Dynamic Analysis
User’s Guide
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 thirdparty 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. LSDYNA 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 © 19922005, 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. 359392, 1999.
Revision 0. June 25, 2010
MDNA*V2010*Z*Z*Z*DCUCR
Corporate Europe Asia Pacific
MSC.Software Corporation
2 MacArthur Place
Santa Ana, CA 92707
Telephone: (800) 3452078
FAX: (714) 7844056
MSC.Software GmbH
Am Moosfeld 13
81829 Munich
GERMANY
Telephone: (49) (89) 43 19 87 0
Fax: (49) (89) 43 61 71 6
Asia Pacific
MSC.Software Japan Ltd.
Shinjuku First West 8F
237 Nishi Shinjuku
1Chome, ShinjukuKu
Tokyo 1600023, JAPAN
Telephone: 0120924832 (toll
free, Japan only)
Mobile phone: 0369111222
Fax: (81) (3)69111201
Main Index
Cont ent s
MD Nastran Dynamic Analysis User’s Guide
Preface
About this Book xii
Acknowledgements xiii
List of Nastran Books xiv
Technical Support xv
Internet Resources xvi
1 Fundamentals of Dynamic Analysis
Overview 2
Equations of Motion 3
Dynamic Analysis Process 13
Dynamic Analysis Types 15
2 Finite Element Input Data
Overview 18
Strategies for Dynamic Analysis 19
Plan the Analysis 20
Mass Input 21
Coupled Mass Matrix Terms 30
Mass Effects in Moving Coordinate Systems 32
Damping Input 35
Units in Dynamic Analysis 40
MD Nastran
Dynamic Analysis
User’s Guide
Main Index
MD Nastran Dynamic Analysis User’s Guide
iv
3 Real Eigenvalue Analysis
Overview 44
Reasons to Compute Normal Modes 46
Mathematical Overview of Normal Modes Analysis 47
Methods of Computation 53
User Interface for Real Eigenvalue Analysis 56
Solution Control for Analysis 62
Massless Mechanism Identification and Control 64
Prestiffened Normal Mode Analysis 66
Examples 70
Rigid Body Modes 99
SUPORT Entry 101
Rigid Body Mode Examples 107
Tools to Help Identify and Validate Normal Modes 114
4 Frequency Response Analysis
Overview 134
Review of Dynamic Excitations 136
Direct Frequency Response Analysis 141
Frequency Response Function (FRF) and FRF Based Assembly (FBA)
143
Modal Frequency Response Analysis 146
Modal Versus Direct Frequency Response 152
FrequencyDependent Structural Elements 153
FrequencyDependent Excitation Definition 160
Solution Frequencies 171
Frequency Response Considerations 175
Main Index
v CONTENTS
Solution Control for Frequency Response Analysis 176
Modal Contribution Fraction (MCFRACTION) 179
Examples 184
5 Transient Response Analysis
Overview 202
Direct Transient Response Analysis 203
Modal Transient Response Analysis 208
Initial Condition Usage in Linear Transient Analysis 214
Dynamic Data Recovery in Modal Transient Response Analysis 217
Modal Versus Direct Transient Response 218
Transient Excitation Definition 219
Integration Time Step 230
Transient Excitation Considerations 231
Solution Control for Transient Response Analysis 232
Restarts in Linear Transient Analysis 235
Examples 236
MAX/MIN Output (New Form) 252
6 Damping
Overview 260
Modeling Damping Effects 261
Damping Effect on Complex Eigenvalues 267
Structural Damping for Shell and Bush Elements 270
Geometric and Material Nonlinear Spring and Damper Element 272
Main Index
MD Nastran Dynamic Analysis User’s Guide
vi
7 Enforced Motion
Overview 282
Direct Enforced Motion in Dynamic Analysis 283
Enforced Motion with Loads 292
Large Mass Method  Direct 304
Large Mass Method  Modal 306
User Interface for the Large Mass Method 308
Examples 311
8 Random Analysis
Random Analysis with Coupled Excitations 326
9 Response Spectrum Analysis
Shock and Response Spectrum Analysis 348
Response Spectrum Examples 358
10 Restarts In Dynamic Analysis
Overview 378
Examples 379
Remarks 381
11 Coupled FluidStructure Analysis
Overview 392
Axisymmetric Hydroelastic Analysis 394
Virtual Fluid Mass 406
Coupled and Uncoupled Acoustic Analysis 414
Main Index
vii CONTENTS
Exterior Acoustics 428
BW Method of FluidStructure Analysis 443
Direct Input of Interface [A] Matrix 455
Acoustic Modal Participation Factors 456
Acoustic XY Plots, Random, and Restarts 469
Acoustic Source Change 477
Rigid Porous Absorber  MAT10 478
12 Modal Augmentation Methods
Overview 484
Data Recovery for Modal Methods 485
Dynamic Response Predictions 491
Modal Versus Direct Frequency Response 493
Residual Vector 494
13 Complex Eigenvalue Analysis
Overview 500
Complex Eigensolutions 501
14 Guidelines and Tools for Effective Dynamic Analysis
Overview 288
Overall Analysis Strategy 289
Units 292
Mass 293
Damping 294
Boundary Conditions 298
Main Index
MD Nastran Dynamic Analysis User’s Guide
viii
Loads 299
Meshing 300
Eigenvalue Analysis 302
Frequency Response Analysis 303
Transient Response Analysis 305
Results Interpretation and Verification 307
Computer Resource Requirements 309
Debugging Dynamic Models 311
Weight Checks 315
Groundcheck 317
Static Loading 318
Thermal Test Case 319
Modal Analysis 320
Modal Effective Mass Computation 321
Modal Kinetic and Strain Energy Output 324
Grid Point Kinetic Energy 327
Element Strain Energy 328
Element Kinetic Energy 330
Element Energy Loss 332
Total Energy Output for Defined SETs 334
15 Results Processing
Overview 338
Structure Plotting 339
XY Plotting 345
Main Index
ix CONTENTS
16 Special Topics
Direct Matrix Input Dynamics 360
DMIGs, Extra Points, and Transfer Functions 361
Dynamic Reduction 371
Superelement Analysis 379
Nonlinear Transient Response Analysis 384
Nonlinear Transient Response Analysis Elements 396
Nonlinear Harmonic Response 410
Design Optimization and Sensitivity Overview 422
Control System Analysis 423
Fourier Transform 424
Squeeze Film Damper Nonlinear Force 430
Aeroelastic Analysis 437
Viscoelastic Material Properties 453
OPENFSI 463
Rotor Dynamics Overview 496
A Glossary of Terms and Nomenclature
Glossary of Terms 498
Nomenclature for Dynamic Analysis 500
B The Set Notation System Used in Dynamic Analysis
Overview 504
C References and Bibliography
Overview 508
Main Index
MD Nastran Dynamic Analysis User’s Guide
x
General References 509
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
MD Nastran Dynamic Analysis User’s Guide
About this Book
xii
About this Book
The MD Nastran Dynamic Analysis User’s Guide is a guide to the proper use of MD Nastran for solving
various dynamic analysis problems. This guide serves as both an introduction to dynamic analysis for
the new user and a reference for the experienced user. The major emphasis focuses on understanding the
physical processes in dynamics and properly applying MD Nastran to model dynamic processes while
restricting mathematical derivations to a minimum.
The basic types of dynamic analysis capabilities available in MD Nastran are described in this guide.
These common dynamic analysis capabilities include normal modes analysis, transient response analysis,
frequency response analysis, and enforced motion. These capabilities are described and illustrative
examples are presented. Theoretical derivations of the mathematics used in dynamic analysis are
presented only as they pertain to the proper understanding of the use of each capability.
To effectively use this guide, it is important for you to be familiar with MD Nastran’s static analysis
capability and the principles of dynamic analysis. Basic finite element modeling and analysis techniques
are covered only as they pertain to MD Nastran dynamic analysis. For more information on static
analysis and modeling, refer to the MD Nastran Linear Static Analysis User’s Guide and to the Getting
Started with MD Nastran User’s Guide.
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, fluidstructure
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) “ “ 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 Open crossdocument links in the same
window; then you would uncheck the and select “OK”.
alt ÷
Preferences Documents ÷ ÷ ÷
Main Index
xiii
CHAPTER
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
MD Nastran Dynamic Analysis User’s Guide
List of Nastran Books
xiv
List of Nastran Books
Below is a list of some of the Nastran documents. You may find any of these documents from
MSC.Software at www.simcompanion.mscsoftware.com.
Installation and Release Guides
• Installation and Operations Guide
• Release Guide
Guides
Reference Books
• Quick Reference Guide
• DMAP Programmer’s Guide
• Reference Manual
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
xv
CHAPTER
Preface
Technical Support
For technical support phone numbers and contact information, please visit:
http://www.mscsoftware.com/Contents/Services/TechnicalSupport/ContactTechnicalSupport.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
MD Nastran Dynamic Analysis User’s Guide
Internet Resources
xvi
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
1
Fundamentals of Dynamic Analysis
Overview
Equations of Motion
Dynamic Analysis Process
Dynamic Analysis Types
Main Index
MD Nastran Dynamic Analysis User’s Guide
Overview
2
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.
Dynamic Analysis Versus Static Analysis
Two basic aspects of dynamic analysis differ from static analysis. First, dynamic loads are applied as a
function of time or frequency. Second, this time or frequencyvarying load application induces time or
frequencyvarying response (displacements, velocities, accelerations, forces, and stresses). These time
or frequencyvarying characteristics make dynamic analysis more complicated and more realistic than
static analysis.
This chapter introduces the equations of motion for a single degreeoffreedom dynamic system (see
Equations of Motion, 3), illustrates the dynamic analysis process (see Dynamic Analysis Process, 13), and
characterizes the types of dynamic analyses described in this guide (see Dynamic Analysis Types, 15).
Those who are familiar with these topics may want to skip to subsequent chapters.
Main Index
3
CHAPTER 1
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:
(11)
Velocity and Acceleration
Velocity is the rate of change in the displacement with respect to time. Velocity can also be described as
the slope of the displacement curve. Similarly, acceleration is the rate of change of the velocity with
respect to time, or the slope of the velocity curve.
Single DegreeofFreedom System
The most simple representation of a dynamic system is a single degreeoffreedom (SDOF) system (see
Figure 11). In an SDOF system, the timevarying displacement of the structure is defined by one
component of motion . Velocity and acceleration are derived from the displacement.
Figure 11 Single DegreeofFreedom (SDOF) System
Dynamic and Static DegreesofFreedom
Mass and damping are associated with the motion of a dynamic system. Degreesoffreedom with mass
or damping are often called dynamic degreesoffreedom; degreesoffreedom with stiffness are called
static degreesoffreedom. It is possible (and often desirable) in models of complex systems to have
fewer dynamic degreesoffreedom than static degreesoffreedom.
u
·
du
dt
 v velocity = = =
u
··
d
2
u
dt
2

dv
dt
 a acceleration = = = =
u t ( ) u
·
t ( ) u
··
t ( )
m = mass (inertia)
b = damping (energy dissipation
k = stiffness (restoring force)
p = applied force
u = displacement of mass
= velocity of mass
= acceleration of mass
u
·
u
··
p t ( )
u t ( )
b
k
m
Main Index
MD Nastran Dynamic Analysis User’s Guide
Equations of Motion
4
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
(12)
The equation of motion accounts for the forces acting on the structure at each instant in time. Typically,
these forces are separated into internal forces and external forces. Internal forces are found on the
lefthand side of the equation, and external forces are specified on the righthand side. The resulting
equation is a secondorder linear differential equation representing the motion of the system as a function
of displacement and higherorder derivatives of the displacement.
Inertia Force
An accelerated mass induces a force that is proportional to the mass and the acceleration. This force is
called the inertia force .
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 . 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 .
Applied Load
The applied load on the righthand side of Eq. (12) 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.
Solution of the Equation of Motion
The solution of the equation of motion for quantities such as displacements, velocities, accelerations,
and/or stresses—all as a function of time—is the objective of a dynamic analysis. The primary task for
the dynamic analyst is to determine the type of analysis to be performed. The nature of the dynamic
analysis in many cases governs the choice of the appropriate mathematical approach. The extent of the
information required from a dynamic analysis also dictates the necessary solution approach and steps.
mu
··
t ( ) bu
·
t ( ) ku t ( ) + + p t ( ) =
mu
··
t ( )
bu
·
t ( )
ku t ( )
p t ( )
Main Index
5
CHAPTER 1
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
righthand side of Eq. (12) set to zero (i.e., no applied load). If damping is neglected, the solution is
known as undamped free vibration analysis.
Free Vibration Analysis
In undamped free vibration analysis, the SDOF equation of motion reduces to
(13)
Eq. (13) has a solution of the form
(14)
The quantity is the solution for the displacement as a function of time . As shown in Eq. (14),
the response is cyclic in nature.
Circular Natural Frequency
One property of the system is termed the circular natural frequency of the structure . The subscript
indicates the “natural” for the SDOF system. In systems having more than one mass degree of freedom
and more than one natural frequency, the subscript may indicate a frequency number. For an SDOF
system, the circular natural frequency is given by
(15)
The circular natural frequency is specified in units of radians per unit time.
Natural Frequency
The natural frequency is defined by
(16)
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 given by
(17)
The period of the response defines the length of time needed to complete one full cycle of response.
mu
··
t ( ) ku t ( ) + 0 =
u t ( ) A e
n
t sin B e
n
t cos + =
u t ( ) t
e
n
n
e
n
k
m
 =
f
n
f
n
e
n
2t
 =
T
n
T
n
1
f
n

2t
e
n
 = =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Equations of Motion
6
In the solution of Eq. (14), and are the integration constants. These constants are determined by
considering the initial conditions in the system. Since the initial displacement of the system
and the initial velocity of the system are known, and are evaluated by substituting their
values into the solution of the equation for displacement and its first derivative (velocity), resulting in
(18)
These initial value constants are substituted into the solution, resulting in
(19)
Eq. (19) 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 12.
Figure 12 SDOF System  Undamped Free Vibrations
If damping is included, the damped free vibration problem is solved. If viscous damping is assumed, the
equation of motion becomes
(110)
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 are
A B
u t 0 = ( )
u
·
t 0 = ( ) A B
B u t 0 = ( ) = and A
u
·
t 0 = ( )
e
n
 =
u t ( )
u
·
0 ( )
e
n
 e
n
t sin u 0 ( ) e
n
t cos + =
Time t
A
m
p
l
i
t
u
d
e
u
t
(
)
mu
··
t ( ) bu
·
t ( ) ku t ( ) + + 0 =
b
Main Index
7
CHAPTER 1
Fundamentals of Dynamic Analysis
• Critically damped
• Overdamped
• Underdamped
Critical damping occurs when the value of damping is equal to a term called critical damping . The
critical damping is defined as
(111)
For the critically damped case, the solution becomes
(112)
Under this condition, the system returns to rest following an exponential decay curve with no oscillation.
A system is overdamped when and no oscillatory motion occurs as the structure returns to its
undisplaced position.
The most common damping case is the underdamped case where . In this case, the solution has
the form
(113)
Again, and are the constants of integration based on the initial conditions of the system. The new
term represents the damped circular natural frequency of the system. This term is related to the
undamped circular natural frequency by the following expression:
(114)
The term is called the damping ratio and is defined by
(115)
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 13. 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.
b
cr
b
cr
2 km 2me
n
= =
u t ( ) A Bt + ( )e
bt – 2m
=
b b
cr
>
b b
cr
<
u t ( ) e
bt – 2m
A e
d
t sin B e
d
t cos + ( ) =
A B
e
d
e
d
e
n
1 .
2
– =
.
.
b
b
cr
 =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Equations of Motion
8
Figure 13 Damped Oscillation, Free Vibration
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. (14) 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).
Forced Vibration Analysis
Forced vibration analysis considers the effect of an applied load on the response of the system. Forced
vibrations analyses can be damped or undamped. Since most structures exhibit damping, damped forced
vibration problems are the most common analysis types.
The type of dynamic loading determines the mathematical solution approach. From a numerical
viewpoint, the simplest loading is simple harmonic (sinusoidal) loading. In the undamped form, the
equation of motion becomes
(116)
In this equation the circular frequency of the applied loading is denoted by . This loading frequency
is entirely independent of the structural natural frequency , although similar notation is used.
Time t
A
m
p
l
i
t
u
d
e
u
t
(
)
mu
··
t ( ) ku t ( ) + p et sin =
e
e
n
Main Index
9
CHAPTER 1
Fundamentals of Dynamic Analysis
This equation of motion is solved to obtain
(117)
where:
Again, and are the constants of integration based on the initial conditions. The third term in
Eq. (117) is the steadystate 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 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 is . In addition, to obtain the steady state solution, the static displacement is scaled
by the denominator.
The denominator of the steadystate solution contains the ratio between the applied loading frequency
and the natural frequency of the structure.
Dynamic Amplification Factor for No Damping
The term
(118)
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
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 14.
A =
B =
u t ( ) A e
n
t sin B e
n
t cos + =
p k
1 e
2
– e
n
2
 et sin +
Initial Condition
Solution
SteadyState
Solution
u
·
t 0 = ( )
e
n

ep k
1 e
2
– e
n
2
( )e
n
 –
u t 0 = ( )
A B
p k
u p k
1
1 e
2
– e
n
2

e e
n
Main Index
MD Nastran Dynamic Analysis User’s Guide
Equations of Motion
10
Figure 14 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
(119)
Time t
A
m
p
l
i
t
u
d
e
u
t
(
)
mu
··
t ( ) bu
·
t ( ) ku t ( ) + + p et sin =
Main Index
11
CHAPTER 1
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 steadystate response is
(120)
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:
(121)
The phase angle is called the phase lead, which describes the amount that the response leads the
applied force.
Dynamic Amplification Factor with Damping
The dynamic amplification factor for the damped case is
(122)
The interrelationship among the natural frequency, the applied load frequency, and the phase angle can
be used to identify important dynamic characteristics. If 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 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 , resonance occurs. In this case, the
magnification factor is , and the phase angle is 270 degrees. The dynamic amplification factor
and phase lead are shown in Figure 15 and are plotted as functions of forcing frequency.
u t ( ) p k
et u + ( ) sin
1 e
2
e
n
2
– ( )
2
2.e e
n
( )
2
+
 =
u
u tan
1 –
2.e e
n
1 e
2
– e
n
2
( )
 – =
u
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. (120) and Eq. (121).
u
u
1
1 e
2
e
n
2
– ( )
2
2.e e
n
( )
2
+

e e
n
e e
n
e e
n
1 =
1 2. ( )
Main Index
MD Nastran Dynamic Analysis User’s Guide
Equations of Motion
12
Figure 15 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.
360°
180°
1
e
n
Forcing Frequency e
P
h
a
s
e
L
e
a
d
u
(
D
e
g
r
e
e
s
)
A
m
p
l
i
f
i
c
a
t
i
o
n
F
a
c
t
o
r
Main Index
13
CHAPTER 1
Fundamentals of Dynamic Analysis
Dynamic Analysis Process
Before conducting a dynamic analysis, it is important to define the goal of the analysis prior to the
formulation of the finite element model. Consider the dynamic analysis process to be represented by the
steps in Figure 16. The analyst must evaluate the finite element model in terms of the type of dynamic
loading to be applied to the structure. This dynamic load is known as the dynamic environment. The
dynamic environment governs the solution approach (i.e., normal modes, transient response, frequency
response, etc.). This environment also indicates the dominant behavior that must be included in the
analysis (i.e., contact, large displacements, etc.). Proper assessment of the dynamic environment leads
to the creation of a more refined finite element model and more meaningful results.
Figure 16 Overview of Dynamic Analysis Process
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).
No
Yes
Yes
No
Finite Element
Model
Dynamic
Environment
Modal
Analysis?
Results
Satisfactory?
Results
Satisfactory?
ForcedResponse
Analysis
End
Yes
No
Main Index
MD Nastran Dynamic Analysis User’s Guide
Dynamic Analysis Process
14
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 forcedresponse 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
15
CHAPTER 1
Fundamentals of Dynamic Analysis
Dynamic Analysis Types
This guide describes the types of dynamic analysis that can be performed with MD Nastran. The basic
types are:
• Real eigenvalue analysis (undamped free vibrations).
• Linear frequency response analysis (steadystate response of linear structures to loads that vary
as a function of frequency).
• Linear transient response analysis (response of linear structures to loads that vary as a function
of time).
Real eigenvalue analysis is used to determine the basic dynamic characteristics of a structure. The results
of an eigenvalue analysis indicate the frequencies and shapes at which a structure naturally tends to
vibrate. Although the results of an eigenvalue analysis are not based on a specific loading, they can be
used to predict the effects of applying various dynamic loads. Real eigenvalue analysis is described in
Real Eigenvalue Analysis, 43.
Frequency response analysis is an efficient method for finding the steadystate 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 timevarying 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 timevarying (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
MD Nastran Dynamic Analysis User’s Guide
Dynamic Analysis Types
16
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 2: Finite Element Input Data
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
MD Nastran Dynamic Analysis User’s Guide
Overview
18
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 builtin 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
19
CHAPTER 2
Finite Element Input Data
Strategies for Dynamic Analysis
The two major approaches to avoid the problems described above are to use the matrix reduction
methods available in MD Nastran or to simply build another dynamic model with fewer grid points and
elements. The matrix reduction methods use sophisticated approximations to reduce the cost of the
dynamic solutions. However, they add complexity to the system. Instead of making decisions about the
element mesh size, the user must consider the frequency band and loading characteristics of the actual
structure.
Although building a separate dynamic model will require extra work, it has several advantages. The
requirements for precise stress results for statics are usually not as important for dynamics, thereby
allowing the use of larger elements. The requirements for accurate dynamic results are not as dependent
on small elements since loads tend to be distributed over wider areas by the inertial and damping effects.
The actual accuracy criteria for dynamic models is related to wavelength size relative to the mesh size.
Also, in a smaller dynamic model the job of interpreting outputs with plotters and other methods
becomes much easier.
It should also be noted that the process of changing finite element meshes is becoming easier with
modern geometrybased modeling systems and automated mesh generation.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Plan the Analysis
20
Plan the Analysis
The following steps are recommended before starting a dynamic analysis:
1. Estimate the frequency range of the structure. The first mode frequency can usually be estimated
by a simple equivalent beam calculation. A normal modes analysis is highly recommended for
dynamics regardless of the approach.
2. Consider the type and frequency content of the loads. Localized high frequency sources such as
brake squeal require different approaches from low frequency distributed loads such as wind
forces on a bridge.
3. Use expected wavelengths to estimate required mesh size. A wavelength can be estimated from
the frequency and sound speed. Six elements per wave for approximately 10% accuracy is a
general ruleofthumb. Also, this method may be used to evaluate the quality of the calculated
eigenvectors.
4. Always use SETs for output requests as a general practice. The use of Case Control requests such
as STRESS = ALL for a large transient analysis could possibly exceed the size limits of your
postprocessor.
5. Estimate the computer resources (CPU time, database size, output requirements) for large
problems before production runs are made. A good method is to run the model with only a few
time steps or frequencies, and examine the major time and size messages in the F04 output file.
The actual costs can then be extrapolated with reasonable accuracy.
Main Index
21
CHAPTER 2
Finite Element Input Data
Mass Input
Basic Definitions of Mass
The mass matrix in MD Nastran may contain much more information than simple structural mass and
inertia terms. In fact, it may contain control system terms, fluid compressibility, or electromagnetic
capacitance. One basic definition is that any term which contributes to the generalized kinetic energy
must create a coefficient in the mass matrix. Another definition is that any generalized force, ,
proportional to an acceleration term produces a mass term , i.e.,
(21)
where each component of the acceleration vector represents a generalized degreeoffreedom.
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.
Lumped and Coupled Mass
Mass is formulated as either lumped mass or coupled mass. Lumped mass matrices contain uncoupled,
translational components of mass. Coupled mass matrices contain translational components of mass
with coupling between the components. The CBAR, CBEAM, and CBEND elements contain rotational
masses in their coupled formulations, although torsional inertias are not considered for the CBAR
element. Coupled mass can be more accurate than lumped mass. However, lumped mass is more
efficient and is preferred for its computational speed in dynamic analysis.
The mass matrix formulation is a userselectable option in MD Nastran. The default mass formulation
is lumped mass for most MD Nastran finite elements. The coupled mass matrix formulation is selected
using PARAM,COUPMASS,1 in the Bulk Data. Table 21 shows the mass options available for each
element type.
F
u
··
M
F { } M   u
··
{ } =
u
··
{ }
Main Index
MD Nastran Dynamic Analysis User’s Guide
Mass Input
22
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:
Table 21 Element Mass Types
Element Type Lumped Mass Coupled Mass*
CBAR X X
CBEAM X X
CBEND X
CFAST X
CONEAX X
CONMi X X
CONROD X X
CRAC2D X X
CRAC3D X X
CHEXA X X
CMASSi X
CPENTA X X
CQUAD4 X X
CQUAD8 X X
CQUADR X X
CROD X X
CSHEAR X
CTETRA X X
CTRIA3 X X
CTRIA6 X X
CTRIAR X X
CTRIAX6 X X
CTUBE X X
*Couple mass is selected by PARAM,COUPMASS,1.
Main Index
23
CHAPTER 2
Finite Element Input Data
CROD Element Stiffness Matrix
The CROD element’s stiffness matrix is given by:
(22)
The zero entries in the matrix create independent (uncoupled) translational and rotational behavior for
the CROD element, although for most other elements these degreesoffreedom are coupled.
CROD Lumped Mass Matrix
The CROD element classical lumped mass matrix is the same as the MD Nastran lumped mass matrix.
This lumped mass matrix is
(23)
The lumped mass matrix is formed by distributing onehalf of the total rod mass to each of the
translational degreesoffreedom. These degreesoffreedom are uncoupled and there are no torsional
mass terms calculated.
L = Length
A = Area
J = Torsional Constant
E = Young’s Modulus
G = Shear Modulus
= Mass Density
= Polar Moment of Inertia I
14 = DegreesofFreedom
p
Ip
L
2 (Torsion)
1 (Translation)
4 (Torsion)
3 (Translation)
K  
K  
AE
L
 0
AE –
L
 0
0
GJ
L
 0
GJ –
L

AE –
L
 0
AE
L
 0
0
GJ –
L
 0
GJ
0

=
1
2
3
4
1 2 3 4
M   pAL
1
2
 0 0 0
0 0 0 0
0 0
1
2
 0
0 0 0 0
=
Main Index
MD Nastran Dynamic Analysis User’s Guide
Mass Input
24
The CROD element classical consistent mass matrix is
(24)
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.
CROD Coupled Mass Matrix
The CROD element MD Nastran coupled mass matrix is
(25)
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.
Lumped Mass Versus Coupled Mass Example
The difference in the axial mass formulations can be demonstrated by considering a fixedfree rod
modeled with a single CROD element (Figure 21). The exact quarterwave natural frequency for the
first axial mode is
M   pAL
1
3
 0
1
6
 0
0
I
p
3A
 0
I
p
6A

1
6
 0
1
3
 0
0
I
p
6A
 0
I
p
3A

=
M   pAL
5
12
 0
1
12
 0
0 0 0 0
1
12
 0
5
12
 0
0 0 0 0
=
1.5708
E p
l

Main Index
25
CHAPTER 2
Finite Element Input Data
Using the lumped mass formulation for the CROD element, the first frequency is predicted to be
which underestimates the frequency by 10%. Using a classical consistent mass approach, the predicted
frequency
is overestimated by 10%. Using the coupled mass formulation in MD Nastran, the frequency
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.
Figure 21 Comparison of Mass Formulations for a ROD
1.414
E p
l

1.732
E p
l

1.549
E p
l

MD Nastran Lumped Mass:
Classical Consistent Mass:
MD Nastran Coupled Mass:
Theoretical Natural Frequency:
e
n
2
E p
l
 1.414
E p
l
 = =
e
n
3
E p
l
 1.732
E p
l
 = =
e
n
12 5
E p
l
 1.549
E p
l
 = =
u t ( )
1
Single Element Model
2
l
e
n
t
2

