is recommended that the matrix update strategy forces
an update on the tangent matrix at every time step.
4. A modal formulation (and therefore modal damping) is not available in a nonlinear solution.
5. Superelements may be used to reduce the size of a nonlinear problem by separating the linear
elements into an upstream component. Component modal synthesis may then be used to maintain
accuracy. Structural damping is allowed on superelements.
Main Index
CHAPTER 6 267
Damping
where:
m = mass
b = viscous damping
g = structural damping coefficient
k = stiffness
p = operator notation for d dt
i = –1
u = displacement
Dividing the above equation by m and making some substitutions, the equation becomes:
2 2
p + 2p + 1 + ig u = 0 (6-15)
where:
2
=km
2 = b m
The structural damping case is obtain by setting to zero, and the viscous damping case is obtained by
setting g to zero. However, the above equation can be easily solved without setting either variable to
zero for its characteristic values, i.e., eigenvalues.
2
p = – – 1 + ig (6-16)
Putting the complex number inside the radical into polar format, yields:
4 2 2 2 i
p = – i 1 – + g e (6-17)
Main Index
268 MD Nastran Dynamic Analysis User’s Guide
Damping Effect on Complex Eigenvalues
g
where = tan ---------------
–1
2
1–
The imaginary part of the above eigenvalue determines the damped natural frequency and the real part
represents the damping coefficent. When the damped natural frequency is viewed in this form, it is
relatively straightforward to observe that viscous damping, g, decreases the damped natural frequency
while the structural damping increases the damped natural frequency. Furthermore, the eigenvalue
problem will never calculate real numbers for the eigenvalues when only structural damping is applied.
Putting the eigenvalue back into Chartesian format:
4 2 2 2
p = – 1 – + g sin 2 – i cos 2 (6-18)
Using the above equation one can compute the complex eigenvalues for any value of g , and to
obtain the results shown in Figure 6-3 as the solid lines. The circle and square symbols represent results
obtain with MD Nastran using the input files Install_dir/mdxxxx/doc/dynamics/bd06sd.dat and
Install_dir/mdxxxx/doc/dynamics/bd06vd.dat.
If one looks at the two cases as separate solutions, i. e., either set g = 0 then the first equation becomes:
2
mp + bp + k u = 0 (6-19)
and the roots are complex conjugate pairs. With = 0 , the equation becomes:
Main Index
CHAPTER 6 269
Damping
2
mp + 1 + ig k u = 0 (6-20)
and the complex eigenvalues are the negative of each other. In this case only the complex eigenvalues
with the positive imaginary part is the one of interest. When the real part of this eigenvalue is positive,
then the model is unstable.
By mixing the two types of damping, i.e., set g = 0.02 and vary the viscous damping value, the
damping comparison is made in Figure 6-4. Again the lines represent the theoretical solution and the
solid symbols are results from MD Nastran.
Main Index
270 MD Nastran Dynamic Analysis User’s Guide
Structural Damping for Shell and Bush Elements
Matrices
PSHELL
If PARAM,SHLDAMP,DIFF is present in the run, a structural damping matrix K 4 is formed for each
shell element based on a relationship of the following form:
4
K = ge m id 1 membrane stiffness (6-21)
+ ge m id2 bending stiffness
+ ge m id3 transverse stiffness
+ ge m id4 combined stiffness
CBUSH
If both stiffness K and structural damping ge are specified, each with its own frequency dependent
tables, matrix terms of the following form are created:
f f f
K j + i ge j K j (6-22)
If stiffness K with frequency dependent tables and non frequency dependent structural damping ge are
specified, matrix terms of the following form are created:
f 0 f
K j + i ge j K j (6-23)
If nonfrequency dependent stiffness K with frequency dependent structural damping ge are specified,
matrix terms of the following form are created:
0 f 0
K j + i ge j K j (6-24)
Main Index
CHAPTER 6 271
Damping
In the above three expressions, the superscript f denotes frequency dependent and the superscript 0
denotes a nominal value. The subscript j implies the j-th degree of freedom of the CBUSH element. The
real term goes into the element stiffness matrix and the imaginary term goes into the element K 4 matrix.
User Interface
The parameter SHLDAMPdefaults to the value SAME, which implies use only the MID1 associated
value of structural damping coefficient. A character value other than SAME such as DIFF implies that
each MIDi entry on the PSHELL can have its own associated value for the structural damping
coefficient.
The PBUSH and PBUSHT Bulk Data entries have five structural damping coefficients and their
corresponding frequency dependent tables.
The resulting K 4 matrices will have appropriate terms accounting for additional structural damping
coefficients
Main Index
272 MD Nastran Dynamic Analysis User’s Guide
Geometric and Material Nonlinear Spring and Damper Element
User Interface
The element is defined with connectivity entry CBUSH1D and a property entry PBUSH1D
described below. The user may define several spring or damping values on the PBUSH1D property
entry. It is assumed that springs and dampers work in parallel. The element force is the sum of all
springs and dampers.
CBUSH1D
1 2 3 4 5 6 7 8 9 10
CBUSH1D EID PID GA GB CID
Main Index
CHAPTER 6 273
Damping
PBUSH1D
Defines linear and nonlinear properties of a one-dimensional spring and damper element.
Format
1 2 3 4 5 6 7 8 9 10
PBUSH1D PID K C M SA SE
“SHOCKA” TYPE CVT CVC EXPVT EXPVC IDTS
IDETS IDECS IDETSD IDECSD
“SPRING” TYPE IDT IDC IDTDU IDCDU
“DAMPER" TYPE IDT IDC IDTDV IDCDV
“GENER” IDT IDC IDTDU IDCDU IDTDV IDCDV
The continuation entries are optional. The four options, SHOCKA, SPRING, DAMPER, and GENER
can appear in any order. See PBUSH1D for definitions.
Output
The CBUSH1D element outputs, if requested, axial force, relative axial displacement and relative axial
velocity. Also stress and strain if stress and strain coefficients are defined. All element related output
(forces, displacements, stresses) is requested with the STRESS Case Control command.
Main Index
274 MD Nastran Dynamic Analysis User’s Guide
Geometric and Material Nonlinear Spring and Damper Element
In linear dynamic solution sequences, the linear stiffness and damping is used. In linear dynamic
solution sequences, the CBUSH1D damping forces are not included in the element force output.
In nonlinear solution sequences, the linear stiffness and damping is used for the initial tangent stiffness
and damping. When nonlinear force functions are defined and the stiffness needs to be updated, the
tangents of the force-displacement and force-velocity curves are used for stiffness and damping. The
CBUSH1D element is considered to be nonlinear if a nonlinear force function is defined or if large
deformation is turned on (PARAM, LGDISP, 1). For a nonlinear CBUSH1D element, the element force
output is the sum of the elastic forces and the damping forces. The element is considered to be a linear
element if only a linear stiffness and a linear damping are defined and large deformation is turned off.
1. The CBUSH1D element nonlinear forces are defined with table look ups and equations.
2. The table look ups are all single precision in MD Nastran. In nonlinear, round-off errors may
accumulate due to single precision table look ups.
3. For linear dynamic solution sequences, the damping forces are not included in the element force
output.
4. The “LOG” option on the TABLED1 is not supported with the CBUSH1D.
Example
Swinging Pendulum
The solution to a swinging pendulum is calculated using SOL 129. The pendulum has a length of
1.0 [m]. and has a concentrated mass of M = 1,000.0 [kg] at the free end, see Figure 6-5. We start
the analysis with the pendulum at rest in horizontal position. The free end is loaded with a gravity
induced load of G = 10,000.0 [N]. The leg of the pendulum is very stiff, K = 1.e+7 [N/m], so that
the relative axial deformation is small compared to the overall motion of the pendulum. Large
deformation effects are turned on with PARAM, LGDISP, 1. The MD Nastran input file is partially
shown in Listing 6-1, see Install_dir/mdxxxx/doc/dynamics/ar29src1.dat for the complete file.
Main Index
CHAPTER 6 275
Damping
Main Index
276 MD Nastran Dynamic Analysis User’s Guide
Geometric and Material Nonlinear Spring and Damper Element
We investigate two different models of the pendulum in one run. In the first model, we use a Rod
element for the stiffness of the leg, K = 1.e+7 [N/m], and a Visc element for 5% equivalent viscous
damping, C = 1,000.0 [Ns/m]. The x- and z-displacements of the free end are shown in Figure 6-6.
The Rod rotates while the viscous damper Visc stays fixed in its initial horizontal position. The Visc
element is linear and does not follow large deformations. The displacements are damped out
because the viscous damper is acting in the x-direction during the whole motion.
Main Index
CHAPTER 6 277
Damping
In the second model, we use a CBUSH1D element. The element has a linear stiffness (K) and a
viscous damper (C) with the same values as in the first model. In the CBUSH1D element, the spring
and damper rotate. The relative axial displacements and velocities are small because of the high
axial stiffness of the element. The rotating damper has no noticeable effect on the overall motion
of the pendulum because it damps only the small relative deformations. The x- and z-
displacements of the free end are not damped, see Figure 6-6.
The example demonstrates how the answers can change if dampers rotate with the deformation
compared to dampers which stay fixed in space.
Simple Model of an Automotive Suspension
A simplified model of an automotive suspension is shown in Figure 6-7. The suspension is modeled
with a stiff beam which has a pinned support at the left end and a shock absorber support in the
middle. The shock absorber is modeled with a CBUSH1D element. The SHOCKA option on the
PBUSH1D property entry is used to model nonlinear damping in tension and compression. A
sinusoidal pulse load is applied at the tip of the overhang. The MD Nastran input file is partially
shown in Listing 6-2 see Install_dir/mdxxxx/doc/dynamics/ar29be01.dat for the complete file. The
beam responds with a rotation of about 45 0 before the motion is damped out. The load and the z-
displacement at the tip of the overhang are shown in Figure 3-4. The example demonstrates that
the CBUSH1D element can undergo large deformation and can simulate complex nonlinear force
deflection laws.
Main Index
278 MD Nastran Dynamic Analysis User’s Guide
Geometric and Material Nonlinear Spring and Damper Element
TIME 1000
SOL 129
$* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
CEND
$
TITLE = Shock absorber and spring loaded with sine pulse
LABEL = CBUSH1D with SHOCKA option and SPRING option
. . . $
$* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
BEGIN BULK
$
PARAM, POST, –1
PARAM, AUTO SPC, YES
PARAM,MAXRATIO, 1 .E+8
Main Index
CHAPTER 6 279
Damping
PARAM LGDISP 1
PARAM LANGLE 2
$
$ CBUSH1D element (SHOCKA option with TABLE for S(u)
$and TABLE for SPRING constant)
$
CBUSH1D, 100, 101, 8, 19
$
$ CVT = 0.2damping for tension
$ CVC = 0.4damping for compression
$ S(u) = 1.0constant scale factor
$ EVT = 1.0exponent of velocity for tension
$ EVC = 0.5exponent of velocity for compression
$
PBUSH1D, 101,10.0, 0.4,,,,, ,+PB1
SHOCKA, TABLE, 0.2, 0.4, 1.0, 0.5, 999, , +PB2
SPRING, TABLE, 998
$
TABLED1,999,,,,,,,, +TB999
+TB999, –1000., 1.0, 0.0, 1.0, 1000., 1.0, ENDT
$
TABLED1,998,,,,,,,, +TB998
+TB998, –1000., –10000., 0.0, 0.0, 1000., 10000., ENDT
$ . . . $ ENDDATA
Main Index
280 MD Nastran Dynamic Analysis User’s Guide
Geometric and Material Nonlinear Spring and Damper Element
Time [sec]
Figure 6-8 Load and Displacement at Tip C for the Automotive Suspension Model
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 7: Enforced Motion
7 Enforced Motion
Overview
Direct Enforced Motion in Dynamic Analysis
Enforced Motion with Loads
Large Mass Method - Direct
Large Mass Method - Modal
User Interface for the Large Mass Method
Examples
Main Index
282 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
Enforced motion specifies the displacements, velocities, and/or accelerations at a set of grid points for
frequency and transient response. Enforced motion is used when motion is specified instead of or in
conjunction with applied loads. A common application is an earthquake excitation applied to a building.
In this case there are no applied loads, instead the base of the building undergoes an enforced
displacement or acceleration time history.
The direct method for enforced motion permits direct specification of displacements, velocities, or
accelerations via the SPC and SPCD data entries. This can eliminate the need for large masses.
Other methods for prescribing enforced motion in dynamics use the procedures described in Frequency
Response Analysis, 133 and Transient Response Analysis, 201 for specifying applied forces in
conjunction with techniques that convert applied forces into enforced motion. One method that is
applicable to both transient response and frequency response is the large mass method. Another method
is the inertial loads method. Both of these methods often lead to computational and numerical problems
due to round-off errors and pseudo rigid body modes. There is also the Lagrange Multiplier technique,
but it is very manual and requires DMAP.
Main Index
CHAPTER 7 283
Enforced Motion
Main Index
284 MD Nastran Dynamic Analysis User’s Guide
Direct Enforced Motion in Dynamic Analysis
If instead of an harmonic force one were to apply an enforced motion, the effect would be similar, that
is, a response with proportional forces of constraint, at the same frequency as that of the enforcing
motion. Furthermore, any one of the enforced displacement, velocity, or acceleration must uniquely
determine the other two (because they differ only by multiples of frequency), with resultant forces of
constraint derived from a solution of the governing equations.
To illustrate this, an applied harmonic forcing function of the form:
it
Pt = P e
will necessarily give rise to the in-phase displacement
it
ut = U e
with the corresponding velocity and acceleration
i t
u· t = i U e
and
2 i t
u·· t = – U e
yielding the familiar equations of frequency response. Here they are written after multipoint constraint
partitioning operations have been performed leaving just the free (f-set) and constrained (s-set) degrees-
of-freedom
– 2 M ff M f s + i B ff B fs + K f f K f s U f = Pf
M sf M ss B s f B ss K sf K ss U s Ps + qs
where P s are the external loads applied to the s-set, and q s are the corresponding forces of constraint.
If the constraints specify zero motion U s = 0 , the solution for the free degrees-of-freedom may be
obtained directly from the upper part of the above equation as:
2
– M ff + iB f f + K f f U f = P f (7-1)
and the corresponding constraint forces from the lower part as:
2
q s = – P s + – M sf + iB sf + K sf U f (7-2)
If enforced displacements, velocities, or accelerations are applied, U s 0 and the free degrees-of-
freedom, again from are:
2 2
– M ff + iB f f + K f f U f = P f – – M fs + iB f s + K fs U s (7-3)
Main Index
CHAPTER 7 285
Enforced Motion
Comparing Eqs. (7-3) and (7-4) with Eqs. (7-1) and (7-2) shows that the enforced motion modifies the
forces applied to the f-set degrees-of-freedom. Note also that, were enforced velocities to be applied
instead, the resultant applied displacements would differ by a factor of 1 i , and by – 1 2 in the
case of enforced accelerations.
·· ·
M ff M fs u f Bf f Bf s uf Kf f Kf s uf Pf t
·· + · + = (7-5)
M s f M ss u s B sf B ss u s K sf K ss us Ps t + qs t
In the case of zero constrained motion, u s = u· s = u·· s = 0 and the solution for the free degrees-of-
freedom is available directly from Eq. (7-5) as:
M f f u·· f + B ff u· f + K f f u f = P f t (7-6)
q s t = – P s t + M sf u·· f + B sf u· f + K s f u f (7-7)
In the case of enforced motion, u s , u· s , u·· s 0 and the solution for the free degrees-of-freedom, again
from Eq. (7-5) is then:
Main Index
286 MD Nastran Dynamic Analysis User’s Guide
Direct Enforced Motion in Dynamic Analysis
·· ·
uf uf uf
q s t = – P s t + M sf M ss + B sf B ss + K sf K ss (7-9)
·· ·
us us us
As with frequency response, note that the effect of enforced motion is to modify the loads on the f-set,
and the s-set forces of constraint.
The character fields may be shortened to as little as a single character on input, if desired.
In order to allow maximum input flexibility and to support existing input data files which had used the
large mass method, the decision regarding whether to use the older, large mass method, or the newer
direct enforced motion will be made based on a search for SPC/SPC1/SPCD data as follows: If the TYPE
field on the TLOAD1/2 entries indicates an enforced motion, the EXCITEID field will first be assumed
to point to SPC-type data. If not present, MD Nastran will then assume the excitation is enforced motion
using large masses and will then look for DAREA and various static loading data, just as in the case of
applied loads. RLOAD1 and RLOAD2 entries will only look for SPC-type data in the case of enforced
Main Index
CHAPTER 7 287
Enforced Motion
motion. For all entries (TLOAD1/2, RLOAD1/2), if the TYPE field indicates an applied load, the
program will search only for static loading data.
With direct enforced motion, data is supplied via the SPC/SPC1/SPCD Bulk Data entries. (Note that
components specified in SPCD data must be also be referenced on SPC or SPC1 entries.) Enforced
motion could be defined solely with SPC data (without the use of SPCD’s), but such usage is discouraged
in favor of SPC/SPC1 data in conjunction with SPCD’s. In this way, SPCD entries are selected via the
EXCITEID field in the dynamic load data, while the SPC/SPC1 sets are used to identify the constrained
displacement set in Case Control. With SPC data alone, enforced motion is applied solely via Case
Control. The entries referenced by the EXCITEID field on the TLOAD1/2 or RLOAD1/2 entries do not
have to exist. This method of specifying enforced motion is less elegant and somewhat more difficult to
interpret.
Diagnostic Messages
Some of the diagnostic messages resulting from a dynamic response analysis execution:
• A user information message indicates the types of excitation specified (applied loads, enforced
displacement, enforced velocity, or enforced acceleration using either SPC-type data or large
masses, or a combination thereof).
• A user warning message is issued if any of the individual dynamic loading data results in a null
loading condition.
• The execution is terminated with a user fatal error if the total excitation is null in a frequency
response analysis, thereby implying a null solution. Such a fatal error has always occurred in
transient response analysis.
Main Index
288 MD Nastran Dynamic Analysis User’s Guide
Direct Enforced Motion in Dynamic Analysis
2. Defines the desired enforced motion using the SPCD Bulk Data entry. The set IDs of these SPCD
data must match the IDs appearing in the EXCITEID fields of the corresponding dynamic load
data in step 1.
3. Ensures that the components referenced in the SPCD Bulk Data entry above are also specified in
the SPC/SPC1 Bulk Data and this SPC/SPC1 data is also selected in Case Control Section.
4. Ensures PARAM,RESVEC is set to YES in the Bulk Data Section if a modal dynamic solution
approach is used.
The following simple examples illustrate these steps for a frequency and transient response analysis,
respectively.
Main Index
CHAPTER 7 289
Enforced Motion
.
.
.
BEGIN BULK
$ THE FOLLOWING SPC1 ENTRY IS SELECTED
$ VIA CASE CONTROL
SPC1,1000,3,75,85
$ THE TYPE FIELD IN THE FOLLOWING TLOAD1 ENTRY
$ SPECIFIES ENFORCED ACCELERATION. ACCORDINGLY,
$ EXCITIED ID 300 HERE REFERS TO SPCD DATA.
$TLOAD1,SID,EXCITIED,DELAY,TYPE,TID
TLOAD1,100,300, ,ACCE, 500
$ THE FOLLOWING SPCD ENTRY IS SELECTED BY THE
$ TLOAD1 DATA ABOVE.
$ NOTE THAT GRID POINT 85 - COMPONENT 3 SPECIFIED
$ HERE ALSO APPEARS IN THE SELECTED SPC1 DATA.
$ NOTE ALSO THAT THE VALUE OF 2.5 SPECIFIED HERE
$ REPRESENTS ACCELERATION
SPCD,300,85,3,2.5
.
.
.
Example
The following example illustrates the application of enforced motion in frequency response. See
Install_dir/mdxxxx/doc/dynamics/bd07enf.dat. The tower (Figure 7-1) is subjected to a 1G (9.81m/sec2)
enforced acceleration applied to the base of the tower. An extra grid point ties the four bottom grid points
together using an RBE2 element.
The selected portion of the input file is shown in Listing 7-1 and illustrates how this enforced motion is
applied. In this case, the ACCE field (field 8) on the RLOAD2,101 entry indicates that the enforced
motion type is acceleration. Field 3 on the RLOAD2 entry points to the SPCD,101 entry with an
enforced motion of 9.81 m/sec2 in the x-direction of grid point 1. The degree-of-freedom with enforced
motion specified must also be constrained, once again x-direction of grid point 1. This is accomplished
by the SPC1,30 entry. Notice that the SPCD/SPC1 combination is similar to the static enforced
displacement procedure.
Main Index
290 MD Nastran Dynamic Analysis User’s Guide
Direct Enforced Motion in Dynamic Analysis
547
1
Figure 7-1 Tower for Enforced Motion.
Main Index
CHAPTER 7 291
Enforced Motion
EIGRL 1 10 0 MASS
$
TABDMP1 2 CRIT +
+ 0. 0.05 1. 0.05ENDT
$
TABLED2 1 0. +
+ 0. 1. 1. 1.ENDT
$
RLOAD2,101,101,,,1,,ACCE
$
SPC1,30,123456,1
$
SPCD,101,1,1,9.81
freq4 3 10. 35. .05 5
freq1 3 10. 1. 25
$
RBE2 82 1 123 22 43 392 101
$
......rest of input
ENDDATA
A 5% critical damping ratio and a frequency combination of FREQ1 and FREQ4 are used for this
analysis. As with all modal approaches, the residual vector (PARAM,RESVEC,YES) improves the
quality of the results. A partial output is shown in Listing 7-2 where grid point 1 is the enforced motion
point and grid point 547 is at the top of the tower. Note that the acceleration at grid point 1 in the x-
direction is 9.81 m/sec2, which is what one would expect, since that is the actual enforced acceleration
value.
FREQUENCY = 1.000000E+01
C O M P L E X A C C E L E R A T I O N
V E C T O R
(MAGNITUDE/PHA
SE)
Main Index
292 MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
Main Index
CHAPTER 7 293
Enforced Motion
Structure
Seismic Mass
M0
B K Filter
3. In nonlinear transient solutions, large springs can affect the error tests and convergence logic.
The internally calculated error ratios are dimensionless numbers obtained by dividing the errors
by an average force or total energy. The forces and energy created by the large mass/spring
approach will dominate these values, resulting in underestimates of the errors and false
convergence. The solution is to decrease the error allowables on the TSTEPNL Bulk Data inputs.
4. Using these methods in a modal formulation requires some attention. For output of total
displacements, the user should retain the zero or low frequency modes that the large masses
produce (i.e., set the parameter LFREQ = – .01). Note that if the low frequency modes are
dropped from the dynamic solution, the output will be the correct relative motion.
Large springs for enforced displacements are not recommended for the modal formulations.
They will generate high frequency modes that are usually missing from the system. The resulting
dynamic solution is not valid since the large springs are not included in the modal stiffness matrix.
5. Numerical conditioning of the matrix solution may be affected by the method used to connect the
large mass or spring. Numerical roundoff of the results may occur. MPCs, RBEi’s, and ASET
operations all use a matrix elimination procedure that may couple many degrees-of-freedom. If a
large mass or spring is not retained in the solution set, its matrix coefficient will be distributed to
other solution points. Then, matrix conditioning for decomposition operations becomes worse,
when the large terms dominate the significant finite element coupling terms. On the other hand,
if the degrees-of-freedom with the large terms remain in the solution set, they remain on the
diagonal of the matrix and the matrix decomposition is unaffected.
In summary, the large mass method is recommended for cases with known accelerations at a single point.
It works well with the modal formulation, providing good stress and forces near the mass, and is easy to
Main Index
294 MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
understand and use. In many cases, the structure is actually excited by the motions of a large, massive,
base (for instance, the geological strata) which can actually be used as a value for the mass.
The large spring method is recommended for cases in which displacements are known at one point and
a direct formulation is used. The primary advantage is its simplicity. However, the inertial loads
approach or Lagrange Multiplier technique, described below, are more general and have better reliability.
Approach
The basic approach is shown in the sketch below:
u
Structure
u
u0 Base
We assume that the motions of the rigid base are known and the structure is fixed only to the base. Define
the displacement vector, u , as a sum of the known base motion, u 0 and the relative displacements, u
by the equation
u = u + D u 0 (7-10)
where D is the rigid body transformation matrix that includes the effects of coordinate systems,
offsets, and multiple directions. If the structure is a free body and aerodynamic effects are neglected, the
base motions should not cause any static or viscous forces and
KD = 0
(7-11)
and B D = 0
Main Index
CHAPTER 7 295
Enforced Motion
where K is the stiffness matrix and B is the damping matrix. These properties allow us to remove
the reference motions from the solution as shown:
The equilibrium equation for the whole system is
M u·· + B u· + K u = P (7-12)
Substituting Eqs. (7-10) and (7-11) into Eq. (7-12), we obtain
The solution matrices have not changed, but the solution, u , must be zero at the base attachment points.
The right-hand term in Eq. (7-13) may be calculated in MD Nastran using one of several options. One
option is the response spectra analysis described in Shock and Response Spectrum Analysis, 260.
Another is a recent MD Nastran implementation that now allows time-dependent gravity loads. By
definition, the GRAV input generates load vectors of the form
Pg = M D g (7-14)
where g is the gravity vector. Then if g were time dependent, it could be used to replace u 0 in Eq. (7-13)
if
g t = – u·· 0 t (7-15)
Input Data
The inertial loads method is available for all dynamic response solutions. The method is valid for both
direct and modal formulations and is automatic in nonlinear and superelement models. Briefly, the
necessary input data is as follows:
1. GRAV -- Bulk Data defining an acceleration vector. Several may be included to define different
functions of time for the different directions.
2. SPC -- remember to fix the base attachments.
3. TLOADi or RLOADi Bulk Data for the dynamic load definition. Reference to the EXCITEID to
the GRAV SIDs.
Other necessary inputs are the conventional DLOAD and TABLEDi data to help define the time
functions.
Recommendations
The following comments describe the benefits and drawbacks to the method:
1. The solution will avoid the problem of rigid body drift that occurs in the other methods when
small errors are present in the input accelerations.
2. By eliminating large constant terms, the stress and force calculations may be numerically more
precise.
Main Index
296 MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
u1
force P 2 applied
m2 u2
Main Index
CHAPTER 7 297
Enforced Motion
u1 = Y1 (7-17)
q contains the unknown constraint forces required to cause the enforced motion and react to the
applied force. Y 1 is the known enforced displacement, and P 2 is the known applied force.
In the LMT method, the constraint equations are included in the system equations. The displacement
constraint can be written in matrix form after q 1 is moved to the left side,
–q1
0 1 0 u1 = Y1 (7-18)
u2
This row and a corresponding column are added to the system matrix to produce
0 1 0 Y1 –q1
K = 1 k –k , P = 0 , u = u1 (7-19)
0 –k k P2 u2
In the equation K u = P , the known vector quantities are now all on the right side, and the
unknown vector quantities on the left side. This equation can be solved directly without the need for
partitioning or constraint elimination. In dynamic analysis, enforced motion can be input on velocity and
acceleration variables directly by similar techniques. The system has a unique solution even though
there are null terms on the diagonal of K . These null terms are an indication of an indefinite matrix.
k – m 1 –k u1
= 0 (7-20)
–k k – m 2 u 2
When the constraint reduction process is used, the first row and column are eliminated. The resulting
equation for the eigenvalue is then
= k m2 (7-21)
Alternatively, using the LMT process, the system is augmented with the constraint variables, with the
resulting determinant
0 1 0
1 k – m 1 –k = 0 (7-22)
0 –k k – m 2
Main Index
298 MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
When this determinant is evaluated, the only non-null product results from the (12)(21)(33) terms,
resulting in the same expression, = k m2 .
–r k
= c1 = 0 (7-23)
c2 1.0
Where c is the constrained eigenvector for the displacements and r is the reaction force. Note the scale
factor of the whole eigenvector may be affected by r if the normalization method is MAX. For a
dynamic solution using a modal formulation, the matrices are transformed using the modes above. In the
stiffness, mass, and damping matrices, the constraint coefficients have no effect. However, if Y 1 is an
enforced displacement, the generalized modal forces are
P = c T P g – r T Y 1 = P 2 – k Y 1 (7-24)
The result of applying an enforced displacement on the LMT coordinates, Y 1 , will be a modal force in
the opposite of the direction of motion. The results of a modal solution will be relative to Y 1 . See below
for a solution that provides total displacements.
The notation of the conventional solutions is described in Constraint and Set Notation (p. 347) in the MD
Nastran Reference Manual. It is maintained with the following changes. The Lagrange Multiplier
Technique requires the definition of enforced motion variables. Their DOFs are placed in the u 1 , u 2 , and
u 3 sets. These sets are members of the USET table but are not used in the conventional solution method.
They are defined on USETi Bulk Data entries. They are regarded as members of the b-set, and have the
following meanings:
New Unique Sets for Enforced Motion
Main Index
CHAPTER 7 299
Enforced Motion
Almost all of the letters of the alphabet are already in use to describe sets in the conventional solutions.
Therefore, sets with integers for names are added to the present set names which use letters of the
alphabet instead.
Direct Solutions in the Frequency and Time Domains
Ad d ud = Pd (7-25)
where:
A d d = K + iB – 2 M d d (7-26)
K , B , and M are the stiffness, damping, and mass matrices, respectively, and is the excitation
frequency. If static or dynamic reduction is not used, most of the constraints are placed in the stiffness
matrix. The advantage of this placement is that the constraints are not multiplied by . This allows the
solution of the equations at an excitation frequency near to zero. The remaining constraints are
associated with enforced motion variables. These constraints are placed in the stiffness, damping, or
mass matrix depending on whether the enforced motion is the displacement, velocity, or acceleration
type, respectively. If static or dynamic reduction is performed, all m- and s-set constraints are eliminated
along with the o-set. The u 3 -set is then the only constraint set in the d-set.
The DOFs listed on USET entries are input points for enforced motion. Data conventionally used to
define dynamic loads applied to these points is interpreted to be time- or frequency-varying enforced
motions. The value of the SETNAME entry on the USET entries determines the type of input. A value
of u 1 indicates an enforced displacement type. A value of u 2 indicates an enforced velocity type, and a
value of u 3 indicates an enforced acceleration type. These dynamic quasi-loads are transferred to the
enforced motion variables by the transformation matrix T 4p as follows:
P4 = T4 p Pp (7-27)
Pp is the load vector output by the FRLG module. In this context it may contain both applied forces
and enforced motions. T 4p has a unit term in each row at the column corresponding to the u i -set point
where enforced motion is applied. The enforced motions are removed from the displacement variables
by the transformation matrix T pp
x ,
Main Index
300 MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
x P
P p = T pp (7-28)
p
where T pp
x is an identity matrix except for the rows associated with enforced motion terms, where it has
a zero row. The p -set and u 4 -set components are merged to form P p .
q3
ud = u3 (7-29)
uc
and
Y3
Pd = P3 (7-30)
Pc
The 3-set contains the constraint force and displacement variables associated with enforced motion. q 3
is the vector of constraint forces and u 3 the vector of constraint displacements. Both are partitions of
the u d vector and u c is their complement. Y 3 is the vector of enforced displacements, P 3
additional forces placed on the enforced motion points, and P c their complement. The enforced
motions of any LMT variables contained in P c are zero by definition. u c contains both displacement
and constraint force unknown variables, while P c contains both known input displacements and
applied forces.
0 R 33 R3 c
K = K 33 K3 c (7-31)
Sym. K cc
Main Index
CHAPTER 7 301
Enforced Motion
0 I 33 0 q3 Y3
M d d u·· d + B d d u· d + I 33 K 33 K 3 c u 3 = P 3 (7-32)
0 K c3 K cc u c Pc
For this case M dd and B dd have no terms for the u 3 -set variables. Their partitions are not shown.
Evaluating the first row of this equation, we obtain
u3 = Y3 (7-33)
Enforced Accelerations
Now consider the case of enforced acceleration. The equation of motion is then
·· ··
0 I 33 q3
0 Y3
I 33 M 33 M 3 c u·· 3 + B d d u· d + K d d u d = P 3 (7-34)
0 M c 3 M cc u·· c Pc
u·· 3 = Y·· 3
The input quantity Y·· 3 is an acceleration vector because the first partition of P d is equated to an
acceleration vector. That is, the type of the matrix where the constraint equations are placed determines
the type of motion of Y 3 . The enforced motion variables may be placed in any or all of the matrices.
Mixed types of enforced motion are therefore supported.
Note that in this case, q·· 3 is the force of constraint.
K a a – 2 M a2a a z = 0 (7-35)
Main Index
302 MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
Each static constraint mode is computed by fixing all constraint DOFs except one and applying a unit
motion to it. The motion of the physical variables is the superposition of the motion of the flexible modes
and the constraint modes.
Let u d = T dh u h , where u d is defined in Eq. (7-29). T dh is the transformation between the modal
and physical dynamic variables. u h , the unknown variable vector for modal analysis, has the
components
uz
q3 = uh (7-36)
u3
is the vector of generalized displacements for the flexible modes. q 3 and u 3 are the unknown
uz
constraint force variables and displacement variables for the constraint modes. The corresponding
known variables are
Pz
Ph = Y3 (7-37)
P3
is the modal force vector and Y 3 and P 3 the known enforced motion variables and applied
Pz
forces of the enforced motion points.
Ph is obtained from the equation
Ph = Td h T Pd (7-38)
The equation relating u d and u h is used to replace u d in Eq. (7-25). The resulting equation is
pre-multiplied by T dh
T to form the reduced basis equations of motion as follows:
T d h T Z d d T dh u h = T d h T P d = P h (7-39)
When the multiplication of Eq. (7-39) is carried out, the components of the modal force vector expressed
in the partitions of Eq. (7-37) become:
P z = q 3 z T Y 3 + 3 z T P 3 + cz T P c (7-40)
u3 = u3 = Y3 (7-41)
P 3 = q 33 T Y 3 + P 3 + u c3 T P c (7-42)
Main Index
CHAPTER 7 303
Enforced Motion
The modal forces P z have components from all three partitions of P d . u 3 is identically u 3 .
P3 includes loads applied directly to 3-set points P 3 , as well as the constraint forces needed to
maintain the motion of the unknown input point variables u 3 at the prescribed input motion Y 3 . (In
this implementation the P 3 forces are moved to Y 3 .)
If there are enforced motion terms, the rows and columns of the constraint terms, u 3 and q 3 , are
moved to the corresponding stiffness, damping or mass matrices. The load vector P d has enforced
motion inputs transferred to the LMT variables by the same technique used in the direct solutions. Note
that the presence of enforced motion variables results in nondiagonal h-set matrices, requires the use of
the coupled solution algorithms for frequency response and transient analysis. The matrices are sparse,
however, with off-diagonal terms occurring only in the columns associated with enforced motion so that
the effect on solution cost is small.
Main Index
304 MD Nastran Dynamic Analysis User’s Guide
Large Mass Method - Direct
1
u·· = ------- p (7-43)
mo
In other words, the load that produces a desired acceleration u·· is approximately
p = m o u·· (7-44)
The accuracy of this approximation increases as m o is made larger in comparison to the mass of the
structure. The only limit for the size of m o is numeric overflow in the computer. MSC recommends that
the value of m o be approximately 10 6 times the mass of the entire structure for an enforced translational
degree-of-freedom and 10 6 times the mass moment of inertia of the entire structure for a rotational DOF.
The factor 10 6 is a safe limit that should produce approximately six digits of numerical accuracy.
The large mass method is implemented in direct transient and frequency response analysis by placing
large masses m o on all enforced degrees-of-freedom and supplying applied dynamic loads specified by
Eq. (7-44); that is, the function u·· is input on entries normally used for the input of loads, and the scale
factor m o can be input on DAREA or DLOAD Bulk Data entries, whichever is more convenient.
CMASSi or CONMi entries should be used to input the large masses.
Equation (7-44) is not directly helpful if enforced displacement or enforced velocity is specified rather
than enforced acceleration. However, Eq. (7-44) can be made serviceable in frequency response analysis
by noting that
2
u·· = iu· = – u (7-45)
so that
2
p = m o u·· = im o u· = – m o u (7-46)
The added factor ( i or – 2 ) can be carried by the function tabulated on the TABLEDi entry used to
specify the frequency dependence of the dynamic load.
In the case of transient analysis, provision is made on the TLOAD1 and TLOAD2 entries for you to
indicate whether an enforced displacement, velocity, or acceleration is supplied (TYPE = 1, 2, or 3).
MD Nastran then automatically differentiates a specified velocity once or a specified displacement twice
to obtain an acceleration. The remaining required user actions are the same as for enforced acceleration.
In summary, the user actions for direct frequency and direct transient response are
• Remove any constraints from the enforced degrees-of-freedom.
Main Index
CHAPTER 7 305
Enforced Motion
• Apply large masses m o with CMASSi or CONMi Bulk Data entries to the DOFs where the
motion is enforced. The magnitude of m o should be approximately 10 6 times the entire mass of
the structure (or approximately 10 6 times the entire mass moment of inertia of the structure if the
component of enforced motion is a rotation).
• In the case of direct frequency response, apply a dynamic load computed according to Eq. (7-46)
to each enforced degree-of-freedom.
• In the case of direct transient response,
• Indicate in field 5 of the TLOAD1 and TLOAD2 entries whether the enforced motion is a
displacement, velocity, or acceleration.
• Apply a dynamic load to each enforced degree-of-freedom equal to m o u , m o u· , or m o u·· ,
depending on whether the enforced motion is a displacement, velocity, or acceleration.
Be careful when using PARAM,WTMASS. The WTMASS parameter multiplies the large mass value,
which changes the effective enforced acceleration to
p
u·· = ------------------------------------
- (7-47)
m o WTMASS
p = ko u (7-48)
where k o is the stiffness of the stiff spring and u is the enforced displacement. The large stiffness method
certainly works, but the large mass method is preferred because it is easier to estimate a good value for
the large mass than to estimate a good value for the stiff spring. In addition and more importantly, the
large mass method is far superior when modal methods are used. If very stiff springs are used for modal
analysis rather than very large masses, the vibration modes corresponding to the very stiff springs have
very high frequencies and in all likelihood, are not included among the modes used in the response
analysis. This is the main reason that large masses should be used instead of stiff springs.
The stiff spring method is advantageous in the case of enforced displacement because it avoids the
roundoff error that occurs while differentiating the displacement to obtain acceleration in the large mass
method. The stiff spring method also avoids the problem of rigid-body drift when applying enforced
motion on statically determinate support points. (Rigid-body drift means that the displacement increases
continuously with time, which is often caused by the accumulation of small numerical errors when
integrating the equations of motion.)
Main Index
306 MD Nastran Dynamic Analysis User’s Guide
Large Mass Method - Modal
The rigid-body mode(s) can be removed from consideration either by not computing them or by using
PARAM,LFREQ,r where r is a small positive number (0.001 Hz, for example). If this is done, the
displacements, velocities, and accelerations obtained are relative to the overall motion of the structure
and are not absolute response quantities. Stresses and element forces are the same as when the rigid-body
modes are included because the rigid-body modes do not contribute to them. Rigid-body modes can be
discarded to remove rigid-body drift.
Rigid-body modes occur when the structure is unconstrained and large masses are applied at the DOFs
which, if constrained, result in a statically determinate structure. Redundant constrained DOFs, which
result in a statically indeterminate structure, present a different situation when the constraints are
removed and large masses are applied at those redundant DOFs. In that case, very low-frequency modes
occur but they are not all rigid-body modes; some are modes that represent the motion of one large mass
relative to the others. These very low-frequency, nonrigid-body modes do contribute to element forces
and stresses, and they must be retained in the solution. In some cases, their frequencies are not
necessarily small; they may be only an order or two in magnitude less than the frequency of the first
flexible mode. If PARAM,LFREQ,r is used to remove the true rigid-body modes, then r must be set
below the frequency of the first relative motion mode.
Consider the two-dimensional clamped bar in Figure 7-2. Assume that each end of the bar is subjected
to the same enforced acceleration time history in the y-direction. One way to model the bar is to use two
large masses (one at each end), which are unconstrained in the y-direction. This model provides two very
low-frequency modes: one that is a rigid-body mode and one that is not. The second mode contributes
to element forces and stresses, and removing its contribution leads to an error because with two such large
masses, one mass can drift over time relative to the other mass. A better way to model the case of
identical inputs at multiple locations is to use one large mass connected to the end points by an RBE2
element. This model provides only one rigid-body mode, which can be safely discarded if only the
answers relative to the structure are desired.
Main Index
CHAPTER 7 307
Enforced Motion
u·· t u·· t
x
Basic Model
pt pt
RB E2
E2 RB
pt
Main Index
308 MD Nastran Dynamic Analysis User’s Guide
User Interface for the Large Mass Method
1 2 3 4 5 6 7 8 9 10
DLOAD SID S S1 L1 S2 L2 -etc.-
DAREA Format:
DAREA SID P1 C1 A1 P2 C2 A2
TABLED1 Format:
TABLED1 TID
x1 y1 x2 y2 x3 y3 -etc.- ENDT
TABLED2 Format:
TABLED2 TID X1
x1 y1 x2 y2 x3 y3 -etc.-
TABLED3 Format:
TABLED3 TID X1 X2
x1 y1 x2 y2 x3 y3 -etc.-
TABLED4 Format:
TABLED4 TID X1 X2 X3 X4
A0 A1 A2 A3 A4 A5 -etc.-
The TABLED4 entry defines a power series and is convenient in frequency response for enforced
constant velocity or displacement.
Main Index
CHAPTER 7 309
Enforced Motion
Frequency Response
If a DLOAD entry is used to scale RLOAD1 input, the applied force magnitude in terms of MD Nastran
input is
Pi f = S Si Ai Ci f + i Di f (7-49)
i
where S and Si are input on the DLOAD Bulk Data entry, A i is input on the DAREA entry, and C i f
and D i f are input on the TABLEDi entries. Note that the “i” non-subscript term in the expression
i Di f is – 1 .
If a DLOAD entry is used to scale RLOAD2 input, the applied force magnitude in terms of MD Nastran
input is
Pi f = S Si Ai Bi f (7-50)
i
where S and S i are input on the DLOAD Bulk Data entry, A i is input on the DAREA entry, and B i f is
input on the TABLEDi entry.
Specification of the large force value depends upon whether acceleration, velocity, or displacement is
enforced.
Enforced Acceleration
Enforced acceleration is the easiest to apply since the required force is directly proportional to the desired
acceleration times the large mass:
p = m o u·· (7-51)
Enforced Velocity
Enforced velocity requires a conversion factor
For constant velocity u· , it may be easiest to use the RLOAD1 and TABLED4 entries because the
imaginary term i D f of Eq. (7-49) and the frequency-dependent term 2f can be specified directly.
Enforced Displacement
Enforced displacement also requires a conversion factor
2 2
p = – m o u = – 2f m o u (7-53)
Main Index
310 MD Nastran Dynamic Analysis User’s Guide
User Interface for the Large Mass Method
For constant displacement u , it may be easiest to use the TABLED4 entry because the
frequency-dependent term – 2f 2 can be specified directly.
Transient Response
For transient response, the type of enforced motion (displacement, velocity, or acceleration) is specified
with the TYPE field (field 5) on the TLOAD1 and TLOAD2 Bulk Data entries.
TLOAD1 Format
1 2 3 4 5 6 7 8 9 10
TLOAD1 SID DAREA DELAY TYPE TID
TLOAD2 Format
MSC.Nastran converts enforced displacements and velocities into accelerations by differentiating once
for velocity and twice for displacement. Note that for enforced acceleration, you can specify either force
(TYPE = 0 or blank) or acceleration (TYPE = 3); they are the same for the large mass method.
You still need to use the large mass when specifying any type of enforced motion in transient response
analysis.
Main Index
CHAPTER 7 311
Enforced Motion
Examples
This section provides several examples showing the input and output. These examples are:
Two-DOF Model
Consider the two-DOF model first introduced in Real Eigenvalue Analysis, 43 and shown below in
Figure 7-3. For this example, apply a constant magnitude base acceleration of 1.0 m/sec 2 over the
frequency range of 2 to 10 Hz and run modal frequency response with 5% critical damping in all modes.
The acceleration input is applied to the large mass (grid point 3). A partial input file for this model is
shown in Listing 7-3. See Install_dir/mdxxxx/doc/dynamics/bd07two.dat.
Grid Point 3
Large 7
Mass m 3 = 10 kg
x
Main Index
312 MD Nastran Dynamic Analysis User’s Guide
Examples
Main Index
CHAPTER 7 313
Enforced Motion
$+TAB1 F1 G1 F2 G2 ETC
TABDMP1 777 CRIT +TABD7
+TABD7 0. 0.05 100. 0.05 ENDT
$
... basic model ...
$
ENDDATA
The large mass value is chosen as 1.0E7 kilograms and is input via the CONM2 entry. The scale factor
for the load (1.0E7) is input on the DLOAD Bulk Data entry. The factor of 1.0E7 is approximately six
orders of magnitude greater than the overall structural mass (10.1 kg). The TABLED4 entry defines the
constant acceleration input. (One of the other TABLEDi entries can also be used, but the TABLED4
entry is chosen to show how to use it for enforced constant velocity and displacement later in this
example.)
Figure 7-4 shows the X-Y plots resulting from the input point (grid point 3) and an output point (grid
point 1). The plots show acceleration and displacement magnitudes. Note that the acceleration input is
not precisely 1.0 m/sec 2 ; there is a very slight variation between 0.9999 and 1.0000 due to the large mass
approximation.
Main Index
314 MD Nastran Dynamic Analysis User’s Guide
Examples
Main Index
CHAPTER 7 315
Enforced Motion
Figure 7-5 Displacements and Accelerations for the Two-DOF Model (continued)
This model was analyzed with several values of large mass. Table 7-1 shows the results. Note that the
model with the 10 6 mass ratio is the model discussed earlier. Peak frequency response results are
compared for each model, and the natural frequencies are compared to those of the constrained model in
Real Eigenvalue Analysis, 43. The table shows that a mass ratio of 10 6 is a good value to use for this
model.
Main Index
316 MD Nastran Dynamic Analysis User’s Guide
Examples
*Resonant frequencies for the constrained model are 4.7876 and 5.2909 Hz.
This model can also be changed to apply constant velocity or constant displacement at its base.
Listing 7-4 is an abridged input file for the model, showing the Bulk Data entries required for enforced
constant acceleration, enforced constant velocity, and enforced constant displacement. Note that only
one of these is usually applied to any model, but all three are shown here for comparison purposes. See
Install_dir/mdxxx/doc/dynamics/bd07twocm.dat.
Main Index
CHAPTER 7 317
Enforced Motion
Each input utilizes the TABLED4 entry. The TABLED4 entry uses the algorithm
N
x – X1 i
Y = A i ---------------- (7-54)
X2
i = 0
where x is input to the table, Y is returned, and N is the degree of the power series. When x X3 , X3
is used for x ; when x X4 , X4 is used for x . This condition has the effect of placing bounds on the
TABLED4 entry; note that there is no extrapolation outside of the table boundaries. There are N+1
entries to this table.
Constant acceleration is the easiest to apply since the force is proportional to the mass for all frequencies.
The power series for this case becomes
x – X1
A0 + A1 ---------------- (7-55)
X2
where:
A0 = 1.0
X1 = 0.0
X2 = 1.0
Main Index
318 MD Nastran Dynamic Analysis User’s Guide
Examples
Constant velocity involves a scale factor that is directly proportional to circular frequency ( 2f ). The
power series for this case becomes
x – X1
A0 + A1 ---------------- (7-56)
X2
where:
A0 = 0.0
A1 = 2 = 6.283185
X1 = 0.0
X2 = 1.0
Note that a phase change of 90 degrees is also required; this change is input using the TD field (field 7)
of the RLOAD1 entry.
Constant displacement involves a scale factor that is proportional to the circular frequency squared 2f 2
with a sign change. The power series for this case becomes
x – X1 x – X1 2
A0 + A1 ---------------- + A2 ---------------- (7-57)
X2 X2
where:
A0 = 0.0
A1 = 0.0
2
A2 = – 2 = – 39.4784
X1 = 0.0
X2 = 1.0
Main Index
CHAPTER 7 319
Enforced Motion
Large Mass
x
1.0E9 kg
Grid Point 1
Figure 7-6 Beam Model with Large Mass
Main Index
320 MD Nastran Dynamic Analysis User’s Guide
Examples
0.15
Idealized Input
Y·· (m/sec )
2
0.15
MD Nastran Input
Y·· (m/sec )
2
Time (sec)
0.05
0.052
Figure 7-7 Idealized Ramp Function Versus MD Nastran Ramp Function
Main Index
CHAPTER 7 321
Enforced Motion
DISPLACEMENT(PLOT) = 11
ACCELERATION(PLOT) = 11
$
$ XYPLOTS
$
... X-Y plot commands ...
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10.$
$
$ LARGE MASS OF 1.0E9
$CONM2 EID G CID M
CONM2 15 1 1.0E9
$
$ CONSTRAIN MASS IN 1,6 DIRECTIONS
SPC 21 1 16
$
$ DYNAMIC LOADING
$DLOAD SID S S1 L1
DLOAD 22 1.0E9 0.102 23
$TLOAD1 SID DAREA DELAY TYPE TID
TLOAD1 23 24 0 25
$DAREA SID P1 C1 A1
DAREA 24 1 2 0.15
$TABLED1 TID +TABL1
$+TABL1 X1 Y1 X2 Y2 ETC.
TABLED1 25 +TABL1
+TABL1 0.0 0.0 0.05 1.0 0.052 0.0 0.1 0.0 +TABL2
+TABL2 ENDT
$
$ CONVERT WEIGHT TO MASS: MASS = (1/G)*WEIGHT
$ G = 9.81 m/sec**2 --> WTMASS = 1/G = 0.102
PARAM WTMASS 0.102
$
$EIGRL SID V1 V2 MSGLVL
EIGRL 10 -1. 3000. 0
$
$TSTEP SID N1 DT1 NO1
TSTEP 27 1000 0.001 1
$
$ MODAL DAMPING OF 5% IN ALL MODES
A large mass of 1.0E9 kg is placed at grid point 1. This grid point is constrained in the T1- and R3-
directions but is free in the T2-direction. The load is scaled to give a peak input acceleration of
2
0.15 m/sec . This scaling is performed by applying a scale factor of 1.0E9 in the S field (field 3) of the
DLOAD entry, a scale factor of 0.102 in the S1 field (field 4) of the DLOAD entry, and a factor of 0.15
in the A1 field (field 5) of the DAREA entry. The applied load is scaled by 0.102 because the large mass
is also scaled by 0.102 due to the PARAM,WTMASS entry (see Eq. (7-47)). The time variation is
specified with the TABLED1 entry. The TLOAD1 entry specifies the type of loading (field 5) as 0
(applied force); this gives the same answers if the type is specified as 3 (enforced acceleration).
Figure 7-8 shows the displacement and acceleration response at grid points 1 (base) and 11 (tip). Note
that at the end of the acceleration pulse the base has a constant velocity, and therefore, a linearly
increasing displacement.
Main Index
322 MD Nastran Dynamic Analysis User’s Guide
Examples
Next, consider the same model with a 0.015 meter displacement imposed instead of an acceleration. The
same ramp time history function is used (with a peak enforced displacement of 0.015 meter) so that the
only change to the input file is to change the excitation type from 0 (applied force) to 1 (enforced
displacement on field 5 of the TLOAD1 entry) and the amplitude in the DAREA entry from 0.15 to 0.015.
See Install_dir/mdxxxx/doc/dynamics/bd07bar2.dat. Figure 7-9 shows the idealized input displacement
time history. Figure 7-10 shows the displacement and acceleration response at grid points 1 and 11.
Main Index
CHAPTER 7 323
Enforced Motion
0.15
Idealized Input
Y(m)
Figure 7-10 Response for Enforced Displacement (With the Rigid-Body Mode)
Now, consider a change to the enforced displacement run. In this case, remove the rigid-body mode's
contribution either by not computing the rigid-body mode (by setting V1 to a small positive value, such
as 0.01 Hz) or by neglecting the rigid-body mode in the transient response (by setting PARAM,LFREQ
to a small positive number, such as 0.01 Hz). See Install_dir/mdxxxx/doc/dynamics/bd07bar3.dat.
Figure 7-11 shows the resulting displacement and acceleration responses at grid points 1 and 11. Note
Main Index
324 MD Nastran Dynamic Analysis User’s Guide
Examples
that the responses are relative to the structure and are not absolute. The relative displacement of grid
point 1 should be zero, and it is very close to zero (i.e., 10 –10 ) as a result of the sufficiently large mass.
Figure 7-11 Response for Enforced Displacement (Without the Rigid-Body Mode)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 8: Random Analysis
8 Random Analysis
Random Analysis with Coupled Excitations
Main Index
326 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
Main Index
CHAPTER 8 327
Random Analysis
Enter
Cross
Yes Spectral No
Density
?
2
Sj = H j a H jb S a b Sj a = Hj a Sa
a b
Yes Combined
Loading
?
No
Sj = Sj a
No
Compute rms
value
Exit
Main Index
328 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
Theory
The application of these frequency response techniques to the analysis of random processes requires that
the system be linear and that the excitation be stationary with respect to time. The theory includes a few
important theorems that will be reviewed. An important quantity in random analysis theory is the
autocorrelation function R j , of a physical variable, u j , which is defined by
lim 1 T
R j = T --- u j t u j t – dt (8-1)
T 0
Note that R j 0 is the time average value of u j2 , which is an important quantity in the analysis of structural
failure. The one-sided power spectral density S j of u j is defined by
lim 2 T 2
S j = T --- e – i t u j t dt (8-2)
T 0
It may be shown (using the theory of Fourier integrals) that the autocorrelation function and the power
spectral density are Fourier transforms of each other. Thus, we define the autocorrelation function in
terms of frequency response functions
1
R j = ------ S j cos d (8-3)
2 0
from the mean-square theorem, the rms (root mean squared) magnitude, u j , is
1
u j2 = R j 0 = ------ S j d (8-4)
2 0
The expected value of the number of zero crossing with positive slope per unit time, or apparent
frequency, is a quantity of interest for fatigue analysis. This mean frequency, N 0 , can be found from the
power spectral density;
2
2 S j d
2 0
N0 = -------------------------------------------------
- (8-5)
S j d
0
The mean frequency, N 0 , is thus the root mean square frequency, where the power spectral density is used
as a weighing function.
The transfer function theorem states that, if H ja is the frequency response of any physical variable, u j ,
due to an excitation source, Q a , which may be a point force, a loading condition or some other form of
excitation, i.e., if
Main Index
CHAPTER 8 329
Random Analysis
u j = H ja Q a (8-6)
where u j and Q a are the Fourier transforms of u j and Q a , then the power spectral density of the
response S j , is related to the power spectral density of the source, S a , by
Sj = Hj a 2 Sa (8-7)
Eq. (8-7) is an important result because it allows the statistical properties (e.g., the Auto-correlation
function) of the response of a system to random excitation to be evaluated via the techniques of
frequency response. Another useful result is that, if sources Q 1 , Q 2 , Q 3 , etc., are statistically
independent, i.e., if the cross-correlation function between any pair of sources
lim 1 T
R a b = T --- q a t q b t – d (8-8)
T 0
is null, then the power spectral density of the total response is equal to the sum of the power spectral
densities of the responses due to individual sources. Thus
Sj = S ja = H ja 2 S a (8-9)
a a
If the sources are statistically correlated, the degree of correlation can be expressed by a cross-spectral
density, S ab , and the spectral density of the response may be evaluated from
Sj = H ja H jb S a b (8-10)
a b
As shown in Eq. (8-8), for a pair of response quantities, u a t and u b t , the cross-correlation function
between them is defined as
T
R a b = lim --1- u a t u b t – dt (8-11)
T T
0
where u a t and u b t can be displacement, velocity or single-point constraint force responses at the same
or different grid points; or stress, strain, and force components in the same or different elements. The
cross-correlation functions have the following relation
R a b – = R ba (8-12)
Main Index
330 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
T T
2 – i t i t
S a b = lim --- u a t e dt u a t e dt
T T
0 0
(8-13)
+
– i
= 2 R a b e d
–
+
1 i
R a b = ------
4 S a b e d (8-14)
–
where S *ab is the complex conjugate of S ab . If we denote S ab by its real and imaginary parts as
R I
S a b = S a b + iS a b
then from Eq. (8-15), we can see that S abR is an even function and S abI is an odd function of , i.e.,
R R
Sa b – = Sa b (8-16)
I I
Sa b – = –Sa b (8-17)
If the system is subject to multiple sources of excitation, the cross-PSD of responses u a t and u b t is
related to the PSDs of excitation sources by the frequency response functions as
*
Sa b = H a j H b k S j k (8-19)
j k
Main Index
CHAPTER 8 331
Random Analysis
The aforementioned equations of cross-PSDs and cross-correlation functions become the ones for auto-
PSDs and auto-correlation functions when u a t = u b t .
The calculation of the cross-correlation function from Eq. (8-18) is carried out by numerical integration
based on the trapezoidal approximation.
1
---
2
1-
2 j
uj = ----- S d (8-20)
0
In numerical calculation, the integral in Eq. (8-20) is computed for a specified frequency range
N–1 i+1
S j d = S j d (8-21)
0 i = 1 i
where log( .) is the logarithmic function to the base 10. The linear interpolation in a sub-domain
i i + 1 in this logarithmic coordinate system yields the following relation
Yi + 1 – Yi
Y – Y i = ------------------------- X – X i (8-22)
Xi + 1 – Xi
Sj i
S j = --------------- i i + 1 (8-23)
i
Main Index
332 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
with
log S j i + 1 S j i
= ---------------------------------------------------------
- (8-24)
log i + 1 i
i+1 i S j i ln i + 1 i = –1
S
S j d = --------------------------
j i +1 +1
- i + 1 – i – 1
(8-25)
i 1 + i
N0 is defined as
r
N 0 = -------j (8-26)
uj
where
1
---
N–1 i+1 2
1 2
rj = -------------- S j d (8-27)
3
2 i = 1
i
3
i+1 i S j i ln i + 1 i = – 3
2
S j d = --------------------------
Sj i +3 +3
- i + 1 – i – 3
(8-28)
i
3 + i
This logarithmic approach yields more accurate results if the PSDF curve is a straight line in an
integration sub-domain in a Log-Log scaled coordinate system. When the PSDF curve is close to a
straight line in a Log-Log scaled coordinate system, it can be expected that the Log-Log approximation
would produce a better result with fewer frequency points as compared to the trapezoidal approximation
of the PSDF using a linear coordinate system.
Inputs
Parameter, RMSINT, Log-Log selects the Log-Log approach for calculating RMS, N0 and CRMS.
Main Index
CHAPTER 8 333
Random Analysis
The cumulative RMS function u j of a random response quantity u j t at a set of selected N frequency
points is defined as
1
i ---
2
1
u j i = ------
2 S j d i = 1 2 N (8-29)
1
Inputs
The RANDOM Case Control command must be used to select the RANDPS/RANDT1 Bulk Data entries
to use.
CRMS function can be output in both the print and punch files. The option of CRMS is selected in Case
Control commands, ACCELERATION, DISPLACEMENT, FORCE, MPCFORCES, OLOAD,
SPCFORCES, STRAIN, STRESS and VELOCITY.
Output
PSDF, AUTO, RMS, N0, and CRMS output in Both Print and Punch Files.
These output requests can be achieved by selecting the RPRINT and RPUNCH options in Case Control
commands, ACCELERATION, DISPLACEMENT, FORCE, MPCFORCES, OLOAD, SPCFORCES,
STRAIN, STRESS and VELOCITY.
Exercise caution when using the RPRINT and RPUNCH options. For real-world problems, an enormous
amount of output may be generated using the RPRINT option. To turn this off, use the NORPRINT
option. The default is RPRINT.
Main Index
334 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
Input:
See Listing 8-1 for a partial input file, the complete input file is
nstall_dir/mdxxxx/doc/dynamics//random_xyz.dat
Main Index
CHAPTER 8 335
Random Analysis
FREQ = 604
SDAMP = 111
ACCELERATION(plot,rprint, psdf)=907
STRESS(Plot, rprint, psdf) = 908
SUBCASE 11 $ Excitation in x direction
DLOAD = 101
SUBCASE 12 $ Excitation in y direction
DLOAD = 102
SUBCASE 13 $ Excitation in z direction
DLOAD = 103
$
OUTPUT(xyplot)
xmin = 10.
xmax = 1000.
xgrid = yes
ygrid = yes
xlog = yes
ylog = yes
$
Xtitle= accel psdf Node 111 (T3)
xypeak,xyplot accel psdf / 111(t3)
$
$ For stress and force component, refer to appendix A item-code
$ in Quick Reference Guide. Use complex item code.
$
xtitle= Stress psdf Elem. 1 Sx
xypeal,xyplot stress psdf /1(3)
BEGIN BULK
$
$ 201 202 203 204 205 206 207 208 209 210 211
$ /*-----*-----*-----*-----*-----*-----*-----*-----*-----*-----*
$ Y /| | | | | | | | | | |
$ ^ /| | | | | | | | | | |
$ | /| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
$ | /| | | | | | | | | | |
$ | /| | | | | | | | | | |
$ +--->X /*-----*-----*-----*-----*-----*-----*-----*-----*-----*-----*
$ 101 102 103 104 105 106 107 108 109 110 111
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......0
$
PARAM AUTOSPC YES
PARAM WTMASS .00259
$
PARAM RMSINT LOGLOG
$
SPC1,666,123456,101,201
$
EIGRL 219 -.1 2000.
$
TABDMP1 111 CRIT
0. .04 99999. 0.04 Endt
$
$ All DOF on SPCD must also be declared in selected SPC set(see SPC 666)
$
$ Define Unit input in X direction
$
$ Since PSDF input in G^2/Hz, we must define Unit G (386.4 inch/sec^2)
$
RLOAD1,101,131, , , 11, ,ACCE
Main Index
336 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
SPCD,131,101,1,386.4
SPCD,131,201,1,386.4
$
$ Define Unit input in Y direction
$
RLOAD1,102,132, , , 11, ,ACCE
SPCD,132,101,2,386.4
SPCD,132,201,2,386.4
$
$ Define Unit input in Z direction
$
RLOAD1,103,133, , , 11, ,ACCE
SPCD,133,101,3,386.4
SPCD,133,201,3,386.4
$
TABLED1,11
+,0.0,1.0,2000.0,1.0,endt
$
$ Select all psdf input break points
$
FREQ, 604, 15.0, 15.1, 25.0, 25.1, 110.0, 110.1
FREQ, 604, 34.9, 35.0, 55.0, 55.1, 310.0, 310.1
$
$ Select more output frequencies - equally spaced based on log scale.
$
FREQ2,604,10.,1000.,50
$
$ Select modal frequencies to capture expected peak at modal frequencies.
$
FREQ4,604,10.,1000.,.1,3
$
$ Define Random Input for all 3 direction using same ID. All RANDPS entry
$ with same ID will be added together.
$
RANDPS,99, 11, 11, 1.0, 0.0, 11
RANDPS,99, 12, 12, 1.0, 0.0, 12
RANDPS,99, 13, 13, 1.0, 0.0, 13
$
TABRND1, 11, Log, Log
+, 10.0, 0.18, 15.0, 0.18, 15.1, 0.12, 25.0, 0.12
+, 25.1, 0.22, 110.0, 0.22, 110.1, 0.12, 1000.0, 0.12
+, Endt
$
TABRND1, 12, Log, Log
+, 10.0, 0.18, 15.0, 0.18, 15.1, 0.12, 25.0, 0.12
+, 25.1, 0.22, 110.0, 0.22, 110.1, 0.12, 1000.0, 0.12
+, Endt
$
TABRND1, 13, Log, Log
+, 10.0, 0.15, 34.9, 0.15, 35.0, 0.25, 55.0, 0.25
+, 55.1, 0.15, 310.0, 0.15, 310.1, .075, 1000.0, .075
+, Endt
$
Main Index
CHAPTER 8 337
Random Analysis
POINT-ID = 105
A C C E L E R A T I O N V E C T O R
( POWER SPECTRAL DENSITY FUNCTION )
FREQUENCY TYPE T1 T2 T3 R1 R2 R3
1.000000E+01 G 2.687507E+04 2.689527E+04 2.426265E+04 2.120942E-03 7.021265E+00 6.942450E-04
1.096478E+01 G 2.687511E+04 2.689941E+04 2.469965E+04 3.224830E-03 1.061186E+01 1.003899E-03
1.202264E+01 G 2.687514E+04 2.690436E+04 2.525660E+04 4.960394E-03 1.620609E+01 1.451786E-03
1.318257E+01 G 2.687521E+04 2.691035E+04 2.597651E+04 7.742561E-03 2.507842E+01 2.099709E-03
1.445440E+01 G 2.687526E+04 2.691752E+04 2.692406E+04 1.231372E-02 3.947413E+01 3.037153E-03
1.500000E+01 G 2.687528E+04 2.692080E+04 2.739068E+04 1.493719E-02 4.765870E+01 3.523486E-03
…
…
7.585776E+02 G 1.875131E+04 8.608298E+03 5.369716E+03 6.176474E+01 1.771618E+02 5.640190E+02
8.317637E+02 G 1.888297E+04 9.738106E+03 4.871404E+03 6.839265E+01 1.015533E+03 5.069272E+02
9.120108E+02 G 1.904800E+04 1.111841E+04 5.791567E+03 9.793225E+01 4.436593E+03 4.546082E+02
9.536902E+02 G 1.914198E+04 1.193009E+04 7.372608E+03 1.359111E+02 1.017060E+04 4.294507E+02
9.733126E+02 G 1.918815E+04 1.234163E+04 8.699032E+03 1.666866E+02 1.579330E+04 4.178163E+02
1.000000E+03 G 1.925294E+04 1.293666E+04 1.161850E+04 2.342402E+02 3.087324E+04 4.020715E+02
POINT-ID = 28
A C C E L E R A T I O N V E C T O R
( POWER SPECTRAL DENSITY FUNCTION )
FREQUENCY TYPE T1 T2 T3 R1 R2 R3
1.000000E+01 G 5.453022E-18 1.056328E-17 3.099549E+02 5.773942E-01 1.416458E+02 6.138918E-19
1.015469E+01 G 5.800392E-18 1.123618E-17 3.296992E+02 6.139689E-01 1.506693E+02 6.529981E-19
1.031177E+01 G 6.169956E-18 1.195208E-17 3.507051E+02 6.528609E-01 1.602694E+02 6.946032E-19
1.047129E+01 G 6.563146E-18 1.271375E-17 3.730536E+02 6.942172E-01 1.704832E+02 7.388679E-19
1.063326E+01 G 6.981471E-18 1.352410E-17 3.968308E+02 7.381930E-01 1.813499E+02 7.859621E-19
1.079775E+01 G 7.426557E-18 1.438629E-17 4.221290E+02 7.849557E-01 1.929119E+02 8.360691E-19
1.096478E+01 G 7.900112E-18 1.530364E-17 4.490453E+02 8.346807E-01 2.052135E+02 8.893813E-19
...
...
9.261187E+02 G 9.289492E-14 1.807436E-13 2.098790E+08 6.027814E+02 5.506023E+05 1.036910E-14
9.404448E+02 G 9.372058E-14 1.824654E-13 1.759028E+08 3.949018E+02 6.941084E+05 1.045697E-14
9.549926E+02 G 9.456252E-14 1.842211E-13 1.505420E+08 2.619290E+02 8.747291E+05 1.054565E-14
9.697653E+02 G 9.542219E-14 1.860153E-13 1.310669E+08 1.819211E+02 1.074708E+06 1.063531E-14
9.847666E+02 G 9.630063E-14 1.878519E-13 1.157493E+08 1.403549E+02 1.285253E+06 1.072605E-14
1.000000E+03 G 9.719865E-14 1.897341E-13 1.034542E+08 1.277149E+02 1.502181E+06 1.081792E-14
Main Index
338 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
A C C E L E R A T I O N V E C T O R
( ROOT MEAN SQUARE )
ELEMENT-ID = 1
S T R E S S E S I N Q U A D R I L A T E R A L E L E M E N T S ( Q U A D 4 )
( POWER SPECTRAL DENSITY FUNCTION )
S T R E S S E S I N Q U A D R I L A T E R A L E L E M E N T S ( Q U A D 4 )
( ROOT MEAN SQUARE )
S T R E S S E S I N Q U A D R I L A T E R A L E L E M E N T S ( Q U A D 4 )
( NUMBER OF ZERO CROSSINGS )
Main Index
CHAPTER 8 339
Random Analysis
Example 2: Plate - subjected to Pressure and point load with Cross-Spectrum Input
Problem Definition: Plate Dimension 5 inch x 2 Inch, Thickness = 0.1 inch.
Material: E = 3.E07, Poison’s Ratio = 0.3, Density = 0.282 lbs/inch3.
Damping = 0.03% (Critical)
Boundary Condition: Fixed at one shorter edge.
Random Input:
Main Index
340 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
Loading:
Main Index
CHAPTER 8 341
Random Analysis
Input
Cross-Spectrum PSD (coupled PSD) usually have Real and Imaginary components. Two RANDPS and
TABRND1 entries are required for complex spectrum input. In the first RANDPS entry, set the real (X)
component to non-zero and imaginary (Y) component to zero. The TID on this RANDPS points to table
TABRND1 with real component of cross-psd input. In the second RANDPS entry, set the real (X)
component to zero and imaginary (y) component to non-zero with TID points to TABRND1 table with
imaginary component of cross-psd input (see Listing 8-2). The complete file is in
Install_dir/mdxxxx/doc/dynamics/random_freq.dat. For uncoupled PSD input, only one RANDPS entry
is required. See the MD Nastran Quick Reference Guide for more information.
Main Index
342 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
Main Index
CHAPTER 8 343
Random Analysis
POINT-ID = 28
A C C E L E R A T I O N V E C T O R
( POWER SPECTRAL DENSITY FUNCTION )
FREQUENCY TYPE T1 T2 T3 R1 R2 R3
1.000000E+01 G 5.453022E-18 1.056328E-17 3.099549E+02 5.773942E-01 1.416458E+02 6.138918E-19
1.015469E+01 G 5.800392E-18 1.123618E-17 3.296992E+02 6.139689E-01 1.506693E+02 6.529981E-19
1.031177E+01 G 6.169956E-18 1.195208E-17 3.507051E+02 6.528609E-01 1.602694E+02 6.946032E-19
1.047129E+01 G 6.563146E-18 1.271375E-17 3.730536E+02 6.942172E-01 1.704832E+02 7.388679E-19
1.063326E+01 G 6.981471E-18 1.352410E-17 3.968308E+02 7.381930E-01 1.813499E+02 7.859621E-19
1.079775E+01 G 7.426557E-18 1.438629E-17 4.221290E+02 7.849557E-01 1.929119E+02 8.360691E-19
1.096478E+01 G 7.900112E-18 1.530364E-17 4.490453E+02 8.346807E-01 2.052135E+02 8.893813E-19
...
...
9.261187E+02 G 9.289492E-14 1.807436E-13 2.098790E+08 6.027814E+02 5.506023E+05 1.036910E-14
9.404448E+02 G 9.372058E-14 1.824654E-13 1.759028E+08 3.949018E+02 6.941084E+05 1.045697E-14
9.549926E+02 G 9.456252E-14 1.842211E-13 1.505420E+08 2.619290E+02 8.747291E+05 1.054565E-14
9.697653E+02 G 9.542219E-14 1.860153E-13 1.310669E+08 1.819211E+02 1.074708E+06 1.063531E-14
9.847666E+02 G 9.630063E-14 1.878519E-13 1.157493E+08 1.403549E+02 1.285253E+06 1.072605E-14
1.000000E+03 G 9.719865E-14 1.897341E-13 1.034542E+08 1.277149E+02 1.502181E+06 1.081792E-14
Main Index
344 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
A C C E L E R A T I O N V E C T O R
( ROOT MEAN SQUARE )
A C C E L E R A T I O N V E C T O R
( NUMBER OF ZERO CROSSINGS )
X Y - O U T P U T S U M M A R Y ( A U T O O R P S D F )
PLOT CURVE FRAME CURVE ID./ RMS NO. POSITIVE XMIN FOR XMAX FOR YMIN FOR X FOR YMAX FOR X FOR
TYPE TYPE NO. PANEL : GRID ID VALUE CROSSINGS ALL DATA ALL DATA ALL DATA YMIN ALL DATA YMAX
AUTO ACCE 1 55( 5) 1.464592E+06 3.549814E+02 0.000E+00 1.000E-01 -1.630E+12 3.150E-03 2.145E+12 0.000E+00
P R I N T E D D A T A F O R T H I S C U R V E F O L L O W S
X Y - O U T P U T S U M M A R Y ( A U T O O R P S D F )
PLOT CURVE FRAME CURVE ID./ RMS NO. POSITIVE XMIN FOR XMAX FOR YMIN FOR X FOR YMAX FOR X FOR
TYPE TYPE NO. PANEL : GRID ID VALUE CROSSINGS ALL DATA ALL DATA ALL DATA YMIN ALL DATA YMAX
PSDF ACCE 2 28( 5) 6.028776E+05 5.728496E+02 1.000E+01 1.000E+03 3.100E+02 1.000E+01 1.529E+10 1.332E+02
P R I N T E D D A T A F O R T H I S C U R V E F O L L O W S
Main Index
CHAPTER 8 345
Random Analysis
Main Index
346 MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 9: Response Spectrum Analysis
Shock and Response Spectrum Analysis
Response Spectrum Examples
Main Index
348 MD Nastran Dynamic Analysis User’s Guide
Shock and Response Spectrum Analysis
Theoretical Background
Starting with a modal transient analysis, the general approximation for a response quantity, u k , is
uk t = i k i t (9-1)
i
where and are the modal outputs and generalized displacements. The actual modal equations are
·· · T
i + g i i i + i2 i = i Pt (9-2)
where P is the vector of loading functions. For loading due to base accelerations, the equivalent inertial
loads are
P t = – M a a D a r u·· r t (9-3)
Main Index
CHAPTER 9 349
Response Spectrum Analysis
where the columns of D ar represent vectors of rigid body motions of the whole structure and the
accelerations correspond to the base motions. Substituting Eq. (9-3) into Eq. (9-2) and combining terms
we can separate the modal quantities from the transient solutions. First we will develop the transient
response functions. We begin by calculating the responses
where x r is a response function in direction r , and is a function of the variables , g , and t . The peak
values of x r , obtained over a range of frequencies and damping factors is called the response spectrum
for the excitation, u·· r .
Next, from the normal mode analysis, we define the participation factors ir , for mode i and direction
r , as
T
ir = – i M aa D ar (9-5)
Then, from Eq. (9-1), the actual transient response at a physical point is
uk t = ik ir x r i ,g i ,t (9-6)
i r
The peak magnitudes of u k in Eq. (9-6) are usually dominated by the peak values of x t occurring at
the natural frequencies. In spectrum analysis the peak values of u k are approximated by combining
functions of the peak values, x ri i ,g i = max x ri i ,g i ,t , in the approximation
uk t i k ir x ri i ,g i (9-7)
i r
ABS Option
Equations (9-6) and (9-7) define the ABS (Absolute Value) option. This method assumes the worst case
scenario in which all of the modal peak values for every point on the structure are assumed to occur at
the same time and in the same phase. Clearly in the case of a sudden impact, this is not very probable
because only a few cycles of each mode will occur. However, in the case of a long term vibration, such
as an earthquake when the peaks occur many times and the phase differences are arbitrary, this method
is acceptable.
A second way of viewing the problem is to assume that the modal magnitudes and phases will combine
in a probabalistic fashion. If the input loads are behaving randomly, the probable (RMS) peak values are
2
uk i k i (9-8)
i
Main Index
350 MD Nastran Dynamic Analysis User’s Guide
Shock and Response Spectrum Analysis
2
i = ir x r i ,g ii (9-9)
r
SRSS Method
This approach is known as the SRSS (square root of sum-squared) method. Note that the results in each
direction are summed in vector fashion for each mode first, followed by an SRSS calculation for all
modes at each selected output quantity u k . It is assumed that the modal responses are uncorrelated and
the peak value for each mode will occur at a different time. These results are optimistic and represent a
lower bound on the dynamic peak values.
The SRSS method may underestimate the actual peaks since the result is actually a probable peak value
for the period of time used in the spectrum analysis. The method is normally augmented with a safety
factor of 1.5 to 2.0 on the critical outputs.
NRL Method
As a compromise between the two methods above, the NRL (Naval Research Laboratories) method was
developed. Here, the peak response is calculated from the equation
2
u k jk j + i k i
ij
where the j-th mode is the mode that produces the largest magnitude in the product jk j . The peak modal
magnitudes, jk j , are calculated with Eq. (9-9).
The rationale for the method is that the peak response will be dominated by one mode and the SRSS
average for the remaining modes could be added directly. The results will fall somewhere between the
ABS and SRSS methods.
Modes that are close in frequency may have their peak response occur at about the same time (and with
the same phase). For this reason, the SRSS and NRL methods contain a provision to sum modal
responses via the ABS method for modes that have closely spaced natural frequencies. Close natural
frequencies are defined by frequencies that meet the following inequality:
f i + 1 CLOSE f i
Main Index
CHAPTER 9 351
Response Spectrum Analysis
Main Index
352 MD Nastran Dynamic Analysis User’s Guide
Shock and Response Spectrum Analysis
(d)
(e)
f 1 f 2 f 3 . . . f max
(c)
Peak
(b)
Response
f 1 f 2 f 3 . . . f max
(a) Transient Series of Oscillators
Response Resonant Frequency
Base
Transient Structure
Excitation
Transient excitation (a) is applied to a base structure (b), from which transient response (c)
is computed for each floor. This response is applied to a series of oscillators (d), for which
the peak response is plotted (e). Steps (d) and (e) are repeated for different damping values
to form response spectra as shown below.
= 5% critical damping
= 2% critical damping
= 0% critical damping
Peak
Response
fn
Resonant Frequency (f)
Note that the peak response for one oscillator does not necessarily occur at the same time as the peak
response for another oscillator. Note, too, that there is no phase information since only the magnitude of
peak response is computed.
Several values of peak response are computed for the oscillators. These include the following response
variable.
Relative velocity and absolute acceleration are approximately related to the relative displacement by
X· r X r
(9-10)
X·· 2 X r
Main Index
CHAPTER 9 353
Response Spectrum Analysis
X0
(9-11)
Xr Ub
X Ub
(9-12)
Xr 0
The approximate relationships between X r , X·· r , and X·· , shown in Eq. (9-10), are not valid at very low or
very high oscillator frequencies or for large values of damping.
It is assumed in this process that the mass of each oscillator is very small relative to the base structure,
so the oscillator’s response does not influence the response of the base structure.
Main Index
354 MD Nastran Dynamic Analysis User’s Guide
Shock and Response Spectrum Analysis
Main Index
CHAPTER 9 355
Response Spectrum Analysis
The time step, DT, and time duration, DT N (where N is the number of time increments), must take
into account the loading, the base structure, and the frequency range of the spectra generation. The time
step must take into account the frequency content of the applied excitation, the frequencies of the base
structure, and the highest frequency for which spectra are to be generated. There must be enough time
steps per cycle of response for both the base structure and the highest frequency oscillator in order to
accurately predict the peak response; 5 to 10 steps per cycle represents a typical value. In addition, the
time duration of the loading, the frequencies of the base structure, and the lowest oscillator frequency
must be considered when defining the time duration. There must be a long enough time duration of
response both for the base structure and the lowest frequency oscillator in order to accurately predict the
peak response. For short duration loadings, the peak response often occurs well after the the load has
peaked.
Initial conditions (specified via the TIC3 - MD Only Bulk Data entry) are not used in response spectrum
generation. Initial conditions are used in the calculation of the transient response of the base structure,
but the calculation of the peak oscillator responses (i.e., the response spectrum calculation) ignores any
initial conditions.
The XYPLOT and XYPUNCH commands are included in the OUTPUT(XYPLOT) section. Further
details about the OUTPUT(XYPLOT) Section are described in Results Processing (Ch. 15).
Main Index
356 MD Nastran Dynamic Analysis User’s Guide
Shock and Response Spectrum Analysis
Relative and absolute spectra are denoted by IP and RM, respectively, in the parentheses of the curve
request.
There are two FREQi entries: one to specify oscillator frequencies (i.e., frequencies for which spectra
will be computed) and the other to specify oscillator damping. (Note that damping for the base structure
is specified in another manner, such as with the TABDMP1 entry used for modal transient response
analysis.) For a more detailed description see the DTI,SPECSEL (p. 1678) in the MD Nastran Quick
Reference Guide and the examples described in this chapter.
Main Index
CHAPTER 9 357
Response Spectrum Analysis
All input listed in the table is required with the exception of PARAM,OPTION and PARAM,CLOSE.
Main Index
358 MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
Figure 9-2 Chimney Model (24 inch radius, 700 inch long and 0.25 inch thick)
Main Index
CHAPTER 9 359
Response Spectrum Analysis
Main Index
360 MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
$
spc1,77,23456,9999
conm2, 9999, 9999, , 1.e8
suport, 9999, 1
Tload1, 66, 44, , acce, 99
darea, 44, 9999, 1, 1.e8
$
tabdmp1, 88, crit
+, 0.0, 0.02, 10000., 0.02, endt
$
$ NASTRAN input file created by the Patran 2008r2 input file translator
$ on May 07, 2010 at 11:50:09.
$
$ Direct Text Input for Bulk Data
$ Elements and Element Properties for region : pshell.11
PSHELL 11 1 .25 1 1
$
$ Material Record : alum
$ Description of Material : Date: 07-May-10 Time: 11:46:54
MAT1 1 1.+7 .25 .1
$ Multipoint Constraints of the Entire Model
RBE2 9999 9999 123456 1 2 3 4 5
6 7 8 9 10 11 12 13
14 15 16 17 18 19 20 21
22 23 24 25 26 27 28 29
30 31 32 33 34 35 36
$$
$ -----------------------------------------------------
$
$ ... rest of model ...
$
ENDDATA
The XYPUNCH ACCE SPECTRAL command in the Case Control Section punches the response
spectrum (which will be used later, in the response spectrum application). In the Bulk Data Section,
PARAM,RSPECTRA,0 requests calculation of the spectrum. The DTI,SPSEL entry references the
FREQ entries; one of the FREQ entries defines the oscillator damping and the other FREQ entry defines
the frequency range over which to compute the spectra. The DTI,SPSEL entry also defines grid points
at which to compute spectra; spectra are computed for grid points 9999, 2593 - though only the one for
grid point 9999 is punched for subsequent use. Note that the SUPORT entry for Grid 9999 is used for
DOF 1 since Response Spectrum is applied at grid 9999 in X-direction only.
Note that a 2% damped spectrum was computed because that is the spectrum that will be applied to a
component model. In practice, however, spectra are often generated for multiple damping values (for
example, 0%, 2%, and 5% damping).
The plotted absolute acceleration response spectra for grid points 9999 and 2593 are shown in Figure 9-3.
Main Index
CHAPTER 9 361
Response Spectrum Analysis
Figure 9-3 Absolute Acceleration Response Spectrum Plots for Grid 9999 (base) and 2593
(tip)
A portion of the printed output file is shown in Listing 9-2. A portion of the absolute acceleration
spectrum output for grid point 9999 is shown. Spectra for the other grid points are also contained in the
output file, as are displacement and velocity spectra. The printed format for each spectrum is similar to
that of frequency response analysis, with the absolute spectrum output in the real location and the relative
spectrum output in the imaginary location. Because relative acceleration spectra are not calculated, those
components are zero. The punched spectrum for grid point 9999 is shown in Listing 9-3; this output is
contained in the punch file.
TRANSIENT ANALYSIS
FRACTION OF CRITICAL DAMPING = .02 ABSOLUTE IN REAL LOCATION, RELATIVE IN IMAG. LOCATION
POINT-ID = 9999
C O M P L E X A C C E L E R A T I O N V E C T O R
(REAL/IMAGINARY)
FREQUENCY TYPE T1 T2 T3 R1 R2 R3
1.000000E+00 G 2.539913E+02 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
1.018615E+00 G 2.624984E+02 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
1.037578E+00 G 2.648515E+02 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
1.056893E+00 G 2.622424E+02 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
1.076567E+00 G 2.911194E+02 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
1.096608E+00 G 3.119053E+02 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
1.117022E+00 G 3.222946E+02 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
Main Index
362 MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
$SUBCASE 1 1
$ACCE 3 9999 3 1 2
$ 2.000000E-02
TABLED1 2
1.0 253.991 1.01862 262.498 1.03758 264.852 1.05689 262.242
1.07657 291.119 1.09661 311.905 1.11702 322.295 1.13782 321.107
1.159 307.908 1.18057 283.899 1.20255 250.723 1.22494 210.996
1.24774 194.307 1.27097 197.562 1.29463 201.576 1.31873 206.812
1.34328 213.406 1.36828 242.229 1.39375 267.792 1.4197 276.681
1.44613 281.898 1.47305 272.642 1.50047 279.163 1.5284 292.666
1.55685 306.362 1.58583 320.748 1.61535 334.938 1.64543 349.981
1.67606 367.169 1.70726 447.014 1.73904 504.022 1.77141 526.362
1.80439 506.608 1.83798 456.479 1.87219 468.317 1.90704 476.508
1.94254 480.528 1.97871 480.239 2.01554 475.988 2.05306 478.137
2.09128 525.215 2.13021 522.18 2.16986 474.84 2.21026 449.432
2.2514 445.012 2.29331 440.6 2.33601 436.617 2.37949 433.057
2.42379 429.622 2.46891 425.95 2.51487 421.863 2.56168 417.533
2.60937 413.535 2.65794 410.717 2.70742 409.838 2.75782 411.399
2.80916 415.659 2.85201 420.869 2.86146 421.976 2.91472 428.647
2.96898 433.677 3.02425 434.769 3.08055 429.637 3.1379 416.123
3.19631 392.679 3.20851 387.077 3.25581 361.06 3.31642 323.823
3.37816 289.818 3.44104 270.674 3.5051 277.729 3.56502 304.895
3.57035 307.97 3.63681 348.876 3.70452 387.358 3.77348 413.504
3.84372 419.772 3.91527 406.071 3.92152 404.055 3.98816 373.676
4.0624 344.687 4.13803 352.075 4.21506 344.042 4.27802 340.393
4.29352 337.824 4.37345 328.908 4.45486 347.942 4.53779 367.092
…
…
18.7775 136.436 19.1271 145.079 19.2869 145.384 19.4831 144.983
19.8458 144.574 20.1562 141.552 20.2152 142.177 20.5916 142.894
20.9749 140.99 21.3653 138.117 21.4299 137.696 21.7631 137.808
22.1682 139.547 22.3958 138.58 22.5809 136.465 23.0012 135.937
23.4294 136.448 23.5729 135.595 23.8656 134.205 24.3098 134.426
24.6354 134.988 24.7177 136.433 24.7624 137.428 24.9116 139.976
25.2233 140.917 25.6929 134.666 25.7159 134.267 25.9545 132.335
26.1712 130.443 26.6584 134.375 26.875 135.894 27.1546 134.85
27.6601 135.962 27.8074 137.239 28.0256 138.53 28.175 138.816
28.6995 135.818 28.8944 135.455 29.1988 134.562 29.2338 134.38
29.778 131.056 30.3323 131.266 30.897 134.551 31.1396 135.406
31.2354 135.552 31.4721 135.882 32.058 135.389 32.4431 135.517
32.5062 135.801 32.6548 136.609 33.2627 137.431 33.8819 131.144
33.9869 129.982 34.2535 127.792 34.5126 126.866 34.6247 126.865
35.1398 127.605 35.1551 127.591 35.6875 124.677 35.8095 123.267
36.118 121.428 36.4761 117.9 37.0766 116.832 37.1551 116.854
37.3675 116.918 37.8468 116.975 38.5513 115.289 38.9318 115.698
38.9528 115.704 39.0442 115.692 39.269 115.296 39.7298 112.701
40. 110.989ENDT
Main Index
CHAPTER 9 363
Response Spectrum Analysis
spc = 77
method = 99
dload = 701
sdamp = 88
param,option,srss
accel(plot) = all
stress = all
force(plot) = all
$
begin bulk
$
param,wtmass,.002589
param,grdpnt,0
param,post,-1
param, scrspec, 0
$
eigrl, 99, , 40.0,
$
$ Multipoint Constraints of the Entire Model
RBE2 9999 9999 123456 1 2 3 4 5
6 7 8 9 10 11 12 13
14 15 16 17 18 19 20 21
22 23 24 25 26 27 28 29
30 31 32 33 34 35 36
$
spc1,77,23456,9999
conm2, 9999, 9999, , 1.e8
suport, 9999, 1
$
Dload, 701, 1.0, 1.0, 71
darea, 44, 9999, 1, 1.e8
$
Dti, Specsel, 71, , A, 2, 0.02
$
$
tabdmp1, 88, crit
+, 0.0, 0.02, 10000., 0.02, endt
$
$ Shock Spectrum at base - Punch file created in Example 1.
$
$SUBCASE 1 1
$ACCE 3 9999 3 1 2
$ 2.000000E-02
TABLED1 2
1.0 253.991 1.01862 262.498 1.03758 264.852 1.05689 262.242
1.07657 291.119 1.09661 311.905 1.11702 322.295 1.13782 321.107
1.159 307.908 1.18057 283.899 1.20255 250.723 1.22494 210.996
1.24774 194.307 1.27097 197.562 1.29463 201.576 1.31873 206.812
1.34328 213.406 1.36828 242.229 1.39375 267.792 1.4197 276.681$
$-------------------------------------------------------
$
$ ... rest of model ...
$
ENDDATA
Main Index
364 MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
defines the modal damping (every mode has 2% critical damping). The TABLED1 entry defines the
input spectrum; this is the punch file that was generated from the spectrum generation run.
Discussion of Results
Listing 9-5 shows a portion of the resulting printed output. The eigenvalue summary shows the computed
natural frequencies; note the rigid body mode, which occurred because the bracket was not constrained
in the x direction (there was a SUPORT for that DOF). Matrix FN is the list of natural frequencies of the
modes used for analysis with response spectrum input (20 in this case—the rigid body mode is not
included because our spectrum started at 1.0 Hz). Matrix PSIT lists the modal participation factors in
transposed form, with one column for each mode and one row for each input point (one in this case).
Note that the some of the modes (e.g. mode 2, 4, 5, 6 etc) cannot be readily excited by base motion in X
direction, since its response to the load is orders of magnitude less than the second (first elastic) mode.
Matrix UHVR occurs once for every analysis subcase with response spectrum input, and it lists the peak
modal response. The first column is displacement, the second is velocity, and the third is acceleration.
There is one row for each mode used in the response spectrum analysis (20 in this case). A portion of
the resulting maximum accelerations, forces and stresses—quantities selected for output via Case
Control—are also shown in Listing 9-5. These quantities are computed using the SRSS method.
Main Index
CHAPTER 9 365
Response Spectrum Analysis
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 0.0 0.0 0.0 1.000000E+00 0.0
2 2 5.017444E+02 2.239965E+01 3.565015E+00 1.000000E+00 5.017444E+02
3 3 5.017527E+02 2.239984E+01 3.565045E+00 1.000000E+00 5.017527E+02
4 4 4.760230E+03 6.899442E+01 1.098080E+01 1.000000E+00 4.760230E+03
5 5 4.760230E+03 6.899442E+01 1.098080E+01 1.000000E+00 4.760230E+03
6 6 6.720657E+03 8.197961E+01 1.304746E+01 1.000000E+00 6.720657E+03
7 7 6.720657E+03 8.197961E+01 1.304746E+01 1.000000E+00 6.720657E+03
...
...
17 17 4.155331E+04 2.038463E+02 3.244314E+01 1.000000E+00 4.155331E+04
18 18 5.150000E+04 2.269361E+02 3.611800E+01 1.000000E+00 5.150000E+04
19 19 5.150000E+04 2.269361E+02 3.611800E+01 1.000000E+00 5.150000E+04
20 20 6.018289E+04 2.453220E+02 3.904421E+01 1.000000E+00 6.018289E+04
21 21 6.018289E+04 2.453220E+02 3.904421E+01 1.000000E+00 6.018289E+04
PSIT
POINT VALUE POINT VALUE POINT VALUE POINT VALUE POINT VALUE
COLUMN 1
9999 T1 -5.08829E+02
COLUMN 2
9999 T1 4.13628E-06
COLUMN 3
9999 T1 2.07668E+00
COLUMN 4
9999 T1 -1.29937E-10
COLUMN 5
9999 T1 8.55402E-11
...
...
COLUMN 18
9999 T1 2.22644E-12
COLUMN 19
9999 T1 -4.85724E-12
COLUMN 20
9999 T1 3.55792E-13
COLUMN 21
9999 T1 -7.24552E-13
Main Index
366 MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
MATRIX UHVR (GINO NAME 101 ) IS A REAL 3 COLUMN X 21 ROW RECTANG MATRIX.
COLUMN 1 ROWS 2 THRU 21 --------------------------------------------------
ROW
2) 2.5135E-06 1.2620E+00 5.3064E-12 -3.4933E-12 1.4455E-12 -3.2655E-13 2.8290E-10 8.8593E-03 1.8225E-13 -2.2641E-13
12) 2.5064E-14 -2.8510E-14 5.5913E-14 -3.4243E-14 8.9867E-15 -2.8723E-14 5.2496E-15 -1.1452E-14 6.8395E-16 -1.3928E-15
COLUMN 2 ROWS 2 THRU 21 --------------------------------------------------
ROW
2) 5.6301E-05 2.8268E+01 3.6612E-10 -2.4102E-10 1.1850E-10 -2.6770E-11 3.8092E-08 1.1929E+00 2.5645E-11 -3.1860E-11
12) 4.8658E-12 -5.5346E-12 1.0940E-11 -6.6998E-12 1.8319E-12 -5.8552E-12 1.1913E-12 -2.5990E-12 1.6779E-13 -3.4169E-13
COLUMN 3 ROWS 2 THRU 21 --------------------------------------------------
ROW
2) 1.2611E-03 6.3320E+02 2.5260E-08 -1.6629E-08 9.7146E-09 -2.1946E-09 5.1290E-06 1.6062E+02 3.6087E-09 -4.4832E-09
12) 9.4460E-10 -1.0745E-09 2.1404E-09 -1.3109E-09 3.7343E-10 -1.1936E-09 2.7035E-10 -5.8980E-10 4.1162E-11 -8.3825E-11
S T R E S S E S I N Q U A D R I L A T E R A L E L E M E N T S ( Q U A D 4 )
ELEMENT FIBER STRESSES IN ELEMENT COORD SYSTEM PRINCIPAL STRESSES (ZERO SHEAR)
ID. DISTANCE NORMAL-X NORMAL-Y SHEAR-XY ANGLE MAJOR MINOR VON MISES
1 -1.250000E-01 1.484016E+03 1.375312E+02 6.375009E+00 0.2713 1.484047E+03 1.375010E+02 1.420297E+03
1.250000E-01 1.526726E+03 1.457677E+02 5.816203E+00 0.2413 1.526751E+03 1.457432E+02 1.459347E+03
2 -1.250000E-01 1.438925E+03 1.333524E+02 1.893104E+01 0.8306 1.439199E+03 1.330779E+02 1.377490E+03
1.250000E-01 1.480337E+03 1.413386E+02 1.727163E+01 0.7389 1.480560E+03 1.411158E+02 1.415288E+03
3 -1.250000E-01 1.350112E+03 1.251217E+02 3.091186E+01 1.4446 1.350892E+03 1.243421E+02 1.293212E+03
1.250000E-01 1.388968E+03 1.326150E+02 2.820226E+01 1.2853 1.389601E+03 1.319822E+02 1.328536E+03
4 -1.250000E-01 1.220277E+03 1.130892E+02 4.195344E+01 2.1669 1.221865E+03 1.115018E+02 1.170105E+03
1.250000E-01 1.255396E+03 1.198619E+02 3.827598E+01 1.9284 1.256685E+03 1.185732E+02 1.201794E+03
...
...
2589 -1.250000E-01 2.738482E-01 2.450364E+00 7.298062E-01 73.0767 2.672420E+00 5.179233E-02 2.646904E+00
1.250000E-01 2.116490E-01 2.492312E+00 7.265424E-01 73.7487 2.704097E+00 -1.363183E-04 2.704165E+00
2590 -1.250000E-01 3.029849E-01 2.711076E+00 5.377296E-01 77.9672 2.825696E+00 1.883649E-01 2.736380E+00
1.250000E-01 2.341678E-01 2.757487E+00 5.353248E-01 78.5042 2.866359E+00 1.252956E-01 2.805810E+00
2591 -1.250000E-01 3.229155E-01 2.889412E+00 3.293143E-01 82.8035 2.930994E+00 2.813340E-01 2.800944E+00
1.250000E-01 2.495715E-01 2.938877E+00 3.278416E-01 83.1490 2.978266E+00 2.101827E-01 2.878934E+00
2592 -1.250000E-01 3.330345E-01 2.979956E+00 1.108930E-01 87.6052 2.984594E+00 3.283968E-01 2.834698E+00
1.250000E-01 2.573921E-01 3.030970E+00 1.103970E-01 87.7243 3.035357E+00 2.530049E-01 2.917095E+00
Following fringe plots shows the comparison between transient analysis and shock spectrum analysis in
Figure 9-4 and Figure 9-5. The maximum (envelope) response across all time steps is plotted using
Main Index
CHAPTER 9 367
Response Spectrum Analysis
MSC.Patran tool, MSC.Explore. Left side of the plot shows the maximum response from transient
analysis and right side shows the response computed using shock spectrum analysis.
Main Index
368 MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
Main Index
CHAPTER 9 369
Response Spectrum Analysis
Model
$
$ Cantilevered Beam Made of Plates Model
$
$ 201 202 203 204 205 206 207 208 209 210 211
$ Y *----*----*----*----*----*----*----*----*----*----*
$ ^ /| | | | | | | | | | |
$ | 99 * | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 |
Main Index
370 MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
$ | \| | | | | | | | | | |
$ +--->X *----*----*----*----*----*----*----*----*----*----*
$ 101 102 103 104 105 106 107 108 109 110 111
$
The spectrum is applied at grid 99.
The spectrum are input in Nastran using DTI, SPECSEL entry.
First entry is always
DTI, SPECSEL, 0
Let’s define actual spectrum input. This is done using pair of DTI, SPECSEL and TABLED1 Bulk Data
entry. Let’s give numerical IDs to spectrum curve.
Define set ID (or Record ID) of 71 for X direction Input Curves
And finally, set ID (or Record ID) of 73 for Z direction Input Curves
Nastran Input
X Direction:
DTI, SPECSEL, 71, , A, 101, 0.02, 102, 0.05
+, 103, 0.07
Tabled1, 101
+, 10.0, 2.0, 20.0, 2.0, 100.0, 8.0, 900.0, 8.0
+, 1100.0, 1.0, 2000.0, 1.0, endt
Tabled1, 102
+, 12.0, 1.5, 15.0, 1.5, 90.0, 4.0, 1200.0, 4.0
+, 1300.0, 0.7, 2500., 0.7, endt
Tabled1, 103
+, 5.0, 1.0, 25.0, 1.0, 110.0, 3.0, 850.0, 3.0
Main Index
CHAPTER 9 371
Response Spectrum Analysis
Note: Each SPECSEL may have different number of Tabled1 for different damping values.
Main Index
372 MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
Main Index
CHAPTER 9 373
Response Spectrum Analysis
param,scrspec,0
SUBCASE 1
subtitle = Excitation in Y Direction : Option : SRSS
Param,Option,SRSS
DLOAD = 701
SUBCASE 2
subtitle = Excitation in X + Z Direction : Option : ABS
Param, Option, ABS
DLOAD = 702
SUBCASE 3
subtitle = Combined X+Y+Z Direction : Option : NRL
Param,Option,NRL
Param,Close,0.1
DLOAD = 703
BEGIN BULK
$
$ Cantilevered Beam Made of Plates Model
$
$ 201 202 203 204 205 206 207 208 209 210 211
$ Y *----*----*----*----*----*----*----*----*----*----*
$ ^ /| | | | | | | | | | |
$ | 99 * | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 |
$ | \| | | | | | | | | | |
$ +--->X *----*----*----*----*----*----*----*----*----*----*
$ 101 102 103 104 105 106 107 108 109 110 111
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......0
$
param,wtmass,.002588
param,autospc,yes
param,grdpnt,0
param,post,-1
$
$ param,scrspec,0 – Selected in case control above all subcases
$ param,option,srss - Selected in case control under each subcase
$
$ Must assign large mass at point of excitation and must have a suport
$ entry with all excitation direction DOF.
$
CONM2,199,99,,1.e8
SUPORT,99,123
SPC1,77,456,99
EIGRL, 66, -.1, 5000.
Main Index
374 MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
$
$
DLOAD, 701, 386.4, 0.0, 71, 1.0, 72, 0.0, 73
DLOAD, 702, 386.4, 1.0, 71, 0.0, 72, 1.0, 73
DLOAD, 703, 386.4, 1.0, 71, 1.0, 72, 1.0, 73
$
$
$ Input Spectrum in direction X
$
DTI, SPECSEL, 71, A, 101, 0.02, 102, 0.05
+, 103, 0.07
Tabled1, 101
+, 10.0, 2.0, 20.0, 2.0, 100.0, 8.0, 900.0, 8.0
+, 1100.0, 1.0, 2000.0, 1.0, endt
Tabled1, 102
+, 12.0, 1.5, 15.0, 1.5, 90.0, 4.0, 1200.0, 4.0
+, 1300.0, 0.7, 2500.,0 0.7, endt
Tabled1, 103
+, 5.0, 1.0, 25.0, 1.0, 110.0, 3.0, 850.0, 3.0
+, 1200.0, 0.5, 3000.0, 0.5, endt
$
$ Input Spectrum in direction Y
$
DTI, SPECSEL, 72, , A, 201, 0.01, 202, 0.04
+, 203, 0.07
Tabled1, 201
+, 10.0, 1.0, 20.0, 1.0, 100.0, 6.0, 900.0, 6.0
+, 1100.0, 1.5, 2000.0, 1.5, endt
Tabled1, 202
+, 12.0, 0.8, 15.0, 0.8, 90.0, 5.0, 1200.0, 5.0
+, 1300.0, 0.9, 2500.0, 0.9, endt
Tabled1, 203
+, 5.0, 0.4, 25.0, 0.4, 110.0, 3.0, 850.0, 3.0
+, 1200.0, 0.5, 3000.0, 0.5, endt
$
$ Input Spectrum in direction Z
$
DTI, SPECSEL, 73, , A, 301, 0.01, 302, 0.03
+, 303, 0.07
Tabled1, 301
+, 10.0, 1.5, 20.0, 1.5, 100.0, 7.0, 900.0, 7.0
+, 1100.0, 1.8, 2000.0, 1.8, endt
Tabled1, 302
+, 12.0, 1.0, 15.0, 1.0, 90.0, 5.0, 1200.0, 5.0
+, 1300.0, 1.4, 2500.0, 1.4, endt
Tabled1, 303
+, 5.0, 0.6, 25.0, 0.6, 110.0, 3.5, 850.0, 3.5
+, 1200.0, 0.7, 3000.0, 0.7, endt
$
$ Select Modal Damping
$
TABDMP1, 88, Crit
+, 0.0, .03, 100.0, .05, 1500., 0.05, 2000., .02
+, 5000., 0.02, Endt
$
GRID 101 0. 0. 0.
GRID 102 1. 0. 0.
GRID 103 2. 0. 0.
GRID 104 3. 0. 0.
Main Index
CHAPTER 9 375
Response Spectrum Analysis
GRID 105 4. 0. 0.
GRID 106 5. 0. 0.
GRID 107 6. 0. 0.
GRID 108 7. 0. 0.
GRID 109 8. 0. 0.
GRID 110 9. 0. 0.
GRID 111 10. 0. 0.
GRID 201 0. 1. 0.
GRID 202 1. 1. 0.
GRID 203 2. 1. 0.
GRID 204 3. 1. 0.
GRID 205 4. 1. 0.
GRID 206 5. 1. 0.
GRID 207 6. 1. 0.
GRID 208 7. 1. 0.
GRID 209 8. 1. 0.
GRID 210 9. 1. 0.
GRID 211 10. 1. 0.
$
CQUAD4 1 1 101 102 202 201
CQUAD4 2 1 102 103 203 202
CQUAD4 3 1 103 104 204 203
CQUAD4 4 1 104 105 205 204
CQUAD4 5 1 105 106 206 205
CQUAD4 6 1 106 107 207 206
CQUAD4 7 1 107 108 208 207
CQUAD4 8 1 108 109 209 208
CQUAD4 9 1 109 110 210 209
CQUAD4 10 1 110 111 211 210
$
Grid, 99, , 0.0, 0.5, 0.0
rbar, 77, 99, 101, 123456
rbar, 78, 99, 201, 123456
$
pshell 1 1 .1 1
mat1 1 10.e6 .3 0.1 1.e-6 0.
$
enddata
Main Index
376 MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 10: Restarts In Dynamic Analysis
Overview
Examples
Remarks
Main Index
378 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
A restart is a logical way of continuing from a previous run without having to start from the beginning.
Restarts (Ch. 14) in the MD Nastran Linear Static Analysis User’s Guide gives the overview of restarts.
In dynamic analysis, the calculation of normal modes is, in general, the most expensive operation.
Therefore, a common application of restart is the performance of a transient or frequency response
analysis by restarting from the normal modes calculation, which was saved in the database from a
previous run. This restart process avoids the recalculation of the normal modes.
Executive Control Section
This section is the same as your normal run with the exception of perhaps the “SOL x” command. For
example, if you are performing a normal modes analysis in run1.dat, then the “SOL x” command in
run1.dat should reference “SOL 103”. In run2.dat, if you are performing a modal transient restart from
run1.dat, then the “SOL x” statement in this case should then reference “SOL 112”.
Case Control Section
The automatic restart logic compares the modifications made to the Case Control and Bulk Data Sections
in determining which operations need to be processed upon restart. Therefore, you must be very careful
with the changes that you make in your restart run. Adhering to the following rules will avoid
unnecessary reprocessing of previously completed operations.
• You must include all “solution-type” related Case Control commands, which are unchanged as
compared to the cold start run, in your restart run. In other words, do not make unnecessary
LOAD, SPC, MPC, or METHOD command changes or remove them from the Case Control
Section unless these are actual changes. This process is clarified later with the example
problems.
• Output requests can be modified. A typical example can be a request of the eigenvector printout
which was not requested in the cold start run.
Bulk Data Section
As mentioned in the previous section, the automatic restart logic compares the changes made in the Bulk
Data Section and determines the path that it follows. A copy of the Bulk Data is stored for each version.
The restart run must not contain any Bulk Data entry that was included in the previous runs and saved in
the database. The Bulk Data Section in the current restart run should contain only new entries, changed
entries, and/or the deletion of old entries from the database. This philosophy is slightly different than the
one used in the Case Control Section.
For conventional dynamic analysis (i.e., non-superelement), restarts involving model changes (e.g.,
changing the thickness of a plate) are not very efficient. Therefore, the savings is probably minimal, if
any. However, in the case of additional output requests or a restart from a modes run to a response run,
the savings can be substantial. This type of restart is covered extensively in Examples, 379. For
superelement analysis, even restarts involving model changes can be beneficial as long as these changes
are localized.
Main Index
CHAPTER 10 379
Restarts In Dynamic Analysis
Examples
The examples perform a typical series of runs starting from a normal modes run and restarting into
transient and frequency response analyses. Table 10-1 summarizes this series of nine runs along with a
brief description. Listings of the ten runs are also included (Listing 10-1 through Listing 10-10). See
Install_dir/mdxxxx/doc/dynamics/bd10bar1.dat through bd10bar10.dat. The way the runs are setup they
must all be copied to your local directory to run correctly.
Main Index
380 MD Nastran Dynamic Analysis User’s Guide
Examples
Main Index
CHAPTER 10 381
Restarts In Dynamic Analysis
Remarks
If the results for run number 1 are not going to be used for any future purposes, then you may consider
making run number 3 as a cold start run instead of a restart run. Model changes do not save you much
time, if any, in a non-superelement analysis. By making run 3 a cold start run, you reduce the total
amount of disk space required. In this case, run number 4 is not necessary since you are starting with a
new database. However, if you want to keep both physical models in the database, then run number 3
should be a restart run as shown in this example. An application of this can be a parametric study of two
different configurations. This type of restart allows you to make efficient data recovery or response
analysis from two different physical models. However, this type of restart is not used often in a
non-superelement analysis since, in general, it is not very efficient. However, in a superelement analysis
(see Superelement Analysis (Ch. 16)), this type of restart can still be very efficient since the changes can
be localized to a small region.
Main Index
382 MD Nastran Dynamic Analysis User’s Guide
Remarks
$
ENDDATA
Listing 10-2 Input File for Requesting Eigenvectors
$ FILE - bd10bar2.dat
$
$ NORMAL MODES RUN
$ REQUEST EIGENVECTOR PRINTOUTS FROM PREVIOUS RUN
$
RESTART VERSION=1,KEEP
ASSIGN MASTER=’bd10bar1.MASTER’
$
ID CANT BEAM
SOL 103
TIME 10
CEND
TITLE = EIGENVECTORS DATA RECOVERY RESTART RUN
SPC = 1
METHOD = 10
DISP = ALL $ PRINT EIGENVECTORS
$
BEGIN BULK
$
ENDDATA
Listing 10-3 Input File for Modifying a Bar Element
$
$ FILE - bd10bar3.dat
$
$ NORMAL MODES RUN
$ MODIFY PBAR
$
RESTART
ASSIGN MASTER='bd10bar1.MASTER'
$
ID CANT BEAM
SOL 103
CEND
TITLE = CANTILEVER BEAM - NORMAL MODES - RESTART RUN
SPC = 1
METHOD = 10
DISP = ALL
$
BEGIN BULK
$
$ DELETE OLD PBAR ENTRY
$
/DELETE PBAR
$ NOT NEEDED IF JUST UPDATING AN ENTRY
$
$ ADD NEW PBAR ENTRY
$
PBAR,1,1,6.158-4,2.9-8,3.1-8,6.-8,2.414
$
ENDDATA
Main Index
CHAPTER 10 383
Restarts In Dynamic Analysis
Main Index
384 MD Nastran Dynamic Analysis User’s Guide
Remarks
Main Index
CHAPTER 10 385
Restarts In Dynamic Analysis
SUBCASE 1
SDAMP = 100
TSTEP = 100
DLOAD = 300
$
$ PLOT RESULTS
$
$...X-Y plot commands ...
$
BEGIN BULK
$
$ SID DAREA DELAY TYPE TID
TLOAD1 300 301 302
$
DAREA,301,11,3,1.0
$
TABLED1,302,,,,,,,,+TBL3
+TBL3,0.0,0.0,.001,1.0,.20,0.0,10.0,0.0,+TBL4
+TBL4,ENDT
$
ENDDATA
Listing 10-8 Input File for Frequency Response Analysis
$
$ FILE - bd10bar8.dat
$
$ THIS IS READ ONLY RESTART RUN TO PERFORM FREQUENCY RESPONSE
$ FROM THE MODES RUN
$
RESTART version=3 logical=bd10bar8
ASSIGN bd10bar8='bd10bar1.MASTER'
ID CANT BEAM
SOL 111
CEND
TITLE = CANTILEVER BEAM - FREQUENCY RESPONSE RESTART
SPC = 1
METHOD = 10
SET 1 = 11
DISP(PHASE) = 1
SUBCASE 1
$
$ A TWO PERCENT CRITICAL DAMPING IS APPLIED TO THIS RUN
$ AS OPPOSED TO ONE PERCENT CRITICAL DAMPING IN THE
$ TRANSIENT ANALYSIS.
$
SDAMP = 1000
DLOAD = 1000
FREQ = 1000
$
$ PLOT RESULTS
$
OUTPUT(XYOUT)
XTGRID LINES=YES
YTGRID LINES=YES
XBGRID LINES = YES
YBGRID LINES = YES
XGRID = YES
YGRID = YES
XLOG = YES
Main Index
386 MD Nastran Dynamic Analysis User’s Guide
Remarks
YBLOG = YES
$XMIN = 1.0
$XMAX = 100.
YTTITLE = D I S P / M A G ( M )
YBTITLE = D I S P / P H A S E ( D E G )
XTITLE = F R E Q U E N C I E S ( H Z )
XYPLOT DISP /11(T3RM,T3IP)
$
BEGIN BULK
$
$ ADDITIONAL ENTRIES FOR FREQUENCY RESPONSE
$
$ SID DAREA M N TC TD
$
RLOAD1 1000 1001 1002
$
DAREA,1001,11,3,0.1
$
TABLED1,1002
,0.,1.,200.,1.,ENDT
$
$ FORCING FREQUENCIES
$
$ RESONANT FREQUENCIES
$
FREQ,1000,2.03174,2.100632,12.59101,13.01795
FREQ,1000,34.90217,36.08563
$
$ SPREAD THROUGHOUT FREQUENCY RANGE OF INTEREST
$ WITH BIASED BETWEEN HALF POWER POINTS
$
FREQ,1000,1.437,1.556,1.675,1.794,1.913
FREQ,1000,2.046,2.059,2.073,2.087
FREQ,1000,2.224,2.347,2.47,2.593,2.716
FREQ,1000,8.903,9.641,10.378,11.116,11.853
FREQ,1000,12.676,12.762,12.847,12.933
FREQ,1000,13.781,14.543,15.306,16.068,16.831
FREQ,1000,24.680,26.724,28.769,30.813,32.858
FREQ,1000,35.139,35.376,35.612,35.849
FREQ,1000,41.189,46.292,51.395,56.499,61.602
$
FREQ1,1000,0.,.5,200
$
$ DAMPING
$
TABDMP1,1000,CRIT,,,,,,,+DAMP
+DAMP,0.,.02,200.,.02,ENDT
$
ENDDATA
Main Index
CHAPTER 10 387
Restarts In Dynamic Analysis
Main Index
388 MD Nastran Dynamic Analysis User’s Guide
Remarks
POINT-ID = 11
D I S P L A C E M E N T V E C T O R
TIME TYPE T1 T2 T3 R1 R2 R3
0.0 G 0.0 0.0 0.0 0.0 0.0 0.0
5.000000E-03 G 1.194303E-16 3.936567E-17 2.504727E-05 0.0 -6.120401E-05 4.001821E-16
1.000000E-02 G 2.737714E-16 6.817522E-17 8.184779E-05 0.0 -1.331090E-04 8.292225E-16
SUBCASE 1
X Y - O U T P U T S U M M A R Y ( R E S P O N S E )
SUBCASE CURVE FRAME CURVE ID./ XMIN-FRAME/ XMAX-FRAME/ YMIN-FRAME/ X FOR YMAX-FRAME/ X FOR
ID TYPE NO. PANEL : GRID ID ALL DATA ALL DATA ALL DATA YMIN ALL DATA YMAX
1 DISP 1 11( 5) 6.000000E-01 1.300000E+00 6.125921E-04 9.450001E-01 7.731026E-03 7.200000E-01
6.000000E-01 1.300000E+00 6.125921E-04 9.450001E-01 7.731026E-03 7.200000E-01
1 ACCE 2 10( 5) 6.000000E-01 1.300000E+00 -7.251530E-01 7.300000E-01 6.462884E-01 9.200000E-01
6.000000E-01 1.300000E+00 -7.251530E-01 7.300000E-01 6.462884E-01 9.200000E-01
Figure 10-1 Partial Output from Transient Analysis with Unit Step Function Input (bd10bar5
and bd10bar10)
POINT-ID = 11
D I S P L A C E M E N T V E C T O R
TIME TYPE T1 T2 T3 R1 R2 R3
0.0 G 0.0 0.0 0.0 0.0 0.0 0.0
5.000000E-03 G 1.187877E-16 3.927214E-17 2.489746E-05 0.0 -6.077061E-05 3.977257E-16
1.000000E-02 G 2.680967E-16 6.663936E-17 8.054515E-05 0.0 -1.304303E-04 8.105531E-16
X Y - O U T P U T S U M M A R Y ( R E S P O N S E )
SUBCASE CURVE FRAME CURVE ID./ XMIN-FRAME/ XMAX-FRAME/ YMIN-FRAME/ X FOR YMAX-FRAME/ X FOR
ID TYPE NO. PANEL : GRID ID ALL DATA ALL DATA ALL DATA YMIN ALL DATA YMAX
1 DISP 1 11( 5) 0.000000E+00 6.000000E-01 -4.039584E-03 4.200000E-01 4.327041E-03 1.850000E-01
0.000000E+00 6.000000E-01 -4.039584E-03 4.200000E-01 4.327041E-03 1.850000E-01
Figure 10-2 Partial Output from Transient Analysis with a Triangular Pulse (bd10bar9)
Main Index
CHAPTER 10 389
Restarts In Dynamic Analysis
POINT-ID = 11
C O M P L E X D I S P L A C E M E N T V E C T O R
(MAGNITUDE/PHASE)
FREQUENCY TYPE T1 T2 T3 R1 R2 R3
0.0 G 1.906914E-15 5.344224E-16 4.089051E-04 0.0 2.044525E-04 5.763767E-15
180.0000 180.0000 0.0 0.0 180.0000 180.0000
5.000000E-01 G 2.017782E-15 5.676515E-16 4.327695E-04 0.0 2.154499E-04 6.100925E-15
179.4355 179.4089 359.4349 0.0 179.4725 179.4328
2.073000E+00 G 3.901148E-14 8.436090E-15 8.402392E-03 0.0 3.875251E-03 1.167835E-13
123.5868 98.5582 303.5746 0.0 123.7787 122.7733
2.087000E+00 G 4.417046E-14 9.296475E-15 9.514233E-03 0.0 4.382797E-03 1.323076E-13
108.1114 89.8882 288.0990 0.0 108.3053 107.5244
2.100632E+00 G 4.612975E-14 9.812769E-15 9.936970E-03 0.0 4.572159E-03 1.383304E-13
90.0803 76.7938 270.0676 0.0 90.2761 89.6427
2.224000E+00 G 1.434210E-14 3.595956E-15 3.091593E-03 0.0 1.406759E-03 4.331211E-14
19.3916 17.3197 199.3766 0.0 19.6055 19.3040
2.347000E+00 G 7.249022E-15 1.926738E-15 1.563781E-03 0.0 7.030737E-04 2.195193E-14
10.2994 9.4297 190.2818 0.0 10.5329 10.2551
X Y - O U T P U T S U M M A R Y ( R E S P O N S E )
SUBCASE CURVE FRAME CURVE ID./ XMIN-FRAME/ XMAX-FRAME/ YMIN-FRAME/ X FOR YMAX-FRAME/ X FOR
ID TYPE NO. PANEL : GRID ID ALL DATA ALL DATA ALL DATA YMIN ALL DATA YMAX
1 DISP 1 11( 5,--) 1.000000E-01 1.000000E+02 2.536850E-07 6.250000E+01 9.936970E-03 2.100632E+00
0.000000E+00 1.000000E+02 2.536850E-07 6.250000E+01 9.936970E-03 2.100632E+00
1 DISP 1 11(--, 11) 1.000000E-01 1.000000E+02 1.815511E+02 5.500000E+00 3.594349E+02 5.000000E-01
0.000000E+00 1.000000E+02 0.000000E+00 0.000000E+00 3.594349E+02 5.000000E-01
N A S T R A N F I L E A N D S Y S T E M P A R A M E T E R E C H O
P R O J E C T V E R S I O N D I R E C T O R Y P R I N T
Main Index
390 MD Nastran Dynamic Analysis User’s Guide
Remarks
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 11: Coupled Fluid-Structure Analysis
Overview
Axisymmetric Hydroelastic Analysis
Virtual Fluid Mass
Coupled and Uncoupled Acoustic Analysis
Exterior Acoustics
BW Method of Fluid-Structure Analysis
Direct Input of Interface [A] Matrix
Acoustic Modal Participation Factors
Acoustic XY Plots, Random, and Restarts
Acoustic Source Change
Rigid Porous Absorber - MAT10
Main Index
392 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
Four major methods are available in MD Nastran to model fluid effects. These methods are described
below.
Main Index
CHAPTER 11 393
Coupled Fluid-Structure Analysis
Several methods are available in MD Nastran for the analysis of normal modes of compressible fluids
bounded by rigid containers and/or free surfaces. One method is the “acoustic cavity” capability, which
uses two-dimensional slot elements and axisymmetric ring elements to define the fluid region. This
method was specifically developed for the acoustic analysis of solid rocket motor cavities. A better
method is to use the three-dimensional fluid elements for the “coupled acoustics” described above and
provide the appropriate boundary conditions.
External Acoustics
With exterior acoustic problems, the acoustic domain is unbounded. The analysis studies the sound
pressure in the vicinity of the vibrating body or far away from the body. A further important result is the
acoustic power radiated from the vibrating structure. A typical example is the determination of the
radiated acoustic power of an engine.
Main Index
394 MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
Overview
MD Nastran’s axisymmetric hydroelastic capability allows the user to solve a variety of fluid problems
with small motion, compressibility, and gravity effects. A complete derivation of the MD Nastran model
and an explanation of the assumptions are given in The NASTRAN Theoretical Manual, Section 16.1.
The input data and the solution logic have many similarities to those for a structural model. The standard
normal modes analysis, transient analysis, complex eigenvalue analysis, and frequency response
solutions are available with minor restrictions. The differences between a fluid model and an ordinary
structural model are due to the physical properties of a fluid. The characteristics of an axisymmetric fluid
model are summarized as follows:
1. The independent degrees-of-freedom for a fluid are the Fourier coefficients of the pressure
function (i.e., harmonic pressure) at fluid points in a cylindrical or spherical coordinate system.
The locations of these points are on the r-z plane.
2. Much like the structural model, the fluid data will produce “stiffness” and “mass” matrices.
Because they now relate pressures and flow instead of displacement and force, their physical
meaning is quite different. The user may not apply loads, constraints, sequencing, or omitted
coordinates directly on the fluid points involved. Instead, the user supplies information related to
the boundaries and MD Nastran internally generates the correct constraints, sequencing, and
matrix terms. Indirect methods, however, are available to the user for utilizing the internally
generated points as normal grid or scalar points. See Hydroelastic Data Processing, 400 for the
identification code.
3. When a physical structure is to be connected to the fluid, the user supplies a list of fluid points on
a plane and a related list of special structural grid points on the circles corresponding to the fluid
parts. Unsymmetric matrix terms define the actual physical relations. A special provision is
included in MD Nastran in the event that the structure has planes of symmetry. The user may, if
so desired, define only a section of the boundary and solve the problem with symmetric or
antisymmetric constraints. The fluid-structure interface will take the missing sections of
structural boundary into account.
4. Because of the special nature of fluid problems, some user convenience options are absent. The
fluid elements and harmonic pressures are not available in the structural plots. Plotting the
harmonic pressures versus frequency or time may not be directly requested. Because mass matrix
terms are automatically generated if compressibility or free surface effects are present, the weight
and C.G. calculations with fluid elements present may not be correct and should be avoided. Also,
the inertia relief rigid format uses the mass matrix to produce internal loads, and if fluids are
included, these special fluid terms in the mass matrix may produce erroneous results.
5. The general acoustic capability in MD Nastran, described in Coupled and Uncoupled Acoustic
Analysis, 414, shares the same small-motion pressure formulation and also produces unsymmetric
matrices. For more generality, the acoustic fluid model is defined by three-dimensional HEXA
and PENTA elements and the structural boundary is defined with conventional GRID points.
Although it lacks an option for gravity waves on the free surface, it can replace the axisymmetric
hydroelastic modeling for most problems.
Main Index
CHAPTER 11 395
Coupled Fluid-Structure Analysis
In spite of the numerous differences between a structural model and a fluid model, the similarities allow
the user to formulate a problem with a minimum of data preparation and obtain efficient solutions to
large-order problems. The similarities of the fluid model to the structural model are as follows:
1. The fluid is described by points in space and finite element connections. The locations of the
axisymmetric fluid points are described by rings (RINGFL) about a polar axis and defined by
their r-z coordinates. The rings are connected by elements (CFLUIDi) which have the properties
of density and bulk modulus of compressibility. Each fluid ring produces, internally, a series of
scalar points p n and p n* (i.e., harmonic pressures), describing the pressure function, P , in the
equation
N N
P = P0 + p n cos n + p n * sin n 0 N 100
n = 1 n = 1
where the set of harmonics 0 , n and n* are selected by the user. If the user desires the output
of pressure at specific points on the circular ring, he may specify them as pressure points
(PRESPT) by giving a point number and an angle on a specified fluid ring. The output data will
have the values of pressure at the angle given in the above equation. The output of free surface
displacements normal to the surface (FREEPT) are also available at specified angles, . The
Case Control command AXISYM = FLUID is necessary when any harmonic fluid
degrees-of-freedom are included. This command may not be used when F = NONE on the AXIF
Bulk Data entry (no harmonics specified).
2. The input file may include all existing options except the axisymmetric structural element data.
All existing Case Control options may be included with some additional fluid Case Control
requests. All structural element and constraint data may be used but not connected to RlNGFL,
PRESPT, or FREEPT fluid points. The structure-fluid boundary is defined with the aid of special
grid points (GRIDB) that may be used for any purpose that a structural grid point is currently
used.
3. The output data options for the structural part of a hydroelastic model are unchanged from the
existing options. The output values of the fluid will be produced in the same form as the
displacement vectors but with format modifications for the harmonic data.
• Pressures and free surface displacements, and their velocities and accelerations, may be
printed with the same request (the Case Control command PRESSURE = SET is equivalent
to DlSP = SET) as structural displacements, velocities, and accelerations.
• Structural plots are restricted to GRID and GRIDB points and any elements connected to
them.
• X-Y plot and Random Analysis capabilities are available for FREEPT and PRESPT points if
they are treated as scalar points.
• The RINGFL point identification numbers may not be used in any plot request; instead, the
special internally generated points used for harmonics may be requested in the X-Y plots and
random analysis. (See Hydroelastic Data Processing, 400 for the identification number code.)
• No element stress or force data is produced for the fluid elements.
Main Index
396 MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
• As in the case of the axisymmetric conical shell problem, the Case Control command
HARMONICS = N is used to select output data up to the Nth harmonic.
Input Data
Several special Bulk Data entries are required for fluid analysis problems. These entries are compatible
with structural entries. A brief description of the uses for each Bulk Data entry follows:
AXIF
AXIF controls the formulation of the axisymmetric fluid problem. It is a required entry if any of the
subsequent fluid-related entries are present. The data references a fluid-related coordinate system to
define the axis of symmetry. The gravity parameter is included on this entry rather than on the GRAV
entry because the direction of gravity must be parallel to the axis of symmetry. The values of density and
elastic bulk modulus are conveniences in the event that these properties are constant throughout the fluid.
A list of harmonics and the request for the nonsymmetric (sine) coefficients are included on this entry to
allow the user to select any of the harmonics without producing extra matrix terms for the missing
harmonics. A change in this list, however, will require a RESTART at the beginning of the problem.
RINGFL
The geometry of the fluid model about the axis of symmetry is defined with RINGFL entries. The
RINGFL data entries serve somewhat the same function for the fluid as the GRID entries serve in the
structural model. In fact, each RINGFL entry will produce, internally, a special grid point for each of the
various harmonics selected on the AXIF data entry. They may not, however, be connected directly to
structural elements (see the GRIDB and BDYLlST entries). No constraints may be applied directly to
RINGFL fluid points.
CFLUIDi
CFLUIDi defines a volume of fluid bounded by the referenced RlNGFL points. The volume is called an
element and logically serves the same purpose as a structural finite element. The physical properties
(density and bulk modulus) of the fluid element may be defined on this entry if they are variables with
respect to the geometry. If a property is not defined, the default value on the AXlF entry is assumed.
Two connected circles (RINGFL) must be used to define fluid elements adjacent to the axis of symmetry.
A choice of three or four points is available in the remainder of the fluid.
GRIDB
GRIDB provides the same functions as the GRID entry for the definition of structural grid points. It will
be attached to a particular RINGFL fluid point. The particular purpose for this entry is to force the user
to place structural boundary points in exactly the same locations as the fluid points on the boundary. The
format of GRIDB is identical to the format of GRID except that one additional field is used to identify
the RINGFL point. The GRDSET entry, however, is not used for GRIDB data, and no superelement
partitioning is allowed.
GRIDB entries may be used without a fluid model. This is convenient in case the user wishes to solve
the structural problem first and to add the fluid effects later without converting GRID entries to GRIDB
entries. The referenced RINGFL point must still be included in a boundary list (BDYLIST; see below),
Main Index
CHAPTER 11 397
Coupled Fluid-Structure Analysis
and the AXIF entry must always be present when GRIDB entries are used. (The fluid effects are
eliminated by specifying no harmonics.)
FREEPT, PRESPT
FREEPT and PRESPT are used to define points on a free surface for displacement output and points in
the fluid for pressure output. No constraints may be applied to these points. Scalar elements and direct
matrix input data may be connected to these points, but the physical meaning of the elements will be
different from in the structural sense.
FSLIST, BDYLIST
FSLIST and BDYLIST define the boundaries of the fluid with a complete freedom of choice. The
FSLIST entry defines a list of fluid points which lie on a free surface. The BDYLIST data make up a
list of fluid points to which structural GRIDB points are connected. Points on the boundary of the fluid
for which BDYLlST or FSLIST data are not defined are assumed to be rigidly restrained from motion in
a direction normal to the surface.
With both of these lists, the sequence of the listed points determines the nature of the boundary. The
following directions will aid the user in producing a list:
1. Draw the z-axis upward and the r-axis to the right. Plot the locations of the fluid points on the
right-hand side of z.
2. If one imagines himself traveling along the free surface or boundary with the fluid on his right
side, the sequence of points encountered is used for the list. If the surface or boundary touches
the axis, the word AXIS is placed in the list. AXIS may be used only for the first and/or last point
in the list.
3. The free surface must be consistent with static equilibrium. With no gravity field, any free
surface consistent with axial symmetry is allowed. With gravity, the free surface must be a plane
perpendicular to the z-axis of the fluid coordinate system.
4. Multiple free surface lists and boundary lists are allowed. A fluid point may be included in any
number of lists.
Figure 11-1 illustrates a typical application of the free surface and structural boundary lists.
Main Index
398 MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
Axis of
Symmetry
FLUID
21
14
7
1 2 3 4 5 6
FLSYM
FLSYM allows the user an option to model a portion of the structure with planes of symmetry containing
the polar axis of the fluid. The first plane of symmetry is assumed at = 0.0 and the second plane of
symmetry is assumed at = 360 M where M is an integer specified on the entry. Also specified
are the types of symmetry for each plane, symmetric (S) or antisymmetric (A). The user must also supply
the relevant constraint data for the structure. The solution is performed correctly only for those harmonic
coefficients that are compatible with the symmetry conditions, as illustrated in the following example for
quarter symmetry, M = 4.
Main Index
CHAPTER 11 399
Coupled Fluid-Structure Analysis
Plane 2
Series Plane 1 S A
Cosine S 0,2,4, 1,3,5,
A none none
Sine S none none
(*) A 1,3,5, 2,4,6,
DMIAX
DMIAX is used for special purposes such as the specification of surface friction effects. DMIAX is
equivalent to DMIG except harmonic numbers are specified for the degrees-of-freedom. A matrix may
be defined with either DMIG or DMlAX entries, but not with both.
Solution Sequences
Restrictions
The characteristics of the fluid analysis problems which cause restrictions on the type of solution are as
follows:
1. The fluid-structure interface is mathematically described by a set of unsymmetric matrices. Since
most solution sequences are restricted to the use of symmetric matrices, the fluid-structure
boundary is ignored. Thus, for any of these solution sequences, MD Nastran solves the problem
for a fluid in a rigid container with an optional free surface and an uncoupled elastic structure with
no fluid present.
2. No means are provided for the direct input of applied loads on the fluid. The only direct means
of exciting the fluid is through the structure-fluid boundary. The fluid problem may be
formulated in any solution sequence. However, only some will provide nontrivial solutions.
The suggested solution sequences for the axisymmetric fluid and the restrictions on each are described
as follows:
SOL 103 or 200 – Normal Modes Analysis
The modes of a fluid in a rigid container may be extracted with a conventional solution request. Free
surface effects with or without gravity may be accounted for. Any structure data in the section will be
treated as a disjoint problem. (The structure may also produce normal modes.) Normalization of the
eigenvectors using the POINT option will cause a fatal error.
SOL 107 – Direct Complex Eigenvalue Analysis
If no damping or direct input matrices are added, the resulting complex roots will be purely imaginary
numbers whose values are the natural frequencies of the system. The mode shape of the combination
may be normalized to the maximum quantity (harmonic pressure or structural displacement) or to a
specified structural point displacement.
Main Index
400 MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
This solution may be used directly if the loads are applied only to the structural points. The use of overall
structural damping (parameter g) is not recommended since the fluid matrices will be affected
incorrectly.
SOL 109 – Direct Transient Response
Transient analysis may be performed directly on the fluid-structure system if the following rules apply:
1. Applied loads and initial conditions are given only to the structural points.
2. All quantities are measured relative to static equilibrium. The initial values of the pressures are
assumed to be in equilibrium.
3. Overall structural damping (parameters 3 and g ) must not be used.
SOLs 110, 111, 112, and 200 – Modal Formulation
Although these solution sequences may be used in a fluid dynamics problem, their practicality is limited.
The modal coordinates used to formulate the dynamic matrices will be the normal modes of both the fluid
and the structure solved as uncoupled systems. Even though the range of natural frequencies would be
typically very different for the fluid than for the structure, the program will select both sets of modes from
a given fixed frequency range. The safest method with the present system is the extraction of all modes
for both systems with the Tridiagonalization Method. This procedure, however, results in a dynamic
system with large full matrices. The Direct Formulation is more efficient in such cases. At present, the
capability for fluid-structure boundary coupling is not provided with the modal formulation. However,
the capability may be provided by means of an alter using the same logic as in the direct formulations.
where:
IN = N + 1 cosine series
IN = N + 1 2 sine series
Main Index
CHAPTER 11 401
Coupled Fluid-Structure Analysis
For example, if the user requested all harmonics from zero to two, including the sine series, each
RINGFL entry will produce five special grid entries internally. If the user’s identification number (in
Field 2 of the RINGFL Bulk Data entry) were 37, the internally generated grid points would have the
following identification numbers:
Harmonic ID
0 1,000,037
1* 1,500,037
1 2,000,037
2* 2,500,037
2 3,000,037
These equivalent grid points are resequenced automatically by MD Nastran to be adjacent to the original
RlNGFL identification number. A RINGFL point may not be resequenced by the user.
The output from matrix printout, table printout, and error messages will have the fluid point labeled in
this form. If the user wishes, he may use these numbers as scalar points for Random Analysis, X-Y
plotting, or for any other purpose.
In addition to the multiple sets of points and connection entries, the program may also generate constraint
sets. For example, if a free surface (FSLIST) is specified in a zero-gravity field, the pressures are
constrained to zero. For this case, the internally generated set of single-point constraints are internally
combined with any user-defined structural constraints and will always be automatically selected.
If pressures at points in the fluid (PRESPT) or gravity-dependent normal displacements on the free
surface (FREEPT) are requested, the program will convert them to scalar points and create a set of
multipoint constraints with the scalar points as dependent variables. The constraint set will be internally
combined with any user-defined sets and will be selected automatically.
The PRESPT and FREEPT scalar points may be used as normal scalar points for purposes such as
plotting versus frequency or time. Although the FREEPT values are displacements, scalar elements
connected to them will have a different meaning than in the structural sense.
Main Index
402 MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
1. The AXISYM = FLUID entry is necessary to control the constraint set selections and the output
formats for a fluid problem. It must appear above the subcase level.
2. DISPLACEMENT and PRESSURE Case Control commands are pseudonyms. DISP = ALL
produces all structure displacements, all free surface displacements, and all fluid pressure values
in the output. The HARMONICS control is a limit on the harmonic data and has the same
function as in an axisymmetrical conical shell problem.
3. The AXIF entry defines the existence of a hydroelastic problem. It is used to define overall
parameters and control the harmonic degrees-of-freedom.
4. The RINGFL entries included define the five points on the fluid cross section.
5. The CFLUIDi entries are used to define the volume of the fluid as finite elements connected by
the RINGFL points. Since parameters p and B are missing, the default values on the AXIF entry
are used.
6. The FSLIST entry is used to define the free surface at z = 10.0. The density factor is placed on
the entry in this case. If blank, the default value on the AXIF entry is used.
7. The fluid-structure boundary is defined on the BDYLIST entry. The AXIF default density is
used.
8. The GRIDB entries define the structure points on the fluid boundary. Points 3 through 6 are
connected to fluid number 2 ring. The rotation in the r direction (“4” in field 8) is constrained.
9. The fact that one-quarter symmetry was used for the structure requires the use of the FLSYM
entry. Symmetric-antisymmetric boundaries indicate that only the cosine terms for the odd
harmonics interact with the structure. If symmetric-symmetric boundary conditions were chosen
on FLSYM, only the even harmonics of the cosine series would interact with the structure.
10. The PRESPT entries define locations of pressure points in the fluid for pressure output.
11. The FREEPT entries define locations on the free surface for displacement output.
Main Index
CHAPTER 11 403
Coupled Fluid-Structure Analysis
3. $AXIF,CID,G,DRHO,DB,NOSYM
AXIF,2,32.2,0.03,,NO
$,N1,N2
1,3
CORD2C,2,,0.,0.,0.,0.,0.,1.0
1.0,0.,0.
4. $RINGFL,IDF,Xl,X2,X3,IDF,X1,X2,X3
RINGFL,1,4.0,,10.0,2,8.0,,10.0
RINGFL,7,4.0,,5.0,8,8.0,,5.0
RINGFL,13,4.0,,0.0
5. $CFLUID,EID,IDF1,IDF2,IDF3,IDF4,RHO,B
CFLUID2,101,1,7
CLFUID2,102,7,13
CFLUID3,103,7,8,13
CFLUID4,104,1,2,7,8
6. $FSLIST,RHO,IDF1,IDF2,IDF3
FSLIST,0.03,AXIS,1,2
7. $BDYLIST,RHO,IDF1,IDF2,IDF3
BDYLIST,,2,8,13
8. $GRIDB,ID,,,PHI,,CD,PS,IDF
GRIDB,3,,,0.0,,2,4,2
GRIDB,4,,,30.0,,2,4,2
GRIDB,5,,,60.0,,2,4,2
GRIDB,6,,,90.0,,2,4,2
GRIDB,9,,,0.0,,2,,8
GRIDB,10,,,30.0,,2,,8
GRIDB,11,,,60.0,,2,,8
GRIDB,12,,,90.0,,2,,8
GRIDB,14,,,0.0,,2,,13
GRIDB,15,,,30.0,,2,,13
GRIDB,16,,,60.0,,2,,13
GRIDB,17,,,90.0,,2,,13
CQUAD4,10,11,3,9,10,4
CQUAD4,11,11,4,10,11,5
CQUAD4,12,11,5,11,12,6
CQUAD4,13,11,9,14,15,10
CQUAD4,14,11,10,15,16,11
CQUAD4,15,11,11,16,17,12
PSHELL,11,12,0.5
MAT1,12,10.6+6,,0.3,0.05
SPC1,3,246,3,9,14
SPC1,3,135,6,12,17
SPC1,3,135,14,15,16
Main Index
404 MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
9. $FLSYM,M,S1,S2
FLSYM,4,S,A
10. $PRESPT,IDF,,IDP,PHI,IDP,PHI,IDP,PHI
PRESPT,7,,27,30.,28,60.
PRESPT,8,,30,30.,31,60.
PRESPT,13,,33,30.,34,60.
11. $FREEPT,IDF,,IDP,PHI,IDP,PHI,IDP,PHI
FREEPT,1,,40,15.,41,30.,42,45.
FREEPT,1,,43,60.,44,75.
FREEPT,2,,45,15.,46,30.,47,45.
FREEPT,2,,48,60.,49,75.
EIGC,1,INV,MAX
0.,0.,0.,5.,3.,2,2
ENDDATA
Main Index
CHAPTER 11 405
Coupled Fluid-Structure Analysis
6 12
5 11
Anti-
symmetric
4 10
17
16
15
14 3 9
Symmetric
4.0 4.0
Free Surface
1 2 3 4 5 6
Fluid
5.0 Points
7 8 9 10 11 12
Structure
Points
5.0
13 14 15 16 17
Fixed
r
Fluid:Density p = 0.03 Structure:Thickness t = 0.5
Bulk Modulus B = 1 Density r = 0.05
Gravity g = 32.2
Main Index
406 MD Nastran Dynamic Analysis User’s Guide
Virtual Fluid Mass
User Interface
The user interface for a virtual fluid mass analysis is simple and straightforward. The sketch below
illustrates some of the features.
Main Index
CHAPTER 11 407
Coupled Fluid-Structure Analysis
Free Surfaces
Element Fluid
List #1 Volume 1
Element
Fluid List #2
Volume 2
Structure
1. The fluid/structure interface is defined with ELIST Bulk Data, which specify a set of wetted
TRIA3 and QUAD4 elements that define the structural portion of the fluid boundary.
2. Each fluid volume is defined on an MFLUID Bulk Data input, which defines the fluid density, the
ELISTs, and other boundaries.
3. A set of MFLUID volumes are requested in the case control request MFLUID = SID.
4. If pressure outputs are desired they will be printed if the case control request MPRES = is used to
define a set of elements. The elements must also be active in an ELIST/MFLUID volume.
An example set of input data for a small problem is shown below:
Case Control
MFLUID = 25
MPRES= ALL
etc.
Bulk Data
1 ,2 ,3 ,4 ,5 ,6 ,7 ,8 ,9 ,10
$MFLUID, SID , CID , ZFS , RHO , ELIST1, ELIST2, PLANE1, PLANE2
MFLUID , 25 , 0 , 115.4, .0246, 255 , , S
$ELIST , ELID ,E1 , E2 , E3 , etc.
ELIST , 255 , 1 , THRU , 25
ETC...
Main Index
408 MD Nastran Dynamic Analysis User’s Guide
Virtual Fluid Mass
In the input above we have defined a fluid volume with a free surface normal to the basic z direction.
The xz plane (PLANE1) is a plane of symmetry and the density is 0.0246. ELIST set 255 defines the
positive faces of elements in the range 1 through 25 (missing numbers are allowed) as the structural/fluid
boundary.
The connected elements may be wetted on one or both sides by the same volume. For instance a baffle
plate extends partially into a tank and have two sides in the same fluid volume. If a plate completely
separates the two parts of the tank, it is recommended that a separate volume be used for each side. Each
side of an element should be listed on a separate ELIST.