E p
l
 1.5708
E p
l
 = =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Mass Input
26
CBAR, CBEAM Lumped Mass
By default, the CBAR element lumped mass matrix is identical to the CROD element lumped mass
matrix. The CBEAM element lumped mass matrix is identical to that of the CROD and CBAR mass
matrices with the exception that torsional inertia is included. The CBAR can also include torsional inertia
if BARMASS (398)=1. Torsional inertia can be removed for the CBAR and CBEAM by setting
SYSTEM(414)=1.
CBAR, CBEAM Coupled Mass
The CBAR element coupled mass matrix is identical to the classical consistent mass formulation except
for two terms: (1) the mass in the axial direction is the average of the lumped and classical consistent
masses, as explained for the CROD element; and (2) there is no torsional inertia. The CBEAM element
coupled mass matrix is also identical to the classical consistent mass formulation except for two terms:
(1) the mass in the axial direction is the lumped mass; and (2) the torsional inertia is the lumped inertia.
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:
(26)
where:
The parameter
PARAM,WTMASS,factor
performs this conversion. The value of the factor should be entered as . 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 , using
PARAM,WTMASS,0.002588 converts the weight density to mass density for the acceleration of gravity
. The mass density, therefore, becomes . If the weight density of steel
= mass or mass density
= acceleration of gravity
= weight or weight density
p
m
1 g ( )p
w
=
p
m
g
p
w
1 g
RHO 0.3 lb/in
3
=
g 386.4 in/sec
2
= 7.765E4 lb
f
sec
2
/in
4
Main Index
27
CHAPTER 2
Finite Element Input Data
is entered as when using metric units, then using PARAM,WTMASS,0.102 converts
the weight density to mass density for the acceleration of gravity . The mass density,
therefore, becomes .
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.
Mass Data Input Options
MD Nastran provides the following means to specify the mass properties of the finite element model:
1. The density (mass per unit volume) of the structural materials, which comprises the finite element
(field RHO on the MATi Bulk Data entries).
2. Nonstructural mass per unit length of line elements or mass per unit area of surface elements
(field NSM on the property Bulk Data entry). Examples of this feature are coatings and thermal
insulating materials.
3. Concentrated mass and inertia terms at a grid point via a CONM2 Bulk Data entry. The
provisions of the CONM2 entry are the mass, the offset of the center of mass from the grid point,
and the moments and products of inertia about the center of mass. As an option, the center of
mass may be measured from the origin of the basic coordinate system rather than as an offset from
the grid point.
4. A full symmetric matrix of mass coefficients at a grid point via the CONM1 Bulk Data
entry.
5. Mass coupling between any two degreesoffreedom via the CMASSi (i = 1,2,3,4) Bulk Data
entries. The form of the relationship is
(27)
where and are the inertia forces acting at degreesoffreedom 1 and 2, respectively, and
is the mass coefficient, specified on the CMASSi entry (or on the PMASS entry if i = 2 or 4).
In most applications, the second degreeoffreedom is not specified. In this case, the entry
generates the inertia force , and is added to the mass matrix in the diagonal position
corresponding to . 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.
RHO 80000 N/m
3
=
g 9.8 m/sec
2
=
8160 kg/m
3
6 6 ×
f
1
f
2
¹ )
´ `
¦ ¹
M M –
M – M
u
··
1
u
··
2
¹ )
´ `
¦ ¹
– =
f
1
f
2
M
f
1
Mu
··
1
– = M
u
1
Main Index
MD Nastran Dynamic Analysis User’s Guide
Mass Input
28
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 is entered as follows:
Nonstructural Mass
An additional way to input mass is to use nonstructural mass, which is mass not associated with the
geometric crosssectional 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 twodimensional geometry. Nonstructural mass is defined on
the element property entry. Such as the NSM field on the PBAR entry for example.
Concentrated Mass and Inertia
Grid point masses can be entered using the CONM1, CONM2, and CMASSi entries. The CONM1 entry
allows input of a fully coupled 6 x 6 mass matrix. You define half of the terms, and symmetry is assumed.
The CONM2 entry defines mass and mass moments of inertia for a rigid body. The CMASSi entries
define scalar masses.
Three Related Parameters
The bulk data input also includes three parameters which relate to the specification of mass properties.
They are:
1. PARAM,WTMASS,V1 – Specifies a factor by which to multiply the mass of the structure to
obtain dimensionally correct mass. For example, if the ftlbsec system is used, and the mass
input unit is pounds, then V1 = 1/32.174 = .031081. This factor operates on all mass terms (except
those defined on DMIG entries; PARAM,CM2,V1 may be used for these instead).
2. PARAM,COUPMASS,1 – Requests that the coupled mass option be used rather than the lumped
mass option for all elements which support the coupled mass option. In the lumped mass option,
the distributed mass of an element is simply divided up and the parts are located at the connected
grid points. In the coupled mass option, the distributed mass is replaced by a nondiagonal mass
matrix which couples the connected grid points. The latter method is sometimes more accurate,
but it uses more computer time. See the Table 21 for descriptions of the elements.
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 7.76E4
7.76E4 lb
f
sec
2
/in
4
Main Index
29
CHAPTER 2
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
MD Nastran Dynamic Analysis User’s Guide
Coupled Mass Matrix Terms
30
Coupled Mass Matrix Terms
A common difficulty for finite element beginners is to comprehend the generation of offdiagonal or
coupled terms in the mass matrix. These are caused by the following actions.
1. With PARAM,COUPMASS,1 the finite elements themselves generate coupled terms to represent
the distributed element mass more accurately than simple lumped masses on the grid points. The
total kinetic energy of the element is represented by shape functions, which in turn, are coupled
functions of displacement and rotation.
2. The internal matrix operations in the solution sequences produce coupled mass terms. MPCs and
RBEtype elements cause the mass to be redistributed from the dependent degreesoffreedom to
the remaining independent points. The ASET, OMIT, and Superelement reduction operations
may automatically fill the entire mass matrix. These operations will maintain exact total mass and
CG properties and provide a good approximation to the inertia effects of the eliminated points,
but they can be costly.
3. Direct user inputs such as CMASSi scalar elements and CONMi mass elements may produce
offdiagonal mass terms. Other inputs that cause coupling are direct matrix inputs (DMIG), and
transfer functions (TF).
An example of a coupled scalar mass is illustrated in the example below:
A spring, , and two masses, and , are attached to a structural model. However, when modeling
fluids or other special connections, the user may wish to use the difference in displacements,
, instead of as the primary degreeoffreedom.
We may find the equivalent coupled mass matrix by the use of energy methods. The
(28)
k
Structure
P
1
m
1
u
1
Au
m
2
P
2
u
2
k m
1
m
2
Au u
1
u
2
– = u
2
V kAu
2
2 =
T m
1
u
··
1
2
m
2
u
··
1
Au
··
– ( )
2
+ =
Main Index
31
CHAPTER 2
Finite Element Input Data
From basic energy principles, the resulting equilibrium equations are
(29)
The mass matrix for the two degreesoffreedom becomes
(210)
where are the active degreesoffreedom. Two CMASSi elements may be used; one will
be coupled to two points. Note that the spring, , is now connected to 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
as the MPC equation, and the reduced mass and stiffness matrices will be identical to
the system above.
. m
1
u
··
1
m
2
u
··
1
Au
··
– ( ) + + P
1
P
2
+ =
kAu m
2
u
··
1
Au
··
– ( ) – P
2
– =
and
m  
m
1
m
2
+ m
2
–
m
2
– m
2
=
u u
1
Au . =
k Au
u
2
u
1
– Au + 0 =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Mass Effects in Moving Coordinate Systems
32
Mass Effects in Moving Coordinate Systems
In most cases the motions of the grid points in MD Nastran are measured in a set of fixed coordinate
systems. Even large displacement nonlinear analysis uses displacements measured in fixed directions.
However, there are several cases in which it is more convenient to use a moving coordinate system.
Examples follow:
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, , the absolute velocity vector of the point, , in fixed coordinates is
(211)
The first term on the righthand side of Eq. (211) is the static centrifugal force; the second term is the
centripetal stiffness; the third term is the Coriolis force; the last term, , is the relative acceleration
vector.
Another term that is calculated for spinning bodies is the socalled 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. (212). (Centrifugal stiffness and differential stiffness terms are of the same magnitude.)
The basic matrix equation for the forces in the moving system is
(212)
The special MD Nastran process assembles the terms in Eq. (211) into matrices, that when substituted
into Eq. (212), result in
(213)
where:
= generates the velocitydependent Coriolis forces. Note that is not symmetric!
= the centripetal stiffness matrix
= the differential stiffness matrix
O
r V
V O r u + ( ) u
·
+ × =
u
M   A { } B   V { } K   u { } + + P { } =
M   u
··
{ } B B
c
+   u
·
{ } K K
c
K
d
+ +   u { } + + P t ( ) { } =
B
c
 
B
c
K
c
 
K
d
 
Main Index
33
CHAPTER 2
Finite Element Input Data
The system defined by Eq. (213) 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 . Unstable conditions usually occur when , where is a positive integer. By
definition, the system is unstable if 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
(214)
where are displacements relative to the moving system and are the steady accelerations to be
determined from the mass and loads. If is a matrix whose columns define the rigid body motions
of the structure, then for a free body,
(215)
where is called the rigid body transformation matrix. However, since the fullsized vector, ,
is
a rigid body motion, it may be defined in terms of accelerations at a set of reference coordinates, ,
by the equation
(216)
Combining Eq. (216) into Eq. (215) and eliminating , we obtain
(217)
where the total mass matrix for the reference coordinates is
(218)
The resulting set of equations defined in Eq. (214) may now be arbitrarily constrained since the total
load is balanced by the inertia forces.
Two different methods are used in MD Nastran to calculate the rigid body matrix 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
degreesoffreedom. Any number of degreesoffreedom that provide a nonredundant set of supports
may be used. This option allows partially free bodies and extra mechanisms. The other method
p
n
o
n
i e
n
± = e
n
PO ~ P
o
n
K   u { } P { } M   a
o
{ } – =
u { } a
o
{ }
D  
D  
T
K   u { } 0 { } D  
T
P { } D  
T
M   a
o
{ } – = =
D   a
o
{ }
a
r
{ }
a
o
{ } D   a
r
{ } =
a
r
{ }
a
o
{ } D   m  
1 –
D  
T
P { } =
m   D  
T
M   D { } =
D  
u
r
Main Index
MD Nastran Dynamic Analysis User’s Guide
Mass Effects in Moving Coordinate Systems
34
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 semirigid base that causes a known stressfree motion,
,
the total
structural motion,
,
is
(219)
where are displacements relative to the base motion. If the structure is not constrained elsewhere,
we may assume that the base displacements produce no force and
(220)
If we move the known base motion to the right hand side it looks almost identical to a gravity load:
(221)
u
0
{ }
u
A
{ }
u
A
{ } u
0
{ } u
g
{ } + =
u
g
{ }
u
0
{ }
M   u
··
g
u
··
0
+ { } K   u
g
{ } + P
g
{ } =
u
··
0
{ }
M   u
··
g
{ } K   u
g
{ } + P
g
{ } M   u
··
0
t ( ) { } – =
Main Index
35
CHAPTER 2
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.
Viscous and Structural Damping
Two types of damping are generally used for linearelastic materials: viscous and structural. The viscous
damping force is proportional to velocity, and the structural damping force is proportional to
displacement. Which type to use depends on the physics of the energy dissipation mechanism(s) and is
sometimes dictated by regulatory standards.
The viscous damping force is proportional to velocity and is given by
(222)
where:
The structural damping force is proportional to displacement and is given by
(223)
where:
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 22 depicts this and also
= viscous damping coefficient
= velocity
= structural damping coefficient
= stiffness
= displacement
= (phase change of 90 degrees)
f
v
f
v
bu
·
=
b
u
·
f
s
f
s
i G k u · · · =
G
k
u
i
1 –
Main Index
MD Nastran Dynamic Analysis User’s Guide
Damping Input
36
shows that for constant amplitude sinusoidal motion the two damping forces are equal at a single
frequency.
At this frequency,
(224)
where is the frequency at which the structural and viscous damping forces are equal for a constant
amplitude of sinusoidal motion.
Figure 22 Structural Damping and Viscous Damping Forces for Constant Amplitude
Sinusoidal Displacement
If the frequency is the circular natural frequency , Eq. (224) becomes
(225)
Recall the definition of critical damping from Eq. (111)
(226)
Some equalities that are true at resonance ( ) for constant amplitude sinusoidal displacement are
(227)
G k b e

or b
Gk
e

 = =
e

f
Forcing Frequency
Structural Damping
Viscous Damping
f
v
bu
·
i b e u = =
f
s
i G k u =
e

e
D
a
m
p
i
n
g
F
o
r
c
e
e

e
n
b
G k
e
n
 G e
n
m = =
b
cr
2 km 2me
n
= =
e
n
b
b
cr
 .
G
2
 = =
Main Index
37
CHAPTER 2
Finite Element Input Data
and (228)
where is the quality or dynamic magnification factor, which is inversely proportional to the energy
dissipated per cycle of vibration.
The Effect of Damping
Damping is the result of many complicated mechanisms. The effect of damping on computed response
depends on the type and loading duration of the dynamic analysis. Damping can often be ignored for
short duration loadings, such as those resulting from a crash impulse or a shock blast, because the
structure reaches its peak response before significant energy has had time to dissipate. Damping is
important for long duration loadings (such as earthquakes), and is critical for loadings (such as rotating
machinery) that continually add energy to the structure. The proper specification of the damping
coefficients can be obtained from structural tests or from published literature that provides damping
values for structures similar to yours.
As is discussed in detail in Frequency Response Analysis, 133 and Transient Response Analysis, 201,
certain solution methods allow specific forms of damping to be defined. The type of damping used in
the analysis is controlled by both the solution being performed and the MD Nastran data entries. In
transient response analysis, for example, structural damping must be converted to equivalent viscous
damping.
Structural Damping Specification
Structural damping is specified on the MATi and PARAM,G, GFL Bulk Data entries. The GE field on
the MATi entry is used to specify overall structural damping for the elements that reference this material
entry. This definition is via the structural damping coefficient GE.
For example, the MAT1 entry:
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.
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 7.764E4 0.10
Q
1
2.