Special terms are calculated if enclosed fluid volumes do not have a free surface or a plane of
anti-symmetry. Otherwise, the incompressible fluid would produce a spurious low frequency mode with
a mode shape corresponding to the volume change. This mode will not affect the structural response, but
it is eliminated. Because the free surface allows the fluid volume to expand, it does not have these
problems.
A free surface is defined as an x-y plane in any local rectangular coordinate system. The user simply
specifies a value of z as the upper limit of the fluid volume. Different fluid volumes may have different
levels and orientation. It is not required that the surface coincide with the element properties since
partially wetted elements are allowed.
The virtual mass fluid option may be used in all MSC.Nastran dynamics solutions, including the
following special approaches:
• Superelements – (Residual Only )
• Nonlinear Analysis – (SOL’s 99 and 129)
• Optimization – (does not create sensitivity matrices.)
Theoretical Summary
The following is a brief overview of the virtual mass approach. For more information, refer to the
standard references on boundary elements.
The Helmholtz method used by MSC.Nastran solves Laplace’s Equation by distributing a set of sources
over the outer boundary, each producing a simple solution to the differential equation. By matching the
assumed known boundary motions to the effective motion caused by the sources, we can solve a linear
matrix equation for the magnitude of the sources. The values of the sources determine the effective
pressures and, thereby, the forces on the grid points. Combining all of these steps into a matrix equation
results in a virtual mass matrix as derived below.
If j is the value of a point source of fluid (units are volume flow rate per area) located at location r j ,
and is assumed acting over an area A j , the vector velocity u· i at any other point r i is
j ei j
u· i = --------------------
- dA (11-1)
r –r 2 j
i j
j A
j
Main Index
CHAPTER 11 409
Coupled Fluid-Structure Analysis
where e ij is the unit vector in the direction from point j to point i . Note that the gradient of the vector
u· i is the potential function which satisfies Laplace’s Equation on a term by term basis.
The other set of necessary equations are the pressures, p i , at any point, i , in terms of the density, ,
sources and geometry, namely
·
j e i j
pi = ------------------ dA j
ri – rj
(11-2)
j A
j
The results of integrating Eqs. (11-1) and (11-2) over the finite element surfaces are collected
respectively in two matrices, and where
u· = (11-3)
and
·
F = (11-4)
where F are the forces at the grid points. The matrix is obtained by integrating Eq. (11-2). An
additional area integration is necessary to convert the pressures to forces. A mass matrix may now be
defined using Eqs. (11-3) and (11-4) as
f
F = M u·· (11-5)
f –1
M = (11-6)
An alternate calculation is performed when the matrix is potentially singular. See Remark 8 on the
MFUID Bulk Data description for the specific instances. An additional source, s 2 , is added near the
center of each MFLUID volume. The expanded versions of Eqs. (11-3) and (11-4) are
u· = + 2 s 2 (11-7)
·
F = + 2 s 2 (11-8)
Main Index
410 MD Nastran Dynamic Analysis User’s Guide
Virtual Fluid Mass
An additional equation provides that the sum of the squares of the ordinary sources in the vector, 1
is minimized and a set of Lagrange multipliers, are used to enforce Eq. (11-7) as constraints. The
function, U , to be minimized is defined as
1 T T
U = --- + u· – – 2 s 2 (11-9)
2
Taking the derivatives of U with respect to , 1 , and s 2 , we obtain three sets of equations. Combining
them, we obtain the following matrix equation:
0 2 u·
T –I 0 = 0 (11-10)
2T 0 0 s 2 0
The vector may be eliminated by solving the second row partition and substituting into the first row
partition of the matrix, with the result:
T 2 u·
s = (11-11)
2T 0 2 0
Equation (11-11) may be solved for and s 2 , which are then used to obtain pressure. To obtain the
pressure, we substitute for in Eq. (11-8) to obtain the matrix equation
·
p = T 2 · (11-12)
s2
Equations (11-11) and (11-12) may then be combined into a single matrix defining the fluid, namely
F = M f u·· (11-13)
where:
–1
T 2 I
Mf = T 2 (11-14)
2T 0 0
Note that the size of the matrix equation is only one term larger than the nonconstrained case. Although
it will be well behaved for the enclosed volume case, the overall incompressible constraint is lost.
Other methods to avoid the singular matrix are as follows:
1. Put one or more small holes in the boundary by removing an element ID in an unimportant area
from the ELIST. The fluid will then leak out to an infinite domain of fluid.
Main Index
CHAPTER 11 411
Coupled Fluid-Structure Analysis
2. Define a free surface near the top of the container and modify the ELIST to remove elements
above the surface.
3. Constrain the structure to eliminate any net change to the enclosed volume. In other words
generate an MPC equation such that
Ai ni ui = 0 (11-15)
i
where i is a boundary grid point and A i and u i are the effective area and normal vector, respectively.
(This is not an easy task.)
Fortunately, there are no problems in which free surfaces or planes of symmetry with antisymmetric
motion are present since the pressure must be zero on these boundaries.
The phantom boundary may be constructed from CTRIA3 and CQUAD4 elements which have zero or
nearly zero stiffness in the normal direction. The PSHELL bending thickness is used for this purpose.
The membrane stiffness is optional. If no other structural elements are present, a small stiffness is
desired for both directions to avoid automatic constraints.
Note that this method will couple only the fluid displacements in the normal direction. Edge/corner
effects and tangential motions will be approximate.
Gravity Effects
The free surfaces produced by the Virtual Fluid Mass option are simple planes of antisymmetric motion
with a null pressure assumed at the location of the free surface plane. Effects such as fluid sloshing due
to gravity waves are assumed to be uncoupled from the higher frequency structural modes.
An approximation to the gravity effects for fluids with finite boundaries may be modeled using a
phantom boundary instead of a free surface. Spread grid points and plate elements over the surface and
constrain the in-plane motions and rotations to zero. Give the plates a small membrane thickness and no
bending material property. Add scalar springs (CELASi) in the direction normal to the plane with
stiffnesses, K i , calculated from
Main Index
412 MD Nastran Dynamic Analysis User’s Guide
Virtual Fluid Mass
K i = A i g (11-16)
where A i is the area under the point, is the density, and g is the gravitational constant. Note that these
springs will affect the rigid body motion of the whole system and this method should be used with
caution. Another drawback is that the extra gravity boundary may result in a completely enclosed fluid
volume.
Examples
The legal and illegal types of fluid boundaries are illustrated in Figure 11-4. The categories are GOOD
= legal configuration, BAD = illegal boundary, and MAYBE = conditionally legal, which could be
permitted if a phantom boundary is used.
Main Index
CHAPTER 11 413
Coupled Fluid-Structure Analysis
Ai
Infinite
Fluid
Fluid Fluid
Air Air
Fs
PB
Fs Fs Fs
Air
Fluid Fluid PB
Fluid
Free PB = Phantom
Surfaces Boundary Between 2
Fs Fluids
Fs Fs
Fluid Fluid
Air
Internal Hole
Only Fluid Fluid Fluid
Fs
Main Index
414 MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
Limitations
Very few limitations are placed on the analyst. Small motion theory prohibits the use of the models in
high velocity flows (more than Mach 0.5) such as mufflers and jet engines. Exterior fluids may be
modeled with extra effort by extending the fluid model out to a large distance and by providing a damped
and/or constrained exterior. Surface sloshing effects caused by gravity such as those in fuel tanks can be
modeled with additional effort by simulating the gravity effects on the free surface with scalar springs.
A formal theoretical discussion is provided in Additional Topics (p. 555) in the MD Nastran Reference
Manual.
Main Index
CHAPTER 11 415
Coupled Fluid-Structure Analysis
It is generally easier to build the three-dimensional fluid mesh first and then use the outer faces
to model the structure (which is usually modeled with plates). Unfortunately, most engineers
start with the structure that has been used for another analysis and encounter difficulties when
they attempt to construct a three-dimensional FE model within the enclosure. In most cases, the
fluid model will have a different number of points on opposite faces and the engineer cannot
construct a simple, matching mesh without having leftover points and lines. In this case, he
should proceed with a simple fluid model and let MD Nastran control the mismatched
connections.
Although MD Nastran is more efficient when the structural GRID points coincide with the fluid
boundaries, it will process misaligned meshes when the fluid faces are larger than the adjacent
structural elements. The basic model interface definitions are defined on the ACMODL Bulk Data
entry. The user may define the sets of points on the interface and a tolerance to define the distance
allowed between a fluid surface element and a connected structural point. This tolerance will be
important when several points are close together and only one is wetted by the fluid.
2. Building the Fluid Finite Element Model
Each GRID point that defines the fluid mesh will have one degree-of-freedom defining the
pressure at a point at a fixed location. The fluid itself is modeled with existing three-dimensional
connection data (CHEXA, CPENTA, and CTETRA Bulk Data). The existence of a fluid is
defined on the property data (Field 8 of the PSOLID Bulk Data) which also references the ID of
the fluid material properties (on a MAT10 Bulk Data entry).
User-input fluid coefficients are density, bulk modulus, and/or compressible wave speed.
Incompressible fluids are approximated by using a large bulk modulus or wave speed. (A typical
value for incompressible fluid would be a wave length of 10 times the size of the structure.) The
integration options on the PSOLID data should be full Gaussian Integration with no shear
reduction scheme.
The higher-order elements (with midside nodes) will generally give better results than the basic
linear elements. The CHEXA, CPENTA and CTETRA elements can be connected together
without the problems that occur in structural elements. However, mixing high-order and
low-order elements is not recommended.
At this stage in the modeling process, it is advisable to set up the loading and desired output points
on the fluid. Acoustic sources such as a small loudspeaker may be specified directly by scalar
loading functions (ACSRCE, SLOAD, DAREA, and/or RLOAD Bulk Data) without the need to
model the structural details. The output displacement printout will actually be the scalar pressure
values.
Guidelines on building the models are different for the fluid elements than for normal MSC
structural elements. The lack of shear locking allows larger aspect ratios and more distorted
shapes for individual elements. Unlike structures, the fluid CHEXA elements have no particular
accuracy advantages over the CTETRA and CPENTA elements. However, in theory, fluid finite
element accuracies will be more sensitive to sharp corners and abrupt openings. In these cases,
a model should be refined using smaller elements around the discontinuity.
Main Index
416 MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
3. Automatic Boundaries
Fluid boundaries may be fixed, flexible, or free. With no applied loads, the outer surfaces of the
fluid elements define a fixed, rigid boundary condition, approximating a normal pressure gradient
of zero. However, if structural finite elements are attached, they will generate flow into the fluid
and affect the pressures. The pressure degrees-of-freedom may be treated exactly like
displacement DOFs, i.e., constraints may be necessary to fix the pressure of an open boundary to
zero. Also, they may be constrained with SPC or MPC Bulk Data for various other reasons, such
as symmetry.
If no structure is connected and the problem involves only fluid, then the effects of the boundary
flexibility may be ignored and the problem is greatly simplified. A stand-alone fluid model may
be analyzed by any of the standard MD Nastran solution sequences. For instance, the natural
frequencies or resonances may be obtained from SOL 103 using symmetric real methods at
greatly reduced costs.
The user controls the structural interfaces with the ACMODL Bulk Data and the locations of the
boundary GRID points. The input defaults are recommended that automatically prompt MD
Nastran to search all elements for potential interfaces. If only a set of fluid or structural points are
desired to be connected, the searching process can be restricted to user-selected points defined on
a SET1 Bulk Data entry. When identical meshes are used, an MD Nastran branch is also provided
to generate area factors more efficiently and accurately.
Problems may occur when plate elements are wetted on both sides by the fluid and a single
structural point must be connected to both sides. The recommended method for this case is to use
separate fluid points for each side of the panel with no physical separation. (The thickness is
ignored.) No provision is made for other special cases such as flow through holes in the plates.
No surface friction, surface tension, gravity, or other indirect boundary effects are calculated for
the simple fluid-structure interface. Damping effects due to acoustic surface materials are
described in the next section.
4. Absorbers, Barriers, and Panels
The acoustic absorber and barrier are used for advanced analysis when special boundary effects
are needed to model soundproofing materials on the structural surface or baffles within the fluid.
The absorber elements (CHACAB and PACABS Bulk Data ) are used to attenuate the reflections
of the acoustic waves when they encounter the structural boundary. They are connected between
the structure and an additional set of displacement GRID points which serve as the actual fluid
interface. They provide a simple tuned circuit with a mass on the point, and a spring and damper
in parallel connected to the structure.
The recommended procedure to implement the CHACAB elements is to move all of the fluid
boundary points a small distance away from the surface to avoid the automatic connections. An
additional set of GRID points corresponding 1:1 with the structural points are placed on the new
fluid boundary to replace the structural points. These absorber points are constrained to allow
only normal displacements and connected to the structure with CHACAB elements. Note that the
damping factors are important in complex eigenvalue and frequency response analysis but will be
ignored in real eigenvalue analysis.
Main Index
CHAPTER 11 417
Coupled Fluid-Structure Analysis
The barrier elements CHACBR are similar except that they provide masses on both sets of points
and have no damping matrix. They may be used without structural elements to simulate heavy
acoustic baffling panels that have little stiffness in the frequency of interest. Either side may be
connected to fluid boundaries or one side may be constrained.
The PANEL Bulk Data entry is a convenient method for combining sets of structural elements
for output purposes.
Setting Up MD Nastran
A number of solution control parameters and commands may be used to guide the acoustic solutions.
These are as follows:
• Executive and Case Control
The solution sequences that currently support the 3-D acoustic boundaries are SOLs 107
through 112. These include both modal and direct formulations for transient and frequency
response, as well as complex eigenvalue analysis. Superelement controls and restarts are
automatic in these solutions and no special data is needed for the acoustics.
The Case Control data should be set up as a normal dynamics solution. A CMETHOD = request
is necessary to obtain coupled eigenvalues. DLOAD and FREQ requests are necessary to
analyze forced response in the frequency domain. DLOAD, IC, and TSTEP requests are
necessary for transient analysis. The main concern here is to avoid costly runs caused by an
excessive number of time steps or frequencies. Be aware that the unsymmetric dynamic
solution matrices with complex terms used in this system may run several times longer per
solution point than a corresponding structural model.
• Estimating Job Costs
Although the fluid GRID points will only add one degree-of-freedom per point to the model a
three-dimensional mesh in the fluid can be several times the size of the basic structural matrix.
Furthermore the boundary coupling terms are unsymmetric and probably damped causing more
numerical processing. As a rough estimate, each matrix solve step will be 4X to 10X the cost of
a structural static solution. In direct frequency analysis each frequency requires a matrix
decomposition. The complex eigenvalues will require approximately 2 to 5 solves. The
transient solutions are dominated by vector operations which are several times faster per time
step. However, most transient problems require many time steps for adequate accuracy.
• Run Strategies
For large models, the superelement (SE) reduction methods are recommended to reduce the size
of the solution matrices. The interior points of the fluid and the non-boundary structure grid
points may each be defined as a separate Tip SEs, which are reduced to a smaller solution size,
while the boundary points remain in the residual SE. The residual SE may also contain
generalized modal displacements in place of the interior of both fluid and structural points by
using the modal synthesis method.
Main Index
418 MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
For moderately large problems, the ASET/OMIT method of reduction is available for both the
fluid and structural degrees-of-freedom. However, the accuracy is highly dependent on the
analyst’s choice of retained degrees-of-freedom. For instance, if all interior fluid points were
omitted, the internal acoustic waves would not be sinusoidal and the waves would instantly
travel from end to end. The most effective application would be a highly refined structural
model connected to a coarser 3-D fluid model. Removing most of the extraneous structural DOF
(i.e., rotations, tangential motion, and nonwetted points) could reduce the costs significantly.
The recommended first analysis for the coupled problem is the Direct Complex Eigenvalue Solution
Sequence (SOL 107). This solution will indicate the overall dynamic behavior dominated by the lowest
frequency natural modes and resonant frequencies. The following are recommendations:
1. The EIGC input to control complex eigensolutions is substantially different from the real mode
EIGR data. Read and understand the input definitions. Avoid the default values.
2. The frequency range definitions are specified in units of Radians per Second. Multiply the
frequencies of cycle per second by 2 .
3. Do not use damping on the initial runs. The results will be easier to debug and the job will run
faster.
4. Beware that an additional mode with zero frequency occurs in closed containers. This mode
causes problems with both INVP and LANC methods. Avoid it by specifying a lower frequency
range nearly as large as the lowest nonzero mode. An alternative is to add a CELAS spring
connected between a fluid point and ground.
• The HESS method should be used only with some form of matrix reduction for typical
problems. See the comments above regarding the use of ASET/OMIT an Superelement
partitioning.
Main Index
CHAPTER 11 419
Coupled Fluid-Structure Analysis
*******************************************************
* *
* MATCHING GRID TABLE *
* *
*******************************************************
* USER GRID IDS. *
* ACOUSTIC GRIDS * STRUCTURAL GRIDS *
* * *
*******************************************************
* * *
* 1 * 1001 *
* * *
* 2 * 1002 *
* * *
* 3 * 1003 *
* * *
* 4 * 1004 *
* * *
* 5 * 1005 *
* * *
* 11 * 1011 *
* * *
ETC
*******************************************************
The following message is a sample of a case when the boundaries do not match:
and:
^^^ DMAP INFORMATION MESSAGE 9055 (SEMG) - THE FL./STR. INTERFACE CHECK IS FORCES AND MOMENTS RESULTING FROM
A UNIT INCREASE IN PRESSURE, OR CHANGES IN THE FLUID PRESSURE RESULTING FROM RIGID BODY MOTIONS OF THE
STRUCTURE. THESE VALUES ARE DIRECTLY PROPORTIONAL TO THE OPEN SURFACE OF THE FLUID.
The following message indicates the error factors in area for free body motions. Nonzero numbers
indicate a hole in the model as indicated by the T2 value.
Main Index
420 MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
• Restarts
The only major reasons to restart a fluid/structure interaction job are to use superelement
processing and/or to change the output requests. In typical jobs the cost in the solution phase is
more than 50% of the run, and a restart for a model or load change is not worth the cost of saving
a large database.
If the uncoupled structural model is large compared to the size of the actual coupled surface, it may be
defined as one or more Tip superelements and run as a separate initial job. The interior points in the fluid
may also be treated in this manner. Many acoustic analysis runs can then be restarted each from a copy
of the initial database without recalculating the large matrices.
Print and plot output requests may need many changes after the initial runs are made. An output-only
restart can be very cost efficient on a large job. Note that most graphical displays do not display complex
numbers on model views (i.e., contour plots or deformed shapes). Printed output of selected grid point
or element sets may be necessary and is recommended.
Diagnosing Problems
Most of the problems in starting to use acoustics are expected to occur in the definition of the area factors
which define the fluid/structure boundary. Some restrictions are necessary to control errors in the case
of overlapping fluid and structural FE meshes, and in most cases the user will be warned of discrepancies.
Some recommended techniques are as follows:
• Carefully check the special diagnostic outputs shown above. Holes in the boundaries will cause
nonzero resultant area factors.
• Run tests without the boundary coupling to estimate the frequencies.
• Temporarily switch to an alternate eigenvalue method and/or a smaller range to ensure modes
are correct.
• If a model is large, try subdividing it into smaller sections that can be checked more
conveniently.
• Try changing the fluid density to a large number (e.g., water) and run with free structural
boundaries. The fluid should not restrain the free-body motions and these modal frequencies
should remain at zero.
Advanced Methods
Many of the other MD Nastran features should be available for acoustic analysis. Some potential
advanced modeling and solution features that will be useful are as follows:
• Modal formulation with superelements: The use of Component Mode Synthesis is
recommended for reducing the size of the matrices and the solution costs. The
degrees-of-freedom of the interior fluid points will be replaced by the modes of the fluid with
rigid container boundaries. The uncoupled structural points may be replaced by modal
coordinates defined by a structure in a vacuum.
Main Index
CHAPTER 11 421
Coupled Fluid-Structure Analysis
• Virtual Fluids: The MFLUID Bulk Data input may be used to represent incompressible fluids
attached to the structure. This may be used to approximate the low frequency effects of an
infinite exterior fluid.
• Decoupled Response Solutions: In most acoustic problems with a load-carrying structure, the
structural modes are not affected by the small mass of the air. In turn, the high stiffness of the
structure is effectively rigid for the purposes of computing acoustic modes in the air. Forced
vibration problems may then be analyzed as two decoupled symmetric solutions:
• Constrain the entire structural model and perform a forced response analysis on the air (use
SLOAD inputs to define the volume inputs from a known source) and save the resulting
boundary pressures.
• Restart the job with the fluid points constrained and a DMAP alter to add the pressure forces
to the frequency- or time-dependent load vectors. The resulting structural model will respond
to the pressures from the first run.
• Random and Response Spectrum Methods: The MD Nastran random analysis options will also
be useful for acoustic analysis. This capability requires a set of unit forced responses in the
frequency domain (SOLs 108 and 111) and spectral densities of the forcing functions (RANDPS
inputs). It will calculate the combination spectral densities and autocorrelations of selected
output quantities. In acoustic problems, it could be used for problems having several correlated
sources such as an automobile interior.
Main Index
422 MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
A simple test problem illustrates the fundamentals of acoustic analysis for enclosed containers with
flexible walls. The physical problem represents an initial attempt at design and analysis of an enclosure
for an acoustic suspension loudspeaker system. The objective was to calculate the resonant frequencies
and responses of the system without the mass and impedance of the speaker coil and magnet components.
Physical Description
The structure consists of a simple rectangular box with wood walls as shown by the plate model in
Figure 11-6. A single cutout is provided for the speaker and a thin polyethylene cone was modeled with
triangular shells. The physical properties of the model are listed below. Note the use of the MKS system
of units, which illustrates the flexibility of MD Nastran.
Main Index
CHAPTER 11 423
Coupled Fluid-Structure Analysis
The Executive and Case Control inputs are used for selecting the type of solution, loads, boundaries, and
outputs. Several different runs using the same model were made to check results and investigate the
effects of structural coupling. The configuration shown below was a Direct Complex Modal solution
using the Lanczos method to extract eigenvalues. An SPC request was used to constrain the structure
and an output set was defined to limit the printed displacements. All other data was normal,
recommended control and label input.
Bulk Data
The key data of note in the Bulk Data Section are as follows:
1. The ACMODL input uses the minimum form which connects fluid and structural points at
identical locations.
2. Several eigenvalue methods were used. It is recommended that complex roots always be checked
by another method.
3. The speaker cone consisted of a light, thin plastic material. The virtual mass of the exterior air
was estimated and added as nonstructural mass on the PSHELL input.
Results from MD Nastran
Three runs (out of many) are described below. They illustrate a recommended sequence for the analysis
process.
Run 1: Real Eigenvalue Analysis
Before the coupled structure is analyzed, it is important to understand the behavior of the structural and
fluid models separately. Fortunately they can be included in the same data file and the Real Eigenvalue
Solution Sequence (SOL 103) may by used. The results for the uncoupled speaker box and the acoustic
modes are shown below. Note that the natural frequencies for both systems occur in the same range,
which indicates that the enclosure modes will interact with the acoustic cavity resonances. Results for
the acoustics were checked by one-dimensional wave solutions. Results for the box were checked
against calculated natural frequencies of simply-supported plates.
Main Index
424 MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
-etc.-
Run 2: Complex Eigenvalue Analysis, Coupled Modes.
The coupled natural frequencies must be obtained from an unsymmetrical matrix equation which requires
a Complex Eigenvalue method even in the undamped case. The CLAN method is recommended for most
large-order non-superelement jobs. The only changes from the Real Modes job was the addition of an
ACMODL input and the changes in eigenvalue method.
Results above are interpreted by examining the frequency shifts from the uncoupled system. The modes
with u(y) fore-aft motion are most likely to be excited by the speaker. Modes 1, 2, 3, 4, 6, and 7 are close
to their uncoupled equivalent. Modes 5 and 8 are new combinations of higher modes. The results were
confirmed by several methods:
1. The job was rerun using the INV method on the EIGC Bulk Data and produced nearly identical
results.
2. A printout of the interface area matrix was obtained and verified by hand calculations.
3. The natural symmetry of the geometry produced symmetric and antisymmetric results relative to
the natural structural planes of symmetry (except for roots with close frequencies).
Run 3: Complex Eigenvalue Analysis, Modal Formulation
A more efficient method for solving a coupled matrix problem is to reduce the size of the matrices by
using modal coordinates. In this case, both the structural displacements and fluid pressures are replaced
in the matrix solution by generalized coordinates representing the uncoupled real modes. For the modal
Main Index
CHAPTER 11 425
Coupled Fluid-Structure Analysis
cases below the real modes (25 total) below 800 Hz. were used. In the coupled cases 13 modes were
obtained below 500 Hz. All results compared to within 1%.
On a Sun SPARCstation 1+ the CPU time comparisons are:
The conclusion is that the modal method (SOL 110) will reduce the costs with very little effect on
accuracy. The user is cautioned to use a liberal number of real modes (twice the number of coupled
modes) to represent the system.
ACOUSTIC TEST PROBLEM DATA FILES
ID SPEAK39F,DNH
SOL 107
DIAG 8,12 $ PRINTS MATRIX TRAILERS AND ROOT-TRACKING MESSAGES.
CEND
TITLE=SPEAKER BOX -WITH CONE, SIMPLE CORNER SUPPORTS
SUBTITLE = COUPLED BOUNDARY, NON-MATCHING ELEMENTS
ECHO= UNSORT
SEALL=ALL
SPC=20
$ USES MKS SYSTEM
CMETHOD = 7 $LANCZOS
SET 20= 3,13,23,43,82,83,84,91,93,95,103,113,123,
131,135,153,163,171,173,175,183,
1013,1023,1043,1082,1083,1084,1091,1095,1113,
1131,1135,1163,1173,1183,1193
DISP= 20 $ FOR MINIMUM PRINTOUT
$ DISP(PLOT)=ALL $ FOR MSC/XL CONTOUR PLOTS
$
BEGIN BULK
$
$ DEFINE FLUID/STRUCTURE INTERFACE: COINCIDENT POINTS
ACMODL,IDENT
$
PARAM,POST,0
PARAM,COUPMASS,1
$
$ BOX PROPERTIES - WOOD
MAT1,11,11.61+9,,0.3,562.0
PSHELL,1000,11,.015,11,,11
$
$ SPEAKER CONE
MAT1,3,3.4+9,,0.3,450.0
PSHELL,10,3,0.1-3,3,,,0.223
$
$ PROPERTIES OF AIR
MAT10,100,131.94+3,1.115
PSOLID,100,100,,2,,1,PFLUID
Main Index
426 MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
$
$ EIGEN METHODS
EIGR,20,MGIV,1.0,600.0
EIGC,7, CLAN,,,,,,,+CLAN
+CLAN,0.0,10.0,0.0,1600.0,100.0,,20
$ USE INVERSE POWER TO CHECK LOW ROOTS
EIGC,107,INV,MAX,,,,,,+EC1
+EC1,0.0,100.0,0.0,1800.0,100.0,12,9
$
$ FIX BOX AT BOTTOM CORNERS
SPC1,20,123,1001,1005,1031,1035
$ COORDINATE SYSTEM AT CENTER OF HOLE
CORD1C,83,283,113,85
$
$
$FLUID GRID POINTS
$ NOTE VALUE OF -1 IN FIELD 7 INDICATES 1 DOF.
GRID,1,, -.25, -.2, -.3,-1
GRID,2,,-.125, -.2, -.3,-1
GRID,3,, 0.0, -.2, -.3,-1
GRID,4,, .125, -.2, -.3,-1
GRID,5,, .25, -.2, -.3,-1
GRID,11,, -.25, -.0667, -.3,-1
GRID,12,,-.125, -.0667, -.3,-1
GRID,13,, 0.0, -.0667, -.3,-1
GRID,14,, .125, -.0667, -.3,-1
GRID,15,, .25, -.0667, -.3,-1
GRID,21,, -.25, 0.0667, -.3,-1
GRID,22,,-.125, 0.0667, -.3,-1
GRID,23,, 0.0, 0.0667, -.3,-1
GRID,24,, .125, 0.0667, -.3,-1
.25, -.2, 0.,-1
-
ETC
-
$ STRUCTURE GRIDS
GRID,1001,, -.25, -.2, -.3
GRID,1002,,-.125, -.2, -.3
GRID,1003,, 0.0, -.2, -.3
GRID,1004,, .125, -.2, -.3
GRID,1005,, .25, -.2, -.3
GRID,1011,, -.25, -.0667, -.3
GRID,1012,,-.125, -.0667, -.3,,6
GRID,1013,, 0.0, -.0667, -.3,,6
-
ETC.
-
$ OPTIONAL ASET DATA TO USE WITH HESS METHOD
$ASET1,1,1,THRU,195
$ASET1,123,1003,1012,1014,1022,1024,1033
$ASET1,123,1163,1172,1174,1182,1184,1193
$ASET1,123,1042,1043,1044,1081,1082,1083,1084
$ASET1,123,1065,1122,1123,1124
$ASET1,123,1072,1074,1112,1113,1114,1152,1154
$ASET1,123,1051,1061,1131,1141
$ASET1,123,1055,1065,1135,1145
$
$
$SOLID ELEMENTS FOR AIR
CHEXA 1 100 1 2 12 11 41 42 +001
Main Index
CHAPTER 11 427
Coupled Fluid-Structure Analysis
+001 52 51
CHEXA 2 100 2 3 13 12 42 43 +011
+011 53 52
CHEXA 3 100 3 4 14 13 43 44 +021
+021 54 53
CHEXA 4 100 4 5 15 14 44 45 +031
+031 55 54
CHEXA 11 100 11 12 22 21 51 52 +041
+041 62 61
CHEXA 12 100 12 13 23 22 52 53 +051
+051 63 62
CHEXA 13 100 13 14 24 23 53 54 +061
+061 64 63
CHEXA 14 100 14 15 25 24 54 55 +071
+071 65 64
CHEXA 21 100 21 22 32 31 61 62
+081
+081 72 71
-
ETC
-
$STRUCTURAL ELEMENTS
CQUAD4 1001 1000 1001 1002 1012 1011
CQUAD4 1002 1000 1002 1003 1013 1012
CQUAD4 1003 1000 1003 1004 1014 1013
CQUAD4 1004 1000 1004 1005 1015 1014
CQUAD4 1011 1000 1011 1012 1022 1021
CQUAD4 1012 1000 1012 1013 1023 1022
CQUAD4 1013 1000 1013 1014 1024 1023
CQUAD4 1014 1000 1014 1015 1025 1024
-
ETC
-
$
$SIMPLE CONE
$
CTRIA3, 5042,10,1042,1043,1083
CTRIA3, 5043,10,1043,1044,1083
CTRIA3, 5044,10,1044,1084,1083
CTRIA3, 5084,10,1084,1124,1083
CTRIA3, 5124,10,1124,1123,1083
CTRIA3, 5123,10,1123,1122,1083
CTRIA3, 5122,10,1122,1082,1083
CTRIA3, 5082,10,1082,1042,1083
ENDDATA
Main Index
428 MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
Exterior Acoustics
Introduction
With interior acoustic problems, the acoustic domain considered is bounded. A typical example is the
determination of the sound pressure inside a car.
With exterior acoustic problems, the acoustic domain is unbounded. The analysis studies the sound
pressure in the vicinity of the vibrating body or far away from the body. A further important result is the
acoustic power radiated from the vibrating structure. A typical example is the determination of the
radiated acoustic power of an engine.
The infinite elements are attached to the boundary of the acoustic finite element mesh to provide the
correct non-reflecting boundary condition.
Apart from standard acoustic results within the finite region it is possible to compute acoustic results at
points within the infinite elements. These points, the so-called field points, may be connected by elements
to form a field point mesh. If a field point mesh is defined, also the acoustic power through this field point
mesh is computed.
Input
Bulk Data entries, CACINF3 and CACINF4, are used to define the connectivity of the infinite elements.
The properties of the infinite elements are defined on PACINF Bulk Data entries.
Field points and field point meshes are defined in separate sections of the bulk data file. These sections
must follow the main bulk data section.
Output of radiated power from the wetted surface and intensities on the wetted surface is controlled by
Case Control commands ACPOWER and INTENSITY. Output of field point mesh results is controlled
by Case Control command ACFPMRESULT.
Main Index
CHAPTER 11 429
Coupled Fluid-Structure Analysis
Within an infinite element, the acoustic pressure is expanded into a power series of (1/r) where r is the
distance from the pole.The radial interpolation order is the number of terms kept in this series.
The connectivity of the base is defined on the CACINF3 and CACINF4 Bulk Data entries. The
orientation should be such that the normal vector on the base points into radial direction. However,
MD Nastran will automatically change the orientation if the normal vector does not point away from the
pole. The location of the pole as well as the radial interpolation order are defined on PACINF Bulk Data
entries.
Main Index
430 MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
1 2 3 4 5 6 7 8 9 10
CACINF3 EID PID G1 G2 G3
Field Contents
EID Element Identification Number (Integer > 0)
PID Property Identification Number of a PACINF entry (Integer > 0)
Gi Grid Point Identification Numbers of Element Base Connection Points (Integer > 0)
1 2 3 4 5 6 7 8 9 10
CACINF4 EID PID G1 G2 G3 G4
Field Contents
EID Element Identification Number (Integer > 0)
PID Property Identification Number of a PACINF entry (Integer > 0)
Gi Grid Point Identification Numbers of Element Base Connection Points (Integer > 0)
Main Index
CHAPTER 11 431
Coupled Fluid-Structure Analysis
1 2 3 4 5 6 7 8 9 10
PACINF PID MID RIO X1 X2 X3
Field Contents
PID Property Identification Number of PACINF entry (Integer > 0)
MID Material Identification Number of a MAT10 entry (Integer > 0)
XP, YP, ZP Coordinates of the Pole of the Infinite Elements (in the Basic Coordinate System)
Main Index
432 MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
$
GRID 1001 2 5. 0. 0.
GRID 1002 2 6. 0. 0.
GRID 1003 2 7. 0. 0.
GRID 1004 2 8. 0. 0.
GRID 1005 2 9. 0. 0.
GRID 1006 2 10. 0. 0.
CORD2R, 2,, 0., 0., 0., -.5, -.5, .707107
, .5, .5, .707107
$
BEGIN AFPM = 200
$
PARAM, POST, -1
$
GRID, 1,, -1., -1., 2.
GRID, 2,, 0., -1., 2.
GRID, 3,, 1., -1., 2.
GRID, 4,, -1., 0., 2.
GRID, 5,, 0., 0., 2.
GRID, 6,, 1., 0., 2.
GRID, 7,, -1., 1., 2.
GRID, 8,, 0., 1., 2.
GRID, 9,, 1., 1., 2.
$
GRID, 11,, -1., -1., 2.
GRID, 12,, 0., -1., 3.
GRID, 13,, 1., -1., 2.
GRID, 14,, -1., 0., 3.
GRID, 16,, 1., 0., 3.
GRID, 17,, -1., 1., 2.
GRID, 18,, 0., 1., 3.
GRID, 19,, 1., 1., 2.
$
CQUAD4, 1, 1, 1, 2, 5, 4
CQUAD4, 2, 1, 2, 3, 6, 5
CQUAD4, 3, 1, 4, 5, 8, 7
CQUAD4, 4, 1, 5, 6, 9, 8
$
CQUAD4, 5, 1, 12, 16, 18, 14
CTRIA3, 6, 1, 11, 12, 14
CTRIA3, 7, 1, 12, 13, 16
CTRIA3, 8, 1, 16, 19, 18
CTRIA3, 9, 1, 14, 18, 17
$
ENDDATA
Main Index
CHAPTER 11 433
Coupled Fluid-Structure Analysis
Format
ALL
ACPOWER SORT1 PRINT,PUNCH CSV = unit = n
SORT2 PLOT
NONE
Describer Meaning
SORT1 Output will be presented as tabular listing of panels for each excitation frequency
SORT2 Output will be presented as tabular listing of excitation frequencies for each panel
(Default)
PRINT The printer will be the output medium (Default).
PUNCH The punch file will be the output medium.
PLOT Results are generated but not output.
CSV Results will be written to a .csv file.
unit Unit of the .csv file as used on the ASSIGN statement
ALL Radiated power will be processed for the wetted surface and all panels.
n Set identification of a previously defined set of panels. Radiated power will be
processed for the wetted surface and all panels in the referenced set.
NONE Radiated power will not be processed.
Main Index
434 MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
ALL
INTENSITY SORT1 PRINT,PUNCH = n
SORT2 PLOT
NONE
Describer Meaning
SORT1 Output will be presented as tabular listing of grid points for each excitation
frequency (Default)
SORT2 Output will be presented as tabular listing of excitation frequencies for each grid
point
PRINT The printer will be the output medium (Default).
PUNCH The punch file will be the output medium.
PLOT Results are generated but not output.
ALL Intensities will be computed for all grid points of the wetted surface.
n Set identification of a previously defined set of grid points. Intensities will be
computed for the grid points in this set only.
NONE Acoustic intensities will not be processed.
Main Index
CHAPTER 11 435
Coupled Fluid-Structure Analysis
ACFPMRESULT SORT1 PRINT,PUNCH VELOCITY = YES
SORT2 PLOT NO
ALL
REAL or IMAG POWER = YES
= n
PHASE NO
NONE
Describer Meaning
SORT1 Output will be presented as tabular listing of grid points for each excitation
frequency (Default)
SORT2 Output will be presented as tabular listing of excitation frequencies for each grid
point
PRINT The printer will be the output medium (Default).
PUNCH The punch file will be the output medium.
PLOT Results are generated but not output.
VELOCITY Requests output of particle velocities (Default: NO)
REAL or IMAG Requests rectangular format (real and imaginary) of complex output. Use of either
REAL of IMAG yields the same output.
PHASE Requests polar format (magnitude and phase) of complex output. Phase output is in
degrees.
POWER Requests output of power through field point mesh (Default: YES)
ALL Results of all field point meshes will be processed
n Set identification of a previously defined set of field point mesh identifiers. Results
will be processed for the field point meshes in this set only.
NONE Field point mesh results will not be processed.
Output
In addition to the standard acoustic results, the following results may be requested:
• Acoustic energy radiated from the wetted surface or from panels
Main Index
436 MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
• Normal component of the acoustic intensity at grid points of the wetted surface
• Acoustic pressure and intensity at the field points
• Acoustic velocity at the field points
• Acoustic energy flowing through the field point mesh
Field point mesh results can be written to the .f06 file or the .op2 file. The .xdb file is not yet supported.
Guidelines
1. The surface the infinite elements are attached to must be convex but it need not be smooth. Infinite
elements have to be connected to acoustic grid points. Thus it is necessary to model the vicinity
of the vibrating structure with acoustic finite elements until a convex surface is reached.
2. The radial interpolation order required depends on the directivity of the pressure field. Usually, a
higher order is needed for higher frequencies.
3. It is recommended to study the sensitivity of the results with respect to the radial interpolation
order by repeating the analysis with a different radial interpolation order.
4. Infinite elements are supported in direct and modal frequency response analysis (SOL 108 and
SOL 111). Experience shows that a large number of fluid modes is needed to get converged
results. Thus, modal reduction of the fluid is not recommended. However, modal reduction can
be applied to the structure.
5. Efficiency can be increased if the iterative solver is used. It is recommended to use the Jacobi
preconditioner with an epsilon of 10-4.
Limitations
1. Infinite elements are supported in a frequency response analysis only (SOL 108 and SOL 111).
2. In this prerelease, case control commands ACPOWER and INTENSITY are not yet supported.
Example
In this example, the sound transmission through an elastic plate embedded in an infinite rigid wall is
analyzed. Figure 11-8 shows the geometry of the plate and of the fluid region meshed with finite
elements. The finite element mesh of the structure and of one quarter of the fluid can be seen in
Figure 11-9.
Main Index
CHAPTER 11 437
Coupled Fluid-Structure Analysis
Infinite Elements are attached to the top and the four lateral faces but not to the bottom. The pole of all
infinite elements is at the center of the plate. The finite element mesh of the structure together with one
quarter of the infinite element mesh is shown in Figure 11-10.
The field point mesh is a cube which completely encloses the fluid mesh. Thus, the power through this
field point mesh equals the total power radiated from the plate. Figure 11-11 shows the fluid mesh and
one quarter of the field point mesh.
Main Index
438 MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
Main Index
CHAPTER 11 439
Coupled Fluid-Structure Analysis
The excitation is a uniform pressure applied to the plate. The acoustic pressure in the finite element mesh
and acoustic results at some discrete field points located along the z-axis and at the field points of the
field point mesh are computed. Acoustic results at the field points include the acoustic pressure and the
acoustic intensities. See Install_dir/mdxxxx/doc/dynamics/bd11aie.dat for the full file, a portion of the
input file is shown in Listing 11-1.
Main Index
440 MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
ACMODL, IDENT
$
RLOAD1, 10, 200,,, 300
PLOAD2, 200, 1., 1, THRU, 150
TABLED1, 300
, 0., 1., 1000., 1., ENDT
$
FREQ1, 20, 2., 2., 7
$
ITER, 30
PRECOND = J,
ITSEPS = 1.0E-4
$
GRDSET,,,,,, -1
$
$ fluid
$
PACINF 10 20 5 0. 0. 0.
$
CACINF4 13901 10 3641 3642 3658 3657
CACINF4 13902 10 3642 3643 3659 3658
CACINF4 13903 10 3643 3644 3660 3659
CACINF4 13904 10 3644 3645 3661 3660
CACINF4 13905 10 3645 3646 3662 3661
CACINF4 13906 10 3646 3647 3663 3662
.......
$
$ structure
$
PSOLID 2 20 PFLUID
MAT10 20 1.21 340.
$
CHEXA 1001 2 1001 1002 1018 1017 1177 1178
1194 1193
CHEXA 1002 2 1002 1003 1019 1018 1178 1179
1195 1194
CHEXA 1003 2 1003 1004 1020 1019 1179 1180
1196 1195
CHEXA 1004 2 1004 1005 1021 1020 1180 1181
1197 1196
.......
$
$ plates
$
PSHELL 1 1 .005 1 1
$
CQUAD4 1 1 1 2 18 17
CQUAD4 2 1 2 3 19 18
CQUAD4 3 1 3 4 20 19
CQUAD4 4 1 4 5 21 20
.......
$ Material Record : steel_iso_SI
MAT1 1 2.1+11 .3 7800. 1.2-5
.......
$ Loads for Load Case : Case_1
$ Displacement Constraints of Load Set : Constraints.10
SPC1 1 123 1 THRU 17
SPC1 1 123 32 33 48 49 64 65
80 81 96 97 112 113 128 129
144 145
Main Index
CHAPTER 11 441
Coupled Fluid-Structure Analysis
Results
Figure 11-12 shows the acoustic pressure in the Finite Element mesh and Figure 11-13 shows some
results printed to the .f06.
Main Index
442 MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
FREQUENCY = 1.000000E+01
C O M P L E X D I S P L A C E M E N T V E C T O R
(MAGNITUDE/PHASE)
FREQUENCY = 1.000000E+01
C O M P L E X A C O U S T I C P R E S S U R E R E S U L T S
(MAGNITUDE/PHASE)
Main Index
CHAPTER 11 443
Coupled Fluid-Structure Analysis
A Body in White method (BW) of calculating fluid structure interface has been implemented. Speed
improvements are such that an automobile model that consists of 120,000 - 150,000 shell elements and
15,000 - 20,000 fluid elements now takes only about 60 seconds for calculating the fluid/structure
interface matrix, which is a 1000 times speed improvement.
For instance, the following is a typical performance increase for a medium sized model:
You can request a .pch file that represents fluid and structure “skin” (param,skinout,punch) in order to
check interface accuracy. A utility program is provided for viewing and inspecting the completeness and
accuracy of the “skin” coverage (see the following example).
Two methods are available on the ACMODL entry for calculating the interface [A] matrix.
Main Index
444 MD Nastran Dynamic Analysis User’s Guide
BW Method of Fluid-Structure Analysis
1 2 3 4 5 6 7 8 9 10
ACMODL INTER INFOR FSET SSET NORMAL METHOD SKNEPS DSKNEPS
INTOL ALLSSET SRCHUNIT
Example
ACMODL .25
For the BW method, no ACMODL entry is necessary if defaults are used (suggested for first run). A
special license is required for this entry.
To use the old CP method the METHOD field (field 7) must be specified as CP:
1 2 3 4 5 6 7 8 9 10
ACMODL INTER INFOR FSET SSET NORMAL METHOD
Example
ACMODL CP
In most cases, the BW (Body in White) method provides a good fluid/structure interface matrix.
However, changes in the interface can be made by either adding or deleting structural elements in the
interface calculation.
A set of utility programs are provided to aid in viewing of the elements in the interface.
BW Coupling Algorithm
The purpose of this algorithm is to overcome the difficulties of many structural elements of various sizes
overlapping each other. Also, this algorithm should be somewhat insensitive to small inadvertent holes
in the structure.
The algorithm proceeds as follows:
1. Use the current search algorithm to locate the fluid free faces and the corresponding structural
element faces.
2. For a fluid free face and its list of structural element faces (that were determined by boxing normal
to the fluid element) do as follows:
a. For each fluid free face establish a face coordinate system.
b. Determine the resultant pressure force for each grid on the fluid element by the relationship
Ri = N f dS p i = 1 ; p j = 0 i = 1 N grid/elem
S
Main Index
CHAPTER 11 445
Coupled Fluid-Structure Analysis
c. Resolve this resultant pressure force for a unit grid pressure to the grids of the fluid element
by the expression (determined by virtual work)
T
Fi = Nf N f dS p i
S
d. Using the origin of the free fluid face, determine the center of pressure X P Y P . The
i i
relationship will be of the form
grids
Fi
XP =
i
----- X j – X 0
Ri
j
grids
Fi
YP =
i ----- Y j – X 0
Ri
j
e. Using rigid relations to consider only a unit motion normal to the fluid face with the
appropriate moment relationships, determine the resulting load distribution at the grids of
each of the structural elements. The area of each structural element projected normal to the
fluid element will be used as a weighting function. The expression is of the form:
R
T i –1
Fj = W R R W R 0
0
where F j is the vector of resulting load distribution at the grids of each of the j structural
elements. W is a diagonal weighting matrix. R is the rigid transformation matrix.
f. Loop over each grid of the fluid element and accumulate the forces at the structural grids.
g. Note, with this algorithm, we do not worry if a structural element is only partially wetted by
the fluid. We always require rigid body equilibrium.
3. Repeat for the next fluid element and its associated group of structural elements. Accumulate the
forces at the structural grids.
Main Index
446 MD Nastran Dynamic Analysis User’s Guide
BW Method of Fluid-Structure Analysis
Example
Figure 11-15
Main Index
CHAPTER 11 447
Coupled Fluid-Structure Analysis
Figure 11-16
The relative location of the fluid face and the structural plates is shown:
Figure 11-17
Main Index
448 MD Nastran Dynamic Analysis User’s Guide
BW Method of Fluid-Structure Analysis
When the model is run, and PARAM,SKINOUT,PUNCH is used, the fluid and structure interface
elements and grids found by the search are written to the acoustic1.pch file shown:
$ ********************
$ FSI - IDS OF FLUID ELEMENTS AT INTERFACE
$ ********************
$ This bulk data entry may by referenced from the FSET field on the
$ ACMODL bulk data entry (with the ELEMENTS option) to remove unwanted
$ fluid faces from the fluid-structure interface.
SET1 1 5001 5001 5001 5002 5002 5003 5003 5004
$ ********************
$ FSI - IDS OF STRUCTURE ELEMENTS AT INTERFACE
$ ********************
$ This bulk data entry may by referenced from the SSET field on the
$ ACMODL bulk data entry (with the ELEMENTS option) to remove unwanted
$ structure faces from the fluid-structure interface.
SET1 2 1041 1037 1038 1040 1039 1083 1094
1092 1091 1090 1082 1095 1093 1063 1081
1088 1064 1065 1062 1066 1053 1054 1055
1087 1052 1056 1057 1086 1051 1096 1058
1019 1028 1085 1073 1050 1072 1076 1084
1020 1070 1059 1013 1077 1071 1060 1012
1010 1061 1009 1008 1046 1047 1007 1006
1044 1045
$ ********************
$ FSI - IDS OF FLUID GRIDS AT INTERFACE
$ ******************** $
This bulk data entry may by referenced from the FSET field on the
$ ACMODL bulk data entry (with the GRIDS option) to remove unwanted
$ fluid faces from the fluid-structure interface.
SET1 3 5001 5002 5003 5004 5005 5006 5007
5008 5009 5010 5011 5012 5013 5016
$ ********************
$ FSI - IDS OF STRUCTURE GRIDS AT INTERFACE
$ ********************
$ This bulk data entry may by referenced from the SSET field on the
$ ACMODL bulk data entry (with the GRIDS option) to remove unwanted
$ structural faces from the fluid-structure interface.
SET1 4 1004 1005 1006 1007 1008 1009 1010
1017 1018 1027 1028 1029 1030 1031 1032
1040 1041 1042 1045 1047 1048 1049 1050
1051 1051 1053 1054 1055 1056 1057 1058
1059 1060 1061 1062 1063 1064 1065 1066
1071 1072 1073 1074 1075 1078 1079 1080
1081 1082 1084 1085 1086 1087 1088 1093
1094 1095 1096 1097 1098 1099 1101 1102
1103 1104 1105 1106 1107 1108 1109 1110
1111 1112 1113 1114 1115 1116 1117 1118
1119 1120 1121 1122 1123 1125 1126 1127
1128 1129
These element and grid SET1 Bulk Data entry lists serve two purposes. First, they can be used in a finite
element preprocessor as an accuracy check by graphically displaying which elements are used in the
analysis, and second they can limit or expand what elements or grids are used in fluid/structure interface.
To check the fluid/structure interface, MSC supplies a perl program named “acnaspat.pl” which reads the
above .pch file and writes a MSC.Patran session file. The perl script is located at “/msc2010/util”. To run
the script on UNIX, type, for instance:
/msc2010/util/acnaspat.pl acoustic1.pch
Main Index
CHAPTER 11 449
Coupled Fluid-Structure Analysis
which creates a MSC.Patran session file named “acoustic1_flstgrp.ses”, which when run in MSC.Patran
will create two groups:
Figure 11-18
Main Index
450 MD Nastran Dynamic Analysis User’s Guide
BW Method of Fluid-Structure Analysis
Which structural elements to be included in the fluid/structure interface can be adjusted in one of two
ways. The first is by increasing the size of the search box by changing the defaults on the NORMAL,
INTOL, and SKNEPS fields of the ACMODL Bulk Data entry.
INTOL L
• NORMAL L - Search box height in positive normal direction from fluid face, where L is
smallest fluid face side length,
• INTOL L - height in negative normal direction,
• 1 + SKNEPS D - in-plane width added to fluid face dimensions where D is the distance
from the center of the fluid face to the grid point.
If the search cannot find any structural faces within the search box, it will widen the search box by using
the value of the DSKNEPS field on the ACMODL Bulk Data entry.
Structural faces, whose centroids fall within the search box, are included in the fluid/structure interface.
The effect of the structural face in the fluid/structure interface matrix is weighed according to areas,
angles and distances between it and the fluid face that it found.
Main Index
CHAPTER 11 451
Coupled Fluid-Structure Analysis
In this case, looking at the interface elements, the normal distance of the search box appears good, but
too many structural elements on the sides of the fluid faces are being selected. By adding an ACMODL
Bulk Data entry with SKNEPS decreased from the default of 0.5 to 0.25:
acmodl,,,,,,,0.25
you get the following interface elements:
Figure 11-20
The structural elements are closer to what is wanted, but there are still too many.
The second way to control which elements are used in the fluid/structure interface is to use the
INFOR=“elements” or “grids” and the FSET and SSET fields on the ACMODL Bulk Data entry.
First the elements not wanted are removed from the “acoustic2_struct” group in MSC.Patran. For this
example, any structure element with less than 50% of it's area projected onto the fluid faces is removed.
The final structural interface elements are shown:
Main Index
452 MD Nastran Dynamic Analysis User’s Guide
BW Method of Fluid-Structure Analysis
Figure 11-21
Other methods of structural interface element selection would also be used at this point such as:
• graphical selection
• searching by property or material type
In MSC.Patran, the “acoustic2_struct” group is then displayed alone, it's elements selected with the Erase
tool, the element list highlighted, and then copied and pasted into a file, “element1.txt” in this case. The
file's contents as shown below:
Elm 1008 1013 1044 1045 1050:1059 1062:1064 1070 1072 1082 1085 1090 1091
Another Perl program from the same install directory is now used to change the above MSC.Patran
element or node list format into a MD Nastran SET1 format. The program is copied to the local directory
and run as follows in a UNIX environment:
/msc2010/util/acpatnas.pl element1.txt
which creates a file named “element1.set1” in the following form:
SET1 1 1008 1013 1044 1045 1050 1051 1052
1053 1054 1055 1056 1057 1058 1059 1062
1063 1064 1070 1072 1082 1085 1090 1091
Main Index
CHAPTER 11 453
Coupled Fluid-Structure Analysis
The SET1 SID is changed from “1” to “2” so as not to conflict with the fluid element SET1 from the
“acoustic1.pch” file. When these two SET1 Bulk Data entries are put in the MD Nastran file along with:
acmodl,,elements,1,2
then only the above structural elements are used in the fluid/structure interface.
Figure 11-22 shows the fluid pressure distribution and the structural deflection.
Figure 11-22
Main Index
454 MD Nastran Dynamic Analysis User’s Guide
BW Method of Fluid-Structure Analysis
The model is loaded with acoustic power on the upper left hand corner of the fluid, thus the high pressure
in that area. The effect of the hole in the inner structure can be seen as a slight reduction in fluid pressure
at the hole.
The structural deformation shows that the outer structure is unaffected by the pressure, as expected.
This concludes the example.
Parallel Structural Element Meshes in the Search Box:
As noted above, the outer parallel structural surface did not have any elements selected for inclusion in
the interface, even though some were in the search box. This is because parallel surfaces are ignored by
using a connectivity technique. The first structural element face found in the search box is the one closest
to the searching fluid face. All elements connected to this first element are checked whether they are in
the search box. Once an element is found to be outside of the box, it is no longer used for the connectivity
check. This tends to eliminate parallel element meshes unless the connection is also in the search box, in
which case the parallel elements must be manually removed using the SET1 technique above.
A problem occurs with this technique if a connectivity discontinuity occurs in the search box. Only one
side of the discontinuity will be selected, causing elements to be missed in the interface search. The
workaround for this is to use the ALLSET=“YES” field on the ACMODL entry in which case all
structural elements or grids in the SET1 entries are forced to be included in the fluid/structure interface
whether the connectivity search finds them or not. You must also set INTER=“DIFF”,
INFOR=“ELEMENTS” or “GRIDS” and SSET must be properly filled out. This method forces an
interface between the listed structural elements and the nearest fluid faces.
Absolute Versus Relative Search Box Dimensions:
If SRCHUNIT=“ABS”, the box dimensions are equal to the value of NORMAL, INTOL, and SKNEPS,
rather than times a fluid face dimension. This is useful when the distance between the structural interface
and the fluid interface is relatively constant instead of relative to the fluid element size.
Main Index
CHAPTER 11 455
Coupled Fluid-Structure Analysis
• A2GG = ADMIG
• A2GG = ADMIG1, ADMIG2, ADMIG3
• A2GG = 1.5*ADMIG1, .5*ADMIG2
Main Index
456 MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
Ms 0 u·· s B s 0 u· s Ks A us Ps
+ + = (11-17)
– A M f p·· 0 B f p·
T 0 Kf p Pf
where:
M = mass
B = damping
K = stiffness
u = displacement
p = pressure
s = structure
f = fluid
A = interface matrix
P = load
us = s s (11-18)
p = f f (11-19)
where s are the uncoupled, undamped structural modes and f are the uncoupled, undamped, rigid-
wall acoustic modes. The vectors s and f are the modal amplitudes. Substituting these relations
into Eq. (11-17) and pre-multiplying by the modal matrices, we get the equation
·· · T
s K s s s A f s s Ps
T T T T
s M s s 0 s s Bs s 0 s
·· + · + = T (11-20)
–f A s Mf f f Bf f f f Kf f f f Pf
T T T T T
0 0
or
ms 0 ·· s ·
b 0 s k a s Qs
+ s + s = (11-21)
T ·· 0 b f · f 0 kf f
–a mf f Qf
Main Index
CHAPTER 11 457
Coupled Fluid-Structure Analysis
s i t
= = e (11-22)
f
· i t
= ie (11-23)
·· 2 i t
= – e (11-24)
For the bottom equation of Eq. (11-21), we get:
2 T 2
a s + – m f + i b f + k f f = Q f (11-25)
Define Z 2 as
2 –1
Z 2 = – m f + i b f + k f (11-26)
then
2 T
f = – Z2 s + Z2 Qf (11-27)
Pf = f f (11-28)
where f is the diagonalized vector of fluid modal amplitudes per excitation frequency.
where s is the diagonalized vector of structural modal amplitudes per excitation frequency.
Pl = f Z2 Qf (11-30)
2 T T
Pg i = – f Z2 f Ab s s (11-32)
Main Index
458 MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
where
T
Ab (11-33)
are the columns extracted from the b-th boundary panel for panel grid i and s are the rows of the
structural modal matrix corresponding to panel grid i.
An adjoint method is used to compute acoustic structural modal participation factors, acoustic panel
participation factors and acoustic grid participation factors, resulting in significant performance
improvements.
Theory
In a linear structural dynamic analysis, the results at a degree of freedom considered are the sum of the
contributions of the different modes, e.g., the accelerations at a degree of freedom considered are the sum
of the accelerations due to the responses of the different structural modes. These contributions are called
structural modal participation factors or modal contribution fractions. The degrees of freedom
considered are called response degrees of freedom. Structural modal participation factors allow to
identify the structural modes that have the largest influence on the response.
Likewise, in an acoustic analysis, the pressure at a grid point considered is the sum of the pressures due
to the responses of the different fluid modes. These contributions are called acoustic fluid modal
participation factors. Acoustic fluid modal participation factors allow to identify the fluid modes that
have the largest influence on the response.
In a coupled analysis, the pressure at a response degree of freedom is the sum of the pressure due to the
acoustic sources in a rigid cavity, and the pressure due to the acceleration of the fluid-structure interface,
the so-called wetted surface. The pressure due to the acoustic sources in a rigid cavity is called the load
participation factor.
Main Index
CHAPTER 11 459
Coupled Fluid-Structure Analysis
The acceleration of the fluid-structure interface is the sum of the accelerations due to the responses of
the different structural modes, obtained from a coupled analysis. These contributions are called acoustic
structural modal participation factors. One acoustic structural modal participation factor equals the
pressure at the response degree of freedom if there are no acoustic sources, and if the acceleration of the
wetted surface consists of the response of one mode only. Thus, the acoustic structural modal
participation factors allow to identify the structural modes that have the largest influence on the pressure
at the grid point considered. In the absence of acoustic sources within the cavity, the acoustic structural
modal participation factors sum up to the total pressure at the response degree of freedom.
Modal participation factors, by their nature, are useful only in the low-frequency range where the
resonance frequencies are well separated, and the response is dominated by a small number of modes.
On the contrary, geometric participation factors are useful also at higher frequencies where the response
has contributions from a large number of modes. There are two types of geometric participation factors,
namely panel participation factors and grid participation factors.
A panel is a set of grid points of the wetted surface. The panel participation factor is that pressure at the
grid point considered that results from the accelerations of the grid points of the panel only, with all other
grid points of the wetted surface kept fixed. Thus, panel participation factors allow to identify the regions
of the wetted surface that have the largest influence on the acoustic pressure at the grid point considered.
Panel participation factors usually do not sum up to the total acoustic pressure. This is only the case if
the panels do not overlap, and if their union equals the complete wetted surface.
The accelerations of the grid points of the panel are the sum of the accelerations due to the different
structural modes. The pressure due to the accelerations at the grid points of the panel that are due to one
mode only is called the acoustic panel modal participation factor. Acoustic panel modal participation
factors sum up to the panel participation factors.
Finally, if the panels consist of one grid point only, acoustic grid participation factors are obtained. For
each structural grid point of the wetted surface, there are six acoustic grid participation factors, i.e. the
pressures at the response degree of freedom due to the accelerations of the six degrees of freedom of this
grid point. The acoustic grid participation factors depend on the mesh size. Thus, their absolute value has
no physical meaning. However, if the mesh of the wetted surface is of comparable size everywhere, the
acoustic grid participation factors allow to quickly identify the regions that make the largest contribution
to the acoustic pressure at the grid point considered, and thus help to define the panels.
Participation factors are complex quantities, summing up to the complex response. Thus, if they are
divided by the response, they sum up to one. The participation factors divided by the response are called
normalized participation factors. Normalized participation factors are complex quantities, too.
The real parts of the normalized participation factors sum up to one whereas the imaginary parts sum up
to zero. Thus, the real part of a normalized participation factor is that part of the participation factor that
is in phase with the total response, divided by the magnitude of the total response. It is called the modal
fraction. The phase of a normalized participation factor is the phase of the participation factor relative to
the total response.
If the modal fraction is multiplied by the magnitude of the total response, the projected participation
factor is obtained. It is that part of the participation factor that is in phase with the total response. The
projected participation factors sum up to the total magnitude.
Main Index
460 MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
Input
Computation and output of modal participation factors is controlled by the PFMODE Case Control
command:
PFMODE STRUCTURE PRINT, PUNCH REAL or IMAG
FLUID PLOT PHASE
ALL
SORT = sorttype KEY = sortitem ITEMS =
itemlist
ALL ALL
FLUIDMP = m f STRUCTMP = m s
NONE NONE
ALL ALL
PANELMP = setp SOLUTION = setf FILTER = fratio
NONE NONE
setdof
NULL = ipower =
NONE
Examples
Main Index
CHAPTER 11 461
Coupled Fluid-Structure Analysis
Computation and output of panel participation factors is controlled by the PFPANEL Case Control
command:
ALL
PFPANEL PRINT, PUNCH REAL or IMAG PANEL = -----------
PLOT PHASE setp
ALL
SORT = sorttype KEY = sortitem ITEMS =
itemlist
ALL
SOLUTION = setf FILTER = fratio NULL = ipower
NONE
= setdof
NONE
Example
ALL
PFGRID PRINT, PUNCH REAL or IMAG GRIDS =
PLOT PHASE setg
ALL
setdof
SOLUTION = setf =
NONE
NONE
Example
SET 20 = 11217
SET 90 = 25., 30., 35.
PFGRID(SOLUTION = 90) = 20
Compute acoustic grid participation factors for the pressure at grid point 11217 at excitation frequencies
25Hz, 30Hz and 35Hz.
Main Index
462 MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
Detailed descriptions of the PFMODE, PFPANEL and PFGRID Case Control Commands can be found
in the MD Nastran Quick Reference Guide.
Panels are defined in the Bulk Data Section using PANEL Bulk Data entries which reference SET1 or
SET3 Bulk Data entries.
• SET1 Bulk Data entries list the grid points of the panels.
• SET3 Bulk Data entries with option ELEM list the elements of the panels. The panels consist of
all grid points associated to these elements.
• SET3 Bulk Data entries with option PROP list the property identifiers of the elements of the
panels. The panels consist of all grid points associated to the elements with the property
identifiers defined.
The FLSTCNT Case Control command combines miscellaneous fluid-structure control parameters in one
place.
YES PEAK
FLSTCNT ACSYM = ----------- ACOUT = ---------------
NO RMS
1.0 YES
PREFDB = --------- ASCOUP = -----------
prp NO
NONE
PUNCH
SKINOUT =
PRINT
ALL
Output
Three different formats are used for printed output. The format for modal participation factors is similar
to the format obtained with the MCFRACTION command. The header contains the type of the
participation factor, information on the grid point and degree of freedom considered, the total response,
information on excitation frequency, subcase and load and information on the maximum modal response,
the sort method and the filter. In case of acoustic panel modal participation factors, the header contains
the name of the panel considered and the panel response instead of the total response.
Main Index
CHAPTER 11 463
Coupled Fluid-Structure Analysis
The format for acoustic panel participation factors is the same as for modal participation factors, except
that the first two columns contain the panel names.
Load participation factors are included in the acoustic structural modal participation factors and the
acoustic panel participation factors, with a mode number of 0 and a panel name –LOAD-. Acoustic
structural modal participation factors, together with the load participation factor, sum up to the total
response.
Acoustic grid participation factors use the output format of complex displacements. Both real and
imaginary part or magnitude and phase format are available. There is one output per excitation frequency
and fluid grid point.
Main Index
464 MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
Example
The numbers printed indicate the normalized acoustic pressure at the selected grid point due to the
accelerations at the corresponding degrees of freedom.
Guidelines
The amount of output produced may be very large. This is especially true for acoustic panel mode
participation factors and for acoustic grid participation factor (e.g., the number of data produced for
acoustic panel modal participation factors equals the number of subcases times the number of excitation
frequencies times the number of response degrees of freedom times the number of panels times the
number of structural modes). Consequently, output should be limited to a small number of excitation
frequencies and to a small number of response degrees of freedom.
Main Index
CHAPTER 11 465
Coupled Fluid-Structure Analysis
Panels are defined for the front window, the rear wall, the left and the right side, the top and the bottom.
See Install_dir/mdxxxx/doc/dynamics/bd11pf8a.dat for the complete input and Listing 11-2 for a partial
input.
Main Index
466 MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
$
$ Illustrates use of
$ o Acoustic Fluid Modal Participation Factors
$ o Acoustic Struct. Modal Part. Factors
$ o Acoustic Panel Part. Factors
$ o Acoustic Grid Part. Factors
$
$ =======================================================
$
Main Index
CHAPTER 11 467
Coupled Fluid-Structure Analysis
DISP(PHAS,SORT2) = 20
$
PFMODE(FLUID, FLUIDMP=ALL, STRUCTMP=ALL,
SOLUTION=90, SORT=ABSD) = 20
PFPANEL(SOLUTION=90, SORT=ABSD) = 20
PFGRID(SOLUTION=90) = 20
$
OUTPUT(XYPLOT)
XPAPER=29.
YPAPER=21.
XGRID=YES
YGRID=YES
XTITLE = Frequency
YTITLE = Pressure
XYPLOT DISP RESPONSE / 11217(T1)
$
BEGIN BULK
$
$ Request OP2 for PATRAN
PARAM, POST, -1
$
$ Define Structural Damping
PARAM, G, 0.02
$
$ Define Fluid Damping
PARAM, GFL, 0.002
$
$ Define Reference Pressure for dB (in Pa)
PARAM, PREFDB, 2.8284-5
$
PARAM, GRDPNT, 0 $ Request Weight Output
$
$ Structural and Acoustic Modes up to 300Hz
EIGRL, 1,,300.
EIGRL, 2,,300.
$
$ Excitation Frequencies
FREQ1, 100, 10., 5., 18
$
$ Constraints
SPC1, 300, 123456, 1, 7, 29, 35
$
$ Excitation
RLOAD1, 200, 210,,, 220,, ACCE
SPCD, 210, 1, 3, 1., 7, 3, 1.
SPCD, 210, 29, 3, 1., 35, 3, 1.
TABLED1, 220
, 0., 1., 1000., 1., ENDT
$
$ Nonmatching Fluid-Structure Interface
ACMODL, DIFF
$
$ Panels
$
SET3, 101, ELEM, 127, THRU, 162, 667, THRU, 738
SET3, 201, ELEM, 37, THRU, 72, 739, THRU, 810
SET3, 301, ELEM, 331, THRU, 384
SET3, 401, ELEM, 25, THRU, 36, 73, THRU, 126
SET3, 501, ELEM, 271, THRU, 294, 601, THRU, 612
SET3, 601, ELEM, 1, THRU, 24, 455, THRU, 478,
Main Index
468 MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
A C O U S T I C S T R U C T U R E M O D A L P A R T I C I P A T I O N F A C T O R S
GRID POINT = 11217, TOTAL RESPONSE (R/I) = -9.42534E-01 / -1.03865E-02, (M/P) = 9.42591E-01 / 180.63
LOAD FREQUENCY = 4.00000E+01, (SUBCASE 1, DLOAD = 200)
MAXIMUM MODAL RESP = 1.13741E+00 FOR MODE ID = 9, SORTKEY = FRACTION, SORT = ABS VALUE DESCENDING, FILTER = 1.00000E-03
MODE NATURAL MODAL RESPONSE MODAL RESPONSE PROJECTION REL. MODAL SCALED RESPONSE
ID FREQ (HZ) REAL IMAGINARY MAGNITUDE PHASE MAGNITUDE PHASE FRACTION MAGNITUDE
9 8.30861E+01 1.13739E+00 -6.25017E-03 1.13741E+00 359.69 -1.13725E+00 179.05 -1.20652E+00 -9.99864E-01
8 8.18779E+01 -6.32888E-01 4.67901E-03 6.32905E-01 179.58 6.32798E-01 -1.05 6.71338E-01 5.56350E-01
3 6.82472E+01 -6.25327E-01 6.91886E-03 6.25365E-01 179.37 6.25213E-01 -1.27 6.63291E-01 5.49681E-01
21 1.33089E+02 -5.71325E-01 1.54503E-03 5.71327E-01 179.85 5.71273E-01 -0.79 6.06066E-01 5.02258E-01
84 2.42589E+03 -4.37526E-01 -1.03197E-04 4.37526E-01 180.01 4.37501E-01 -0.62 4.64147E-01 3.84647E-01
31 1.60995E+02 3.90763E-01 -2.29250E-04 3.90763E-01 359.97 -3.90737E-01 179.34 -4.14534E-01 -3.43532E-01
14 1.07144E+02 -2.38134E-01 1.45832E-03 2.38139E-01 179.65 2.38104E-01 -0.98 2.52605E-01 2.09338E-01
79 4.55425E+02 2.27102E-01 -8.55791E-05 2.27102E-01 359.98 -2.27088E-01 179.35 -2.40918E-01 -1.99653E-01
5 7.74596E+01 1.86438E-01 -1.13864E-03 1.86441E-01 359.65 -1.86414E-01 179.02 -1.97768E-01 -1.63894E-01
1 3.61553E+01 -1.61339E-01 -1.79486E-02 1.62334E-01 186.35 1.61527E-01 5.72 1.71364E-01 1.42013E-01
13 1.05870E+02 -1.54939E-01 5.68473E-04 1.54940E-01 179.79 1.54923E-01 -0.84 1.64359E-01 1.36207E-01
32 1.70901E+02 -1.41683E-01 1.99820E-04 1.41683E-01 179.92 1.41672E-01 -0.71 1.50301E-01 1.24557E-01
29 1.51451E+02 1.20258E-01 -1.69764E-04 1.20258E-01 359.92 -1.20249E-01 179.29 -1.27573E-01 -1.05722E-01
50 2.09592E+02 -9.86342E-02 1.15611E-04 9.86342E-02 179.93 9.86269E-02 -0.70 1.04634E-01 8.67119E-02
11 9.45225E+01 9.81448E-02 -3.61858E-04 9.81454E-02 359.79 -9.81348E-02 179.16 -1.04112E-01 -8.62792E-02
80 4.99581E+02 -9.78103E-02 5.94714E-05 9.78103E-02 179.97 9.78037E-02 -0.67 1.03760E-01 8.59881E-02
36 1.79440E+02 -7.51393E-02 -7.29040E-05 7.51393E-02 180.06 7.51355E-02 -0.58 7.97116E-02 6.60584E-02
43 2.01720E+02 7.43311E-02 -5.88513E-05 7.43311E-02 359.95 -7.43259E-02 179.32 -7.88528E-02 -6.53467E-02
23 1.33661E+02 -7.42764E-02 1.22552E-04 7.42765E-02 179.91 7.42705E-02 -0.73 7.87940E-02 6.52980E-02
A C O U S T I C P A N E L P A R T I C I P A T I O N F A C T O R S
GRID POINT = 11217, TOTAL RESPONSE (R/I) = -9.42534E-01 / -1.03865E-02, (M/P) = 9.42591E-01 / 180.63
LOAD FREQUENCY = 4.00000E+01, (SUBCASE 1, DLOAD = 200)
MAXIMUM PANEL RESP = 2.63941E+00 FOR PANEL = BOTTOM , SORTKEY = FRACTION, SORT = ABS VALUE DESCENDING, FILTER = 1.00000E-
03
PANEL NAME PANEL RESPONSE PANEL RESPONSE PROJECTION REL. PANEL SCALED RESPONSE
REAL IMAGINARY MAGNITUDE PHASE MAGNITUDE PHASE FRACTION MAGNITUDE
BOTTOM -2.63940E+00 8.05866E-03 2.63941E+00 179.83 2.63915E+00 -0.81 2.79989E+00 9.99901E-01
FRONT 1.11163E+00 -4.59632E-03 1.11164E+00 359.76 -1.11152E+00 179.13 -1.17921E+00 -4.21123E-01
TOP 1.09365E+00 -3.39035E-03 1.09365E+00 359.82 -1.09354E+00 179.19 -1.16015E+00 -4.14314E-01
REAR -2.33400E-01 -2.06344E-02 2.34310E-01 185.05 2.33613E-01 4.42 2.47841E-01 8.85095E-02
RIGHT -6.01577E-02 5.20980E-03 6.03829E-02 175.05 6.00966E-02 -5.58 6.37568E-02 2.27690E-02
LEFT -6.01573E-02 5.20977E-03 6.03825E-02 175.05 6.00963E-02 -5.58 6.37564E-02 2.27688E-02
-LOAD- -3.82035E-02 6.22966E-06 3.82035E-02 179.99 3.82011E-02 -0.64 4.05278E-02 1.44734E-02
The acoustic structural mode participation factors show that the largest contribution comes from the 9th
structural mode which has a resonance at 83.09Hz. The acoustic panel participation factors show that the
rear wall and the bottom make the largest positive contribution whereas the other panels make smaller or
negative contributions.