1
G
 = =
Q
Main Index
MD Nastran Dynamic Analysis User’s Guide
Damping Input
38
Viscous Damping Specification
Viscous damping is defined by the following elements:
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
pelements. The same, including forces, have been implemented for shell and beam pelements.
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 pelements 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 pdistribution may
CDAMP1 entry Scalar damper between two degreesoffreedom (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 springanddamper structural element that may be nonlinear or
frequency dependent. It references a PBUSH entry.
Main Index
39
CHAPTER 2
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 pelements. The appropriate functions for
pelements 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 pelements 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
MD Nastran Dynamic Analysis User’s Guide
Units in Dynamic Analysis
40
Units in Dynamic Analysis
Because MD Nastran does not assume a particular set of units, you must ensure that the units in your
MD Nastran model are consistent. Because there is more input in dynamic analysis than in static
analysis, it is easier to make a mistake in units when performing a dynamic analysis. The most frequent
source of error in dynamic analysis is incorrect specification of the units, especially for mass and
damping.
Table 22 shows typical dynamic analysis variables, fundamental and derived units, and common English
and metric units. Note that for English units all “lb” designations are . The use of “lb” for mass (i.e.,
) is avoided.
Table 22 Engineering Units for Common Variables
Variable Dimensions*
Common
English Units
Common
Metric Units
Length L in m
Mass M kg
Time T sec sec
Area
Volume
Velocity in / sec m / sec
Acceleration
Rotation  rad rad
Rotational Velocity rad / sec rad / sec
Rotational Acceleration
Circular Frequency rad / sec rad / sec
Frequency cps; Hz cps; Hz
Eigenvalue
Phase Angle  deg deg
Force lb N
Weight lb N
Moment inlb Nm
Mass Density
Young’s Modulus
Poisson’s Ratio   
Shear Modulus
Area Moment of Inertia
lb
f
lb
m
lbsec
2
in
L
2
in
2
m
2
L
3
in
3
m
3
LT
1 –
LT
2 –
in sec
2
m sec
2
T
1 –
T
2 –
rad sec
2
rad sec
2
T
1 –
T
1 –
T
2 –
rad
2
sec
2
rad
2
sec
2
MLT
2 –
MLT
2 –
ML
2
T
2 –
ML
3 –
lbsec
3
in
4
kg m
3
ML
1 –
T
2 –
lb in
2
Pa; N m
2
ML
1 –
T
2 –
lb in
2
Pa; N m
2
L
4
in
4
m
4
Main Index
41
CHAPTER 2
Finite Element Input Data
Torsional Constant
Mass Moment of Inertia
Stiffness N / m
Viscous Damping Coefficient lbsec / in Nsec / m
Stress
Strain   
* L Denotes length
M Denotes mass
T Denotes time
 Denotes dimensionless
Table 22 Engineering Units for Common Variables (continued)
Variable Dimensions*
Common
English Units
Common
Metric Units
L
4
in
4
m
4
ML
2
inlbsec
2
kgm
2
MT
2 –
lb in
MT
1 –
ML
1 –
T
2 –
lb in
2
Pa; N m
2
Main Index
MD Nastran Dynamic Analysis User’s Guide
Units in Dynamic Analysis
42
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 3: Real Eigenvalue Analysis
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
MD Nastran Dynamic Analysis User’s Guide
Overview
44
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 31). 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 32).
Figure 31 The First Four Mode Shapes of a Cantilever Beam
x
y
z
4
x
y
z
1
x
y
z
2
x
y
z
3
Main Index
45
CHAPTER 3
Real Eigenvalue Analysis
Figure 32 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.
x
y
z
4
x
y
z
1
x
y
z
2
x
y
z
3
Main Index
MD Nastran Dynamic Analysis User’s Guide
Reasons to Compute Normal Modes
46
Reasons to Compute Normal Modes
There are many reasons to compute the natural frequencies and mode shapes of a structure. One reason
is to assess the dynamic interaction between a component and its supporting structure. For example, if a
rotating machine, such as an air conditioner fan, is to be installed on the roof of a building, it is necessary
to determine if the operating frequency of the rotating fan is close to one of the natural frequencies of the
building. If the frequencies are close, the operation of the fan may lead to structural damage or failure.
Decisions regarding subsequent dynamic analyses (i.e., transient response, frequency response, response
spectrum analysis, etc.) can be based on the results of a natural frequency analysis. The important modes
can be evaluated and used to select the appropriate time or frequency step for integrating the equations
of motion. Similarly, the results of the eigenvalue analysisthe natural frequencies and mode shapescan
be used in modal frequency and modal transient response analyses (see Frequency Response Analysis,
133 and Transient Response Analysis, 201).
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
47
CHAPTER 3
Real Eigenvalue Analysis
Mathematical Overview of Normal Modes Analysis
The solution of the equation of motion for natural frequencies and normal modes requires a special
reduced form of the equation of motion. If there is no damping and no applied loading, the equation of
motion in matrix form reduces to
(31)
where:
This is the equation of motion for undamped free vibration. To solve Eq. (31) assume a harmonic
solution of the form
(32)
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 degreesoffreedom 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:
(33)
which after simplifying becomes
(34)
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
(35)
= mass matrix
= stiffness matrix
= displacement
= acceleration
= the eigenvector or mode shape
= is the circular natural frequency
M   u
··
{ } K   u   + 0 =
M  
K  
u  
u
··
{ }
u { } o { } et sin =
o { }
e
e
2
M   o { } et sin K   o { } et sin + – 0 =
K   e
2
M   – ( ) o { } 0 =
A ìI –  x 0 =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Mathematical Overview of Normal Modes Analysis
48
where:
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 , , and as shown in Eq. (34) with .
There are two possible solution forms for Eq. (34):
1. If det , the only possible solution is
(36)
This is the trivial solution, which does not provide any valuable information from a physical point
of view, since it represents the case of no motion. (“det” denotes the determinant of a matrix.)
2. If det , then a nontrivial solution ( ) is obtained for
From a structural engineering point of view, the general mathematical eigenvalue problem
reduces to one of solving the equation of the form
(37)
or
(38)
where
The determinant is zero only at a set of discrete eigenvalues or . There is an eigenvector which
satisfies and corresponds to each eigenvalue. Therefore, can be rewritten as
(39)
Each eigenvalue and eigenvector define a free vibration mode of the structure. The ith eigenvalue is
related to the ith natural frequency as follows:
(310)
A
= square matrix
= eigenvalues
I
= identity matrix
x
= eigenvector
ì
K e M e
2
ì =
K   e
2
M   – ( ) 0 =
o { } 0 =
K   e
2
M   – ( ) 0 = o { } 0 =
K   e
2
M   – ( ) o { } 0 =
det K   e
2
M   – ( ) 0 =
det K   ì M   – ( ) 0 =
ì e
2
=
ì
i
e
i
2
o
i
{ }
K e
i
2
M –   o
i
{ } 0 i 1 2 3. . . = =
ì
i
f
i
e
i
2t
 =
Main Index
49
CHAPTER 3
Real Eigenvalue Analysis
where:
The number of possible eigenvalues and eigenvectors is equal to the number of degreesoffreedom that
have mass or the number of dynamic degreesoffreedom.
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
(311)
where:
Second, if and are symmetric and real (as is the case for all the common structural finite
elements), the following mathematical properties hold:
(312)
(313)
and
(314)
(315)
Also, from Eq. (313) and Eq. (315) Rayleigh’s equation is obtained
(316)
Eq. (312) and Eq. (314) 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
= ith natural frequency
=
= vector of physical displacements
= ith mode shape
= ith modal displacement
f
i
e
i
ì
i
u { } o
i
( )c
i
i
¯
=
u { }
o
i
{ }
c
i
K   M  
o
i
{ }
T
M   o
j
{ } 0 if i j = =
o
j
{ }
T
M   o
j
{ } m
j
jth generalized mass = =
o
j
{ }
T
K   o
j
{ } 0 if i j = =
o
j
{ }
T
K   o
j
{ } k
j
jth generalized stiffness e
2
m
j
= = =
e
j
2
o
j
{ }
T
K   o
j
{ }
o
j
{ }
T
M   o
j
{ }
 =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Mathematical Overview of Normal Modes Analysis
50
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).
Rigid Body Modes Introduction
If a structure is not totally constrained in space, it is possible for the structure to displace (move) as a rigid
body or as a partial or complete mechanism. For each possible component of rigidbody motion or
mechanism, there exists one natural frequency that is equal to zero. The zerofrequency modes are called
rigidbody modes. Rigidbody motion of all or part of a structure represents the motion of the structure
in a stressfree condition. Stressfree, rigidbody modes are useful in conducting dynamic analyses of
unconstrained structures, such as aircraft and satellites. Also, rigidbody modes can be indicative of
modeling errors or an inadequate constraint set.
For example, the simple unconstrained structure in Figure 33 has a rigidbody mode.
Figure 33 RigidBody Mode of a Simple Structure
When both masses move the same amount (as a rigid body), there is no force induced in the connecting
spring. A detailed discussion of rigidbody 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 timevarying 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 twoDOF structure are shown in
Figure 34. The graphical representation of the eigenvectors in the figure shows the modal displacements
rotated by 90 degrees in order to view the deformation better.
e
1
0 =
o
1
{ }
u
1
u
2
¹ )
´ `
¦ ¹ 1
1
¹ )
´ `
¦ ¹
= =
k
m
2
m
1
u
1
u
2
Main Index
51
CHAPTER 3
Real Eigenvalue Analysis
Figure 34 Representations of Mode Shapes for a TwoDOF System
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.
Mode Shape Normalization
Although the scaling of normal modes is arbitrary, for practical considerations mode shapes should be
scaled (i.e., normalized) by a chosen convention. In MD Nastran there are three normalization choices,
MASS, MAX, and POINT normalization.
MASS normalization is the default method of eigenvector normalization. This method scales each
eigenvector to result in a unit value of generalized mass
(317)
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
Mode 1 Mode 2
u
1
u
1
u
1
u
2
u
2
u
2
m
1
m
2
k
2
k
1
o
1
u
1
u
2 ¹ )
´ `
¦ ¹ 3
1
¹ )
´ `
¦ ¹ 300
100
¹ )
´ `
¦ ¹ 0.6 –
0.2 –
¹ )
´ `
¦ ¹
o
2
{ }
u
1
u
2 ¹ )
´ `
¦ ¹ 1.6 –
0.8
¹ )
´ `
¦ ¹ 160
80 –
¹ )
´ `
¦ ¹ 1.0
0.5 –
¹ )
´ `
¦ ¹
= = = = = = = =
o
j
{ }
T
M   o
j
{ } 1.0 =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Mathematical Overview of Normal Modes Analysis
52
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 aset component.
(Dynamic Reduction, 371 and The Set Notation System Used in Dynamic Analysis, 503 provide
discussions of the aset.) This normalization results in the largest aset 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 nonnormalized 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. Rigidbody modes (see Rigid Body Modes, 99) represent a special case of repeated roots.
Main Index
53
CHAPTER 3
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 largesized problems, since it has a performance advantage over other methods.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Methods of Computation
54
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.
Givens and Householder Methods
The Givens and Householder modal extraction methods require a positive definite mass matrix (all
degreesoffreedom must have mass). There is no restriction on the stiffness matrix except that it must
be symmetric. These matrices always result in real (positive) eigenvalues. The Givens and Householder
methods are the most efficient methods for small problems and problems with dense matrices when a
large portion of the eigenvectors are needed. These methods find all of the eigenvalues and as many
eigenvectors as requested. While these methods do not take advantage of sparse matrices, they are
efficient with the dense matrices sometimes created using dynamic reduction (see Dynamic Reduction,
371).
The Givens and Householder methods fail if the mass matrix is not positive definite. To minimize this
problem, degreesoffreedom with null columns are removed by the automatic application of static
condensation (see Dynamic Reduction, 371) called autoomit. Applying the autoomit 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.
Modified Givens and Modified Householder Methods
The modified Givens and modified Householder methods are similar to their standard methods with the
exception that the mass matrix can be singular. Although the mass matrix is not required to be
nonsingular in the modified methods, a singular mass matrix can produce one or more infinite
eigenvalues. Due to roundoff error, these infinite eigenvalues appear in the output as very large positive
or negative eigenvalues. To reduce the incidence of such meaningless results, degreesoffreedom with
null masses are eliminated by automatic static condensation as in the case of the unmodified methods.
The modified methods require more computer time than the standard methods.
Automatic Givens and Automatic Householder Methods
Many times you may not know whether the mass matrix is singular. To assist you in choosing the
appropriate method, two options–automatic Givens and automatic Householder–are available. Initially
the automatic methods use the standard methods. In the first step of the method, if the mass matrix is not
wellconditioned for decomposition, the method shifts to the corresponding modified method. The
modified methods are more expensive and may introduce numerical noise due to the shift, but they
resolve most of the numerical problems of the illconditioned mass matrix. The automatic methods,
Main Index
55
CHAPTER 3
Real Eigenvalue Analysis
therefore, use the modified methods when necessary for numerical stability but use the standard methods
when the numerical stability is accurate.
Inverse Power Method
The inverse power method is a tracking method since the lowest eigenvalue and eigenvector in the
desired range are found first. Then their effects are “swept” out of the dynamic matrix, the next higher
mode is found, and its effects are “swept” out, and so on. Hence, the term “tracking,” which means that
one root at a time is found. In addition, each root is found via an iterative procedure. (The classical
literature often refers to this method as the inverse iteration method with sweeping.)
However, the inverse power method can miss modes, making it unreliable. The Sturm modified inverse
power method is a more reliable tracking method.
Sturm Modified Inverse Power Method
This method is similar to the inverse power method except that it 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 designed
range are computed. This process helps to ensure that modes are not missed.
The Sturm modified inverse power method is useful for models in which only the lowest few modes are
needed. This method is also useful as a backup method to verify the accuracy of other methods.
Main Index
MD Nastran Dynamic Analysis User’s Guide
User Interface for Real Eigenvalue Analysis
56
User Interface for Real Eigenvalue Analysis
The EIGR and EIGRL Bulk Data entries define the method and select the parameters that control the
eigenvalue extraction procedure. The EIGRL entry is used for the Lanczos method, and the EIGR entry
is used for all of the other methods.
User Interface for the Lanczos Method
The Bulk Data entry EIGRL has the following format:
Format:
Example:
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.
EIGRL 1 0.1 3.2 10
NORM=MAX NUMS=2
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, )
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.)
5 – 10
16
× V1 s V2 < 5. s 10
16
×
Main Index
57
CHAPTER 3
Real Eigenvalue Analysis
Examples of the results of using explicit or default values for the V1, V2, and ND fields are shown in
Table 31. The defaults on the EIGRL entry are designed to provide the minimum number of roots in
cases where the input is ambiguous.
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
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):
NUMS Number of frequency segments for the parallel method. (Integer > 0; Default = 1)
Fi Frequency at the upper boundary of the ith segment. See Remark 13. (Real or blank;
)
option_i=
value_i
Assigns a value to the fields above except for SID. ALPH, NUMS, and Fi must be
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.
Table 31 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
Field Contents
Fi V2 V1 – ( )
1 ALPH
i
–
1. ALPH
NUMS
–
 =
V1 F1 F2 .F15 V2 < < < <
Main Index
MD Nastran Dynamic Analysis User’s Guide
User Interface for Real Eigenvalue Analysis
58
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 userdesignated 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., ). The MAX method normalizes to a unit value of the largest
component in the aset (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.
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.
User Interface for the Other Methods
The data entered on the EIGR entry selects the eigenvalue method and the frequency range or number of
required roots. The basic format of the Bulk Data entry EIGR is as follows:
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:
m
j
1.0 =
Fi V2 V1 – ( ) 1.0 ALPH
i
– ( ) 1.0 ALPH
NUMS
– ( ) · =
Main Index
59
CHAPTER 3
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
MD Nastran Dynamic Analysis User’s Guide
User Interface for Real Eigenvalue Analysis
60
There is an interrelationship among the F1, F2, and ND fields on the EIGR entry as defined in Table 32.
.
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 rigidbody 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.
Table 32 Relationship Between METHOD Field and Other Fields for Obsolete Methods
Field
METHOD Field
INV or SINV GIV, MGIV, HOU, or MHOU
F1, F2 Frequency range of interest. F1 must
be input. If METHOD = “SINV” and
ND, is blank, then F2 must be input.
See also Remark 21.
(Real > 0.0)
Frequency range of interest. If ND is not blank, F1
and F2 are ignored. If ND is blank, eigenvectors
are found with natural frequencies that lie in the
range between F1 and F2. (Real > 0.0; F1 < F2)
NE Estimate of number of roots in range
(Required for METHOD = “INV”).
Not used by “SINV” method.
(Integer > 0)
Not used.
ND Desired number of roots. If this field
is blank and METHOD = “SINV”,
then all roots between F1 and F2 are
searched and the limit is 600 roots.
(Integer > 0, Default is for
METHOD = “INV” only.)
Desired number of eigenvectors. If ND is zero, the
number of eigenvectors is determined from F1 and
F2. If all three are blank, then ND is automatically
set to one more than the number of degreesof
freedom listed on SUPORTi entries. (Integer > 0;
Default = 0)
1 2 3 4 5 6 7 8 9 10
EIGR 13 SINV 0.0 0.01 1
3 NE ·
Main Index
61
CHAPTER 3
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.
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.
In this example, the same method is requested, but all the modes below 100 cycles per unit time are
requested with MAX vector normalization.
In this example, the Sturm modified inverse power method is requested for the first six modes found in
the range specified (0.1 to 100 Hz). The POINT normalization method is requested with each
eigenvector scaled such that grid point 32 in the T3 direction has a magnitude of 1.0. Note that this
degreeoffreedom must be in the aset.
1 2 3 4 5 6 7 8 9 10
EIGR 1 AHOU 10
EIGR 2 AHOU 100.
MAX
EIGR 3 SINV 0.1 100.0 6
POINT 32 3
Main Index
MD Nastran Dynamic Analysis User’s Guide
Solution Control for Analysis
62
Solution Control for Analysis
This section describes input required for the selection and control of normal modes analysis.
Executive Control Section
You can run a normal modes analysis in SOL 103 of the Structured Solution Sequences. The Executive
Control Section can also contain diagnostic DIAG 16, which prints the iteration information used in the
INV or SINV method. You may also run a normal modes analysis as part of the other solution sequences
(such as modal transient response, modal frequency response, design optimization, and aeroelasticity).
Case Control Section
The most important eigenanalysis command in the Case Control is the METHOD command. This
command is required. The set identification number specified by the METHOD Case Control command
refers to the set identification number of an EIGR or EIGRL entry in the Bulk Data.
When a modal analysis is performed, the MD Nastran output file contains various diagnostic messages
and an eigenvalue analysis summary. Optional grid and element output are available using standard Case
Control output requests. Eigenvectors are printed only if a DISPLACEMENT or VECTOR command is
included. These requests are summarized in Table 33.
Table 33 Eigenvalue Extraction Output Requests
Grid Output
DISPLACEMENT
(or VECTOR)
Requests the eigenvector (mode shape) for a set of grid points.
GPFORCE Requests the modal grid point force balance table to be computed for each mode
for a set of grid points.
GPKE Requests the output of the modal kinetic energy at selected grid points in normal
modes analysis only.
GPSTRESS Requests modal grid point stresses to be computed for a set of grid points. This
request must be accompanied by the ELSTRESS Case Control request and the
definition of stress surfaces and/or stress volumes in the OUTPUT(POST) section
of the Case Control. This request also requires the use of Rigid Format Alter
RF3D81 when used in SOL 3.
SPCFORCES Requests modal forces of singlepoint constraint to be computed for a set of grid
points for each mode.
Element Output
ELSTRESS (or
STRESS)
Requests the computation of modal stresses for a set of elements for each mode.
ESE Requests the computation of modal element strain energies for a set of elements
for each mode.
Main Index
63
CHAPTER 3
Real Eigenvalue Analysis
Bulk Data Section
In addition to Bulk Data entries required to define the structural model, the only other required Bulk Data
entry is the eigenvalue selection entry EIGR or EIGRL. The EIGR entry is used to select the modal
extraction parameters for the inverse power, Lanczos, and automatic Householder methods. The EIGRL
entry is used to select the modal extraction parameters for the Lanczos method, see the User Interface for
Real Eigenvalue Analysis, 56
ELFORCE (or
FORCE)
Requests the computation of modal element forces for a set of elements for each
mode.
STRAIN Requests the computation of modal element strains for a set of elements.
Miscellaneous
MODES A special Case Control request that permits selective output requests to be
processed on selective modes.
Table 33 Eigenvalue Extraction Output Requests (continued)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Massless Mechanism Identification and Control
64
Massless Mechanism Identification and Control
A massless mechanism (MM) mode is defined as a modeshape that results in very little strain energy and
kinetic energy for the mode. The ratio of strain energy to kinetic energy is proportional to the eigenvalue
of this mode. This result is an indeterminate quantity produced by two terms approaching zero. If
applied loads excite this shape, small loads can cause very large deflections. If a normal mode analysis
can be forced to produce a solution in the presence of this condition, the mode produced with this shape
can have any eigenvalue from positive to negative infinity. These frequencies may change drastically for
small changes in the model or when run on a different computer. Eigensolution failures due to Sturm
number inconsistencies are often due to the presence of massless mechanisms.
The presence of such modes reduces the reliability and repeatability of eigensolutions and dynamic
analysis solutions. A method is now provided to automatically constrain massless mechanisms for
eigensolutions, controlled by the parameter PARAM,MECHFIX,AUTO, which is turned on by default.
When this method is used, it is unlikely that the eigensolution will fail because of three failed shifts, and
the solution produced will be repeatable for small changes in the model, loading condition, or computer
type used in the solution. The goal of this capability is to allow the real eigensolvers to produce an
eigensolution every time a proper nonzero stiffness and mass matrix are provided.
Input
The massless mechanism feature is controlled by several parameters. The following is a list of the
primary parameters along with their functions:
MECHFIX Default = AUTO
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
65
CHAPTER 3
Real Eigenvalue Analysis
The following secondary parameters have default values that work well in most circumstances. They are
provided for tuning unusual circumstances.
MAXRATIO Default = 1.E7
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.
MECHFIL Default = 1.E6
Criterion for discarding massless mechanism modes with small generalized mass. A
smaller value will result in more marginal constraint modes being retained.
MMFIL Default = 1.e10
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
MD Nastran Dynamic Analysis User’s Guide
Prestiffened Normal Mode Analysis
66
Prestiffened Normal Mode Analysis
So far, all the discussion about normal modes has been for unloaded structures. In this section, normal
modes analysis of structures which have a static preload applied to them will be considered. Typical
realworld examples are:
• Thin walled pressure vessels where the pressurization dominates stiffness and hence, the natural
frequencies, such as rocket fuel tanks and satellite oxygen tanks
• Structures subject to significant dead weight loading, such as suspension cable bridges and oil
tankers
• Tensioning of strings in musical instruments to achieve required frequencies
• Structures subject to centrifugal loading such as jet engine turbine and fan blades
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 prestiffened structures is:
(318)
where 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. elasticplastic behavior
• Geometric stiffness, due to changes in the structural shape... these can be due to small or large
displacement behavior
• Followerforce 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 preloaded 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 preload
• Internal pressurization with 100 psi
Geometry of the Cylinder
e
2
M   – K   K  
D
+ ( ) + ( ) o { } 0 { } =
K  
D
Main Index
67
CHAPTER 3
Real Eigenvalue Analysis
Install_dir/mdxxxx/doc/dynamics/prestiff103.dat is the normal modes analysis without prestiffening.
Install_dir/mdxxxx/doc/dynamics/prestiff103pl.dat is the normal modes analysis with prestiffening.
Install_dir/mdxxxx/doc/dynamics/prestiff106.dat is the nonlinear statics analysis run for comparison.
Listing 31 shows part of the Nastran input file for prestiffening for 103pl.dat.
Listing 31 Part of prestiff103pl.dat
SOL 103
CEND
TITLE = Prestiffened Normal Modes Analysis
ECHO = NONE
RESVEC = NO
SUBCASE 1
SUBTITLE=This is the static subcase.
SPC = 2
LOAD = 2
DISPLACEMENT(PLOT,SORT1,REAL)=ALL
SPCFORCES(PLOT,SORT1,REAL)=ALL
STRESS(SORT1,REAL,VONMISES,BILIN)=ALL
SUBCASE 2
SUBTITLE=This is the Normal Modes subcase.
STATSUB = 1
METHOD = 1
SPC = 2
VECTOR(PLOT,SORT1,REAL)=ALL
SPCFORCES(PLOT,SORT1,REAL)=ALL
$
BEGIN BULK
PARAM POST 0
PARAM PRTMAXIM YES
EIGRL 1 10 0 MASS
Cylinder Wall Thickness 0.01625 inches
Cap Thickness 0.25 inches
Length (l) 10 inches
Diameter (d) 2 inches
Elastic Modulus 1.0E7 psi
Material Density 0.101 lbm/in
3
Poisson Ratio 0.33
t
l
d
Main Index
MD Nastran Dynamic Analysis User’s Guide
Prestiffened Normal Mode Analysis
68
Listing 32 Eigenvalue table from the SOL 103 normal modes analysis, without
prestiffening
Listing 33 Eigenvalue table from the SOL 103 normal modes analysis, with prestiffening
Listing 34 Eigenvalue table from the nonlinear prestiffened normal modes analysis
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.
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
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
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
Main Index
69
CHAPTER 3
Real Eigenvalue Analysis
First mode for normal modes analysis for
prestiffened structure
First mode for normal modes analysis
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
70
Examples
This section provides several normalmodes analysis examples showing the input and output. These
examples are as follows:
These examples are described in the sections that follow.
TwoDOF Model
This example is a restrained twoDOF model with two springs and two masses as illustrated in Figure 35.
Model
Number of
Grid Points
Element
Types
Output
Requests
Analysis
Method Units
bd03two 3 CELAS2
CONM2
DISPLACEMENT
SPCFORCE
ELFORCE
AHOU Metric
bd03bar1,
bd03bar2
11 CBAR DISPLACEMENT SINV Metric
bd03bkt 236 CQUAD4
CONM2
RBE2
DISPLACEMENT
STRESS
ESE
MODES
Lanczos English
bd03car 972 CQUAD4
CTRIA3
CELAS2
DISPLACEMENT
ESE
Lanczos English
bd03fix 8157 CHEXA
CPENTA
DISPLACEMENT Lanczos English
bd03plt1, bd03plt2 81 CQUAD4 DISPLACEMENT Lanczos Metric
bd03dmi 7 CBAR
DMIG
None Lanczos Metric
Main Index
71
CHAPTER 3
Real Eigenvalue Analysis
Figure 35 TwoDOF Model
The masses are constrained to deflect in only the ydirection. 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 35, see Install_dir/mdxxxx/doc/dynamics/bd03two.dat.
Listing 35
$ 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
$
Grid Point 1
Grid Point 2
y
z
m
1
0.1 kg =
k
1
100 N/m =
m
2
10 kg =
k
2
1.0E4 N/m =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
72
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 36. The eigenvalue summary lists the eigenvalue , circular
frequency (radians per second), natural frequency (cycles per second), generalized mass (see
Eq. (313)), and generalized stiffness (see Eq. (315)) for each mode. The eigenvectors, SPC forces, and
spring forces are shown for each mode.
e
n
2
e
n
f
n
Main Index
73
CHAPTER 3
Real Eigenvalue Analysis
Cantilever Beam Model
This example is a fixedfree aluminum cantilever beam with properties as shown in Figure 36.
Listing 36 Output from the TwoDOF System
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 9.048751E+02 3.008114E+01 4.787562E+00 1.904875E01 1.723674E+02
2 2 1.105125E+03 3.324342E+01 5.290854E+00 2.105125E01 2.326426E+02
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 9.512492E02 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
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 1.051249E01 0.0 0.0 0.0 0.0
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
3 G 0.0 9.512491E+02 0.0 0.0 0.0 0.0
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
3 G 0.0 1.051249E+03 0.0 0.0 0.0 0.0
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
L
r
y
x z
y
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
74
Figure 36 Cantilever Beam Model
The term is the weight density and must be converted to mass density for consistency of units.
PARAM,WTMASS is used to convert this weight density to mass density.
where is the acceleration of gravity in . Therefore,
. The nonstructural weight of 2.414 N/m is added to
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, and . The
input file is shown in Listing 37. (See Install_dir/mdxxxx/doc/dynamics/bd03bar1.dat)
Listing 37 Input File for the First Beam Model
$ FILE bd03bar1.dat
$
$ CANTILEVER BEAM MODEL
$ CHAPTER 3, NORMAL MODES
$
SOL 103 $ NORMAL MODES ANALYSIS
TIME 10
CEND
$
TITLE = CANTILEVER BEAM
SUBTITLE = NORMAL MODES
LABEL = MODEL 1 (I1 NE I2)
$
SPC = 1
$
$ OUTPUT REQUEST
DISPLACEMENT = ALL
$
$ SELECT EIGR ENTRY
METHOD = 10
$
BEGIN BULK
L 3.0 m =
A 6.158E4 m
2
=
E 7.1E10 N/m
2
=
Nonstructural
Weight
2.414 N/m =
r 0.014 m =
I1 I2 3.0E8 m
4
= =
v 0.33 =
J 6.0E8 m
4
=
p
w
2.65E4 N/m
3
=
p
w
p
m
WTMASS 1 g 1 9.81 0.102 sec
2
m = = = g m/sec
2
p
m
p
w
WTMASS · 2.65E4 0.102 ( ) · 2703 kg/m
3
= = =
I1 2.9E8 m
4
= I2 3.1E8 m
4
=
Main Index
75
CHAPTER 3
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 6.1584 2.98 3.18 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 ydirection modes are illustrated in Figure 37. Displacements in the ydirection
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
MD Nastran Dynamic Analysis User’s Guide
Examples
76
Figure 37 First Two Mode Shapes in the YDirection
Printed output is shown in Listing 38. Note that modes 1 and 3 are ydirection (T2) modes and modes 2
and 4 are zdirection (T3) modes.
Main Index
77
CHAPTER 3
Real Eigenvalue Analysis
Now, consider the second model for which I1 and I2 are identical. Printed output is shown in Listing 39.
Note that modes 1 and 2; 3 and 4; 5 and 6; etc., have identical frequencies; this is a case of repeated roots.
Listing 38 Printed Results from the First Model
E I G E N V A L U E A N A L Y S I S S U M M A R Y (STURM INVERSE POWER)
NUMBER OF EIGENVALUES EXTRACTED . . . . . . 6
NUMBER OF TRIANGULAR DECOMPOSITIONS . . . . 9
TOTAL NUMBER OF VECTOR ITERATIONS . . . . . 59
REASON FOR TERMINATION: ALL EIGENVALUES FOUND IN RANGE.
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 0.0 0.0 0.0 0.0 0.0
2 G 0.0 1.391872E02 2.259793E10 0.0 1.470464E09 9.057010E02
3 G 0.0 5.301210E02 8.606855E10 0.0 2.725000E09 1.678406E01
4 G 0.0 1.133067E01 1.839608E09 0.0 3.765826E09 2.319480E01
5 G 0.0 1.908986E01 3.099362E09 0.0 4.598456E09 2.832321E01
6 G 0.0 2.820258E01 4.578870E09 0.0 5.232973E09 3.223138E01
7 G 0.0 3.831632E01 6.220901E09 0.0 5.685052E09 3.501587E01
8 G 0.0 4.911331E01 7.973862E09 0.0 5.976744E09 3.681249E01
9 G 0.0 6.032288E01 9.793808E09 0.0 6.137029E09 3.779972E01
10 G 0.0 7.173455E01 1.164657E08 0.0 6.202180E09 3.820101E01
11 G 0.0 8.321185E01 1.350998E08 0.0 6.215970E09 3.828595E01
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 0.0 0.0 0.0 0.0 0.0
2 G 1.363453E33 2.249974E10 1.391872E02 0.0 9.057010E02 1.464532E09
3 G 2.693333E33 8.574880E10 5.301210E02 0.0 1.678406E01 2.716750E09
4 G 3.956893E33 1.834005E09 1.133067E01 0.0 2.319480E01 3.758543E09
5 G 5.123021E33 3.092054E09 1.908986E01 0.0 2.832321E01 4.594735E09
6 G 6.163003E33 4.571193E09 2.820258E01 0.0 3.223138E01 5.234397E09
7 G 7.051232E33 6.214489E09 3.831632E01 0.0 3.501587E01 5.692031E09
8 G 7.765834E33 7.970303E09 4.911331E01 0.0 3.681249E01 5.988579E09
9 G 8.289217E33 9.794348E09 6.032288E01 0.0 3.779972E01 6.152242E09
10 G 8.608492E33 1.165196E08 7.173455E01 0.0 3.820101E01 6.219055E09
11 G 8.715798E33 1.352052E08 8.321185E01 0.0 3.828595E01 6.233260E09
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 0.0 0.0 0.0 0.0 0.0
2 G 0.0 7.568120E02 9.828895E08 0.0 5.937006E07 4.571418E01
3 G 0.0 2.464387E01 3.200556E07 0.0 8.257535E07 6.358200E01
4 G 0.0 4.318525E01 5.608564E07 0.0 7.289605E07 5.612913E01
5 G 0.0 5.632111E01 7.314543E07 0.0 3.723918E07 2.867391E01
6 G 0.0 5.916957E01 7.684467E07 0.0 1.426249E07 1.098159E01
7 G 0.0 4.941767E01 6.417950E07 0.0 6.984853E07 5.378201E01
8 G 0.0 2.743728E01 3.563287E07 0.0 1.184615E06 9.121327E01
9 G 0.0 4.170797E02 5.417504E08 0.0 1.522753E06 1.172494E+00
10 G 0.0 4.159041E01 5.401546E07 0.0 1.689206E06 1.300660E+00
11 G 0.0 8.124724E01 1.055190E06 0.0 1.730573E06 1.332512E+00
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 0.0 0.0 0.0 0.0 0.0
2 G 3.698388E22 1.207686E08 7.568131E02 0.0 4.571424E01 1.427123E08
3 G 7.305710E22 2.529794E08 2.464390E01 0.0 6.358204E01 3.100643E07
4 G 1.073314E21 1.866785E07 4.318528E01 0.0 5.612913E01 7.615683E07
5 G 1.389628E21 4.615435E07 5.632114E01 0.0 2.867385E01 1.006117E06
6 G 1.671725E21 7.409781E07 5.916957E01 0.0 1.098168E01 7.569575E07
7 G 1.912658E21 8.602044E07 4.941765E01 0.0 5.378208E01 4.733732E08
8 G 2.106496E21 6.796384E07 2.743725E01 0.0 9.121327E01 1.181005E06
9 G 2.248464E21 1.591279E07 4.170818E02 0.0 1.172493E+00 2.239405E06
10 G 2.335068E21 6.217935E07 4.159040E01 0.0 1.300659E+00 2.875692E06
11 G 2.364175E21 1.520816E06 8.124720E01 0.0 1.332511E+00 3.057266E0
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
78
(See Install_dir/mdxxxx/doc/dynamics/bd03bar2.dat) Also note that the eigenvectors are not pure y or
pure ztranslation (as they were in the first model); the eigenvectors are linear combinations of the y and
z modes since this model has repeated roots.
Listing 39 Printed Results from the Second Model
E I G E N V A L U E A N A L Y S I S S U M M A R Y (STURM INVERSE POWER)
NUMBER OF EIGENVALUES EXTRACTED . . . . . . 8
NUMBER OF TRIANGULAR DECOMPOSITIONS . . . . 6
TOTAL NUMBER OF VECTOR ITERATIONS . . . . . 68
REASON FOR TERMINATION: ALL EIGENVALUES FOUND IN RANGE.
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 0.0 0.0 0.0 0.0 0.0
2 G 4.198658E20 9.109908E03 1.052335E02 0.0 6.847622E02 5.927882E02
3 G 8.293447E20 3.469681E02 4.008020E02 0.0 1.268971E01 1.098529E01
4 G 1.218317E19 7.416008E02 8.566642E02 0.0 1.753660E01 1.518116E01
5 G 1.577188E19 1.249444E01 1.443303E01 0.0 2.141397E01 1.853772E01
6 G 1.897125E19 1.845877E01 2.132276E01 0.0 2.436876E01 2.109563E01
7 G 2.170278E19 2.507827E01 2.896933E01 0.0 2.647398E01 2.291808E01
8 G 2.389961E19 3.214496E01 3.713246E01 0.0 2.783232E01 2.409396E01
9 G 2.550813E19 3.948168E01 4.560753E01 0.0 2.857873E01 2.474010E01
10 G 2.648916E19 4.695067E01 5.423540E01 0.0 2.888212E01 2.500274E01
11 G 2.681884E19 5.446261E01 6.291288E01 0.0 2.894634E01 2.505833E01
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 . 2
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 0.0 0.0 0.0 0.0 0.0
2 G 1.219015E33 1.052334E02 9.109882E03 0.0 5.927866E02 6.847615E02
3 G 2.407892E33 4.008016E02 3.469673E02 0.0 1.098526E01 1.268970E01
4 G 3.537263E33 8.566635E02 7.415994E02 0.0 1.518114E01 1.753659E01
5 G 4.579275E33 1.443302E01 1.249442E01 0.0 1.853771E01 2.141396E01
6 G 5.508284E33 2.132276E01 1.845875E01 0.0 2.109563E01 2.436876E01
7 G 6.301480E33 2.896932E01 2.507825E01 0.0 2.291809E01 2.647399E01
8 G 6.939436E33 3.713246E01 3.214495E01 0.0 2.409399E01 2.783234E01
9 G 7.406566E33 4.560753E01 3.948168E01 0.0 2.474014E01 2.857874E01
10 G 7.691474E33 5.423541E01 4.695069E01 0.0 2.500278E01 2.888214E01
11 G 7.787220E33 6.291289E01 5.446264E01 0.0 2.505838E01 2.894636E01
EIGENVALUE = 6.474471E+03
CYCLES = 1.280626E+01 R E A L E I G E N V E C T O R N O . 3
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 0.0 0.0 0.0 0.0 0.0
2 G 1.606174E26 6.893317E02 3.123881E02 0.0 1.886937E01 4.163812E01
3 G 3.172614E26 2.244652E01 1.017221E01 0.0 2.624463E01 5.791277E01
4 G 4.660607E26 3.933468E01 1.782550E01 0.0 2.316833E01 5.112443E01
5 G 6.033448E26 5.129929E01 2.324757E01 0.0 1.183568E01 2.611722E01
6 G 7.257351E26 5.389377E01 2.442332E01 0.0 4.532863E02 1.000244E01
7 G 8.302282E26 4.501138E01 2.039804E01 0.0 2.219953E01 4.898660E01
8 G 9.142668E26 2.499085E01 1.132523E01 0.0 3.764997E01 8.308035E01
9 G 9.757999E26 3.798927E02 1.721591E02 0.0 4.839685E01 1.067950E+00
10 G 1.013329E25 3.788206E01 1.716723E01 0.0 5.368715E01 1.184688E+00
11 G 1.025940E25 7.400293E01 3.353632E01 0.0 5.500190E01 1.213700E+00
EIGENVALUE = 6.474471E+03
CYCLES = 1.280626E+01 R E A L E I G E N V E C T O R N O . 4
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 0.0 0.0 0.0 0.0 0.0
2 G 6.181717E19 3.123822E02 6.893279E02 0.0 4.163791E01 1.886905E01
3 G 1.221121E18 1.017205E01 2.244643E01 0.0 5.791260E01 2.624437E01
4 G 1.794002E18 1.782531E01 3.933455E01 0.0 5.112445E01 2.316835E01
5 G 2.322707E18 2.324743E01 5.129921E01 0.0 2.611744E01 1.183602E01
6 G 2.794219E18 2.442331E01 5.389376E01 0.0 1.000213E01 4.532376E02
7 G 3.196926E18 2.039817E01 4.501146E01 0.0 4.898637E01 2.219917E01
8 G 3.520915E18 1.132542E01 2.499097E01 0.0 8.308033E01 3.764996E01
9 G 3.758207E18 1.721460E02 3.798841E02 0.0 1.067952E+00 4.839722E01
10 G 3.902959E18 1.716725E01 3.788207E01 0.0 1.184692E+00 5.368777E01
11 G 3.951610E18 3.353654E01 7.400307E01 0.0 1.213704E+00 5.500259E01
Main Index
79
CHAPTER 3
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
productiontype models (i.e., complex threedimensional structures), this difference is negligible.
Bracket Model
This example is a steel bracket as shown in Figure 38.
Figure 38 Bracket Model
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
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
80
A concentrated mass is suspended from the center of the hole in bracket. This mass has the following
properties:
The concentrated mass (grid point 999) is connected to the bracket by an RBE2 element connecting 24
grid points, as shown in Figure 39.
Figure 39 Bracket Model Showing RBE2 Element (Dashed Lines)
The bracket is clamped by constraining six degreesoffreedom 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 310. (See Install_dir/mdxxxxmdxxxx/doc/dynamics/bd03bkt.dat)
m 0.0906 lbsec
2
/in =
I11 0.35 inlbsec
2
=
I22 0.56 inlbsec
2
=
I33 0.07 inlbsec
2
=
Main Index
81
CHAPTER 3
Real Eigenvalue Analysis
Listing 310 Abridged Input File for the Bracket Model
$ FILE bd03bkt.dat
$
$ BRACKET MODEL
$ CHAPTER 3, NORMAL MODES
$
SOL 103 $ NORMAL MODES ANALYSIS
CEND
$
TITLE = BRACKET MODEL
SUBTITLE = NORMAL MODES ANALYSIS
$
SPC = 1
$
$ SELECT EIGRL
METHOD = 777
$
$ OUTPUT REQUESTS
DISPLACEMENT = ALL
SUBCASE 1
MODES = 2 $ USE FOR FIRST TWO MODES
STRESS(CORNER) = ALL
SUBCASE 3
ESE = ALL
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10.$
$
$EIGRL SID V1 V2 ND MSGLVL MAXSET SHFSCL NORM
EIGRL 777 100.
$
$ CONCENTRATED MASS, SUPPORTED WITH AN RBE2 ELEMENT
$
GRID 999 3.0 3.0 3.8
$
$CONM2 EID G CID M X1 X2 X3 +CON1
$+CON1 I11 I21 I22 I31 I32 I33
CONM2 999 999 0.0906 +CM1
+CM1 0.35 0.56 0.07
$
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
$
$ STEEL, M A S S DENSITY FOR RHO
MAT1 1 3.+7 1.153+7 7.764
$
... basic model ...
$
ENDDATA
Listing 311 shows an abridged version of the resulting MD Nastran output. The circular total element
strain energy (ESE) for each mode of the entire model is for the ith mode when .
The frequency of the third mode is 230.34 radians per second squared; therefore, the total strain energy
of the model is for the third mode. The printed . The
ESE e
i
2
2 = o
i
T
Mo
i
1 =
ESE
total
230.34 ( )
2
2 27493 = = ESE
total
26527 =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
82
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.)
The deformed shape resulting from the first mode is illustrated in Figure 310 and is overlaid on the
undeformed shape. Figure 311 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 312.
Listing 311 Abridged Output from the Bracket Model
E I G E N V A L U E A N A L Y S I S S U M M A R Y (LANCZOS ITERATION)
BLOCK SIZE USED ...................... 6
NUMBER OF DECOMPOSITIONS ............. 2
NUMBER OF ROOTS FOUND ................ 3
NUMBER OF SOLVES REQUIRED ............ 3
TERMINATION MESSAGE : REQUIRED NUMBER OF EIGENVALUES FOUND.
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
ELEMENTTYPE = QUAD4 * TOTAL ENERGY OF ALL ELEMENTS IN PROBLEM = 2.749221E+04
MODE 3 TOTAL ENERGY OF ALL ELEMENTS IN SET 1 = 2.749221E+04
*
TYPE = QUAD4 SUBTOTAL 2.749221E+04 100.0000
Main Index
83
CHAPTER 3
Real Eigenvalue Analysis
Figure 310 Deformed Shape of the First Mode
Figure 311 Second Mode Deformation with Element Stress Contours
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
84
Figure 312 Element Strain Energy Contours for the Third Mode
Car Frame Model
Figure 313 shows a model of an aluminum car frame. The frame model is comprised of plate elements
(CQUAD4 and CTRIA3), with springs (CELAS2) representing the suspension. Spring stiffnesses are
input in the three translational directions; a stiffness of 500 lb/in is used in the vertical direction (T2), and
stiffnesses of 1000 lb/in are used in the other translational directions (T1 and T3). When using CELASi
elements to connect two grid points, it is recommended that the coordinates of the two grid points be
identical in order to represent coaxial springs (noncoincident coordinates can lead to errors). The goal
of the analysis is to compute resonant frequencies up to 50 Hz using the Lanczos method. Element strain
energies are computed for the springs in order to help characterize the resulting modes.
Listing 311 shows the input file. Modal displacements are written to the plot file and are not printed.
The rigid body mass matrix is computed via the PARAM,GRDPNT,0 entry. “Include” files are used to
partition the input file into several smaller files. The INCLUDE statement inserts an external file into
the input file. The basic file is bd03car.dat. The springs are contained in file bd03cars.dat (Listing 312),
and the rest of the input file is contained in file bd03carb.dat (not shown). (See
Install_dir/mdxxxx/doc/dynamics/bd03car.dat, bd03cars.dat, and bd03carb.dat)
Main Index
85
CHAPTER 3
Real Eigenvalue Analysis
Figure 313 Car Frame Model
Listing 312 Basic Input File for the Car Model
$ FILE bd03car.dat
$
$ CAR FRAME MODEL
$ CHAPTER 3, NORMAL MODES
$
$ MODEL COURTESY LAPCAD ENGINEERING
$ CHULA VISTA, CALIFORNIA
$
SOL 103 $ NORMAL MODES ANALYSIS
TIME 30
CEND
$
TITLE = CAR MODEL WITH SUSPENSION SPRINGS
SUBTITLE = MODAL ANALYSIS CASE CONTROL
ECHO = UNSORT
$
METHOD = 1
$
DISPLACEMENT(PLOT) = ALL
SET 99 = 1001,1002,1003,1011,1012,1013,
1021,1022,1023,1031,1032,1033
ESE = 99
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10..
...$
$
INCLUDE 'bd03cars.dat' $ Car springs
$
$EIGRL SID V1 V2 ND MSGLVL MAXSET SHFSCL NORM
EIGRL 1 1. 50.
$
$MAT1 MID E G NU RHO
MAT1 1 1.0E7 0.33 2.59E4
$
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
86
$ PRINT RIGIDBODY MASS
PARAM,GRDPNT,0
$
ECHOOFF
INCLUDE 'bd03carb.dat' $ Rest of Bulk Data
$
ENDDATA
Listing 313 The Input File for the Springs
$ FILE bd03cars.dat
$
$ SPRINGS CONNECTED TO GROUND
$ GRIDS 1059,1562,1428,1895 HAVE THE SAME COORDS. AS 59,562,428,895
$
GRID 1059 152.012 32.7964 2.90000 123456
GRID 1562 152.012 32.7964 67.1000 123456
GRID 1428 35.6119 30.8257 0.66667 123456
GRID 1895 35.6119 30.8257 69.3333 123456
$
CELAS2 1001 1000. 59 1 1059 1
CELAS2 1002 500. 59 2 1059 2
CELAS2 1003 1000. 59 3 1059 3
CELAS2 1011 1000. 562 1 1562 1
CELAS2 1012 500. 562 2 1562 2
CELAS2 1013 1000. 562 3 1562 3
CELAS2 1021 1000. 428 1 1428 1
CELAS2 1022 500. 428 2 1428 2
CELAS2 1023 1000. 428 3 1428 3
CELAS2 1031 1000. 895 1 1895 1
CELAS2 1032 500. 895 2 1895 2
CELAS2 1033 1000. 895 3 1895 3
Listing 314 shows the grid point weight generator output. The grid point weight generator indicates that
the translational mass is . This type of information is useful in verifying your model.
Listing 315 shows some of the rest of the output. The eigenvalue analysis summary indicates that there
are 12 modes below 50 Hz. The large element strain energies in the first six modes indicate that these
are primarily suspension modes, comprised of the car frame acting as a rigid body supported by the
flexible springs. Approximately 94% of the total strain energy is represented by spring deformation as
opposed to frame deformation. Modes 7 and above show insignificant spring strain energy, indicating
that these are primarily frame modes.
0.8027 lbsec
2
in.
Main Index
87
CHAPTER 3
Real Eigenvalue Analysis
Listing 314 Output from the Grid Point Weight Generator
O U T P U T F R O M G R I D P O I N T W E I G H T G E N E R A T O R
REFERENCE POINT = 0
M O
* 8.027376E01 0.000000E+00 0.000000E+00 0.000000E+00 2.808118E+01 2.179473E+01 *
* 0.000000E+00 8.027376E01 0.000000E+00 2.808118E+01 0.000000E+00 7.610537E+01 *
* 0.000000E+00 0.000000E+00 8.027376E01 2.179473E+01 7.610537E+01 0.000000E+00 *
* 0.000000E+00 2.808118E+01 2.179473E+01 2.325008E+03 2.153940E+03 2.662697E+03 *
* 2.808118E+01 0.000000E+00 7.610537E+01 2.153940E+03 1.020870E+04 7.626398E+02 *
* 2.179473E+01 7.610537E+01 0.000000E+00 2.662697E+03 7.626398E+02 9.437676E+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 XC.G. YC.G. ZC.G.
X 8.027376E01 0.000000E+00 2.715050E+01 3.498177E+01
Y 8.027376E01 9.480728E+01 0.000000E+00 3.498177E+01
Z 8.027376E01 9.480728E+01 2.715050E+01 0.000000E+00
I(S)
* 7.509408E+02 8.764180E+01 3.962963E01 *
* 8.764180E+01 2.011031E+03 2.217236E01 *
* 3.962963E01 2.217236E01 1.630595E+03 *
I(Q)
* 2.017097E+03 *
* 7.448741E+02 *
* 1.630595E+03 *
Q
* 6.905332E02 9.976128E01 5.012719E04 *
* 9.976128E01 6.905355E02 4.555999E04 *
* 4.891269E04 4.686146E04 9.999998E01 *
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
88
Mode shapes for modes 7, 8, 9, and 10 are shown in Figure 314. 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.
Listing 315 Abridged Output from the Car Model
E I G E N V A L U E A N A L Y S I S S U M M A R Y (LANCZOS ITERATION)
BLOCK SIZE USED ...................... 7
NUMBER OF DECOMPOSITIONS ............. 3
NUMBER OF ROOTS FOUND ................ 12
NUMBER OF SOLVES REQUIRED ............ 11
TERMINATION MESSAGE : REQUIRED NUMBER OF EIGENVALUES FOUND.
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
ELEMENTTYPE = ELAS2 * TOTAL ENERGY OF ALL ELEMENTS IN PROBLEM = 1.173240E+03
MODE 1 TOTAL ENERGY OF ALL ELEMENTS IN SET 99 = 1.104569E+03
*
ELEMENTID STRAINENERGY PERCENT OF TOTAL STRAINENERGYDENSITY
1002 2.735009E+02 23.3116
1003 4.059090E02 .0035
1012 2.685884E+02 22.8929
1013 3.240471E02 .0028
1021 8.017746E02 .0068
1022 2.833448E+02 24.1506
1023 3.963123E01 .0338
1031 7.903841E02 .0067
1032 2.781467E+02 23.7076
1033 3.573737E01 .0305
TYPE = ELAS2 SUBTOTAL 1.104569E+03 94.1469
E L E M E N T S T R A I N E N E R G I E S
ELEMENTTYPE = ELAS2 * TOTAL ENERGY OF ALL ELEMENTS IN PROBLEM = 1.402770E+04
MODE 7 TOTAL ENERGY OF ALL ELEMENTS IN SET 99 = 3.172818E+03
*
ELEMENTID STRAINENERGY PERCENT OF TOTAL STRAINENERGYDENSITY
1001 1.278503E+02 .9114
1002 5.884620E+02 4.1950
1003 8.020268E+01 .5717
1011 1.278420E+02 .9114
1012 5.883779E+02 4.1944
1013 8.022697E+01 .5719
1021 7.525866E+01 .5365
1022 4.885996E+02 3.4831
1023 2.255234E+02 1.6077
1031 7.528390E+01 .5367
1032 4.896509E+02 3.4906
1033 2.255398E+02 1.6078
TYPE = ELAS2 SUBTOTAL 3.172818E+03 22.6182
Main Index
89
CHAPTER 3
Real Eigenvalue Analysis
Figure 314 Mode Shapes for Modes 7, 8, 9, and 10
Test Fixture Model
This example is an aluminum test fixture, which is shown in Figure 315. The model is comprised of
8157 grid points, 5070 CHEXA elements, and 122 CPENTA elements. The primary plates are 1 inch
thick, and the gusset plates are 0.5 inch thick. The base of the fixture is constrained to have no vertical
(y) motion, and the bolt holes at the base are constrained to also have no horizontal (x and z) motion.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
90
Figure 315 Test Fixture Model
A portion of the input file is shown in Listing 316. (See Install_dir/mdxxxx/doc/dynamics/bd03fix.dat)
The Lanczos method is used to compute the modes. The first six modes are requested (ND is 6, with V1
and V2 blank). The GRDSET Bulk Data entry removes the rotational DOFs (456) from the analysis since
the solid elements have no rotational stiffness. The Bulk Data is in free format.
Listing 316 Abridged Input File for Test Fixture Model
$ FILE bd03fix.dat
$
$ TEST FIXTURE
$ CHAPTER 3, NORMAL MODES
$
TIME 240
SOL 103
CEND
ECHO = NONE
DISPLACEMENT(PLOT) = ALL
SPC = 1
$
METHOD = 1
$
BEGIN BULK
Main Index
91
CHAPTER 3
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, 2.53834
$
... basic model ...
$
ENDDATA
Figure 316 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.
Figure 316 Test Fixture Mode Shapes
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
92
Quarter Plate Model
This example is a quarter model of a simply supported flat plate, shown in Figure 317. This example
illustrates the use of multiple boundary conditions for modeling symmetric structures. In this case the
plate is doubly symmetric.
Figure 317 Derivation of Quarter Plate Model
A portion of the input file is shown in Listing 317. (See Install_dir/mdxxxx/doc/dynamics/bd03plt1.dat)
Four subcases are used—one for each of the following sets of boundary conditions for the quarter model:
• Symmetricantisymmetric
• Antisymmetricsymmetric
• Symmetricsymmetric
• Antisymmetricantisymmetric
The BC Case Control command identifies multiple boundary conditions. The SPCADD Bulk Data entry
defines a union of SPC sets.
Listing 317 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
Lines of Symmetry
Quarter Model
Full Model
ss
ss
ss
ss
ss
ss
ss
ss
ss
ss
1
2
x
y
z
Main Index
93
CHAPTER 3
Real Eigenvalue Analysis
SUBTITLE = NORMAL MODES CASE CONTROL
LABEL = QUARTER PLATE MODEL
$
DISPLACEMENT = ALL
$
SUBCASE 1
LABEL = SYMASYM
BC = 1
METHOD = 1
SPC = 101
SUBCASE 2
LABEL = ASYMSYM
BC = 2
SPC = 102
METHOD = 1
SUBCASE 3
LABEL = SYMSYM
BC = 3
SPC = 103
METHOD = 1
SUBCASE 4
LABEL = ASYMASYM
BC = 4
SPC = 104
METHOD = 1
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10..
$
$ SYMASYM
SPCADD 101 11 1 4
$ ASYMSYM
SPCADD 102 11 2 3
$ SYMSYM
SPCADD 103 11 1 3
$ ASYMASYM
SPCADD 104 11 2 4
$
$EIGRL SID V1 V2
EIGRL 1 0.1 100.
$
SPC 1 1 246 0.00
SPC 1 2 246 0.00
... etc. ...
SPC 11 80 12356 0.00
SPC 11 81 123456 0.00
$
... basic model ...
$
ENDDATA
Figure 318 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
MD Nastran Dynamic Analysis User’s Guide
Examples
94
Figure 318 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 319 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
95
CHAPTER 3
Real Eigenvalue Analysis
Figure 319 Planar Cantilever Beam
The stiffness and mass matrices from the contractor for element 5 are as follows:
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 319 see
Install_dir/mdxxxx/doc/dynamics/bd03dmi.dat.
Listing 318 Input File for the DMIG Example
$ FILE bd03dmi.dat
$
$ DMIG EXAMPLE
$ CHAPTER 3, NORMAL MODES
$
SOL 103 $ NORMAL MODES ANALYSIS
TIME 10
CEND
TITLE = DMIG TO READ STIFFNESS AND MASS FOR ELEM 5
SUBTITLE = PLANAR PROBLEM
$
1 2 3 4 5 6
1 2 3 4 5
Z
X
L
= Grid Point ID
= Element ID
K
5
5, 3
5, 5
6, 3
6, 5
500038.8
250019.4 – 166679.6
500038.8 – 250019.4 500038.8
250019.4 – 83339.8 250019.4 166679.6
\ .