Main Index
CHAPTER 11 469
Coupled Fluid-Structure Analysis
XY Plots
The following “yvtype”s are on the XYPLOT Case Control commands.
• FMPF (mode_id/frequency_id)
• Fluid mode participation factors
• GMPF (mode_id/frequency_id, panel_name, panel_grid_id
• Panel grid Mode participation factors
• LMPF
• Load mode participation factors
• PMPF (mode_id/frequency_id, panel_name)
• Panel mode participation factors
• SMPF (mode_id/frequency_id)
• Structural mode participation factors
Random Analysis
A “AUTO” and “PSDF” XYPLOT random “ptype”s support the new acoustic modal participation
“yvtype”s:
• FMPF
• SMPF
• PMPF
Main Index
470 MD Nastran Dynamic Analysis User’s Guide
Acoustic XY Plots, Random, and Restarts
• LMPF
• GMPF
Root mean square random results may be a convenient way to calculate the effect of a mode on a fluid
response across a frequency range. Weigh the effect of the frequencies on the RANDPS Bulk Data entry
and use one of the above “ptype” in an PSDF XYPLOT command and a root mean square result will be
calculated.
Restart
The base line and restart input files below represents a simple fluid-structure problem for the purpose of
showing salient input.
Model Description
The model consists of a CQUAD4 and four CELAS2 to represent structure, a CHEXA to represent the
fluid, and a CHACAB structural element to represent an acoustic absorber. In the baseline Case Control,
note the use of the FLSFSEL to select lower limits on the fluid and structure frequency range modes. Also
note the use of FLSTCNT to set the reference pressure. Some fictitious random entries are input to
demonstrate the XYPLOT features for structural and fluid grids. On large models of automobiles and
aircraft, the modal solutions, especially on the structure, can take significant time and disk space. Also
modal participation can, on large models, generate gigabytes of data. Thus no participation information
is requested for the base line run, as participation results will be requested on the restart run. Since a data
base is required, the MD Nastran base line submittal contains SCR=NO.
In the restart file, note the use of the ASSIGN and RESTART commands. Notice, also, that Case Control
commands such as FLSFSEL, FLSTCNT, FREQ, SPC, METHOD(STRUCT) and METHOD(FLUID)
are copied over from the base line run. To change or omit these would cause the restart logic to attempt
to do the modal analysis over again. To calculate mode participation information, the FLSPOUT
command is added to case control along with three set definitions directly above.
To define random data appropriate to modal participation, a different RANDOM command is provided.
Also, a different set of XYPLOT commands are provided that are to be used for plotting of participation
results.
For the restart run, the original bulk data entries are removed and some PANEL and SET1 entries are
added to define panels for the participation run. Also, new RANDPS and RANDT1 entries are added for
the participation analysis. Note that no TABRND1 is included because the original one from the base line
run is used.
See Listing 11-5 for a printout of the base line file and see
Install_dir/mdxxxx/doc/dynamics/bd11acxybl.dat for the base line file. The restart file is shown as
Listing 11-6 and see Install_dir/mdxxxx/doc/dynamics/bd11acxx.dat for the data file.
Main Index
CHAPTER 11 471
Coupled Fluid-Structure Analysis
Main Index
472 MD Nastran Dynamic Analysis User’s Guide
Acoustic XY Plots, Random, and Restarts
Main Index
CHAPTER 11 473
Coupled Fluid-Structure Analysis
$
$ FLUID ELEMENTS
CHEXA 1000 15 100 104 124 120 200 204
224 220
$
$ FLUID PROPERTIES
PSOLID 15 25 1 PFLUID
MAT10 25 1. 0.1
$
$ SPC ALL THE STRUCTRUAL POINTS NOT TO ROTATE ABOUT X OR Z
SPC1 1313 1246 41 THRU 61
SPC1 1313 146 65
SPC113131200
$ SPC THE END STRUCT. POINTS NOT TO ROATE ABOUT Y (STATED IN THE PROBLEM)
SPC1 1313 5 41 61 65 45
$ 2 3 4 5 6 7 8 9 0
$ DEFINE ACOUSTIC ABSORBER
CHACAB 9 9 41 45 65 61 141 145
+ 265 261
PACABS 9 YES 11 22 33 1. 1000.
TABLED1 11
+ 2. .15 50. .15 100. .15 150. .15
+ 200. .15 250. .15 300. .15 ENDT
TABLED1 22
+ 2. -40. 50. 14. 100. 30. 150. 46.
+ 200. 62. 250. 78. 300. 94. ENDT
TABLED1 33
+ 2. 1. 50. 1. 100. 1. 150. 1.
+ 200. 1. 250. 1. 300. 1. ENDT
$ RANDOM INPUT
RANDPS,10,100,100,4.,,1
RANDPS 10,200,200,8.,,1
RANDPS 10,100,200 2.,,1
TABRND1,1
+,0.,100.,300.,100.,ENDT
RANDT1,10,4,0.,.004
$
ENDDATA
Listing 11-6 Restart Run
ASSIGN MASTER='bd11acxybL.MASTER'
RESTART VERSION=1 KEEP
SOL 111 $ MODAL frequency response
CEND
TITLE= FLUID-STRUCTURE INTERACTION
$
SET 29 = bndy
SET 224 = 224
FLSPOUT fluidmp=all,structmp=all,panelmp=29,gridmp=all,
gridfmp=224,seps=0.,ars=0.,
outfmp=all,outsmp=all,psort=(absolute,descending)
FLSFSEL LFREQFL=1.0,LFREQ=2.0
FLSTCNT PREFDB=20.-6 , SKINOUT=ALL
$
DISP(PLOT)=ALL
RANDOM=50
$
SPC =1313
FREQ=200
Main Index
474 MD Nastran Dynamic Analysis User’s Guide
Acoustic XY Plots, Random, and Restarts
METHOD(struct)=30
METHOD(fluid)=20
$
SUBCASE 100
DLOAD=100
$
SUBCASE 200
DLOAD=200
$
OUTPUT(XYPLOT)
XGRID=YES
YGRID=YES
$
$-1 plt 2nd sc
XTITLE=EXCITATION FREQUENCY FROM 2.5 to 250 hertz sc 200 #1
YTITLE=fluid mode pf at fluid point 224 for mode 2
XYPLOT,XYPEAK FMPF(2) MODE 200 /224
$-2,3 plt 1st & 2nd sc
XTITLE=EXCITATION FREQUENCY FROM 2.5 to 250 hertz default sc 100 200 #2 and #3
YTITLE= load pf at fluid point 224
XYPLOT,XYPEAK LMPF MODE /224
$-4 1st sc
XTITLE= natural modes from .4 to 2 hertz #4
YTITLE= panel mode pf at fluid point 224 for mode 9 sc 100
XYPLOT,XYPEAK PMPF(9,BNDY) MODE 100 /224
$-5 1st sc
XTITLE= natural modes from .4 to 2 hertz sc 100 #5
YTITLE=structure mode pf at fluid point 224 for mode 3
XYPLOT,XYPEAK SMPF(3) MODE 100 /224
$-6,7 1st & 2nd sc
XTITLE= natural modes from .4 to 2 hertz default sc 100 200 #6 and #7
YTITLE=grid panel mode pf at fluid point 224 for mode 9 grid 61
XYPLOT,XYPEAK GMPF(9,BNDY,61) MODE /224
$-8,9 1st & 2nd sc
XTITLE= natural modes from .4 to 2 hertz try 100,200 #8 and #9
YTITLE=grid panel mode pf at fluid point 224 for mode 9 grid 61
XYPLOT,XYPEAK GMPF(9,BNDY,61) MODE 100,200 /224
$-10
XTITLE=frequency #10
YTITLE=psd mode participation for grid 224
XYPLOT,XYPEAK FMPF(3) PSDF /224
$-11
XTITLE=frequency #11
YTITLE=auto mode participation for grid 224
XYPLOT,XYPEAK FMPF(3) AUTO /224
$-12
XTITLE=frequency #12
YTITLE=psd grid mode participation grid 224 at grid 61 first decend mode
XYPLOT,XYPEAK GMPF(9,BNDY,61) PSDF /224
$-13
XTITLE=frequency #13
YTITLE=auto grid mode participation grid 224 at grid 61 first decend mode
XYPLOT,XYPEAK GMPF(9,BNDY,61) AUTO /224
$ --------------------------------------
BEGIN BULK
$ DEFINE STRUCTURAL PANELS
PANEL,BNDY,777
SET1,777,SKIN
$ NEW RANDOM INPUT FOR PARTICIPATION
RANDPS,50,100,100,2.,,1
Main Index
CHAPTER 11 475
Coupled Fluid-Structure Analysis
RANDPS,50,200,200,5.,,1
RANDPS,50,100,200,1.,,1
RANDT1,50,6,0.,.009
$
ENDDATA
The following XYPLOT commands create the plot below:
XTITLE=FREQUENCY #12
YTITLE=PSD GRID MODE PARTICIPATION FOR GRID 224 FIRSTONE AT GRID 61 FIRST
DECEND
XYPLOT,XYPEAK GMPF(9,BNDY,61) PSDF /224
Figure 11-23
This is the PSD of structural modal participation of grid point 61, mode 9, on fluid grid 224. In order to
get structural grid participation factors, the structural grid must be defined on a panel and called with
GMPF in the XYPLOT command. Panels are defined with PANEL and SET1 Bulk Data entries. In this
case, the “SKIN” option on the SET1 Bulk Data entry selects all of the structural interface grids to be in
panel “BNDY”:
PANEL BNDY 777
SET1 777 SKIN
Main Index
476 MD Nastran Dynamic Analysis User’s Guide
Acoustic XY Plots, Random, and Restarts
For PSD curves, the root-mean-square value across frequencies is calculated and printed in the .f06 file:
X Y - O U T P U T S U M M A R Y ( A U T O O R P S D F )
PLOT CURVE FRAME CURVE ID./ RMS NO. POSITIVE XMIN FOR XMAX FOR YMIN FOR X FOR YMAX FOR X FOR*
TYPE TYPE NO. PANEL : GRID ID VALUE CROSSINGS ALL DATA ALL DATA ALL DATA YMIN ALL DATA YMAX
This RMS value, 5.075398E+6, can be used as a measure of the combined relative effect of mode 9 across
all the frequencies.
Main Index
CHAPTER 11 477
Coupled Fluid-Structure Analysis
i t ·
Qe = u n dS (11-34)
s
Also, similar to the Maxwell-Betti reciprocity law in structures, they define a principle of acoustic
reciprocity which states that in an unchanging environment if the locations of a small receiver and small
source are interchanged, the received signal will remain the same.
With these definitions, they arrive at the expression for power P f in terms of complex source strength,
which for a pulsating sphere is a real value Q . The resulting expression as used in MD Nastran is:
1 8cP f
Q = -------- --------------------- (11-35)
2f
Acoustic power is applied by referencing the ACSRCE Bulk Data entry from a DLOAD Case Control
command or Bulk Data entry. The ACSRCE entry specifies the power from a TABLEDi Bulk Data entry
and also points to DAREA Bulk Data entries to specify the degree of freedom to be loaded as well as a
scale factor. In Eq. (11-35) it should be noted that source strength is proportional to the square root of
power.
The power is translated to source strength from the values on the TABLEDi entries. This source strength
is then distributed linearly to the degrees of freedom using the factors on the DAREA entries. Acoustic
engineers need to know that when using the ACSRCE Bulk Data entry that if power is to be scaled, do
it within the TABLEDi. If source strength is to be scaled, do it on DAREA. This may be confusing to
the beginner, but actually follows the definition of power point sources given previously.
A way to avoid this confusion is to change your power to source strength using Eq. (11-35), and apply it
as needed on the RLOADi and TLOADi Bulk Data entries.
Main Index
478 MD Nastran Dynamic Analysis User’s Guide
Rigid Porous Absorber - MAT10
There is a capability to model basic rigid skeleton porous absorber characteristics in acoustic response
analysis. The capability allows some types of absorbent material to be modelled, such as vehicle seat
structures or lining materials which exhibit stiff carcasses. The absorber material is considered using an
equivalent fluid analogy and so is modelled in the same manner as a typical fluid, using solid CHEXA,
CPENTA or CTETRA elements, the GRID points for which have their CD field set to -1. The porous
absorber elements reference a PSOLID property entry with field 8 set to PFLUID in the usual way, with
field 3 of the PSOLID entry referencing a MAT10 entry, where field 7 pertains to porous absorber
materials.
Porous Materials
If a material is not completely solid, but contains voids or air pockets, then it is said to be porous. There
are a great many materials which exhibit porosity, the term given to the degree of openness of the
material, including materials generally considered “solid” like brick or stone. If the voids in the
substance are large enough, they may form an interconnected maze of passage ways allowing air (or any
other fluid) to pass through the material. However, depending on the degree of convolution in the
passage ways (known as tortuosity), the air will encounter some resistance as it passes through the
substance, requiring pressure to be exerted to push the air against the resistance. Sound waves striking
the material do not cause air to flow through the material, but they do cause local perturbations that exert
pressure and cause the air to move in the vicinity of the material; the oscillating movement of air caused
by the sound waves encounters resistance (called impedance) which uses up some of the sound energy
and damps the level of sound. This energy is eventually dissipated as heat.
If the porous material is enclosed in a frame which is considered as rigid, as will be the case for example
for a porous medium which has a high skeleton density or very large elastic modulus or weak fluid-
structure coupling, the porous material can be considered as an equivalent fluid with both density and
bulk modulus being complex frequency dependent parameters. It is possible to obtain values for these
parameters by empirical methods as introduced, for example, by Delany and Bazley1, methods which
have been widely used to describe sound propagation in fibrous materials.
In the equivalent fluid approach, the equation of motion reads
2
1 2
----- P + ------ P = 0
e Be
where e is the equivalent density, B e the equivalent bulk modulus, P the complex pressure amplitude
and the circular excitation frequency, in which it can be shown
1- = --1- 1 + iGE
----
e
and
Main Index
CHAPTER 11 479
Coupled Fluid-Structure Analysis
1 1 A 1 i
------ = --- – i ---- = --- 1 – -----
Be B B
if it is assumed that the parameters are not frequency dependent, a reasonable assumption for the study
of frequencies in a narrow band. Here, , B and GE are the values of RHO, BULK and GE
respectively of the MAT10 entry for the porous absorber material. A new field 7 has been added to the
MAT10 entry to allow the value of , the normalized admittance coefficient2, to be entered.
Inputs
1 2 3 4 5 6 7 8 9 10
MAT10 MID BULK RHO C GE ALPHA
The following may be used to calculate the equivalent fluid property values to be entered on the MAT10
entry starting from the complex density and complex bulk modulus describing the rigid porous absorber.
If a complex density and complex speed of sound are determined for the porous material, the complex
bulk modulus must first be calculated.
MAT10 Density
GE = – -----i
r
Bi
= -----
Br
Main Index
480 MD Nastran Dynamic Analysis User’s Guide
Rigid Porous Absorber - MAT10
Discussion
The implementation implies that if the complex density and bulk modulus are constant, the normalized
admittance coefficient is a function of frequency. However, frequency dependent is not supported, so
a reference frequency must be selected. Typically, this frequency will be either in the mid-range of the
desired frequency range to be studied, or will correspond to the frequency at a response peak of interest.
As frequencies progressively further away from the reference frequency are considered, the response
becomes increasingly subject to the limitations of the frequency independent formulation; the extent will
depend somewhat on the nature of the problem, and it may be necessary to study discrete frequency bands
in order to mitigate against this effect.
The use of a non-zero value in field 7 of the MAT10 entry causes the generation of a damping matrix
because the normalized admittance coefficient is multiplied by the imaginary operator i. Consequently,
the use of modal methods on the fluid are not appropriate and frequency response analysis must be carried
out using the direct method, at least for the fluid.
References
1. M.E. Delany & E.N. Bazley, Acoustical Characteristics of Fibrous Absorbent Materials,
National Physics Laboratory, Aerodynamics Division, NPL Aero Report Ac 37, March 1969.
2. J. Wandinger, Possible Implementations of Porous Absorbers in Nastran, MSC internal memo,
April 2006.
Example
Consider the following unbounded fluid (air) and porous absorber medium domains as in Figure 11-24.
An acoustic source is placed at the location indicated and the acoustic response (pressure) at the centre
of the fluid is monitored.
Porous medium
Air
Response
Acoustic source
Main Index
CHAPTER 11 481
Coupled Fluid-Structure Analysis
Density 1.225 + i0
Speed of sound 340.0 + i3.4
Porous Absorber
Porous Absorber
A frequency of 250 Hz was selected to calculate the values of alpha for air and the porous absorber.
Using the equations illustrated above, the following data is entered on the MAT10 entries.
MAT10 for air
Notice that the values of bulk modulus, GE damping coefficient and alpha are all negative; this is a
normal characteristic of the implementation.
Main Index
482 MD Nastran Dynamic Analysis User’s Guide
Rigid Porous Absorber - MAT10
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 12: Modal Augmentation Methods
Overview
Data Recovery for Modal Methods
Dynamic Response Predictions
Modal Versus Direct Frequency Response
Residual Vector
th
Main Index
484 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
In dynamic analysis, the size of the calculated output data could become several orders of magnitude
larger than the size of output for the equivalent statics analysis job. For instance, a typical run could
produce thousands of solution vectors. If caution is not taken, the cost of processing the outputs for
printing or plotting could exceed that of the entire solution process. Given these concerns, the dynamic
engineer should use the SET logic in the Case Control Section of the MD Nastran input file to reduce the
number of outputs. Wherever possible, use a graphical postprocessor to visualize the results.
This chapter describes methods for calculating physical results from a solution using modal degrees-of-
freedom. The default method is the so-called matrix method, which is the most effective technique for
the majority of dynamics problems. An alternative method uses the static approach to calculate results
one vector at a time.
The mode acceleration method, also detailed in this chapter, is useful in some cases where accurate stress
results are required for only a few peak time steps or frequencies. However, this option may be more
expensive if many solution vectors exist.
In addition to the standard displacement, stress, and force outputs, dynamics engineers frequently use
other types of output that combine the effects of random or spectral inputs. shock and response spectral
analysis provides a quick calculation of the probable peak values in a modal solution for base
accelerations. Random analysis is a more formal method of determining power spectral densities, RMS
averages, and other data for a structure in random load environment.
Main Index
CHAPTER 12 485
Modal Augmentation Methods
ua = a (12-1)
Free Set
G0
uf = ua (12-2)
I
SPC Forces
Q s = K sf u f – P s (12-3)
Independents
uf
un = (12-4)
0
Grid Set
Gm
ug = un (12-5)
I
Main Index
486 MD Nastran Dynamic Analysis User’s Guide
Data Recovery for Modal Methods
In the actual calculations, each vector at a particular frequency or time step becomes a column of the
solution matrix u g . Note that for all modal methods, no provisions are made for thermal loads or
enforced displacements (which are only calculated in statics analysis.)
The element stresses and forces are calculated directly from the displacement vectors, one element and
one vector at a time. Note that if a large number of solutions exist (for instance 10000 time steps), the
size of the matrices defined above may be quite large and the calculation speed will be very slow.
u gm = g (12-6)
SPC Forces
Q sm = K sf f (12-7)
Stresses
Sm = (12-8)
Forces
Fm = f (12-9)
Note that the number of rows of these matrices will correspond only to the requested output sets (DISP
=, STRESS =, etc.). The number of columns is equal to the number of modes (which is usually smaller
than the number of time steps of frequencies). For the final output printing or xy-plotting, the equations
for calculating the transient solution outputs are as follows:
Displacements
u go t = u gm t (12-10)
Forces of SPC
Q s t = Q sm t (12-11)
Similar equations are used for SPC forces, stresses, and element forces, as well as for frequency response,
by simply combining the modal responses with the modal factors as indicated above.
Main Index
CHAPTER 12 487
Modal Augmentation Methods
– 2 M + jB + K u = P (12-12)
where j is the imaginary unit value.
We may obtain an exact solution if we use all of the modes in the system; namely, if N is the order of the
problem, then:
N
u = i i (12-13)
i = 1
Note that this set of modes also includes modes with infinite frequencies. These are present with singular
mass matrices. The modal identities are
i2 M i = K i (12-14)
Main Index
488 MD Nastran Dynamic Analysis User’s Guide
Data Recovery for Modal Methods
g
B i = ------ K i (12-15)
3
N
jg
--------
j B u = 3
- K i i (12-17)
i = 1
and
N
Ku = K i i (12-18)
i = 1
If all modes were included, we know that the results would be identical to a direct solution. For the high
frequency modes ( i m ), it is apparent that their contribution will be much smaller for mass and
damping effects than from their effect on the static solution of Eq. (12-18). Therefore, if we truncate the
modes such that the primary error will be due to the forces generated by the missing modes in Eq. (12-18).
In the frequency domain the approximations to Eq. (12-16) through Eq. (12-18) become
M
–2 M u – 2 M i i (12-19)
i = 1
M
j B u j B i i (12-20)
i = 1
M
Ku = K i i + K u (12-21)
i = 1
Substituting Eqs. (12-19) and (12-20) into Eq. (12-15), we obtain a corrected solution:
Main Index
CHAPTER 12 489
Modal Augmentation Methods
where:
R = M M
j i i – 2 i i
i =1 i =1
An alternate method would be to solve the static part of Eq. (12-15) using Eq. (12-21). Since the residual
vector is orthogonal to the retained modes, then:
T K u = 0 (12-23)
We can assume that u contains static response, if K is nonsingular, plus modal response in the
form: (12-24)
u = – K – 1 P + x (12-25)
–1
x i = – m i i2 i T P (12-26)
and, then from Eqs. (12-25) and (12-26), we obtain a decoupled residual solution
u = Z P (12-27)
where:
Z = .
–1 T
K – m i i2
.
The matrix, Z , is known as the residual flexibility matrix and has also been used for modal synthesis
modeling. Here it could be used as a data recovery step.
In the MD Nastran design, Eq. (12-22) has been chosen over Eq. (12-27) for the calculation of the
improved solution of large problems because the Z matrix could be very dense and the singular free
body case was easier to process.
Note that in the actual mode acceleration process other nonstructural effects, such as direct input
matrices or transfer functions, are not included in the matrix, K , but are treated as terms on the
right-hand side and added to P . For free bodies, the right hand loads are converted to equilibrium
loads to permit a decomposition of the singular stiffness matrix, identically to the inertia relief solution.
In fact, the entire process may be viewed as if all the dynamic modal solutions were converted into
equivalent static loads, and linear static solutions were generated using the symmetric structural stiffness
matrix.
Main Index
490 MD Nastran Dynamic Analysis User’s Guide
Data Recovery for Modal Methods
uo = Go a ua + o – Go a a q (12-28)
where:
Main Index
CHAPTER 12 491
Modal Augmentation Methods
The time step, DT, and time duration, DT N (where N is the number of time increments), must take
into account the loading, the base structure, and the frequency range of the spectra generation. The time
step must take into account the frequency content of the applied excitation, the frequencies of the base
structure, and the highest frequency for which spectra are to be generated. There must be enough time
steps per cycle of response for both the base structure and the highest frequency oscillator in order to
accurately predict the peak response; 5 to 10 steps per cycle represents a typical value. In addition, the
time duration of the loading, the frequencies of the base structure, and the lowest oscillator frequency
must be considered when defining the time duration. There must be a long enough time duration of
response both for the base structure and the lowest frequency oscillator in order to accurately predict the
peak response. For short duration loadings, the peak response often occurs well after the the load has
peaked.
Initial conditions (specified via the TIC Bulk Data entry) are not used in response spectrum generation.
Initial conditions are used in the calculation of the transient response of the base structure, but the
calculation of the peak oscillator responses (i.e., the response spectrum calculation) ignores any initial
conditions.
The XYPLOT and XYPUNCH commands are included in the OUTPUT(XYPLOT) section. Further
details about the OUTPUT(XYPLOT) section are described in Results Processing (Ch. 15).
Main Index
492 MD Nastran Dynamic Analysis User’s Guide
Dynamic Response Predictions
Relative and absolute spectra are denoted by IP and RM, respectively, in the parentheses of the curve
request.
There are two FREQi entries: one to specify oscillator frequencies (i.e., frequencies for which spectra
will be computed) and the other to specify oscillator damping. (Note that damping for the base structure
is specified in another manner, such as with the TABDMP1 entry used for modal transient response
analysis.)
Main Index
CHAPTER 12 493
Modal Augmentation Methods
Modal Direct
Small Model X
Large Model X
Few Excitation Frequencies X
Many Excitation Frequencies X
High Frequency Excitation X
Nonmodal Damping X
Higher Accuracy X
In general, larger models may be solved more efficiently in modal frequency response because the
numerical solution is a solution of a smaller system of uncoupled equations. The modal method is
particularly advantageous if the natural frequencies and mode shapes were computed during a previous
stage of the analysis. In that case, you simply perform a restart (see Restarts In Dynamic Analysis, 377).
Using the modal approach to solve the uncoupled equations is very efficient, even for very large numbers
of excitation frequencies. On the other hand, the major portion of the effort in a modal frequency
response analysis is the calculation of the modes. For large systems with a large number of modes, this
operation can be as costly as a direct solution. This result is especially true for high-frequency excitation.
To capture high frequency response in a modal solution, less accurate, high-frequency modes must be
computed. For small models with a few excitation frequencies, the direct method may be the most
efficient because it solves the equations without first computing the modes. The direct method is more
accurate than the modal method because the direct method is not concerned with mode truncation.
Table 12-3 provides an overview of which method to use. Many additional factors may be involved in
the choice of a method, such as contractual obligations or local standards of practice.
Main Index
494 MD Nastran Dynamic Analysis User’s Guide
Residual Vector
Residual Vector
Main Index
CHAPTER 12 495
Modal Augmentation Methods
Case Control command, RESVEC, allows the user to specify the loads used for residual vector
calculation. The format for the RESVEC is as follows:
APPLOD ADJLOD DYNRSP
RESVEC INRLOD RVDOF DAMPLOD
=
NOINRL NOAPPL NOADJLOD NORVDO NODAMP NODYNRSP
SYSTEM/NOSYSTEM
COMPONENT/NOCOMPONENT
BOTH or YES
NO
Examples:
RESVEC=SYSTEM
RESVEC(NOINRL)=COMPONENT
RESVEC=NO
Describer Meaning
INRLOD/ Controls calculation of residual vectors based on inertia relief (Default =
NOINRL INRLOD).
APPLOD/ Controls calculation of residual vectors based on applied loads (Default =
NOAPPL APPLOD).
ADJLOD/ NOADJLOD Controls calculation of residual vectors based on adjoint load vectors
(SOL 200 only; Default = ADJLOD).
RVDOF/ Controls calculation of residual vectors based on RVDOFi entries (Default
NORVDOF = RVDOF).
DAMPLOD/ Controls calculation of residual vectors based on viscous damping (Default
NODAMP = DAMPLOD).
DYNRSP/ Controls whether the residual vectors will be allowed to respond
NODYNRSP dynamically in the modal transient or frequency response solution. See
Remark 5. (Default = DYNRSP).
SYSTEM/ Controls calculation of residual vectors for system (a-set) modes. For
NOSYSTEM NOSYSTEM, describers inside the parentheses are ignored. See Remark
2. for default.
COMPONENT/ Controls calculation of residual vectors for component (superelement or o-
NOCOMPONENT set) modes. For NOCOMPONENT, describers inside the parentheses are
ignored. See Remark 2. for default.
Main Index
496 MD Nastran Dynamic Analysis User’s Guide
Residual Vector
Describer Meaning
BOTH or YES Requests calculation of residual vectors for both system modes and
component modes. See Remark 2. for default.
NO Turns off calculation of residual vectors for both system and component
modes, and describers inside the parentheses are ignored. See Remark 2.
for default.
The inertial forces due to the normal mode shapes are removed from the loads used for residual vector
calculation using the following:
–1
P = P – MM P
The modified loads are applied to the structure to determine the deformations that serve as the base
vectors for residual vector calculation.
K Vl o a d = P
Additionally, for transient response, initial conditions are also added to the set of base vectors. Before
the initial condition vectors are added, any mode shape content is removed.
–1
V init ial = V in i ti al – M M V ini tial
The initial condition vectors are appended to the load vectors and the complete set of base vectors are
made linearly independent. This process removes vectors that are linear combinations of other vectors
in the base vector set. The remaining base vectors are then made orthogonal with respect to the mass and
stiffness matrices. The resulting orthogonal vectors are the residual vectors.
The residual vectors are appended to the modes and the reduced mass and stiffness matrices are modified
to include the residual vector additions.
= modes residual
T
M = M
T
K = K
Residual vectors which produce ‘zero’ mass terms in the reduced mass matrix are treated as ‘massless’
modes. The diagonal mass is set to 1.0E-36 and the diagonal stiffness is set to 1.0. The augmented set of
modes/residual vectors is used to reduce the damping, applied loads, and any external matrix input.
Residual vector calculation is available for all modal solutions in MD Nastran. They can be calculated
for the residual structure and superelements to improve the analysis results.
Main Index
CHAPTER 12 497
Modal Augmentation Methods
Main Index
498 MD Nastran Dynamic Analysis User’s Guide
Residual Vector
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 13: Complex Eigenvalue Analysis
Overview
Complex Eigensolutions
Main Index
500 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
Complex eigenvalue analysis is used to compute the damped modes of structures and assess the stability
of systems modeled with transfer functions (including servomechanisms and rotating systems).
Complex eigenvalue analysis solves for the eigenvalues and mode shapes based on the following
equation in operator notation:
2
p M + pB + K u = 0
where p = + i
This equation is similar to that for normal modes analysis (see Real Eigenvalue Analysis (Ch. 3)) except
that damping is added and the eigenvalue is now complex. In addition, the mass, damping, and stiffness
matrices may be unsymmetric, and they may contain complex coefficients.
Complex eigenvalue analysis is controlled with the EIGC Bulk Data entry (similar to the EIGRL or EIGR
Bulk Data entries for normal modes analysis). There are three methods of solution: upper Hessenberg,
complex Lanczos, and inverse power. Complex eigenvalue analysis is available as a direct method
(SOL 107), in which the equations are of the same size as the number of physical variables. Complex
eigenvalue analysis is also available as a modal method (SOL 110), in which undamped modes are first
computed and then are used to transform the matrices from physical to modal variables.
Main Index
CHAPTER 13 501
Complex Eigenvalue Analysis
Complex Eigensolutions
Complex eigenvalue analysis is necessary when the matrices contain unsymmetric terms, damping
effects, or complex numbers where real modes analysis cannot be used. It is used for the analysis of
aeroelastic flutter, acoustics, rotating bodies, and many other physical effects.
The unforced motion of a system of equations can be expressed as the sum of the motion of its
eigenvectors, oscillating and decaying or expanding with terms of the form
+ i i t
u t = Real i e i (13-1)
The value of i gives a measure of the rate of decay or divergence of the i-th natural dynamic mode. If
the value of i is nonzero, it gives a measure of the rate of oscillation of the solution. For the most
general case, i , which represents the shape of the mode, contains complex numbers. The relative size
of these numbers indicates which parts of the structure are most active in this mode of motion. The
imaginary parts of u signify phase differences or lag times between the degrees-of-freedom.
The results of complex eigenvalue analysis are used for tasks such as measuring the effect of damping
materials on system performance and determining the stability of a system when it contains sources of
energy such as rotating components. The complex eigensolution is an end product in MD Nastran.
There are no provisions for using these shapes for modal solutions of frequency response analysis or
transient response analysis, as is done for the real eigensolution capability.
K – i M i = 0 (13-2)
where K and M are real and symmetric, and M is positive semi-definite. This type of eigenvalue
problem is discussed in the Real Eigenvalue Analysis (Ch. 3). All damping effects are ignored. It can
be shown that the eigenvalues i and eigenvectors i of this problem must be real, not complex, and that
the eigenvectors can be made orthonormal with respect to the mass matrix and orthogonal to the stiffness
matrix; that is,
t M = I (13-3)
t K = (13-4)
where has the eigenvalues stored along its diagonal. This property is used to separate eigenvectors
for repeated eigenvalues into linearly independent sets.
Main Index
502 MD Nastran Dynamic Analysis User’s Guide
Complex Eigensolutions
Any problem whose input matrices do not meet these restrictions is called a complex eigenvalue problem
in MD Nastran. In general, the solution may include real or complex numbers, and the orthogonality
properties of this more general case are less powerful.
The matrix equation used for the complex eigenvalue problem is
K + pB + p 2 M = 0 (13-5)
and
p = + i
where any of the matrices may contain real or complex numbers, and may also be symmetric or
unsymmetric. If all matrices are symmetric, there is a transformation similar to Eqs. (13-3) and (13-4)
that can transform all three matrices to a diagonal form. If any of the matrices is complex or
unsymmetric, only a weaker orthogonality condition applies. If all of the eigenvalues are distinct (that
is, there are no repeated roots), each matrix can be reduced to diagonal form by also computing the left
hand eigenvectors L , that is, the eigenvectors of Eq. (13-5) when all of its matrices are transposed. One
orthogonality condition for any two roots of the system happens to be
L i T p i + p j M + B R j = 0 if p i p j (13-6)
where R are the right-hand eigenvectors of Eq. (13-5). Some of the eigensolution methods use these
orthogonality conditions to improve the linear independence of the eigenvectors, while others do not.
None of the MD Nastran complex eigensolution methods has a special provision for repeated roots. They
may return the same eigenvector for each of the repeated roots.
A special form of Eq. (13-5), called the shifted form, is used by some methods to provide faster
convergence in the iterations. If we introduce a shift point, p 0 , the relative distance of the root is
= p – p0 (13-7)
K + p 0 B + p 02 M = – p 0 B + 2p 0 M + M (13-8)
One may iterate Eq. (13-8) by solving for trial vectors on the left side using previous estimates on the
right side to produce vectors corresponding to the smallest values of . This procedure is used in the
Lanczos method to obtain sets of approximate vectors and in Inverse Power to obtain actual vectors.
Main Index
CHAPTER 13 503
Complex Eigenvalue Analysis
A – i I ui = 0 (13-9)
A is transformed to an upper Hessenberg form. In this form all terms below the subdiagonal are null.
All of the roots of this transformed problem are computed by the double QR algorithm of Francis. The
eigenvectors are then found by inverse iteration. Convergence is usually within one or two iterations
because the roots are known at this point. There are similarities in this solution process with the
tridiagonal methods of real eigenvalue analysis.
The three-matrix form is transformed to the single matrix form by premultiplying by the inverse of the
mass matrix,
M – 1 K + pM – 1 B + p 2 I u = 0 (13-10)
If B is null then A = M –1 K and i = – p 2 . If B is not null the velocity vector v is defined from u ,
v = pu (13-11)
Equations (13-10) and (13-11) are combined to form
M –1 B M –1 K + p I 0 v
= 0 (13-12)
–I 0 0 I u
If B is null Eq. (13-10) is used to produce n eigenvalues (only roots with 0 are calculated), while
if B exists Eq. (13-12) will produce 2n eigenvalues. The user can control the number of eigenvectors
computed in the Inverse and Lanczos methods.
The fact that M must have an inverse means that M may not contain null rows or columns, or other
sources of singularity. The rows and columns for these singular DOFs may be removed by omitting them
with OMITi entries, or, what is often less labor for large models, by retaining important DOFs with
independent mass with ASETi entries. All other unconstrained DOFs are then placed in the omitted set.
In general, translation DOFs with attached masses are safe candidates for the a-set. Care should be taken
with rotational DOFs. An example of a DOF with nonnull mass but which is not independent is a rotation
DOF to which a CONM2 element is attached, and the CONM2 element has a point mass and an offset,
but zero moments of inertia. The rotation DOF will be singular when the matrix is inverted.
No special logic is used for repeated roots, which return the same eigenvector. There is a system cell
which requests that a newer method with spill capability be used. The default value for this system cell
uses an algorithm with no spill capability. System cells can be set on the NASTRAN command. Refer
to the EIGC (p. 1770) in the MD Nastran Quick Reference Guide for details.
This method does not take advantage of matrix sparsity or banding. Solution costs are proportional to n 4 .
Main Index
504 MD Nastran Dynamic Analysis User’s Guide
Complex Eigensolutions
Solution costs are proportional to r 10 nb 2 where b is the semi-bandwidth and r the number of
eigenvectors computed. Many roots can be computed from each shift.
Main Index
CHAPTER 13 505
Complex Eigenvalue Analysis
4. The h-set mass matrix is guaranteed to be well conditioned for the inverse operation unless very
strange effects are input with the M2PP matrix or the counterpoint terms of transfer functions, so
there is no need to omit massless physical DOF, a task which can be laborious on a complicated
model. (It is necessary, however, to ensure that any extra points that are used have nonnull mass
terms.)
5. All eigensolutions are then found in this reduced basis. The user is assured that there are no
solutions which have been skipped over.
6. The economic trends are also favorable because the real eigensolutions do take advantage of
matrix sparsity and banding (the Real Lanczos Method is the best in this respect today), and they
reduce the problem to a much smaller size before the Hessenberg solution is attempted.
However, if strong nonstructural effects are present, such as those caused by direct input matrices or
transfer functions, a large number of real modes may be necessary to account for these forces. An
example would be a servo control that acts as a large mass on the structure. This mass could reduce the
frequency of some local high frequency modes and move them into the important range.
The Complex Lanczos and Inverse Power methods may also be used in the modal method, but the
economic rewards are less, and they may also miss roots. They should be used as backup methods in
this context, used to check the solution produced by the Hessenberg method when there is some question
about its correctness.
User Interface
Various form of the eigenvalues are used as inputs and are provided as outputs. The real eigensolution
outputs the eigenvalues in three forms. The complex eigenvalue, p , is formed from the real quantities
and where p = + i . If = 0. , the radian frequency, of complex eigenvalue analysis is
the same as that of real eigensolutions. The real part, , is a measure of the decay rate of a damped
structure, or if negative, the rate of divergence of an unstable system. The imaginary part, , is the
modified frequency in radians/unit time. However, roots with negative values of should be treated as
special terms. The output labeled f is the circular frequency in cycles per unit time. It is equal to
2 .
The CMETHOD = [SID] Case Control command selects the EIGC Bulk Data entry, which has the form
1 2 3 4 5 6 7 8 9 10
EIGC SID METHOD NORM G C E ND0
Main Index
506 MD Nastran Dynamic Analysis User’s Guide
Complex Eigensolutions
The shift points for the Lanczos method are defined on the continuation entries. The recommended
practice is to specify one point 1 = 0.0 and 1 at the lower bound of the expected range of eigenvalues,
but not at 0.0 A second shift may be input at 2 = 0.0 and 2 at the upper bound of the expected range.
All ALPHABj and OMEGABj must be blank. The number of Ritz vectors computed for the J-th shift is
three times the number entered in NDJ. While each Ritz vector may be capable of defining an
eigenvector, it is more likely that some of the Ritz vectors will be discarded because they are not
orthogonal to lower-numbered Ritz vectors during the Lanczos reduction process. After the poor vectors
are discarded, the remaining vectors are used to compute the eigensolutions.
The number of modes computed from all shifts up to the j-th shift may be more or less than NDJ. If there
are less, the processing continues with the next shift; otherwise, the process ends and all eigensolutions
are output. The ratio of discarded vectors to retained vectors is problem dependent. However, the
recommended practice is to ask for all of the desired eigenvectors for ND1 on the first shift, but put in
additional shifts. If it is possible to compute all solutions with one shift, this is the most economical
option. If more shifts are needed, they will be used.
The subregions for searching for roots with the Inverse Power method are also specified on the
continuation entries. (ALPHAAJ, OMEGAAJ) define one point in the complex plane, and (ALPHABJ,
OMEGABJ) a second point. A line is drawn between these points, and a box of width LJ is placed around
this line. NEJ is larger than the number of roots expected in the subregion, and NDJ the number desired.
This defines the first subregion for searching for eigenvalues. More continuation entries may be used to
define more subregions. Again, if more eigenvalues are computed than are requested, all are output.
The eigenvalue output for a sample problem is shown below:
C O M P L E X E I G E N V A L U E S U M M A R Y
ROOT EXTRACTION EIGENVALUE FREQUENCY DAMPING
NO. ORDER (REAL) (IMAG) (CYCLES) COEFFICIENT
1 9 -5.806441E+01 5.750383E+03 9.152019E+02 2.019497E-02
2 8 -6.294888E+01 6.293917E+03 1.001708E+03 2.000309E-02
3 7 -6.910709E+01 6.844852E+03 1.089392E+03 2.019243E-02
4 1 -3.300980E+02 1.667092E+04 2.653260E+03 3.960164E-02
5 5 -3.565692E+02 1.823559E+04 2.902285E+03 3.910694E-02
. . .
The column labeled (REAL) contains 1 , and the column labeled (IMAG) contains 1 . The column
labeled (FREQUENCY) contains the circular frequency. The last column is the damping coefficient
computed from the equation
g = – 2
which is approximately twice the value of the conventional modal damping ratio. This form was more
popular with the aeroelastic flutter specialists who were the primary users of this capability.
Note that if the magnitude of this term is computed to be less than 5.0 10 –4 , it is reset to zero.
For small values, the damping coefficient is twice the fraction of critical damping for the mode. The
eigenvalues are sorted on , with the negative values sorted first (there are none in this example), sorted
on increasing magnitude, followed by the eigenvalues with positive , again sorted on magnitude.
Roots with equal values are sorted next on .
Main Index
CHAPTER 13 507
Complex Eigenvalue Analysis
Modeling Techniques
All Methods
Repeated roots are usually due to planes of symmetry of a structure. For example, a pencil has a pair of
repeated roots for each bending mode, with their eigenvectors being at 90 degrees to each other. In
physical structures there is never absolute symmetry because of the effects of manufacturing tolerances,
material inconsistencies, etc. The user may adopt this point of view when repeated roots cause problems
by making small perturbations to the structure to destroy symmetry. For example, when doing an
analysis for the damped modes of a pencil, a small mass term can be added in one direction. A small
perturbation is usually adequate to separate the modes enough to eliminate numerical problems.
Hessenberg Method
1. If extra points or other DOFs are connected by only transfer functions or x2PP matrix terms, they
may need to be differentiated to provide a mass term on these DOFs. For example, if the
following equations are written for extra points e 1 and e 2 , which are coupled to the structural
DOFs u 11 and u 12 ,
c 1 e 1 = u 11
c 2 e· 1 = u 12
In these forms, c 1 would appear in the stiffness term of the transfer function, and c 2 in the
damping term.
The first equation can be differentiated twice, and the second once, to provide
c e·· = u··
1 1 11
c 2 e··1 = u· 12
The c i terms are placed in mass terms slots in the transfer function, or in the M2PP matrix.
2. Avoid coupled mass. One apparent choice for obtaining mass terms on rotation DOFs is to
request coupled mass by use of the PARAM entry that requests it (but this is usually a poor
choice). The theories for coupled mass for some elements in MD Nastran do not provide
rotational mass while other elements have theories that produce rotational mass in a manner that
does not provide an independent mass term, leading to either failure in the inversion of M , or
poor numerical conditioning. Null masses on rotational DOFs cause no problems in modal
analysis.
In direct analysis, if all unconstrained rotations are omitted, this potential singularity problem is
avoided, with little or no loss in accuracy, and the cost of static reduction of the system is less
than the savings of solving a smaller sized eigenproblem.
Main Index
508 MD Nastran Dynamic Analysis User’s Guide
Complex Eigensolutions
k Housing
Pad
Y
V Moving Surface
The objective of the analysis is to determine whether the friction effects can cause the onset of a divergent
squeal mode. The following assumptions are made in the modeling:
1. The speed of the sliding surface is assumed to be much less than the speed of the traveling
vibrational waves. Therefore, the elements representing the surface may be limited to small
motions and the traveling wave effects are ignored.
2. Pure sliding friction is assumed. The magnitude of the pad vibration may be very small for the
onset of the unstable mode. The analysis will be invalid when the vibrational velocities exceed
the surface velocity.
3. A static preload is assumed to be large enough to maintain full contact on the pad surface. The
frictional coefficient is assumed to be constant. (However, it could be varied over the contact
region.)
Frictional forces on the pad and sliding surface are proportional to the normal contact forces, which in
turn may vary with dynamic response. For small dynamic motions the relationship is
F p y = – F s y = N (13-13)
where:
Main Index
CHAPTER 13 509
Complex Eigenvalue Analysis
The value of N is only available indirectly from the element forces or by the use of Lagrange multipliers.
The solution variables in a conventional model consist only of displacements, velocities, and
accelerations.
An approximation to N may be generated from a penalty function method using large springs to
represent the contact constraint. If a large spring, K (CELAS2 data), is connected to the two normal
displacements at a point, the normal force is a function of the displacements
N = –K up z – us z (13-14)
Combining Eqs. (13-13) and (13-14), we obtain the following matrix relationship:
Fp y u
–1 1 p z
= K etc. (13-15)
Fs y 1 – 1 u sz
Special Input
The matrix terms in Eq. (13-15) may be input to the problem for each contact point directly as DMIG
data. The resulting matrix will be unsymmetric. The Case Control command required for DMIG input is
K2PP = FRIC * FRICTN
where FRIC will be replaced with user selected value to multilply the matrix by and
where FRICTN is the name of a DMIG Bulk Data entry, which starts with the header entry
DMIG, FRICTN, 0, 1, 1, 0 . . .
where the inputs and their functions are as follows:
Listing 13-1 shows the resulting MD Nastran input file. RBEi and CONM2 elements are used to simplify
the model and eliminate spurious roots. The unique inputs are the DMIG data and the associated
CELAS2 springs, (Ids = 311-314).
Main Index
510 MD Nastran Dynamic Analysis User’s Guide
Complex Eigensolutions
The Case Control for a complex LANCZOS method is shown here, with the provisions for the other
methods preceded by a $ symbol. (Lines starting with $ are treated as comments, but are otherwise
ignored.)
CMETHOD = 10 $CLAN
$CMETHOD = 11 $INVP
$CMETHOD = 12 $HESS
A METHOD command is required when using the modal method of SOL 110. This input is for the direct
method of SOL 107. The CMETHOD command is required for complex eigensolution method selection.
Eigenvector output is requested by the DISP command.
The Bulk Data entries selected by these commands are as follows:
EIGRL,1, , 4000.0, 12 $ real lanczos
EIGC, 10, CLAN, MAX, , , , , ,+ELAN1
+ELAN1, 0.0, 6000.0, , , , , 13,, +ELAN2
EIGC, 11, INV , MAX, , , , , ,+EINV
+EINV, 0.0, 10000.0, 0.0, 50000., 20000., 20, 12
EIGC, 12, HESS , MAX, , , , 13
Main Index
CHAPTER 13 511
Complex Eigenvalue Analysis
GRID,103,,50.0,30.0,7.5,,1456
RBE2,101,100,23,101,102,103
$ LUMPED MASS AT CG
CONM2,100,100,,2.0 , , , , ,+CM1
+CM1,7200.0,,7200.0 , , ,7200.0
$ SPRINGS TO GROUND
CELAS2,102, 78.96+6, 100, 2
CELAS2,103, 78.96+6, 100, 3
CELAS2,104, 2.846+11,100,4
$
$ PAD IS A SINGLE HEXA WITH A POINT FOR THE HOUSING
CHEXA,200,200, 211, 212, 214, 213, 221, 222,+CHX1
+CHX1, 224, 223
RBE3,201,,201,123456,1.0,123,221,222,+RBE3
+RBE3,223,224
CONM2,202,201,,0.2
GRID, 201, ,0.0,0.0,15.0
GRID, 211, , -50., -30., 0.0,,1456
GRID, 212, , 50., -30., 0.0,,1456
GRID, 213, , -50., 30., 0.0,,1456
GRID, 214, , 50., 30., 0.0,,1456
GRID, 221, , -50., -30., 15.0,,1456
GRID, 222, , 50., -30., 15.0,,1456
GRID, 223, , -50., 30., 15.0,,1456
GRID, 224, , 50., 30., 15.0,,1456
PSOLID, 200, 200, , THREE
MAT1,200,1.0+6,,0.3,1.111-6,,,0.02
$
$ RIGID LINKS FROM PAD FOR TANGENTIAL STOPS
MPC, 200, 213,2, 1.0, 223,2, 1.0, , +MPC21
+MPC21,, 102,2, -2.0
$MPC, 200, 213,3, 1.0, 223,3, 1.0, , +MPC22
$+MPC22,, 102,3, -2.0
MPC, 200, 214,2, 1.0, 224,2, 1.0, , +MPC31
+MPC31,, 103,2, -2.0
$MPC, 200, 214,3, 1.0, 224,3, 1.0, , +MPC32
$+MPC32,, 103,3, -2.0
$
$RBE3,202,,102, 234, 1.0, 123, 213 ,223,+RB31
$+RB31,UM ,213,23, 223,2
$RBE3,203,,103, 234, 1.0, 23 ,214, 224,+RB32
$+RB32,UM ,214,23, 224,2
$ SPRINGS CONNECTING HOUSING TO CENTER OF PAD
CELAS2,201, 15.791+6, 201,3 ,101,3
CELAS2,202, 15.791+6, 201,2 ,101,2
$
$ MOVING SURFACE SIMULATED WITH PLATES
$
CQUAD4,301,300,301,302,304,303
CQUAD4,302,300,303,304,312,311
CQUAD4,303,300,311,312,314,313
CQUAD4,304,300,313,314,322,321
CQUAD4,305,300,321,322,332,331
$
GRID, 301,,-50., -150.0, 0.0,, 12356
GRID, 302,, 50., -150.0, 0.0,, 12356
GRID, 303,,-50., -90.0, 0.0,, 1256
GRID, 304,, 50., -90.0, 0.0,, 1256
GRID, 311,,-50., -30.0, 0.0,, 1256
GRID, 312,, 50., -30.0, 0.0,, 1256
Main Index
512 MD Nastran Dynamic Analysis User’s Guide
Complex Eigensolutions
Results
This model was run with three different types of analysis. For the modal complex method, 20 real modes
were requested to produce the reduced problem for complex eigensolution analysis. It was found that the
eleventh mode was quite divergent, while all of the other modes were reasonably stable. The eigenvalues
of this mode and the preceding and following modes by this method are given:
Main Index
CHAPTER 13 513
Complex Eigenvalue Analysis
C O M P L E X E I G E N V A L U E S U M M A R Y
ROOT EXTRACTION EIGENVALUE FREQUENCY DAMPING
NO. ORDER (REAL) (IMAG) (CYCLES) COEFFICIENT
.....
10 10 -9.998501E+02 5.001250E+04 7.959736E+03 3.998401E-02
11 11 3.294983E+03 5.022236E+04 7.993137E+03 -1.312158E-01
12 12 -5.339673E+03 5.010448E+04 7.974375E+03 2.131415E-01
13 13 -1.200202E+03 5.960435E+04 9.486327E+03 4.027230E-02
.....
When the same model is analyzed by the Lanczos method, without modal reduction, the roots in the same
region agreed. This agreement occurs because almost all of the real modes were found, including those
at much higher frequencies than that of the eleventh complex mode. When the friction coefficient
parameter was reduced from a value of 0.5 to 0.05, all roots became stable. The roots in the region of
the eleventh mode are as follows:
It would appear that there is a stability threshold somewhere between the friction coefficient values of
0.05 and 0.5. The precise threshold can be determined by resetting the value of this parameter and
rerunning the model until the real value of this mode approaches zero. Hardware elements that affect
this unstable mode can be identified by inspecting the eigenvector. The affect of hardware changes on
the stability of this mode can be evaluated by changing the most active elements and rerunning the
analysis.
This study shows that the prudent method for checking this answer is the Lanczos method, but the less
expensive modal Hessenberg method can be used for the parameter variation studies to find the stability
boundary, with a final check made by the Lanczos method.
The eigenvector computed for this mode from the Direct Lanczos results is shown below. This format
contains two lines per grid point. The first line is the real value of displacement. The second line
indicates the imaginary part. Magnitude and phase angles are an alternative choice.
Main Index
514 MD Nastran Dynamic Analysis User’s Guide
Complex Eigensolutions
Note that two points have a value of 1.0, except for a very small imaginary part. These are the analysis
set points with the largest value. (See points 223, 224.) Note that one point has a value which exceeds
unity. (See point 214.) This point is the dependent point on a selected MPC entry. Only analysis set
points are considered when performing the normalization. Dependent points for MPC equations and
omitted points may be smaller or larger than unity.
The meaning of the imaginary parts of the displacement at the other points is a measure of the phase angle
between motion at the normalizing point and the other points. A more meaningful result for the
imaginary part can be obtained from the modal analysis by requesting SDISP output. This prints the
modal contributions of the real modes which are used to synthesize the complex modes. The output for
the eleventh mode, from a Modal Hessenberg analysis, is as follows:
Although the columns are labeled T1, T2, etc., they really represent the scalar modal variables printed
six per row. For example, the largest value occurs for the eleventh real mode, printed in the row labeled
“7 M” under R2. (M symbolizes modal variable in this context.) If the objective of the analysis is to find
Main Index
CHAPTER 13 515
Complex Eigenvalue Analysis
the parts of the structure that have the largest effect on stability, when changed, it is often easier to
understand the physics from the real mode shapes rather than the displacement in a complex mode shape
of physical variables.
The a-set values of the real mode shape can be printed by the SVECT Case Control command. The
values of nonanalysis set DOFs, such as m-set or omitted points, appear as blank spaces. If the complete
real mode shapes are needed for plotting, or if element strain energy is desired for the real modes, this
input stream can be run in SOL 103, the real modes analysis, by changing only the SOL entry. Inputs
not related to real eigenvalue analysis are ignored.
Main Index
516 MD Nastran Dynamic Analysis User’s Guide
Complex Eigensolutions
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 14: Guidelines and Tools for Effective Dynamic Analysis
Main Index
288 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
Dynamic analysis is more complicated than static analysis because of more input (mass, damping, and
time- and frequency-varying loads) and more output (time- and frequency-varying results). Results from
static analysis are usually easier to interpret, and there are numerous textbook solutions for static analysis
that make it relatively easy to verify certain static analyses. Nevertheless, the guidelines in this chapter
help you to perform dynamic analysis in a manner that will give you the same level of confidence in the
dynamic results that you would have with static results.
Main Index
CHAPTER 14 289
Guidelines and Tools for Effective Dynamic Analysis
Once you have confidence in a small model and are ready to analyze your actual model, again do the
analysis in steps. The following is a suggested order for performing dynamic analysis on any structure:
1. Create the initial model only; do not apply any loads. Verify the model's connectivity, element
and material properties, and boundary conditions. Use a graphical tool such as Patran to assist in
this. Make sure that mass is specified for this model.
2. Perform a static analysis (SOL 101) first in order to verify proper load paths and overall model
integrity. (Note that you have to constrain the structure for static analysis even if you were not
planning to do so for dynamic analysis.) For a three-dimensional model, you should run three
load cases, each with a 1g gravity load applied in a different direction. Compute displacements
and SPC forces, and verify the results. Check for unusually large grid point displacements and
unreasonable SPC forces. The use of a graphical postprocessor can aid you at this step.
Next, apply static loads that have the same spatial distribution that your subsequent dynamic
loads have. Verify the results for reasonableness. Do not go to dynamic analysis until you are
satisfied with the results from your static analysis.
It is recommended at this stage that the model contain PARAM,GRDPNT,n (where n is a
reference grid point or 0, the origin of the basic coordinate system). Verify the results from the
grid point weight generator in order to ensure that the model's rigid-body mass and inertia look
reasonable. This step, in conjunction with the static analysis results, helps to ensure that the
proper mass units are specified.
3. Perform an eigenvalue analysis (SOL 103) next. Compute only a few modes first, verify their
frequencies, and view their mode shapes for reasonableness. If your graphical postprocessor can
animate the mode shapes, do so because that helps you to visualize them. Things to check at this
step are local mode shapes, in which one or a few grid points are moving a very large amount
relative to the rest of the model (this can indicate poor stiffness modeling in that region), and
unwanted rigid-body modes (which can arise due to improper specification of the boundary
conditions or a mechanism).
Once you are satisfied with these results, perform the full eigenvalue analysis (for as many modes
as you need).
4. If you have frequency-dependent loads, perform frequency response analysis (SOL 108 or
SOL 111) using the dynamic load spatial distribution. If your structure is constrained, then apply
the dynamic load at only one frequency, which should be 0.0 Hz. Compare the 0.0 Hz
displacement results to the static analysis displacement results. The results should be the same if
Main Index
290 MD Nastran Dynamic Analysis User’s Guide
Overall Analysis Strategy
direct frequency response (without structural damping) is used. If the results are not equal, then
there is probably an error in the specification of the dynamic load, and you should check the LSEQ
and DAREA entries. If modal frequency response (without structural damping) is used, then the
0.0 Hz results should be close to the static results; the difference is due to modal truncation.
Next apply the load across the entire frequency range of interest. If you are running modal
frequency response, then make sure that you have enough modes to ensure accurate results for
even the highest forcing frequency. Also be sure to have a small enough f in order to accurately
capture the peak response. Verify these results for reasonableness (it may be easier to look at
magnitude and phase results instead of real and imaginary results, which are the default values).
If your ultimate goal is a transient response analysis for which damping is to be neglected, then
the frequency response analysis can also omit damping. However, if damping is to be included,
then use the correct damping in your frequency response analysis. The proper specification of
damping can be verified by looking at the half-power bandwidth.
Plots are important at this stage to assist in results interpretation. X-Y plots are necessary in order
to see the variation in response versus frequency. Deformed structure plots at a frequency near a
resonant frequency can also help to interpret the results. If structure plots are made, look at the
imaginary component because the single degree-of-freedom (SDOF) displacement response at
resonance is purely imaginary when damping is present (this response does not occur in practice
because the response is usually due to several modes).
5. If you have time-dependent loads, perform transient response (SOL 109 or SOL 112) analysis. If
your structure is constrained, apply the load “very quickly” (over one or two time steps) as a step
function and look at the displacement results. The duration of the analysis needs to be as long as
the period of the lowest frequency mode. For an SDOF system, a quickly applied load results in
a peak displacement response that is twice the response resulting from the same load applied
statically. This peak response does not occur in your actual model because of multiple modes and
damping, but the results should be close. If your structure is unconstrained, the displacements will
grow with time (unless the rigid-body modes are excluded in a modal transient response analysis).
However, the stresses should be roughly twice those from the static analysis. In any event,
examine the results via X-Y plots to ensure reasonableness.
Once you are satisfied, apply the correct time variation to the load and compute the results. Again,
use X-Y plots to verify the accuracy of the results.
6. Finally, perform any other dynamic analyses, such as response spectrum analysis, random
response, nonlinear transient response, or dynamic response optimization. The confidence gained
by using the previous steps first helps to ensure that you have an accurate model at this stage.
These and other guidelines are described further in the remainder of this chapter.
Main Index
CHAPTER 14 291
Guidelines and Tools for Effective Dynamic Analysis
Main Index
292 MD Nastran Dynamic Analysis User’s Guide
Units
Units
Mistakes in units and boundary conditions (see Boundary Conditions, 298) are among the most common
mistakes made when performing finite element analysis. MD Nastran does not assume a particular set
of units, but it does assume that they are consistent.
It is up to you to ensure that the units that you use are both consistent and correct. Table 2-2 lists
consistent units for common variables.
There are several ways to verify units. For mass, you can print the results from the grid point weight
generator and verify that the mass is correct. For stiffness, you can apply a simple load and verify that
the resulting static displacements seem reasonable. For both, you can verify that the natural frequencies
are reasonable.
These checks assume that you have enough knowledge about your structure to know when the results are
reasonable and when a mistake has been made. In other words, running small models and/or proceeding
through dynamic analysis via the steps outlined in the previous section are necessary in order to be
confident that the correct units are specified.
Main Index
CHAPTER 14 293
Guidelines and Tools for Effective Dynamic Analysis
Mass
Mistakes with mass primarily involve mistakes in mass units as described in the previous section. A
common mistake is to mix mass and weight units. Using PARAM,WTMASS does not solve this problem
because it scales all mass and weight input (except certain types of direct input matrices) and still leaves
the mixture of units. Also, the use of PARAM,WTMASS can have unwanted effects as well because it
also scales the large mass used in enforced motion, thereby scaling the value of the enforced motion
input. In order to reduce the chances for error, it is recommended that only mass units be used
everywhere. Therefore, avoid the use of weight input wherever possible as well as the use of
PARAM,WTMASS.
One way to verify mass input is to apply gravity in several directions and print the SPC forces, which
shall match the weight of the structure. This verification ensures that the mass is correctly applied and
that the units are correct.
Main Index
294 MD Nastran Dynamic Analysis User’s Guide
Damping
Damping
Proper specification of damping is probably the most difficult modeling input to verify because its
verification can only be done via dynamic response analysis. In addition, its units are not necessarily
familiar because damping is normally not a modeling input that you use frequently, unless dynamic
response analysis is all you run. Also, there are several ways to specify damping in MD Nastran, which
increase the chances of making a mistake. Finally, even though the damping units are correct and the
damping is input correctly in MD Nastran, it is difficult to know that the damping specification itself is
correct from a physical standpoint. While there are relatively easy checks for correctness in mass and
stiffness input, including comparison to static and modal test data, there are no such easy checks for
damping input.
The easiest way to specify damping is to use modal damping, which is often specified as the percentage
of critical damping. The TABDMP1 Bulk Data entry is used to specify modal damping. The following
are several potential mistakes associated with using modal damping.
• Forgetting to select the TABDMP1 entry using the SDAMPING Case Control command.
• Forgetting to specify the damping TYPE (field 3). The default damping type is structural
damping, which is twice the critical damping percentage.
• Not making the table cover an adequate range of frequencies. Like almost all of the MD Nastran
tables, the TABDMP1 entry extrapolates beyond the endpoints of the table by using the first two
or the last two entered values. As a rule, you should try to provide enough input points so that
the table lookup point is always an interpolated value.
Modal damping can be used only in modal frequency response and modal transient response. Other
forms of damping have to be used for the direct methods of response. For frequency response analysis,
GE (field 9 of the MAT1 entry) and PARAM,G, GFL define structural damping. These variables are also
used to specify structural damping for transient response analysis but are not activated unless
PARAM,W3, W4, W3FL, W4FL are set to nonzero values. A common mistake is to forget to set these
values.
Note that damping is additive, that is, the damping matrix is comprised of all of the input damping
sources. For example, damping due to CDAMPi elements adds to damping due to PARAM,G and GE.
Mixing of damping types can increase the chances for error, and you are cautioned against the mixing of
damping types unless it is necessary.
In many cases damping is not an important consideration. For example, a structure's peak response due
to an impulsive load is relatively unaffected by damping since the peak response occurs during the first
cycle of response. Damping in a long duration transient excitation, such as an earthquake, can make a
difference in the peak response on the order of 10 to 20% or so, but this difference is small when
compared to the other modeling uncertainties. Therefore, it is often conservative to ignore damping in a
transient response analysis.
For frequency response analysis, however, the value of damping is critical to the value of the computed
response at resonance since the dynamic magnification factor is inversely proportional to the damping
value. One way to verify the accuracy of the modal damping ratio input is to run the modal frequency
response across the half-power bandwidth of the modal frequencies of interest as shown in Figure 14-2.
Main Index
CHAPTER 14 295
Guidelines and Tools for Effective Dynamic Analysis
Peak Response
Peak
----------- = Half-Power Point
2
Response
f1 fn f2 Frequency
Half-Power
Bandwidth
For lightly damped structures ( 0.1 ), the approximate relationship between the critical damping
ratio, the half-power bandwidth ( f 2 – f 1 ), and the resonant frequency f n is
f2 – f1
--------------- (14-1)
2f n
Another approximate way to verify damping is to run transient response analysis and look at the
successive peak values of the free vibration response as shown in Figure 14-3. In direct transient
response analysis, this plot can be generated by using a displacement initial condition; in modal transient
response analysis, it can be generated by applying an impulsive force that ramps up and down quickly
with respect to the dominant period of response. The logarithmic decrement is the natural log of the
amplitude ratio of two successive cycles of free vibration response given by
x n – 1
= ln ----------------
- (14-2)
xn
The logarithmic decrement provides an approximate damping relationship for lightly damped structures
as given by
------ (14-3)
2
In transient response analysis, remember to use PARAM,W3 or PARAM,W4 to include structural
damping if GE or PARAM,G is used.
Main Index
296 MD Nastran Dynamic Analysis User’s Guide
Damping
Both the half-power bandwidth method and the logarithmic decrement method assume an SDOF
response. These approximations are less accurate when there are multiple modes of response; however,
they are useful for verifying that the damping input is within a factor of two or three of the desired
damping.
Main Index
CHAPTER 14 297
Guidelines and Tools for Effective Dynamic Analysis
You can prevent the program from terminating the job as above by specifying MDAMPEXT=1 [or
SYSTEM(426)=1] on the NASTRAN statement. The user fatal message mentioned above does inform
the user of this avoidance scheme.
Main Index
298 MD Nastran Dynamic Analysis User’s Guide
Boundary Conditions
Boundary Conditions
The proper specification of boundary conditions is just as important for dynamic analysis as it is for static
analysis. The improper specification of the boundary conditions leads to incorrect answers. One such
improper specification of boundary conditions is forgetting to fully constrain the structure. Unlike static
analysis, for which an under-constrained model does not run, an under-constrained model does run in
dynamic analysis. You should perform an eigenvalue analysis first and verify that there are no unwanted
rigid-body modes.
The large mass used for enforced motion simulates a constrained condition as well as adds a rigid-body
mode. The value of the large mass is important for obtaining accurate answers. The large mass value
must be large enough to properly simulate the constrained condition, but it must not be so large as to
create numerical difficulties when solving the equations. A range of 10 3 to 10 6 times the overall
structural mass is adequate to meet both conditions. One way to verify that a proper value is chosen is
to run a normal modes analysis with the enforced DOF constrained via SPCs. Then, run a normal modes
analysis with your choice of the large mass(es) and compare the frequencies of the flexible modes. If the
frequencies compare favorably (i.e., to within four or five significant digits), then the large mass value is
accurate. If the frequencies do not compare, then increase the value of the large mass(es). A common
mistake is to use too low a value (or omit it entirely) for the rotational components. Using the grid point
weight generator is very important to obtain the overall structural mass and inertias so that you can
specify a good value for the large mass(es).
Main Index
CHAPTER 14 299
Guidelines and Tools for Effective Dynamic Analysis
Loads
Because of their time- or frequency-varying nature, it is more complicated to apply dynamic loads than
it is to apply static loads. Therefore, it is important to verify that the dynamic loads are correctly
specified and that there are no discontinuous loads.
The best way to verify proper dynamic load specification is to plot the loads as a function of time or
frequency. Another way to verify proper dynamic load specification is to print the loads as a function of
time or frequency. Use the OLOAD Case Control command to plot or print the loads.
Main Index
300 MD Nastran Dynamic Analysis User’s Guide
Meshing
Meshing
An adequate element mesh is required to create an accurate model. For static analysis the mesh density
is primarily controlled by the load paths; the element mesh must be fine enough so that there is a smooth
transition of stress from one element to another in the region of interest.
Load paths are also important for dynamic analysis, but there is an additional consideration: the mesh
must be fine enough to accurately represent the deformed shape of the modes of interest. If many modes
are to be considered in the analysis, then the model must be fine enough to accurately represent the mode
shapes of the highest modes of interest. Table 14-1 shows the frequencies resulting from several
cantilever beam models; each column represents the same structure but with a different number of
elements. Note that the frequencies are closer to the theoretical results as the model fineness increased.
In addition, note that the error is greatest in the higher modes. This table shows the computed frequencies
compared to theory; it does not show the computed mode shapes compared to theory. However, the
comparison for mode shapes shows even more error than is shown for the frequencies.
A general rule is to use at least five to ten grid points per half-cycle of response amplitude. Figure 14-4
shows the theoretical response shape for the fifth mode. Note that there are four half-cycles in the mode
shape, which means that 20 to 40 grid points at a minimum are recommended to accurately represent that
mode shape. This modeling guideline is also reflected in Table 14-1, which shows that the 40-element
model is much more accurate than the 10-element model.
Four Half-Cycles
Main Index
CHAPTER 14 301
Guidelines and Tools for Effective Dynamic Analysis
Another way to verify the accuracy of the mesh density is to apply static loads that give a deformed shape
the same as the mode of interest and perform stress discontinuity calculations. This process can be
laborious and is not recommended as a general checkout procedure.
Main Index
302 MD Nastran Dynamic Analysis User’s Guide
Eigenvalue Analysis
Eigenvalue Analysis
In addition to the meshing guidelines described earlier, the other primary factor in eigenvalue analysis is
the proper selection of the eigenvalue solution method. The Lanczos method is the best overall method
and is the recommended method to use. The automatic Householder method is useful for small, dense
matrices. The SINV method is useful when only a few modes are required. The other methods should
be regarded as backup methods.
Carefully examine the computed frequencies and mode shapes. Viewing only one or the other is usually
not enough to verify accuracy of your model. Modes with 0.0 Hz (or computational zero) frequencies
indicate rigid-body or mechanism modes. If these modes are unintended, then there is a mistake in
boundary conditions or connectivity. The existence and cause of unintended zero-frequency modes can
also be ascertained from the mode shapes. In addition, mode shape plots are useful for assessing local
modes, in which a group of one or a few grid points displaces and the rest of the structure does not. Local
modes may also be unintended and are often the result of incorrect connectivity or element properties.
Main Index
CHAPTER 14 303
Guidelines and Tools for Effective Dynamic Analysis
The frequency increment is defined by f on the FREQ1 Bulk Data entry. Note that FREQ and FREQ2
entries can be used in conjunction with FREQ1 to define more solutions in the areas of resonance (the
frequencies of these solutions should have been determined by a prior normal modes analysis). A
nonuniform f imposes no cost increase relative to a uniform f .
f 2 – f 1 = m – 1 f (14-4)
f2 – f1 m – 1 f
--------------- = ------------------------- (14-5)
2f n 2f n
2f n
f = ------------------
- (14-6)
m – 1
Main Index
304 MD Nastran Dynamic Analysis User’s Guide
Frequency Response Analysis
where m is the number of frequency points within the half-power bandwidth. For example, with 1%
critical damping and six points used to define the half-power bandwidth, the maximum frequency
increment is 0.004 f n . The frequency increment is smaller for lighter damped structures.
Another good check is looking at the X-Y plots. If the response curves are not smooth, there is a good
chance that f is too large.
Main Index
CHAPTER 14 305
Guidelines and Tools for Effective Dynamic Analysis
• t must be small enough to accurately capture the magnitude of the peak response, which
means that at least ten time steps per cycle of the highest mode be used. For example, if the
highest frequency of interest is 100 Hz, then t should be 0.001 second or smaller.
• t must be small enough to accurately represent the frequency content of the applied loading.
If the applied loading has a frequency content of 1000 Hz, then t must be 0.001 second or less
(preferably much less in order to represent the applied loading with more than one point per
cycle).
The integration time step can be changed in a transient response analysis, but it is not recommended.
Much of the cost of direct transient response occurs with the decomposition of the dynamic matrix,
which occurs only once if there is a constant t . A new decomposition is performed every time t
changes, thereby increasing the cost of the analysis if the time increment is changed. Therefore, the
recommendation is to determine the smallest t required based on the loading and frequency range and
then use that t throughout the entire transient response analysis.
Main Index
306 MD Nastran Dynamic Analysis User’s Guide
Transient Response Analysis
Value of Damping
The proper selection of the damping value is relatively unimportant for analyses that are of very short
duration, such as a crash impulse or a shock blast. The specification of the damping value is most
important for long duration loadings (such as earthquakes) and is critical for loadings (such as sine
dwells) that continually add energy into the system.
Main Index
CHAPTER 14 307
Guidelines and Tools for Effective Dynamic Analysis
6 11
1N 2N
1.0
Load
Factor
0.003 0.006
Time (sec)
Main Index
308 MD Nastran Dynamic Analysis User’s Guide
Results Interpretation and Verification
Table 14-2 shows the y-displacements for grid points 6 and 11. As expected, the frequency response
results at 0.0 Hz are the same as the static analysis results (see Frequency Response Analysis, 303). The
frequency response results at 2.05 Hz (very near the first natural frequency) are approximately 25 times
the static analysis results. The factor of 25 is the dynamic amplification factor at resonance for a damping
ratio of 2% critical damping.
1 1
Amplification Factor at Resonance = ------ = ---------------- = 25
2 2 .02
The transient response results are approximately twice the static analysis results. The factor of two is the
amplification of response for a transient load applied suddenly (see Transient Response Analysis, 305).
Main Index
CHAPTER 14 309
Guidelines and Tools for Effective Dynamic Analysis
Note that these memory requirements are for no spill conditions. The symmetric solver can run with less
memory, but because of spill conditions, the performance is degraded in both computer runtime and disk
space usage.
Empirical formulas were developed for these benchmark problems to estimate the disk space
requirement given the number of degrees-of-freedom (DOF) contained in the model.
For the equation
Main Index
310 MD Nastran Dynamic Analysis User’s Guide
Computer Resource Requirements
These values were obtained from a 32-bit computer and therefore should be doubled for a 64-bit
computer. If you wish to calculate more than ten modes, then multiply the space calculated above by
#modes 10 – 1
1.18** . For example, for 30 modes you would multiply the space value calculated above
by 1.18**2.
Main Index
CHAPTER 14 311
Guidelines and Tools for Effective Dynamic Analysis
Modeling Hints
There is no single checklist that will ensure a complete check of a comprehensive finite element model.
Also, there is no substitute for actual test correlation with the model, nor is there a substitute for the
analyst’s engineering interpretation of the output and one’s intuition. A results prediction, which is
determining gross results before the analysis, can be used to good advantage. Simple load paths or
natural frequencies of equivalent simple beam/mass systems, etc., can be used to remove redundancies
and predict the results. In fact, this must be accomplished to some degree to size the model initially. This
will also provide baseline data, and the effects of finer modeling will then be known.
Relative to preparation of the pilot model, the following are a few suggestions for eliminating or reducing
modeling problems.
• Start construction of a simple pilot model in which the user should:
• Use beams and plates instead of solids.
• Use RBE2s and RBE3s where they will simplify.
• Simplify modeling offsets and local modeling details.
• Ignore minor discontinuities such as holes and fillets.
Further refinement after this initial modeling should yield acceptable changes in the results.
• Do not rely on bending capability of thin plates and long, thin axial members to render the
model kinematically stable.
• Make an initial run with membrane-only properties and pinned ended bars, and check for
irregularities.
• Avoid use of AUTOSPC in the final model.
• For shells, use the parameters K6ROT or SNORM.
The post-analysis assessment should include a check of the physical significance of the loads and of the
load path. Offsets whose moments are not properly accounted for may overly weaken a very stiff load
path. Also, large moments in relatively weak bending members or plates may indicate modeling
problems.
Stress analysis should be performed at the detailed part level with the loads from the model. The use of
element stresses directly from the output of the model requires detailed review in most cases. In fact,
Main Index
312 MD Nastran Dynamic Analysis User’s Guide
Debugging Dynamic Models
model properties may be intentionally different from the actual hardware to obtain correct load
distributions, and to match test data or dynamic characteristics. Effective thicknesses or reduced bending
properties may have been used to reflect panel cutouts or partial beam and fixity. In this event, the finite
element model loads should be used with the actual drawing or as-built dimensions for detail stress
analysis. This piece-part assessment ensures a check and balance of the finite element model and the
stress distributions visualized and treated by the element selection. Also, the source of the components
of stress are known, that is, whether the predominant stress component is due to bending or axial loads.
Load transformation matrices are useful for isolating critical design conditions but are not necessarily a
sufficient basis for computing the margin of safety.
One area in which an underestimation of load could occur is the local response of small masses during a
dynamic analysis. These should be addressed in the detailed stress analysis with both the model
predictions and an alternate loading such as a specified loading condition. For the model to give correct
loads for the local response of a mass, one needs all of the following:
• Mass must be represented by enough points to characterize the energy of the critical local mode
(a single-point mass may not be sufficient).
• Mass must be supported by proper elastic elements to represent the local mode (RBE2 or RBE3
may not be sufficient).
• Mass must be in the ASET.
• Model and all analysis (input spectra, etc.) must be carried beyond this local critical mode (as far
as frequency is concerned).
Geometry Plots
The MD Nastran plotting package Patran, or another preprocessor graphics package should be used to
obtain visual images of the finite element model from many views in such a way as to provide a clear
representation of each element in at least one view and to verify overall geometry and placement of
elements.A shrink option should be used if possible to make sure all elements are present (see
Figure 14-7). This is particularly helpful when bars or beams are used to model stringers along the edges
of plate elements. Discontinuities show up only when the shrink option is invoked.
Main Index
CHAPTER 14 313
Guidelines and Tools for Effective Dynamic Analysis
Mass Distribution
PARAM,GRDPNT uses the Grid Point Weight Generator (GPWG), which gives the mass, the CG, the
moments of inertia, and the principal moments of inertia and their direction cosines. Full use should be
made of this diagnostic tool to correlate the model with existing hardware or mass properties
calculations. The GPWG uses only the weight properties and geometry to calculate mass properties.
The resultant mass properties are also called the rigid body mass properties. See Grid Point Weight
Main Index
314 MD Nastran Dynamic Analysis User’s Guide
Debugging Dynamic Models
Generator (Ch. 19) in the MD Nastran Linear Static Analysis User’s Guide for a detailed description.
Note that PARAM,WTMASS does not affect the GPWG output—it is factored out of the printed results.
Main Index
CHAPTER 14 315
Guidelines and Tools for Effective Dynamic Analysis
Weight Checks
A weight check is similar to the GRID POINT WEIGHT GENERATOR output (requested with
PARAM,GRDPNT) but with additional options also see Weightcheck (Ch. 3) and Weight Checks
(PARAM,GRDPNT,WEIGHTCHECK) (Ch. 10) in the MD Nastran Linear Static Analysis User’s Guide.
The weight check is requested with the WEIGHTCHECK Case Control command.
PRINT
PUNCH, SET = G, N, N + AUTOSPC, F, A,V
NOPRINT ALL
WEIGHTCHECK = YES
NO
GRID = gid, CGI = YES WEIGHT
NO MASS
SET
Keywords DOF Set Description
G g-set before single point, multipoint constraints, and rigid elements
are applied
N n-set after multipoint constraints and rigid elements are applied
N+AUTO n-set with same as the n-set with the rows/columns in the mass matrix
AUTOSPC corresponding to degrees-of-freedom constrained by the
PARAM,AUTOSPC operation zeroed out
F f-set after single point, multipoint constraints, and rigid elements are
applied
A a-set after static condensation
V v-set Component mode mass which combines the o-set, r-set, and c-
set
The GRID keyword references the grid point for the calculation of the rigid body motion. For SET<>G,
the keyword CGI=YES requests output of center gravity and mass moments of intertia calculation.
It should be noted that the q-set degrees-of-freedom in the current superelement and those from upstream
superelements are eliminated from the check. It should also be noted that in superelement analysis, the
weight check includes the weight of upstream superelements.
At DOF sets below the g-set, the weight check also computes weight loss or gain from the g-set.
Main Index
316 MD Nastran Dynamic Analysis User’s Guide
Weight Checks
O U T P U T F R O M W E I G H T C H E C K
DEGREES OF FREEDOM SET = A
REFERENCE POINT = 0
M O
* 1.559152E-02 -1.308159E-18 0.000000E+00 0.000000E+00 0.000000E+00 -4.915227E-02 *
* -6.540795E-19 1.410569E-02 0.000000E+00 0.000000E+00 0.000000E+00 9.119664E-20 *
* 0.000000E+00 0.000000E+00 1.516500E-02 4.761092E-02 3.401213E-17 0.000000E+00 *
* 0.000000E+00 0.000000E+00 4.761092E-02 1.513152E-01 9.075353E-17 0.000000E+00 *
* 0.000000E+00 0.000000E+00 3.401213E-17 9.189817E-17 2.319450E-03 0.000000E+00 *
* -4.915227E-02 0.000000E+00 0.000000E+00 0.000000E+00 0.000000E+00 1.578139E-01 *
The first matrix printed out is simply the rigid body mass matrix for the a-set mass. The second matrix is
the comparison of the diagonal terms with respect to the g-set WEIGHTCHECK output.
G must be one of the selected sets in the SET keyword in order to get the comparison. If G is not one of
the selected sets, then “N/A” will be printed for the comparison.
WEIGHTCHECK is available in all solution sequences. However, in SOLs 101, 105, 114, and 116,
because no mass reduction is performed, only WEIGHTCHECK(SET=G) is available.
Main Index
CHAPTER 14 317
Guidelines and Tools for Effective Dynamic Analysis
Groundcheck
The GROUNDCHECK Case Control command can help identify unintentional constraints and ill-
conditioned stiffness matrix. See Groundcheck (Ch. 3) or Unintentional Grounding (GROUNDCHECK)
(Ch. 10) in the MD Nastran Linear Static Analysis User’s Guide for an overview of this capability.
Main Index
318 MD Nastran Dynamic Analysis User’s Guide
Static Loading
Static Loading
Static loading, such as a simple gravity load, can be helpful in checking out various properties of finite
element models. Displacements, element forces, and support reactions (SPC forces) derived from 1G
loading conditions provide a first check on mass, stiffness, and determinacy of supports. Weight and CG
can be calculated from SPC forces, which should also be compared to any applied loads or weight. Load
paths can also be assessed using the element forces. Epsilon, Max Ratio, and SPC forces (at grids other
than legitimate boundary conditions) describe the overall health of the stiffness matrix.
Allowable values for these quantities are as follows:
–6
Epsilon* 1.0 10 (large model)
–9
1.0 10 (small model)
+5
Max Ratio 1.0 10
–5
SPC Forces (at internal points) 1.0 10 (model weight)
–3
SPC Moments (at internal points) 1.0 10 (model weight) (unit length)
*Epsilon is machine dependent. The above data is for CDC 64-bitword. Other machines
should give smaller numbers ( 1.0 10 – 8 and 1.0 10 –11 ). Mechanisms or symmetry
conditions may require reevaluation of SPC force limits.
The 1G cases also provide a rough approximation of the frequency of the first mode. This approximation
1
can be accomplished by using the displacement ( D ) at the CG in the equation FN ------ GD.
2
A 1G static load case can easily be obtained through the GRAV entry. This is preferred over the inertia
relief type method.
Static forces and moments can be applied to generate displacements, element, and SPC forces. The
magnitude and point of application of the forces should be representative of typical structural loading,
thereby allowing the analyst a good feel for the size of displacements and forces as in the 1G cases. It is
also helpful to use element strain energy and grid point force balance with these runs. For information
on the details of requesting these capabilities, see Case Control Commands, ESE, 301 and GPFORCE
(p. 343) in the MD Nastran Quick Reference Guide.
Main Index
CHAPTER 14 319
Guidelines and Tools for Effective Dynamic Analysis
Main Index
320 MD Nastran Dynamic Analysis User’s Guide
Modal Analysis
Modal Analysis
There are several diagnostic tools that can be used to further assess the integrity of the model. These tools
(effective mass, strain energy, kinetic energy, deformed plots) are outlined in the following paragraphs.
Main Index
CHAPTER 14 321
Guidelines and Tools for Effective Dynamic Analysis
where:
This participation factor has some interesting properties, one of which is its use in Response Spectrum
analysis. However, a more interesting use—at least for the general engineering public—may be
described as follows:
If all modes of a model are obtained, then a linear combination of the eigenvectors can describe any
possible motion.
Therefore, we may state:
DR = (14-10)
where:
D R = a rigid-body vector
= the eigenvectors
= scaling factors for the eigenvectors
T T
MD R = M (14-11)
which becomes:
T
MD R = M i i (14-12)
where:
Main Index
322 MD Nastran Dynamic Analysis User’s Guide
Modal Effective Mass Computation
Therefore, we may state that the participation factor is simply the scaling factor multiplied by the
generalized mass of the mode. If the modes are scaled to unit generalized mass, the participation factor
is equal to the scaling factor (this is the most often used case).
Taking this a step further, the `rigid-body' mass M r of a structure in the direction we are considering is
T
M r = D R MD R (14-13)
but
DR = (14-14)
or
T
M r = M ii (14-16)
Since M i i is a diagonal matrix, the contribution (or amount of the rigid-body mass), which each mode
provides, is:
2
i Mi i (14-17)
The proportion of the total mass for each mode is known as its modal effective mass and the modal
effective weight is simply the modal effective mass multiplied by the acceleration of gravity (or
1/WTMASS).
The Case Control command, MEFFMASS, has the option to calculate and print these values as standard
output whenever the system modes are calculated. The command is a very general one, with options for
printed and/or punched output. The user may select individual items or the ALL option. The output
occurs immediately following the Eigenvalue Summary Table.
The command has the following form:
SUMMARY, PARTFAC,
MEFFMASS PRINT PUNCH GRID = gid, = YES
MEFFM, MEFFW,
NOPRINT NOPUNCH NO
FRACSUM, ALL
Examples
MEFFMASS
MEFFMASS(GRID=12,SUMMARY,PARTFAC)
Main Index
CHAPTER 14 323
Guidelines and Tools for Effective Dynamic Analysis
Describers Meaning
PRINT Write output to the print file. (Default)
NOPRINT Do not write output to the print file.
PUNCH Write output to the punch file.
NOPUNCH Do not write output to the punch file. (Default)
gid Reference grid point for the calculation of the Rigid Body Mass Matrix.
SUMMARY Requests calculation of the Total Effective Mass Fraction, Modal Effective Mass
Matrix, and the A-set Rigid Body Mass Matrix. (Default)
PARTFAC Requests calculation of Modal Participation Factors.
MEFFM Requests calculation of Modal Effective Mass in units of mass.
MEFFW Requests calculation of the Modal Effective Mass in units of weight.
FRACSUM Requests calculation of the Modal Effective Mass Fraction.
Main Index
324 MD Nastran Dynamic Analysis User’s Guide
Modal Kinetic and Strain Energy Output
MODE
SORT1 PRINT REAL or IMAG
MODALKE ( PUNCH ESORT = ASCEND
SORT2 NOPRINT PHASE
DESCENT
ALL
TIME ALL AVERAGE
THRESH = e = AMPLITUDE ) = n
FREQ r
PEAK NONE
MODE
SORT1 PRINT REAL or IMAG
MODALSE ( PUNCH ESORT = ASCEND
SORT2 NOPRINT PHASE
DESCENT
ALL
TIME ALL AVERAGE
THRESH = e
= AMPLITUDE ) = n
FREQ r
PEAK NONE
Main Index
CHAPTER 14 325
Guidelines and Tools for Effective Dynamic Analysis
Examples
MODALSE= ALL
SET 100= 1, 3, 4, 5, 7
MODALKE(ESORT=ASCEND,THRESH=.0001)= 100
Describer Meaning
SORT1 Output will be presented as a tabular listing of modes for each frequency or time
step.
SORT2 Output will be presented as a tabular listing of frequencies or time steps for each
mode.
PRINT Write the results to the .f06 file (Default).
NOPRINT Do not write the results to the .f06 file.
PUNCH Write the results to the punch (.f07) file.
ESORT Present the modal energies sorted by mode number, ascending energy value or
descending energy value (Default is MODE)
THRESH Write out only those energies greater than e (Default = 0.001).
MODES Compute energies for all modes or the set of mode numbers defined in SET n
(Default = ALL).
TIME or FREQ Compute energies at all time steps, or frequencies, or the set of frequencies
defined by SET r (Default = ALL).
AVERAGE Requests average energy in frequency response analysis only (Default).
AMPLITUDE Requests amplitude of energy in frequency response analysis only.
PEAK Requests peak energy for frequency response analysis only. PEAK is the sum
of AVERAGE and AMPLITUDE.
ALL, n, NONE Compute modal energies for (1) all modes, (2) the modes defined on SET n, or
(3) no modes.
MODE NUMBERActualNormalizedFractional
ixxx.xxxxx.xxxxx.xx
i+1...
....
....
i+n...
Above form repeated for each time or frequency.
Main Index
326 MD Nastran Dynamic Analysis User’s Guide
Modal Kinetic and Strain Energy Output
Mode Number: i
FREQUENCY = 1.000000E+00
M O D A L S T R A I N E N E R G Y
MODE NUMBER = 1
M O D A L K I N E T I C E N E R G Y
Main Index
CHAPTER 14 327
Guidelines and Tools for Effective Dynamic Analysis
ALL
PRINT
GPKE PUNCH, THRESH = e = n
NOPRINT
NONE
The results of the Grid Point Kinetic Energy request GPKE have the same format as a DISPLACEMENT
output:
EIGENVALUE = 1.914339E+01
CYCLES = 6.963531E-01 G R I D P O I N T K I N E T I C E N E R G Y ( P E R C E N T )
MODE NUMBER = 4
However, the values represent the relative amount of kinetic energy in each degree-of-freedom at each
GRID point. Note that the values are scaled such that the total kinetic energy is 1.0 and the printed values
are further scaled to be a percentage of the total.
For models using the lumped mass formulation, the Grid Point Kinetic energy can be used to examine
the distribution of kinetic energy among the grid points and is computed as:
mass mass
Ek = g Mg g g
g
Where mass
g indicates that the mass-normalized eigenvectors so that the total grid point kinetic energy
is scaled to be unity and the operator indicates term-wise matrix multiplication. The Grid Point
Kinetic Energy output has limited meaning for a coupled mass formulation. Because the coupled mass
formulation produces a coupling of mass across grid points, the sharing of kinetic energy among grid
points can occur. In general, this obscures the meaning of the computation as a means of identifying
important model parameters to control modal behavior.
Main Index
328 MD Nastran Dynamic Analysis User’s Guide
Element Strain Energy
AVERAGE ALL
PRINT, PUNCH
ESE AMPLITUDE THRESH = p = n
PLOT
PEAK NONE
Examples
ESE=ALL
ESE (PUNCH, THRESH=.0001)=19
Describer Meaning
PLOT Do not write energies to either the punch file or the print file.
PEAK Requests peak energy for frequency response analysis only. PEAK is the sum of
AVERAGE and AMPLITUDE.
THRESH Energies for elements having an energy value of less than p% will be suppressed in
all output files–print, punch, plot, op2, and xdb. THRESH overrides the value of
TINY described in Remark 1. (Default=0.001)
n Set identification number. Energy for all elements specified on the SET n command
will be computed. The SET n command must be specified in the same subcase as the
ESE command or above all subcases. (Integer>0)
Main Index
CHAPTER 14 329
Guidelines and Tools for Effective Dynamic Analysis
For frequency response analysis, the energy may be computed in one of three ways as selected by
• AVERAGE:
1 T T
E o = --- u r K e u r + u i K e u i
4
• AMPLITUDE:
1 T T 2 T 2
E a = --- u r K e u r – u i K e u i + 2 u r K e u i
4
• PEAK:
Ep e a k = Eo + Ea
where:
E = elemental energy
u r = displacement - real part
u i = displacement - imaginary part
K e = elemental stiffness
Limitations
In SOL 111, ESE can only be requested if PARAM,DDRMM,-1 is used. Element data recovery for
thermal loads is not currently implemented in dynamics.
Following is a .f06 excerpt of average element strain energy output in frequency response analysis:
FREQUENCY = 1.000000E+00
E L E M E N T S T R A I N E N E R G I E S ( A V E R A G E )
Main Index
330 MD Nastran Dynamic Analysis User’s Guide
Element Kinetic Energy
AVERAGE ALL
PRINT, PUNCH
EKE AMPLITUDE THRESH = p = n
PLOT
PEAK NONE
Examples
EKE=ALL
EKE (PUNCH, THRESH=.0001)=19
Describer Meaning
PLOT Do not write energies to either the punch file or the print file.
PEAK Requests peak energy for frequency response analysis only. PEAK is the sum
of AVERAGE and AMPLITUDE.
THRESH Strain energies for elements having a energy value of less than p% will be
suppressed in all output files–print, punch, plot, op2, and xdb. THRESH
overrides the value of TINY described in Remark 1. (Default=0.001)
n Set identification number. Energy for all elements specified on the SET n
command will be computed. The SET n command must be specified in the
same subcase as the EKE command or above all subcases. (Integer>0)
Main Index
CHAPTER 14 331
Guidelines and Tools for Effective Dynamic Analysis
For frequency response analysis, the energy may be computed in one of three ways as selected by
• AVERAGE:
1 T T
E o = --- v r M e v r + v i M e v i
4
• AMPLITUDE:
1 T T 2 T 2
E a = --- v r M e v r – v i M e v i + 2 v r M e v i
4
• PEAK:
Ep e a k = Eo + Ea
where:
E = elemental energy
v r = velocity - real part
v i = velocity - imaginary part
M e = elemental mass
Limitations
In SOL 111, EKE can only be requested if PARAM,DDRMM,-1 is used. Element data recovery for
thermal loads is not currently implemented in dynamics.
Following is a .f06 excerpt of average element kinetic energy output in frequency response analysis:
FREQUENCY = 1.000000E+00
E L E M E N T K I N E T I C E N E R G I E S ( A V E R A G E )
Main Index
332 MD Nastran Dynamic Analysis User’s Guide
Element Energy Loss
AVERAGE ALL
PRINT, PUNCH
EDE AMPLITUDE THRESH = p = n
PLOT
PEAK NONE
Examples
EDE=ALL
EDE (PUNCH, THRESH=.0001)=19
Describer Meaning
PLOT Do not write energies to either the punch file or the print file.
PEAK Requests peak energy for frequency response analysis only. PEAK is the sum of
AVERAGE and AMPLITUDE.
THRESH Energies for elements having an energy value of less than p% will be suppressed
in all output files–print, punch, plot, op2, and xdb. THRESH overrides the value
of TINY described in Remark 1. (Default = 0.001)
n Set identification number. Energy for all elements specified on the SET n
command will be computed. The SET n command must be specified in the same
subcase as the EDE command or above all subcases. (Integer > 0)
Main Index
CHAPTER 14 333
Guidelines and Tools for Effective Dynamic Analysis
For frequency response analysis, the energy may be computed in one of three ways as selected by
• AVERAGE:
T T
E o = u r B e u r + u i B e u i
• AMPLITUDE:
T T 2 T 2
E a = u r B e u r – u i B e u i + 2 u r B e u i
• PEAK:
Ep e a k = Eo + Ea
where:
E = elemental energy
u r = displacement - real part
u i = displacement - imaginary part
Be = elemental damping
Limitations
FREQUENCY = 1.000000E+00
E L E M E N T E N E R G Y L O S S P E R C Y C L E ( A V E R A G E )
Main Index
334 MD Nastran Dynamic Analysis User’s Guide
Total Energy Output for Defined SETs
Describer Meaning
n SETP identification number. Any SETP may be redefined by reassigning its
identification number. SETPs specified under a SUBCASE command are
recognized for that SUBCASE only. (Integer > 0)
SET Identification numbers. If no such identification number exists, the request
is ignored. (Integer > 0)
EXCEPT Set identification numbers following EXCEPT will be deleted from output list as
long as they are in the range of the set defined by the immediately preceding
THRU. An EXCEPT list may not include a THRU list or ALL.
Remarks
1. A SETP command may be more than one physical command. A comma at the end of a physical
command signifies a continuation command. Commas may not end a set. THRU may not be used
for continuation. Place a number after the THRU.
2. Set identification numbers following EXCEPT within the range of the THRU must be in
ascending order.
In SET 88 above, the numbers 77, 78, etc., are included in the set because they are outside the
prior THRU range.
3. SETP usage is limited to the EDE, EKE, and ESE Case Control Commands.
Main Index
CHAPTER 14 335
Guidelines and Tools for Effective Dynamic Analysis
Output Example
0 SUBCASE 1001
E L E M E N T S T R A I N E N E R G I E S
0 SUBCASE 1001
E L E M E N T S T R A I N E N E R G I E S
Main Index
336 MD Nastran Dynamic Analysis User’s Guide
Total Energy Output for Defined SETs
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 15: Results Processing
15 Results Processing
Overview
Structure Plotting
X-Y Plotting
Main Index
338 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
Plotted output is important in verifying your model and understanding its results. Plots show information
in a format that is much easier to interpret than printed output. Plots are especially important for dynamic
analysis because the analysis can produce voluminous output. For example, consider a transient response
analysis for which there are 1000 output time steps, 100 grid points of interest, and 10 elements of
interest. Printed output is too large to interpret efficiently and effectively, and it does not easily show the
time variation. Plotted output overcomes these problems.
There are two kinds of plotted output: structure plots and X-Y plots. Structure plots can depict the entire
structure or a portion of it. Structure plots are useful for verifying proper geometry and connectivity.
They also can be used to show the deformed shape or stress contours at a specified time or frequency.
X-Y plots, on the other hand, show how a single-response quantity, such as a grid point displacement or
element stress, varies across a portion or all of the time or frequency range.
There are numerous commercial and in-house plotting programs that interface to MD Nastran for
structure and/or X-Y plotting. PATRAN and SimX are two good interactive programs for plotting.
The commercial programs are similar because they operate interactively. The MD Nastran plot
capabilities, on the other hand, are performed as a batch operation, which means that you predefine your
plots when you make your MD Nastran run.
Interactive plots are easier to change and to zoom or fly around the model. The batch operations can be
very useful for design studies where the plot requests stay the same over many jobs.
This chapter briefly describes the kinds of plots available in MD Nastran.
Main Index
CHAPTER 15 339
Results Processing
Structure Plotting
Structure plotting is performed to verify the model’s geometry and element connectivity prior to
performing a dynamic analysis.
After (or during) the analysis, structure plotting is performed to view deformed shapes and contours. For
dynamic response, deformed shape and contour plots can be made for normal modes analysis (for which
there is a plot, or set of plots, per mode), frequency response analysis (for which there is a plot, or set of
plots, per output frequency), and transient response analysis (for which there is a plot, or set of plots, per
output time).
Structure plot commands are described in the MD Nastran Reference Manual. In the MD Nastran input
file, structure plotting commands are listed in the OUTPUT(PLOT) Commands, which immediately
precedes the Bulk Data Section. The structure plotting commands define the set of elements to be plotted
(SET), the viewing axes (AXES), the viewing angles (VIEW) as well as the plot type and parameters
(PLOT). Optionally, the scale of the plotted deformation (MAXIMUM DEFORM) can be specified; if not
specified, the plotted deformation is scaled such that the maximum deformation is 5% of the maximum
dimension of the structure.
Listing 15-1 shows the structure plotting commands applied to a normal modes analysis of the bracket
model. See Install_dir/mdxxxx/dynamics/doc/bd15bkt.dat. Figure 15-1 shows the resulting structure
plots. The first plot shows the undeformed shape, and the next two plots show the undeformed shape
overlaid on the deformed shapes for modes 1 and 2. The default is chosen such that the maximum plotted
deformation is 5% of the maximum dimension of the bracket; the actual maximum deformation is printed
at the top of the plot. The plots shown in this chapter were converted to PostScript format for printing
on a PostScript-compatible printer.
Listing 15-1 Normal Modes Structure Plot Commands for the Bracket Model
$ PLOT COMMANDS FOR BRACKET NORMAL MODES ANALYSIS
... Executive, Case Control ...
$
$ OUTPUT REQUESTS
DISPLACEMENT(PLOT) = ALL
$
$ STRUCTURE PLOTS
OUTPUT(PLOT)
CSCALE = 3.0
SET 333 = ALL
AXES MX,MY,Z
VIEW 20.,20.,20.
FIND SCALE, ORIGIN 5, SET 333
$ PLOT UNDEFORMED SHAPE
PLOT SET 333, ORIGIN 5
$ PLOT DEFORMED, UNDEFORMED SHAPES
PLOT MODAL DEFORMATION 0,1
PLOT MODAL DEFORMATION 0,2
$
BEGIN BULK
$
... Bulk Data ...
$
ENDDATA
Main Index
340 MD Nastran Dynamic Analysis User’s Guide
Structure Plotting
Undeformed Shape
Mode 1
Mode 2
Figure 15-1 Normal Modes Structure Plots for the Bracket Model
Main Index
CHAPTER 15 341
Results Processing
Listing 15-2 shows the structure plotting commands applied to a modal frequency response analysis of
the cantilever beam model. See Install_dir/mdxxxx/doc/dynamics/bd15bar1.dat. The displacements are
computed in magnitude/phase form. PARAM,DDRMM,-1 and PARAM,CURVPLOT,1 are required in
the Bulk Data to create structure plots at specified frequencies. (Note that PARAM,DDRMM,-1
generally increases the amount of computer time and is not recommended unless otherwise required.)
Figure 15-2 shows the resulting plots. The first plot shows the undeformed shape, and the next plot
shows the magnitude of response at 2.05 Hz, which is overlaid on the undeformed shape. The default is
chosen such that the maximum plotted deformation is 5% of the length of the bar; the actual maximum
deformation is printed at the top of the plot.
Listing 15-2 Frequency Response Structure Plot Commands for the Bar Model –
Magnitude/Phase
$ PLOT COMMANDS FOR BAR MODAL FREQUENCY RESPONSE
$ --- MAGNITUDE, PHASE ---
... Executive, Case Control ...
$
$ OUTPUT REQUESTS
$ MAGNITUDE, PHASE
DISPLACEMENT(PHASE,PLOT) = ALL
$
$ STRUCTURE PLOTS
OUTPUT(PLOT)
$ DEFINE ELEMENTS IN PLOT SET
SET 333 = ALL
$ PLOT AXES: R=-Z, S=X, T=Y
AXES MZ,X,Y
VIEW 0.,0.,0.
FIND SCALE, ORIGIN 5, SET 333
$ PLOT UNDEFORMED SHAPE
PLOT SET 333, ORIGIN 5
$ PLOT DEFORMED SHAPE AT 2.05 HZ
PLOT FREQUENCY DEFORMATION 0 RANGE 2.05,2.051 MAGNITUDE SET 333
$
BEGIN BULK
$
$ REQUIRED FOR "FREQUENCY FROZEN" STRUCTURE PLOTS
PARAM,DDRMM,-1
PARAM,CURVPLOT,1
$
... rest of Bulk Data ...
$
ENDDATA
Undeformed Shape
Frequency 2.05
Figure 15-2 Frequency Response Structure Plots for the Bar Model -- Magnitude/Phase
Main Index
342 MD Nastran Dynamic Analysis User’s Guide
Structure Plotting
The same plots are regenerated except that now the displacements are in real/imaginary format (the
default). Listing 15-3 shows the plot commands. See Install_dir/mdxxxx/doc/dynamics/bd15bar2.dat.
Note that the imaginary component is selected by PHASE LAG -90. Figure 15-3 shows the resulting
plots. The default is chosen such that the maximum plotted deformation is 5% of the length of the bar;
the actual maximum deformation is printed at the top of the plot.
Listing 15-3 Frequency Response Structure Plot Commands for the Bar Model --
Real/Imaginary
$ PLOT COMMANDS FOR BAR MODAL FREQUENCY RESPONSE
$ --- REAL, IMAGINARY ---
... Executive, Case Control ...
$
$ OUTPUT REQUESTS
$ REAL, IMAGINARY
DISPLACEMENT(PLOT) = ALL
$
$ STRUCTURE PLOTS
OUTPUT(PLOT)
$ DEFINE ELEMENTS IN PLOT SET
CSCALE = 1.8
SET 333 = ALL
$ PLOT AXES: R=-Z, S=X, T=Y
AXES MZ,X,Y
VIEW 0.,0.,0.
FIND SCALE, ORIGIN 5, SET 333
$ PLOT UNDEFORMED SHAPE
PLOT SET 333, ORIGIN 5
$ PLOT DEFORMED SHAPE--REAL
PLOT FREQ DEFORM 0 RANGE 2.05,2.051 SET 333
$ PLOT DEFORMED SHAPE--IMAGINARY
PLOT FREQ DEFORM 0 RANGE 2.05,2.051 PHASE LAG -90. SET 333
$
BEGIN BULK
$
$ REQUIRED FOR "FREQUENCY FROZEN" STRUCTURE PLOTS
PARAM,DDRMM,-1
PARAM,CURVPLOT,1
$
... rest of Bulk Data ...
$
ENDDATA
Main Index
CHAPTER 15 343
Results Processing
Undeformed Shape
Listing 15-4 shows the structure plotting commands applied to a modal transient response analysis of the
cantilever beam model. See Install_dir/mdxxxx/doc/dynamics/bd15bar3.dat. PARAM,DDRMM,-1 is
required in the Bulk Data in order to create structure plots at various times and/or frequencies. (Note that
PARAM,DDRMM,-1 generally increases the amount of computer time and is not recommended unless
otherwise required.) Figure 15-4 shows the resulting plots. The plots show the displacements at the
following times: 0.25, 0.50, and 0.75 seconds, which are overlaid on the undeformed shape. The default
is chosen such that the maximum plotted deformation is 5% of the length of the bar; the actual maximum
deformation is printed at the top of the plot. For this case it may be better to specify a maximum
deformation so that the plots will show relative amplitudes.
Listing 15-4 Transient Response Structure Plot Commands for the Bar Model
$ PLOT COMMANDS FOR BAR MODAL TRANSIENT RESPONSE
... Executive, Case Control ...
$
$ OUTPUT REQUESTS
DISPLACEMENT(PLOT) = ALL
$
$ STRUCTURE PLOTS
OUTPUT(PLOT)
$ DEFINE ELEMENTS IN PLOT SET
SET 333 = ALL
$ PLOT AXES: R=-Z, S=X, T=Y
AXES MZ,X,Y
VIEW O.,O.,O.
FIND SCALE, ORIGIN 5, SET 333
$ PLOT DEFORMED SHAPE AT TIMES 0.25,0.5,0.75
PLOT TRANSIENT DEFORMATION 0 RANGE 0.25,0.251 SET 333
PLOT TRANSIENT DEFORMATION 0 RANGE 0.50,0.501 SET 333
Main Index
344 MD Nastran Dynamic Analysis User’s Guide
Structure Plotting
Time 0.25
Time 0.50
Time 0.75
Figure 15-4 Transient Response Structure Plots for the Bar Model
Although structure plotting is best performed in an interactive environment outside of MD Nastran, the
batch structure plotting capability in MD Nastran is nevertheless a useful tool for model verification and
results processing. The batch plotting capability can save time and effort when many plots are required
for a model that is run repeatedly.
Main Index
CHAPTER 15 345
Results Processing
X-Y Plotting
X-Y plots are used to display frequency and transient response results where the x-axis is frequency or
time and the y-axis is any output quantity. Unlike structure plotting, which is often performed in an
interactive environment, X-Y plotting is ideal for the MD Nastran batch environment due to the large
volume of data.
X-Y plot commands are contained in the X-Y PLOT Commands Section that immediately precedes the
Bulk Data Section. You define the titles (XTITLE and YTITLE) and plots (XYPLOT). You can specify the
plots to be generated in log format (XLOG and YLOG), and you can specify different line styles. You can
also specify that the plots are to be made in pairs (with a top and a bottom plot), which is particularly
useful for frequency response results when you want to display magnitude/phase or real/imaginary pairs.
Details about X-Y plotting and its commands are located in Plotting (Ch. 11) in the MD Nastran
Reference Manual.
Element force and stress component numbers are also described in the manual. The X-Y plot commands
use numbers to identify a single component of grid point and element data. Note that frequency response
requires the use of complex force and stress components.
Once a good set of X-Y plot commands is established, it is wise to use this set repeatedly. The examples
that follow provide a good starting point.
Listing 15-5 shows X-Y plot commands for a modal frequency response analysis of the cantilever beam
model. See Install_dir/mdxxxx/dir/dynamics/bd15bar4.dat. Plots are made in pairs in magnitude/phase
format. The “t” as the second letter in XTGRID, YTGRID, YTLOG, and YTTITLE corresponds to the
top plot of each pair; the letter “b” corresponds to the bottom plot (XBGRID, YBGRID, etc.) Plots are
made for the applied loads (OLOAD) at grid points 6 and 11, displacements (DISP) of grid points 6 and
11, bending moment (ELFORCE) at end A in plane 1 for element 6, and modal displacements (SDISP)
for modes 1 and 2. For CBAR elements, force component 2 is the bending moment component at end
A in plane 1 for real or magnitude output, and force component 10 is the similar component for
imaginary or phase output. Figure 15-5 shows the resulting plots. The plots in this chapter were
converted to PostScript format for printing on a PostScript-compatible printer. The plots in the output do
not exactly follow the input request order due to sorting for processing efficiency.
The X-Y plotter makes reasonable choices of upper and lower bounds for the axes for both the x- and
y-axes on most plots. If it does not, the bounds can be fixed with the XMIN, XMAX, YMIN, YMAX, and
their variations for half-frame curves. Instances where setting bounds explicitly results in better plots
include the following situations:
• If you expect a variable to be constant or vary only slightly but want to plot it to confirm that it
indeed does not vary, set the YMIN and YMAX to include the expected value, but separate them
by at least 10 percent of their average value.
Main Index
346 MD Nastran Dynamic Analysis User’s Guide
X-Y Plotting
For this case the automatic bound selection chooses bounds very close to each other to make the
data fill up the plot. This selection causes the bounds to be nearly equal and magnifies the scale
of the plot orders of magnitude larger than other plots made of varying functions. The response
appears to be erratic when, in fact, it is smooth within engineering criteria. The extreme cases
occur when the function varies only in its last digit. Then the function appears to be oscillating
between the upper and lower limits or it can even cause a fatal error due to numerical overflow
when it attempts to divide numbers by the difference YMAX-YMIN.
• The automatic bound selector tends to round up the bounds to integer multiples of 100. When
plotting phase angles, bounds that cause grid lines at 90 degrees are more readable. For
example, for a plot that traverses the range of 0 to 360 degrees, the usual selected bounds are 0 to
400 degrees. If you prefer to have grid lines drawn at integer multiples of 90 degrees, set YMIN
to 0.0 and YMAX at 360.0 degrees.
• When plotting log plots, any bounds you input may be rounded up or down to a value that the
plotter considers more reasonable. In general, changing bounds on log plots sometimes requires
experimentation before a reasonable set can be found.
• Producing good quality plots is an interactive process whether the plot is produced by an
interactive or batch plotter. The restart feature discussed in Restarts In Dynamic Analysis
(Ch. 10) can reduce the computer costs for this iteration since restarts performed to change only
plot requests are made efficiently.
Listing 15-5 X-Y Plot Commands for the Bar Frequency Response Analysis
$ X-Y PLOT COMMANDS FOR BAR MODAL FREQUENCY RESPONSE
$ --- MAGNITUDE, PHASE ---
...Executive, Case Control ...
$
$ APPLIED LOAD OUTPUT
SET 15 = 6,11
OLOAD(PHASE,PLOT) = 15
$
$ PHYSICAL OUTPUT REQUEST
SET 11 = 6,11
DISPLACEMENT(PHASE,PLOT) = 11
$
$ MODAL SOLUTION SET OUTPUT
SET 12 = 1,2
SDISP(PHASE,PLOT) = 12
$
$ ELEMENT FORCE OUTPUT
SET 13 = 6
ELFORCE(PHASE,PLOT) = 13
$
OUTPUT(XYPLOT)
XTGRID = YES
YTGRID = YES
XBGRID = YES
YBGRID = YES
$
$ PLOT RESULTS
XTITLE = FREQUENCY
$
YTLOG = YES
YTTITLE = DISPL. MAG. 6
Main Index
CHAPTER 15 347
Results Processing
YTMAX = 8.0
YTMIN = 0.0
YTTITLE = LOAD MAG. 11
YBTITLE = LOAD PHASE 11
XYPLOT OLOAD /11(T2RM,T2IP)
$
BEGIN BULK
$
... Bulk Data ...
$
ENDDATA
Main Index
348 MD Nastran Dynamic Analysis User’s Guide
X-Y Plotting
1.0
Displ. Mag. 6
1.0E-5
400
Displ. Phase 6
1.0
Displ. Mag. 11
1.0E-4
400
Displ. Phase 11
0
0 Frequency (Hz) 20
Figure 15-5 X-Y Plots for the Bar Frequency Response Analysis
Main Index
CHAPTER 15 349
Results Processing
1.0
Sdisp. Mag. Mode1
1.0E-4
200
Sdisp. Phase Mode1
1.0E-2
Sdisp. Mag. Mode 2
1.0E-4
360
Sdisp. Phase Mode 2
160
0 Frequency (Hz) 20
Figure 15-5 X-Y Plots for the Bar Frequency Response Analysis (continued)
Main Index
350 MD Nastran Dynamic Analysis User’s Guide
X-Y Plotting
1.0E3
Bend. Moment A1 El. 6 Mag.
1.0E-1
400
Bend. Moment A1 El. 6 Phase
0
0 Frequency (Hz) 20
Figure 15-5 X-Y Plots for the Bar Frequency Response Analysis (continued)
Main Index
CHAPTER 15 351
Results Processing
4.0
Load Mag. 6
0
100
Load Phase 6
0
8.0
Load Mag. 11
0
100
LoadP hase 11
0
0 20
Frequency (Hz)
Figure 15-5 X-Y Plots for the Bar Frequency Response Analysis (continued)
Listing 15-6 shows X-Y plot commands for a modal transient response analysis of the cantilever beam
model. See Install_dir/mdxxxx/doc/dynamics/bd15bar5.dat. Plots are made for the applied loads
(OLOAD) at grid points 6 and 11, displacements (DISP) of grid points 6 and 11, accelerations (ACCE)
for grid points 6 and 11, bending moment (ELFORCE) at end A in plane 1 for element 6, and modal
displacements (SDISP) for modes 1 and 2. Figure 15-6 shows the resulting plots.
Main Index
352 MD Nastran Dynamic Analysis User’s Guide
X-Y Plotting
Listing 15-6 X-Y Plot Commands for the Bar Transient Response Analysis
$ X-Y PLOT COMMANDS FOR BAR MODAL TRANSIENT RESPONSE
... Executive, Case Control ...
$
$ PHYSICAL OUTPUT REQUEST
SET 11 = 6,11
DISPLACEMENT(PLOT) = 11
ACCELERATION(PLOT) = 11
$
$ MODAL SOLUTION SET OUTPUT
SET 12 = 1,2
SDISP(PLOT) = 12
$
$ ELEMENT FORCE OUTPUT
SET 13 = 6
ELFORCE(PLOT) = 13
$
$ APPLIED LOAD OUTPUT
SET 15 = 6,11
OLOAD(PLOT) = 15
$
OUTPUT(XYPLOT)
XGRID = YES
YGRID = YES
$
$ PLOT RESULTS
XTITLE = TIME
$
YTITLE = DISPL. 6
XYPLOT DISP /6(T2)
YTITLE = DISPL. 11
XYPLOT DISP /11(T2)
YTITLE = ACCEL. 6
XYPLOT ACCE /6(T2)
YTITLE = ACCEL. 11
XYPLOT ACCE /11(T2)
$
YTITLE = SDISP. 1
XYPLOT SDISP /1(T1)
YTITLE = SDISP. 2
XYPLOT SDISP /2(T1)
$
YTITLE = BEND. MOMENT A1, EL. 6
XYPLOT ELFORCE /6(2)
$
YTITLE = LOAD 6
XYPLOT OLOAD /6(T2)
YTITLE = LOAD 11
XYPLOT OLOAD /11(T2)
$
BEGIN BULK
$
... Bulk Data ...
$
ENDDATA
Main Index
CHAPTER 15 353
Results Processing
0.03
Displ. 6
0.0
-0.02
0.08
Displ. 11
0.0
-0.06
0.0 2.0
Time (sec)
Figure 15-6 X-Y Plots for the Bar Transient Response Analysis
Main Index
354 MD Nastran Dynamic Analysis User’s Guide
X-Y Plotting
4.0
0.0
Accel. 6
-6.0
10.0
Accel. 11
0.0
-15.0
0.0 2.0
Time (sec)
Figure 15-6 X-Y Plots for the Bar Transient Response Analysis (continued)
Main Index
CHAPTER 15 355
Results Processing
25.0
Bend. Moment A1, El. 6
0.0
-20.0
0.0 2.0
Time (sec)
Figure 15-6 X-Y Plots for the Bar Transient Response Analysis (continued)
Main Index
356 MD Nastran Dynamic Analysis User’s Guide
X-Y Plotting
0.08
Sdisp. 1
0.0
-0.10
0.0015
Sdisp. 2
0.0
-0.0010
0.0 2.0
Time (sec)
Figure 15-6 X-Y Plots for the Bar Transient Response Analysis (continued)
Main Index
CHAPTER 15 357
Results Processing
3.0
Load 6
0.0
-3.0
6.0
Load 11
0.0
-6.0
0.0 2.0
Time (sec)
Figure 15-6 X-Y Plots for the Bar Transient Response Analysis (continued)
Main Index
358 MD Nastran Dynamic Analysis User’s Guide
X-Y Plotting
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 16: Special Topics
16 Special Topics
Direct Matrix Input Dynamics
DMIGs, Extra Points, and Transfer Functions
Dynamic Reduction
Superelement Analysis
Nonlinear Transient Response Analysis
Nonlinear Transient Response Analysis Elements
Nonlinear Harmonic Response
Design Optimization and Sensitivity Overview
Control System Analysis
Fourier Transform
Squeeze Film Damper Nonlinear Force
Aeroelastic Analysis
Viscoelastic Material Properties
OPENFSI
Rotor Dynamics Overview
Main Index
Direct Matrix Input Dynamics
If structural matrices are available externally, you can input the matrices directly into MD Nastran
without providing all the modeling information. Direct Matrix Input (Ch. 12) in the MD Nastran Linear
Static Analysis User’s Guide provides the basic overview of this capability. The extensions into dynamics
will be discussed here.
The direct matrix input feature can be used to input stiffness, mass, damping, and load matrices attached
to the grid and/or scalar points in dynamic analysis. These matrices are referenced in terms of their
external grid IDs and are input via DMIG Bulk Data entries. As shown in Table 16-1, there are seven
standard kinds of DMIG matrices available in dynamic analysis, but only three in statics.
The symbols for g-type matrices in mathematical format are K 2gg , M 2gg , B 2gg , and { P 2g }. The three
matrices K2GG, M2GG, and B2GG must be real and symmetric. These matrices are implemented at the
g-set level (see The Set Notation System Used in Dynamic Analysis, 503 for a description of the set
notation for dynamic analysis). In other words, these terms are added to the corresponding structural
matrices at the specified DOFs prior to the application of constraints (MPCs, SPCs, etc.).
The symbols for p-type matrices in standard mathematical format are K 2pp , M 2pp , and B 2pp . The p-set
is a union of the g-set and extra points. These matrices need not be real or symmetric. The p-type
matrices are used in applications such as control systems. Only the g-type DMIG input matrices are
covered in this guide.
The DMIG Bulk Data user interface is described in the DMIG Bulk Data User Interface and the Case
Control user interface in the DMIG Case Control User Interface (Ch. 12) in the MD Nastran Linear Static
Analysis User’s Guide.
.Use of the DMIG entry for inputting mass and stiffness is illustrated in the DMIG Example (Ch. 3) and
in the Stiffness Matrix Using DMIG, Mass Matrix Using DMIG, and Load Matrix Using DMIG (Ch. 12) in
the MD Nastran Linear Static Analysis User’s Guide.
Main Index
CHAPTER 16 361
Special Topics
m
.
.
.
.
etc
.
g
p
q
r a
c
b d
In direct solutions, the system matrices are formed by merging structural points and e-points into the
p-set. The actual solution matrices are assembled and added together at the reduced, d-set level.
The set logic for modal formulations is as follows:
0 h
f
Main Index
362 MD Nastran Dynamic Analysis User’s Guide
DMIGs, Extra Points, and Transfer Functions
In modal solutions the system matrices are transformed to the modal coordinates and merged with the
extra point degrees-of-freedom. The modal coordinates include free body modes, 0 , and flexible
modes, f . Note that the coupling between the structural motions and the extra points is only
approximated by the truncated modal solution.
Modal Transformations
When extra points are included in the modal formulation, a transformation is needed between the
physical displacements and the modal coordinates. The displacements, u p , are defined by the
transformation
ug
u p = ----- = p h (16-1)
u
e ue
where:
p h =
g 0
0 I
and g are the real eigenvectors. The loads and direct input matrices are transformed consistently with
the displacements. For the applied loads, P ,
T
Ph = p h Pp (16-2)
Note: DMIG matrices are selected by the K2PP =, B2PP =, and the M2PP = Case Control
commands.
Main Index
CHAPTER 16 363
Special Topics
For dynamics modeling, the direct input matrices are defined for the p-set of degrees-of-freedom for
mass, damping, and stiffness matrices. The actual values are defined by the user and may represent any
type of linear solution, including fluids, electrical circuits, and external structures. However, some
practical recommendations are as follows:
1. Use the double field bulk data format (DMIG*) if more than three significant digits of the input
values are desired. Note that these solutions are sensitive to the precision of the matrix terms.
Use DMAP modules, INPUTi, for full double-precision input.
2. Use the automatic selection of single- or double-precision data for the system. The program sets
the size depending on the word size of the computer.
3. Use the symmetric matrix option if possible. The unsymmetric flag will force all operations into
the unsymmetric mode, costing time and storage space.
An example problem that uses DMIG data for generating friction forces is given in Complex
Eigensolutions, 501.
Transfer Functions
The MD Nastran transfer functions (TF inputs) are actually a convenient method for generating special
unsymmetric matrix input. When used with extra points they can define second order operators with one
output and multiple inputs. The basic equation defining individual TF input is
2 2
B 0 + B 1 p + B 2 p u d + A o i + A 1 i p + A 2 i p u i = 0 (16-4)
i
Here u d is the dependent degree-of-freedom, u i are the selected input degrees-of-freedom, and the
coefficients A and B are user-specified.
Internally, these coefficients are simply added to a single row in the matrix equation:
2
Mp + Bp + K u = P (16-5)
Limitations
If the TF terms are the only occupants of the u d row of the matrix equation, the dynamic solution will
include the basic equation. However, no internal checks are made for additional terms in the TF row in
the matrices or loads. To avoid conflicts, the following rules apply:
Main Index
364 MD Nastran Dynamic Analysis User’s Guide
DMIGs, Extra Points, and Transfer Functions
2
F j = A 0 + A 1 p + A 2 p u e (16-6)
The B j coefficients must be zero since the structural point u j is already connected to the finite elements.
This transfer function is equivalent to adding a positive force on the right-hand side of the matrix
equation.
If the force or moment is acting on an actuator with a feedback loop, include the feedback as another
transfer function.
2 3 4
A + Bp + Cp + Dp + Ep u b = G p u a (16-8)
Main Index
CHAPTER 16 365
Special Topics
this may be subdivided into two equations suitable for the TF format:
2 2
A + Bp + Cp u b + p u e1 – G p u a = 0 (16-9)
and
2
u e1 – Dp + Ep u b = 0 (16-10)
Here, u e1 is an intermediate extra point which is coupled to the u b degree-of-freedom in both equations.
Note that this method may produce a singular mass matrix that causes problems for certain complex
eigenvalue options. This would occur in the example above if E = 0 . A solution is discussed in
Complex Eigensolutions, 501.
Alternate Method
If the polynomial is available in factored form, i.e., G 1 p G 2 p G 3 p , a series of TF transfer
functions can be used as illustrated in the following sketch.
ua ue 1 u e2 ub
G1 G2 G3
u e1 – G 1 u a = 0
u e2 – G 2 u e1 = 0
ub – G3 ue 2 = 0
For polynomials in the denominator of a transfer function, use a feedback loop or coupled equations as
shown in the previous example. See the example below for a realistic application.
Example
As an example of the general approach, consider the control system shown in Figure 16-1, which might
represent a simplified model of an optical device. It contains many of the components found in control
systems including attitude and rate sensors, signal conditioners, and a nonlinear mechanical actuator
with local feedback. The structure is represented in Figure 16-1 by the displacement components, which
are sensed by the control system, or to which loads are applied.
The transfer functions (TFs) of the control system are listed in Table 16-2. The first five transfer
functions give the outputs of the devices labeled 1 to 5 in Figure 16-1 in terms of their inputs. TF 6
defines the force on the structure, at point u 11 , which is a function of the input to the mechanical actuator.
Main Index
366 MD Nastran Dynamic Analysis User’s Guide
DMIGs, Extra Points, and Transfer Functions
In the analysis, an extra point is assigned to each of the new variables, u 4 , ..., u 10 . The coefficients of
the transfer functions, expressed by TFs 1 to 6, are listed in Table 16-3. Note the coefficient of 0.0 for
the structural degree-of-freedom, u 11 , on TF 6. The existing structure provides the stiffness and mass for
this displacement.
Main Index
CHAPTER 16 367
Special Topics
t
Input
Signal 3
u6 Signal
Conditioner
Filter
u7
2 4
u1 u5
Attitude
Sensor
u2
u4
1
Rate
Gyros
u3
Structure
Actuator
Mechanical
Force
6
u 11
5 u8
Displacement
Feedback
Main Index
368 MD Nastran Dynamic Analysis User’s Guide
DMIGs, Extra Points, and Transfer Functions
1
u 5 = ------------------- Ru 1
1 + 2 p 2
1
u 6 = ------------------- u 10
1 + 3 p 3
A0 + A1 p
u 7 = ------------------------------------------- Cu 4 + Du 5 + Eu 6 4
2
B0 + B1 p + B2 p
Gu 11
u 8 = ------------------
- 5
1 + 4 p
0 u 11 = – k u 7 – u 8 6
u3 –B
2 u5 1 2 u1 –R
Main Index
CHAPTER 16 369
Special Topics
4 u7 B0 B1 B2 u4 – CA 0 – CA 1
u5 – DA 0 – DA 1
u6 – EA 0 – EA 1
5 u8 1 4 u 11 –G
6 u 11 0.0 u7 –k
+k
u8
Main Index
370 MD Nastran Dynamic Analysis User’s Guide
DMIGs, Extra Points, and Transfer Functions
GRID,1,,500.0
GRID,3,,250.0
GRID,14,,0.0
GRID,15,,-250.0
GRID,11,,-500.0
CBEAM,1,1,1,3,,,1.0
CBEAM,2,1,3,14,,,1.0
CBEAM,3,1,14,15,,,1.0
CBEAM,4,1,15,11,,,1.0
PBEAM,1,1,1000.,125+6,125+6,,250.+6
MAT1,1,69.0+6,,0.3,5.0-6
$ OPTICAL DEVICE ON THE END
CONM2,6,1,,10.0,,,,,+CNM2
+CNM2,8.0+5,,4.0+5,,,4.0+5
$ PIVOT ON THE CENTER
SPC,10,14,12
$ EXTRA POINTS ARE VOLTAGES
EPOINT,4,5,6,7,8,10
$ TRANSFER FUNCTIONS IN ORDER
$ RATE GYROS PICK UP VELOCITIES
TF,6,4,,1.0,0.015,,,,+TF101
+TF101,1,2,,-0.25,,,,,+TF102
+TF102,3,2,,-0.5
$ ATTITUDE SENSOR MEASURES ROTATION RZ
TF,6,5,,1.0,0.02,,,,+TF201
+TF201,1,6,-1.2
$ INPUT SIGNAL, E10, IS FILTERED
TF,6,6,,1.0,0.001,,,,+TF301
+TF301,10,,-1.0
$ SIGNAL CONDITIONER COMBINES THE VOLTAGES
TF,6,7,,39.48,8.885-2,1.0-4,,,+TF401
+TF401,4,,-20.0,-0.4,,,,,+TF402
+TF402,5,,-200.0,-4.0,,,,,+TF403
+TF403,6,,100.,-20.0
$ DISPLACEMENT FEEDBACK TO THE ACTUATOR
TF,6,8,,1.0,0.005,,,,+TF501
+TF501,11,2,-1.0
$ ACTUATOR FORCES ARE ADDED TO ROW OF STRUCTURAL MATRICES
TF,6,11,2,0.0,,,,,+TF601
+TF601,7,,-1.0+5,,,,,,+TF602
+TF602,8,,1.0+5
$ TEMPORARILY GROUND POINT 10 WITH A DIAGONAL TERM
TF,6,10,,1.0
ENDDATA
Main Index
CHAPTER 16 371
Special Topics
Dynamic Reduction
Dynamic reduction is an optional numerical approach that can be used in MD Nastran to reduce a
dynamic model to one with fewer degrees-of-freedom. Typically, the intent of dynamic reduction is to
increase the efficiency of a dynamic solution by working with smaller matrices while maintaining the
dynamic characteristics of the system.
Dynamic reduction is used for a number of reasons. One possible reason may be that a particular model
may be too large from a computer resource standpoint (computer runtime and/or disk space) to be solved
without using reduction. A related issue is that the model may have more detail than required. Many
times dynamic analyses are performed on models that were initially created for detailed static stress
analyses, which typically require refined meshes to accurately predict stresses. When these static models
are used in a dynamic analysis, the detailed meshes result in significantly more detail than is reasonably
required to predict the natural frequencies and mode shapes accurately.
Two methods are available in MD Nastran for the dynamic reduction of the eigenequation prior to modal
extraction: static condensation (also called Guyan reduction) and dynamic reduction (also called
component mode systhesis). Once the natural frequencies and mode shapes are calculated for the
reduced model, these modes can be used in the transient or frequency response solution process.
Static Condensation
In the static condensation process you select a set of dynamic DOFs called the a-set; these are the retained
DOFs that form the analysis set. The complementary set is called the o-set and is the set of DOFs that
are omitted from the dynamic analysis through a reduction process. The process distributes the o-set
mass, stiffness, and damping to the a-set DOFs by using a transformation that is based on a partition of
the stiffness matrix (hence the term static condensation). This reduction process is exact only for static
analysis, which leads to approximations in the dynamic response. The a-set DOFs are defined by the
ASET or ASET1 Bulk Data entires, and the o-set DOFs are defined by the OMIT or OMIT1 Bulk Data
entries.
Dynamic Reduction
Dynamic reduction offers a more accurate solution by including an approximation of the dynamic effects
when forming the transformation. Bulk Data entries for dynamic reduction are BSET, BSET1, CSET,
CSET1, QSET, and QSET1.
It is emphasized that dynamic reduction is an optional technique and is best left to the dynamic analysis
specialist. Whereas dynamic reduction was required in the days of small computer memory and disk
space, now it is no longer required due to increased computer resources and better eigenanalysis methods
(in particular, the Lanczos method).
Main Index
372 MD Nastran Dynamic Analysis User’s Guide
Dynamic Reduction
Definition
The basic dynamic equation before reduction is given in the u f set (after SPC and MPC constraints have
been applied, but before DMIGs and extra points). The standard matrix equation to be reduced is:
·· ·
M aa M a o u a Ba a Ba o ua Ka a Ka o ua Pa
·· + · + = (16-11)
M oa M o o u o Bo a Bo o uo Ko a Ko o uo Po
where:
u a , u· a , u·· a = the displacements, velocities, and accelerations of the analysis (a) set, to be retained.
u o , u· o , u·· o = the displacements, velocities, and accelerations of the omit (o) set, to be eliminated.
M, B, K = the mass, damping, and stiffness matrices (assumed to be real and symmetric).
Pa , Po = the applied loads.
Note that all free-body motions must be included in the u a partition. Otherwise, K o o will be singular.
The bar quantities ( P , etc.) indicate unreduced values.
Statics
For statics problems, we may ignore the mass and damping effects and solve the lower partition of
Eq. (16-11) for u o :
–1
u o = – K o o K oa u a + P o (16-12)
The two parts of Eq. (16-12) become the Guyan matrix G o and the static corrective displacement u oo :
–1
Go = – Ko o Ko a (16-13)
–1
u oo = K o o P o (16-14)
The exact static solution system is obtained by substituting Eq. (16-12) through Eq. (16-14) into the upper
partition terms of Eq. (16-11), resulting in the reduced equations used in the static solution
Ka a ua = Pa (16-15)
and
Main Index
CHAPTER 16 373
Special Topics
u o = G o u a + u oo (16-16)
where:
Ka a = Ka a + Ka o Go (16-17)
P a = P a + G oT P o (16-18)
In actual practice the size of the u a set is usually small compared to u o , but the reduced matrices are
dense, resulting in no savings in cost. The savings in solving Eq. (16-15) are usually offset by the costs
of calculating G o and u oo .
However, for dynamics, we also may approximate the vectors u·· o and u· o to reduce the order of the
system. A good place to start is to use the static properties. From Eq. (16-16), define the transformation
ua
u f = = H f u f (16-19)
uo
where:
ua
u f = (16-20)
u oo
I 0
Hf = (16-21)
Go I
Here u oo are the incremental displacements relative to the static shape. The system described in
Eq. (16-11) may be transformed to the new coordinates with no loss of accuracy. The stiffness matrix in
the transformed system is
I G oT Ka a Ka o I 0
K ff = (16-22)
0 I Ko a Ko o Go I
Ko o 0
K ff = (16-23)
0 Ko o
Although the stiffness matrix becomes decoupled, the mass and damping matrices tend to have more
coupling than the original system. Since the damping terms have the same form as the mass, we will not
include them here. The exact transformed system becomes
Main Index
374 MD Nastran Dynamic Analysis User’s Guide
Dynamic Reduction
··
M a a M a o u a Ka a 0 ua Pa
·· + =
M o a M o o u o 0 Ko o uo Po (16-24)
where:
T
M a a = M a a + M a o G o + G o M o a + M o o G o (16-25)
M a o = M oTa = M a o + G oT M o o (16-26)
M o o = M o o (16-27)
The damping matrix terms of B ff are similar in form to the mass matrix partitions. An alternative
derivation which does not rely on symmetric transformation is given below. Starting from Eq. (16-11)
through Eq. (16-18), we may estimate the acceleration effects of the omitted points by the equation
Substituting Eq. (16-28) into the lower partition of Eq. (16-11), and solving for u o , with damping
neglected, we obtain the approximation
Substituting Eq. (16-13) for K oa and Eq. (16-26) for the mass terms into Eq. (16-29), we obtain
Substituting Eqs. (16-28) and (16-30) into the upper half of Eq. (16-11) (ignoring damping), we obtain
M aa + M ao G o u·· a + K aa + K ao G o u a – K ao K oo
–1 M ··
oa + M oo G o u a (16-31)
= P a – K ao K oo–1 P
o
Combining the terms, we obtain the same results as Eq. (16-24) through Eq. (16-27).
The significance of this exercise is to show that the Guyan transformation has very interesting properties,
namely:
1. The approximation occurs only on the acceleration terms Eq. (16-28).
2. The stiffness portion of the reduced system is exact.
3. The interior displacements defined by Eqs. (16-30) and (16-24) are nearly identical.
The significant aspects of the partially decoupled system described by Eqs. (16-19) through (16-31) are
that most of the MD Nastran reduction methods are easily developed from this form and the
approximations are conveniently explained in these terms. The Guyan reduction, the Dynamic reduction
process, and the modal synthesis methods are described below.
Main Index
CHAPTER 16 375
Special Topics
Guyan Reduction
In the MSC Guyan reduction process, the omitted relative accelerations, u·· oo , in Eq. (16-24) are
approximated by Eq. (16-28) and the solution system is
M a a u·· a + K a a u a = P a (16-32)
The system described by Eq. (16-28) has several desirable properties. The overall mass and center of
gravity properties are preserved in the mass matrix. Also note that the static stiffness is exact. If
accelerations occur, the errors may be estimated by solving for u oo after u a is obtained. The omitted
points, u oo , could be recovered by solving the lower part of Eq. (16-24).
However, in most applications, the stiffness terms of Eq. (16-24) dominate and the normal modes of the
omitted degrees-of-freedom are of higher frequency than the solution set. Therefore, we may ignore the
left-hand mass and solve for u o directly from Eq. (16-12) by the equation
uo = Go ua (16-33)
The errors in Eq. (16-33) are proportional to the vector u oo , which becomes small when the individual
masses and applied loads in the omit set are small. Because of the same assumption, the errors of the
upper half of Eq. (16-24) are also small. (In MD Nastran, u oo is neglected in the dynamic data recovery
process.)
u oo o q u q (16-34)
where oq are a set of approximate eigenvectors or natural shapes. In its simple form1, with u t replacing
u a , the dynamic reduction transformation corresponding to Eq. (16-29) is
ut I 0 ut
= (16-35)
uo Go o q uq
Main Index
376 MD Nastran Dynamic Analysis User’s Guide
Dynamic Reduction
··
M tt M tq u t K tt 0 u t Pt
·· + = (16-36)
Mq t Mq q uq 0 Kq q uq Pq
M t q = M oTt o q + G oT M o o o q (16-37)
M q q = ot q M o o o q (16-38)
K tt = K t t + K to G o (16-39)
K q q = oTq K o o o q (16-40)
P t = P t + G oT P o (16-41)
P q = oTq P o (16-42)
Error Analysis
Note that, as in Guyan Reduction, the exact static stiffness is retained. If accelerations exist, errors will
occur when the modal displacements do not match the exact solution for u oo . Assume that the
displacement error, o , is:
o = q u q – u oo (16-43)
and from the lower half of Eqs. (16-36) and (16-40) through (16-42), we obtain
1In these solution sequences u includes u q and the previous equations for the u a set are replaced by the
a
ut = ub + uc + ur set.
Main Index
CHAPTER 16 377
Special Topics
o
q = oTq M o o u·· o – o q u·· q (16-47)
Therefore, we observe that the loading errors for dynamic reduction will be proportional to the size of
the omitted masses, M oo , and with the quality of the approximation
Relative to Eq. (16-47), the equivalent load error for Guyan reduction was
o = M o o u·· oo (16-49)
Even when the mode shapes are only approximate, as in Dynamic reduction, we conclude that the results
(for the same u o set) will be improved from the Guyan results. However, note that when a u t set is not
used, the modes of the u oo system have the same lower frequencies as the system modes and the
approximation might not improve. In other words, the use of physical u t points in the dynamic reduction
process combines the best parts of both Guyan and Rayleight-Ritz reductions.
Note that in Guyan Reduction the relative displacements, u oo , for the omitted point are only approximated
in dynamics. In the dynamic reduction and modal synthesis methods, interior results are obtained by
solving for its eigenvector displacements and added directly into the matrix solution system.
A special case, known as the inertia relief effect, occurs when the structural component is excited at low
frequencies (below the first modal frequency). The effect is defined by Eq. (16-30) when u·· a is a low
frequency excitation, having little effect on the vibration modes, resulting in a relative quasi-static
displacement, u oo , where
–1
u oo = – K o o M o a u·· a (16-50)
The displacement vector is called the inertial relief shape when u·· a is a rigid body acceleration vector. It
may cause a significant redistribution of the internal forces in the structure.
Main Index
378 MD Nastran Dynamic Analysis User’s Guide
Dynamic Reduction
In the solution of a single structural component, the inertia relief effects may be obtained after the
solution, u a , is obtained. However, for multiple superelements connected at the u t boundary points, the
load distributions on the boundary will be affected by these internal loads. Note that the errors defined
in Eqs. (16-33) and (16-44) will be affected since the eigenvectors may only approximate the vectors
defined in Eq. (16-47).
The solution used in MD Nastran provides for inertia relief modes by adding six vectors to the
approximation set o . These vectors are obtained from Eq. (16-50) and use the equation:
–1
oir = K o o M o a D a (16-51)
where D a is a matrix of six rigid body motions defined by the grid point geometry. Each column of the
matrix oir is scaled to a reasonable value and added to the o matrix, and a corresponding u q
displacement is also included.
.
Main Index
CHAPTER 16 379
Special Topics
Superelement Analysis
Superelement analysis is a form of substructuring wherein a model is divided into pieces with each piece
represented by a reduced stiffness, mass, and damping matrix. Each superelement is processed
individually, and the solutions are then combined to solve the entire model. The final analysis (in which
all of the individual superelement solutions are combined) involves much smaller matrices than would
be required if the entire model were solved in a one-step process. An example of a superelement model
is shown in Figure 16-2.
Superelement analysis has the advantage of reducing computer resource requirements, especially if
changes are made to only one portion (superelement) of the model; in this case, only the affected
superelement needs to be reanalyzed and the final analysis repeated.
Superelement analysis presents procedural advantages as well, particularly when multiple engineering
contractors are involved in an analysis. Imagine a model of a rocket and payload: one contractor models
the booster, another contractor models the engines, and another contractor models the satellite. Each
contractor can reduce his model to its boundary degrees-of-freedom, which is suitable for superelement
analysis. The systems integrator then combines these reduced models into one model for a liftoff
analysis. Superelement analysis has the advantage that matrices can be passed from one organization to
another without revealing proprietary modeling details or concern about whether the same superelement
interior grid point and element numbers are used by every participant.