 
=
5, 3 5, 5 6, 3 6, 5
M
5
5, 3
5, 5
6, 3
6, 5
3.5829 0. 0. 0.
0. 0. 0. 0.
0. 0. 3.5829 0.
0. 0. 0. 0.
\ .




 
=
5, 3 5, 5 6, 3 6, 5
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
96
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 319 Input File for the DMIG Example
PBAR 1 1 2.6543 5.8697
SPC1 10 123456 1
$
ENDDATA
Main Index
97
CHAPTER 3
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 nonnull 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 ( ) represents the induced reactive
load in the ith degreeoffreedom due to a unit displacement in the jth direction with all other
displacement degreesoffreedom 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
offreedom, respectively. In this case, it is grid point5, degreeoffreedom 3 (ztranslation 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 rowbyrow. These four terms
are defined by sets of three fields. They are the external grid ID number, corresponding degreeof
freedom, and the actual matrix term, respectively. The first term of column one is defined by external
grid ID 5, degreeoffreedom 3 (ztranslation at grid point 5) with a stiffness value of 500039. The
second term of column one is defined by external grid ID 5, degreeoffreedom 5 (yrotation at grid point
5) with a stiffness value of 250019. The third term of column one is defined by external grid ID 6,
degreeoffreedom 3 (ztranslation at grid point 6) with a stiffness value of 500039. The fourth term of
column one is defined by external grid ID 6, degreeoffreedom 5 (yrotation 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 degreeoffreedom, respectively. In this
case, it is grid point 5, degreeoffreedom 5 (yrotation 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.
K
i j
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
98
• 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 nonnull 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 smallfield 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
largefield 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
99
CHAPTER 3
Real Eigenvalue Analysis
Rigid Body Modes
A structure or a portion of a structure can displace without developing internal loads or stresses if it is
not sufficiently tied to ground (constrained). These stressfree displacements are categorized as
rigidbody modes or mechanism modes.
Free body motion in a structure occurs when a structure may move freely without applied forces.
Although the stiffness matrix will have one or more singularities, the combined dynamic system, with
mass and damping, may not be singular. Examples are flying objects such as aircraft or rockets, and
structures with kinematic mechanisms such as a gyroscope or a pendulum. In most cases, the dynamic
response solutions in MD Nastran do not require any special attention for free body motions. In general,
MD Nastran will decompose matrix combinations that are not singular. For example, an unbalanced load
applied to a free body has a unique dynamic solution, namely a constant acceleration. The only exception
is that of a frequency response analysis at a frequency of zero. In this case it is recommended that a small
frequency be used instead of zero for free bodies.
Rigidbody Modes
Rigidbody modes occur in unconstrained structures, such as satellites and aircraft in flight. For a
general, unconstrained 3D structure without mechanisms, there are six rigidbody modes often
described as T1, T2, T3, R1, R2, and R3, or combinations thereof. Rigidbody 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. Rigidbody 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 rigidbody 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 1.0E4 Hz or
less for typical structures. The same unconstrained model may give different values of the rigidbody
frequencies when run on different computer types.
Rigidbody modes generated by MD Nastran are orthogonal with respect to each other and with respect
to the flexible (or elastic) modes. Each rigidbody mode has the following property:
Main Index
MD Nastran Dynamic Analysis User’s Guide
Rigid Body Modes
100
(319)
where denotes the rigidbody mode shapes, denotes the mass matrix, and denotes the
stiffness matrix. The rigidbody modes contain no strain energy.
Figure 320 depicts the rigidbody modes for a twodimensional model of a straight beam. (See
Install_dir/mdxxxx/doc/dynamics/bd03rmbar1.dat) Because the twodimensional model has three DOFs
per grid point (xtranslation, ytranslation, and zrotation), there are three rigidbody modes. The figure
shows classical rigidbody modes, whereby one mode is purely xtranslation, another is purely
ytranslation, and another is purely zrotation about the center of the beam. Because rigidbody modes
are a special case of repeated roots, any linear combination of these displacement shapes also comprises
a valid set of rigidbody modes.
Figure 320 RigidBody Modes of a 2D Beam
o
ri g
{ }
T
M   o
ri g
{ } 0 >
o
ri g
{ }
T
K   o
ri g
{ } 0 =
o
ri g
{ } M   K  
x
y
Main Index
101
CHAPTER 3
Real Eigenvalue Analysis
SUPORT Entry
Rigidbody 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 rigidbody 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 rigidbody shape calculations by defining the rset 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 rigidbody mode or mechanism. The
format of the Bulk Data entry SUPORT is shown below:
Treatment of SUPORT by Eigenvalue Analysis Methods
The eigenvalue extraction methods treat the SUPORT entry differently as described below.
EIGR Methods
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 modes
(where is the number of SUPORT DOFs) are replaced with a value of 0.0 Hz. The first
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 rigidbody modes (as
opposed to the rigidbody modes computed without the use of the SUPORT entry).
Note that MD Nastran has no builtin checks to warn if any of the eigenvalues are not rigidbody
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
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).
N
r
N
r
N
r
N
r
Main Index
MD Nastran Dynamic Analysis User’s Guide
SUPORT Entry
102
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 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 Lanczoscomputed, rigidbody eigenvectors are used, i.e., the rigidbody
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
rigidbody modes. There is some logic to determine the relationship of these candidate rigidbody 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
1.0E6, for example). On the other hand, if your SUPORT entry defines four DOFs, then only the first
three are treated as rigidbody 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 rigidbody eigenvectors, there is probably little
to be gained by using the SUPORT entry with the Lanczos method unless response spectrum analysis is
being performed.
SUPORT Entry  Theoretical Considerations
Degreesoffreedom defined on the SUPORT entry are placed in the rset. When an rset is present, static
rigidbody vectors are calculated in MD Nastran by first partitioning the aset into the r and lsets
(320)
Introducing this partitioning in the stiffness matrix results in
(321)
for the rigidbody modes defined in the rset.
There is no load on lset DOFs. The load on the rset is not needed in subsequent equations. Then
solve for in terms of
N
r
u
a
{ }
u
l
u
r
¹ )
´ `
¦ ¹
=
K
l l
K
l r
K
rl
K
rr
u
l
u
r
¹ )
´ `
¦ ¹
0
P
r ¹ )
´ `
¦ ¹
=
P
r
{ }
u
l
u
r
Main Index
103
CHAPTER 3
Real Eigenvalue Analysis
(322)
where:
The matrix is used to construct a set of rigidbody vectors
(323)
where represents the motion of the aset for a unit motion of each SUPORT DOF with all other
SUPORT DOFs constrained and is an identity matrix.
The rigidbody vectors can be used to create a rigidbody mass matrix
To improve the quality of the rigidbody mode shapes, orthogonalization is applied to create a diagonal
mass matrix by
(324)
where: is a transformation matrix.
This transformation matrix is used to construct the final set of rigidbody mode shape vectors by
(325)
such that
(326)
where:
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 rigidbody vectors. If
you do not specify the rset DOFs, the rigidbody modes are calculated directly by the method selected
for the flexible frequency modes. If an insufficient number of rset DOFs are specified, the calculation
of rigidbody modes is unreliable.
u
l
{ } D   u
r
{ } =
D   K
l l
  –
1 –
K
l r
  =
D   v
ri g
 
v
ri g
 
D
I
r
=
v
ri g
 
I
r
  r r ×
M
r
 
M
r
  v
ri g
 
T
M
aa
  v
ri g
 
D
I
r
T
M
aa
 
D
I
r
= =
M
o
 
M
o
  o
ro
T
  M
r
  o
ro
  = =
o
ro
 
o
ri g
  v
ri g
  o
ro
 
Do
ro
o
ro
= =
o
ri g
 
T
K
aa
  o
ri g
  0 ~
o
ri g
 
T
M
aa
  o
ri g
  M
o
=
M
o
Main Index
MD Nastran Dynamic Analysis User’s Guide
SUPORT Entry
104
As a modeling aid, MD Nastran calculates equivalent internal strain energy (work) for each rigidbody
vector as follows:
(327)
which can be simplified as
(328)
When rset DOFs exist, the printed strain energies are the diagonal elements of divided by 2 and
should be approximately zero.
Note that is the transformation of the stiffness matrix to rset coordinates, which by definition
of rigidbody (i.e., zero frequency) vector properties should be null. If this is not the case, the equilibrium
may be violated by the rset choice or other modeling errors may exist. The matrix is also called the
rigidbody check matrix.
MD Nastran also calculates a rigidbody error ratio
(329)
where: = Euclidean norm of the matrix
One value of is calculated using Eq. (329) based on all SUPORT DOFs. Therefore, in UIM3035 the
same is printed for every supported DOF.
The rigidbody 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 ( , for example) that normally is 0.0 except for numerical
roundoff.) The rigidbody error ratio and strain energy may be significantly nonzero for any of the
following reasons:
• Roundoff error accumulation.
• The set is overdetermined, leading to redundant supports. The condition gives high strain
energy.
• The set is underspecified, leading to a singular reducedstiffness matrix and a MAXRATIO
error. This condition gives a high rigidbody error ratio.
• The multipoint constraints are statically indeterminate. This condition gives high strain energy
and a high rigidbody error ratio.
X  
D
T
I
r
K
l l
K
l r
K
rl
K
rr
D
I
r
=
X   D  
T
K
l l
  D   K
rr
  + =
X  
X   K
aa
 
X  
c
K
rr
  K
l r
 
T
D   +
K
rr
 =
x
i j
2
j
¯
i
¯
=
c
c
10
5 –
u
r
u
r
Main Index
105
CHAPTER 3
Real Eigenvalue Analysis
• There are too many singlepoint constraints. This condition gives high strain energy and a high
rigidbody error ratio.
• is null. This condition gives a unit value for the rigidbody error but low strain energy.
This is an acceptable condition and may occur when generalized dynamic reduction is used (see
Special Topics, 359).
SUPORT Entry  Modeling Considerations
When using a SUPORT you must select a set of DOFs that is capable of constraining all the rigidbody
modes. Another way to state this requirement is that the rset must be able to constrain the structure in
a statically determinate manner. There are usually many choices of DOFs that satisfy this requirement.
Two choices that work for simple threedimensional structures are:
• Six DOFs on one grid point when all its degreesoffreedom have stiffness.
• Three translation DOFs normal to one plane, two translation DOFs normal to an orthogonal
plane, and then one translational DOF normal to the last orthogonal plane. Such a system can be
used, for instance, on a model composed entirely of solid elements that have no inherent
stiffness for grid point rotation. See Figure 321.
Figure 321 Statically Determinate rset
There are special cases where a model need not have six rigidbody modes. A planar model has only
three rigidbody 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 rigidbody motion, then provide rset
DOFs that define these rigidbody 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,
K
rr
 