Main Index
380 MD Nastran Dynamic Analysis User’s Guide
Superelement Analysis
coordinates (corresponding to the grid points on the superelement boundaries). CMS is advantageous
because there are fewer modal coordinates than physical coordinates—perhaps only one percent as many.
In addition, CMS can utilize modal test data, thereby increasing the accuracy of the overall analysis.
Acceleration Method
If the acceleration is uniform over the structure, a time-dependent gravity load has exactly the same form
as the last term of Eq. (2-21), and the GRAV input load data may be used.
An example data input for a dynamic acceleration load is shown below.
Case Control
LOADSET = 20 $ Requests LSEQ Id. 20 Process
DLOAD = 200 $ Requests Dynamic Load #200
Bulk Data
GRAV, 386, , 386.4, -1.0 $ Defines Gravity Load in -x direction
LSEQ, 20, 201, 386 $ Assembles GRAV load vector
$ Added to DAREA Id. 201
TLOAD1, 200, 201, etc. $ Dynamic Load using DAREA Id. 201
The net result is that the time-dependent inertia loads are applied to all points on the structure in the -x
direction in proportion to the time-dependent function specified on the TLOADi or RLOADi Bulk Data
entry. The base points should be constrained and the displacements will be calculated relative to the
moving base. However, note that the accelerations output from the solution will also be relative to the
base motion and should be corrected before being compared with accelerometer data.
A simpler alternative to base motion is the large mass approach, described in Enforced Motion with Loads,
292.
Main Index
CHAPTER 16 381
Special Topics
vectors. If too few q-set DOFs are defined then modal truncation occurs and accuracy may suffer. If too
many then the dynamically reduced matrices will have null columns for the unused q-set DOFs and may
result in a performance degradation.
The user may replace all q-set related Bulk Data entries with the user parameter
PARAM,AUTOQSET,YES. The number of component modes computed is determined by the
frequency range and/or number of desired engenvectors specified on the selected EIGR or EIGRL Bulk
Data entry.
Since the generalized coordinates are automatically defined, the following entries may not be specified:
QSETi, SEQSETi, SENQSET, or PARAM,NQSET. Also, those GRID and/or SPOINT entries used to
define the q-set may be left in the Bulk Data Section but it is recommended that they be removed.
In superelement analysis, the calculation of component modes is attempted on all superelements
including the residual structure. Also, all generalized coordinates for all superelements will become
interior to the residual structure and also assigned to the q-set in the residual structure. In other words,
component modes may not be assigned interior to a superelement and they may not be removed
(constrained).
This feature is currently not supported with:
1. Multiple boundary conditions
2. Design optimization (SOL 200)
3. Aerodynamic analyses (SOLs 144, 145, 146)
4. Cyclic symmetry analyses (SOLs 114, 115, 116, 118)
5. Restarts
Example
In the following example the user defines six q-set DOFs for natural frequencies up to 1200 cycles per
unit time.
SOL 103
DIAG 8,15
CEND
TITLE= AUTOQSET DEMONSTRATION PROBLEM
SUBTITLE= TWENTY CELL BEAM
SPC=1002
METHOD=1
BEGIN BULK
EIGRL11200.
QSET1 0 101 THRU 106
SPOINT 101 THRU 106
GRID 10000 0.0 0.0 0.0 1246
= *(1) = *(5.) == $
=(19)
CBAR 101 100 10000 10001 0.0 0.0 1. 1
= *(1) = *(1) *(1) == $
=(18)
PBAR 100 1000 0.31416 0.15708
MAT1 1000 3.+7 .3 7.764-4
Main Index
382 MD Nastran Dynamic Analysis User’s Guide
Superelement Analysis
R E A L E I G E N V A L U E S
(BEFORE AUGMENTATION OF RESIDUAL VECTORS)
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 1.881936E+04 1.371837E+02 2.183346E+01 1.000000E+00 1.881936E+04
2 2 3.011058E+05 5.487311E+02 8.733327E+01 1.000000E+00 3.011058E+05
3 3 1.524259E+06 1.234609E+03 1.964941E+02 1.000000E+00 1.524259E+06
4 4 4.816616E+06 2.194679E+03 3.492940E+02 1.000000E+00 4.816616E+06
5 5 1.175494E+07 3.428547E+03 5.456702E+02 1.000000E+00 1.175494E+07
6 6 2.435711E+07 4.935292E+03 7.854762E+02 1.000000E+00 2.435711E+07
7 7 4.506449E+07 6.713009E+03 1.068409E+03 1.000000E+00 4.506449E+07
^^^
^^^ USER WARNING MESSAGE 9144 (SEMR4)
^^^ THERE ARE NO Q-SET DEGREES-OF-FREEDOM LEFT TO ACCOMMODATE ANY RESIDUAL VECTORS.
^^^ USER INFORMATION: NO RESIDUAL VECTORS WILL BE COMPUTED.
^^^ USER ACTION: SPECIFY AT LEAST 6 MORE Q-SET DEGREES-OF-FREEDOM.
^^^
^^^
^^^ USER WARNING MESSAGE 9145 ( RESLOAD )
^^^ THERE ARE NOT ENOUGH Q-SET DEGREES-OF-FREEDOM DEFINED TO ACCOMMODATE ALL OF THE COMPUTED EIGENVECTORS AND/OR RESIDUAL
VECTORS.
^^^ USER INFORMATION: THE LAST 1 MODE(S) ABOVE WILL BE TRUNCATED.
^^^ USER ACTION: SPECIFY AT LEAST 1 MORE Q-SET DEGREES-OF-FREEDOM.
R E A L E I G E N V A L U E S
(BEFORE AUGMENTATION OF RESIDUAL VECTORS)
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 1.881936E+04 1.371837E+02 2.183346E+01 1.000000E+00 1.881936E+04
2 2 3.011058E+05 5.487311E+02 8.733327E+01 1.000000E+00 3.011058E+05
3 3 1.524259E+06 1.234609E+03 1.964941E+02 1.000000E+00 1.524259E+06
4 4 4.816616E+06 2.194679E+03 3.492940E+02 1.000000E+00 4.816616E+06
5 5 1.175494E+07 3.428547E+03 5.456702E+02 1.000000E+00 1.175494E+07
6 6 2.435711E+07 4.935292E+03 7.854762E+02 1.000000E+00 2.435711E+07
Main Index
CHAPTER 16 383
Special Topics
R E A L E I G E N V A L U E S
(BEFORE AUGMENTATION OF RESIDUAL VECTORS)
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 1.881936E+04 1.371837E+02 2.183346E+01 1.000000E+00 1.881936E+04
2 2 3.011058E+05 5.487311E+02 8.733327E+01 1.000000E+00 3.011058E+05
3 3 1.524259E+06 1.234609E+03 1.964941E+02 1.000000E+00 1.524259E+06
4 4 4.816616E+06 2.194679E+03 3.492940E+02 1.000000E+00 4.816616E+06
5 5 1.175494E+07 3.428547E+03 5.456702E+02 1.000000E+00 1.175494E+07
6 6 2.435711E+07 4.935292E+03 7.854762E+02 1.000000E+00 2.435711E+07
7 7 4.506449E+07 6.713009E+03 1.068409E+03 1.000000E+00 4.506449E+07
R E A L E I G E N V A L U E S
(AFTER AUGMENTATION OF RESIDUAL VECTORS)
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 1.881936E+04 1.371837E+02 2.183346E+01 1.000000E+00 1.881936E+04
2 2 3.011059E+05 5.487311E+02 8.733327E+01 1.000000E+00 3.011059E+05
3 3 1.524259E+06 1.234609E+03 1.964941E+02 1.000000E+00 1.524259E+06
4 4 4.816615E+06 2.194679E+03 3.492940E+02 1.000000E+00 4.816615E+06
5 5 1.175493E+07 3.428547E+03 5.456702E+02 1.000000E+00 1.175493E+07
6 6 2.435711E+07 4.935292E+03 7.854761E+02 1.000000E+00 2.435711E+07
7 7 4.506449E+07 6.713009E+03 1.068409E+03 1.000000E+00 4.506449E+07
8 8 9.003539E+07 9.488698E+03 1.510173E+03 1.000000E+00 9.003539E+07
9 9 1.442988E+08 1.201244E+04 1.911840E+03 1.000000E+00 1.442988E+08
Main Index
384 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis
Main Index
CHAPTER 16 385
Special Topics
Geometric Nonlinearity
Geometrically nonlinear problems involve large displacements; “large” means that the displacements
invalidate the small displacement assumptions inherent in the equations of linear analysis. For example,
consider a thin plate subject to an out-of-plane load. If the deflection of the plate’s midplane is
approximately equal to the thickness of the plate, then the displacement is considered large, and a linear
analysis is not applicable.
Another aspect of geometric nonlinear analysis involves follower forces. Consider a clamped plate
subject to a pressure load. As shown in Figure 16-3, the load has followed the plate to its deformed
position. Capturing this behavior requires the iterative update techniques of nonlinear analysis.
Material Nonlinearity
Material nonlinear analysis can be used to analyze problems where the stress-strain relationship of the
material is nonlinear. In addition, moderately large strain values can be analyzed. Examples of material
nonlinearities include metal plasticity, materials (such as soils and concrete), and rubbery materials
(where the stress-strain relationship is nonlinear elastic). Various yield criteria, such as von Mises or
Tresca (for metals) and Mohr-Coulomb or Drucker-Prager (for frictional materials, such as soils or
concrete), can be selected. Three hardening rules are available in MD Nastran: isotropic hardening,
kinematic hardening, or combined isotropic and kinematic hardening. With such generality, most plastic
material behavior with or without the Bauschinger effect can be modeled. In addition, gaps can be used
to model the effects due to structural separation.
Contact
Contact occurs when two or more structures (or portions of structures) collide. Contact can be modeled
as point-to-point contact (CGAP) or as contact along a line (BLSEG).
Main Index
386 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis
piping systems supported by nonlinear springs, chains (stiffness in tension only), or constant force
devices; base-mounted equipment and structure supported by nonlinear shock isolation systems;
structures with gaps (piping systems and buildings); and soil or concrete structures (which only exhibit
stiffness when in compression). Other systems exhibit nonlinearities that are proportional to the velocity,
such as seat belts and piping supports.
Force
Force
Displacement Displacement
Force
Displacement Displacement
Force
Displacement Displacement
There are several methods in MD Nastran for solving nonlinear elastic problems. A general nonlinear
solution scheme can be used, in which the material properties are specified as nonlinear elastic
(NLELAST on the MATS1 entry). Nonlinear element stiffness matrices are generated, and equilibrium
Main Index
CHAPTER 16 387
Special Topics
iterations and convergence checks are required at each load step. The CBUSH element adds some
capabilities in this area, also. In addition, gap elements (CGAP) can be used in conjunction with elastic
elements to produce systems with piece-wise linear force-deflection curves. This process also requires
the additional computations of nonlinear element stiffness matrix generation, equilibrium iteration, and
convergence testing. Nonlinear analyses requiring these additional computations can be substantially
more costly than an equivalent linear analysis even if the nonlinearities are relatively few in number.
An efficient technique called the pseudoforce method exists in MD Nastran, in which the localized
nonlinearities are treated so that they require no additional computer time when compared to a purely
linear analysis. In this method, which is available only for transient response solutions (SOLs 109 and
112, for example), deviations from linearity are treated as additional applied loads. The dynamic
equations of motion are written as
M u·· t + B u· t + K u t = P t + N t (16-52)
where M , B , and K denote the system mass, damping, and stiffness matrices, respectively. The
vectors P t and u t denote applied nodal loads and system displacements, respectively, as
functions of time. The vector N t denotes the nonlinear forces, which are added to the right-hand
side of Eq. (16-52) (and hence are treated as additional applied loads).
The nonlinear forces are evaluated at the end of one time step for use in the successive time step. The
equations of motion therefore become the following:
M u·· t + C u· t + K u t = P t + N t – t (16-53)
Note that the nonlinear force lags the true solution by a time step, which may require using small
integration time steps (i.e., smaller than those required for a purely linear analysis). Equation (16-53) can
be solved in physical or modal coordinates (the nonlinearity itself must be expressed in terms of physical
coordinates).
A nonlinear force can be used in conjunction with a linear elastic element to produce the desired
force-deflection curve as illustrated in Figure 16-5. The nonlinear dynamic force is formulated using a
NOLINi entry and a TABLEDi entry, which contains a force-versus-deflection table describing the
nonlinear force. For desired force-deflection curves more complicated than the bilinear stiffness shown
in the figure, the nonlinear force is made correspondingly more complex.
Main Index
388 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis
Force
Force
Force
= +
Displacement Displacement Displacement
Main Index
CHAPTER 16 389
Special Topics
The variables, u , may be displacement or velocity components of grid, scalar, or extra points in the
solution set. In the DIRECT TRANSIENT solutions, the connected degrees-of-freedom, u i and u j , etc.,
must remain in the solution set, u d . In the modal transient solutions, only extra points are available for
use by the NOLINi entries. In nonlinear transient solutions all degrees-of-freedom are available, but
unfortunately, extra points are not supported.
Note that these functions mimic the basic nonlinear function generators used in passive analog
computers (MSC.Software’s original business). They may be added together and combined with other
functions such as MPCs and scalar elements to handle a variety of problems.
where N are the nonlinear forces which are dependent on variable displacements, u , and
velocities, v , of the unknowns. Note that if these functions were used to replace a stiffness term, the
proper definition of N would be N = – Ku .
There are several options in MD Nastran for transient integration. For the basic purposes of this
introduction, we will discuss the simplest form, the three-point method which performs a step-by-step
calculation. At time step t n , the solution is u n and the step size is h . Dropping the brackets, the
averaged values of displacement, u , velocity, v , acceleration, a , and load, P , in terms of the three
discrete steps are as follows:
u = u n + 1 + 1 – 2 u n + u n – 1 (16-55)
v = u n + 1 – u n – 1 2h (16-56)
2
a = u n + 1 – 2u n + u n + 1 h (16-57)
P = P n + 1 + 1 – 2 P n + P n (16-58)
In the solution, the vectors at step n + 1 are obtained by substituting Eq. (2-7) through Eq. (2-10) in
Eq. (2-40). This method is actually a variation of the Newmark-Beta method and is guaranteed to be
Main Index
390 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis
stable if 0.25 . For consistency and stability, it would be desirable to have the N vector also in this
form. Ideally, the nonlinear loads would be consistent with the the linear displacements and loads, i.e.,
N = N n + 1 + 1 – 2 N n + N n – 1 (16-59)
N n + 1 2N n – N n – 1 (16-60)
Substituting this approximation into Eq. (16-59) we obtain the term used in MD Nastran:
N Nn (16-61)
In other words, calculating the nonlinear function at the center step is a valid approximation to the value
averaged over three steps. Unfortunately, when u is rapidly changing, N will lag behind and create
unstable oscillations. The results of an unstable integration usually grow with a period of two or three
time steps per cycle. An alternate option for averaging nonlinear loads is available with DIAG 10,
although it is not recommended.
N u M
For small motions, if we define the function as an equivalent spring with N = ku , the integration
equation for u n + 1 is
2
M u n + 1 – 2u n + u n – 1 h = N = – ku n (16-62)
For stability analysis, we will assume a uniform growth rate, r , where u n = ru n – 1 and u n + 1 = ru n .
Substituting into Eq. (16-62) we obtain a quadratic equation for r :
2
kh
r 2 + 2 -------- – 1 r + 1 = 0 (16-63)
2M
If the magnitude of r is greater than 1, the solution will potentially diverge ( r may be complex). It is
easy to show that if kh 2 4M the system will have a real root r – 1 , which is an unstable oscillation that
changes sign every step.
Main Index
CHAPTER 16 391
Special Topics
In other words, the stability limit for the nonlinear stiffness for this case is
N 4M
– ------ = k -------- (16-64)
u h
2
The cure for divergence is to either reduce the time step, h , add linear stiffness to replace some of the
nonlinear portion, or add enough mass, M , in parallel to the nonlinear function to satisfy the criteria for
every possible value of k . Scalar mass elements CMASSi may be added to couple two scalar points, in
parallel with the nonlinear spring, without affecting the total mass of the structure.
N v = – B u,v u n + 1 – u n – 1 2h (16-65)
However, the current displacements u n + 1 have not been calculated at this stage, so we again use the
assumption of Eq. (16-60) and obtain the approximation:
N v – B u,v u n – u n – 1 h N u n – u n – 1 h (16-66)
The stability of velocity-dependent NOLINi systems is very marginal for most applications. Replacing
the nonlinear spring ( k ) in the example problem above with a nonlinear damper, b , we obtain the
following system equation:
2
M u n + 1 – 2u n + u n – 1 h = N = –b un – un – 1 h (16-67)
For stability analysis we may substitute ru n and r 2 u n as in the stiffness example above, to obtain a
quadratic equation:
r + bh
------ – 2 r + 1 – bh
------ = 0
2
(16-68)
M M
b 2M h (16-69)
Again, the problem will show up as a spurious diverging oscillation. Divergence may be cured by
decreasing the time step size, h , by changing part of the nonlinear function, b , into a linear damper, or
by adding mass in parallel with the nonlinear function. SOL 129 is better suited than SOL 109 for
unstable or badly conditioned problems.
Main Index
392 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis
Main Index
CHAPTER 16 393
Special Topics
4. The forces, N , must have the same sign and direction as an externally applied load on the
structure. Note that this will have the opposite sign as the forces applied to the GRID points by
the finite elements. (The N force for a spring element will have a negative value for a positive
displacement.) This frequently causes some difficulty in debugging the input data.
5. Use MPC or TF data to simplify the NOLINi inputs. A single nonlinear spring in three
dimensions connecting two points could require up to 6 6 = 36 nonlinear functions.
Alternatively, one MPC equation could be used to define the strain as an added scalar point, and
one nonlinear function would define the generalized force (force times length) on the scalar point.
Because the MPC constraints are consistent, the force will automatically be distributed to all six
degrees-of-freedom.
6. Plan ahead to retain the nonlinear points in the solution set. Remember that in SOLs 109 and 112
the NOLINi functions may not reference any dependent or eliminated degree-of-freedom. Note
that in a modal analysis, this leaves only extra points.
7. Use the full nonlinear solution (SOL 129) for small to medium-sized problems.
u r1
u r2
t
101
rotating
u f1
201 fixed
Figure 16-6
Main Index
394 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis
A simple MPC equation may be used to connect the axial coordinates u 3 . The rotations may have
similar definitions but are not shown. A similar equation defines the forces on the fixed point N f in
terms of the rotating forces Q r .
For the actual connection, we will use the Lagrange Multiplier technique (see Lagrange Multiplier
Technique, 296 for details) by defining another GRID Q r which defines the constraint forces on point
101. The resulting matrix partition equation for the three points is
Mf f 0 u··
0 K ff 0 0 Uf Pf + Nf
··
u
0 m rr 0 + 0 K rr I ur = Pr (16-72)
··
0 0 0 rQ 0 I 0 Qr Nq
where:
Nf =
P r1 cos t – P r2 sin t
P r1 sin t + P r 2 cos t
NQ =
u f 1 cos t – u r2 sin t
– u f1 sin t + u r2 cos t
In order to generate the sine and cosine terms, we will add two scalar points and connect unit masses and
springs k = 2 . Using initial conditions, these are the sine and cosine functions defined as scalar
displacements. The following input data will solve the connection:
In Case Control
IC = 100 $ For sine and cosine
K2PP = RCOUP $ For matrix terms
NONLINEAR=10 $ For NOLIN2
In Bulk Data
$ 50 radian/sec oscillators ( Note: Don’t use PARAM,G)
SPOINT,2001,2002 $ sine and cosine
CMASS4,2001,1.0,2001
CMASS4,2002,1.0,2002
CELAS4,2003,2500.0,2001
CELAS4,2004,2500.0,2002
TIC,100,2001,,,50.0 $ sine
TIC,100,2002,,1.0 $ cosine
Main Index
CHAPTER 16 395
Special Topics
DMIG,RCOUP,101,1,,3001,1,1.0
DMIG,RCOUP,101,2,,3001,2,1.0
$ Note that other half is generated when using the SYM option
Main Index
396 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis Elements
Nonlinear elements may be combined with linear elements for computational efficiency if the nonlinear
effects can be localized. Primary operations for nonlinear elements are the updating of element
coordinates and applied loads for large displacements and the internal relaxation iteration for material
nonlinearity. Refer to Table 16-4 for a summary of the nonlinear element properties.
Main Index
CHAPTER 16 397
Special Topics
Case Control
Each subcase defines a time interval starting from the last time step of the previous subcase, subdivided
into smaller time steps. The output time is labeled by the cumulative time, including all previous
subcases. There are advantages in dividing the total duration of analysis into many subcases so that each
subcase does not have excessive number of time steps. The data blocks containing solutions are
generated at the end of each subcase for storage in the database for output processing and restarts. As
such, converged solutions are apt to be saved at many intermediate steps in case of divergence and more
flexible control becomes possible with multiple subcases.
The input loading functions may be changed for each subcase or continued by repeating the same
DLOAD request. However, it is recommended that one use the same TLOAD Bulk Data for all the
subcases in order to maintain the continuity between subcases, because TLOADi data defines the loading
Main Index
398 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis Elements
history as a function of cumulative time. Static loads (PLOADi, FORCEi, MOMENTi) may be
associated with time-dependent functions by matching the EXCITEID on the TLOADi entries.
However, no thermal loads or enforced displacements (SPCD) are allowed in the nonlinear transient
analysis. Nonlinear forces as functions of displacements or velocities (NOLINi) may be selected and
printed by the Case Control commands NONLINEAR and NLLOAD, respectively. Each subcase may
have a different time step size, time interval, and iteration control selected by the TSTEPNL request. The
Case Control requests that may not be changed after the first subcase are SPC, MPC, DMIG, and TF.
Output requests for each subcase are processed independently. Requested output quantities for all the
subcases are appended after the computational process for actual output operation. Available outputs are
DISPLACEMENT, VELOCITY, ACCELERATION, OLOAD, STRESS, FORCE,
SDISPLACEMENT, SVELOCITY, SACCELERATION, NONLINEAR (NLLOAD), and SPCFORCE.
However, element force output and GRID point stresses are not available for nonlinear elements.
Initial conditions (displacement or velocity) can be specified by the Bulk Data input, TIC, selectable by
the Case Control command IC. If initial conditions are given, all of the nonlinear element forces and
stresses must be computed to satisfy equilibrium with the prescribed initial displacements. On the other
hand, initial conditions can be generated by applying static analysis for the preload using
PARAM,TSTATIC in the first subcase. Then the transient analysis can be performed in the ensuing
subcases. Associated with the adaptive time stepping method, the PARAM,NDAMP is used to control
the stability in the ADAPT method. The parameter NDAMP represents the numerical damping (a
recommended value for usual cases is 0.01), which is often required to improve the stability and
convergence in contact problems.
All the superelement model generation options and matrix reduction options are allowed for the linear
portion of the structure. The generalized dynamic reduction, component mode synthesis, and Guyan
reduction may be performed for upstream superelements. The residual superelement may contain scalar
degrees-of-freedom representing linear modal formulations.
Main Index
CHAPTER 16 399
Special Topics
duration. It is advised that a drastic change in the time step size between subcases be avoided. A drastic
change—e.g., ratio of the two time steps exceeding 1000—could cause a loss of accuracy upon subcase
switch. In this case, an intermediate subcase should be provided for a transition period of short interval
to reduce the ratio.
The parameters for controlling the accuracy and stability of the incremental and iterative process can be
specified in the TSTEPNL Bulk Data entry. The controls are applicable to the automatic time step
adjustment and bisection process in addition to stiffness matrix updates, BFGS updates, and line searches
similar to those on the NLPARM Bulk Data entry. Since default values have been selected from
numerous test runs, the analysis should be started with the default setting and changed if necessary. The
TSTEPNL data format is shown below with default values:
Format
1 2 3 4 5 6 7 8 9 10
TSTEPNL ID NDT DT NO METHOD KSTEP MAXITER CONV +TNL1
+TNL1 EPSU EPSP EPSW MAXDIV MAXQN MAXLS FSTRESS +TNL2
+TNL2 MAXBIS ADJUST MSTEP RB MAXR UTOL RTOLB
1 2 3 4 5 6 7 8 9 10
TSTEPNL x x x 1 ADAPT 2 10 PW +TNL1
+TNL1 1.E-2 1.E-3 1.E-6 2 10 2 0.2 +TNL2
+TNL2 5 5 0 0.75 16.0 0.1 20.
The TSTEPNL Bulk Data entry is selected using ID via the Case Control command TSTEPNL. Each
subcase (residual superelement solutions only) requires a TSTEPNL entry. Multiple subcases are
assumed to occur sequentially in time. Therefore, the initial conditions of each subcase are defined by
the end conditions of the previous subcase.
The NDT field specifies the number of time steps with DT as the size of each time step. The total
duration for the subcase can be assessed by multiplying NDT and DT (i.e., NDT*DT). The time
increment ( t ) remains constant during the analysis in AUTO and TSTEP options, and is equal to DT.
However, the time increment ( t ) changes during the analysis in the ADAPT option and the actual
number of time steps will not be equal to NDT. In the ADAPT option, DT is used as an initial value for
t .
The NO field specifies the time step interval for output; i.e., every NO-th step solution is saved for
output. The data will be output at steps 0, NO, 2NO, ..., etc., and the last converged step for printing and
plotting purposes. The Case Control command OTIME may also be used to control the output points.
The METHOD field selects an option for direct time integration and the stiffness matrix update strategies
among ADAPT, AUTO and TSTEP. If the AUTO option is selected, MD Nastran automatically updates
the stiffness matrix to improve convergence while the KSTEP value is ignored. If the TSTEP option is
selected, MD Nastran updates the stiffness matrix every KSTEP-th increment of time. If the ADAPT
Main Index
400 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis Elements
option is selected, MD Nastran automatically adjusts the incremental time and uses the bisection
algorithm in case of divergence. During the bisection process in the ADAPT option, stiffness is updated
at every KSTEP-th successful bisection. The ADAPT method allows linear transient analysis, but AUTO
or TSTEP will abort the run if the model does not have any data representing nonlinearity. The stiffness
matrix is always updated for a new subcase or restart, irrespective of the option selected.
The number of iterations for a time step is limited to MAXITER. If the solution does not converge in
MAXITER iterations, the process is treated as a divergent process; i.e., either a bisection or stiffness
matrix update takes place based on the value of MAXBIS. The sign of MAXITER provides a control
over reiteration in case of failure in convergence or bisection. If MAXITER is negative, the analysis is
terminated when the divergence condition is encountered twice during the same time step or the solution
diverges for five consecutive time steps. If MAXITER is positive, MD Nastran computes the best
attainable solution and continues the analysis.
The convergence test is controlled by convergence test flags (U for displacement error test, P for load
equilibrium error test, W for work error test) and the error tolerances (EPSU, EPSP and EPSW) which
define the convergence criteria. All requested criteria (a combination of U, P, and/or W) are satisfied
upon convergence. It should be noted that at least two iterations are necessary to check the displacement
convergence criterion.
The MAXDIV field provides control over diverging solutions. Depending on the rate of divergence, the
number of diverging solutions (NDIV) is incremented by 1 or 2. The solution is assumed to be divergent
when NDIV reaches MAXDIV during the iteration. If the bisection option is used with the ADAPT
method, the time step is bisected upon divergence. Otherwise, the solution for the time step is repeated
with a new stiffness based on the converged state at the beginning of the time step. If NDIV reaches
MAXDIV twice within the same time step, the analysis is terminated with a fatal message.
In transient analysis, the BFGS quasi-Newton updates and the line search process work in the same way
as in static analysis (except for the default settings). The MAXQN field defines the maximum number
of quasi-Newton vectors to be saved on the database and the MAXLS defines the number of line searches
allowed per iteration. Nonzero values of MAXQN and MAXLS activate the quasi-Newton update and
the line search process, respectively.
The FSTRESS field defines a fraction of the effective stress ( ) which is used to limit the subincrement
size in the material routine. The number of subincrements in the material routines is determined such
that the subincrement size is approximately FSTRESS (equivalent stress). FSTRESS is also used
to establish a tolerance for error correction in the elasto-plastic material, i.e.,
Main Index
CHAPTER 16 401
Special Topics
continued to the next time step. If MAXBIS is negative and the solution does not converge in MAXBIS
bisections, the analysis is terminated.
The parameter ADJUST allows the user to control the automatic time stepping in the ADAPT option. A
value of zero for ADJUST turns off the automatic adjustment completely. If ADJUST is positive, the
time increment is continually adjusted for the first few steps until a good value of t is obtained. After
this initial adjustment, the time increment is adjusted every ADJUST time steps only. A value of
ADJUST an order greater than NDT will turn off adjustments after the initial adjustment. Since the
automatic time step adjustment is based on the mode of response and not on the loading pattern, it may
be necessary to limit the adjustable step size when the period of the forcing function is much shorter than
the period of dominant response frequency of the structure. It is the user’s responsibility to ensure that
the loading history is properly traced with the ADJUST option. The ADJUST option should be
suppressed for the duration of a short pulse loading. If unsure, the user should start with a value of DT
that is much smaller than the pulse duration in order to properly represent the loading pattern.
MSTEP defines the desired number of time steps to obtain the dominant period response accurately
( 10 Integer 200 ). RB defines bounds for maintaining the same time step for the stepping function
in the automatic time step adjustment method ( 0.1 Real 1.0 ). Parameters MSTEP and RB are used
to adjust the time increment during the analysis in the ADAPT option. The adjustment is based on the
number of time steps desired to capture the dominant frequency response accurately. The time increment
is adjusted as
t n + 1 = f r t n (16-73)
where:
r =
1 2 1
-------------------- ------ --------
MSTEP n t n
with:
f = 1.0 for RB r 2
f = 2.0 for 2. r 3. RB
f = 4.0 for r 3. RB
The recommended value of MSTEP for nearly linear problems is 20. A larger value (e.g., 40) is required
for highly nonlinear problems. In the default options, MD Nastran automatically computes the value of
MSTEP based on the changes in the stiffness.
Main Index
402 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis Elements
The MAXR field defines the maximum ratio for the adjusted incremental time relative to DT allowed for
time step adjustment ( 1.0 Real 32.0 ). MAXR is used to define the upper and lower bounds for
adjusted time step size, i.e.,
DT DT
MIN ---------------------- , ----------------- t MAXR DT (16-74)
2 M A X BI S MAXR
The UTOL field defines the tolerance on displacement increment below which there is no time step
adjustment ( 0.001 Real 1.0 ). UTOL is used to filter undesirable time step adjustment; i.e., no time
step adjustment is performed if
U· n
- UTOL
------------------ (16-75)
U· m a x
The RTOLB field defines the maximum value of incremental rotation (in degrees) allowed per iteration
to activate bisection ( Real 2.0 ). The bisection is activated if the incremental rotation for any
degree-of-freedom ( x , y , or z ) exceeds the value specified for RTOLB. This bisection strategy
based on the incremental rotation is controlled by the MAXBIS field.
Main Index
CHAPTER 16 403
Special Topics
ITR DIV Number of occurrences of divergence detected during the adaptive iteration by
the iteration module NLTRD2
MAT DIV Number of occurrences of bisection conditions in the material routine (excessive
stress increment) or in the rotation angle (excessive rotation) during the iteration
using the ADAPT method
NO. BIS Number of bisections executed for the current time interval
ADJUST Ratio of time step adjustment relative to DT within a subcase
Basic Equations
We may calculate the load equilibrium error vector, R n at time step n by the equation
where:
The above equation is solved at the reduced ( u d ) displacement vector size. The approximation errors
due to dynamic reduction methods are not included in the error vector R n . Applying Newmark’s
method over a finite time period, t n – 1 t t n + 1 , the average static forces are
F = F u n + 1 + 1 – 2 F u n + F u n – 1 (16-77)
where is the Newmark Beta operator and F u n is the nonlinear force due to a generalized
displacement vector u n . An identical definition occurs for P from the applied loads at each time
step.
In summary, at each time step MD Nastran will iterate the displacements and forces until R ni passes
the convergence tests or the number of passes reaches an iteration limit. With a single step, i = i,
Main Index
404 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis Elements
calculating only R 0 , the results will be identical to the existing NOLINi results in MD Nastran. For faster
convergence the iterations may continue, the matrices may be updated, and/or the time step size may be
reduced. For more details on this algorithm, refer to Chapter 9 of the MD Nastran Nonlinear Handbook.
Method ADAPT
The multistep implicit integration method has difficulties when changing time step size. Therefore, to
allow self-adapting time step sizes, the two-point integration method is introduced with module
NLTRD2. This method can be selected by specifying the ADAPT method in the TSTEPNL Bulk Data
entry.
Newmark Integration
For the adaptive scheme, Newmark’s method is employed with the two-point recurrence formula for
one-step integration, i.e.,
1
U n + 1 = U n + t U· n + --- t 2 U·· n + t 2 U·· n + 1 – U·· n (16-78)
2
and
where U , U· , U·· and t denote displacement, velocity, acceleration and the time step
increment, respectively. The subscript n designates the time step and the parameters ( and ) are to
be selected for the best solution. An equilibrium equation to be satisfied at time step ( n + 1 ) is
M U·· n + 1 + C U· n + 1 + F U n + 1 = P n + 1 (16-80)
where M and C denote mass and damping matrices, and F and P n + 1 denote internal and external
forces, respectively.
An alternative expression for the load vector can be derived for = 0.5 by introducing
M U·· n + C U· n = P n – F n (16-81)
By virtue of Eq. (16-81) the residual load error at each time step is effectively carried over to the next
step and the error propagation is reduced, while the computation is significantly simplified.
4
R ni + 1 = P n + 1 – F ni + 1 + ----- M U· n + P n – F n (16-82)
t
4 2
– -------- M + ----- C U ni + 1 – U n
t 2 t
Main Index
CHAPTER 16 405
Special Topics
The iteration method calculates new values of R ni + 1 until it is sufficiently small. The results then become
the starting point for the next time step.
Main Index
406 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis Elements
reiteration procedure will end at the i-th iteration and the normal time stepping procedure will be
resumed.
Main Index
CHAPTER 16 407
Special Topics
Restarts
Since SOL 106 and SOL 129 share the same database storage formats for nonlinear tables and matrices,
the restart system for transient analysis can use either a previous static or transient nonlinear analysis as
its initial conditions.
where NDT i and DT i are the number of time steps and the time increment of the i-th subcase,
respectively.
If a SOL 129 run is terminated abnormally in the middle of a subcase, it may or may not be restartable
depending upon the cause of the abnormal exit. If the job is stopped due to a diverging solution, it can
be restarted either from the end of a previous subcase or from the last saved solution step. The restart
procedure for the former is identical to that for the normal restart as described in the preceding paragraph.
Main Index
408 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Transient Response Analysis Elements
The latter case also requires parameters LOOPID and STIME; however, the input value for STIME
differs depending on the value of METHOD specified on the TSTEPNL entry.
If METHOD = AUTO or TSTEP, STIME is the time corresponding to the last output step which may be
calculated based on the output skip factor (i.e., the NO on the TSTEPNL entry). If METHOD = ADAPT,
the last converged solution is always treated as an output step and is always saved for the restart so that
STIME can be the time of the last converged step. The values of STIME and LOOPID can also be found
in the printout, if the ADAPT method is used.
Once STIME and LOOPID are known, determine the number of remaining time steps in the subcase and
create a new TSTEPNL entry for the remaining time. Insert a new subcase that references the new
TSTEPNL entry prior to the remaining subcases in the Case Control Section.
A solution may be terminated in the middle of a subcase due to insufficient CPU time: (1) the CPU time
specified in the Executive Control Section is insufficient so the run is forced to exit by MD Nastran, or
(2) the CPU time exceeds the limit specified in the computer system which leads to a sudden job abortion
by the system. In the first case, MD Nastran is able to detect the specified CPU time in the Executive
Control Section and automatically activate the exit procedure before time expiration. When completed,
the solution can be restarted from the termination point as in the solution diverging case. In the second
case, the solution can only be restarted from the end of a subcase.
Restarts may also be performed solely for data recovery by providing the following parameters:
SDATA = -1 : Recover data without running the solution module
LOOPID = N : from the 1st through the N-th subcases
Note that solution sets DISP, VELO, ACCE, OLOAD, SPCF (printout and plotting) and NLLOAD
(plotting only) are recoverable while STRESS, SDISP, SVELO and SACCE sets are not.
Main Index
CHAPTER 16 409
Special Topics
Main Index
410 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Harmonic Response
Introduction
There is a class of dynamic response analyses where a structure exhibiting nonlinearities is subjected to
a harmonic excitation in which the response is essentially periodic. That is to say that the degree of
nonlinearity in the system is light enough that the response may be described as sufficiently accurate by
a combination of harmonic responses, i.e. periodic; this is achieved using a Fourier series.
Some examples of this class of dynamic response problems are rotor/stator contact under abnormal
running conditions, or an overload condition in an oscillating mechanism causing periodic contact. The
response of such systems may exhibit multiple solutions in a steady-state vibration response scenario,
possibly with amplitude jumps as the system moves from one frequency to another, such as might occur
in a rotor that is increasing or decreasing in speed. These jumps reveal different behavior of the dynamic
system with increasing or decreasing excitation frequency.
The nonlinear harmonic response solution sequence uses the harmonic balance method to calculate the
periodic response of a non-linear system under harmonic excitation. This requires the definition of a
frequency domain problem in the presence of nonlinearities. The harmonic balance method assumes the
steady-state response consists of a sum of sinusoidal responses finding the coefficients of the sinusoids
to satisfy the equations of motion. Harmonic balance is only efficient if a small number of sinusoids are
necessary to approximate the solution to a desired accuracy. This is why the nonlinearities in the system
must be mild.
As with any nonlinearity, it must only be present in the residual structure. However, this does not
preclude the use of superelements or ASET degrees of freedom to perform dynamic reduction using
CMS. In fact, this is a recommended technique in order to keep the number of degrees-of-freedom for
harmonic balance to a minimum.
Nonlinear harmonic response is available in the presence or absence of rotors, but there must be at least
one degree-of-freedom defined on nonlinear force type entries such as the NLRGAP, NLRSFD, and
NOLINi entries. Other types of nonlinearity that may be defined include the CBUSH2D element or
indeed any elements having frequency dependent properties. With MD Nastran’s user-defined service,
the NLRSFD entry is able to call an external user-defined behavior to replace the standard NLRSFD
behavior.
Input
The FMS Section
The FMS statement, related to the SCA UDS, is required if a user-defined service is to be used. When a
user-defined service (UDS) is to be utilized to describe the behavior of the NLRSFD type nonlinearity in
nonlinear harmonic response, the CONNECT (p. 48) in the MD Nastran Quick Reference Guide entry is
used with the SERVICE qualifier to make the connection between the GROUP name on the NLRSFD
Bulk Data entry and the name of the external service.
Main Index
CHAPTER 16 411
Special Topics
Executive Control
Nonlinear harmonic response is available as SOL 128 or SOL SENLHARM. This solution sequence is
based around linear direct frequency response (SOL 108), but in which nonlinearities may be taken into
account.
Case Control
The Case Control command NLHARM (p. 413) in the MD Nastran Quick Reference Guide for nonlinear
harmonic response references the NLHARM Bulk Data entry. The NONLINEAR (p. 424) in the MD
Nastran Quick Reference Guide Case Control command may also be used in nonlinear harmonic
response to reference nonlinear force Bulk Data entries (NOLINi, NLRGAP, NLRSFD,…).
Bulk Data
The three Bulk Data entries (NLHARM, NLFREQ, and NLFREQ1) relate specifically to a nonlinear
harmonic response, and so does the table Bulk Data entry (TABLED5). Bulk Data entries NLRGAP,
NLRSFD and NOLIN1 (p. 2624) in the MD Nastran Quick Reference Guide are also used in a nonlinear
harmonic response.
NLHARM
The NLHARM Bulk Data entry is used to define the parameters for nonlinear harmonic response.
1 2 3 4 5 6 7 8 9 10
NLHARM ID SUBFAC NHARM NLFREQ
Field Contents
ID Identification number referenced by the NLHARM Case Control command. (Integer
> 0)
SUBFAC Factor for capturing sub-harmonic response. See Remark 3. (Integer > 1, Default = 1)
NHARM The number of harmonics to include in the solution. See Remark 2. (Integer > 0)
NLFREQ Identification number of the NLFREQ or NLFREQ1 entry specifying the forcing
frequency list. (Integer > 0)
NLFREQ
The NLFREQ entry is used to define the forcing frequencies for a nonlinear harmonic response.
1 2 3 4 5 6 7 8 9 10
NLFREQ ID F1 F2 F3 ... Fn
Main Index
412 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Harmonic Response
Field Contents
ID Identification number referenced by the NLFREQ field (field 5) of an NLHARM
Bulk Data entry. (Integer > 0)
F1...Fn Forcing frequency values in cycles per unit time. (Real > 0.0)
NLFREQ1
The NLFREQ entry is used to define the forcing frequencies for a nonlinear harmonic response by using
a start frequency, a frequency interval and a number of intervals.
1 2 3 4 5 6 7 8 9 10
NLFREQ1 ID F1 DF NDF
Field Contents
ID Identification number referenced by the NLFREQ field (field 5) of an NLHARM
Bulk Data entry. (Integer > 0).
F1 First forcing frequency in the set. (Real > 0.0)
DF Frequency increment. See Remark 1. (Real < > 0.0; Required)
NDF Number of frequency increments/decrements. (Integer > 0, Default = 1)
TABLED5
The TABLED5 entry is used generally to define a value as a function of two variables for use in generating
frequency-dependent and time-dependent dynamic loads. In nonlinear harmonic response it is used to
define an NLRGAP whose force-penetration characteristics vary with frequency.
1 2 3 4 5 6 7 8 9 10
TABLED5 TID
X(1) TID(1) X(2) TID(2) X(3) TID(3) X(4) TID(4)
... ... ENDT
Field Contents
TID Table identification number. (Integer > 0)
X(i) X value for the function specified by TID(i) (Real; no Default).
TID(i) ID of a TABLED1, TABLED2, TABLED3 or TABLED4 defining the function Y for
the given value of X. (Integer > 0; no Default).
Main Index
CHAPTER 16 413
Special Topics
Main Index
414 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Harmonic Response
PARAM, MXICODE0, 5
In nonlinear harmonic response analysis, if the solution fails to converge more than MXICODE0 times
in succession, a new trial displacement vector is calculated. MXICODE0 allows the number of
successive failed convergences to be modified before a new trial displacement vector is calculated.
(Default is 5).
PARAM, NHPLUS, 20
In nonlinear harmonic response analysis, in order to avoid aliasing in the calculation of the Fourier
coefficients, a certain number of extra evaluation points are used. NHPLUS allows the number of extra
points to be defined.
A literature search suggests this technique comes from reference [6] where the number of time steps, S,
selected should be in the range (2K-1) S 3(2K-1) and K is the number of frequencies present in the signal.
Main Index
CHAPTER 16 415
Special Topics
In most cases of nonconvergence that is not attributed to dynamically unstable conditions, either the
number of harmonics or the subharmonic content is insufficient, or the system has reached a bifurcation
or turning point. Try increasing the number of harmonics or subharmonics (NLHARM entry) as well as
adjusting the parameters described above. If none of these are successful, the system may have struck a
bifurcation point. Try adjusting the excitation frequencies slightly. For example, if using the following
excitation frequencies: 5.0, 10.0, 15.0, 20.0…etc., try adjusting these to 5.1, 10.1, 15.1, 20.1. In the
current implementation, there is no automatic treatment for possible bifurcation points, and the subject
of using a continuation strategy is under discussion for a future development.
In the case of nonconvergence, the response quantities are set to zero, and the calculation continues to
the next excitation frequency in the sequence retaining the initial conditions of the solution from the last
converged frequency. If the solution at the next excitation frequency does not converge, the same
procedure is followed until a converged solution is found.
There may come a time when it is judicious to change the initial conditions, particularly when the next
excitation frequency becomes distanced from the last converged frequency after a sequence of failed
attempts to converge excitation frequencies. In this situation, where the initial conditions for an
excitation frequency have become somewhat distanced from the previously converged excitation
frequency, it may not mean very much physically to continue to use the initial conditions from that
previously converged excitation frequency.
Therefore, it may be favorable to start from zero conditions as always happens for the first frequency of
an analysis starting from scratch. There is some difficulty in deciding how far away from the previously
converged excitation frequency is acceptable to return to zero initial conditions. There is presently no
logic for handling this situation and the only course available is to reset the initial conditions to zero by
starting a completely new analysis with a starting frequency somewhere after the instability point.
Example
A nonlinear harmonic response was developed to study rotor/stator contact problems, but the presence
of a rotor is not obligatory. The capability may be used to study any periodic response to a harmonic
excitation in the presence of light nonlinearities.
The following example Figure 16-7 shows a rotor bearing system in which an out-of-balance load excites
a rigid overhung disk mounted on a flexible shaft turning in bearings exhibiting nonlinear stiffness.
Main Index
416 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Harmonic Response
The round solid shaft of diameter 0.1 metres is 1.0 metre long and runs in two bearings, one located at
one end of the shaft while the other is positioned just inboard of the disk such that the disk is overhung.
The bearings are mounted on an isolation material that exhibits nonlinear stiffness varying with a cubic
law. The overhung massive rigid disk exhibits a small eccentricity in its mass distribution.
Disk Properties:
First, the analysis is run in a linear direct frequency response with linear-bearing properties. The analysis
is then repeated in the nonlinear harmonic solution sequence, still with linear bearing properties and the
answers compared. Finally, the nonlinear bearing stiffness properties are added, and the response is
compared against theory.
Main Index
CHAPTER 16 417
Special Topics
The bulk data for the direct linear frequency response analysis model is shown as Listing 16-2 and can
found in Install_dir/mdxxxx/doc/dynamics/nlrb108.dat
SOL 108
CEND
DISP(SORT2,PHASE)=ALL
SPC=1
RGYRO=66
DLOAD=77
FREQ=88
BEGIN BULK
$
$ BEARING STIFFNESS
CELAS2,10442,1.5+7,105,2,5,2
CELAS2,10443,1.5+7,105,3,5,3
CELAS2,10552,1.5+7,106,2,6,2
CELAS2,10553,1.5+7,106,3,6,3
$
FREQ1,88,9.549296,.0530516,240
$
$ Same for all runs
PARAM,GRDPNT,0
PARAM,COUPMASS,1
$
$ ROTOR
ROTORG,1,1,2,3,4
RGYRO,66,SYNC,1,FREQ
RSPINR,1,1,4,FREQ,1.
GRID,1
GRID,2,,.5
GRID,3,,1.
GRID,4,,.99
CBEAM,1,1,1,2,,1.
CBEAM,2,1,2,4,,1.
CBEAM,3,1,4,3,,1.
PBEAM*,1,1,7.8539820-3,4.9087390-6
*,4.9087390-6,,9.8174780-6
+
+,.68,.68 $ k1 k2
MAT1,1,2.07+11,,.27,7.75+3
$
$ DISK MASS & INERTIA
CONM2,6,3,,2000.
,200.,,100.,,,100.
$
$ STATOR
GRID,5
GRID,6,,.99
$
$ ROTOR TO STATOR CONNECTION
RBE2,941,5,123456,1
RBE2,953,6,123456,4
Main Index
418 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Harmonic Response
$
$ GROUND
GRID,105
GRID,106,,.99
$
$ BEARING DAMPING
CDAMP2,20442,1.72+4,105,2,5,2
CDAMP2,20443,1.72+4,105,3,5,3
CDAMP2,20552,1.72+4,106,2,6,2
CDAMP2,20553,1.72+4,106,3,6,3
$
$ REMOVE SINGULAR DOFS
SPC1,1,14,2,3,5,6
SPC1,1,123456,105,106
$
$ OUT OF BALANCE FOR ROTOR
DLOAD,77,1.,60.,1001,60.,1002
RLOAD2,1001,1001,,,1000
RLOAD2,1002,1002,,1002,1000
DAREA,1001,2,2,9.4286-5
DAREA,1002,2,3,9.4286-5
DPHASE,1002,2,3,-90.
TABLED4,1000,0.,1.,0.,1000.
,0.,0.,39.47842,ENDT
ENDDATA
This is to be compared with the nonlinear harmonic response with nonlinear forces defined with linear
behaviour. The partial input showing the necessary changes to use SOL 128 is show in Listing 16-3 and
the complete input can be found in Install_dir/mdxxxx/doc/dynamics/nlrb128.dat
The same problem is now run as a nonlinear harmonic response which solves the problem in an iterative
manner. Half the stiffness of the linear bearings is replaced with nonlinear force definition on NOLIN1
entries. The force-displacement relationship is defined as linear.
SOL 128
CEND
DISP(SORT2,PHASE)=ALL
SPC=1
RGYRO=66
DLOAD=77
NONLINEAR=1000
NLHARM=2000
BEGIN BULK
$
$ LINEAR BAERING STIFFNESS (HALVED)
CELAS2,10442,7.5+6,105,2,5,2
CELAS2,10443,7.5+6,105,3,5,3
CELAS2,10552,7.5+6,106,2,6,2
CELAS2,10553,7.5+6,106,3,6,3
$
$ LINEAR BEARING STIFFNESS (HALF PROVIDED BY NOLIN1)
NOLIN1,1000,5,2,-1.,5,2,1001
NOLIN1,1000,5,3,-1.,5,3,1001
Main Index
CHAPTER 16 419
Special Topics
NOLIN1,1000,6,2,-1.,6,2,1001
NOLIN1,1000,6,3,-1.,6,3,1001
TABLED1, 1001,
, -1.0, -7.5+6, 1.0, 7.5+6, ENDT
$
NLFREQ1,88,9.549296,.0530516,240
NLHARM,2000,1,1,88
$
When these two linear analyses are run, the magnitude response of GRID point 3 in the Y direction looks
like Figure 16-8:
The linear frequency response curve and the nonlinear harmonic response curve are superposed. This
shows the linear problem can be solved using the 2 different methods (SOLs 108 and 128) and the
response is the same.
Now the bearing stiffness is replaced by a cubic stiffness defined by NOLIN3 and NOLIN4 entries. The
partial input required to change the example to cubic stiffness for the nonlinear harmonic response in
SOL 128. is shown as Listing 16-4 and the complete input can be found in
Install_dir/mdxxxx/doc/dynamics/nlrb128cs.dat
Main Index
420 MD Nastran Dynamic Analysis User’s Guide
Nonlinear Harmonic Response
SOL 128
CEND
DISP(SORT2,PHASE)=ALL
SPC=1
RGYRO=66
DLOAD=77
NONLINEAR=1000
NLHARM=2000
BEGIN BULK
$
$ BEARING STIFFNESS
CELAS2,10442,1.5+7,105,2,5,2
CELAS2,10443,1.5+7,105,3,5,3
CELAS2,10552,1.5+7,106,2,6,2
CELAS2,10553,1.5+7,106,3,6,3
$
$ CUBIC NONLINEAR STIFFNESS (TENSION)
NOLIN3,1000,5,2,-1+12,5,2,3.
NOLIN3,1000,5,3,-1+12,5,3,3.
NOLIN3,1000,6,2,-1+12,6,2,3.
NOLIN3,1000,6,3,-1+12,6,3,3.
$ CUBIC NONLINEAR STIFFNESS (COMPRESSION)
NOLIN4,1000,5,2,-1+12,5,2,3.
NOLIN4,1000,5,3,-1+12,5,3,3.
NOLIN4,1000,6,2,-1+12,6,2,3.
NOLIN4,1000,6,3,-1+12,6,3,3.
$
$ Spin up
NLFREQ1,88,9.549296,.1591549,240
$ Spin down
$NLFREQ1,88,47.74648,-.1591549,240
NLHARM,2000,1,1,88
$
The analysis is run in two parts, the first starting from a cyclical frequency of 9.55 Hz (60 radians/second)
with an increasing frequency up to a frequency of 47.7 Hz (300 radians/second); this simulates a spin-up
event where each frequency is considered in its steady state condition. The second part of the analysis
simulates a spin-down event starting from a cyclical frequency of 47.7 Hz with a decreasing frequency.
The magnitude of the Y direction response of GRID point 3 is plotted against rotational frequency with
the resulting two curves is show in Figure 16-9:
Main Index
CHAPTER 16 421
Special Topics
Clearly there is a zone of bifurcation just before 23 Hz; that is to say, after 23 Hz, two possible states
exist. In the spin-down case, the solution jumps from one solution to the other and then retraces the spin-
up response curve; an unstable condition exists between these two.
The results are in good agreement with those reported in Reference [For references please see page 542]
Main Index
422 MD Nastran Dynamic Analysis User’s Guide
Design Optimization and Sensitivity Overview
Plus many other uses, so many that this material has its own guide, the MD Nastran Design Sensitivity
and Optimization User’s Guide has complete details on the use of the capability.
Main Index
CHAPTER 16 423
Special Topics
Main Index
424 MD Nastran Dynamic Analysis User’s Guide
Fourier Transform
Fourier Transform
The Fourier transform capability in MD Nastran allows a transient analysis to be performed using a
frequency response solution. Time-dependent applied loads are transformed to the frequency domain
and all frequency dependent matrix calculations are completed. The frequency response solution
variables are then transformed back into the time domain.
Fourier transform methods have been implemented in MD Nastran to integrate the equations of motion
in order to obtain the aeroelastic response of fixed wing aircraft. This capability is especially important
for this type of analysis since the unsteady aerodynamic matrices are known only in the frequency
domain. The Fourier transform method may also be used to solve for the transient response of
conventional structural models (no aerodynamic effects) subjected to periodic loads.
This capability is available in SOLs 108 and 111 for frequency response output data. For transient type
output, SOL 146 must be used. The transformation is performed when the requested load is the TLOADi
form.
Theory
Two forms of the Fourier transform are available: the Fourier series and the Fourier integral. Both
methods require necessary numerical compromises and hence produce numerical approximations. The
inverse transform includes an infinite sum, for which only a finite number of terms are numerically
evaluated. The inverse Fourier integral must be numerically integrated, which may result in integration
errors. The number of frequencies at which the integrand is evaluated is limited by the cost of
calculations.
In the Fourier series, the basic time interval is 0 t T , with the function periodic. The circular
frequencies are given by
n = 2nf
1
f = ---
T
where T is a large time equal to the period of the lowest forcing frequency.
T
– i t
P̃ a n = P a t e n dt (16-85)
0
ũ j n = H j a n P̃ a n (16-86)
Main Index
CHAPTER 16 425
Special Topics
where H ja n is the frequency response of any physical variable due to unit load. The response in the
time domain is given by
1 i t
u j t = -------- --- ũ j 0 + Re ũ j n e n (16-87)
2
n = 1
In the Fourier Integral, the time interval is the limit as T , f 0 ,and 2nf of the Fourier
series. Here, is a continuous variable. Equations (16-85), (16-86), and (16-87) take the form
P̃ a = P a t e –i t dt (16-88)
0
ũ j = H ja P̃ a (16-89)
1
u j t = --- Re ũ j e i t d (16-90)
0
For piecewise linear tabular functions (TLOAD1), a table of pairs ( x i , Y i )( i = 1,N ) prescribes N – 1
time intervals. If an X1 shift and an X2 scale factor are included, the time-dependent load at point a is
given by
t – a – X1
P a t = A a Y T --------------------------- (16-91)
X2
where A a is an amplitude factor and a is a delay factor that may depend upon the loading point.
Applying finite step-by-step integration to Eq. (16-85), the transformed load, P̃ a , is obtained for each
requested frequency.
Likewise, the general function (TLOAD2) is defined by
where t = t – T 1 – a
Main Index
426 MD Nastran Dynamic Analysis User’s Guide
Fourier Transform
The value of n must be an integer for transient analysis by the Fourier method. The transformation to
the frequency domain is also obtained by numerical integration.
These loads, which appear in the form required for frequency response, are transformed to the modal
coordinates exactly as in the modal frequency response method.
One other source of loads for aeroelastic problems is a one-dimensional gust. The same time
dependencies are allowed as defined above; however, the amplitude ( A a ) and delays ( a ) for the
aerodynamic elements are computed from areas, dihedrals, and coordinates in the flow direction.
Method 0 Approximate ũ e it as a constant in each interval (the default method). For equal
frequency intervals, this method reduces to the Fourier series approximation of Eq. 3.
Method 1 Fit ũ with a piecewise linear function, and do not approximate e it .
Method 2 Fit ũ with a cubic spline function, and do not approximate e it .
Consider Method 2. Solving the three-moment equations, the second derivatives ũ , can be found for
each for which a frequency response has been computed. Then, in any interval i i + 1 ,
i + 1 – i 2
ũ = ũ i s + ũ i + 1 r – ---------------------------------- (16-93)
6
ũ i s – s 3 + u i + 1 r – r 3
where:
– i
r = -----------------------------
-
i + 1 – i
s = 1–r
Main Index
CHAPTER 16 427
Special Topics
Integrate Eq. (16-90) using Eq. (16-93) for ũ , and sum over the integrals. Then collect the terms for
each n with the result,
N
1 i t
u t = --- Re C n t ũ n + D n t ũ n e n (16-94)
n = 1
n – n – 1 n + 1 – n
C n t = ---------------------------- E 2 – it n – n – 1 + ---------------------------- E 2 it n + 1 – n (16-95)
2 2
n – n – 1 3 n + 1 – n 3
D n t = ------------------------------------ G it n – n – 1 – --------------------------------- G it n + 1 – n (16-96)
24 24
For the first terms in Eq. (16-94) ( n = 1 ), use only the second terms on the right side of Eqs. (16-95)
and (16-96). For the last term in Eq. (16-94) ( n = N ), use only the first terms on the right side of Eqs.
(16-95) and (16-96).
G z = 2E 2 z – E 4 z (16-97)
k = K
K! zk
------
zk
ez – ----
k!
EK z = k = 0 (16-98)
z z2 z3
1 + -------------- + --------------------------------------- + -----------------------------------------------------------
K + 1 K + 1K + 2 K + 1K + 2K + 3
The above form of the inverse transform has two advantages. First, numerical problems for small values
of t are efficiently evaluated by choosing the series form of Eq. (16-98). Also, the other two methods
are easily derived as subcases. If u terms are removed from Eq. (16-93), a piecewise linear fit occurs.
Thus, Method 1 results by deleting ũ from Eq. (16-94), i.e.,
N
1 i t
u t = --- Re C n t u n e n (16-99)
n = 1
with C n defined by Eq. (16-96). Method 0 results if we replace E 2 by 1.0 in Eq. (16-96).
The above procedure for Method 0 always multiplies the first and last terms in the series by one-half. In
order to force agreement with the Fourier series, which is the limiting case of equal frequency intervals,
the first term in the series is multiplied by one-half only if the value of the first frequency is zero.
Some special considerations are given in the equal frequency interval case. When all f ’s are equal and
the first frequency is an integer multiple of f , the time step t is adjusted to make
f t = 1 integer , reducing the number of distinct values of sin n t and cos n t used in Eq. (16-99).
Main Index
428 MD Nastran Dynamic Analysis User’s Guide
Fourier Transform
Also, C n t and D n t Eqs. (16-95) and (16-96) become independent of n and do not need to be computed
at every frequency.
Other important practical considerations must be observed to use these methods successfully. to
illustrate one problem, consider the response of a simple damped oscillator to a pulse (Figure 16-10). The
upper three curves show the pulse and the response of the system if it is very stable and slightly stable.
Using the Fourier method, the pulse is replaced by a series of pulses, with period 1 f .
1 t t t
-----
f
(a) True Transient
P(t)
t t t
1 – cos t 1
- = 1 – --- t 2 +
--------------------------------- Method 0
Cn t = 1 2 t 2 6 (16-100)
Method 1
Thus, Method 1 (and also Method 2) produces a decaying envelope that the user may incorrectly interpret
as additional damping.
Main Index
CHAPTER 16 429
Special Topics
The use of equal frequency intervals versus unequal intervals has been studied and results are shown in
the MSC.Nastran Aeroelastic Analysis User’s Guide.
Main Index
430 MD Nastran Dynamic Analysis User’s Guide
Squeeze Film Damper Nonlinear Force
SFD uses as input the relative displacements and velocities x x· y y· at the connecting grids and outputs
the forces F x x x· y y· and F y x x· y y· acting on the SFD damper journal grid point. Equal and opposite
forces - F x x x· y y· and - F y x x· y y· are applied to the stator (SFD housing) grid point.
Referring to Figure 16-11, GRID I is on the damper journal and GRID J is on the damper housing. The
two grids should be coincident and have parallel Cartesian coordinate systems. The forces applied to the
grids are based on the relative displacements and velocities of the grids determined from the previous
time steps in the NASTRAN implicit time integration. If a parallel centering spring is used, then this
separate spring is entered using the CELAS2 two-ended element.
Main Index
CHAPTER 16 431
Special Topics
Housing
Journal
Z
GRIDS
I&J
Figure 16-11 Imbedding the SFD Model in MD Nastran: Grid I is on the Damper Journal and
Grid J is on the Damper Housing
Main Index
432 MD Nastran Dynamic Analysis User’s Guide
Squeeze Film Damper Nonlinear Force
The Bulk Data entry for the NLRSFD has the following form:
1 2 3 4 5 6 7 8 9 10
NLRSFD SID GA GB PLANE BDIA BLEN BCLR SOLN
VISCO PVAPCO NPORT PRES1 THETA1 PRES2 THETA2 NPNT
OFFSET1 OFFSET2
Field Contents
SID Nonlinear load set identification number. (Integer > 0, Required)
GA Inner (e.g., damper journal) grid for squeeze film damper. (Integer > 0, Required)
GB Outer (e.g., housing) grid for squeeze film damper. (Integer > 0, Required)
PLANE Radial gap orientation plane: XY, XZ, or ZX. See Remark 1. (Character, Default =
XY)
BDIA Inner journal diameter. (Real > 0.0, Required)
BLEN Damper length. (Real > 0.0, Required)
BCLR Damper radial clearance. (Real > 0.0, Required)
SOLN Solution option: LONG or SHORT bearing. (Character, Default = LONG)
VISCO Lubricant viscosity. (Real > 0.0, Required)
PVAPCO Lubricant vapor pressure. (Real > 0.0, Required)
NPORT Number of lubrication ports: 1 or 2 (Integer, no default)
PRES1 Boundary pressure for port 1. (Real > 0.0, Required if NPORT = 1 or 2)
THETA1 Angular position for port 1. See Remark 2. (0.0 < Real > 360.0, Required if NPORT
= 1 or 2).
PRES2 Boundary pressure for port 2. (Real > 0.0, Required if NPORT = 2).
THETA2 Angular position for port 2. See Remark 2. (0.0 < Real < 360.0, Required if NPORT
= 2)
NPNT Number of finite difference points for damper arc. (Odd Integer < 201, Default =
101)
OFFSET1 Offset in the SFD direction 1. (Real, Default = 0.0)
OFFSET2 Offset in the SFD direction 2. (Real, Default = 0.0)
Remarks
1. The XY, YZ, and ZX planes are relative to the displacement coordinates of GA and GB. The plane
coordinates correspond to the NLRSFD directions 1 and 2. GA and GB should be coincident grids
with parallel displacement coordinate systems. Wrong answers will be produced if this rule is not
followed.
Main Index
CHAPTER 16 433
Special Topics
2. The angular measurement is counterclockwise from the displacement x-axis for the XY plane, the
y-axis for the YZ plane, and the z-axis for the ZX plane.
3. OFFSET1 = Damper housing ID center offset displacement relative to OD center in the
horizontal direction. Entered as a positive value for horizontally to the left (negative x-direction)
displacement.
4. OFFSET2 = Damper housing ID center offset displacement relative to OD center in the vertical
direction. Entered as a positive value for downward (negative y-direction) displacement. Positive
entry typically used for -1 g compensation.
Note: The OFFSET2 value represents an eccentric damper housing in the vertical direction and is
typically used to compensate for the -1g displacement of damper supported by a centering
spring.
NLRSFD
Spring to Ground
Main Index
434 MD Nastran Dynamic Analysis User’s Guide
Squeeze Film Damper Nonlinear Force
Listing 16-5
$ file sfd109ex.dat
$
SOL 109
CEND
TITLE = Simple test model, SOL 109, No damping
$
ECHO= UNSORT
$
$------------- Results requests --------------------
SET 101 = 100,101
DISP (PRINT,SORT2) = 101
SET 102 = 101,102
ELFORCE (PRINT,SORT2) = 102
SET 103 = 100
OLOAD = 103
$
TSTEP = 999
NONLINEAR=1
SUBCASE 200
LABEL = 1G down static load + 20 gm-in unbalance
$ LOADSET=100
DLOAD = 200
$
OUTPUT(XYPLOT)
XGRID= YES
YGRID= YES
XTITLE= TIME (SEC)
YTITLE= SFD FORCE (X)
XYPLOT NONLINER/ 101(T1)
YTITLE= SFD FORCE (Y)
XYPLOT NONLINEAR/ 101(T2)
BEGIN BULK
$
$ 1/386.4
PARAM WTMASS258799-8
PARAM GRDPNT 0
$
TSTEP 999 30001 .000010 100
$
$===========================================================================
$ Set up transient 1G down Load
$
$LSEQ 100 150 102
DLOAD 200 1.0 1.0 201 20.0 301 20.0 302
$
GRAV 102 0 386.4 0.0 -1.0 0.0
$TLOAD1 201 150 99
TLOAD1 201 102 99
TLOAD2 301 301 LOAD 0.0 100.0166.6667 270.0
TLOAD2 302 302 LOAD 0.0 100.0166.6667 0.0
$
$ DAREA card used to 'bridge' static load case into transient solution
DAREA 150 100 1 0.0
$
$ F(f) = UNBAL * f**2 * (1/453.5924 lbm/gm) * (2*pi)**2 / 386.08858 in/sec**2
$ = UNBAL * f**2 * 2.25243e-4 (lb)
$ (where UNBAL is given in GM-IN, 'freq' in Hertz)
$ = 1.0 * (10000.*2*pi/60)^2 /453.6/386.4
Main Index
CHAPTER 16 435
Special Topics
Main Index
436 MD Nastran Dynamic Analysis User’s Guide
Squeeze Film Damper Nonlinear Force
.
Figure 16-13 SFD Force, X and Y Direction
Main Index
CHAPTER 16 437
Special Topics
Aeroelastic Analysis
The MD Nastran: Aeroelastic Analysis User’s Guide describes the theoretical aspects; the numerical
techniques used to perform aeroelastic analyses; and how to setup, run and interpret the results using MD
Nastran. The Aeroelastic Guide is the best source for information, only an overview of this important
capability is presented here.
Each is described in the text that follows. In addition, information is provided on the ability to include
aeroelastic responses within the MD Nastran optimization capability, the aerodynamic methods
available, and on special features that make the MD Nastran aeroelastic capability unique.
Main Index
438 MD Nastran Dynamic Analysis User’s Guide
Aeroelastic Analysis
Static aeroelastic divergence is a non-oscillatory instability condition that can occur when the
aerodynamic forces overpower the stiffness of the structure. For free-flying vehicles, this phenomenon
is typically not of concern, but it can be critical in the structural design of restrained wind tunnel models.
Static aeroelastic divergence analysis can be performed as an option within the overall static aeroelastic
capability.
Aerodynamic Flutter
Flutter is the oscillatory aeroelastic instability that occurs at some airspeed at which energy extracted
from the airstream during a period of oscillation is exactly dissipated by the hysteretic damping of the
structure. The motion is divergent in a range of speeds above the flutter speed. Flutter analysis utilizes
complex eigenvalue analysis to determine the combination of airspeed and frequency for which the
neutrally damped motion is sustained (see Figure 16-14).
0.2
1st Mode
2nd Mode
0.1
Damping, g
0.0
Flutter Divergence
-0.1
-0.2
0 500 1000 1500 2000
Velocity (ft/sec)
Figure 16-14 Flutter Stability Curve
Three methods of flutter analysis are provided: the American flutter method (called the K-method in MD
Nastran), an efficient K-method (called the KE-method) for rapid flutter evaluations, and the British
flutter method (called the PK-method) for more realistic representation of the unsteady aerodynamic
loads as frequency dependent stiffness and damping terms. The complex eigenvalue analysis is specified
Main Index
CHAPTER 16 439
Special Topics
by the user with the K-method, and the QR-transformation method is used with the KE- and
PK-methods. Again, linear and/or surface splines may be used to connect the aerodynamic and structural
grid points.
3.0E8
Root Bending Moment (in-lbs)
-5.0E8
0 8
Time (sec)
Aeroelastic Optimization
The integration of the aeroelastic analysis capability contained in MD Nastran with a design sensitivity
and optimization capability provides a design tool for the aeroelastician. Sensitivity analysis entails the
determination of the effects that changes in structural properties have on response quantities, such as
displacements or stresses. Optimization utilizes information on the response values and their
sensitivities to automatically determine a design that meets a design objective, such as limits on stresses,
Main Index
440 MD Nastran Dynamic Analysis User’s Guide
Aeroelastic Analysis
deformations, or flutter characteristics. The static aeroelastic and flutter analyses are available within the
MD Nastran optimization capability and can be used in multidisciplinary fashion along with standard
static analysis, normal modes analysis, and dynamic response analysis. Dynamic aeroelastic response is
not available for optimization. Aeroelastic responses available for the sensitivity and optimization
include stability derivatives and trim settings from the static aeroelastic analysis and flutter damping level
from the flutter analysis.
Aerodynamic Methods
Five oscillatory aerodynamic theories are available for flutter analysis. There is one subsonic method,
the Doublet-Lattice Method with body interference, and three supersonic methods: the Mach Box
Method, Piston Theory, and a new (Aero II option) multiple interfering surface method called ZONA51.
The fifth method is rudimentary Strip Theory, which can be applied at any Mach number. The static and
dynamic aeroelastic response solutions use both the subsonic Doublet-Lattice and supersonic ZONA51
aerodynamic methods.
Special Features
Aeroelastic analysis in MD Nastran provides several advances in the state of-the art. The fundamental
problem of interconnecting the aerodynamic and structural grids in the finite element models is solved
by a closed form solution to an infinite plate over multiple supports. This two-dimensional interpolation
was developed in addition to a generalization of the one-dimensional spline for a bending, twisting beam
(elastic axis) on multiple collinear supports.
The implementation of the lined-up British flutter method, called the PK-method in MD Nastran, was the
first attempt to popularize the British approach to flutter analysis in the United States. This, along with
the transfer function capability for control systems, makes analysis of aeroservoelastic problems a routine
matter. The transfer function representation is for second order systems: a single output from multiple
inputs.
The analysis of response to a discrete gust requires Fourier transform methods, because the aerodynamics
assume harmonic motion. First, a direct transform of the discrete gust profile is necessary to place the
forcing function in the frequency domain. Second, an inverse Fourier transform of the forced frequency
response is necessary to obtain the transient response of vehicles to the gust. Both the direct and inverse
Fourier transform calculations have been implemented in MD Nastran.
The quasi-steady equations of motion of a free-flying vehicle require consideration of the inertial relief
effects. For the unrestrained vehicle, the inertial effects are contained in the basic stability and control
derivatives. However, derivatives that are independent of weight distribution are desirable for use in
flight simulators, and are obtained by assuming the aircraft to be restrained in some reference support
configuration. The equations of motion using restrained aeroelastic derivatives require not only
additional inertial derivatives, but also the rotations of the mean axes relative to the support for each
aerodynamic variable (e.g., angle of attack, elevator rotation, pitch rate). These additional aeroelastic
coefficients permit the support to be unloaded and angular momentum to be conserved. MD Nastran
provides both restrained and unrestrained aeroelastic derivatives, and in the restrained case, the inertial
derivatives and mean axis rotations.