Plane 1
Plane 2
Plane 3
1
2
3
4
5
6
1, 2, 3 Normal to Plane 1
4, 5 Normal to Plane 2
6 Normal to Plane 3
Main Index
MD Nastran Dynamic Analysis User’s Guide
SUPORT Entry
106
diagnostics are produced that indicate overconstraint. For this case, these diagnostics may be safely
ignored.
Poorlyconstrained rigidbody 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 degreesoffreedom normal to the shell and their
associated bending degreesoffreedom 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 rset. The RBE3 element does not affect
the flexible modes when applied in this manner.
An example of a structure whose rset shows poor linear independence is a slightly curved bar modeled
using coordinate systems that follow the curve of the bar such that the xaxis 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 rset DOFs, including their
locations in space and the orientation of their global coordinates. Then apply the threeplane test
described earlier.
Using a physical analogy, a good rset 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 rigidbody modes without affecting the flexible
modes.
Main Index
107
CHAPTER 3
Real Eigenvalue Analysis
Rigid Body Mode Examples
This section provides several rigidbody modes examples showing input and output. These examples
are as follows:
These examples are described in the sections that follow.
Unconstrained Beam Model
The constraints (SPCs) on the example cantilever beam model from Cantilever Beam Model, 191 are
removed to create an unconstrained structure as shown in Figure 322. A GRDSET entry is added with
the ztranslation, xrotation, and yrotation directions constrained to make the problem twodimensional.
Therefore, there are three DOFs per grid point (xtranslation, ytranslation, and zrotation) and three
rigidbody modes.
Figure 322 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)
Table 34 Unconstrained Beam Model Summary
Model
Analysis
Method
SUPORT/
No SUPORT
Redundancy of
SUPORT
bd03rmbar1 Lanczos No SUPORT –
bd03rmbar2 Lanczos SUPORT Statically determinate
bd03rmbar3 Lanczos SUPORT Underdetermined
bd03rmbar4 Lanczos SUPORT Overdetermined
bd03rmbar5 SINV No SUPORT –
bd03rmbar6 SINV SUPORT Statically determinate
bd03rmbar7 SINV SUPORT Underdetermined
bd03rmbar8 SINV SUPORT Overdetermined
bd03rmbkt Lanczos No SUPORT –
X
Y
Z
Main Index
MD Nastran Dynamic Analysis User’s Guide
Rigid Body Mode Examples
108
Listing 320 shows a portion of the input file for the statically determinate SUPORT and the Lanczos
method. (See Install_dir/mdxxxx/doc/dynamics/bd03rmbar2.dat)
Listing 320 Input File for Cantilever Beam Model
$ FILE bd03rmbar2.dat
$
$ CANTILEVER BEAM MODEL
$ CHAPTER 3, RIGIDBODY MODES
$
SOL 103
TIME 10
CEND
TITLE = CANTILEVER BEAM
SUBTITLE = NORMAL MODES
LABEL = USE SUPORT, STATICALLY DETERMINATE
$
$ OUTPUT REQUEST
DISPLACEMENT = ALL
$
$ SELECT EIGRL ENTRY
METHOD = 10
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10..
...$
$
$ STATICALLY DETERMINATE SUPORT
$SUPORT G C
SUPORT 1 126
$
$ MAKE 2D MODEL
GRDSET 345
$
$EIGRL SID V1 V2
EIGRL 10 0.1 50.
$
... basic model ...
$
ENDDATA
Table 35 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 xdirection; therefore, two rigidbody modes are
computed in this direction when using the SINV method. In all cases the flexible (greater than 0)
frequencies are correct.
Main Index
109
CHAPTER 3
Real Eigenvalue Analysis
Listing 321 shows the output for the rigidbody modes computed for the SINV method when using no
SUPORT. The rigidbody frequencies are denoted by computational zeroes on the order of Hz or
less. Note that the magnitude may be different when the same problem is run on a different computer
type.
Listing 322 shows the output for the rigidbody 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
rigidbody modes have frequencies of 0.0 Hz. The SUPORT entry provides cleaner mode shapes than
those shown in Listing 321 as illustrated by the purely xtranslation, ytranslation, and zrotation
eigenvectors shown in Listing 322.
Table 35 Frequencies for the Unconstrained Beam Models
Mode
Lanczos Method Frequencies (Hz) SINV Method Frequencies (Hz)
No
SUPORT
Stat. Det.
SUPORT
Under
Det.
Over Det.
SUPORT
No
SUPORT
Stat. Det.
SUPORT
Under
Det.
Over Det.
SUPORT
1 2.46E6 0.0 0.0 0.0 1.72E6 0.0 0.0 0.0
2 1.77E6 0.0 0.0 0.0 3.58E6 0.0 0.0 0.0
3 1.26E5 0.0 1.26E5 0.0 4.76E6 0.0 1.01E6 0.0
4 12.82 12.82 12.82 12.82 12.82 12.82 12.82 0.0
5 34.62 34.62 34.62 34.62 34.62 34.62 34.62 12.82
6 66.60 66.60 66.60 34.62
7 66.60
10
5 –
Main Index
MD Nastran Dynamic Analysis User’s Guide
Rigid Body Mode Examples
110
Listing 321 Unconstrained Beam Modes Without SUPORT (SINV Method)
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 1.866121E11 4.319862E06 6.875275E07 1.000000E+00 1.866121E11
2 2 1.819082E09 4.265070E05 6.788069E06 1.000000E+00 1.819082E09
3 3 2.000299E09 4.472470E05 7.118156E06 1.000000E+00 2.000299E09
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
EIGENVALUE = 1.866121E11
CYCLES = 6.875275E07 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 4.273806E34 2.470432E01 0.0 0.0 0.0 1.067134E01
2 G 4.273806E34 2.790573E01 0.0 0.0 0.0 1.067134E01
3 G 4.273806E34 3.110713E01 0.0 0.0 0.0 1.067134E01
4 G 4.273806E34 3.430853E01 0.0 0.0 0.0 1.067134E01
5 G 4.273806E34 3.750993E01 0.0 0.0 0.0 1.067134E01
6 G 4.273806E34 4.071134E01 0.0 0.0 0.0 1.067134E01
7 G 4.273806E34 4.391274E01 0.0 0.0 0.0 1.067134E01
8 G 4.273806E34 4.711414E01 0.0 0.0 0.0 1.067134E01
9 G 4.273806E34 5.031555E01 0.0 0.0 0.0 1.067134E01
10 G 4.273806E34 5.351695E01 0.0 0.0 0.0 1.067134E01
11 G 4.273806E34 5.671835E01 0.0 0.0 0.0 1.067134E01
EIGENVALUE = 1.819082E09
CYCLES = 6.788069E06 R E A L E I G E N V E C T O R N O . 2
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 2.953214E01 5.597336E01 0.0 0.0 0.0 3.291537E01
2 G 2.953214E01 4.609875E01 0.0 0.0 0.0 3.291537E01
3 G 2.953214E01 3.622414E01 0.0 0.0 0.0 3.291537E01
4 G 2.953214E01 2.634953E01 0.0 0.0 0.0 3.291537E01
5 G 2.953214E01 1.647492E01 0.0 0.0 0.0 3.291537E01
6 G 2.953214E01 6.600305E02 0.0 0.0 0.0 3.291537E01
7 G 2.953214E01 3.274305E02 0.0 0.0 0.0 3.291537E01
8 G 2.953214E01 1.314892E01 0.0 0.0 0.0 3.291538E01
9 G 2.953214E01 2.302354E01 0.0 0.0 0.0 3.291538E01
10 G 2.953214E01 3.289815E01 0.0 0.0 0.0 3.291538E01
11 G 2.953214E01 4.277276E01 0.0 0.0 0.0 3.291538E01
EIGENVALUE = 2.000299E09
CYCLES = 7.118156E06 R E A L E I G E N V E C T O R N O . 3
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 2.953611E01 5.596586E01 0.0 0.0 0.0 3.291096E01
2 G 2.953611E01 4.609257E01 0.0 0.0 0.0 3.291096E01
3 G 2.953611E01 3.621928E01 0.0 0.0 0.0 3.291096E01
4 G 2.953611E01 2.634599E01 0.0 0.0 0.0 3.291096E01
5 G 2.953611E01 1.647270E01 0.0 0.0 0.0 3.291096E01
6 G 2.953611E01 6.599414E02 0.0 0.0 0.0 3.291095E01
7 G 2.953611E01 3.273870E02 0.0 0.0 0.0 3.291095E01
8 G 2.953611E01 1.314715E01 0.0 0.0 0.0 3.291095E01
9 G 2.953611E01 2.302044E01 0.0 0.0 0.0 3.291095E01
10 G 2.953611E01 3.289373E01 0.0 0.0 0.0 3.291095E01
11 G 2.953611E01 4.276701E01 0.0 0.0 0.0 3.291095E01
Main Index
111
CHAPTER 3
Real Eigenvalue Analysis
Listing 323 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 rigidbody vectors there is no advantage to using
a SUPORT entry to compute rigidbody modes.
Listing 322 Unconstrained Beam Modes With Statically Determinate SUPORT
(SINV Method)
*** USER INFORMATION MESSAGE 3035 FOR DATA BLOCK KLR
SUPPORT PT.NO. EPSILON STRAIN ENERGY EPSILONS LARGER THAN 0.001 ARE FLAGGED WITH ASTERISKS
1 7.7496606E17 5.5879354E09
2 7.7496606E17 0.0000000E+00
3 7.7496606E17 7.1622708E11
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
2 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
3 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
4 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
5 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
6 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
7 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
8 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
9 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
10 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
11 G 4.176756E01 0.0 0.0 0.0 0.0 0.0
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 4.176756E01 0.0 0.0 0.0 0.0
2 G 0.0 4.176756E01 0.0 0.0 0.0 3.593777E15
3 G 0.0 4.176756E01 0.0 0.0 0.0 6.723840E15
4 G 0.0 4.176756E01 0.0 0.0 0.0 9.575675E15
5 G 0.0 4.176756E01 0.0 0.0 0.0 1.220145E14
6 G 0.0 4.176756E01 0.0 0.0 0.0 1.439249E14
7 G 0.0 4.176756E01 0.0 0.0 0.0 1.596912E14
8 G 0.0 4.176756E01 0.0 0.0 0.0 1.698929E14
9 G 0.0 4.176756E01 0.0 0.0 0.0 1.745155E14
10 G 0.0 4.176756E01 0.0 0.0 0.0 1.770949E14
11 G 0.0 4.176756E01 0.0 0.0 0.0 1.758923E14
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 7.163078E01 0.0 0.0 0.0 4.775385E01
2 G 0.0 5.730462E01 0.0 0.0 0.0 4.775385E01
3 G 0.0 4.297847E01 0.0 0.0 0.0 4.775385E01
4 G 0.0 2.865231E01 0.0 0.0 0.0 4.775385E01
5 G 0.0 1.432615E01 0.0 0.0 0.0 4.775385E01
6 G 0.0 1.508516E14 0.0 0.0 0.0 4.775385E01
7 G 0.0 1.432615E01 0.0 0.0 0.0 4.775385E01
8 G 0.0 2.865231E01 0.0 0.0 0.0 4.775385E01
9 G 0.0 4.297847E01 0.0 0.0 0.0 4.775385E01
10 G 0.0 5.730463E01 0.0 0.0 0.0 4.775385E01
11 G 0.0 7.163078E01 0.0 0.0 0.0 4.775385E01
Main Index
MD Nastran Dynamic Analysis User’s Guide
Rigid Body Mode Examples
112
Listing 323 UIM 3035 Results
Unconstrained Bracket Example
The constraints (SPCs) on the example bracket model from Real Eigenvalue Analysis, 43 are removed to
create an unconstrained model (see Examples, 70 for a description of the model). Figure 323 shows the
bracket model. See Install_dir/mdxxxx/doc/dynamics/bd03rmbkt.dat. The model is a threedimensional
model and therefore produces six rigidbody modes. The MD Nastran results are shown in Listing 324
for the first eight modes. The Lanczos method is used.
Figure 323 Unconstrained Bracket Model
Statically determinate SUPORT:
*** USER INFORMATION MESSAGE 3035 FOR DATA BLOCK KLR
SUPPORT PT.NO. EPSILON STRAIN ENERGY EPSILONS LARGER THAN 0.001 ARE FLAGGED WITH ASTERISKS
1 7.7496606E17 5.5879354E09
2 7.7496606E17 0.0000000E+00
3 7.7496606E17 7.1622708E11
Overdetermined SUPORT:
*** USER INFORMATION MESSAGE 3035 FOR DATA BLOCK KLR
SUPPORT PT.NO. EPSILON STRAIN ENERGY EPSILONS LARGER THAN 0.001 ARE FLAGGED WITH ASTERISKS
1 1.9913979E01 7.2869660E+06 ****
2 1.9913979E01 0.0000000E+00 ****
3 1.9913979E01 7.1850081E11 ****
4 1.9913979E01 7.2869660E+06 ****
Underdetermined SUPORT:
*** USER INFORMATION MESSAGE 4158STATISTICS FOR SYMMETRIC DECOMPOSITION OF DATA BLOCK KLL FOLLOW
MAXIMUM RATIO OF MATRIX DIAGONAL TO FACTOR DIAGONAL = 9.1E+13 AT ROW NUMBER 31
*** USER WARNING MESSAGE 4698. STATISTICS FOR DECOMPOSITION OF MATRIX KLL .
THE FOLLOWING DEGREES OF FREEDOM HAVE FACTOR DIAGONAL RATIOS GREATER THAN
1.00000E+05 OR HAVE NEGATIVE TERMS ON THE FACTOR DIAGONAL.
GRID POINT ID DEGREE OF FREEDOM MATRIX/FACTOR DIAGONAL RATIO MATRIX DIAGONAL
11 R3 9.49483E+13 2.84000E+04
*** USER INFORMATION MESSAGE 3035 FOR DATA BLOCK KLR
SUPPORT PT.NO. EPSILON STRAIN ENERGY EPSILONS LARGER THAN 0.001 ARE FLAGGED WITH ASTERISKS
1 7.6288287E17 5.5879354E09
2 7.6288287E17 7.2759576E12
Main Index
113
CHAPTER 3
Real Eigenvalue Analysis
The six rigidbody modes have computational zero frequencies on the order of Hz. Note that the
magnitudes of the rigidbody 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.
Listing 324 Unconstrained Bracket Frequencies
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 1.690642E07 4.111741E04 6.544039E05 1.000000E+00 1.690642E07
2 2 9.807991E09 9.903530E05 1.576196E05 1.000000E+00 9.807991E09
3 3 5.515176E09 7.426423E05 1.181952E05 1.000000E+00 5.515176E09
4 4 3.390596E09 5.822883E05 9.267405E06 1.000000E+00 3.390596E09
5 5 1.266017E08 1.125174E04 1.790770E05 1.000000E+00 1.266017E08
6 6 1.726585E08 1.313996E04 2.091289E05 1.000000E+00 1.726585E08
7 7 2.649932E+06 1.627861E+03 2.590821E+02 1.000000E+00 2.649932E+06
8 8 4.279463E+06 2.068686E+03 3.292416E+02 1.000000E+00 4.279463E+06
10
5 –
Main Index
MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
114
Tools to Help Identify and Validate Normal Modes
While it is relatively easy to have NASTRAN solve for the normal modes of a structure, making sure you
know what you are getting can be more challenging. This section will identify various tools available to
the analyst to aid in the identification and validation of the normal modes.
• Estimation of the natural frequency
• Animation of the modeshapes
• Modal Effective Mass
• Modal Grid Point Kinetic Energy
• Modal Strain Energy
• Weightcheck
• Groundcheck
Many of these are also described in Debugging Dynamic Models (Ch. 14).
Estimation of Natural Frequency
There are handbooks available that provide closed form solutions to common simple structures, such as
beam and plates. In the real world, you can quickly move away from such simple structures.
Sometimes you are working with a model that you may not have much familiarity with and aren’t sure if
the frequencies solved are even in the correct range. You can use the following formula to get a rough
estimate of the primary modes:
Estimation of Natural Frequency:
where
For the above simple model (L=10in, E=30E6, =.3, =7.4E4 and cross section as shown), using the
above equation and NASTRAN to compute the natural frequencies result in a pretty good estimation. Of
= center of gravity displacement in direction of applied gload
= acceleration due to gravity
f
1
2t

g
d
 =
d
g
u p
Main Index
115
CHAPTER 3
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 xy plane bending mode should be near 100Hz instead of 2000 Hz.
Animation for Mode Identification
When it comes to characterizing a mode, one of the best tools is visualization. Using a postprocessor
to animate the mode shapes can be the most effective means of characterization. The ability to rotate the
display during animation helps to see what is moving even if the initial view obscures the motion.
There are a couple conditions in which an animation can be misleading. If there is a point mass rotating
or a line element torsioning about its own axis, there isn’t anything for the postprocessor to display. In
this case, the postprocessor will amplify any other motion even if it is ‘noise’. This can be very hard to
recognize, and good modeling techniques should be used to eliminate this risk.
For example, using the beam model in the previous section, the 4
th
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:
Mode Direction Prediction From FEM
Fx 5236 Hz 5028 Hz
Fy 109 Hz 81 Hz
Fz 38 Hz 28 Hz
Main Index
MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
116
As can be seen, there is some very small outofplane 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 crosssectional connectivity to ensure the
postprocessor can display the appropriate deformation.
Modal Effective Mass
Since much of the quality of a dynamics solution depends on the accuracy of the mass, having a better
understanding of how the mass participates in each of the modes is important. Being able to identify what
direction the mass in a mode is moving, and also, how much mass each mode contributes to the total
modal mass of a structure can offer much insight into the characteristics of the structure being analyzed.
Until some kind of loading is applied, either transient or frequency response, it is very difficult to predict
which modes will play a dominant role in the response of a structure. One method to help predict what
the important modes are is to calculate what is called the modal participation factors. Linear
combinations of eigenvectors can be used to define any vector. This is because eigenvectors calculated
in a normal modes analysis are linearly independent of each other and span the vector space used to
define the model response. A rigid body vector can be constructed from a set of flexible body
eigenvectors, with it having motion in a desired direction.
Rigid body vector is defined as
Use where is a vector of scaling factors for the eigenvectors in .
Premultiply the expression for by
where is the diagonal matrix of generalized masses for the normal modes.
D { }
R
o   D { }
R
o   c { } = c { } o  
D { }
R
o  
T
M  
o  
T
M   D { }
R
o  
T
M   o   c { } =
o  
T
M   D { }
R
m   c { } =
m  
Main Index
117
CHAPTER 3
Real Eigenvalue Analysis
The term is commonly known as the participation factor, .
The scaling factor multiplies the generalized mass to define the participation factor .
The relationship between the rigid body vector and the corresponding rigid body mass, , is
Using
Then
So the contribution which the ith mode provides to the rigid body mass is . This is known as
the modal effective mass. If you mass normalize the eigenvectors, , then the
participation factors are , and the modal effective mass is . 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:
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.
o  
T
M   D { }
R
I { }
I { } o  
T
M   D { }
R
m   c { } = = I
i
m
i i
c
i
= ¬
c
i
m
i i
I
i
D { }
R
M
R
M
R
D { }
R
T
M   D { }
R
=
D { }
R
o   c { } =
M
R
c { }
T
o  
T
M   o   c { } c { }
T
m   c { } = =
M
R
c
i
2
m
i i
o  
T
M   o   I   =
I
i
c
i
= c
i
2
MEFFMASS
PRINT
NOPRINT
PUNCH
NOPUNCH
GRID gid,
SUMMARY, PARTFAC,
MEFFM, MEFFW,
FRACSUM, ALL
= .
YES
NO ¹ )
´ `
¦ ¹
=
Main Index
MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
118
The values calculated are used by different industries in different ways:
For example, in Civil Engineering seismic analysis, the contribution from each mode is assessed as a
percentage and the total is summed. Any shortfall from 100% is classified as ‘missing mass’. If the
missing mass is significant, then it may indicate errors in the analysis, typically insufficient modes being
used in a modal method. ‘Missing mass’ is often characterized as higher frequency body type loading
and can be simulated by applying a 1g inertia load in the appropriate direction, factored by the % missing
mass, then added as a static load.
In testing, many times, the modal effective mass is used as a way of measuring whether sufficient modes
have been retained.
Let’s look at the output found in the f06 file for a eigensolution solving for the first 10 modes of the
cantilevered beam. See Install_dir/mdxxxx/doc/dynamics/beam103m10.dat.
The Effective Mass Matrix is the 6x6 mass represented by the modes, while the ASET 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
MEFFW Requests calculation of the modal effective mass in units of weight.
FRACSUM Requests calculation of the modal effective mass fraction.
Describer Meaning
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
TOTAL EFFECTIVE MASS FRACTION
REFERENCE POINT AT ORIGIN OF BASIC COORDINATE SYSTEM
T1 T2 T3 R1 R2 R3
9.654936E18 9.106090E01 9.657329E01 9.544338E01 9.996986E01 9.985170E01
EFFECTIVE MASS MATRIX
*** ***
* 4.140326E22 3.485344E14 2.780172E16 8.022095E17 1.546341E16 4.481505E14 *
* 3.485344E14 3.904965E05 5.777767E13 4.721861E12 3.825362E13 2.228940E04 *
* 2.780172E16 5.777767E13 4.141352E05 1.762713E06 2.248915E04 4.310256E13 *
* 8.022095E17 4.721861E12 1.762713E06 3.104371E07 9.389775E06 2.560531E11 *
* 1.546341E16 3.825362E13 2.248915E04 9.389775E06 1.511734E03 2.523439E13 *
* 4.481505E14 2.228940E04 4.310256E13 2.560531E11 2.523439E13 1.510022E03 *
*** ***
ASET RIGID BODY MASS MATRIX
*** ***
* 4.288300E05 0.000000E+00 0.000000E+00 0.000000E+00 4.872074E12 1.784896E06 *
* 0.000000E+00 4.288300E05 0.000000E+00 4.872074E12 0.000000E+00 2.257000E04 *
* 0.000000E+00 0.000000E+00 4.288300E05 1.784896E06 2.257000E04 0.000000E+00 *
* 0.000000E+00 4.872074E12 1.784896E06 3.252578E07 9.394190E06 2.564249E11 *
* 4.872074E12 0.000000E+00 2.257000E04 9.394190E06 1.512190E03 2.027877E13 *
* 1.784896E06 2.257000E04 0.000000E+00 2.564249E11 2.027877E13 1.512264E03 *
*** ***
Main Index
119
CHAPTER 3
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
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.
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.
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
TOTAL EFFECTIVE MASS FRACTION
REFERENCE POINT AT ORIGIN OF BASIC COORDINATE SYSTEM
T1 T2 T3 R1 R2 R3
8.497244E01 9.990354E01 1.000000E+00 1.000000E+00 1.000000E+00 9.999877E01
MODAL EFFECTIVE MASS FRACTION
(FOR TRANSLATIONAL DEGREES OF FREEDOM)
MODE FREQUENCY T1 T2 T3
NO.
FRACTION SUM FRACTION SUM FRACTION SUM
1 2.784550E+01 0.000000E+00 0.000000E+00 1.489817E17 1.489817E17 6.432269E01 6.432269E01
2 8.110487E+01 1.416777E33 1.416777E33 6.431246E01 6.431246E01 4.896923E18 6.432269E01
3 1.716828E+02 0.000000E+00 1.416777E33 4.558935E17 6.431246E01 1.990311E01 8.422580E01
4 3.988035E+02 1.505143E38 1.416792E33 2.485408E13 6.431246E01 4.099026E04 8.426679E01
5 4.732146E+02 1.784275E37 1.416970E33 6.710384E17 6.431246E01 6.780655E02 9.104744E01
6 5.009878E+02 1.028366E32 1.170063E32 1.990124E01 8.421370E01 4.815021E16 9.104744E01
7 9.043718E+02 1.564099E36 1.170219E32 3.722504E15 8.421370E01 3.474705E02 9.452215E01
MODAL EFFECTIVE MASS FRACTION
(FOR ROTATIONAL DEGREES OF FREEDOM)
MODE FREQUENCY R1 R2 R3
NO. FRACTION SUM FRACTION SUM FRACTION SUM
1 2.784550E+01 1.484940E01 1.484940E01 9.704365E01 9.704365E01 1.821291E17 1.821291E17
2 8.110487E+01 1.202327E12 1.484940E01 5.794401E18 9.704365E01 9.703533E01 9.703533E01
3 1.716828E+02 5.929535E02 2.077893E01 2.488552E02 9.953220E01 5.050688E17 9.703533E01
4 3.988035E+02 6.633550E01 8.711444E01 3.122559E07 9.953223E01 1.569335E14 9.703533E01
5 4.732146E+02 2.670249E03 8.738146E01 3.223244E03 9.985456E01 3.438163E18 9.703533E01
6 5.009878E+02 7.208571E14 8.738146E01 5.449564E18 9.985456E01 2.495353E02 9.953068E01
7 9.043718E+02 8.491335E03 8.823059E01 8.347176E04 9.993803E01 1.199591E16 9.953068E01
Main Index
MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
120
By reviewing the printout of the full, 30mode solution, you can see the 23
rd
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 9
th
mode and the T3 direction by the 5
th
mode.
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.
Element Strain Energy (ESE)
Element Strain Energy is a terrific tool for determining which elements to modify the least, but have the
greatest potential for impacting results. In the context of a normal modes solution, the modal element
strain energy cannot be compared across modes. But within a mode, the modal strain energy can still
provide useful relative information. If you find the same elements having a large amount of modal strain
energy across several modes, these elements are good candidates for modification. Changing the
stiffness of the elements with the most ESE in a mode is often the most efficient way to shift the
frequency of that mode.
Element Strain Energy can be defined as .
Example–two springs in series. Which is the best element to stiffen to reduce the tip deflection?
MODAL EFFECTIVE MASS FRACTION
(FOR TRANSLATIONAL DEGREES OF FREEDOM)
MODE FREQUENCY T1 T2 T3
NO. FRACTION SUM FRACTION SUM FRACTION SUM
1 2.784550E+01 0.000000E+00 0.000000E+00 1.489817E17 1.489817E17 6.432269E01 6.432269E01
2 8.110487E+01 1.416777E33 1.416777E33 6.431246E01 6.431246E01 4.896923E18 6.432269E01
3 1.716828E+02 0.000000E+00 1.416777E33 4.558935E17 6.431246E01 1.990311E01 8.422580E01
4 3.988035E+02 1.505143E38 1.416792E33 2.485408E13 6.431246E01 4.099026E04 8.426679E01
5 4.732146E+02 1.784275E37 1.416970E33 6.710384E17 6.431246E01 6.780655E02 9.104744E01
6 5.009878E+02 1.028366E32 1.170063E32 1.990124E01 8.421370E01 4.815021E16 9.104744E01
7 9.043718E+02 1.564099E36 1.170219E32 3.722504E15 8.421370E01 3.474705E02 9.452215E01
8 1.189065E+03 6.334474E34 1.233564E32 2.883686E13 8.421370E01 3.731063E05 9.452588E01
9 1.381379E+03 4.071596E34 1.274280E32 6.847200E02 9.106090E01 4.501012E14 9.452588E01
10 1.455066E+03 3.527966E34 1.309559E32 1.138931E13 9.106090E01 2.047412E02 9.657329E01
11 1.936458E+03 6.510201E32 7.819760E32 1.070572E15 9.106090E01 1.602071E03 9.673350E01
12 2.105694E+03 1.884441E32 9.704201E32 1.023835E15 9.106090E01 1.134227E02 9.786772E01
13 2.612863E+03 1.380257E31 2.350677E31 1.640671E11 9.106090E01 3.072468E03 9.817497E01
14 2.657074E+03 1.387716E33 2.364555E31 3.503040E02 9.456394E01 2.751159E12 9.817497E01
15 2.814768E+03 1.057610E31 3.422164E31 1.474569E12 9.456394E01 5.159775E03 9.869095E01
16 3.195672E+03 2.073240E30 2.415457E30 1.317168E13 9.456394E01 2.593142E03 9.895026E01
17 3.517616E+03 7.847688E29 8.089234E29 4.580544E13 9.456394E01 2.399510E03 9.919021E01
18 3.628539E+03 8.663078E29 1.675231E28 7.837101E16 9.456394E01 1.583888E03 9.934860E01
19 3.881244E+03 1.573308E30 1.690964E28 9.830500E15 9.456394E01 5.597762E04 9.940458E01
20 4.206083E+03 1.264551E28 2.955515E28 2.845621E12 9.456394E01 1.768755E03 9.958146E01
21 4.294404E+03 1.335030E27 1.630581E27 2.106229E02 9.667017E01 1.901734E13 9.958146E01
22 4.861434E+03 1.209969E31 1.630702E27 2.754001E16 9.667017E01 1.410443E03 9.972250E01
23 5.028497E+03 8.497244E01 8.497244E01 2.459785E27 9.667017E01 2.808369E28 9.972250E01
24 5.518441E+03 8.815934E27 8.497244E01 3.039200E14 9.667017E01 1.342143E03 9.985672E01
25 6.183626E+03 1.066271E26 8.497244E01 2.506868E13 9.667017E01 1.045135E03 9.996123E01
26 6.232633E+03 2.659922E27 8.497244E01 1.379041E02 9.804921E01 1.683167E14 9.996123E01
27 6.699681E+03 1.637151E28 8.497244E01 1.153580E15 9.804921E01 3.877237E04 1.000000E+00
28 8.367291E+03 1.238390E27 8.497244E01 9.297304E03 9.897894E01 2.567073E16 1.000000E+00
29 1.051664E+04 3.297015E26 8.497244E01 6.011833E03 9.958013E01 1.701874E16 1.000000E+00
30 1.240296E+04 1.137721E26 8.497244E01 3.234186E03 9.990354E01 9.269108E17 1.000000E+00
ESE
e
0.5 = u
e
{ }
T
K
ee
  u
e
{ } ·
Main Index
121
CHAPTER 3
Real Eigenvalue Analysis
where:
for ,
where
for
where:
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 postprocessor.
= 10
= 1
= 1
1 2
K
2
K
1
P
ESE
1
2
 KA
2
=
K
1
K
2
P
K
1
ESE
1
1
2
 K
1
A
1
2 1
2

P
2
K
1

1
20
 = = =
A
1
P
K
1
 =
K
2
ESE .
2
1
2
 K
2
A
2
A
1
– ( )
2 1
2
 K
2
P
K
1

P
K
2

P
K
1
 – +
\ .
 
= =
1
2
 K
2
P
K
2

\ .
 