Main Index
CHAPTER 16 441
Special Topics
Control systems can also be included in dynamic response analysis using MD Nastran transfer functions.
With this feature, the aeroservoelastic interactions of ride comfort and load alleviation systems can be
investigated.
The MD Nastran procedure has general capabilities that are beyond those listed here, and their
application to aeroelastic design is limited only by the analyst’s ingenuity. For example,
aerothermoelasticity considers the effects of thermal stresses on structural stiffness and the subsequent
aeroelastic interactions. MD Nastran provides a capability for nonlinear static analysis that includes
temperature loadings. Aerothermoelastic problems of high speed flight can therefore be addressed by
restarting any of the three aeroelastic analyses from a database created by the nonlinear analysis that has
generated the stiffness of the heated structure.
Aeroelastic Solutions
The MD Nastran Aeroelastic Analysis User’s Guide describes the theoretical aspects and the numerical
techniques used to perform aeroelastic analyses with MD Nastran. As described in Overview of
Aeroelastic Analysis, 437, the system is used for flutter, frequency response, gust response, and static
analysis of aerodynamically loaded structures. An outline of the capability is given here.
The aeroelastic analyses use the following features:
Structural Model
Any of the existing MD Nastran structural finite elements (except axisymmetric and p-elements) can be
used to build the structural model. The structural stiffness, mass, and damping matrices required by the
aeroelastic analyses are generated by MD Nastran from the user input of geometric, structural, inertial,
and damping data, for subsequent use in the various aeroelastic analyses.
Fluid/Structure Connections
Matrices of aerodynamic influence coefficients are computed only from the data describing the geometry
of the aerodynamic finite elements. The choice of aerodynamic grid points for the aerodynamic model
is independent of the location of the structural grid points. An automated interpolation procedure is
provided to relate the aerodynamic to the structural degrees-of-freedom. Splining techniques for both
lines and surfaces are used to generate the transformation matrix from structural grid point deflections
to aerodynamic grid point deflections where local streamwise slopes are also computed. The transpose
of this matrix transfers the aerodynamic forces and moments at aerodynamic boxes to structural grid
points.
Aerodynamic Theories
One subsonic and three supersonic lifting surface aerodynamic theories are available in MD Nastran, as
well as Strip Theory. The subsonic theory is the Doublet-Lattice method, which can account for
interference among multiple lifting surfaces and bodies. The supersonic theories are the Mach Box
method, Piston Theory, and the ZONA51 method for multiple interfering lifting surfaces.
Main Index
442 MD Nastran Dynamic Analysis User’s Guide
Aeroelastic Analysis
Modal Formulation
Dynamic aero solutions provide for modal reduction of the system matrices. The number of
degrees-of-freedom required for accurate solutions to dynamic aeroelastic problems is generally far less
than the number of physical degrees-of-freedom used in the finite element structural model. The number
of independent degrees-of-freedom can be greatly reduced by using the (complex) amplitudes of a series
of vibration modes as generalized coordinates, e.g., by Galerkin’s method. MD Nastran can compute the
vibration modes and frequencies and make the transformation to modal coordinates. The matrices of
aerodynamic influence coefficients are also transformed to generalized aerodynamic forces by use of the
vibration eigenvectors.
Flutter Analysis
The dynamic aeroelastic stability problem, flutter, is solved in SOL 145, by any of three methods. The
traditional American flutter method developed by the Air Materiel Command (AMC) in 1942 is available
in the first two methods. The first method is called the K-method and is a variation of the AMC method.
The second method, called the KE-method, is more efficient from the point of view of tracking roots, but
is limited in input (no viscous damping) and output (no eigenvectors). The third method, called the
PK-method, is similar to the British flutter method, which was developed by the Royal Aircraft
Establishment.
Frequency Response
The coupling with aerodynamic loads has also been added to the existing MD Nastran structural modal
frequency response capability, SOL 146. Analyses of frequency response to arbitrarily specified forcing
functions can be carried out using the oscillatory aerodynamic loads from any of the available
aerodynamic theories. Frequency response to a harmonic gust field can be calculated at subsonic speeds
using the Doublet-Lattice method for wing/body interference, and by the ZONA51 method for interfering
lifting surfaces at supersonic speeds.
Transient Response
Because unsteady aerodynamic loads are obtained only for steady-state harmonic motion, they are known
only in the frequency- and not the time-domain. In SOL 146, Inverse Fourier Transform techniques
provide the appropriate methods by which transient response is obtained from the frequency response.
Both forward and inverse Fourier transforms are provided so that the time-varying forcing function or
the gust profile can be transformed into the frequency domain. Then, after convolution with the system
frequency response, the inverse transform leads to the transient response of the system to the specified
forcing function or gust profile.
Main Index
CHAPTER 16 443
Special Topics
Random Response
Stationary random response of the system, is available in SOL 146 from specified loadings and the power
spectral densities of loads. Loads may be either specified force distributions or harmonic gust fields.
The statistical quantities of interest in the response are A , the ratio of standard deviations (rms values)
of the response to that of the input loading, and N o , the mean frequency of zero crossings (with a positive
slope) of the response. The capability to compute these quantities was added to MD Nastran by
modifying the existing random response module to include options to generate various atmospheric
turbulence power spectra and to perform the calculation of N o .
Design Sensitivities
The sensitivities of response parameters to changes in design variables are calculated by the perturbation
techniques developed for structural optimization in MD Nastran and extended to include static
aeroelasticity and flutter in SOL 200. The basic aeroelastic sensitivities that can be obtained include
stability derivatives, trim variables, and flutter system dampings. The synthetic response technique of
MD Nastran optimization also permits the calculation of sensitivities of user-specified functions of those
standard response quantities.
Aeroelastic Optimization
Optimization of aeroelastic characteristics can be combined with the other optimization features of
MD Nastran in SOL 200, and vehicles can now be designed optimally for aeroelastic loads, flying
qualities, and flutter, as well as for strength, vibration frequencies, and buckling characteristics.
Aeroelastic Modules
Options are available to:
• Generate aerodynamic grid points.
• Compute aerodynamic matrices.
• Provide connection (interpolation) between the structural and aerodynamic grid points.
• Solve the equations for static aeroelasticity.
• Solve the equations for flutter.
• Solve the equations for dynamic aeroelastic response.
• Calculate aeroelastic design sensitivities.
• Optimize aeroelastic and related structural characteristics.
Main Index
444 MD Nastran Dynamic Analysis User’s Guide
Aeroelastic Analysis
Aerodynamic Theories
MD Nastran has implemented six aerodynamic theories:
1. Doublet-Lattice subsonic lifting surface theory (DLM)
2. ZONA51 supersonic lifting surface theory
3. Subsonic wing-body interference theory (DLM with slender bodies)
4. Mach Box method
5. Strip Theory
6. Piston Theory
Each of these methods is described in the MSC.Nastran Aeroelastic Analysis User’s Guide. They all
share a common matrix structure.
Three matrix equations summarize the relationships required to define a set of aerodynamic influence
coefficients [see Rodden and Revell (1962)]. These are the basic relationships between the lifting
pressure and the dimensionless vertical or normal velocity induced by the inclination of the surface to the
airstream; i.e., the downwash (or normalwash),
w j = A jj f j q (16-101)
P k = S kj f j (16-103)
where:
Main Index
CHAPTER 16 445
Special Topics
wj = downwash (dimensionless)
w jg = static aerodynamic downwash; it includes, primarily, the static incidence distribution
that may arise from an initial angle of attack, camber, or twist
fj = pressure on lifting element j
q = flight dynamic pressure
k = reduced frequency, k = b V where is the angular frequency, b is a reference
semichord, and V is the free-stream velocity
A j j m,k = aerodynamic influence coefficient matrix, a function of Mach number ( m ), and
reduced frequency ( k )
uk , Pk = displacements and forces at aerodynamic grid points
1 2
D jk , D jk = real and imaginary parts of substantial differentiation matrix, respectively
(dimensionless)
S kj = integration matrix
All aerodynamic methods compute the S , D 1 and D 2 matrices at user-supplied Mach numbers and
reduced frequencies. The Doublet-Lattice and ZONA51 theories compute the A matrix. Then, matrix
decomposition and forward and backward substitution are used in the computation of the Q matrix. The
remaining methods compute A – 1 directly and use matrix multiplications to form Q . Details of the
various methods are described in the MSC.Nastran Aeroelastic Analysis User’s Guide.
w j = D j k u k + D j x u x + w jg (16-104)
where:
Main Index
446 MD Nastran Dynamic Analysis User’s Guide
Aeroelastic Analysis
u x = a vector of extra aerodynamic points used to describe, e.g., aerodynamic control surface
deflections and overall rigid body motions
w jg = represents an initial static aerodynamic downwash. It includes, primarily, the static
incidence distribution that may arise from an initial angle of attack, camber, or washout
(twist)
D jk = a substantial derivative matrix for the aerodynamic displacements. This is the D jk1 term of
Eq. (16-102). The D jk2 term is not used for this quasi-steady analysis.
D jx = a substantial derivative matrix for the extra aerodynamic points
Fa = Qa a ua
and a second matrix, Q ax , which provides forces at the structural grid points due to unit deflections of the
aerodynamic extra points, u x :
Fx = Qa x ux
K a a – q Q a a u a + M a a U·· a = q Q a x u x + P a (16-105)
This is the basic set of equations used for static aeroelastic analysis. In the general case, rigid body
motions are included in the equations to represent the free-flying characteristic of an air vehicle. This is
addressed in MD Nastran by a requirement that the user identify reference degrees-of-freedom equal in
number to the number of rigid body motions using the SUPORT Bulk Data entry. Equation (16-105) is
then partitioned into r-set (supported) and l-set (left over) degrees-of-freedom, yielding
··
K lal K lar u l M ll M lr u l K lax Pl
+ ·· = – ux + (16-106)
a K a ur
K rl M r l M rr u r a
K rx Pr
rr
Main Index
CHAPTER 16 447
Special Topics
K aaa = K a a – q Q a a
K aax = – q Q a x
where:
Main Index
448 MD Nastran Dynamic Analysis User’s Guide
Aeroelastic Analysis
A level of complexity is added if the flutter analysis includes the use of extra points. Extra points are
used for the representation of control systems and are therefore required in aeroservoelastic analyses.
The flutter analysis then uses a merged matrix
Q i i Q ie
Qh h = (16-108)
0 0
in which the h-set is a combination of the i-set normal modes and the e-set extra points. It is seen that
the lower e-set rows in the matrix are null. Physically, this indicates that the normal mode deflections do
not produce aerodynamic forces on the extra points ( Q ei = 0 ) and that the extra point deflections do not
produce aerodynamic loads on the extra points ( Q ee = 0 ).
1 2
– M h h 2 + iB h h + 1 + ig K h h – --- V Q hh m,k u h = 0 (16-109)
2
where:
Main Index
CHAPTER 16 449
Special Topics
For the K-method of solution, the aerodynamic term is converted to an equivalent aerodynamic mass
c 2 2 i
– M h h + --- ------ Q h h m,k --------------- + B hh ------------------- + K h h u h = 0 (16-110)
2 2k 1 + ig 1 + ig
The term involving B hh in Eq. (16-110) has been multiplied by 1 + ig for mathematical convenience,
and is valid only at flutter, i.e., when g = 0 . Equation (16-110) is solved as an eigenvalue problem for
a series of values for parameters m , k , and . The complex eigenvalue is 2 1 + ig , which can be
interpreted as real values of and g . The velocity, V , is recovered from V = c 2k . Flutter occurs
for values of m , k , and for which g = 0 . The solutions are not valid except when g = 0 , since
the aerodynamic force terms are valid only for sinusoidal motion and g is not a physical damping.
The K-method of flutter analysis is a looping procedure. The values of V , g , and f = 2 are
solved for various values of m , k , and . Plots of V versus g can be used to determine the flutter
speed(s) (where g goes through zero to positive values). The KE-method and the PK-method are the
other major flutter options. These are discussed in the MSC.Nastran Aeroelastic Analysis User’s Guide.
Typical flutter plots are shown in Figure 16-16 and Figure 16-17 for the output from the PK-method.
Main Index
450 MD Nastran Dynamic Analysis User’s Guide
Aeroelastic Analysis
4.0
3.0
Frequency, f (Hz)
2.0
1.0
Velocity (ft/s)
Figure 16-16 V-f Curve for BAH Wing
Main Index
CHAPTER 16 451
Special Topics
0.2
0.1
Damping, g
0.0
Flutter Divergence
-0.1
-0.2
0 500 1000 1500 2000
Velocity (ft/s)
Figure 16-17 V-g Curve for BAH Wing
Note that the bending branch goes to zero frequency before the mode goes unstable.
1
– M h h 2 + iB h h + 1 + ig K h h – --- V 2 Q h h m,k u h = P (16-111)
2
Main Index
452 MD Nastran Dynamic Analysis User’s Guide
Aeroelastic Analysis
where all terms on the left-hand side are identical to those of Eq. (16-109) and are defined with that
equation. The right-hand side provides the loading in modal coordinates, which can be aerodynamic or
nonaerodynamic in nature and is a function of the analysis frequency. Nonaerodynamic generalized
loads, designated PHF , are obtained in the standard fashion from the loadings applied to physical
coordinates.
The solution of Eq. (16-111) entails solving for the generalized displacements by
decomposition/forward-backward substitution techniques applied to the coupled set of complex
equations. Because modal reduction techniques have been applied, the solution costs are typically
modest. Once the generalized displacements have been computed, standard data recovery techniques can
be used to determine physical displacements, velocities, stress, etc.
Transformation of Loads
The user specifies loads in the same manner as given in Modal Versus Direct Transient Response, 218.
The two general forms are the tabular, piecewise linear function and the general purpose function.
These loads, which are in the form required for frequency response, are transformed to the modal
coordinates exactly as in the modal frequency response method.
Main Index
CHAPTER 16 453
Special Topics
G f = G f + iG f (16-112)
where:
The ratio
G f - = tan
------------- (16-113)
G f
is denoted as the shear loss tangent.
The above formulation of viscoelastic (frequency-dependent) material properties may be used in direct
frequency analysis (SOL 108).
The stiffness and damping components of the dynamic matrices for direct frequency response analysis
are documented in the MD Nastran Dynamic Analysis User’s Guide in the following form:
where:
g = overall structural damping specified through the PARAM,G Bulk Data entry
K d1d = stiffness matrix for structural elements
K d2d = stiffness terms generated through direct matrix input, e.g., DMIG Bulk Data entries
Main Index
454 MD Nastran Dynamic Analysis User’s Guide
Viscoelastic Material Properties
K d4d = element damping matrix generated by the multiplication of individual element stiffness
matrices by an element damping, g e , entered on the MATi Bulk Data entry associated with
the element or elements in question
B d1d = damping matrix generated through CVISC and CDAMP1 Bulk Data entries
B d2d = damping terms generated through direct matrix input, e.g., DMIG Bulk Data entries
Equation (16-3) is of particular interest in the current discussion of viscoelastic material properties
because the presence of these properties will be reflected in terms of this equation. For discussion
purposes, frequency-dependent material properties will be denoted as viscoelastic materials and those
material properties that are independent of frequency will be denoted as elastic materials. Thus, if the
stiffness properties for the viscoelastic elements are initially computed on the basis of a representative
reference modulus, G REF , the stiffness matrix for the viscoelastic elements (denoted by the subscript V )
may be written in the form
G f + iG f
Kd d f V = ------------------------------------- K d1d V (16-116)
GR E F
Input Description
To use the viscoelastic capability, the following conditions are necessary:
1. Assume the K dd1 matrix will be restricted only to the viscoelastic elements. This restriction
implies that elastic elements will have a blank or zero entry for g e on their associated MATi Bulk
Data entries. Conversely, all viscoelastic materials must have representative reference values of
g e , and G REF entered on their associated MATi Bulk Data entries. Then, by definition,
K d4d V = g R EF K d1d V
2. The TABLEDi tabular functions TR f and TI f are defined to represent the complex moduli
of all viscoelastic materials.
These two conditions may be combined in Eq. (16-3) to provide the following expression:
1 + TR f + iTI f K 4
K d d V = 1 + ig K dd V dd V
(16-117)
= 1 + g R E F TR f + i g + g RE F TI f K d1d V
A comparison of Eqs. (16-5) and (16-6) yields the form of the tabular functions TR f and TI f :
1 G f
TR f = ------------- -------------- – 1 (16-118)
g RE F G R EF
1 G f
TI f = ------------- -------------
-–g (16-119)
g R E F G R EF
Main Index
CHAPTER 16 455
Special Topics
of viscoelasticity.
Direct frequency response analyses that involve viscoelastic materials require some special input data
relative to analyses that involve only elastic materials. These special input requirements are given
below:
1. Executive Control Section:
None
2. Case Control Section:
SDAMPING = n reference TABLEDi Bulk Data entry that defines the alternate tabular form of
TR f
3. Bulk Data Section:
a. MATi Bulk Data entry
• G = G RE F , the reference modulus
• NU = Poissons ratio for the viscoelastic material
• GE = g R E F , the reference element damping
• All other entries on the MATi Bulk Data entry are utilized in the standard manner.
b. TABLEDi Bulk Data entries:
• A TABLEDi Bulk Data entry with an ID = n is used to define the function TR f of
Eq. (16-7).
• A TABLEDi Bulk Data entry with an ID = n + 1 is used to define the function TI f of
Eq. (16-8).
All other input requirements to the MD Nastran Bulk Data entry are typical of direct frequency response
analysis. Note that the overall structural damping, g , entered through the PARAM Bulk Data entry
(PARAM,G,XX) applies to all elastic materials.
G R EF « G
G
G R EF « -------
g
Main Index
456 MD Nastran Dynamic Analysis User’s Guide
Viscoelastic Material Properties
G R EF g R E F = 1
TR f = G f (16-120)
TI f = G f (16-121)
The above simplifications should not be used if OMIT calculations are involved in the analysis in order
to avoid possible matrix ill-conditioning.
Note that stress and force data recovery calculations are performed with the reference moduli irrespective
of frequency.
Example
To illustrate the representation of viscoelastic material properties in MD Nastran, consider the following
structure that may undergo both axial extension along the z-axis and torsion about the z-axis:
where:
f J
torsional stiffness, K = G
--------------
z l
Main Index
CHAPTER 16 457
Special Topics
length, l = 2.0
area, A = .9
area moment, J = 2.0
The symbols E f and G f imply that the extensional and torsional moduli are functions of frequency,
i.e., viscoelastic. For simplicity, it is assumed that E f = G f and that these quantities have the
following frequency-dependent characteristics:
f, hz G(f) G(f)
.8 1800. 180.
1.1 1850. 185.
1.4 1910. 191.
1.7 1970. 197.
2.0 2030. 203.
2.3 2070. 207.
2.6 2140. 214.
2.9 2210. 221.
Main Index
458 MD Nastran Dynamic Analysis User’s Guide
Viscoelastic Material Properties
2, 3 2, 6
z
21 Iz 24
K z 23 K 23
z
3 3
i = Grid Point ID
k = Element ID
Figure 16-18
This model can be generated with the following MD Nastran Bulk Data entries:
1 2 3 4 5 6 7 8 9 10
GRID ID CP X1 X2 X3 CD PS
GRID 2 2. 1245
GRID 3 123456
CMASS2 EID M G1 C1 G2 C2
CMASS2 21 2. 2 3 3 3
CMASS2 24 10. 2 6 3 6
Main Index
CHAPTER 16 459
Special Topics
Note that the reference values of 2000. for both E and G are specified on the MAT1 Bulk Data entry.
The reference value for structural damping, g REF , is set to .09 under the GE field of the MAT1 Bulk Data
entry. Once the reference values G REF and g REF have been assigned, one can evaluate Eqs. (16-7) and
(16-8) for the values to be assigned to TR f and TI f . The values for these functions are entered on
TABLEDi Bulk Data entries. As elastic portions of the structure may exist in addition to viscoelastic
portions, assume that a value of overall structural damping, g, is to be utilized for these elastic portions
of the model. The overall structural damping value of .06 is assigned through the following PARAM
Bulk Data entry.
1 2 3 4 5 6 7 8 9 10
PARAM N V1 V2
PARAM G .06
This value of g must be considered in Eq. (16-8). The evaluation of Eq. (16-7) and Eq. (16-8) will result
in the values shown in the following TABLED1 Bulk Data entries:
1 2 3 4 5 6 7 8 9 10
TABLED1 ID
TABLED1 10 TABR1
X1 Y1 X2 Y2 X3 Y3 X4 Y4
+ABR1 .0 .0 .8 –1.11111 1.1 –.833333 1.4 –.5 TABR2
X5 Y5 X6 Y6 X7 Y7 X8 Y8
+ABR2 1.7 –.166667 2. .1666667 2.3 .3888889 2.6 .7777777 TABR3
X9 Y9
+ABR3 2.9 .5611111 ENDT
TABLED1 ID
TABLED1 11 TABI1
X1 Y1 X2 Y2 X3 Y3 X4 Y4
+ABI1 0. 0. .8 .3333333 1.1 .3611111 1.4 .3944444 TABI2
X5 Y5 X6 Y6 X7 Y7 X8 Y8
+ABI2 1.7 .4277778 2. .4611111 2.3 .4833333 2.6 .5222222 TABI3
Main Index
460 MD Nastran Dynamic Analysis User’s Guide
Viscoelastic Material Properties
X9 Y9
+ABI3 2.9 .5611111 ENDT
To demonstrate that elastic as well as viscoelastic elements may be included in the same analysis, the
following single degree-of-freedom is added to the Bulk Data Section:
U 1
m = 10.
K = 10000.
3
1 B = 2.
g = .06
i = Scalar Point ID
j = Element ID
The following Bulk Data entries are required to represent the foregoing single degree-of-freedom
oscillator:
1 2 3 4 5 6 7 8 9 10
CELAS2 EID K G1 C1 G2 C2 GE S
CELAS2 1 1000. 1
CDAMP2 EID B G1 C1 G2 C2
CDAMP2 2 2. 1
CMASS2 EID M G1 C1 G2 C2
CMASS2 3 10. 1
The excitation for both disjoint models is a force (moment for the torsional system) with a magnitude of
cos t . This function can be generated with the following Bulk Data entries:
Main Index
CHAPTER 16 461
Special Topics
1 2 3 4 5 6 7 8 9 10
RLOAD2 SID L M N TB TP
RLOAD2 1 1 0 0 1 0
TABLED1 ID
TABLED1 1 +ABC
X1 Y1 X2 Y2 X3 Y3
+ABC 0. 1. 1. 1. 100. 1. ENDT
DAREA SID P C A
DAREA 1 2 3 1.
DAREA 1 2 6 1.
DAREA 1 1 1.
To perform a frequency response analysis is necessary to provide a list of frequencies (Hz) at which
solutions are desired. The following FREQ1 Bulk Data entry is used for this purpose.
Main Index
462 MD Nastran Dynamic Analysis User’s Guide
Viscoelastic Material Properties
$ Structural damping
PARAM G .06
$ Frequency-dependent
TABLED1 10 +
+ 0. 0. .8 -1.111111.1 -.8333331.4 -.5 +
+ 1.7 -.1666672. .16666672.3 .38888892.6 .7777777+
+ 2.9 .5611111ENDT
TABLED1 11 +
+ 0. 0. .8 .33333331.1 .36111111.4 .3944444+
+ 1.7 .42777782. .46111112.3 .48333332.6 .5222222+
+ 2.9 .5611111ENDT
$ Oscillator
CELAS2 1 1000. 1
CDAMP2 2 2. 1
CMASS2 3 10. 1
$
$ Load
$
RLOAD2 1 1 0 0 1 0
TABLED1 1 +
+ 0. 1. 1. 1. 100. 1. ENDT
DAREA 1 2 3 1.
DAREA 1 1 1.
$ frequencies
FREQ1 1 .5 .3 10
$
ENDDATA
Main Index
CHAPTER 16 463
Special Topics
OPENFSI
Introduction
The MD OpenFSI service provides the ability to solve coupled fluid structure interaction (FSI) problems
as well as to generally access forces calculated by an external service. MSC Software has partnered with
select CFD vendors to provide services for fluid structure interaction problems. Users can similarly
utilize the published APIs and build environment to create custom OpenFSI services. Custom OpenFSI
services can be as simple as a lookup tables for forces or as extensive as an interface to an in-house CFD
code. This document describes how to utilize available OpenFSI services as well as how users can create
their own OpenFSI services.
OpenFSI is intended for problem where the boundary conditions cannot be specified simply by a table
or field. Instead, OpenFSI boundary conditions must be solved by coupling to an external code or
application.
A common OpenFSI multidisciplinary application is where the fluid flow affects the structural response
and the structural response in turn affects the fluid flow. In such applications the structural model must
be coupled to a flow field solution in an external CFD code or user defined application.
• MD Nastran supports CDFD structural applications where MD Nastran provides the
displacements and velocities on the OpenFSI boundary while the service returns the calculated
forces on the wetted surface node. Some applications in MD Nastran include: Automotive –
Door seal aspiration, shock absorbers, hydraulic engine mounts, convertible top; Aerospace –
Flexible wings, Latch loads; and Biomed – Heart valves, Flow regulators
For CFD services, the vendor has implemented the Open FSI APIs and the service is delivered as part of
the CFD software installation. The CFD code should support a model with boundary conditions that can
be tagged by the OpenFSI service. A list of current commercially available OpenFSI services from CFD
partners can be found on the MSC website:
http://www.mscsoftware.com/partners/technology.cfm?Q=434&Z=436.
In the case of look-up table services, the user creates a SCA service that complies with the APIs
(OpenFSI IDL). The tools for building these custom SCA services are delivered in the MD Nastran
solver SDK. An OpenFSI lookup table use case is described in Example, 474.
Feature Description
OpenFSI is based on the Simulation Component Architecture (SCA) framework. It allows the MD
nonlinear solver to communicate with a CFD code or other external code to access forces computed by
the CFD code and send structural displacements and velocities computed by MD Nastran during
dynamic simulations. MSC has partnered with several CFD vendors who have agreed to implement the
published OpenFSI APIs. These services are delivered in the form of a library and SCA catalog entry
that enables the communication.
Main Index
464 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
To use a commercial CFD OpenFSI service, you simply point to the location of these SCA services files
on the network. You can similarly create OpenFSI services for in-house CFD codes or other application
by implementing the OpenFSI APIs with the SCA build environment included in MD Nastran. The SCA
service has to be built on the same platform as MD Nastran. However, the implementation can
accommodate the flexibility to run MD Nastran and the CFD solutions on a different platform; e.g., one
on Windows the other on Linux. This allows OpenFSI simulations where the CFD code may reside on a
platform not supported by MD Nastran.
For coupling with CFD codes, the OpenFSI approach assumes a pre-existing CFD model with wall
boundaries corresponding to the wetted surface of the structural model. The CFD model should be ready
to run except for displacement information to be passed by OpenFSI. MD Nastran only needs to be aware
of results on the wetted surface nodes. The CFD wall surfaces and the MD Nastran wetted surfaces
participating in the FSI solution should have similar geometric form, although their respective meshes
will likely be different.
In MD Nastran 2010, data interpolation of coupling regions will be performed by the CFD code.
MD Nastran obtains the required coupling definition from SimXpert and the CFD solver accesses the
structural wetted surface via the API. The CFD coupling region should be collocated and in the same
units as the MD Nastran structural model.
MD Nastran and the CFD code exchange data on the on the wetted surfaces to account for the viscous
and pressure loads, and corresponding displacements during the simulation. The Open FSI process is
illustrated in Figure 16-19.
Main Index
CHAPTER 16 465
Special Topics
The OpenFSI service is called from the MD nonlinear solution (SOL 400) at each solution time step or
at a user specified iterative frequency. The basic dataflow is illustrated in Figure 16-20.
The OpenFSI interface communicates the data on wetted surfaces, see Figure 16-21, which are the
surfaces where the fluid is in contact with the structures. The wetted surfaces are defined in MD Nastran
as meshes consisting of triangular and quadrilateral elements. The MD Nastran wetted surface mesh
coordinates and elements are sent to the CFD code (or external code) in an OpenFSI initialize call. The
matching surfaces in the CFD code can take any form chosen by the CFD vendor, and the mapping
between the possibly discrepant wetted surface representations is performed by the CFD code. Note that
in this implementation, the mesh topology is constant during the simulation, which means that no mesh
adaptivity is supported.
Main Index
466 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
Table 16-6 Nastran input file structure for single wetted load WL1, and a single wetted
surface WS1.
Main Index
CHAPTER 16 467
Special Topics
The dependency between the entries used for OpenFSI using a single load (with input file structure as
Table 16-7) is illustrated in the diagram.
Table 16-7
WETSURF
WETELMG WETELME
Main Index
468 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
If a linear combination of TLOAD1 entries are specified in the DLOAD Bulk Data entry (as in
Table 16-7), we have the dependency between the entries used for OpenFSI as follows:
WETSURF
WETELMG WETELME
The definitions of the OpenFSI Bulk Data entries are explained below. The FSICTRL and WETLOAD
entries reference SERV ID, an OpenFSI SCA service. If more than one FSICTRL and WETLOAD
entries are used, they may reference different OpenFSI SCA services, which is useful if different external
codes are used for different sets of wetted surfaces.
Two different approaches may be used to define the wetted surface elements:
• Using the WETELMG entry, the grid points making up a wetted element are specified;
• Using the WETELME entry, a wetted element is defined by referencing a face (or a side) of a
parent structural element.
If a shell has wetted surfaces on both faces, separate WETELMG or WETELME are provided on each
face.
OpenFSI support an implicit or explicit type coupling with the external service. Explicit coupling is the
simplest type of service, as the nodal forces from the external code are only read at the beginning of the
time step, and the nodal results, the displacement and velocity, are only sent at the end of the time step.
The data flow for an explicit service is illustrated in Figure 16-22,
Main Index
CHAPTER 16 469
Special Topics
In the implicit service coupling, the data is communicated inside the Newton-Raphson loop at a
frequency specified by the FSICTRL entry in the Bulk Data Section. Note that the nodal forces are also
read at the beginning of the time step before entering the Newton-Raphson loop (not shown here), in
which case the FSI forces do not have to be read at the first iteration. The data flow for an implicit method
service coupling is illustrated in Figure 16-23.
Main Index
470 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
The formats of the Bulk Data entries FSICTRL, WETLOAD, WETSURF, WETELME, and WETELMG
to support OpenFSI are described here:
FSICTRL - MD Only
1 2 3 4 5 6 7 8 9 10
FSICTRL SERV_ID TYPE FREQ
Examples:
Main Index
CHAPTER 16 471
Special Topics
WETLOAD - MD Only
1 2 3 4 5 6 7 8 9 10
WETLOAD WLID WSID SERV_ID
Example:
Field Contents
WLID Load set ID, referenced by the EXCITEID field in the TLOAD entry. (Integer > 0; no
Default)
WSID Wetted surface identification number. The wetted surface must be defined in the
WETSURF Bulk Data entry. (Integer > 0; no Default)
SERV_ID OpenFSI SCA service name associated with the wetted surface loads. The OpenFSI
SCA service is defined using the CONNECT SERVICE FMS entry. (Character; no
Default)
Main Index
472 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
WETSURF - MD Only
1 2 3 4 5 6 7 8 9 10
WETSURF WSID WTAG
WEID1 WEID2 WEID3 WEID4 WEID5 WEID6 WEID7 WEID8
WEID9 WEID10 -etc.-
Alternate Format:
Field Contents
WSID Wetted surface identification number. (Integer > 0; no Default)
WTAG Wetted surface tag name exported to an external code using the OpenFSI SCA
interface. (Character; no Default)
WEID1, Wetted element identification numbers defined using the WETELMG or
WEID2, ... WETELME Bulk Data entries. (Integer > 0; no Default)
THRU, BY Keywords to specify a range of wetted elements. (Character; no Default)
INC Increment to use with the “THRU” and “BY” keywords. (Integer; Default = 1)
WETELME - MD Only
1 2 3 4 5 6 7 8 9 10
WETELME WEID EID SIDE
Example:
WETELME 10001 34 3
Main Index
CHAPTER 16 473
Special Topics
Field Contents
WEID Wetted element identification number. (Integer > 0; no Default)
EID Structural element identification number, which corresponds to a surface element
CQUAD4, CQUAD8, CQUADR, CQUAD, CTRIA3, CTRIA6, CTRIAR; or a solid
element CTETRA, CPENTA, or CHEXA. (Integer > 0; no Default)
SIDE Side identification number of element EID. (1 < Integer < 6; no Default)
WETELMG - MD Only
1 2 3 4 5 6 7 8 9 10
WETELMG WEID TYPE
G1 G2 G3 G4 G5 G6 G7 G8
Example:
Field Contents
WEID Wetted element identification number. (Integer > 0; no Default)
TYPE Wetted element type, which can be any of TRIA3, TRAI6, QUAD4 or QUAD8.
(Character; no Default)
G1, ..., G8 Grid point identification numbers for the wetted surface element WEID. (Integer > 0;
no Default)
Limitations
The following limitations should be noted for MD Nastran 2010:
1. OpenFSI is limited to structural applications, where MD Nastran solves for displacements and
velocities on the wetted surface and the service calculates the forces on the wetted surface nodes.
The plan is to extend OpenFSI to heat transfer applications in a future release.
2. The wetted surface is limited to triangular or quadrilateral faces of 3D elements or 2D faces.
3. Interpolation of dissimilar meshes are performed by the CFD code. A mapping component is
planned in a future release.
4. The CFD coupling region should be collocated and in the same units as the MD Nastran structural
model. This restriction may be relaxed in a future release.
Main Index
474 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
Example
Two OpenFSI examples are presented in this section. The first example utilizes a predefined OpenFSI
CFD service (AcuSolveAcuSolve from Acusim). MD Nastran and the CFD code are coupled for the
transient FSI simulation. The second example involves building a user defined OpenFSI service that
calculates nodal forces on the wetted surface nodes based on an expression that is a function of time and
node number.
This example illustrates setting up a transient FSI simulation using one of the available OpenFSI CFD
partner services. MD Nastran calculates the baffle deformation and nodal velocities while the CFD code
calculates the flow induced loads on the baffle wetted surfaces. The initial condition is a converged CFD
solution on the undeformed baffle. The SOL 400 nonlinear transient solution references the connected
OpenFSI service. This is transient FSI problem as indicated by the XY-plot.
The steps in this analysis follow a typical scenario in industry. The CFD group has created the CFD
model ignoring the baffle deformation. The structural analyst has an existing model of the baffle that may
have been solved based on the undeformed pressure distribution. The baffle geometry and spatial
location are the same in both models. The objective is to get the true baffle behavior based on the fluid
structure interaction. This example illustrates the OpenFSI-AcuSolve service from Acusim (the steps
should be similar for other CFD partners). The basic steps are:
1. The CFD partner will deliver the OpenFSI service in the form of a library (.dll or .so) and a
SCAServiceCatalog (.xml). Make sure that the partner CFD code has been installed and the SCA
service environment variables are set to locate the OpenFSI CFD service
(SCA_LIBRARY_PATH, SCA_RESOURCE_DIR, SCA_SERVICE_CATALOG). See the
User Defined Services guide for more detail on user services.
2. Obtain the ready to run CFD model. The input file may need to be edited to enable FSI coupling.
3. The structural file for this example is included in the tpl folder (tentatively named
plate_baffle_cfd.dat)
4. The SCA OpenFSI library and service catalog need to be moved under the Nastran installation
directory or included in the environmental variables for finding SCA services.
Main Index
CHAPTER 16 475
Special Topics
5. The user creates a model in SimXpert that references the OpenFSI service for loads on wetted
surfaces.
6. The model is submitted to MD Nastran. MD Nastran will use the SCA environment path to locate
and load the service.
7. Displacement results appear in the standard MD Nastran output files and are postprocessed by
SimXpert.
The CFD model should be ready to run except for a few input file changes involving the coupling to the
structural code.
The task for the structural engineer is to determine the deformation and stresses in the baffle. The fluid
loads on the wetted surfaces of the baffle are obtained by the CFD solver.
The user should obtain the CFD model for the baffle in a duct. The geometry in the CFD model is
illustrated in Figure 16-24.
The CFD model consists of approximately 800K tets. Note the baffle surface in the CFD model is made
up of triangular faces while in the structural model the baffle wetted surfaces are quadrilateral. The
mapping will be handled by AcuSolve. Flow enters the channel on the left face of the volume. The flow
conditions should create sufficient pressure to deform the baffle (~2000 N/m2).
The coupling feature has to be enabled in the CFD input file. The following illustrates this step in
AcuSolve:
EXTERNAL_CODE {
communication = socket
socket_initiate = off
socket_host = "name_of_nastran_host"
socket_port = 10000
}
The specified host is where the MD Nastran model will be running on the network. It can be a different
platform than where the CFD code is running (e.g., MD Nastran on Linux, AcuSolve on Windows). The
socket host name must include the quotes. The port number (10000) has been set in the AcuSolve service
but could be changed if there conflicts through the environment variable
“"ACUSIM_NASTRAN_PORT".
Main Index
476 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
The set of CFD surface elements corresponding to the baffle structural wetted surface are also identified
in the input file.
EXTERNAL_CODE_SURFACE( "baffle" ) {
surfaces = Read( "baffle.srf.tri" )
shape = three_node_triangle
element_seT = "interior"
mesh_displacement_type = tied
velocity_type = wall
gap = 0.0
gap_factor = 0.0
}
Once the CFD part has been prepared, the rest of the OpenFSI problem can be set up through SimXpert.
The OpenCFD service is delivered by the CFD vendor and the location of the service should be set as
described earlier.
Main Index
CHAPTER 16 477
Special Topics
Main Index
478 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
Note: The form above shows a generic name (myService.openFSI) for the OpenFSI Service
name. This will be translated to the Connect Service entry in the MD Nastran input file
along with the alias Name (8 characters or less). Check for the actual service in the
SCAServiceCatalog.xml as defined by the vendor, for example,
'acuSolveService.openFSI'
Main Index
CHAPTER 16 479
Special Topics
Main Index
480 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
Main Index
CHAPTER 16 481
Special Topics
Main Index
482 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
Main Index
CHAPTER 16 483
Special Topics
Main Index
484 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
Main Index
CHAPTER 16 485
Special Topics
Main Index
486 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
Main Index
CHAPTER 16 487
Special Topics
Main Index
488 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
The forces returned by the service on the wetted surface nodes are a function of time based on the
following formula:
force_x = 0.00
force_y = 0.01*sin( 2.0*PI * fsiTime / 2.0 )
force_z = -10.0*cos( 2.0*PI * fsiTime / 2.0 )
where “fsitime” is the transient time in the nonlinear solution. The nonlinear simulation is to determine
the transient deformations resulting from the external service load.
Main Index
CHAPTER 16 489
Special Topics
7. Displacement results appear in the standard MD Nastran output files and are postprocessed by
SimXpert.
In many situations, the service creation in steps 1 and 2 above will have been done by a commercial
vendor (e.g., CFD code) or a methods group in the company. In these cases the user would start with Step
3. The example below will first illustrate the case where the service exists. The second part will illustrate
how to create the service.
Main Index
490 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
SCA Service allows selection from a list of OpenFSI services that have been defined under the User
Services menu.
Wetted Surface defines the surfaces where loads are to be applied from the OpenFSI external service.
The user can select either surface(s) or element face(s) as wetted surface. If the user selects a surface,
then all the elements faces lying on the surface will be exported as WETELMG*.
Coupling Region Tag Name is passed to the OpenFSI service and usually represent the BC name in the
CFD code corresponding to the OpenFSI wetted surface. In the case of a lookup table, this tagname can
be used to define an entry point in the service. The default name of the companion region will be given
as WS_<ID>.
Load Scale Factor can be used to scale the loads coming from the OpenFSI service (default is 1.0).
This Open FSI object will create the TLOAD1 along with the WETLOAD.
OpenFSI allows pointing to different services for each OpenFSI LBC. Only one FSICTRL entry per
service will be exported in the bulk data entry, irrespective of the number of WETLOAD entries.
Main Index
CHAPTER 16 491
Special Topics
Main Index
492 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
Main Index
CHAPTER 16 493
Special Topics
The desired service will return forces returned on the wetted surface nodes are a function of time based
on the following formula:
force_x = 0.00
force_y = 0.01*sin( 2.0*PI * fsiTime / 2.0 )
force_z = -10.0*cos( 2.0*PI * fsiTime / 2.0 )
This example OpenFSI service along with the source structure is included in the delivery. See the
OpenFSI.idl file for the details of the interface, and openFsi.cpp for the implementation example.
The following methods are implemented in an OpenFSI service:
• For the initialization stage:
• initialize
• For the solver stage:
• initializeTimeStep
• getWettedNodeForces
• putWettedNodeDisplacementsAndVelocities
• finalizeTimeStep
• For the termination stage:
• terminate
The abbreviated source for this user-defined service is given below (OpenFSI.cpp file).
Main Index
494 MD Nastran Dynamic Analysis User’s Guide
OPENFSI
in the OpenFSI root directory. This will generate an Apps directory containing the library (dll or so) and
the SCA service catalog.
Main Index
CHAPTER 16 495
Special Topics
If the service has been built in a directory not in the default path, the user will have to set the following
environment variables (Windows example shown).
• SCA_LIBRARY_PATH <path>/Apps/WINNT/lib/
• SCA_SERVICE_CATALOG
<path>/Apps/res/SCAServiceCatalog.xml
• SCA_RESOURCE_DIR <path>/Apps/res/
For a detailed description of how to build a SCA service, see the MD Nastran 2010 User Supplied
Subroutines and SCA Service Guide.
Product Dependencies
CFD services requires an OpenFSI enabled CFD code. These are either commercial codes that have
implemented the OpenFSI APIs or in-house codes that have created custom services.
In MD Nastran 2010, SimXpert provides the graphical user interface for defining the structural wetted
surface application regions and selecting the OpenFSI service. For CFD OpenFSI services, the CFD
model and coupling boundary are defined in the CFD preprocessor.
In the case of the OpenFSI-UVLM (Unsteady Vortex Lattice Method) service from Zona Technologies,
the aero model is included in the Nastran input file.
Documentation Dependencies
User defined OpenFSI services are created using the SCA build environment delivered with
MD Nastran. For more information on creating SCA services see the following documents:
• MD Nastran 2010 - SCA Service Guide
• MD Nastran 2010 - User Defined Services
Main Index
496 MD Nastran Dynamic Analysis User’s Guide
Rotor Dynamics Overview
Main Index
MD Nastran Dynamic Analysis User’s Guide
Ap. A: Glossary of Terms and Nomenclature
Glossary of Terms
Nomenclature for Dynamic Analysis
Main Index
498 MD Nastran Dynamic Analysis User’s Guide
Glossary of Terms
Glossary of Terms
Main Index
APPENDIX A 499
Glossary of Terms and Nomenclature
Main Index
500 MD Nastran Dynamic Analysis User’s Guide
Nomenclature for Dynamic Analysis
General
Þ Multiplication
ª Approximately
[] Matrix
{} Vector
i –1
g Acceleration of Gravity
t Time Step
f Frequency Step
i j k Subscripts (Indices)
Infinity
u Displacement
u0 Initial Displacement
u· Velocity
u· 0 Initial Velocity
u·· Acceleration
m Mass
m0 Large Mass
b Damping
b cr Critical Damping
k Stiffness
p Applied Force
Circular Frequency
n Circular Natural Frequency
2
Eigenvalue
Eigenvalue
f Frequency
fn Natural Frequency
Tn Period
Main Index
APPENDIX A 501
Glossary of Terms and Nomenclature
Structural Properties
Geometry
L Length
A Area
Stiffness
E Young’s Modulus
G Shear Modulus
J Torsional Constant
Poisson’s Ratio
I Area Moment of Inertia
Mass
Mass Density
m Mass Density
w Weight Density
Ip Polar Moment of Inertia
Damping
Main Index
502 MD Nastran Dynamic Analysis User’s Guide
Nomenclature for Dynamic Analysis
Main Index
MD Nastran Dynamic Analysis User’s Guide
Overview
Main Index
504 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
The set notation system used for dynamic analysis in MD Nastran continues and expands the set notation
system for static analysis. Because of the great variety of physical quantities and displacement sets used
in dynamic analysis, becoming familiar with the set notation system in MD Nastran aids its efficient use
for dynamic analysis. Good reference material is available in Set Definition (Ch. 12) in the MD Nastran
Linear Static Analysis User’s Guide, Degree-of-Freedom Set Definitions (Ch. 7) in the MD Nastran
Quick Reference Guide and Constraint and Set Notation (Ch. 6) in the MD Nastran Reference Manual.
Dynamic Sets
In addition to the basic static sets there are a number of additional, mutually independent sets of physical
displacements (namely, q-set, c-set, b-set and e-set) that are used in dynamic analysis to supplement the
sets used in static analysis. The q-, c-, and b-sets facilitate generalized dynamic reduction and component
mode synthesis. The e-set is used to represent control systems and other nonstructural physical variables.
The p-set is created by combining the g-set with the e-set variables.
In addition to the combined sets described above, the v-set is a combined set created by combining the
c-, r- and o-sets. The DOFs in these sets are the DOFs free to vibrate during component mode synthesis
or generalized dynamic reduction.
Some additional sets (sa-, k-, ps-, and pa-sets) are defined and used in aeroelastic analysis.
The modal coordinate set is separated into zero frequency modes o and elastic (finite frequency)
modes f . For dynamic analysis by the modal method, the extra points u e are added to the modal
coordinate set i to form the h-set as shown in Figure A-1.
See the Degree-of-Freedom Set Definitions (Ch. 7) in the MD Nastran Quick Reference Guide for a
definition of supersets, combined sets and set names. The parameter PARAM,USETPRT can be used to
print of lists of degrees-of-freedom and the sets to which they belong. Different values of the PARAM
create various tables in the printed output.
Physical Set
e
Modal Sets h
o
i
f
Main Index
APPENDIX B 505
The Set Notation System Used in Dynamic Analysis
where:
Main Index
506 MD Nastran Dynamic Analysis User’s Guide
Overview
Main Index
MD Nastran Dynamic Analysis User’s Guide
Ap. C: References and Bibliography
Overview
General References
Bibliography
Main Index
508 MD Nastran Dynamic Analysis User’s Guide
Overview
Overview
This appendix includes references of interest in the field of dynamic analysis. Two categories are
included. The first category, General References, lists books that cover the general range of structural
dynamic analysis. The second category, Bibliography, is an excerpt from the dynamic analysis section
of the MSC.Nastran Bibliography, Second Edition.
Main Index
APPENDIX C 509
References and Bibliography
General References
1. Paz, M., Structural Dynamics: Theory and Computation, Van Nostrand Reinhold, New York,
N.Y., 1985.
2. Bathe, K. J. and Wilson, E. L., Numerical Methods in Finite Element Analysis, Prentice-Hall,
Englewood Cliffs, N.J., 1976.
3. Harris, C. M. and Crede, C. E., Shock and Vibration Handbook, McGraw-Hill, New York, N.Y.,
1976.
4. Clough, R. W. and Penzien, J., Dynamics of Structures, McGraw-Hill, New York, N.Y., 1975.
5. Timoshenko, S., Young, D. H., and Weaver Jr., W., Vibration Problems in Engineering, John
Wiley and Sons, New York, N.Y., 1974.
6. Hurty, W. C. and Rubinstein, M. F., Dynamics of Structures, Prentice-Hall, Englewood Cliffs,
N.J., 1964.
7. Miura, H., MSC/NASTRAN Handbook for Structural Optimization, The MacNeal-Schwendler
Corporation, 1988.
8. Moore, G.J. MSC/NASTRAN Design Sensitivity and Optimization User’s Guide, Version 67, The
MacNeal-Schwendler Corporation, 1992.
9. Patel, H., “Multidisciplinary Design with Superelements in MSC/NASTRAN,” AIAA
CP-92-4732, Fourth AIAA/USAF/NASA/OAI Symposium on Multidisciplinary Analysis and
Optimization, Cleveland, Ohio, September1992.
10. Johnson, E.H., and Reymond, M.A., “Multidisciplinary Aeroelastic Analysis and Design using
MSC/NASTRAN,” AIAA CP-91-1097, 32nd SDM Conference, Baltimore, MD, April 1991.
11. Kodiyalam, S., Vanderplaats, G., Miura, H., Nagendra, G. and Wallerstein, D. “Structural Shape
Optimization with MSC/NASTRAN,” AIAA CP 90-1007, 31st SDM Conference, Long Beach,
CA, April 1990.
12. Johnson, E.H., “Disjoint Design Spaces in the Optimization of Harmonically Excited Structures,”
AIAA Journal, Vol. 14, No. 2, February 1976, pp 259-261.
13. “Introduction to Version 67 of MSC/NASTRAN,” The MacNeal-Schwendler Corporation,
September 1991.
14. Cassis, J.H., “Optimum Design of Structures Subjected to Dynamic Loads,” PhD. Thesis
UCLA-ENG-7451, June 1974.
15. Adams, M. L., Padovan, J., Fertis, D. G., “Engine Dynamic Analysis With General Nonlinear
Finite-Element Codes, Part 1: Overall Approach and Development of Bearing Damper Element”,
ASME Journal of Engineering for Power, Vol. 104, July 1982, pp. 586-593.
16. Castelli, V., and Shapiro, W., “Improved Method for Numerical Solutions of the General
Incompressible Fluid Film Lubrication Problem”, ASME Journal of Lubrication Technology,
Vol. 89, No. 2, 1967, pp. 211-218.
17. Adams, M. L., Padovan, J., Fertis, D. G., “Finite Elements for Rotor/Stator Interactive Forces in
General Dynamic Simulation, Part 1: Development of Bearing Damper Element”, NASA CR-
165214, EDA 201-3A, October 1980.
Main Index
510 MD Nastran Dynamic Analysis User’s Guide
General References
18. Ghaby, R., “Transient/Nonlinear Vibration of Gas Turbine Engines With Squeeze Film Dampers
Due to Blade Loss”, May 1984 Master of Science Thesis, Case Western Reserve University.
19. Black, G., Gallardo, V., “Blade Loss Transient Dynamics Analysis Task II- TETRA 2 User’s
Manual”, NASA CR-179633, November 1986.
Main Index
APPENDIX C 511
References and Bibliography
Bibliography
DYNAMICS – GENERAL
• Abdallah, Ayman A.; Barnett, Alan R.; Widrick, Timothy W.; Manella, Richard T.; Miller,
Robert P. Stiffness-Generated Rigid-Body Mode Shapes for Lanczos Eigensolution with
Support DOF Via a MSC/NASTRAN DMAP Alter, MSC 1994 World Users’ Conf. Proc., Paper
No. 10, June, 1994.
• Anderson, William J.; Kim, Ki-Ook; Zhi, Bingchen; Bernitsas, Michael M.; Hoff, Curtis; Cho,
Kyu-Nam. Nonlinear Perturbation Methods in Dynamic Redesign, MSC/NASTRAN Users’
Conf. Proc., Paper No. 16, March, 1983.
• Barber, Pam; Arden, Kevin. Dynamic Design Analysis Method (DDAM) Using
MSC/NASTRAN, MSC 1994 World Users’ Conf. Proc., Paper No. 31, June, 1994.
• Bedrossian, Herand; Veikos, Nicholas. Rotor-Disk System Gyroscopic Effect in
MSC/NASTRAN Dynamic Solutions, MSC/NASTRAN Users’ Conf. Proc., Paper No. 12,
March, 1982.
• Bernstein, Murray; Mason, Philip W.; Zalesak, Joseph; Gregory, David J.; Levy, Alvin.
NASTRAN Analysis of the 1/8-Scale Space Shuttle Dynamic Model, NASTRAN: Users’
Exper., pp. 169-242, September, 1973, (NASA TM X-2893).
• Berthelon, T.; Capitaine, A. Improvements for Interpretation of Structural Dynamics
Calculation Using Effective Parameters for Substructures, Proc. of the 18th MSC Eur. Users’
Conf., Paper No. 9, June, 1991.
• Birkholz, E. Dynamic Investigation of Automobile Body Parts, Proc. of the 15th
MSC/NASTRAN Eur. Users’ Conf., October, 1988.
• Bishop, N. W. M.; Lack, L. W.; Li, T.; Kerr, S. C. Analytical Fatigue Life Assessment of
Vibration Induced Fatigue Damage, MSC 1995 World Users’ Conf. Proc., Paper No. 18, May,
1995.
• Blakely, Ken; Howard, G. E.; Walton, W. B.; Johnson, B. A.; Chitty, D. E. Pipe Damping
Studies and Nonlinear Pipe Benchmarks from Snapback Tests at the Heissdampfreaktor,
NUREG/CR-3180, March, 1983.
• Blakely, Ken. Dynamic Analysis: Application and Modeling Considerations, J. of
Engineering Computing and Applications, Fall, 1987.
• Bramante, A.; Paolozzi, A; Peroni, I. Effective Mass Sensitivity: A DMAP Procedure, MSC
1995 World Users’ Conf. Proc., Paper No. 39, May, 1995.
• Brutti, C.; Conte, M.; Linari, M. Reduction of Dynamic Environment to Equivalent Static
Loads by a NASTRAN DMAP Procedure, MSC 1995 European Users’ Conf. Proc., Italian
Session, September, 1995.
• Butler, Thomas G. Dynamic Structural Responses to Rigid Base Acceleration, Proc. of the
Conf. on Finite Element Methods and Technology, Paper No. 8, March, 1981.
Main Index
512 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Butler, Thomas G. Telescoping Robot Arms, MSC/NASTRAN Users’ Conf. Proc., Paper
No. 10, March, 1984.
• Butler, T. G. Experience with Free Bodies, Thirteenth NASTRAN Users’ Colloq., pp. 378-388,
May, 1985, (NASA CP-2373).
• Butler, Thomas G. Mass Modeling for Bars, Fifteenth NASTRAN Users’ Colloq., pp. 136-165,
August, 1987, (NASA CP-2481).
• Butler, T. G. Coupled Mass for Prismatical Bars, Sixteenth NASTRAN Users’ Colloq.,
pp. 44-63, April, 1988, (NASA CP-2505).
• Caldwell, Steve P.; Wang, B. P. An Improved Approximate Method for Computing
Eigenvector Derivatives in MSC/NASTRAN, The MSC 1992 World Users’ Conf. Proc., Vol. I,
Paper No. 22, May, 1992.
• Case, William R. Dynamic Substructure Analysis of the International Ultraviolet Explorer
(IUE) Spacecraft, NASTRAN: Users’ Exper., pp. 221-248, September, 1975, (NASA TM
X-3278).
• Chang, H. T.; Cao, Tim; Hua, Tuyen. SSF Flexible Multi-Body Control/Structure Interaction
Simulation, The MSC 1993 World Users’ Conf. Proc., Paper No. 15, May, 1993.
• Chang, W. M.; Lai, J. S.; Chyuan, S. W.; Application of the MSC/NASTRAN Design
Optimization Capability to Identify Joint Dynamic Properties of Structure, The Sixth Annual
MSC Taiwan Users’ Conf. Proc., Paper No. 1, November, 1994.
• Chargin, M.; Miura, H.; Clifford, Gregory A. Dynamic Response Optimization Using
MSC/NASTRAN, The MSC 1987 World Users Conf. Proc., Vol. I, Paper No. 14, March, 1987.
• Chen, J. T.; Chyuan, S. W.; You, D. W.; Wong, H. T. A New Method for Determining the Modal
Participation Factor in Support Motion Problems Using MSC/NASTRAN, The Seventh
Annual MSC/NASTRAN Users’ Conf. Proc., Taiwan, 1995.
• Chen, J. T.; Wong, H. T. Applications of Modal Reaction Method in Support Motion Problems,
Techniques in Civil Engineering, Vol. 4, pp 17 - 30, March, 1996, in Chinese.
• Chen, J. T.; Hong, H. K.; Chyuan, S. W.; Yeh, C. S. A Note on the Application of Large Mass
and Large Stiffness Techniques for Multi-Support Motion, The Fifth Annual MSC Taiwan
Users’ Conf. Proc., November, 1993.
• Chen, Yohchia. Improved Free-Field Analysis for Dynamic Medium-Structure Interaction
Problems, The MSC 1992 World Users’ Conf. Proc., Vol. I, Paper No. 13, May, 1992.
• Chen, Yohchia. Dynamic Response of Reinforced Concrete Box-Type Structures, The MSC
1992 World Users’ Conf. Proc., Vol. I, Paper No. 24, May, 1992.
• Chiu, Chi-Wai. Spacecraft Dynamics During Solar Array Panel Deployment Motion, The
Fifth Annual MSC Taiwan Users’ Conf. Proc., November, 1993.
• Chung, Y. T.; Kahre, L. L. A General Procedure for Finite Element Model Check and Model
Identification, MSC 1995 World Users’ Conf. Proc., Paper No. 38, May, 1995.
• Cicia, C. Static, Thermal and Dynamic Analysis of the Liquid Argon Cryostat for the ICARUS
Experiment, Proc. of the 15th MSC/NASTRAN Eur. Users’ Conf., October, 1988.
Main Index
APPENDIX C 513
References and Bibliography
• Cifuentes, Arturo O. Dynamic Analysis of Railway Bridges Using MSC/NASTRAN, The MSC
1988 World Users Conf. Proc., Vol. II, Paper No. 44, March, 1988.
• Cifuentes, A. O. Dynamic Response of a Beam Excited by a Moving Mass, Finite Elements in
Analysis and Design, Vol. 5, pp. 237-246, 1989.
• Citerley, R. L.; Woytowitz, P. J. Ritz Procedure for COSMIC/ NASTRAN, Thirteenth
NASTRAN Users’ Colloq., pp. 225-233, May, 1985, (NASA CP-2373).
• Ciuti, Gianluca. Avionic Equipment Dynamic Analysis, MSC 1995 European Users’ Conf.
Proc., Italian Session, September, 1995.
• Coates, Dr. Tim; Matthews, Peter. Transient Response in Dynamic and Thermal Behaviour,
The Second Australasian MSC Users Conf. Proc., Paper No. 4, November, 1988.
• Coppolino, Robert N.; Bella, David F. Employment of MSC/STI-VAMP for Dynamic
Response Post-Processing, The MSC 1987 World Users Conf. Proc., Vol. I, Paper No. 12,
March, 1987.
• Corder, P. R.; Persh, R. Castigliano and Symbolic Programming in Finite Element Analysis,
Proceedings of the 16th Annual Energy - Sources Technology Conference and Exhibition,
Houston, 1993.
• Coyette, J. P.; Wijker, J. J. The Combined Use of MSC/NASTRAN and Sysnoise for
Evaluating the Dynamic Behavior of Solar Array Panels, Proc. of the 20th MSC European
Users’ Conf., Paper No. 16, September, 1993.
• Curti, G.; Chiandussi, G.; Scarpa, F. Calculation of Eigenvalue Derivatives of
Acousto-Structural Systems with a Numerical Comparison, MSC 23rd European Users’ Conf.
Proc., Italian Session, September, 1996.
• Defosse, H.; Sergent, A. Vibro-Acoustic Modal Response Analysis of Aerospace Structures,
Proc. of the MSC/NASTRAN Eur. Users’ Conf., April, 1985.
• Deloo, Ph.; Dunne, L.; Klein, M. Alter DMAPS for the Generation, Assembly and Recovery of
Craig-Bampton Models in Dynamic Analyses, Actes de la 2ème Confèrence Française
Utilisateurs des Logiciels MSC, Toulouse, France, September, 1995.
• Denver, Richard E.; Menichello, Joseph M. Alternate Approaches to Vibration and Shock
Analysis Using NASTRAN, Sixth NASTRAN Users’ Colloq., pp. 199-212, October, 1977,
(NASA CP-2018).
• Detroux, P.; Geraets, L. H. Instability at Restart or Change of Time Step with NASTRAN in
the Presence of Nonlinear Loads, Proc. of the MSC/NASTRAN Eur. Users’ Conf., June, 1983.
• Deuermeyer, D. W.; Clifford, G. A.; Petesch, D. J. Traditional Finite Element Analysis:
Opportunities for Parallelism?, Computing Systems in Engineering, Vol. 2, No. 2-3,
pp. 157-165, 1991.
• Dirschmid, Dr. W.; Nolte, Dr. F.; Dunne, L. W. Application of an FRF-Based Update Method
to the Model Parameter Tuning of an Hydraulic Engine Mounting, Proc. of the 18th MSC
Eur. Users’ Conf., Paper No. 10, June, 1991.
Main Index
514 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Drago, Raymond J.; Margasahayam, Ravi N. Resonant Response of Helicopter Gears Using
3-D Finite Element Analysis, The MSC 1988 World Users Conf. Proc., Vol. I, Paper No. 20,
March, 1988.
• Elchuri, V.; Smith, G. C. C.; Gallo, A. Michael. An Alternative Method of Analysis for Base
Accelerated Dynamic Response in NASTRAN, Eleventh NASTRAN Users’ Colloq.,
pp. 89-112, May, 1983.
• Everstine, Gordon C.; Schroeder, Erwin A. The Dynamic Analysis of Submerged Structures,
NASTRAN: Users’ Exper., pp. 419-430, September, 1975, (NASA TM X-3278).
• Everstine, Gordon C. Structural Analogies for Scalar Field Problems, Int. J. for Numerical
Methods in Engineering, Vol. 17, No. 3, pp. 471-476, March, 1981.
• Everstine, G. C. Dynamic Analysis of Fluid-Filled Piping Systems Using Finite Element
Techniques, J. of Pressure Vessel Technology, Vol. 108, pp. 57-61, February, 1986.
• Flanigan, Christopher C. Accurate and Efficient Mode Acceleration Data Recovery for
Superelement Models, The MSC 1988 World Users Conf. Proc., Vol. I, Paper No. 38, March,
1988.
• Flanigan, Christopher C.; Manella, Richard T. Advanced Coupled Loads Analysis Using
MSC/NASTRAN, The MSC 1991 World Users’ Conf. Proc., Vol. I, Paper No. 14, March, 1991.
• Fox, Gary L. Solution of Enforced Boundary Motion in Direct Transient and Harmonic
Problems, Ninth NASTRAN Users’ Colloq., pp. 96-105, October, 1980, (NASA CP-2151).
• Geyer, A.; Schweiger, W. Multiple Support Excitation for NASTRAN Piping Analysis, Proc.
of the MSC/NASTRAN Eur. Users’ Conf., April, 1982.
• Geyer, A.; Schweiger, W. Aeroelastic and Stress Analysis of the CHIWEC Chinese Wind
Energy Converter Using MSC/NASTRAN, Proc. of the MSC/NASTRAN Eur. Users’ Conf.,
April, 1985.
• Ghofranian, S.; Dimmagio, O. D. Space Station Dynamic Analysis with Active Control
Systems Using MSC/NASTRAN, The MSC 1988 World Users Conf. Proc., Vol. I, Paper No. 17,
March, 1988.
• Gibson, Warren C. Experiences with Optimization Using ASD/NASOPT and MSC/NASTRAN
for Structural Dynamics, The MSC 1987 World Users Conf. Proc., Vol. I, Paper No. 13, March,
1987.
• Gibson, Warren C.; Austin, Eric. Analysis and Design of Damped Structures Using
MSC/NASTRAN, The MSC 1992 World Users’ Conf. Proc., Vol. I, Paper No. 25, May, 1992.
• Gielen, L.; Brughmans, M.; Petellat, C. A Stepwise Approach for Fatigue Evaluation of
Engine Accessories Prior to Prototyping Using Hybrid Modelling Technology, MSC 1996
World Users’ Conf. Proc., Vol. III, Paper No. 29, June, 1996.
• Go, James Chi-Dian. Structural Dynamic and Thermal Stress Analysis of Nuclear Reactor
Vessel Support System, NASTRAN: Users’ Exper., pp. 465-476, September, 1972, (NASA TM
X-2637).
• Grasso, A.; Tomaselli, L. Whirling Speed Analysis of Multispool Systems, Proc. of the
MSC/NASTRAN Eur. Users’ Conf., May, 1984.
Main Index
APPENDIX C 515
References and Bibliography
• Grimes, Roger G.; Lewis, John G.; Simon, Horst D.; Komzsik, Louis; Scott, David S. Shifted
Block Lanczos Algorithm in MSC/NASTRAN, MSC/NASTRAN Users’ Conf. Proc., Paper
No. 12, March, 1985.
• Harn, Wen-Ren; Lin, Shyang-Kuang; Chen, Jeng-Tzong. Localization of Dynamic Model
Modification Based on Constrained Minimization Method, The 2nd Annual MSC Taiwan
Users Conf., Paper No. 14, October, 1990.
• Herting, David N.; Bella, David F.; Kimbrough, Patty A. Finite Element Simulation of
Coupled Automobile Engine Dynamics, The MSC 1987 World Users Conf. Proc., Vol. I, Paper
No. 10, March, 1987.
• High, Gerald D. An Iterative Method for Eigenvector Derivatives, The MSC 1990 World Users
Conf. Proc., Vol. I, Paper No. 17, March, 1990.
• Hill, R. G. Transient Analysis of an IVHM Grapple Impact Test, NASTRAN: Users’ Exper.,
pp. 161-178, September, 1972, (NASA TM X-2637).
• Howells, R. W.; Sciarra, J. J. Finite Element Analysis Using NASTRAN Applied to Helicopter
Transmission Vibration/Noise Reduction, NASTRAN: Users’ Exper., pp. 321-340, September,
1975, (NASA TM X-3278).
• Howlett, James T. Applications of NASTRAN to Coupled Structural and Hydrodynamic
Responses in Aircraft Hydraulic Systems, NASTRAN: Users’ Exper., pp. 407-420, September,
1971, (NASA TM X-2378).
• Huang, S. L.; Rubin, H. Static and Dynamic Analysis, F-14A Boron Horizontal Stabilizer,
NASTRAN: Users’ Exper., pp. 251-264, September, 1971, (NASA TM X-2378).
• Hurwitz, Myles M. New Large Deflection Analysis for NASTRAN, Sixth NASTRAN Users’
Colloq., pp. 235-256, October, 1977, (NASA CP-2018).
• Hussain, M. A.; Pu, S. L.; Lorensen, W. E. Singular Plastic Element: NASTRAN
Implementation and Application, Sixth NASTRAN Users’ Colloq., pp. 257-274, October,
1977, (NASA CP-2018).
• Ishikawa, Masanori; Iwahara, Mitsuo; Nagamatsu, Akio. Dynamic Optimization Applied to
Engine Structure, The MSC 1990 World Users Conf. Proc., Vol. I, Paper No. 31, March, 1990.
• Iwahara, Mitsuo. Dynamic Optimization Using Quasi Least Square Method, The Fifth
MSC/NASTRAN User’s Conf. in Japan, October, 1987, in Japanese.
• Jakovich, John; Van Benschoten, John. SDRC SUPERTAB Interactive Graphics as a
Front-End to MSC/NASTRAN Dynamic Analysis, Proc. of the MSC/NASTRAN Users’ Conf.,
March, 1979.
• Jones, Gary K. The Response of Shells to Distributed Random Loads Using NASTRAN,
NASTRAN: Users’ Exper., pp. 393-406, September, 1971, (NASA TM X-2378).
• Kalinowski, Anthony J. Steady Solutions to Dynamically Loaded Periodic Structures, Eighth
NASTRAN Users’ Colloq., pp. 131-164, October, 1979, (NASA CP-2131).
• Kalinowski, A. J. Solution Sensitivity and Accuracy Study of NASTRAN for Large Dynamic
Problems Involving Structural Damping, Ninth NASTRAN Users’ Colloq., pp. 49-62,
October, 1980, (NASA CP-2151).
Main Index
516 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Kasai, Manabu. Real Eigenvalue Analysis by Modal Synthesis Method Taking Differential
Stiffness into Account, The First MSC/NASTRAN User’s Conf. in Japan, October, 1983, in
Japanese.
• Kasai, Manabu. Recovery Method for Components by DMAP of Constrained Modal Type, The
Fifth MSC/NASTRAN User’s Conf. in Japan, October, 1987, in Japanese.
• Kasai, Manabu. DMAP Program for Modal Mass and Momentum, The Sixth
MSC/NASTRAN User’s Conf. in Japan, October, 1988, in Japanese.
• Kienholz, Dave K.; Johnson, Conor D.; Parekh, Jatin C. Design Methods for Viscoelastically
Damped Sandwich Plates, AIAA/ASME/ASCE/AHS 24th Structures, Structural Dynamics and
Materials Conf., Part 2, pp. 334-343, May, 1983.
• Lambert, Nancy; Tucchio, Michael. Ring Element Dynamic Stresses, Ninth NASTRAN Users’
Colloq., pp. 63-78, October, 1980, (NASA CP-2151).
• Lee, Jyh-Chiang. Investigation for the Large Stiffness Method, The Fifth Annual MSC Taiwan
Users’ Conf. Proc., November, 1993.
• Lee, Ting-Yuan; Lee, Jyh-Chiang. Modal Analysis and Structural Modification for a Harpoon
Launcher, The Fifth Annual MSC Taiwan Users’ Conf. Proc., November, 1993.
• LeMaster, R. A.; Runyan, R. B. Dynamic Certification of a Thrust-Measuring System for
Large Solid Rocket Motors, Eleventh NASTRAN Users’ Colloq., pp. 207-225, May, 1983.
• Lewis, J.; Komzsik, L. Symmetric Generalized Eigenproblems in Structural Engineering,
SIAM Conf. on Applied Numerical Analysis, 1985.
• Lin, Chih-Kai; Harn, Wen-Ren; Lin, Shyang-Kuang. The Dynamic Response of Bridge Due to
Passing of Vehicle, The 2nd Annual MSC Taiwan Users Conf., Paper No. 6, October, 1990.
• Lin, S. L.; Yang, T. W.; Chen, J. T. MSC/NASTRAN Application in Inertia Relief, The 1st
MSC Taiwan Users’ Conf., Paper No. 13, October, 1989, in Chinese.
• Lin, Shan. Time-Dependent Restrained Boundary Condition Simulation, The MSC 1988
World Users Conf. Proc., Vol. I, Paper No. 9, March, 1988.
• Lu, Ming-Ying; Yang, Joe-Ming. Analysis of Static and Dynamic Responses on Ship
Structures Under Wave Loadings, The Sixth Annual MSC Taiwan Users’ Conf. Proc., Paper
No. 14, November, 1994.
• Magari, P. J.; Shultz, L. A.; Murthy, V. R. Dynamics of Helicopter Rotor Blades, Computers
and Structures, Vol. 29, No. 5, pp. 763-776, 1988.
• Malcolm, D. J. Dynamic Response of a Darrieus Rotor Wind Turbine Subject to Turbulent
Flow, Engineering Structures, Vol. 10, No. 2, pp. 125-134, April, 1988.
• Maritan, M.; Micelli, D. Dynamic Behaviour of a High-Speed Crankshaft, MSC 23rd
European Users’ Conf. Proc., Italian Session, September, 1996.
• Masters, Steven G. Plant Troubleshooting with MSC/NASTRAN, Proc. of the Conf. on Finite
Element Methods and Technology, Paper No. 12, March, 1981.