2
=
1
2
 =
2 1
2
1
1
K
P
K
P
K
P
+ = A
= A
A
1
P
K
1
 =
A
2
P
K
1

P
K
2
 + =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
122
Requests the output of the strain energy in selected elements.
Format:
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.
ESE Element Strain Energy Output Request
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.
ESE (
PRINT, PUNCH
PLOT
AVERAGE
AMPLITUDE
PEAK
THRESH p =   )
ALL
n ¹ )
´ `
¦ ¹
=
Main Index
123
CHAPTER 3
Real Eigenvalue Analysis
Grid Point Kinetic Energy (GPKE)
In normal modes, Grid Point Kinetic Energy information can be very helpful in identifying how much
each DOF is participating in the motion of a particular mode. Like modal element strain energy, modal
grid point kinetic energy cannot be compared across modes. But, within a mode, the participation of
DOF in the mode’s motion can be seen. If the GPKE is shared across many DOF in the model, this points
towards the mode being a global mode, while having a smaller number of DOF showing up with
significant GPKE in a mode indicates a local mode.
While Element Strain Energy is a stiffnessbiased calculation, Grid Point Kinetic Energy is a mass
biased calculation. This can help provide insight into which DOF are moving the most and have the most
mass participating in that motion.
The results of the Grid Point Kinetic Energy request GPKE have the same format as a DISPLACEMENT
output. However, the values represent the relative amount of kinetic energy in each degreeoffreedom
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:
Where indicates that the massnormalized eigenvectors so that the total grid
point kinetic energy is scaled to be unity and the operator indicates termwise 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.
E L E M E N T S T R A I N E N E R G I E S
ELEMENTTYPE = BEAM * TOTAL ENERGY OF ALL ELEMENTS IN PROBLEM = 1.530523E+04
MODE 1 * TOTAL ENERGY OF ALL ELEMENTS IN SET 1 = 1.530523E+04
0
ELEMENTID STRAINENERGY PERCENT OF TOTAL STRAINENERGYDENSITY
1 5.293249E+03 34.5846 8.677457E+05
2 3.856094E+03 25.1946 6.321465E+05
3 2.657727E+03 17.3648 4.356930E+05
4 1.704405E+03 11.1361 2.794107E+05
5 9.933071E+02 6.4900 1.628372E+05
6 5.074111E+02 3.3153 8.318216E+04
7 2.138926E+02 1.3975 3.506436E+04
8 6.648590E+01 0.4344 1.089933E+04
9 1.197225E+01 0.0782 1.962665E+03
10 6.816160E01 0.0045 1.117403E+02
E
k
g
u
g
mass
= M
gg
u
g
mass
  ©
u
g
mass
©
Main Index
MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
124
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.
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:
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
2 G 7.326856E30 1.330765E19 1.119430E02 9.668289E05 7.297595E30 1.350100E33
3 G 2.651446E29 1.021578E18 1.615364E01 7.219911E04 2.642917E29 4.841049E33
4 G 4.687660E29 3.287765E18 7.366567E01 2.286668E03 4.702317E29 8.516130E33
5 G 6.409843E29 7.363459E18 2.089146E+00 5.060130E03 6.409102E29 1.167118E32
6 G 7.439181E29 1.344727E17 4.557218E+00 9.147877E03 7.560056E29 1.332900E32
7 G 9.776883E29 2.148098E17 8.408554E+00 1.447662E02 9.688099E29 1.732981E32
8 G 7.389515E29 3.114971E17 1.381095E+01 2.079251E02 7.530929E29 1.326966E32
9 G 5.441464E29 4.190508E17 2.082973E+01 2.767412E02 5.501916E29 9.839795E33
10 G 6.935092E29 5.300563E17 2.944991E+01 3.455821E02 7.066868E29 1.226697E32
11 G 3.286211E29 3.178247E17 1.980990E+01 2.038623E02 3.317488E29 5.708364E33
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
POINT ID. TYPE T1 T2 T3 R1 R2 R3
2 G 2.615311E27 3.624955E13 1.319956E02 6.299611E01 1.009845E27 1.605532E27
3 G 1.205652E26 2.209151E12 1.249419E01 2.313763E+00 7.865180E27 4.191535E27
4 G 1.987451E26 4.536370E12 2.882752E01 4.589959E+00 1.850854E26 1.365973E27
5 G 1.930245E26 4.420710E12 1.649861E01 7.029121E+00 2.717585E26 7.872930E27
6 G 5.526213E27 1.182677E13 4.420250E01 9.450140E+00 2.790627E26 2.238171E26
7 G 1.964410E26 6.056167E12 1.128664E+00 1.192419E+01 1.722487E26 3.687122E26
8 G 4.904750E26 7.357273E12 1.210967E+00 1.457577E+01 1.771801E27 5.082134E26
9 G 9.178819E26 4.910897E12 5.034333E01 1.733273E+01 1.623074E26 7.556239E26
10 G 1.169130E25 3.781622E11 4.962833E01 1.981653E+01 2.917142E26 8.774242E26
11 G 5.705810E26 4.742649E11 6.010876E01 1.074152E+01 1.518653E26 4.187053E26
GPKE Grid Point Kinetic Energy Output Request
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.
GPKE
PRINT
NOPRINT
PUNCH, THRESH e = ( ) .
ALL
n
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
Main Index
125
CHAPTER 3
Real Eigenvalue Analysis
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 gset.
Format:
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 Rigid Body Mass Reduction Check
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).
Describer Meaning
WEIGHTCHECK
(
PRINT
NOPRINT
, SET (
G, N, N AUTOSPC, F, A,V +
ALL ¹ )
´ `
¦ ¹
) =
GRID gid, CGI
YES
NO
WEIGHT
MASS
) . = =
YES
NO ¹ )
´ `
¦ ¹
=
Main Index
MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
126
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).
gid Reference grid point for the calculation of rigid body motion. The default
is the origin of the basic coordinate system.
CGI For , 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).
SET
Keywords DOF Set Description
G gset before single point, multipoint constraints, and rigid elements are
applied
N nset after multipoint constraints and rigid elements are applied
N+AUTO nset with
AUTOSPC
same as the nset with the rows/columns in the mass matrix
corresponding to degrees of freedom constrained by the
PARAM,AUTOSPC operation zeroed out
F Fset after single point, multipoint constraints, and rigid elements are
applied
A aset after static condensation
V vset Component mode mass which combines the oset, rset and cset
Describer Meaning
SET G =
Note: The units of weight or mass output depends on input, it is independent of param,wtmass,x.
Main Index
127
CHAPTER 3
Real Eigenvalue Analysis
Using the sample beam model, here are some different set level outputs.
Below, you see the printout for the NSet. Since there are no MPC’s in this model, the NSet is the same
as the GSet. If it was desired to see the CG and inertia terms, CGI=YES could have been added to the
WEIGHTCHECK command.
Now, the FSet printout is shown. Here you can see a change with respect to the GSet. As you move
down the set processing, the FSet 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.
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
* 4.514000E05 0.000000E+00 0.000000E+00 0.000000E+00 5.128499E12 1.878838E06 *
* 0.000000E+00 4.514000E05 0.000000E+00 5.128499E12 0.000000E+00 2.257000E04 *
* 0.000000E+00 0.000000E+00 4.514000E05 1.878838E06 2.257000E04 0.000000E+00 *
* 0.000000E+00 5.128499E12 1.878838E06 3.423766E07 9.394190E06 2.564249E11 *
* 5.128499E12 0.000000E+00 2.257000E04 9.394190E06 1.512190E03 2.134607E13 *
* 1.878838E06 2.257000E04 0.000000E+00 2.564249E11 2.134607E13 1.512268E03 *
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 XC.G. YC.G. ZC.G.
X 4.514000E05 0.000000E+00 4.162246E02 1.136132E07
Y 4.514000E05 5.000000E+00 0.000000E+00 1.136132E07
Z 4.514000E05 5.000000E+00 4.162246E02 0.000000E+00
I(S)
* 2.641748E07 0.000000E+00 3.231174E27 *
* 0.000000E+00 3.836900E04 2.524355E29 *
* 3.231174E27 2.524355E29 3.836900E04 *
I(Q)
* 2.641748E07 *
* 3.836900E04 *
* 3.836900E04 *
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 *
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
* 4.514000E05 0.000000E+00 0.000000E+00 0.000000E+00 5.128499E12 1.878838E06 *
* 0.000000E+00 4.514000E05 0.000000E+00 5.128499E12 0.000000E+00 2.257000E04 *
* 0.000000E+00 0.000000E+00 4.514000E05 1.878838E06 2.257000E04 0.000000E+00 *
* 0.000000E+00 5.128499E12 1.878838E06 3.423766E07 9.394190E06 2.564249E11 *
* 5.128499E12 0.000000E+00 2.257000E04 9.394190E06 1.512190E03 2.134607E13 *
* 1.878838E06 2.257000E04 0.000000E+00 2.564249E11 2.134607E13 1.512268E03 *
0 CHANGE WITH RESPECT TO GSET RIGID BODY MASS/WEIGHT
* 0 % *
* 0 % *
* 0 % *
* 0 % *
* 0 % *
* 0 % *
Main Index
MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
128
After the FSet, we see the ASet. Since there were no further reduction in the sample cantilevered beam
(no ASET or OSET definitions), the ASet is no different from the FSet.
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.
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
* 4.288300E05 0.000000E+00 0.000000E+00 0.000000E+00 4.872074E12 1.784896E06 *
* 0.000000E+00 4.288300E05 0.000000E+00 4.872074E12 0.000000E+00 2.257000E04 *
* 0.000000E+00 0.000000E+00 4.288300E05 1.784896E06 2.257000E04 0.000000E+00 *
* 0.000000E+00 4.872074E12 1.784896E06 3.252578E07 9.394190E06 2.564249E11 *
* 4.872074E12 0.000000E+00 2.257000E04 9.394190E06 1.512190E03 2.027877E13 *
* 1.784896E06 2.257000E04 0.000000E+00 2.564249E11 2.027877E13 1.512264E03 *
0 CHANGE WITH RESPECT TO GSET RIGID BODY MASS/WEIGHT
* 5 % *
* 5 % *
* 5 % *
* 5 % *
* 0 % *
* 0 % *
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
* 4.288300E05 0.000000E+00 0.000000E+00 0.000000E+00 4.872074E12 1.784896E06 *
* 0.000000E+00 4.288300E05 0.000000E+00 4.872074E12 0.000000E+00 2.257000E04 *
* 0.000000E+00 0.000000E+00 4.288300E05 1.784896E06 2.257000E04 0.000000E+00 *
* 0.000000E+00 4.872074E12 1.784896E06 3.252578E07 9.394190E06 2.564249E11 *
* 4.872074E12 0.000000E+00 2.257000E04 9.394190E06 1.512190E03 2.027877E13 *
* 1.784896E06 2.257000E04 0.000000E+00 2.564249E11 2.027877E13 1.512264E03 *
0 CHANGE WITH RESPECT TO GSET RIGID BODY MASS/WEIGHT
* 5 % *
* 5 % *
* 5 % *
* 5 % *
* 0 % *
* 0 % *
Main Index
129
CHAPTER 3
Real Eigenvalue Analysis
Perform grounding check analysis on the stiffness matrix to expose unintentional constraints by moving
the model rigidly.
Format:
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 GSet. As shown in the informational
message, common reasons for failing at this level are CELAS elements connected to only one grid, or
connected between noncoincident grids or improper DMIG matrices.
GROUNDCHECK Rigid Body Motion Grounding Check
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 degreeoffreedom 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).
GROUNDCHECK
(
PRINT
NOPRINT
PUNCH, SET . (
G, N, N AUTOSPC, F, A +
ALL ¹ )
´ `
¦ ¹
) =
GRID gid, THRESH e,DATAREC =
YES
NO
RTHRESH r = ( )) . = =
YES
NO ¹ )
´ `
¦ ¹
=
Main Index
MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
130
Looking at the FSet, 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 FSet, grid 1 doesn’t exist. So grid 2 shows the grounding force.
*** USER INFORMATION MESSAGE 7570 (GPWG1D)
RESULTS OF RIGID BODY CHECKS OF MATRIX KGG (GSET) FOLLOW:
PRINT RESULTS IN ALL SIX DIRECTIONS AGAINST THE LIMIT OF 3.660000E05
DIRECTION STRAIN ENERGY PASS/FAIL
  
1 0.000000E+00 PASS
2 0.000000E+00 PASS
3 0.000000E+00 PASS
4 0.000000E+00 PASS
5 8.640200E12 PASS
6 2.478373E11 PASS
SOME POSSIBLE REASONS MAY LEAD TO THE FAILURE:
1. CELASI ELEMENTS CONNECTING TO ONLY ONE GRID POINT;
2. CELASI ELEMENTS CONNECTING TO NONCOINCIDENT POINTS;
3. CELASI ELEMENTS CONNECTING TO NONCOLINEAR DOF;
4. IMPROPERLY DEFINED DMIG MATRICES;
*** USER INFORMATION MESSAGE 7570 (GPWG1D)
RESULTS OF RIGID BODY CHECKS OF MATRIX KFF (FSET) FOLLOW:
PRINT RESULTS IN ALL SIX DIRECTIONS AGAINST THE LIMIT OF 3.660000E05
DIRECTION STRAIN ENERGY PASS/FAIL
  
1 9.150000E+04 FAIL
2 4.467632E+03 FAIL
3 6.424691E+02 FAIL
4 3.360738E+00 FAIL
5 2.171048E+02 FAIL
6 1.754427E+03 FAIL
SOME POSSIBLE REASONS MAY LEAD TO THE FAILURE:
1. CONSTRAINTS WHICH PREVENT RIGIDBODY MOTION.
Main Index
131
CHAPTER 3
Real Eigenvalue Analysis
G R O U N D C H E C K F O R C E S ( F  S E T )
POINT ID. TYPE T1 T2 T3 R1 R2 R3
2 G 1.830000E+05 0.0 0.0 0.0 0.0 0.0
DIRECTION 2
G R O U N D C H E C K F O R C E S ( F  S E T )
POINT ID. TYPE T1 T2 T3 R1 R2 R3
2 G 0.0 8.935264E+03 0.0 0.0 0.0 4.467632E+03
DIRECTION 3
G R O U N D C H E C K F O R C E S ( F  S E T )
POINT ID. TYPE T1 T2 T3 R1 R2 R3
2 G 0.0 0.0 1.284938E+03 0.0 6.424691E+02 0.0
DIRECTION 4
G R O U N D C H E C K F O R C E S ( F  S E T )
POINT ID. TYPE T1 T2 T3 R1 R2 R3
2 G 0.0 0.0 0.0 6.721477E+00 0.0 0.0
DIRECTION 5
G R O U N D C H E C K F O R C E S ( F  S E T )
POINT ID. TYPE T1 T2 T3 R1 R2 R3
2 G 0.0 0.0 6.424691E+02 0.0 2.082595E+02 0.0
DIRECTION 6
G R O U N D C H E C K F O R C E S ( F  S E T )
POINT ID. TYPE T1 T2 T3 R1 R2 R3
2 G 7.616911E+03 4.467632E+03 0.0 0.0 0.0 9.587776E+02
Main Index
MD Nastran Dynamic Analysis User’s Guide
Tools to Help Identify and Validate Normal Modes
132
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 4: Frequency Response Analysis
4
Frequency Response Analysis
Overview
Review of Dynamic Excitations
Direct Frequency Response Analysis
Frequency Response Function (FRF) and FRF Based Assembly (FBA)
Modal Frequency Response Analysis
Modal Versus Direct Frequency Response
FrequencyDependent Structural Elements
FrequencyDependent Excitation Definition
Solution Frequencies
Frequency Response Considerations
Solution Control for Frequency Response Analysis
Modal Contribution Fraction (MCFRACTION)
Examples
Main Index
Overview
Frequency response analysis is a method used to compute structural response to steadystate 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 steadystate 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 41.
Figure 41 Phase Shift
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
135
CHAPTER 4
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 42.
Figure 42 Complex Plane
where:
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.
= magnitude =
= phase angle =
= real component =
= imaginary component =
r (Real)
i (Imaginary)
u
u
i
u
r
u
u
u
r
2
u
i
2
+
u
tan
1 –
u
i
u
r
( )
u
r
u u cos
u
i
u u sin
Main Index
MD Nastran Dynamic Analysis User’s Guide
Review of Dynamic Excitations
136
Review of Dynamic Excitations
The methods used for generating dynamic loads in MD Nastran are very different from those used for
static loads. Dynamic loads generally vary with time or frequency. They may also be applied with
different phases or time lags to different portions of the structure. The most general case would be a
structure having a different load history for each direction on each point on the model, which could
require thousands of tabular inputs.
Fortunately, most applications require only a simple pattern of load variations with time and geometry.
The engineer usually encounters a single time function applied to a portion of the structure or a general
time function applied to a few structural points. An example of the first case would be wind gust loads
on an aircraft. An example of the second case woul be an automobile traveling over a rough road.
MD Nastran requires only a minimum input for either case.
Because of the many possible types of dynamic loading conditions, the input options are also numerous.
Most of the basic options are described in Modal Versus Direct Frequency Response, 152 and Modal
Versus Direct Transient Response, 218. Some of the advanced methods are described below.
Subcases in Dynamic Analysis
The Case Control options in dynamic analysis are somewhat less flexible than statics and nonlinear
analysis. Some of the statics options that are not available in dynamic response analysis and complex
eigenvalue analysis include:
Changes in boundary conditions between subcases for transient response analysis.
SUBSEQ and SUBCOM solution combinations.
Grid point forces outputs for frequency response analysis.
Otherwise, each type of solution uses the Case Control subcase for a different purpose. Some uses for
subcases in dynamics analysis are given below.
1. Subcases are essential for most dynamic optimization problems (SOL 200). Each subcase may
define a different type of analysis or a different load.
2. In a superelement analysis, the subcases may be used to specify different upstream LOADSET
data and output processing for different SE components.
3. Subcases are recommended in nonlinear transient analysis (SOL 129) for applying loads in a
sequence or changing the time step method. A single load may be applied in steps or a second
load may be added with new subcases. Changing time step size or the solution method may be
necessary to overcome a troublesome divergence region. Subcases also define the allowable
solutions that may be used in restarts. Use them liberally.
4. Multiple subcases are available for frequency response analysis (SOLs 108, 111 and 118) for the
purpose of solving multiple loading conditions more efficiently. (Each frequency requires a
matrix decomposition and each additional load vector may be processed at this time with small
cost.) Another use is in random analysis where several loads need to be combined, each with a
different spectral density distribution.
Main Index
137
CHAPTER 4
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 43. 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
MD Nastran Dynamic Analysis User’s Guide
Review of Dynamic Excitations
138
Figure 43 Input Data Relationships for Dynamic Loads
• 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.
Case Control
Bulk Data
FORCEi
PLOADi
GRAV
etc.
..
LSEQ
DAREA
DPHASE
DELAY
Load Vectors
P(x,y,z)
TLOADi
RLOADi
DLOAD
TABLEDi
F(t) or F(f)
Dynamic Loads
P(t) or P(f)
Available to Superelements
LOADSET =
DLOAD =
Dynamic
Coefficients
Static Loads
Time or Frequency
Functions
DAREA
without LSEQ
Main Index
139
CHAPTER 4
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 degreeoffreedom, 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 LSEQtype 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 longperiod 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
MD Nastran Dynamic Analysis User’s Guide
Review of Dynamic Excitations
140
Setup for Random Analysis
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
141
CHAPTER 4
Frequency Response Analysis
Direct Frequency Response Analysis
In direct frequency response analysis, structural response is computed at discrete excitation frequencies
by solving a set of coupled matrix equations using complex algebra. Begin with the damped forced
vibration equation of motion with harmonic excitation
(41)
The load in Eq. (41) is introduced as a complex vector, which is convenient for the mathematical
solution of the problem. From a physical point of view, the load can be real or imaginary, or both. The
same interpretation is used for response quantities.
For harmonic motion (which is the basis of a frequency response analysis), assume a harmonic solution
of the form:
(42)
where is a complex displacement vector. Taking the first and second derivatives of Eq. (42),
the following is obtained:
(43)
When the above expressions are substituted into Eq. (41), the following is obtained:
(44)
which after dividing by simplifies to
(45)
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.
Damping in Direct Frequency Response
Damping simulates the energy dissipation characteristics of a structure. Damping in direct frequency
response is represented by the damping matrix and additions to the stiffness matrix .
The damping matrix is comprised of several matrices
(46)
M   x
··
t ( ) { } B   x
·
t ( ) { } K   x t ( ) { } + + P e ( ) { }e
i et
=
x { } u e ( ) { }e
i et
=
u e ( ) { }
x
·
{ } ie u e ( ) { }e
i et
=
x
··
{ } e –
2
u e ( ) { }e
i et
=
e
2
– M   u e ( ) { }e
i et
ie B   u e ( ) { }e
i et
K   u e ( ) { }e
i et
+ + P e ( ) { }e
i et
=
e
i et
e
2
– M ieB K + +   u e ( ) { } P e ( ) { } =
e
B   K  
B   B
1
  B
2
  + =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Direct Frequency Response Analysis
142
where:
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:
(47)
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.
= damping elements (CVISC, CDAMPi and hybrid damping*) and B2GG
*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.
= B2PP direct input matrix and transfer functions
= global stiffness matrix
= overall structural damping coefficient (PARAM,G)
= element stiffness matrix
= element structural damping coefficient (GE on the MATi entry)
B
1
 
B
2
 
K   1 iG + ( ) K   i G
E
K
E
 
¯
+ =
K  
G
K
E
 
G
E
Main Index
143
CHAPTER 4
Frequency Response Analysis
Frequency Response Function (FRF) and FRF Based
Assembly (FBA)
Introduction
Frequency Response Functions (FRFs) can be used to represent components. The FRFs of such
components can then be combined to yield the FRFs of assemblies of such components. Such an FRF
based assembly (FBA) process represents an alternative to component mode synthesis (or modal
substructuring). This method also lends itself to tracking load paths or energy flows through a structure,
commonly referred as Transfer Path Analysis (TPA). This approach may be regarded as an alternative
to mode participation studies. The FRF/FBA capability provides an excellent tool for NVH studies.
FRF Concept
Using the usual symbols, the equation of motion for frequency response analysis is given by:
(48)
where is the dynamic stiffness matrix of the system.
Solving the above equation for the response displacements , we obtain
(49)
The 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 of the FRF matrix represents the response displacement at the ith DOF due to a unit load
applied to the jth DOF. Thus, the rows in , the jth column of the FRF, represent the displacements
at all response DOFs due to a unit load applied to the jth DOF. Similarly, the columns in the ith row of
the FRF represent the response displacements at the ith DOF due to unit loads applied to all excitation
DOFs.
In practice, the FRF matrix is computed by solving Eq. (49) for a series of unit loads. Thus, consider the
following equation which represents Eq. (49) for a series of unit loads on the right hand side:
(410)
The solution of the above equation for the matrix of displacements yields the desired FRF matrix
.
e
2
M   – ie B   K   + + ( ) u { } Z   u { } P { } = =
Z  
u { }
u { } Z  
1 –
P { } H   P { } = =
H  
e
H
i j
H { }
j
U   Z  
1 –
I   =
U  
H  
Main Index
MD Nastran Dynamic Analysis User’s Guide
Frequency Response Function (FRF) and FRF Based Assembly (FBA)
144
The matrix of displacements resulting from the solution of Eq. (410), which is equivalent to the
receptance or FRF matrix of Eq. (49), is square because Eq. (410) 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 given by Eq. (410). 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.
Usage of the FRF/FBA Feature
The FRF/FBA capability is available in both SOL 108 (Direct Frequency Response) and SOL 111(Modal
Frequency Response). The generation of FRFs for a compoent and their use in a subsequent FBA process
starts with the FRF  MD Only Case Control command and running a SOL 108 or SOL 111 job.
In addition to the FRF command, the following Bulk Data entries are used to define the components
connectivity and loads. FBADLAY  MD Only, FBALOAD  MD Only, FBAPHAS  MD Only, FRFCONN 
MD Only, FRFFLEX  MD Only, FRFRELS  MD Only, FRFSPC1  MD Only, FRFCOMP  MD Only,
FRFXIT  MD Only and FRFXIT1  MD Only.
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.
U  
H  
U  
Main Index
145
CHAPTER 4
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
MD Nastran Dynamic Analysis User’s Guide
Modal Frequency Response Analysis
146
Modal Frequency Response Analysis
Modal frequency response analysis is an alternate approach to computing the frequency response of a
structure. This method uses the mode shapes of the structure to reduce the size, uncouple the equations
of motion (when modal or no damping is used), and make the numerical solution more efficient. Since
the mode shapes are typically computed as part of the characterization of the structure, modal frequency
response is a natural extension of a normal modes analysis.
As a first step in the formulation, transform the variables from physical coordinates to modal
coordinates by assuming
(411)
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 (411) 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
(412)
at forcing frequency .
Substituting the modal coordinates in Eq. (411) for the physical coordinates in Eq. (412) and dividing
by , the following is obtained:
(413)
Now this is the equation of motion in terms of the modal coordinates. At this point, however, the
equations remain coupled.
To uncouple the equations, premultiply by to obtain
(414)
where:
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 offdiagonal terms that couple the equations of motion. Therefore, in this form
= modal (generalized) mass matrix
= modal (generalized) stiffness matrix
= modal force vector
u e ( ) { }
c e ( ) { }
x { } o   c e ( ) { }e
i et
=
o  
e
2
M   x { } K   x { } + – P e ( ) { } =
e
e
i et
e
2
– M   o   c e ( ) { } K   o   c e ( ) { } + P e ( ) { } =
o
T
 
e
2
o  
T
M   o   c e ( ) { } – o  
T
K   o   c e ( ) { } + o  
T
P e ( ) { } =
o  
T
M   o  
o  
T
K   o  
o  
T
P { }
Main Index
147
CHAPTER 4
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 degreeoffreedom systems as
(415)
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 degreeoffreedom systems.
Once the individual modal responses are computed, physical responses are recovered as the
summation of the modal responses using
(416)
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.
Damping in Modal Frequency Response
If a damping matrix exists, the orthogonality property (see Mathematical Overview of Normal Modes
Analysis, 47) of the modes does not, in general, diagonalize the generalized damping matrix
(417)
If structural damping is used, the orthogonality property does not, in general, diagonalize the generalized
stiffness matrix
(418)
where
In the presence of a 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
(419)
Equation (419) is similar to Eq. (45) for the direct frequency response analysis method except that
Eq. (419) is expressed in terms of modal coordinates . Since the number of modes used in a solution
= ith modal mass
= ith modal stiffness
= ith modal force
e
2
– m
i
c
i
e ( ) k
i
c e ( ) + p
i
e ( ) =
m
i
k
i
p
i
c
i
e ( )
x { } o   c e ( ) { }e
i et
=
B  
o  
T
B   o   diagonal =
o  
T
K   o   diagonal =
K 1 i G + ( ) K   i G
E
K
E
 
¯
+ =
B  
e –
2
o  
T
M   o   ie o  
T
B   o   o  
T
K   o   + +   c e ( ) { } o  
T
P e ( ) { } =
c
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modal Frequency Response Analysis
148
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 where . The equations of motion
remain uncoupled and have the form
(420)
for each mode.
Each of the modal responses is computed using
(421)
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.
b
i
b
i
2m
i
e
i
.
i
=
e
2
– m
i
c
i
e ( ) ieb
i
c
i
e ( ) k
i
c
i
e ( ) + + p
i
e ( ) =
c
i
e ( )
p
i
e ( )
m
i
– e
2
ib
i
e k
i
+ +
 =
Main Index
149
CHAPTER 4
Frequency Response Analysis
At resonance, the three types of damping are related by the following equations:
(422)
Note that the subscript is for the ith mode, and not the ith excitation frequency.
The values of and define pairs of frequencies and dampings. Note that 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. (422). Straightline interpolation is used for
modal frequencies between consecutive 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 41 and modes exist at 1.0, 2.5, 3.6, and 5.5 Hz,
MD Nastran interpolates and extrapolates as shown in Figure 44 and in the table. Note that there is no
table entry at 1.0 Hz; MD Nastran uses the first two table entries at and to extrapolate
the value for .
Figure 44 Example TABDMP1
b
cr
2m
i
e
i
=
.
i
b
i
b
cr

G
i
2
 = =
Q
i
1
2.
i
( )

1
G
i
 = =
i
fi gi gi
fi
f 2.0 = f 3.0 =
f 1.0 =
0.10
0.20
2.0 4.0 6.0
f (Hz)
0.0
= entered value
= computed value
+
+
+ +
.
+
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modal Frequency Response Analysis
150
Modal damping is processed as a complex stiffness when PARAM,KDAMP, KDAMPFL is entered as 1.
The uncoupled equation of motion becomes
(423)
The default for PARAM,KDAMP is 1, which processes modal damping as a damping matrix as shown
in Eq. (420).
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.
Select Modes to be Included for Response Analysis
It is possible that not all of the computed modes are required in the frequency response solution. You
need to retain, at a minimum, all the modes whose resonant frequencies lie within the range of forcing
frequencies. For example, if the frequency response analysis must be between 200 and 2000 Hz, all
modes whose resonant frequencies are in this range should be retained. This guideline is only a minimum
requirement, however. For better accuracy, all modes up to at least two to three times the highest forcing
frequency should be retained. In the example where a structure is excited to between 200 and 2000 Hz,
all modes from 0 to at least 4000 Hz should be retained.
Modal dynamic analysis involves the use of mode shapes in the dynamic analysis. Normally, the analysis
use all of the computed mode shapes. However, the user has the option of selecting a subset of the
computed mode shapes for use in the dynamic analysis. The frequency range selected on the eigenvalue
entry (EIGRL or EIGR) is one means to control the modes used in the modal frequency response solution.
Also, three parameters are available to limit the number of modes included in the solution.
Table 41 Example TABDMP1 Interpolation/Extrapolation
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
e
2
– m
i
c
i
e ( ) 1 iG e ( ) + ( )k
i
c
i
e ( ) + p
i
e ( ) =
Main Index
151
CHAPTER 4
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.
Dynamic Data Recovery in Modal Frequency Response
Analysis
In modal frequency response analysis, two options are available for recovering displacements and
stresses: the mode displacement method and the matrix method. Both methods give the same answers,
although with differences in cost.
The mode displacement method computes the total physical displacements for each excitation frequency
from the modal displacements, and then computes element stresses from the total physical
displacements. The number of operations is proportional to the number of excitation frequencies.
The matrix method computes displacements per mode and element stresses per mode, and then computes
physical displacements and element stresses as the summation of modal displacements and element
stresses. Costly operations are proportional to the number of modes.
Since the number of modes is usually much less that the number of excitation frequencies, the matrix
method is usually more efficient and is the default. The mode displacement method can be selected by
using PARAM,DDRMM,1 in the Bulk Data. The mode displacement method is required when
“frequencyfrozen” structural plots are requested (see Results Processing (Ch. 15)).
The mode acceleration method (Mode Acceleration Method (Ch. 12)) is another data recovery method
for modal frequency response analysis. This method can provide better 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.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modal Versus Direct Frequency Response
152
Modal Versus Direct Frequency Response
Some general guidelines can be used when selecting modal frequency response analysis versus direct
frequency response analysis. These guidelines are summarized in Table 42.
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 highfrequency excitation.
To capture high frequency response in a modal solution, less accurate, highfrequency 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 42 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.
Table 42 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
Main Index
153
CHAPTER 4
Frequency Response Analysis
FrequencyDependent Structural Elements
Vibration control devices, such as engine and component vibration isolation mounts, can have
impedance (stiffness and damping) values that vary with excitation frequency. You can also have models
of the frequencydependent mounts included in a system frequency response analysis. The CBUSH,
CELAS1, CELAS3, CDAMP1, and CDAMP3 elements provide impedance values that change with
frequency to allow modeling such devices for frequency response analysis. The frequencydependent
impedance values are input on table entries. The frequencydependent features may be used in SOLs 108,
111, and 200. These elements may also be used with frequency dependence ignored in any solution
sequence.
The BUSH element is also a useful modeling tool when used with only constant properties. It allows
specification of stiffness and damping values like the present CELASi and CDAMPi elements, but it also
meets all criterion for avoiding internal constraints while having an extensive capability for modeling
both geometry and coordinate systems. When the CELASi and CDAMPi elements are used for this
purpose it is the user's responsibility to align geometry and coordinate systems to avoid internal
constraints. The BUSH element automatically performs this alignment. In this sense, the BUSH element
may be considered as an extended scalar element with the advantage of avoiding the internal constraints
that are possible with the older scalar elements. See the The BUSH Element (Ch. 4) in the MD Nastran
Linear Static Analysis User’s Guide for a good description of the CBUSH element and a comparison to
the CELAS elements.
Input
Use CBUSH, CELAS1, CELAS3, CDAMP1, and CDAMP3 entries to describe element connectivity
and geometry. CELAS2, CELAS4, CDAMP2, and CDAMP4 cannot be used for frequencydependent
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 realmode analysis and for cases where no tabular input is used to define frequencydependent
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
MD Nastran Dynamic Analysis User’s Guide
FrequencyDependent Structural Elements
154
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 elementoriented force output in static analysis. If present,
the frequencydependent 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 frequencydependent 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 frequencydependent 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 rigidbody modes have
nonzero natural frequencies or in cases where the mode shapes for the rigidbody 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 46 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 42.
Main Index
155
CHAPTER 4
Frequency Response Analysis
Figure 45 Model Diagram
Listing 41 Model Input File
SOL 108
CEND
TITLE = VERIFICATION PROBLEM, FREQ. DEP. IMPEDANCE BUSHVER
SUBTITLE = SINGLE DOF, CRITICAL DAMPING, 3 EXCITATION FREQUENCIES
ECHO=BOTH
SPC = 1002
METHOD = 1
DLOAD = 1
DISP = ALL
FREQ = 10
ELFO = ALL
BEGIN BULK
$ CONVENTIONAL INPUT FOR MOUNT
GRDSET, , , , , , , 23456 $ PS
$ TIE DOWN EVERYTHING EXCEPT THE 1 DOF
GRID, 11, , 0., 0., 0.0 $ GROUND
GRID 12, , 0., 0., 0.0 $ ISOLATED DOF
SPC1, 1002, 123456, 11 $ GROUND
CONM2, 12, 12, , 1.0 $ THE ISOLATED MASS
CELAS1, 1112, 1112, 11, 1, 12, 1
PELAS, 1112, 1.0
$PELAST, PID TKID TGEID
PELAST, 1112, 1112
CDAMP1, 11120, 11120, 11, 1, 12, 1
PDAMP, 11120, 0.0 $ NOMINAL VALUE REQUIRED
$PDAMPT P1D1 TBID1
PDAMPT, 11120, 11120
TABLED1, 1112 $ STIFFNESS TABLE
, 0.9 0.81, 1.0, 1.0, 1.1, 1.21 ENDT
TABLED1, 11120 $ DAMPING TABLE
, 0.9 .2864789, 1.0, .3183099, 1.1, .3501409 ENDT
Main Index
MD Nastran Dynamic Analysis User’s Guide
FrequencyDependent Structural Elements
156
$ CONVENTIONAL INPUT FOR FREQUENCY RESPONSE
PARAM, WTMASS, .0253303 $ 1/(2*PI)**2. GIVES FN=1.0
DAREA, 1, 12, 1, 2. $ CAUSES UNIT DEFLECTION
$FREQ, 10, 0.9, 1.0, 1.1 $ BRACKET THE NATURAL FREQENCY
$
$ Perturb the input frequency values so that they do not lie
$ exactly on the natural frequencies.
$
FREQ, 10, 0.92, 1.02, 1.12 $ BRACKET THE NATURAL FREQENCY
RLOAD1, 1, 1, , , 3
TABLED1,3 $ TABLE FOR FORCE VS. FREQUENCY
, 0.9, 0.81, 1., 1., 1.1, 1.21,ENDT $ P = K
ENDDATA
The model is excited at frequencies of 0.9, 1.0, and 1.1 Hz. The impedance is adjusted by TABLED1
entries referenced by a PELAST entry to provide undamped natural frequencies at the excitation
frequencies, and critical damping at all three frequencies. The RLOAD1 and TABLED1 entries describe
a loading function that is equal to the stiffness at each excitation frequency. The DAREA entry requests
that the load defined on the remaining entries be multiplied by 2.0. As the ratio of the dynamic response
at resonance to the static deflection is  0. 5i for critically damped systems, the expected dynamic
response for this loading condition is 1.0i. The actual output values match these results.
EngineBody Coupling
A series of example problems are based on the planar model shown in Figure 46.
.
Figure 46 EngineBody Demonstration Model
25 35 45
Example Problem for Bushing Element Frequency Dependent Impedance
44
24
34
43
Engine Model
Bush Element
Body
Model
Bush Element
RBAR Element
23
33
15
13
55
62
52
42 32 22 12 2
1 11 21 31 41 51
61
Main Index
157
CHAPTER 4
Frequency Response Analysis
These example problems are stored in Install_dir/mdxxxx/doc/dynamics/bushd108.dat and bushd111.dat.
A carbody superelement is attached to an engine superelement at two mount points. The mounts connect
from the points at the end of the triangular brackets attached to the engine to points at the top of the body
model. In frequency response analysis, the model is excited by a rotating force vector acting at the center
of the engine representing forces due to crankshaft unbalance. The mounts are designed to provide a low
natural frequency for the engine translation modes and a higher frequency for its torsion mode. This
variance is accomplished by providing different stiffness and damping in different directions for each
mount. A coordinate system is provided for each mount whose z axis aligns with the element's axis, and
the mounts are aligned on a line from their attachment point to the engine's center of gravity. Each mount
type has impedance in its local 1 and 3 directions. The righthand mount is connected directly to the two
components. The elastomer center of elasticity is at the center of the element. Listing 42 shows the input
entries for the elastomer center of elasticity.
Listing 42 Input Entries for the Elastomer Center of Elasticity
$ RH ENGINE MOUNT
grid,55,,4500.,0.,1500.,,0,0
$ POINT 52 IS ATTACHED TO LOWER STRUCTURE.
$ POINT 55 IS ATTACHED TO UPPER STRUCTURE.
$ AIM Z NEAR ENGINE CG PARALLEL TO BUSH
CORD2R,52,0,5000.,0.,1000.,4000.,0.,2000.
,4000.,0.,1000.
cbush,5001,1001,52,55,,,,52
$ default value of halfspan used for elastomer center
The default value of “S”, which is the location of the elastomer center of elasticity, is used. “S” is the
ratio of the distance between the connected points to the location of the center of elasticity of the mount.
The default value of 0.5 places the elasticity center at the midpoint between the connected points.
Listing 43 shows the element property values for the element.
Listing 43 Element Property Values
$ LH ENGINE MOUNT
$ grid point 12 is the attachment point of the lower mount to structure.
$ a rbar replaces one end of the bush element modeling the mount.
$ The other end of the rbar attaches to grid point 13.
$ grid point 13 is at the elastomer center of elasticity
$ grid point 15 is attached to the upper structure.
GRID, 13,,1250.,,1250.,12,0 $ LOWERMOUNT AXIS
CORD2R, 12,0,1000.,0.,1000.,2000.,0.,2000.
,2000.,0.,1000. $ AIM Z NEAR ENGINE CG
cbush,1001,1001,13,15,,,,12
$ move the center of elasticity to grid point 13 end.
RBAR,12,12,13,123456 $
plotel,12012,13 $ show where rbars are on plot
$ the rbar is instrumentation to measure force in bush element. It
$ is the only element attached to the lower end of the mount
The only elements connecting to the center point are the RBAR and the CBUSH element. Both elements
are massless, and there are no external forces applied to this point. The only sources of force acting on
the point therefore result from the RBAR and CBUSH elements. The forces in the CBUSH element are
then equal and opposite to the forces in the RBAR element at the point in common between them.
Main Index
MD Nastran Dynamic Analysis User’s Guide
FrequencyDependent Structural Elements
158
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 frequencydependent 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 frequencydependent 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 frequencydependent
effects.
In direct analysis, there is no DDRMM method of data recovery and the addition of the incremental
matrices for frequencydependent 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 constraintforce
output. GE damping is considerably more expensive than viscous damping. The approximations of the
modal method are more pronounced for a large number of frequencydependent elements, while
computation costs approach direct analysis costs. These trends indicate that a direct analysis offers
Main Index
159
CHAPTER 4
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 frequencydependent elements is small.
Main Index
MD Nastran Dynamic Analysis User’s Guide
FrequencyDependent Excitation Definition
160
FrequencyDependent Excitation Definition
An important aspect of a frequency response analysis is the definition of the loading function. In a
frequency response analysis, the force must be defined as a function of frequency. Forces are defined in
the same manner regardless of whether the direct or modal method is used.
The following Bulk Data entries are used for the frequencydependent load 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 frequencydependent excitation. The
description is given in terms of the coefficients that define the dynamic load.
FrequencyDependent Loads – RLOAD1 Entry
The RLOAD1 Bulk Data entry is a general form in which to define a frequencydependent load. It defines
a dynamic loading of the form
(424)
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.
RLOAD1 Tabular inputreal and imaginary
RLOAD2 Tabular inputmagnitude and phase
DAREA Spatial distribution of dynamic load
DLOAD Combines dynamic load sets
TABLEDi Tabular values versus frequency
DELAY Time delay
DPHASE Phase lead
P f ( ) { } A C f ( ) iD f ( ) +  e
i u 2tf t – { }
{ } =
Main Index
161
CHAPTER 4
Frequency Response Analysis
Note that f is the frequency in cycles per unit time and that .
FrequencyDependent Loads – RLOAD2 Entry
The RLOAD2 Bulk Data entry is a variation of the RLOAD1 entry used for defining a
frequencydependent load. Whereas the RLOAD1 entry defines the real and imaginary parts of the
complex load, the RLOAD2 entry defines the magnitude and phase.
The RLOAD2 entry defines dynamic excitation in the form
(425)
The RLOAD2 definition may be related to the RLOAD1 definition by
1 2 3 4 5 6 7 8 9 10
RLOAD1 SID
EXCITEID DELAYI/
DELAYR
DPHASEI/
DPHASER
TC TD TYPE
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 . 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 degreesoffreedom 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 degreesoffreedom that
are excited by this dynamic load entry. See Remark 2. (Real or blank)
TC Set identification number of the TABLEDi entry that gives . See Remark 2.
(Integer > 0)
TD Set identification number of the TABLEDi entry that gives . See Remark 2.
(Integer > 0)
TYPE Defines the type of the dynamic excitation. See Remarks 5. and 6. (Integer,
character or blank; Default = 0)
A { }
t
t
u
u
C f ( )
D f ( )
e
i wt
wt cot i wt sin + ( ) =
P f ( ) { } AB f ( )e
i o f ( ) u 2tf t – + { }
{ } =
Main Index
MD Nastran Dynamic Analysis User’s Guide
FrequencyDependent Excitation Definition
162
(426)
Note that is the frequency in cycles per unit time.
Spatial Distribution of Loading  DAREA Entry
The DAREA Bulk Data entry defines the degreesoffreedom where the dynamic load is to be applied and
the scale factor to be applied to the loading. The DAREA entry provides the basic spatial distribution of
the dynamic loading.
1 2 3 4 5 6 7 8 9 10
RLOAD2 SID
EXCITEID DELAYI/
DELAYR
DPHASEI/
DPHASER
TB TP TYPE
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 . 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 degreesoffreedom 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 degreesoffreedom that
are excited by this dynamic load entry. See Remark 2. (Real or blank)
TB Set identification number of the TABLEDi entry that gives . (Integer > 0)
TP Set identification number of the TABLEDi entry that gives 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
RLOAD1
Definition
RLOAD2
Definition
C f ( ) iD f ( ) + B f ( )e
i o f ( )
=
A { }
t
t
u
u
B f ( )
o f ( )
f
Main Index
163
CHAPTER 4
Frequency Response Analysis
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.
Time Delay – DELAY Entry
The DELAY Bulk Data entry defines the time delay in an applied load.
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.
Phase Lead – DPHASE Entry
The DPHASE Bulk Data entry defines the phase lead .
Field Contents
SID Set ID specified by RLOADi entires.
Pi Grid, extra, or scalar point ID.
Ci Component number.
Ai Scale factor.
1 2 3 4 5 6 7 8 9 10
DELAY
SID P1 C1 P2 C2
Field Contents
SID Set ID specified by an RLOADi entry.
Pi Grid, extra, or scalar point ID.
Ci Component number.
Time delay for Pi, Ci. (Default = 0.0)
1 2 3 4 5 6 7 8 9 10
DPHASE
SID P1 C1 P2 C2
Field Contents
SID Set ID specified by an RLOADi entry.
Pi Grid, extra, or scalar point ID.
t
t
2
t
i
u
u
1
u
2
Main Index
MD Nastran Dynamic Analysis User’s Guide
FrequencyDependent Excitation Definition
164
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.
Dynamic Load Tabular Function  TABLEDi Entries
The TABLEDi entries (i = 1 through 4) each define a tabular function for use in generating
frequencydependent dynamic loads. The form of each TABLEDi entry varies slightly, depending on
the value of i, as does the algorithm for y(x). The x values need not be evenly spaced.
The TABLED1, TABLED2, and TABLED3 entries linearly interpolate between the end points and
linearly extrapolate outside of the endpoints, as shown in Figure 47. The TABLED1 entry also performs
logrithmic interpolation between points. The TABLED4 entry assigns the endpoint values to any value
beyond the endpoints.
Figure 47 Interpolation and Extrapolation for TABLED1, TABLED2, and TABLED3 Entries
The TABLED1 Bulk Data entry has the following format:
Ci Component number.
Phase lead (in degrees) for Pi, Ci. (Default = 0.0)
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 xaxis. (Character: “LINEAR” or
“LOG”; default = “LINEAR”)
Field Contents
u
i
y
x
Linear Interpolation Between Endpoints
x
1
x
2
x
3
x
4
x
5
x
6
x
7
Linear Extrapolation of Segment x
6
x
7
–
Linear Extrapolation of Segment x
2
x
1
–
Main Index
165
CHAPTER 4
Frequency Response Analysis
The TABLED1 entry uses the algorithm
(427)
The algorithms used for interpolation and extrapolation are as follows:
The TABLED2 Bulk Data entry has the following format:
YAXIS Specifies a linear or logarithmic interpolation for the yaxis. (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.
XAXIS YAXIS
LINEAR LINEAR
LOG LINEAR
LINEAR LOG
LOG LOG
1 2 3 4 5 6 7 8 9 10
TABLED2
TID X1
x1 y1 x2 y2 x3 y3 etc. ENDT
Field Contents
y y
T
x ( ) =
y
T
x ( )
xj x –
xj xi –
 yi
x xi –
xj xi –
 yj +
xj x ( ) ln
xj xi ( ) ln
 yi
x x i ( ) ln
xj xi ( ) ln
yj +
xj x –
xj xi –
 yi
x xi –
xj xi –
 yj ln + ln exp
xj x ( ) ln
xj xi ( ) ln
 y ln i
x x i ( ) ln
xj xi ( ) ln
 yj + exp
Main Index
MD Nastran Dynamic Analysis User’s Guide
FrequencyDependent Excitation Definition
166
The TABLED2 entry uses the algorithm
(428)
ENDT ends the table input.
The TABLED3 Bulk Data entry has the following format:
The TABLED3 entry uses the algorithm
(429)
ENDT ends the table input.
The TABLED4 Bulk Data entry has the following format:
Field Contents
TID Table identification number.
X1 Table parameter.
xi, yi Tabular values. Values of x are frequency in cycles per unit time.
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.
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.
y y
T
x X1 – ( ) =
y y
T
x X1 –
X2

\ .
 
=
Main Index
167
CHAPTER 4
Frequency Response Analysis
The TABLED4 entry uses the algorithm
(430)
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:
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
Xi Table parameters (X2 = 0.0; X3 < X4).
Ai Coefficients.
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
Field Contents
y Ai
x X1 –
X2

\ .
 
i
i 0 =
N
¯
=
Main Index
MD Nastran Dynamic Analysis User’s Guide
FrequencyDependent Excitation Definition
168
form. This table is shown graphically in Figure 48. 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.
Figure 48 TABLED1  Amplitude Versus Frequency
Static Load Sets
Only concentrated forces and moments can be specified directly using DAREA entries. To
accommodate more complicated loadings conveniently, the EXCITEID on the RLOADi entry is used to
define static load entries that define the spatial distribution of dynamic loads. EXCITEID may refer to
one or more static load entries (FORCE, PLOADi, GRAV, etc.). All static loads with the Set ID
referenced on the entry define the spatial distribution of the dynamic loading. MD Nastran converts this
information to equivalent dynamic loading.
The static loads are combined with any DAREA entries in the referenced set. The DAREA Set ID is
referenced by an RLOADi entry. This reference defines the temporal distribution of the dynamic
loading.
Figure 49 demonstrates the relationships of these entries. To activate a load set defined in this manner,
the DLOAD Case Control command refers to the Set ID of the selected DLOAD or RLOADi entry. The
EXCITEID entries point to the static loading entries that are used to define dynamic loadings and/or
DAREA Set ID. Together this relationship defines a complete dynamic loading. To apply dynamic
loadings in this manner, the DLOAD Case Control command and the RLOADi Bulk Data entry must be
defined.
5
2 4 6
Frequency (Hz)
Amplitude
on
TABLED1
10
0
Main Index
169
CHAPTER 4
Frequency Response Analysis
Figure 49 Relationship of Dynamic and Static Load Entries
Dynamic Load Set Combination – DLOAD
One of the requirements of frequencydependent loads is that RLOAD1s and RLOAD2s must have
unique SIDs. If they are to be applied in the same analysis, they must be combined using the DLOAD
Bulk Data entry. The total applied load is constructed from a linear combination of component load sets
as follows:
(431)
where:
The DLOAD Bulk Data entry has the following format:
= overall scale factor
= scale factor for the ith load set
= ith load set
= total applied load
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.
DLOAD
RLOADi
DAREA Static Load Entries Dynamic Load
(Temporal
Distribution)
(Spatial
Distribution)
Case Control
Bulk Data
(ReferenceLink)
P { } S S
i
P
i
{ }
i
¯
=
S
S
i
P
i
P
Main Index
MD Nastran Dynamic Analysis User’s Guide
FrequencyDependent Excitation Definition
170
As an example, in the following DLOAD entry:
\
a dynamic Load Set ID of 33 is created by taking 0.5 times the loads in the Load Set ID of 14, adding to
it 2.0 times the loads in the Load Set ID of 27, and multiplying that sum by an overall scale factor of 3.25.
As with other frequencydependent loads, a dynamic load combination defined by the DLOAD Bulk
Data entry is selected by the DLOAD Case Control command.
$DLOAD SID S S1 L1 S2 L2 etc.
DLOAD 33 3.25 0.5 14 2.0 27
Main Index
171
CHAPTER 4
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.
FREQ Defines descrete excitation frequencies.
FREQ1 Defines a starting frequency , a frequency incitement , and the number of
frequency increments to solve NDF.
FREQ2 Defines a starting frequency , and ending frequency , and the number of
logarithmic intervals, NF, to be used in the frequency range.
FREQ3*
* Used for modal solution only.
Defines the number of excitation frequencies used between modal pairs in a given
range.
FREQ4* Defines excitation frequencies using a spread about each normal mode within a
range.
FREQ5* Defines excitation frequencies as all frequencies in a given range as a defined
fraction of the normal modes.
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
F
start
Af
F
start
F
end
Main Index
MD Nastran Dynamic Analysis User’s Guide
Solution Frequencies
172
The FREQ1 Bulk Data example specifies 14 frequencies between 2.9 Hz and 9.4 Hz in increments of
0.5 Hz.
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.
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.
Field Contents
SID Set ID specified by a FREQUENCY Case Control command.
F Frequency value (cycles per unit time).
$FREQ1
SID NDF
FREQ1
6 2.9 0.5 13
Field Contents
SID Set ID specified by a FREQUENCY Case Control command.
Starting frequency in set (cycles per unit time).
Frequency increment (cycles per unit time).
NDF Number of frequency increments.
1 2 3 4 5 6 7 8 9 10
$FREQ2
SID NF
FREQ2
9 1.0 8.0 6
Field Contents
SID Set ID specified by a FREQUENCY Case Control command.
Starting frequency (cycles per unit time).
Ending frequency (cycles per unit time).
NF Number of logarithmic intervals.
F
start
Af
F
start
Af
F
start
F
end
F
start
F
end
Main Index
173
CHAPTER 4
Frequency Response Analysis
The FREQ4 Bulk Data entry example chooses 21 equally spaced frequencies across a frequency band of
to for each natural frequency between 20 and 2000.
1 2 3 4 5 6 7 8 9 10
$FREQ3
SID F1 F2 TYPE NEF
CLUSTER
FREQ3
6 20.0 2000.0 LINEAR 10 2.0
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)
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.)
0.7 f
N
· 1.3 f
N
·
Main Index
MD Nastran Dynamic Analysis User’s Guide
Solution Frequencies
174
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
175
CHAPTER 4
Frequency Response Analysis
Frequency Response Considerations
Exciting an undamped (or modal or viscous damped) system at 0.0 Hz using direct frequency response
analysis gives the same results as a static analysis and also gives almost the same results when using
modal frequency response (depending on the number of retained modes). Therefore, if the maximum
excitation frequency is much less than the lowest resonant frequency of the system, a static analysis is
probably sufficient.
Undamped or very lightly damped structures exhibit large dynamic responses for excitation frequencies
near resonant frequencies. A small change in the model (or running it on another computer) may result
in large changes in such responses.
Use a fine enough frequency step size ( ) to adequately predict peak response. Use at least five points
across the halfpower bandwidth (which is approximately for an SDOF system) as shown in
Figure 410.
Figure 410 HalfPower Bandwidth
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.
Af
2.f
n
Response
Peak Response
Frequency
HalfPower Bandwidth
= Excitation Frequency
Peak
2
 HalfPower Point =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Solution Control for Frequency Response Analysis
176
Solution Control for Frequency Response Analysis
In the Case Control Section of the MD Nastran input file, you must select the solution parameters
associated with the current analysis (i.e., frequencies, loads, and boundary conditions), and also the
output quantities required from the analysis. The Case Control commands directly related to frequency
response analysis are listed in Table 43. They can be combined in the standard fashion with the more
generic entries, such as SPC, MPC, etc.
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, reallife 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
Table 43 Case Control Commands for Frequency Response Solution Control
Case
Control
Command
Direct
or
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
Data
Optional
OFREQUENCY Both Select the frequencies for output (Default =
all)
Optional
OMODES Both Selects mode for output requests (Default =
all)
Optional
Main Index
177
CHAPTER 4
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 44 and Table 45.
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.
Table 44 Grid Output from a Frequency Response Analysis
Case Control
Command Description
ACCELERATION Grid point acceleration results for a set of grid points.
DISPLACEMENT
(or VECTOR)
Grid point displacement results for a set of grid points.
OLOAD Requests the applied load table to be output for a set of grid points.
SACCELERATION Requests the solution set acceleration output: dset in direct solutions and
modal variables in modal solutions.
SDISPLACEMENT Requests the solution set displacement output: dset in direct solutions
and modal variables in modal solutions.
SVECTOR Requests the real eigenvector output for the aset in modal solutions.
SVELOCITY Requests the solution set velocity output: dset in direct solutions and
modal variables in modal solutions.
SPCFORCES Requests the forces of a singlepoint constraint for a set of grid points.
VELOCITY Grid point velocity results for a set of grid points.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Solution Control for Frequency Response Analysis
178
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 46.
Table 45 Element Output from a Frequency Response Analysis
Case Control Command Description
ELSTRESS (or STRESS) Element stress results for a set of elements.
ELFORCE (or FORCE) Element force results for a set of elements.
STRAIN Element strain results for a set of elements.
Table 46 Bulk Data Entries for Frequency Response Analysis
Bulk Data
Entry
Direct
or Modal Description
Required/
Optional
FREQ, FREQi Both Excitation frequencies Required
RLOADi Both Dynamic loading Required
EIGR or EIGRL Modal Eigenvalue analysis parameters Required
LSEQ Both Dynamic loading from static loads Optional
TABLEDi Both Frequencydependent tables Both*
*Required for RLOAD1; optional for RLOAD2.
DAREA Both Load component and scale factor Required
DELAY Both Time delay on dynamic load Optional
DPHASE Both Phase angle on dynamic load Optional
DLOAD Both Dynamic load combination, required if RLOAD1
and RLOAD2 are used
Optional
TABDMP1 Modal Modal damping table Optional
Main Index
179
CHAPTER 4
Frequency Response Analysis
Modal Contribution Fraction (MCFRACTION)
Introduction
The response of large structural components to loads in the frequency or time domain can be determined
using either a direct or modal solution approach. The modal approach is typically employed for large
models when the response of the structure can be approximated using only a small number of its natural
modes. During this process, data recovery and postprocessing of large quantities of response data usually
occurs. Often, the user desires detailed information for several points of interest in the model. One item
of particular interest is often referred to as the modal contribution fraction or the modal participation
factor As these names imply, the value for this item is defined as that portion of the total response at the
point that can be attributed to a particular natural mode.
The MFRACTION command provides more control over the amount of output produced during the
modal contribution fraction calculation.
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, , at a degree of freedom can be found by summing the individual modal
responses.
(432)
In Eq. (432), is a real one row by h column physical response data recovery matrix and
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, , is a complex
U
U rphi   uh { } =
rphi   uh { }
U
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modal Contribution Fraction (MCFRACTION)
180
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
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 411.
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:
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 XYPLOT 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 47.
rphi uh · ( )
MCFRACTION
STRUCTURE
(
PRINT, PUNCH
PLOT
REAL or IMAG
PHASE
SORT sorttype = ( )
. . .
KEY
FRACTION
sortitem
= ITEMS
FRACTION
ALL
(itemlist)
= SOLUTION
ALL
m
NONE
= FILTER
0.001
fratio
= .
NULL
12
ipowr
=
ALL
n
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
Main Index
181
CHAPTER 4
Frequency Response Analysis
Figure 411 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.
Figure 411 Sample Modal Frequency Response Solution Output
The headings in Figure 411 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 412.
Figure 412 Sample Modal Transient Response Solution Output
Table 47 MCFRACTION Item List Descriptions
Item Identifier Description
RESPONSE each mode’s response at each degree of freedom selected
PROJECTION projection of modal response on solution
FRACTION fraction of total displacement per mode (PROJECTION divided by total)
SCALED scaled magnitudes (PROJECTION divided by largest term in FRACTION)
MODEDISP modal displacements (complex solution at each dof by mode number)
MODERESP modal response for each mode (polar format wrt total displacement)
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) = 4.73302E01 / 5.20364E03, (M/P) = 4.73330E01 / 0.63
LOAD FREQUENCY = 2.00000E+01, (SUBCASE 1, DLOAD = 15)
MAXIMUM MODAL RESP = 4.67546E01 FOR MODE ID = 2, SORTKEY = FRACTION, SORT = ABS VALUE DESCENDING, FILTER = 1.00000E02
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 4.67512E01 5.59608E03 4.67546E01 0.69 4.67546E01 0.06 9.87778E01 9.99999E01
3 3.17429E+01 2.36155E02 4.93491E04 2.36207E02 358.80 2.36087E02 358.17 4.98778E02 5.04949E02
6 7.63429E+01 1.77298E02 9.97411E05 1.77301E02 179.68 1.77276E02 179.05 3.74530E02 3.79164E02
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 = 3.73650E08, MAGNITUDE = 3.73650E08
TIME STEP = 1.00000E02, (SUBCASE 1, DLOAD = 15)
MAXIMUM MODAL RESP = 3.30842E08 FOR MODE ID = 2, SORTKEY = SCALED , SORT = ALGEBRAIC ASCENDING, FILTER = 1.00000E03
MODE ID NATURAL FREQ. (HZ) MODAL RESPONSE MODAL FRACTION SCALED MAGNITUDE
2 9.35245E+00 3.30842E08 8.85432E01 1.00000E+00
6 7.63429E+01 6.92352E09 1.85294E01 2.09270E01
9 2.54595E+02 5.27676E10 1.41222E02 1.59495E02
13 5.30493E+02 4.79519E10 1.28334E02 1.44939E02
3 3.17429E+01 2.69084E09 7.20151E02 8.13333E02
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modal Contribution Fraction (MCFRACTION)
182
Guidelines and Limitations
1. The MCFRACTION command is available in modal frequency response (SOL 111), modal
transient response (SOL 112) and modal complex eigenvalue analysis (SOL 110) solution
sequences only. If superelements are used, its use is restricted to residual structure data recovery
operations only.
2. The modal contribution fraction computation is intended for a small subset of the degrees of
freedom present in the model. The calculations are limited by the amount of memory available
to store the physical response recovery matrix. Its columns represent a response for each mode
and its rows each represent one of the grid point components for which a physical solution is
requested. This matrix must fit in the available memory.
3. Only the STRUCTURE option is currently supported.
4. Both PRINT and PUNCH may be requested.
5. Printed output includes results for all of the data items described in Table 47.
6. Punched output includes results for only the data items selected by the ITEMS keyword.
7. Modal Contribution Fractions are sorted by increasing order of mode number unless the SORT
keyword specifies a particular sorting order. If a sorting order is specified, the KEY keyword
selects the particular data item in the printed results tabular output listing that is sorted.
8. The SOLUTION keyword can be used to select a subset of the available solutions. If SET m is
specified, the items in the SET list are forcing frequency values, time step values, or complex
eigenvalue mode numbers depending upon the solution sequence used.
9. The FILTER keyword specifies a filter ratio value that is used to limit the amount of printed
output produced. It applies to the data item selected by the KEY keyword if it is specified. If no
KEY keyword is present, the default value of KEY=FRACTION will be used. The maximum
value for the selected data item across all natural modes is determined. If the ratio of the data item
value to the maximum data item value is less than fratio for any natural mode, no output for that
natural mode is produced.
10. The NULL keyword can be used to establish the null response threshold value. If the magnitude
of the total response at a selected grid point component is less than , no modal
contribution fraction output is generated for that degree of freedom.
11. For modal transient response solution sequences, response quantities are real numbers. There are
no imaginary terms. Therefore, polar representations of the data have no meaning. Furthermore,
projections of responses onto the total response are simply the individual modal contribution to
the total response at a degree of freedom. Thus, the only items available for output are the
individual modal response magnitude (PROJECTION), the modal fraction (FRACTION) and the
scaled response magnitude (SCALED). Selection of any of the other items from the item list of
Table 47 causes selection of the modal response magnitude (PROJECTION) item.
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
1.0 10
ipowr –
×
Main Index
183
CHAPTER 4
Frequency Response Analysis
solution; mcfr110.dat (Install_dir/mdxxxx/doc/dynamics/mcfr110.dat) demonstrates usage in a modal
complex eigenvalue response solution; and mcfr112.dat
(Install_dir/mdxxxx/doc/dynamics/mcfr112.dat) demonstrates usage in a modal transient response
solution.
In all of these examples, the MCFRACTION Case Control command is introduced to request calculation
and output of the modal contribution fraction data. Each example problem uses different options to sort
and filter the output.
The model used for these examples is the doubleheaded flyswatter model consisting of 67 CQUAD4
shell elements and 104 grid points cantilevered at the base.
PFMODE  Modal Participation Factor Output
PFMODE and MCFRACTION produce very similar output. PFMODE includes fluid and acoustic
response as will the structural mode participation factors. See Acoustic Modal Participation Factors
(Ch. 11). PFMODE calculations can be more efficient and will appear after standard output requests such
as DISP.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
184
Examples
This section provides several examples showing the input and output. These examples are:
These examples are described in the sections that follow.
TwoDOF Model
Consider the twoDOF system shown in Figure 413. 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 44 shows the input file. (See Install_dir/mdxxxx/doc/dynamics/bd04two.dat)
Figure 413 TwoDOF Model
Model
Frequency Response Bulk Data
Entries Output
bd04two EIGRL, FREQ1, TABDMP1, RLOAD1,
DAREA, TABLED1
XY plots (linear), printed results
SORT2
bd04twos1 EIGRL, FREQ1, TABDMP1, RLOAD1,
DAREA, TABLED1
Printed results SORT1
bd04twona EIGRL, FREQ1, TABDMP1, RLOAD1,
DAREA, TABLED1
XY plots (linear)
bd04bar EIGRL, FREQ1, TABDMP1, DLOAD,
RLOAD2, DAREA, DPHASE, TABLED1
XY plots (log)
bd04bkt EIGRL, FREQ1, TABDMP1, RLOAD1,
LSEQ, TABLED1, PLOAD4
XY plot (log)
y
x
Auxiliary Structure, Grid
Point 1
Primary Structure, Grid
Point 2
P = 20 N sin et
Main Index
185
CHAPTER 4
Frequency Response Analysis
Listing 44 Input File (Abridged) for the TwoDOF Example
$ FILE bd04two.dat
$
$ TWODOF SYSTEM
$ CHAPTER 4, FREQUENCY RESPONSE
$
SOL 111 $ MODAL FREQUENCY RESPONSE
CEND
TITLE = TWODOF SYSTEM
SUBTITLE = MODAL FREQUENCY RESPONSE
LABEL = 20 N FORCE APPLIED TO PRIMARY MASS
$
$ SPECIFY SPC
SPC = 996
$
$ SPECIFY MODAL EXTRACTION
METHOD = 10
$
$ SPECIFY DYNAMIC INPUT
DLOAD = 999
FREQ = 888
SDAMPING = 777
$
$ SELECT OUTPUT
SET 10 = 2.0, 2.05, 2.1, 2.15, 2.2, 2.25
SET 11 = 1,2
OFREQ = ALL
$
SUBCASE 1
DISPLACEMENT(PHASE,PLOT) = 11
SUBCASE 2
OFREQ=10
DISPLACEMENT(SORT2,REAL,PRINT,PLOT) = 11
SUBCASE 3
OFREQ=10
DISPLACEMENT(SORT2,PHASE,PRINT,PLOT) = 11
$
$ XYPLOTS
OUTPUT(XYPLOT)
XTGRID = YES
YTGRID = YES
XBGRID = YES
YBGRID = YES
XTITLE = FREQUENCY
YTTITLE = DISP MAGNITUDE GRID 1
YBTITLE = PHASE GRID 1
XYPLOT DISP 1 /1(T2RM,T2IP)
YTTITLE = DISP MAGNITUDE GRID 2
YBTITLE = PHASE GRID 2
XYPLOT DISP 1 /2(T2RM,T2IP)
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10..
...$
$
$ ENTRIES FOR FREQUENCY RESPONSE
$
$ LOAD DEFINITION
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
186
$RLOAD1 SID DAREA TC
RLOAD1 999 997 901
$DAREA SID P1 C1 A1
DAREA 997 2 2 20.0
$TABLED1 TID +TABL1
$+TABL1 X1 Y1 X2 Y3 ETC.
TABLED1 901 +TAB901
+TAB901 0.0 1.0 10.0 1.0 ENDT
$
$ ALTERNATE LOAD DEFINITION USING DLOAD
$DLOAD SID S S1 RLOAD1
$DLOAD 999 1.0 1.0 998
$RLOAD1 SID DAREA TC
$RLOAD1 998 997 901
$
$ FREQUENCY RANGE 210 HZ
$FREQ1 SID F1 DF NDF
FREQ1 888 2. 0.05 160
$
$ MODAL DAMPING OF 5% CRITICAL
$TABDMP1 TID TYPE +TABD1
$+TABD1 F1 G1 F2 G2 ETC.
TABDMP1 777 CRIT +TABD7
+TABD7 0. 0.05 100. 0.05 ENDT
$
$ MODAL EXTRACTION
$EIGRL SID V1 V2 ND MSGLVL
EIGRL 10 0.1 20. 0
$
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 996 3 2
$
ENDDATA
Main Index
187
CHAPTER 4
Frequency Response Analysis
Figure 414 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.
The RLOAD1 entry describes a sinusoidal load in the form
(433)
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 46, Listing 47, and Listing 47 showing a portion
of their printed output.
Figure 414 Relationship Between the Case Control Commands and Bulk Data Entries for
the TwoDOF Model
Case Control Bulk Data
METHOD EIGRL
FREQUENCY FREQ1
SDAMPING TABDMP1
DLOAD RLOAD1
DAREA
TABLED1
A = 20.0 (entered on the DAREA entry)
C = 1.0 for all frequencies entered on the TABLED1 entry
D = 0.0 (field 7 of the RLOAD1 entry is blank)
= 0.0 (field 5 of the RLOAD1 entry is blank)
= 0.0 (field 4 of the RLOAD1 entry is blank)
P f ( ) { } A C f ( ) iD f ( ) +  e
i u 2tf t – { }
{ } =
u
t
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
188
Listing 45 Real/Imaginary Output in SORT2 Format
Listing 46 Magnitude/Phase Output in SORT2 Format
POINTID = 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 2.813048E03 0.0 0.0 0.0 0.0
0.0 2.107981E04 0.0 0.0 0.0 0.0
2.050000E+00 G 0.0 2.866637E03 0.0 0.0 0.0 0.0
0.0 2.229159E04 0.0 0.0 0.0 0.0
2.100000E+00 G 0.0 2.923136E03 0.0 0.0 0.0 0.0
0.0 2.358377E04 0.0 0.0 0.0 0.0
2.150000E+00 G 0.0 2.982726E03 0.0 0.0 0.0 0.0
0.0 2.496357E04 0.0 0.0 0.0 0.0
2.200000E+00 G 0.0 3.045604E03 0.0 0.0 0.0 0.0
0.0 2.643902E04 0.0 0.0 0.0 0.0
2.250000E+00 G 0.0 3.111983E03 0.0 0.0 0.0 0.0
0.0 2.801898E04 0.0 0.0 0.0 0.0
POINTID = 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 2.374953E03 0.0 0.0 0.0 0.0
0.0 1.129933E04 0.0 0.0 0.0 0.0
2.050000E+00 G 0.0 2.397706E03 0.0 0.0 0.0 0.0
0.0 1.180853E04 0.0 0.0 0.0 0.0
2.100000E+00 G 0.0 2.421475E03 0.0 0.0 0.0 0.0
0.0 1.234172E04 0.0 0.0 0.0 0.0
2.150000E+00 G 0.0 2.446310E03 0.0 0.0 0.0 0.0
0.0 1.290071E04 0.0 0.0 0.0 0.0
2.200000E+00 G 0.0 2.472262E03 0.0 0.0 0.0 0.0
0.0 1.348744E04 0.0 0.0 0.0 0.0
2.250000E+00 G 0.0 2.499386E03 0.0 0.0 0.0 0.0
0.0 1.410402E04 0.0 0.0 0.0 0.0
POINTID = 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 2.820935E03 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 2.875291E03 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 2.932634E03 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 2.993155E03 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 3.057058E03 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 3.124571E03 0.0 0.0 0.0 0.0
0.0 354.8552 0.0 0.0 0.0 0.0
POINTID = 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 2.377640E03 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 2.400612E03 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 2.424618E03 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 2.449710E03 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 2.475939E03 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 2.503362E03 0.0 0.0 0.0 0.0
0.0 356.7702 0.0 0.0 0.0 0.0
Main Index
189
CHAPTER 4
Frequency Response Analysis
Listing 47 Real/Imaginary and Magnitude/Phase Output in SORT1 Format
Figure 415 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 416. See
Install_dir/mdxxxx/doc/dynamics/bd04twona.dat.
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)
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 2.813048E03 0.0 0.0 0.0 0.0
0.0 2.107981E04 0.0 0.0 0.0 0.0
2 G 0.0 2.374953E03 0.0 0.0 0.0 0.0
0.0 1.129933E04 0.0 0.0 0.0 0.0
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)
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 2.866637E03 0.0 0.0 0.0 0.0
0.0 2.229159E04 0.0 0.0 0.0 0.0
2 G 0.0 2.397706E03 0.0 0.0 0.0 0.0
0.0 1.180853E04 0.0 0.0 0.0 0.0
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)
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 2.820935E03 0.0 0.0 0.0 0.0
0.0 355.7145 0.0 0.0 0.0 0.0
2 G 0.0 2.377640E03 0.0 0.0 0.0 0.0
0.0 357.2761 0.0 0.0 0.0 0.0
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)
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1 G 0.0 2.875291E03 0.0 0.0 0.0 0.0
0.0 355.5535 0.0 0.0 0.0 0.0
2 G 0.0 2.400612E03 0.0 0.0 0.0 0.0
0.0 357.1805 0.0 0.0 0.0 0.0
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
190
Figure 415 Displacement Response Magnitudes With the Auxiliary Structure
Figure 416 Displacement Response Magnitude Without the Auxiliary Structure
0.12
0
0.014
0
2 10 Frequency (Hz)
D
i
s
p
l
a
c
e
m
e
n
t
1
(
m
)
D
i
s
p
l
a
c
e
m
e
n
t
2
(
m
)
0.02
0
2 10 Frequency (Hz)
D
i
s
p
l
a
c
e
m
e
n
t
2
(
m
)
Main Index
191
CHAPTER 4
Frequency Response Analysis
Cantilever Beam Model
Consider the cantilever beam shown in Figure 417. This model is a planar model of the cantilever beam
introduced in Real Eigenvalue Analysis, 43 with unrestrained DOFs in the T2 and R3 directions. Two
loads are applied: one at grid point 6 and the other at grid point 11. The loads have the frequency
variation shown in Figure 418. The loads in the figure are indicated with a heavy line in order to
emphasize their values. The load at grid point 6 has a 45degree phase lead, and the load at grid point
11 is scaled to be twice that of the load at grid point 6. Modal frequency response is run across a
frequency range of 0 to 20 Hz. Modal damping is used with 2% critical damping between 0 and 10 Hz
and 5% critical damping above 10 Hz. Modes to 500 Hz are computed using the Lanczos method.
Figure 417 Cantilever Beam Model with Applied Loads
y
x
6
p(f)
11
2p(f)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
192
Figure 418 Applied Loads
0 20 Frequency (Hz)
100
8.0
100
4.0
0
0
0
0
3.0
45
6.0
L
o
a
d
M
a
g
n
i
t
u
d
e
6
(
N
)
L
o
a
d
P
h
a
s
e
6
(
d
e
g
)
L
o
a
d
M
a
g
n
i
t
u
d
e
1
1
(
N
)
L
o
a
d
P
h
a
s
e
1
1
(
d
e
g
)
Main Index
193
CHAPTER 4
Frequency Response Analysis
Listing 48 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.
Listing 48 Input File (Abridged) for the Beam Example
$ FILE bd04bar.dat
$
$ CANTILEVER BEAM MODEL
$ CHAPTER 4, FREQUENCY RESPONSE
$
SOL 111 $ MODAL FREQUENCY RESPONSE
TIME 10
CEND
TITLE = CANTILEVER BEAM
SUBTITLE = MODAL FREQUENCY RESPONSE
$
SPC = 21
$
DLOAD = 22
FREQ = 27
SDAMPING = 20
$
METHOD = 10
$
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
$
$ XYPLOTS
$
... XY plot commands ...
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10.$
$
EIGRL 10 0.1 500. 0
$
FREQ1 27 0.0 0.05 400
$
TABDMP1 20 CRIT +TABD1
+TABD1 0.0 0.02 10.0 0.02 10.01 0.05 25.0 0.05 +TABD2
+TABD2 ENDT
$
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
194
$ 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 48 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 45degree phase lead.
Table 48 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
DAREA
TABLED1
RLOAD2 231
DPHASE
DAREA
RLOAD2 232
Main Index
195
CHAPTER 4
Frequency Response Analysis
The RLOAD2 entry describes a sinusoidal load in the form
(434)
where:
Logarithmic plots of the output are shown in the following figures. Figure 419 shows the magnitude of
the displacements for grid points 6 and 11. Figure 420 shows the magnitude of the modal displacements
for modes 1 and 2. Figure 421 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.
Figure 419 Displacement Magnitude (Log)
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)
P f ( ) { } AB f ( )e
i o f ( ) u 2tf t – + { }
{ } =
o
u
t
1.0
1.0E4
0 20 Frequency (Hz)
1.0
1.0E5
D
i
s
p
l
a
c
e
m
e
n
t
6
(
m
)
D
i
s
p
l
a
c
e
m
e
n
t
1
1
(
m
)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
196
Figure 420 Modal Displacement Magnitude (Log)
Figure 421 Bending Moment Magnitude at End A, Plane 1 (Log)
S
D
i
s
p
l
a
c
e
m
e
n
t
2
1.0E4
0
20 Frequency (Hz)
S
D
i
s
p
l
a
c
e
m
e
n
t
1
1.0
1.0E4
1.0E2
M
o
m
e
n
t
A
1
(
N