• Mastrorocco, David T. Predicting Dynamic Environments for Space Structure Appendages,
The MSC 1992 World Users’ Conf. Proc., Vol. II, Paper No. 56, May, 1992.
Main Index
APPENDIX C 517
References and Bibliography
• Mayer, Lee S.; Zeischka, Johann; Scherens, Marc; Maessen, Frank. Analysis of Flexible
Rotating Crankshaft with Flexible Engine Block Using MSC/NASTRAN and DADS, MSC
1995 World Users’ Conf. Proc., Paper No. 35, May, 1995.
• McLaughlin, A. Finite Element Dynamic Analysis of Production Aircraft, 4th Eur. Rotorcraft
and Powered Lift Aircraft Forum, Assoc. Ital di Aeronaut ed Astronaut, pp. 20.1-20.7,
September, 1978.
• Melli, R.; Rispoli, F.; Sciubba, E.; Tavani, F. Structural and Thermal Analysis of Avionic
Instruments for an Advanced Concept Helicopter, Proc. of the 15th MSC/NASTRAN Eur.
Users’ Conf., October, 1988.
• Mikami, Kouichi. Dynamic Stress Analysis System for Ship’s Hull Structure Under Wave
Loads, The Second MSC/NASTRAN User’s Conf. in Japan, October, 1984, in Japanese.
• Moharir, M. M. NASTRAN Nonlinear Capabilities in Dynamic Solutions, MSC/NASTRAN
Users’ Conf. Proc., Paper No. 9, March, 1985.
• Moore, Gregory J.; Nagendra, Gopal K. Dynamic Response Sensitivities in MSC/NASTRAN,
The MSC 1991 World Users’ Conf. Proc., Vol. I, Paper No. 4, March, 1991.
• Mulcahy, T. M.; Turula, P.; Chung, H.; Jendrzejczyk, A. Analytical and Experimental Study of
Two Concentric Cylinders Coupled by a Fluid Gap, NASTRAN: Users’ Exper., pp. 249-258,
September, 1975, (NASA TM X-3278).
• Murthy, P. L. N.; Chamis, C. C. Dynamic Stress Analysis of Smooth and Notched Fiber
Composite Flexural Specimens, National Aeronautics and Space Administration, April, 1984,
(NASA TM-83694).
• Murthy, P. L. N.; Chamis, C. C. Dynamic Stress Analysis of Smooth and Notched Fiber
Composite Flexural Specimens, Composite Materials: Testing and Design (Seventh Conf.),
ASTM, pp. 368-391, 1986, (ASTM STP 893).
• Neal, M. Vibration Analysis of a Printed Wiring Board Assembly, Proc. of the
MSC/NASTRAN Eur. Users’ Conf., May, 1984.
• Nefske, D. J.; Sung, S. H. Power Flow Finite Element Analysis of Dynamic Systems: Basic
Theory and Application to Beams, American Soc. of Mechanical Engineers, Noise Control and
Acoustics Division, Vol. 3, pp. 47-54, December, 1987.
• Nowak, Bill. The Analysis of Structural Dynamic Effects on Image Motion in Laser Printers
Using MSC/NASTRAN, The MSC 1988 World Users Conf. Proc., Vol. I, Paper No. 10, March,
1988.
• Nowak, Bill. Structural Dynamics Analysis of Laser Printers, Sound and Vibration, Vol. 23,
No. 1, pp. 22-26, January, 1989.
• Nowak, William J. Dynamic Analysis of Optical Scan Systems Using MSC/NASTRAN, Proc.
of the Conf. on Finite Element Methods and Technology, Paper No. 10, March, 1981.
• Nowak, William; James, Courtney. Dynamic Modeling and Analysis of Spinning Polygon
Assemblies Using MSC/NASTRAN, The MSC 1993 World Users’ Conf. Proc., Paper No. 66,
May, 1993.
Main Index
518 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Oei, T. H.; Broerse, G. Reduction of Forced Vibration Levels on Ro-Ro Car Ferry-Type Ships
by Means of Minor Changes of the Inner Aft-Body Construction, Proc. of the
MSC/NASTRAN Eur. Users’ Conf., April, 1982.
• Ojalvo, I. U. Extensions of MSC/NASTRAN to Solve Flexible Rotor Problems,
MSC/NASTRAN Users’ Conf. Proc., Paper No. 13, March, 1982.
• Palmieri, F. Nonlinear Dynamic Analysis of STS Main Engine Heat Exchanger, Proc. of the
15th MSC/NASTRAN Eur. Users’ Conf., October, 1988.
• Palmieri, F. W. Analyzing Deployment of Spacecraft Appendages Using MSC/NASTRAN, The
MSC 1990 World Users Conf. Proc., Vol. I, Paper No. 6, March, 1990.
• Pamidi, P. R.; Brown, W. K. On Eigenvectors of Multiple Eigenvalues Obtained in
NASTRAN, NASTRAN: Users’ Exper., pp. 285-300, September, 1975, (NASA TM X-3278).
• Paolozzi, A. Structural Dynamics Modification with MSC/NASTRAN, Proc. of the 19th MSC
European Users’ Conf., Paper No. 14, September, 1992.
• Parthasarathy, Alwar. Force-Sum Method for Dynamic Stresses in MSC/NASTRAN
Aeroelastic Analysis, The MSC 1991 World Users’ Conf. Proc., Vol. I, Paper No. 8, March,
1991.
• Parthasarathy, Alwar; Elzeki, Mohamed; Abramovici, Vivianne. PSDTOOL-A DMAP
Enhancement to Harmonic/Random Response Analysis in MSC/NASTRAN, The MSC 1993
World Users’ Conf. Proc., Paper No. 36, May, 1993.
• Patel, Jayant S.; Seltzer, S. M. Complex Eigenvalue Solution to a Spinning Skylab Problem,
NASTRAN: Users’ Exper., pp. 439-450, September, 1971, (NASA TM X-2378).
• Patel, Jayant S.; Seltzer, S. M. Complex Eigenvalue Analysis of Rotating Structures,
NASTRAN: Users’ Exper., pp. 197-234, September, 1972, (NASA TM X-2637).
• Patel, Kirit V. Stress Analysis of Hybrid Pins in a Warped Printed Wiring Board Using
MSC/NASTRAN, MSC 1995 World Users’ Conf. Proc., Paper No. 20, May, 1995.
• Paxson, Ernest B., Jr. Simulation of Small Structures- Optics-Controls Systems with
MSC/NASTRAN, The 1989 MSC World Users Conf. Proc., Vol. II, Paper No. 39, March, 1989.
• Pinnament, Murthy. Mode Acceleration Data Recovery in MSC/NASTRAN Dynamic Analysis
with Generalized Dynamic Reduction, MSC/NASTRAN Users’ Conf. Proc., Paper No. 24,
March, 1985.
• Raney, John P.; Kaszubowski, M.; Ayers, J. Kirk. Analysis of Space Station Dynamics Using
MSC/NASTRAN, The MSC 1987 World Users Conf. Proc., Vol. I, Paper No. 11, March, 1987.
• Reyer, H. A Crash-Down Calculated with NASTRAN, Proc. of the MSC/NASTRAN Eur.
Users’ Conf., April, 1982.
• Rose, Ted L. Using Superelements to Identify the Dynamic Properties of a Structure, The
MSC 1988 World Users Conf. Proc., Vol. I, Paper No. 41, March, 1988.
• Rose, Ted L. Creation of and Use of ‘Craig-Bampton’ Models Using MSC/NASTRAN, The
MSC 1990 World Users Conf. Proc., Vol. II, Paper No. 51, March, 1990.
• Rose, Ted. Using Residual Vectors in MSC/NASTRAN Dynamic Analysis to Improve
Accuracy, The MSC 1991 World Users’ Conf. Proc., Vol. I, Paper No. 12, March, 1991.
Main Index
APPENDIX C 519
References and Bibliography
• Rose, Ted. DMAP Alters to Apply Modal Damping and Obtain Dynamic Loading Output for
Superelements, The MSC 1993 World Users’ Conf. Proc., Paper No. 24, May, 1993.
• Rose, Ted.; McNamee, Martin. A DMAP Alter to Allow Amplitude-Dependent Modal
Damping in a Transient Solution, MSC 1996 World Users’ Conf. Proc., Vol. V, Paper No. 50,
June, 1996.
• Ross, Robert W. Prediction and Elimination of Resonance in Structural Steel Frames, The
MSC 1988 World Users Conf. Proc., Vol. II, Paper No. 45, March, 1988.
• Russo, A.; Mocchetti, R. Dynamic Analysis of Loaded Structures in the Helicopter Field,
Proc. of the MSC/NASTRAN Eur. Users’ Conf., May, 1984.
• Salus, W. L.; Jones, R. E.; Ice, M. W. Dynamic Analysis of a Long Span, Cable-Stayed
Freeway Bridge Using NASTRAN, NASTRAN: Users’ Exper., pp. 143-168, September, 1973,
(NASA TM X-2893).
• Sauer, G.; Wolf, M. Gyroscopic Effects in the Dynamic Response of Rotating Structures, Proc.
of the MSC/NASTRAN Eur. Users’ Conf., Paper No. 11, May, 1986.
• Schips, C. Aero-Engine Turbine Dynamic Analysis, Proc. of the 18th MSC Eur. Users’ Conf.,
Paper No. 8, June, 1991.
• Schmitz, Ronald P. Structural Dynamic Analysis of Electronic Assemblies Using NASTRAN
Restart/Format Change Capability, NASTRAN: Users’ Exper., pp. 363-392, September, 1971,
(NASA TM X-2378).
• Schweiger, W.; de Bruyne, F.; Dirschmid, W. Fluid Structure Interaction of Car Fuel Tanks,
Proc. of the MSC/NASTRAN Eur. Users’ Conf., May, 1984.
• Shein, Shya-Ling. Generation of the Space Station Freedom On-Orbit Dynamic Loads
Analysis Model Using MSC/NASTRAN V66A Superelements, The 2nd Annual MSC Taiwan
Users Conf., Paper No. 7, October, 1990.
• Shiraki, K.; Hashimoto, H.; Sato, N.; Nasu, S.; Kinno, M. Japanese Experiment Module
(JEM): On-Orbit Structural Dynamic Analysis, 1993 MSC Japan’s 11th User’s Conf. Proc.,
Paper No. 10.
• Shivaji, M.; Raju, V. S. N. Dynamic Analysis of R. C. C. Chimneys, MSC 1995 World Users’
Conf. Proc., Paper No. 34, May, 1995.
• Singh, Ashok K.; Nichols, Christian W. Derivation of an Equivalent Beam Model From a
Structural Finite Element Model, The MSC 1988 World Users Conf. Proc., Vol. I, Paper
No. 14, March, 1988.
• Singh, Sudeep K.; Engelhardt, Charlie. Dynamic Analysis of a Large Space Structure Using
External and Internal Superelements, The MSC 1991 World Users’ Conf. Proc., Vol. I, Paper
No. 27, March, 1991.
• Skattum, Knut S. Modeling Techniques of Thin-Walled Beams with Open Cross Sections,
NASTRAN: Users’ Exper., pp. 179-196, September, 1972, (NASA TM X-2637).
Main Index
520 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Smith, Michael R.; Rangacharyulu, M.; Wang, Bo P.; Chang, Y. K. Application of Optimization
Techniques to Helicopter Structural Dynamics, AIAA/ASME/ASCE/AHS/ASC 32nd
Structures, Structural Dynamics, and Materials Conf., Part 1, Paper No. 91-0924, pp. 227-237,
April, 1991.
• Stockwell, Alan E.; Perez, Sharon E.; Pappa, Richard S. Integrated Modeling and Analysis of a
Space-Truss Test Article, The MSC 1990 World Users Conf. Proc., Vol. I, Paper No. 16, March,
1990.
• Struschka, M.; Goldstein, H. Approximation of Frequency Dependant Nonlinearities in
Linear FE-Models, Proc. of the 15th MSC/NASTRAN Eur. Users’ Conf., October, 1988.
• Subrahmanyam, K. B.; Kaza, K. R. V.; Brown, G. V.; Lawrence, C. Nonlinear
Bending-Torsional Vibration and Stability of Rotating, Pre-Twisted, Preconed Blades
Including Coriolis Effects, National Aeronautics and Space Administration, January, 1986,
(NASA TM-87207).
• Tecco, T. C. Analyzing Frequency Dependent Stiffness and Damping with MSC/NASTRAN,
MSC/NASTRAN Users’ Conf. Proc., Paper No. 25, March, 1985.
• Thornton, E. A. Application of NASTRAN to a Space Shuttle Dynamics Model, NASTRAN:
Users’ Exper., pp. 421-438, September, 1971, (NASA TM X-2378).
• Ting, Tienko. Test/Analysis Correlation for Multiple Configurations, The MSC 1993 World
Users’ Conf. Proc., Paper No. 74, May, 1993.
• Tinti, Francesco Carlo; Scaffidi, Costantino. Structural Dynamics and Acoustic Design of
Engine Component in View of Exterior Noise Reduction Using Numerical Techniques, MSC
1995 European Users’ Conf. Proc., Italian Session, September, 1995.
• Turner, Patrick Ryan. Integrating Finite Element Analysis with Quasi-Static Loadings from a
Large Displacement Dynamic Analysis, The 1989 MSC World Users Conf. Proc., Vol. II, Paper
No. 37, March, 1989.
• Tzong, George T. J.; Sikes, Gregory D.; Dodd, Alan J. Large Order Modal Analysis Module in
the Aeroelastic Design Optimization Program (ADOP), The MSC 1991 World Users’ Conf.
Proc., Vol. II, Paper No. 36, March, 1991.
• Unger, B.; Eichlseder, Wilfried; Schuch, F. Predicting the Lifetime of Dynamically Stressed
Components, Proc. of the 20th MSC European Users’ Conf., Paper No. 36, September, 1993.
• Vance, Judy; Bernard, James E. Approximating Eigenvectors and Eigenvalues Across a Wide
Range of Design, The MSC 1992 World Users’ Conf. Proc., Vol. II, Paper No. 46, May, 1992.
• Visintainer, Randal H.; Aslani, Farhang. Shake Test Simulation Using MSC/NASTRAN, MSC
1994 World Users’ Conf. Proc., Paper No. 32, June, 1994.
• Vitiello, P.; Quaranta, V. SEA Investigation Via a FEM Based Substructuring Technique,
MSC 1995 European Users’ Conf. Proc., Italian Session, September, 1995.
• Walton, William B.; Blakely, Ken. Modeling of Nonlinear Elastic Structures Using
MSC/NASTRAN, MSC/NASTRAN Users’ Conf. Proc., Paper No. 11, March, 1983.
• Wamsler, M.; Blanck, N.; Kern, G. On the Enforced Relative Motion Inside a Structure, Proc.
of the 20th MSC European Users’ Conf., September, 1993.
Main Index
APPENDIX C 521
References and Bibliography
Main Index
522 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Blakely, Ken. Revising MSC/NASTRAN Models to Match Test Data, Proc. of the 9th Int.
Modal Analysis Conf., April, 1991.
• Blakely, Ken. Get the Model Right, then Run the Analysis, Machine Design, October 24, 1991.
• Blakely, Ken; Rose, Ted. Cross-Orthogonality Calculations for Pre-Test Planning and Model
Verification, The MSC 1993 World Users’ Conf. Proc., Paper No. 72, May, 1993.
• Blakely, Ken; Bush, Richard. Using MSC/NASTRAN to Match Dynamic Test Data, Proc. of
the Int. Conf. on Structural Dynamics Modelling, July, 1993.
• Blakely, Ken; Rose, Ted. Cross-Orthogonality Calculations for Pre-Test Planning and Model
Verification, Proc. of the 20th MSC European Users’ Conf., September, 1993.
• Blakely, Ken. Matching Frequency Response Test Data with MSC/NASTRAN, MSC 1994
World Users’ Conf. Proc., Paper No. 17, June 1994.
• Blakely, Ken. Matching Frequency Response Test Data with MSC/NASTRAN, Proc. of the
21st MSC European User’s Conf., Italian Session, September, 1994.
• Brillhart, Ralph; Hunt, David L.; Kammer, Daniel C.; Jensen, Brent M.; Mason, Donald R.
Modal Survey and Test-Analysis Correlation of the Space Shuttle SRM, Proc. of the 6th Int.
Modal Analysis Conf., pp. 863-870, February, 1988.
• Brughmans, Marc; Leuridan, Jan; Blauwkamp, Kevin. The Application of FEM-EMA
Correlation and Validation Techniques on a Body-in-White, The MSC 1993 World Users’
Conf. Proc., Paper No. 6, May, 1993.
• Brughmans, M.; Lembregts, PhD. F.; Furini, PhD. F.; Storrer, O. Modal Test on the Pininfarina
Concept Car Body “ETHOS 1", Actes de la 2ème Confèrence Française Utilisateurs des
Logiciels MSC, Toulouse, France, September, 1995.
• Brughmans, M.; Lembregts, F, Ph.D.; Furini, F., Ph.D. Modal Test on the Pininfarina Concept
Car Body “ETHOS 1", MSC 1995 World Users’ Conf. Proc., Paper No. 5, May, 1995.
• Budynas, R.; Kolhatkar, S. Modal Analysis of a Robot Arm Using Finite Element Analysis and
Modal Testing, Proc. of the 8th Int. Modal Analysis Conf., Vol. I, pp. 67-70, January, 1990.
• Budynas, R. G.; Krebs, D. Modal Correlation of Test and Finite Element Results Using Cross
Orthogonality with a Reduced Mass Matrix Obtained by Modal Reduction and NASTRAN’s
Generalized Dynamic Reduction Solution, Proc. of the 9th Int. Modal Analysis Conf., Vol. I,
pp. 549-554, April, 1991.
• Butler, Thomas G. Test vs. Analysis: A Discussion of Methods, Fourteenth NASTRAN Users’
Colloq., pp. 173-186, May, 1986, (NASA CP-2419).
• Call, V.; Mason, D. Space Shuttle Redesigned Solid Rocket Booster Structural Dynamic
Predictions and Correlations of Liftoff, AIAA/SAE/ASME/ASEE 26th Joint Propulsion Conf.,
Paper No. AIAA 90-2081, July, 1990.
• Chung, Y. T. Model Reduction and Model Correlation Using MSC/NASTRAN, MSC 1995
World Users’ Conf. Proc., Paper No. 8, May, 1995.
• Coladonato, Robert J. Development of Structural Dynamic Test Evnironments for Subsystems
and Components, Seventh NASTRAN Users’ Colloq., pp. 85-110, October, 1978, (NASA
CP-2062).
Main Index
APPENDIX C 523
References and Bibliography
Main Index
524 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Kammer, Daniel C.; Jensen, Brent M.; Mason, Donald R. Test-Analysis Correlation of the
Space Shuttle Solid Rocket Motor Center Segment, J. of Spacecraft, Vol. 26, No. 4,
pp. 266-273, March, 1988.
• Kelley, William R.; Isley, L. D. Using MSC/NASTRAN for the Correlation of Experimental
Modal Models for Automotive Powertrain Structures, The MSC 1993 World Users’ Conf.
Proc., Paper No. 8, May, 1993.
• Kelley, William R.; Isley, L. Dean; Foster, Thomas J. Dynamic Correlation Study Transfer
Case Housings, MSC 1996 World Users’ Conf. Proc., Vol. II, Paper No. 15, June, 1996.
• Kientzy, Donald; Richardson, Mark; Blakely, Ken. Using Finite Element Data to Set Up Modal
Tests, Sound and Vibration, June, 1989.
• Lammens, Stefan; Brughmans, Marc; Leuridan, Jan; Sas, Paul. Application of a FRF Based
Model Updating Technique for the Validation of Finite Element MOdels of Components of the
Automotive Industry, MSC 1995 World Users’ Conf. Proc., Paper No. 7, May, 1995.
• Lee, John M.; Parker, Grant R. Application of Design Sensitivity Analysis to Improve
Correlations Between Analytical and Test Modes, The 1989 MSC World Users Conf. Proc.,
Vol. I, Paper No. 21, March, 1989.
• Linari, M.; Mancino, E. Application of the MSC/NASTRAN Program to the Study of a Simple
Reinforced Concrete Structure in Nonlinear Material Field, Proc. of the 15th
MSC/NASTRAN Eur. Users’ Conf., October, 1988.
• Lowrey, Richard D. Calculating Final Mesh Size Before Mesh Completion, The MSC 1990
World Users Conf. Proc., Vol. II, Paper No. 44, March, 1990.
• Marlow, Jill M.; Lindell, Michael C. NASSTAR: An Instructional Link Between
MSC/NASTRAN and STAR, Proceedings of the 11th International Modal Analysis Conference,
Florida, 1993.
• Masse, Barnard; Pastorel, Henri. Stress Calculation for the Sandia 34-Meter Wind Turbine
Using the Local Circulation Method and Turbulent Wind, The MSC 1991 World Users’ Conf.
Proc., Vol. II, Paper No. 53, March, 1991.
• Mindle, Wayne L.; Torvik, Peter J. A Comparison of NASTRAN (COSMIC) and Experimental
Results for the Vibration of Thick Open Cylindrical Cantilevered Shells, Fourteenth
NASTRAN Users’ Colloq., pp. 187-204, May, 1986, (NASA CP-2419).
• Morton, Mark H. Application of MSC/NASTRAN for Assurance of Flight Safety and Mission
Effectiveness with Regard to Vibration upon Installation of the Stinger Missile on the
AH-64A, The MSC 1991 World Users’ Conf. Proc., Vol. II, Paper No. 52, March, 1991.
• Neads, M. A.; Eustace, K. I. The Solution of Complex Structural Systems by NASTRAN
within the Building Block Approach, NASTRAN User’s Conf., May, 1979.
• Nowak, William. Electro-Mechanical Response Simulation of Electrostatic Voltmeters Using
MSC/NASTRAN, The MSC 1993 World Users’ Conf. Proc., Paper No. 65, May, 1993.
• O’Callahan, Dr. John; Avitabile, Peter; Reimer, Robert. An Application of New Techniques for
Integrating Analytical and Experimental Structural Dynamic Models, The 1989 MSC World
Users Conf. Proc., Vol. II, Paper No. 47, March, 1989.
Main Index
APPENDIX C 525
References and Bibliography
• Ott, Walter; Kaiser, Hans-Jurgen; Meyer, Jurgen. Finite Element Analysis of the Dynamic
Behaviour of an Engine Block and Comparison with Experimental Modal Test Results, The
MSC 1990 World Users Conf. Proc., Vol. I, Paper No. 14, March, 1990.
• Paolozzi, A. Structural Dynamics Modification with MSC/NASTRAN, Proc. of the 19th MSC
European Users’ Conf., Paper No. 14, September, 1992.
• Park, H. B.; Suh, J. K.; Cho, H. G.; Jung, G. S. A Study on Idle Vibration Analysis Technique
Using Total Vehicle Model, MSC 1995 World Users’ Conf. Proc., Paper No. 6, May, 1995.
• Parker, Grant R.; Rose, Ted L.; Brown, John J. Kinetic Energy Calculation as an Aid to
Instrumentation Location in Modal Testing, The MSC 1990 World Users Conf. Proc., Vol. II,
Paper No. 47, March, 1990.
• Preve, A.; Meneguzzo, M.; Merlo, A.; Zimmer, H. Simulation of Vehicles’ Structural Noise:
Numerical/Experimental Correlation in the Acoustic Simulation of the Internal Noise, Proc.
of the 21st MSC European Users’ Conf., Italian Session, September, 1994.
• Rabani, Hadi. Static and Dynamic FEM/Test Correlation of an Automobile Body, The 1989
MSC World Users Conf. Proc., Vol. II, Paper No. 49, March, 1989.
• Rainer, I. G. MSC/NASTRAN as a Key Tool to Satisfy Increasing Demand for Numerical
Simulation Techniques, Proc. of the 20th MSC European Users’ Conf., September, 1993.
• Ray, William F. The Use of MSC/NASTRAN and Empirical Data to Verify a Design,
MSC/NASTRAN Users’ Conf. Proc., Paper No. 11, March, 1984.
• Scapinello, F.; Colombo, E. An Approach for Detailed Analysis of Complex Structures
Avoiding Complete Models, Proc. of the MSC/NASTRAN Eur. Users’ Conf., May, 1987.
• Sok-chu, Park; Ishii, Tetsu; Honda, Shigeki; Nagamatsu, Akio. Vibration Analysis and
Optimum Design of Press Machines, 1994 MSC Japan Users’ Conf. Proc.
• Stack, Charles P.; Cunningham, Timothy J. Design and Analysis of Coriolis Mass Flowmeters
Using MSC/NASTRAN, The MSC 1993 World Users’ Conf. Proc., Paper No. 54, May, 1993.
• Su, Hong. Structural Analysis of Ka-BAND Gimbaled Antennas for a Communications
Satellite System, MSC 1996 World Users’ Conf. Proc., Vol. IV, Paper No. 33, June, 1996.
• Tawekal, Ricky; Budiyanto, M. Agus. Finite Element Model Correlation for Structures, The
MSC 1993 World Users’ Conf. Proc., Paper No. 73, May, 1993.
• Ting, T.; Ojalvo, I. U. Dynamic Structural Correlation via Nonlinear Programming
Techniques, The MSC 1988 World Users Conf. Proc., Vol. II, Paper No. 57, March, 1988.
• Ting, Tienko; Chen, Timothy L. C. FE Model Refinement with Actual Forced Responses of
Aerospace Structures, The MSC 1991 World Users’ Conf. Proc., Vol. II, Paper No. 51, March,
1991.
• Ting, Tienko. Test/Analysis Correlation for Multiple Configurations, The MSC 1993 World
Users’ Conf. Proc., Paper No. 74, May, 1993.
• Ujihara, B. H.; Dosoky, M. M.; Tong, E. T. Improving a NASTRAN Dynamic Model with Test
Data Using Linwood, Tenth NASTRAN Users’ Colloq., pp. 74-86, May, 1982, (NASA
CP-2249).
Main Index
526 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Zeischka, H. LMS/Link Correlating and Validating F.E.A. for Dynamic Structure Behaviour
with Experimental Modal Analysis, Proc. of the MSC/NASTRAN Eur. Users’ Conf., May,
1987.
Main Index
APPENDIX C 527
References and Bibliography
• Ghosh, Tarun. MSC/NASTRAN Based Component Mode Synthesis Analysis Without the Use
of DMAPS, MSC 1996 World Users’ Conf. Proc., Vol. II, Paper No. 18, June, 1996.
• Gieseke, R. K. Analysis of Nonlinear Structures via Mode Synthesis, NASTRAN: Users’
Exper., pp. 341-360, September, 1975, (NASA TM X-3278).
• Graves, Roger W. Interfacing MSC/NASTRAN with SDRC-IDEAS to Perform Component
Mode Synthesis Combining Test, Analytical, and F. E. Data, The MSC 1988 World Users
Conf. Proc., Vol. II, Paper No. 58, March, 1988.
• Halcomb, J. R. Application of Component Modes to the Analysis of a Helicopter, Proc. of the
MSC/NASTRAN Users’ Conf., March, 1979.
• Hambric, Stephen A. Power Flow and Mechanical Intensity Calculations in Structural Finite
Element Analysis, ASME J. of Vibration and Acoustics, Vol. 112, pp. 542-549, October, 1990.
• Herting, David N.; Hoesly, R. L. Development of an Automated Multi-Stage Modal Synthesis
System for NASTRAN, Sixth NASTRAN Users’ Colloq., pp. 435-448, October, 1977, (NASA
CP-2018).
• Herting, David N. Accuracy of Results with NASTRAN Modal Synthesis, Seventh NASTRAN
Users’ Colloq., pp. 389-404, October, 1978, (NASA CP-2062).
• Herting, D. N. A General Purpose, Multi-Stage, Component Modal Synthesis Method, Finite
Elements in Analysis and Design, Vol. 1, No. 2, 1985.
• Hill, R. G.; Merckx, K. R. Seismic Response Evaluation of a Reactor Core Using Modal
Synthesis, Proc. of the 3rd Int. Modal Analysis Conf., Vol. 2, pp. 996-1000, June, 1985.
• Hodgetts, P. A.; Maitimo, F. M.; Wijker, J. J. Dynamic Analysis of the Polar Platform Solar
Array Using a Multilevel Component Mode Synthesis Technique, Proc. of the 19th MSC
European Users’ Conf., Paper No. 15, September, 1992.
• Ichikawa, Tetsuji; Hagiwara, Ichiro. Frequency Response Analysis of Large-Scale Damped
Structures Using Component Mode Synthesis, Nippon Kikai Gakkai Ronbunshu, C
Hen/Transactions of the Japan Society of Mechanical Engineers, Part C v 60 n 569 Jan 1994.
• Jasuja, S. C.; Borowski, V. J.; Anderson, D. H. Finite Element Modeling Techniques for the
Simulation of Automotive Powertrain Dynamics, Proc. of the 5th Int. Modal Analysis Conf.,
Vol. II, pp. 1520-1530, April, 1987.
• Jay, Andrew; Lewis, Bryan; Stakolich, Ed. Effect of Time Dependent Flight Loads on
Turbofan Engine Performance Deterioration, ASME J. of Engineering for Power, Vol. 104,
No. 3, July, 1982.
• Kammer, Daniel C.; Jensen, Brent M.; Mason, Donald R. Test- Analysis Correlation of the
Space Shuttle Solid Rocket Motor Center Segment, J. of Spacecraft, Vol. 26, No. 4,
pp. 266-273, March, 1988.
• Kasai, Manabu. Generalized CMS Employing External Superelements, The Fourth
MSC/NASTRAN User’s Conf. in Japan, October, 1986, in Japanese.
• Kasai, Manabu. Representation of DMAP by Generalized CMS in System Identification, The
Fourth MSC/NASTRAN User’s Conf. in Japan, October, 1986, in Japanese.
Main Index
528 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Kasai, Manabu. Approach to CMS Subjected to the Boundary Constrained at Single Point,
The Sixth MSC/NASTRAN User’s Conf. in Japan, October, 1988, in Japanese.
• Kim, Hyoung M.; Bartkowicz, Theodoore J.; Van Horn, David A. Data Recovery and Model
Reduction Methods for Large Structures, The MSC 1993 World Users’ Conf. Proc., Paper
No. 23, May, 1993.
• Kubota, Minoru. Efficient Use of Component Mode Sysnthesis Using Image Superelements
Applied to Dynamic Analysis of Crankshaft, MSC/NASTRAN Users’ Conf. Proc., Paper
No. 22, March, 1986.
• Lee, W. M. Substructure Mode Synthesis with External Superelement, The 2nd Annual MSC
Taiwan User’s Conf., Paper No. 16, October, 1990, in Chinese.
• MacNeal, R. H. A Hybrid Method of Component Mode Synthesis, Computers and Structures,
Vol. 1, No. 4, pp. 581-601, 1971.
• Martinez, D. R.; Gregory, D. L. A Comparison of Free Component Mode Synthesis
Techniques Using MSC/NASTRAN, MSC/NASTRAN Users’ Conf. Proc., Paper No. 18,
March, 1983.
• Martinez, David R.; Gregory, Danny L. A Comparison of Free Component Mode Synthesis
Techniques Using MSC/NASTRAN, Sandia National Laboratories, June, 1984,
(SAND83-0025).
• Murakawa, Osamu. Hull Vibration Analysis by Modal Synthesis Method, The First
MSC/NASTRAN User’s Conf. in Japan, October, 1983, in Japanese.
• Murakawa, Osamu; Iwahashi, Yoshio; Sakato, Tsuneo. Ship Vibration Analysis Using Modal
Synthesis Technique, MSC/NASTRAN Users’ Conf. Proc., Paper No. 23, March, 1984.
• Nasu, Syouichi. Modal Synthesis of Experimental Vibration Characteristics Using
MSC/NASTRAN Results as the Reference Model, The Sixth MSC/NASTRAN User’s Conf. in
Japan, October, 1988, in Japanese.
• Nefske, D. J.; Sung, S. H.; Duncan, A. E. Applications of Finite Element Methods to Vehicle
Interior Acoustic Design, Proc. of the 1984 Noise and Vibration Conf., Paper No. 840743, 1984.
• Ookuma, Masaaki; Nagamatsu, Akio. Comparison of Component Mode Synthesis Method
with MSC-NASTRAN, Nippon Kikai Gakkai Ronbunshu, C Hen, Vol. 49, No. 446,
pp. 1883-1889, October, 1983, in Japanese.
• Ookuma, Masaaki; Nagamatsu, Akio. Comparison of Component Mode Synthesis Method
with MSC-NASTRAN, Bulletin of the JSME, Vol. 27, No. 228, pp. 1294-1298, June, 1984.
• Parekh, Jatin C.; Harris, Steve G. The Application of the Ritz Procedure to Damping Prediction
Using a Modal Strain Energy Approach, Damping ’89, Paper No. CCB, November, 1989.
• Philippopoulos, V. G. Dynamic Analysis of an Engine-Transmission Assembly-Superelement
and Component Mode Synthesis, Proc. of the Conf. on Finite Element Methods and
Technology, Paper No. 3, March, 1981.
• Reyer, H. Modal Synthesis with External Superelements in MSC/NASTRAN, Proc. of the
MSC/NASTRAN Eur. Users’ Conf., May, 1984.
Main Index
APPENDIX C 529
References and Bibliography
• Sabahi, Dara; Rose, Ted. Special Applications of Global-Local Analysis, The MSC 1990
World Users Conf. Proc., Vol. II, Paper No. 49, March, 1990.
• Sabahi, Dara; Rose, Ted. MSC/NASTRAN Superelement Analysis of the NASA/AMES
Pressurized Wind Tunnel, The MSC 1990 World Users Conf. Proc., Vol. II, Paper No. 50,
March, 1990.
• Shein, Shya-Ling; Marquette, Brian; Rose, Ted. Superelement Technology Application and
Development in Dynamic Analysis of Large Space Structures, The MSC 1991 World Users’
Conf. Proc., Vol. I, Paper No. 26, March, 1991.
• Suzukiri, Yoshihiro. Component Mode Synthesis Application of MSC/NASTRAN V66, The
2nd Annual MSC Taiwan Users Conf., Paper No. 10, October, 1990.
• Suzukiri, Yoshihiro. Component Mode Synthesis Application of MSC/NASTRAN V66, Proc.
of the First MSC/NASTRAN Users’ Conf. in Korea, Paper No. 17, October, 1990.
• Tong, Edward T.; Chang, Craig C. J. An Efficient Procedure for Data Recovery of a
Craig-Bampton Component, MSC 1994 World Users’ Conf. Proc., Paper No. 26, June 1994.
• Wamsler, M.; Komzsik, L.; Rose, T. Combination of Quasi-Static and Dynamic System Mode
Shapes, Proc. of the 19th MSC European Users’ Conf., Paper No. 13, September, 1992.
• Wang, Bo Ping. Synthesis of Structures with Multiple Frequency Constraints,
AIAA/ASME/ASCE/AHS 27th Structures, Structural Dynamics and Materials Conf., Part 1,
pp. 394-397, May, 1986.
DYNAMICS – DAMPING
• El Maddah, M.; Imbert, J. F. A Comparison of Damping Synthesis Methods for Space Vehicle
Dynamic Analysis, NASTRAN User’s Conf., May, 1979.
• Everstine, Gordon C.; Marcus, Melvyn S. Finite Element Prediction of Loss Factors for
Structures with Frequency- Dependent Damping Treatments, Thirteenth NASTRAN Users’
Colloq., pp. 419-430, May, 1985, (NASA CP-2373).
• Gibson, W. C.; Johnson, C. D. Optimization Methods for Design of Viscoelastic Damping
Treatments, ASME Design Engineering Division Publication, Vol. 5, pp. 279-286, September,
1987.
• Gibson, Warren C.; Austin, Eric. Analysis and Design of Damped Structures Using
MSC/NASTRAN, The MSC 1992 World Users’ Conf. Proc., Vol. I, Paper No. 25, May, 1992.
• Johnson, Conor D.; Keinholz, David A. Prediction of Damping in Structures with Viscoelastic
Materials Using MSC/NASTRAN, MSC/NASTRAN Users’ Conf. Proc., Paper No. 17, March,
1983.
• Kalinowski, A. J. Modeling Structural Damping for Solids Having Distinct Shear and
Dilational Loss Factors, Seventh NASTRAN Users’ Colloq., pp. 193-206, October, 1978,
(NASA CP-2062).
Main Index
530 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Kalinowski, A. J. Solution Sensitivity and Accuracy Study of NASTRAN for Large Dynamic
Problems Involving Structural Damping, Ninth NASTRAN Users’ Colloq., pp. 49-62, October,
1980, (NASA CP-2151).
• Kienholz, Dave K.; Johnson, Conor D.; Parekh, Jatin C. Design Methods for Viscoelastically
Damped Sandwich Plates, AIAA/ASME/ASCE/AHS 24th Structures, Structural Dynamics and
Materials Conf., Part 2, pp. 334-343, May, 1983.
• Li, Tsung-hsiun; Bernard, James. Optimization of Damped Structures in the Frequency
Domain, The MSC 1993 World Users’ Conf. Proc., Paper No. 28, May, 1993.
• Lu, Y. P.; Everstine, G. C. More on Finite Element Modeling of Damped Composite Systems, J.
of Sound and Vibration, Vol. 69, No. 2, pp. 199-205, 1980.
• Mace, M. Damping of Beam Vibrations by Means of a Thin Constrained Viscoelastic Layer:
Evaluation of a New Theory, Journal of Sound and Vibration v 172 n 5 May 19 1994.
• Merchant, D. H.; Gates, R. M.; Ice, M. W.; Van Derlinden, J. W. The Effects of Localized
Damping on Structural Response, NASTRAN: Users’ Exper., pp. 301-320, September, 1975,
(NASA TM X-3278).
• Parekh, Jatin C.; Harris, Steve G. The Application of the Ritz Procedure to Damping Prediction
Using a Modal Strain Energy Approach, Damping ’89, Paper No. CCB, November, 1989.
• Rose, Ted. DMAP Alters to Apply Modal Damping and Obtain Dynamic Loading Output for
Superelements, The MSC 1993 World Users’ Conf. Proc., Paper No. 24, May, 1993.
• Rose, Ted.; McNamee, Martin. A DMAP Alter to Allow Amplitude-Dependent Modal
Damping in a Transient Solution, MSC 1996 World Users’ Conf. Proc., Vol. V, Paper No. 50,
June, 1996.
• Shieh, Rong C. A Superefficient, MSC/NASTRAN-Interfaced Computer Code System for
Dynamic Response Analysis of Nonproportionally Damped Elastic Systems, The MSC 1993
World Users’ Conf. Proc., Paper No. 14, May, 1993.
• Tonin, Renzo. Vibration Isolation of Impacts in High-Rise Structures, The Second
Australasian MSC Users Conf., Paper No. 11, November, 1988.
Main Index
APPENDIX C 531
References and Bibliography
Main Index
532 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Wamsler, Manfred; Krusemann, Rolf. Calculating and Interpreting Contact Forces Between
Brake Disc and Linings in Frequency Response Analysis, Proc. of the 18th MSC Eur. Users’
Conf., Paper No. 7, June, 1991.
• Yen, K. Z. Y.; Hsueh, W. C.; Hsui, T. C. Chatter Suppression of a CNC Lathe in
Inside-Diameter Cutting, The Sixth Annual MSC Taiwan Users’ Conf. Proc., Paper No. 4,
November, 1994.
Main Index
APPENDIX C 533
References and Bibliography
• Barnett, Alan R.; Abdallah, Ayma A.; Ibrahim, Omar M.; Sullivan, Timothy L. Solving Modal
Equations of Motion with Initial Conditions Using MSC/NASTRAN DMAP Part 2: Coupled
vs. Uncoupled Integration, The MSC 1993 World Users’ Conf. Proc., Paper No. 13, May, 1993.
• Bella, David F.; Steinhard, E. Critical Frequency Determination of a Flexible Rotating
Structure Attached to a Flexible Support, Proc. of the 18th MSC Eur. Users’ Conf., Paper
No. 28, June, 1991.
• Bella, David; Hartmueller, Hans; Muehlenfeld, Karsten; Tokar, Gabriel. Identification of
Critical Speeds of Rotors Attached to Flexible Supports, The MSC 1993 World Users’ Conf.
Proc., Paper No. 34, May, 1993.
• Blakely, Ken; Rose, Ted. Cross-Orthogonality Calculations for Pre-Test Planning and Model
Verification, The MSC 1993 World Users’ Conf. Proc., Paper No. 72, May, 1993.
• Blakely, Ken; Rose, Ted. Cross-Orthogonality Calculations for Pre-Test Planning and Model
Verification, Proc. of the 20th MSC European Users’ Conf., September, 1993.
• Brughmans, M.; Lembregts, PhD. F.; Furini, PhD. F.; Storrer, O. Modal Test on the Pininfarina
Concept Car Body “ETHOS 1", Actes de la 2ème Confèrence Française Utilisateurs des
Logiciels MSC, Toulouse, France, September, 1995.
• Brughmans, M.; Lembregts, F, Ph.D.; Furini, F., Ph.D. Modal Test on the Pininfarina Concept
Car Body “ETHOS 1", MSC 1995 World Users’ Conf. Proc., Paper No. 5, May, 1995.
• Buchanan, Guy. Superelement Data Recovery via the Modal Acceleration Method, The MSC
1988 World Users Conf. Proc., Vol. I, Paper No. 40, March, 1988.
• Budynas, R.; Kolhatkar, S. Modal Analysis of a Robot Arm Using Finite Element Analysis
and Modal Testing, Proc. of the 8th Int. Modal Analysis Conf., Vol. I, pp. 67-70, January, 1990.
• Budynas, R. G.; Krebs, D. Modal Correlation of Test and Finite Element Results Using Cross
Orthogonality with a Reduced Mass Matrix Obtained by Modal Reduction and NASTRAN’s
Generalized Dynamic Reduction Solution, Proc. of the 9th Int. Modal Analysis Conf., Vol. I,
pp. 549-554, April, 1991.
• Butler, Thomas G.; Muskivitch, John C. Application of Flanigan’s Mode Acceleration in
MSC/NASTRAN Version 66, The 1989 MSC World Users Conf. Proc., Vol. I, Paper No. 25,
March, 1989.
• Caldwell, Steve; Wang, B. P. Application of Approximate Techniques in the Estimation of
Eigenvalue Quality, The MSC 1993 World Users’ Conf. Proc., Paper No. 11, May, 1993.
• Campanile, P.; Pisino, E.; Testi, R.; Manzilli, G.; Minen, D. Flexible Structures in Adams
Using Modal Data from NASTRAN, Proc. of the 21st MSC European Users’ Conf., Italian
Session, September, 1994.
• Carlson, Mark. Applications of Finite Element Analysis for an Improved Musical Instrument
Design, MSC 1996 World Users’ Conf. Proc., Vol. I, Paper No. 8, June, 1996.
• Carneiro, S. H. S.; Duarte, J. A. A.; Mendonca, C. B. Theoretical and Experimental Modal
Analysis of the VLS (Satellite Launcher Vehicle) Bent, Proc. of the 2nd MSC/NASTRAN
Users’ Conf. in Brazil, Paper No. 3, March, 1993, in Portuguese.
Main Index
534 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Case, William R. Jr. NASTRAN DMAP Alter for Determining a Local Stiffness Modification
to Obtain a Specified Eigenvalue, NASTRAN: Users’ Exper., pp. 269-284, September, 1973,
(NASA TM X-2893).
• Case, William R. A NASTRAN DMAP Procedure for Calculation of Base Excitation Modal
Participation Factors, Eleventh NASTRAN Users’ Colloq., pp. 113-140, May, 1983.
• Cattani, E.; Micelli, D.; Sereni, L.; Cocordano, S. Cylinder Block Eigenfrequencies and
Eigenvectors Prediction with a Linear Brick and Wedge Finite Element Model, Proc. of the
19th MSC European Users’ Conf., Paper No. 16, September, 1992.
• Chang, Cuann-yeu; Chang, Yuan-bing. Using MSC/NASTRAN to Obtain Modal Parameters,
The MSC 1988 World Users Conf. Proc., Vol. I, Paper No. 21, March, 1988.
• Chargin, M. L.; Dunne, L. W.; Herting, D. N. Nonlinear Dynamics of Brake Squeal, MSC
1996 World Users’ Conf. Proc., Vol. V, Paper No. 47, June, 1996.
• Chen, J. T.; Chyuan, S. W.; You, D. W.; Wong, H. T. A New Method for Determining the Modal
Participation Factor in Support Motion Problems Using MSC/NASTRAN, The Seventh
Annual MSC/NASTRAN Users’ Conf. Proc., Taiwan, 1995.
• Clary, Robert R. Practical Analysis of Plate Vibrations Using NASTRAN, NASTRAN: Users’
Exper., pp. 325-342, September, 1971, (NASA TM X-2378).
• Cohen, Allan R.; Laurenson, Robert M. Application of a Substructure Technique for
STS/Payload Coupled Modal Analysis, Proc. of the MSC/NASTRAN Users’ Conf., March,
1979.
• Concilio, A.; Del Gatto, S.; Lecce, L.; Miccoli, G. Simple and Cheap Noise and Vibration
Active Control System Using Collocated Piezoelectric Devices on a Panel, Proceedings of the
11th International Modal Analysis Conference, Florida, 1993.
• Courtney, Roy Leon. NASTRAN Modeling Studies in the Normal-Mode Method and
Normal-Mode Synthesis, NASTRAN: Users’ Exper., pp. 181-200, September, 1971, (NASA
TM X-2378).
• Cronkhite, James D.; Smith, Michael R. Experiences in NASTRAN Airframe Vibration
Prediction at Bell Helicopter Textron, American Helicopter Soc. Dynamics Specialists Mtg.,
Section 6, Vibrations Session I, Paper No. 1, November, 1989.
• Cross, C.; Rao, A. Comparison of Modal Performance of Alternate Compressor Bracket
Design, Seventh Australasian Users Conf. Proc., Sydney, October, 1993.
• Deger, Yasar. Modal Analysis of a Concrete Gravity Dam - Linking FE Analysis and Test
Results, Proc. of the 20th MSC European Users’ Conf., September, 1993.
• Deutschel, Brian William. A Systematic Approach Using Finite Elements for Improving
Vehicle Ride, CAD/CAM Robotics and Factories of the Future Integration of Design, Analysis,
and Manufacturing (Proc.), Springer-Verlag Berlin, Heidelberg, Vol. I, pp. 150-154, 1989.
• Deutschel, Brian W.; Katnik, Richard B.; Bijlani, Mohan; Cherukuri, Ravi. Improving Vehicle
Response to Engine and Road Excitation Using Interactive Graphics and Modal Reanalysis
Methods, SAE Trans., Paper No. 900817, September, 1991.
Main Index
APPENDIX C 535
References and Bibliography
• Egashira, Yuji. Large Scale Vibration Analysis of Car Body Using Superelement Method, The
Second MSC/NASTRAN User’s Conf. in Japan, October, 1984, in Japanese.
• El-Bayoumy, Lotfi. Identification and Correction of Damaging Resonances in Gear Drives,
Gear Technology, Vol. 1, No. 2, pp. 14-19, August/September, 1984.
• Ferg, D.; Foote, L.; Korkosz, G.; Straub, F.; Toossi, M.; Weisenburger, R. Plan, Execute, and
Discuss Vibration Measurements, and Correlations to Evaluate a NASTRAN Finite Element
Model of the AH-64 Helicopter Airframe, National Aeronautics and Space Administration,
January, 1990, (NASA CR-181973).
• Flanigan, Chris. Methods for Calculating and Using Modal Initial Conditions in
MSC/NASTRAN, Proc. of the Conf. on Finite Element Methods and Technology, March, 1980.
• Flanigan, C. An Alternate Method for Mode Acceleration Data Recovery in MSC/NASTRAN,
Proc. of the Conf. on Finite Element Methods and Technology, Paper No. 7, March, 1981.
• Flanigan, Christopher C. Efficient and Accurate Procedures for Calculating Data Recovery
Matrices for Superelement Models, The 1989 MSC World Users Conf. Proc., Vol. II, Paper
No. 44, March, 1989.
• Flanigan, Christopher C. Implementation of the IRS Dynamic Reduction Method in
MSC/NASTRAN, The MSC 1990 World Users Conf. Proc., Vol. I, Paper No. 13, March, 1990.
• Friberg, Olof; Karlsson, Rune; Akesson, Bengt. Linking of Modal and Finite Elements in
Structural Vibration Analysis, Proc. of the 2nd Int. Modal Analysis Conf. and Exhibit, Vol. 1,
pp. 330-339, February, 1984.
• Gallaher, Bruce. Determination of Structural Dynamic Response Sensitivity to Modal
Truncation, MSC/NASTRAN Users’ Conf. Proc., Paper No. 10, March, 1986.
• Ghosh, Tarun; Nall, Marsha; Muniz, Ben; Cheng, Joseph. Space Station Solar Array Pointing
System Control/Structure Interaction Study Using CO-ST-IN for Modal Reduction, The MSC
1993 World Users’ Conf. Proc., Paper No. 68, May, 1993.
• Gieseke, R. K. Modal Analysis of the Mated Space Shuttle Configuration, NASTRAN:
Users’ Exper., pp. 221-236, September, 1971, (NASA TM X-2378).
• Girard, A.; Boullet, A.; Dardel, R. Dynamic Analysis of a Satellite Using the Normal Modes of
the Appendages, Proc. of the MSC/NASTRAN Eur. Users’ Conf., April, 1985.
• Grandle, Robert E.; Rucker, Carl E. Modal Analysis of a Nine-Bay Skin-Stringer Panel,
NASTRAN: Users’ Exper., pp. 343-362, September, 1971, (NASA TM X-2378).
• Gupta, Viney K.; Zillmer, Scott D.; Allison, Robert E. Solving Large-Scale Dynamic Systems
Using Band Lanczos Method in Rockwell NASTRAN on Cray X-MP, Fourteenth NASTRAN
Users’ Colloq., pp. 236-246, May, 1986, (NASA CP-2419).
• Hardman, E. S. Static and Normal Modes Analysis of an Aircraft Structure Using the
NASTRAN External Superelement Method, Proc. of the MSC/NASTRAN Eur. Users’ Conf.,
May, 1986.
• Harn, Wen-Ren; Hwang, Chi-Ching. Evaluation of Direct Model Modification Methods via
MSC/NASTRAN DMAP Procedures, The MSC 1990 World Users Conf. Proc., Vol. II, Paper
No. 43, March, 1990.
Main Index
536 MD Nastran Dynamic Analysis User’s Guide
Bibliography
Main Index
APPENDIX C 537
References and Bibliography
Main Index
538 MD Nastran Dynamic Analysis User’s Guide
Bibliography
Main Index
APPENDIX C 539
References and Bibliography
• Nagy, Lajos I.; Cheng, James; Hu, Yu-Kan. A New Method Development to Predict Brake
Squeal Occurence, MSC 1994 World Users’ Conf. Proc., Paper No. 14, June, 1994.
• Neads, M. A.; Eustace, K. I. The Solution of Complex Structural Systems by NASTRAN
within the Building Block Approach, NASTRAN User’s Conf., May, 1979.
• Newman, Malcolm; Pipano, Aaron. Fast Modal Extraction in NASTRAN via the FEER
Computer Program, NASTRAN: Users’ Exper., pp. 485-506, September, 1973, (NASA TM
X-2893).
• Nishiwaki, Nobukiyo. Coupled Vibration of Rotating Disc and Blades, The Second
MSC/NASTRAN User’s Conf. in Japan, October, 1984, in Japanese.
• Overbye, Vern D. MSC/NASTRAN Dynamic Analysis: Modal or Direct?, MSC/NASTRAN
Users’ Conf. Proc., Paper No. 6, March, 1986.
• Pamidi, M. R.; Pamidi, P. R. Modal Seismic Analysis of a Nuclear Power Plant Control Panel
and Comparison with SAP IV, NASTRAN: Users’ Exper., pp. 515-530, October, 1976, (NASA
TM X-3428).
• Pamidi, P. R. On the Append and Continue Features in NASTRAN, Seventh NASTRAN
Users’ Colloq., pp. 405-418, October, 1978, (NASA CP-2062).
• Paolozzi, A. Interfacing MSC/NASTRAN with a Structural Modification Code, Proc. of the
18th MSC Eur. Users’ Conf., Paper No. 30, June, 1991.
• Park, H. B.; Suh, J. K.; Cho, H. G.; Jung, G. S. A Study on Idle Vibration Analysis Technique
Using Total Vehicle Model, MSC 1995 World Users’ Conf. Proc., Paper No. 6, May, 1995.
• Parker, G. R.; Brown, J. J. Kinetic Energy DMAP for Mode Identification, MSC/NASTRAN
Users’ Conf. Proc., Paper No. 8, March, 1982.
• Parker, G. R.; Brown, J. J. Evaluating Modal Contributors in a NASTRAN Frequency
Response Analysis, MSC/NASTRAN Users’ Conf. Proc., Paper No. 14, March, 1983.
• Pulgrano, Louis J.; Masters, Steven G. Self-Excited Oscillation of a 165 Foot Water Tower,
MSC 1995 World Users’ Conf. Proc., Paper No. 32, May, 1995.
• Reyer, H. Modal Synthesis with External Superelements in MSC/NASTRAN, Proc. of the
MSC/NASTRAN Eur. Users’ Conf., May, 1984.
• Rose, Ted. A Method to Apply Initial Conditions in Modal Transient Solutions, The MSC
1991 World Users’ Conf. Proc., Vol. I, Paper No. 13, March, 1991.
• Rose, Ted. Some Suggestions for Evaluating Modal Solutions, The MSC 1992 World Users’
Conf. Proc., Vol. I, Paper No. 10, May, 1992.
• Rose, Ted. Using Optimization in MSC/NASTRAN to Minimize Response to a Rotating
Imbalance, MSC 1994 Korea Users’ Conf. Proc., December, 1994.
• Rose, Ted. Using Dynamic Optimization to Minimize Driver Response to a Tire
Out-of-Balance, MSC 1994 Korea Users’ Conf. Proc., December, 1994.
• Rose, Ted. Using Optimization in MSC/NASTRAN to Minimize Response to a Rotating
Imbalance, 1994 MSC Japan Users’ Conf. Proc., Paper No. 28, December 1994.
Main Index
540 MD Nastran Dynamic Analysis User’s Guide
Bibliography
Main Index
APPENDIX C 541
References and Bibliography
• Tawekal, Ricky; Budiyanto, M. Agus. Finite Element Model Correlation for Structures, The
MSC 1993 World Users’ Conf. Proc., Paper No. 73, May, 1993.
• Tawekal, Ricky L.; Miharjana, N. P. Validation of 3650 DWT Semi Containe Ship Finite
Element Model by Full Scale Measurements, MSC 1994 World Users’ Conf. Proc., Paper No.
19, June, 1994.
• Thornton, Earl A. A NASTRAN Correlation Study for Vibrations of a Cross-Stiffened Ship’s
Deck, NASTRAN: Users’ Exper., pp. 145-160, September, 1972, (NASA TM X-2637).
• Ting, Tienko; Chen, Timothy L. C.; Twomey, William. A Practical Solution to Mode Crossing
Problem in Continuous Iterative Procedure, The 1989 MSC World Users Conf. Proc., Vol. I,
Paper No. 14, March, 1989.
• Ting, Tienko; Chen, T.; Twomey, W. Correlating Mode Shapes Based on Modal Assurance
Criterion, The MSC 1992 World Users’ Conf. Proc., Vol. I, Paper No. 21, May, 1992.
• Tokuda, Naoaki; Mitikami, Shinsuke; Sakata, Yoshiuki. Accuracy of Vibration Analysis for
Thin Cylindrical Shell by MSC/NASTRAN, MSC/NASTRAN Users’ Conf. Proc., Paper
No. 28, March, 1984.
• Vaillette, David. Evaluation of the Modal Response of a Pressure Vessel Filled with a Fluid,
The MSC 1991 World Users’ Conf. Proc., Vol. I, Paper No. 24, March, 1991.
• Vance, Judy, M.; Bernard, James E. Approximating Eigenvectors and Eigenvalues Across a
Wide Range of Design, Finite Elements in Analysis and Design v 14 n 4 Nov 1993.
• Vandepitte, D.; Wijker, J. J.; Appel, S.; Spiele, H. Normal Modes Analysis of Large Models,
and Applications to Ariane 5 Engine Frame, Proc. of the 18th MSC Eur. Users’ Conf., Paper
No. 6, June, 1991.
• Wamsler, M.; Komzsik, L.; Rose, T. Combination of Quasi-Static and Dynamic System Mode
Shapes, Proc. of the 19th MSC European Users’ Conf., Paper No. 13, September, 1992.
• Wang, B. P.; Cheu, T. C.; Chen, T. Y. Optimal Design of Compressor Blades with Multiple
Natural Frequency Constraints, ASME Design Engineering Division Publication, Vol. 5,
pp. 113-117, September, 1987.
• Wang, B. P.; Lu, C. M.; Yang, R. J. Topology Optimization Using MSC/NASTRAN, MSC 1994
World Users’ Conf. Proc., Paper No. 12, June, 1994.
• Wang, Bo Ping. Minimum Weight Design of Structures with Natural Frequency Constraints
Using MSC/NASTRAN, The MSC 1988 World Users Conf. Proc., Vol. II, Paper No. 60, March,
1988.
• Welte, Y. Vibration Analysis of an 8MW Diesel Engine, Proc. of the MSC/NASTRAN Eur.
Users’ Conf., May, 1986.
• West, Timothy S. Approximate Dynamic Model Sensitivity Analysis For Large, Complex
Space Structues, MSC 1996 World Users’ Conf. Proc., Vol. I, Paper No. 6, June, 1996.
• Wijker, J. J. Substructuring Technique Using the Modal Constraint Method, Proc. of the
MSC/NASTRAN Eur. Users’ Conf., June, 1983.
• Wijker, J. J. MSC/NASTRAN Normal Mode Analysis on CRAY Computers, Proc. of the
MSC/NASTRAN Eur. Users’ Conf., June, 1983.
Main Index
542 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Yang, Howard J. Sorted Output in MSC/NASTRAN, Proc. of the Conf. on Finite Element
Methods and Technology, Paper No. 4, March, 1981.
• Yen, K. Z. Y.; Hsueh, W. C.; Hsui, T. C. Chatter Suppression of a CNC Lathe in
Inside-Diameter Cutting, The Sixth Annual MSC Taiwan Users’ Conf. Proc., Paper No. 4,
November, 1994.
• Zhu, H.; Knight, D. Finite Element Forced Response Analysis on the Mondeo Front End
Accessory Drive System, Proc. of the 20th MSC European Users’ Conf., September, 1993.
• Shock and Vibration Handbook, Page 4-8, 3rd Edition, Cyril M. Harris, McGraw Hill, 1987.
• Frequency-domain analysis of nonlinear circuits driven by multi-tone signals, A.Ushida and
L.O.Chua, IEEE Trans. Circuits Syst., Vol. CAS-31, pp. 766-778, Sept. 1984.
• Steady-State Response of Continuous Nonlinear Rotor-Bearing Systems Using Analytical
Approach, J.W. Zu and Z.Y. Ji, Journal of Engineering for Gas Turbines and Power, ASME, 120,
pp 751 - 758, 1998
Main Index
APPENDIX C 543
References and Bibliography
Main Index
544 MD Nastran Dynamic Analysis User’s Guide
Bibliography
Main Index
APPENDIX C 545
References and Bibliography
DYNAMICS – SEISMIC
• Bonaldi, P.; Peano, A.; Ruggeri, G.; Venturuzzo, M. Seismic and Impact Analyses of Nuclear
Island Buildings of Italian Unified Nuclear Design, Proc. of the 15th MSC/NASTRAN Eur.
Users’ Conf., October, 1988.
• Burroughs, John W. An Enhancement of NASTRAN for the Seismic Analysis of Structures,
Ninth NASTRAN Users’ Colloq., pp. 79-96, October, 1980, (NASA CP-2151).
• Chen, J. T.; Chyuan, S. W.; Yeh, C. S.; Hong, H. K. Comparisons of Analytical Solution and
MSC/NASTRAN Results on Multiple-Support Motion of Long Bridge During Earthquake,
The 4th MSC Taiwan Users’ Conf., Paper No. 20, November, 1992.
• Chen, J. T.; Hong, H. K.; Yen, C. S.; Chyuan, S. W. Integral Representations and
Regularizations for a Divergent Series Solution of a Beam Subjected to Support Motions,
Earthquake Engineering and Structural Dynamics, Vol. 25, pp. 909-925, 1996.
• Chen, Yohchia. Nonlinear Seismic Analysis of Bridges: Practical Approach and
Comparative Study, The MSC 1993 World Users’ Conf. Proc., Paper No. 57, May, 1993.
• Chen, Yohchia. Refined Analysis for Soil-Pipe Systems, MSC 1994 World Users’ Conf. Proc.,
Paper No. 38, June, 1994.
• Dahlgren, F.; Citrolo, J.; Knutson, D.; Kalish, M. Dynamic Analysis of the BPX Machine
Structure, Proc. of the 14th IEEE/NPSS Symp. on Fusion Engineering, Vol. 1, pp. 47-49, 1992.
• Fallet, P.; Derivery, J. C. MSC/NASTRAN Earthquake Analysis, NASTRAN User’s Conf.,
May, 1979.
• Hill, Ronald G. Nonlinear Seismic Analysis of a Reactor Structure with Impact Between Core
Components, NASTRAN: Users’ Exper., pp. 397-418, September, 1975, (NASA TM X-3278).
• Hirata, M.; Ishikawa, K.; Korosawa, M.; Fukushima, S.; Hoshina, H. Seismic Analysis of
Plutonium Glovebox by MSC/NASTRAN, January, 1993.
• Liepins, Atis A.; Nazemi, Hamid. Virtual Mass of Fluid in Egg-Shaped Digesters, The MSC
1993 World Users’ Conf. Proc., Paper No. 77, May, 1993.
• Nomura, Yoshio. Seismic Response Analysis by MSC/NASTRAN for Coupled Structure -
Ground and Pile, The Fifth MSC/NASTRAN User’s Conf. in Japan, October, 1987, in
Japanese.
• Pamidi, M. R.; Pamidi, P. R. Modal Seismic Analysis of a Nuclear Power Plant Control Panel
and Comparison with SAP IV, NASTRAN: Users’ Exper., pp. 515-530, October, 1976, (NASA
TM X-3428).
• Tsaur, D. H.; Chyuan, S. W.; Chen, J. T. On the Accuracy of MSC/NASTRAN on Response of
Two-Span Beams to Spatially Varying Seismic Excitation, The 4th MSC Taiwan Users’ Conf.,
Paper No. 9, November, 1992.
Main Index
546 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Yiak, K. C.; Pezeshk, S. Seismic Study of a Cable-Stayed Bridge, Proceedings of the Structures
Congrees ’94, Atlanta, 1994.
• Zhou, Hongye; Chen, Youping. The Influence of Phase-Difference Effects on Earthquake
Response of Cable-Stayed Bridges, MSC 1994 World Users’ Conf. Proc., Paper No. 37, June,
1994.
Main Index
APPENDIX C 547
References and Bibliography
• Frye, John W. Transient Analysis of Bodies with Moving Boundaries Using NASTRAN,
NASTRAN: Users’ Exper., pp. 377-388, September, 1975, (NASA TM X-3278).
• Hirata, M.; Ishikawa, K.; Korosawa, M.; Fukushima, S.; Hoshina, H. Seismic Analysis of
Plutonium Glovebox by MSC/NASTRAN, January, 1993.
• Katnik, Richard B.; Deutschel, Brian; Cherukuri, Ravi. Transient Response of a Vehicle Over
Road Bumps Using the Fourier Transform in a Modal Subspace, The MSC 1992 World
Users’ Conf. Proc., Vol. I, Paper No. 6, May, 1992.
• Kim, Hyoung M.; Bartkowicz, Theodoore J.; Van Horn, David A. Data Recovery and Model
Reduction Methods for Large Structures, The MSC 1993 World Users’ Conf. Proc., Paper
No. 23, May, 1993.
• Larkin, Paul A.; Miller, Michael W. STS Coupled Loads Analysis Using MSC/NASTRAN,
MSC/NASTRAN Users’ Conf. Proc., Paper No. 18, March, 1985.
• Lee, J. H.; Tang, J. H. K. Dynamic Response of Containments Due to Shock Wave, Proc. of the
Int. Conf. on Containment Design, pp. 25-32, June, 1984.
• Lee, Sang H.; Bock, Tim L.; Hsieh, Steve S. Adaptive Time Stepping Algorithm for Nonlinear
Transient Analysis, The MSC 1988 World Users Conf. Proc., Vol. II, Paper No. 54, March,
1988.
• Leifer, Joel; Gross, Michael. Non-Linear Shipboard Shock Analysis of the Tomahawk Missile
Shock Isolation System, 58th Shock and Vibration Symp., Vol. 1, pp. 97-117, October, 1978,
(NASA CP-2488).
• Lipman, Robert R. Computer Animation of Modal and Transient Vibrations, Fifteenth
NASTRAN Users’ Colloq., pp. 111-117, August, 1987, (NASA CP-2481).
• Mattana, G.; Miranda, D. MSC/NASTRAN Applications in P-180 Analysis, Proc. of the
MSC/NASTRAN First Italian Users’ Conf., October, 1987.
• McMeekin, Michael; Kirchman, Paul. An Advanced Post Processing Methodology for
Viewing MSC/NASTRAN Generated Analyses Results, MSC 1994 World Users’ Conf. Proc.,
Paper No. 21, June, 1994.
• McNamee, Martin J.; Zavareh, Parviz. Nonlinear Transient Analysis of a Shock Isolated
Mechanical Fuse, The MSC 1990 World Users Conf. Proc., Vol. I, Paper No. 21, March, 1990.
• Neilson, H. C.; Everstine, G. C.; Wang, Y. F. Transient Response of Submerged Fluid-Coupled
Double-Walled Shell Structure to a Pressure Pulse, J. of the Acoustic Soc. of America, Vol. 70,
No. 6, pp. 1776-1782, December, 1981.
• Pamidi, P. R. On the Append and Continue Features in NASTRAN, Seventh NASTRAN
Users’ Colloq., pp. 405-418, October, 1978, (NASA CP-2062).
• Rose, Ted. A Method to Apply Initial Conditions in Modal Transient Solutions, The MSC
1991 World Users’ Conf. Proc., Vol. I, Paper No. 13, March, 1991.
• Rose, Ted.; McNamee, Martin. A DMAP Alter to Allow Amplitude- Dependent Modal
Damping in a Transient Solution, MSC 1996 World Users’ Conf. Proc., Vol. V, Paper No. 50,
June, 1996.
Main Index
548 MD Nastran Dynamic Analysis User’s Guide
Bibliography
• Swan, Jim. A DMAP Alter for Interface Loads Across Superelements in Dynamic Analyses,
The MSC 1992 World Users’ Conf. Proc., Vol. I, Paper No. 23, May, 1992.
• Tang, C. C. Space Station Freedom Solar Array Wing: Nonlinear Transient Analysis of
Plume Impingement Load, MSC 1994 World Users’ Conf. Proc., Paper No. 35, June, 1994.
• Urban, Michael R.; Dobyns, Alan. MSC/NASTRAN Transient Analysis of Cannon Recoil
Loads on Composite Helicopters, The MSC 1991 World Users’ Conf. Proc., Vol. I, Paper
No. 28, March, 1991.
• West, Timothy S. Approximate Dynamic Model Sensitivity Analysis For Large, Complex
Space Structues, MSC 1996 World Users’ Conf. Proc., Vol. I, Paper No. 6, June, 1996.
• Wingate, Robert T.; Jones, Thomas C.; Stephens, Maria V. NASTRAN Postprocessor Program
for Transient Response to Input Accelerations, NASTRAN: Users’ Exper., pp. 707-734,
September, 1973, (NASA TM X-2893).
• Yang, Jackson C. S.; Goeller, Jack E.; Messick William T. Transient Analysis Using Conical
Shell Elements, NASTRAN: Users’ Exper., pp. 125-142, September, 1973, (NASA TM
X-2893).
Main Index
MSC.Fatigue Quick Start Guide
Index
MD Nastran Dynamic Analysis User’s Guide
Main Index
550 MD Nastran Dynamic Analysis User’s Guide
CFAST, 22 CTETRA, 22
CHACAB, 470 CTRIA3, 22, 84
Characteristic frequency, 44 CTRIA6, 22
Characteristic shape, 44 CTRIAR, 22
CHEXA, 22, 89, 470 CTRIAX6, 22
Circular natural frequency, 5, 47 CTUBE, 22
CMASSi, 22, 27, 304, 305 CVISC, 150, 204, 212, 261
Complex arithmetic, 141
Complex eigensolutions, 501, 502 D
Complex Lanczos method, 504 Damped circular natural frequency, 7
example, 508 damper elements, 487
Hessenberg method, 503, 507 Damping, 35, 260, 303
Inverse power method, 504 critical, 264
Modal Hessenberg method, 504 modal, 263
modeling techniques, 507 nonlinear, 266
user interface, 505 structural, 261, 266
Complex eigenvalue analysis, 500 viscuous, 261, 264, 266
complex eigenvalue analysis, 179 Damping ratio, 7
complex source strength, 477 DAREA, 162, 167, 178, 184, 194, 222, 225,
Component mode synthesis, 420, 372, 379 236, 242, 304, 321
Computer resource requirements, 309 Data recovery
Computer resources, 20 and superelements, 490
CONEAX, 22 for modal methods, 485
CONMi, 22, 27, 304, 305 Decoupled response solutions, 421
CONROD, 22 DELAY, 163, 167, 178, 222, 226, 236
Consistent mass, 22, 24, 25 Design optimization, 422
Control system, 423 Design sensitivity, 443
Control systems, 440, 441 DIAG 10, 390
example, 365 Direct frequency response analysis, 141, 152,
Coriolis forces, 32, 395 493
Coupled acoustic analysis damping, 141
job output, 418 excitation definition, 160
Coupled mass, 21, 22, 24, 25 Direct matrix input, 360
Coupled mass matrix terms, 30 Direct matrix input (DMIGs), 361, 362
Coupled scalar mass Direct transient response analysis, 203, 218
example, 30 damping, 204
CPENTA, 22, 89 excitation definition, 219
CQUAD4, 22, 84, 470 initial conditions, 206
CQUAD8, 22 DISPLACEMENT, 62, 80, 177, 193
CQUADR, 22 displacements, 284
CRAC2D, 22 DLOAD, 304
CRAC3D, 22 DLOAD (Bulk Data), 169, 178, 184, 194, 228,
Critical damping, 7 236, 242, 321
CROD, 22 DLOAD (Case Control), 167, 168, 176, 187,
cross-PSDs, 331 194, 225, 242
CSHEAR, 22 DMIG, 94
Main Index
INDEX 551
Main Index
552 MD Nastran Dynamic Analysis User’s Guide
Main Index
INDEX 553
Main Index
554 MD Nastran Dynamic Analysis User’s Guide
Main Index
INDEX 555
Main Index
556 MD Nastran Dynamic Analysis User’s Guide
U
Underdamped system, 7
Units of measure, 40
Unstable integration, 390
User messages
3035, 101, 109, 111
USETi,U6, 497
V
VECTOR, 62, 177, 233
velocities, 284
VELOCITY, 177
Velocity, 3
Virtual fluid mass, 392, 406, 421
example, 407
theory, 408
user interface, 406
Virtual mass
phantom boundaries, 411
singularities for enclosed volumes, 409
Main Index