m
)
0.1
0 20
1000
Frequency (Hz)
Main Index
197
CHAPTER 4
Frequency Response Analysis
Bracket Model
Consider the bracket model shown in Figure 422. An oscillating pressure load of 3 psi is applied to the
elements on the top face in the zdirection. 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.
Figure 422 Bracket Model
Listing 49 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 49 shows the relationship
between the Case Control commands and the Bulk Data entries.
Listing 49 Input File (Abridged) for the Bracket Model
$
$ FILE bd04bkt.dat
$
$ BRACKET MODEL
$ CHAPTER 4, FREQUENCY RESPONSE
$
SOL 111 $ MODAL FREQUENCY RESPONSE
CEND
TITLE = BRACKET MODEL
SUBTITLE = MODAL FREQUENCY RESPONSE ANALYSIS
$
SPC = 1
$
METHOD = 777
DLOAD = 2
SDAMPING = 4
FREQUENCY = 5
$
$ OUTPUT REQUEST
SET 123 = 999
DISPLACEMENT(PHASE,PLOT)=123
$
$ XYPLOTS
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
198
$
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
199
CHAPTER 4
Frequency Response Analysis
ENDDATA
Figure 423 shows a logarithmic plot of the zdisplacement magnitude of grid point 999, which is the
concentrated mass at the center of the cutout.
Figure 423 Displacement Magnitude (Log)
Table 49 Relationship Between Case Control Commands and
Bulk Data Entries for the Bracket Model
Case Control Bulk Data
METHOD EIGRL
FREQUENCY FREQ1
SDAMPING TABDMP1
DLOAD
RLOAD1
TABLED1
PLOAD4
1.000
0.100
0.010
0.001
10
0.0001
0 Frequency (Hz)
100
D
i
s
p
l
a
c
e
m
e
n
t
(
m
)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
200
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 5: Transient Response Analysis
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
MD Nastran Dynamic Analysis User’s Guide
Overview
202
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
timevarying 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
203
CHAPTER 5
Transient Response Analysis
Direct Transient Response Analysis
In direct transient response (SOL 109), structural response is computed by solving a set of coupled
equations using direct numerical integration. Begin with the dynamic equation of motion in matrix form
(51)
The fundamental structural response (displacement) is solved at discrete times, typically with a fixed
integration time step .
By using a central finite difference representation for the velocity and the acceleration
at discrete times,
(52)
and averaging the applied force over three adjacent time points, the equation of motion can be rewritten
as:
(53)
Collecting terms, the equation of motion can be rewritten as:
(54)
where:
Matrix is termed the dynamic matrix, and is the applied force (averaged over three adjacent
time points). This approach is similar to the classical NewmarkBeta direct integration method except
that is averaged over three time points and is modified such that the dynamic equation of
motion reduces to a static solution if no or exists.
=
=
=
=
M   u
··
t ( ) { } B   u
·
t ( ) { } K   u t ( ) { } + + P t ( ) { } =
At
u
·
t ( ) { } u
··
t ( ) { }
u
·
n
{ }
1
2At
 u
n 1 +
u
n 1 –
– { } =
u
··
n
{ }
1
At
2
 u
n 1 +
2u
n
u +
n 1 –
– { } =
M
At
2
 u
n 1 +
2u
n
– u
n 1 –
+ ( )
B
2At
 u
n 1 +
u
n 1 –
– ( )
K
3
  u
n 1 +
u
n
u
n 1 –
+ + ( )
+
+
1
3
 P
n 1 +
P
n
P
n 1 –
+ + ( ) =
A
1
  u
n 1 +
{ } A
2
  A
3
  u
n
{ } A
4
  u
n 1 –
{ } + + =
A
1
M
At
2

B
2At

K
3
 + +
A
2
 
1
3
 P
n 1 +
P
n
P
n 1 –
+ + { }
A
3
 
2M
At
2

K
3
  –
A
4
 
M
At
2
 –
B
2At

K
3
 – +
A
1
  A
2
 
P t ( ) { } K  
K   u
n
{ } P
n
{ } = M   B  
Main Index
MD Nastran Dynamic Analysis User’s Guide
Direct Transient Response Analysis
204
The transient solution is obtained by decomposing and applying it to the righthand side of the above
equation. In this form, the solution behaves like a succession of static solutions with each time step
performing a forwardbackward substitution (FBS) on a new load vector. Note that the transient nature
of the solution is carried through by modifying the applied force matrix with the and
terms.
In its simplest form, the , , and 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 remains constant during the analysis. With a constant , the
matrix needs to be decomposed only once. Each progressive step in the analysis is only an FBS of
a new load vector. If is changed, 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.
Damping in Direct Transient Response
The damping matrix is used to represent the energy dissipation characteristics of a structure. In the
general case, the damping matrix is comprised of several matrices
(55)
where:
= damping elements (CVISC, CDAMPi) + B2GG
= B2PP direct input matrix + transfer functions
G = overall structural damping coefficient (PARAM,G)
= frequency of interest in radians per unit time (PARAM,W3) for the conversion of overall
structural damping into equivalent viscous damping
= global stiffness matrix
= element structural damping coefficient (GE on the MATi entry)
= frequency of interest in radians per unit time (PARAM,W4) for conversion of element
structural damping into equivalent viscous damping
= element stiffness matrix
A
1
 
A
2
  A
3
  A
4
 
M   B   K  
At At
A
1
 
At A
1
 
B  
B   B
1
  B
2
 
G
W
3
 K  
1
W
4
 G
E
K
E
 
¯
+ + + =
B
1
 
B
2
 
W
3
K  
G
E
W
4
K
E
 
Main Index
205
CHAPTER 5
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.
(56)
The structural damping force is a displacementdependent damping. The structural damping force is a
function of a damping coefficient and a complex component of the structural stiffness matrix.
(57)
Assuming constant amplitude oscillatory response for an SDOF system, the two damping forces are
identical if
(58)
or
(59)
Therefore, if structural damping is to be modeled using equivalent viscous damping , then the
equality Eq. (59) holds at only one frequency (see Figure 51).
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.
M   u
··
t ( ) { } B   u
·
t ( ) { } K   u t ( ) { } + + P t ( ) { } =
G
M   u
··
t ( ) { } 1 iG + ( ) K   u t ( ) { } + P t ( ) { } =
Gk be =
b
Gk
e
 =
G b
Main Index
MD Nastran Dynamic Analysis User’s Guide
Direct Transient Response Analysis
206
Figure 51 Structural Damping Versus Viscous Damping (Constant Oscillatory
Displacement)
Initial Conditions in Direct Transient Response
You may impose initial displacements and/or velocities in direct transient response. The TIC Bulk Data
entry is used to define initial conditions on the components of grid points. The IC Case Control command
is used to select TIC entries from the Bulk Data.
If initial conditions are used, initial conditions should be specified for all DOFs having nonzero values.
Initial conditions for any unspecified DOFs are set to zero.
Initial conditions and are used to determine the values of , , and used in
Eq. (54) to calculate .
(510)
(511)
In the presence of initial conditions, the applied load specified at is replaced by
(512)
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
Structural Damping f
s
= iGku
Damping
Force
Equivalent
Viscous
Damping
b Gk e
3
(or e
4
) =
f
v
bu
·
ibeu = =
e e
3
(or e
4
)
u
0
{ } u
·
0
{ } u
1 –
{ } P
0
{ } P
1 –
{ }
u
1
{ }
u
1 –
{ } u
0
{ } u
·
0
{ }At – =
P
1 –
{ } K   u
1 –
{ } B   u
·
0
{ } + =
t 0 =
P
0
{ } K   u
0
{ } B   u
·
0
{ } + =
Main Index
207
CHAPTER 5
Transient Response Analysis
Initial conditions may be specified only in the aset (see The Set Notation System Used in Dynamic
Analysis (App. B).
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.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modal Transient Response Analysis
208
Modal Transient Response Analysis
Modal transient response (SOL 112) is an alternate approach to computing the transient response of a
structure. This method uses the mode shapes of the structure to reduce the size, uncouple the equations
of motion (when modal or no damping is used), and make the numerical integration more efficient. Since
the mode shapes are typically computed as part of the characterization of the structure, modal transient
response is a natural extension of a normal modes analysis.
As a first step in the formulation, transform the variables from physical coordinates to modal
coordinates by
(513)
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 (513) 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
(514)
If the physical coordinates in terms of the modal coordinates (Eq. (513) is substituted into Eq. (514)),
the following equation is obtained:
(515)
This is now the equation of motion in terms of the modal coordinates. At this point, however, the
equations remain coupled.
To uncouple the equations, premultiply by to obtain
(516)
where:
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 offdiagonal 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
(517)
= modal (generalized) mass matrix
= modal (generalized) stiffness matrix
= modal force vector
u { }
c { }
u t ( ) { } o   c t ( ) { } =
o  
M   u
··
t ( ) { } K   u t ( ) { } + P t ( ) { } =
M   o   c
··
t ( ) { } K   o { } c t ( ) { } + P t ( ) { } =
o  
T
o  
T
M   o   c
··
{ } o  
T
K   o   c { } + o  
T
P t ( ) { } =
o  
T
M   o  
o  
T
K   o  
o  
T
P { }
m
i
c
··
i
t ( ) k
i
c
i
t ( ) + p
i
t ( ) =
Main Index
209
CHAPTER 5
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 are computed, physical responses are recovered as the
summation of the modal responses
(518)
Since numerical integration is applied to the relatively small number of uncoupled equations, there is not
as large a computational penalty for changing as there is in direct transient response analysis.
However, a constant is still recommended.
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.
Damping in Modal Transient Response Analysis
If the damping matrix exists, the orthogonality property (see Mathematical Overview of Normal
Modes Analysis, 47) of the modes does not, in general, diagonalize the generalized damping matrix
(519)
In the presence of a 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:
(520)
where:
= ith modal mass
= ith modal stiffness
= ith modal force
=
=
m
i
k
i
p
i
c
i
t ( )
u t ( ) { } o   c t ( ) { } =
At
At
B  
o  
T
B   o   diagonal =
B  
A
1
  c
n 1 +
{ } A
2
  A
3
  c
n
{ } A
4
  c
n 1 –
{ } + + =
A
1
 
o
T
 
M
At
2

B
2At

K
3
 + + o  
A
2
 
1
3
 o  
T
P
n 1 +
P
n
P
n 1 –
+ + { }
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modal Transient Response Analysis
210
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 . The equations of motion remain uncoupled
and have the following form for each mode:
(521)
or
(522)
where:
The TABDMP1 Bulk Data entry defines the modal damping ratios. A table is created by the
frequencydamping 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
etc. ENDT
Field Contents
TID Table identification number.
TYPE Type of damping units:
G (default)
CRIT
Q
A
3
 
o
T
 
2M
At
2

K
3
  – o  
A
4
 
o  
T M
At
2
 –
B
2At

K
3
 – + o  
b
i
m
i
c
··
i
t ( ) b
i
c
·
i
t ( ) k
i
c
i
t ( ) + + p
i
t ( ) =
c
··
i
t ( ) 2.
i
e
i
c
·
i
t ( ) e
i
2
c
i
t ( ) + +
1
m
i
 p
i
t ( ) =
.
i
b
i
2m
i
e
i
( ) modal damping ratio ÷
e
i
2
k
i
m
i
modal frequency (eigenvalue) ÷
f
1
g
1
f
2
g
2
f
3
g
3
Main Index
211
CHAPTER 5
Transient Response Analysis
At resonance, the three types of damping are related by the following equations:
(523)
The values of (units = cycles per unit time) and define pairs of frequencies and dampings. Note
that can be entered as structural damping (default), critical damping, or quality factor. The entered
damping is internally converted to structural damping using Eq. (523). Straightline interpolation is used
for modal frequencies between consecutive 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 51 and if modes exist at 1.0, 2.5, 3.6, and 5.5 Hz,
MD Nastran interpolates and extrapolates as shown in Figure 52 and the table. Note that there is no table
entry at 1.0 Hz; MD Nastran uses the first two table entries at and to extrapolate the
value for .
Figure 52 Example TABDMP1
fi Frequency value (cycles per unit time).
gi Damping value in the units specified.
Field Contents
.
i
b
i
b
cr

G
i
2
 = =
b
cr
2m
i
e
i
=
Q
i
1
2.
i

1
G
i
 = =
fi gi
gi
fi
f 2.0 = f 3.0 =
f 1.0 =
0.10
0.20
2.0 4.0 6.0
f(Hz)
0.0
= entered value
= computed value
+
+
+ +
+
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modal Transient Response Analysis
212
With the modal equations in the form of Eq. (522), 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
(524)
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
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.
Mode Truncation in Modal Transient Response Analysis
It is possible that not all of the computed modes are required in the transient response solution. Often,
only the lowest few suffice for dynamic response calculation. It is quite common to evaluate the
frequency content of transient loads and determine a frequency above which no modes are noticeably
excited. This frequency is called the cutoff frequency. The act of specifically not using all of the modes
of a system in the solution is termed mode truncation. Mode truncation assumes that an accurate solution
can be obtained using a reduced set of modes. The number of modes used in a solution is controlled in a
modal transient response analysis through a number of methods.
Table 51 Example TABDMP1 Interpolation/Extrapolation
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
c t ( ) e
bt – 2m
c
o
e
d
t cos
c
·
o
b 2m ( ) c
o
+
e
d
 e
d
t sin +
\ .

 
e
bt – 2m 1
me
d
 e
bt 2m
p t ( ) e
d
t t – ( )dt sin
0
t
í
+
=
B  
Main Index
213
CHAPTER 5
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, highfrequency modes are
truncated because they are more costly to compute. So, truncating highfrequency modes truncates high
frequency response. In most cases, highfrequency 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 ‘afterthefact’ 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 quasistatic influence of the truncated highfrequency modes.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Initial Condition Usage in Linear Transient Analysis
214
Initial Condition Usage in Linear Transient Analysis
The use of initial conditions in linear transient analysis includes:
• Specification of Initial Conditions in Modal Transient Analysis (SOL 112)
The specification of initial conditions in modal transient analysis (SOL 112) either in physical
coordinates (as is done in direct transient analysis) or in modal coordinates, by use of the IC Case
Control command and to the TIC Bulk Data entry.
• Usage of Static Solutions as Initial Conditions in Linear Transient Analysis (SOLs 109 and 112)
The results of static analysis solutions can be used as the initial conditions in linear transient
analysis (SOLs 109 and 112). Differential stiffness effects may be included if the user so
desires.
IC
The format of the IC Case Control command is as follows:
Format:
Examples:
IC = 10
IC(PHYSICAL) = 100
IC(MODAL) = 200
IC(STATSUB) = 1000
IC(STATSUB,DIFFK) = 2000
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.
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.
IC
PHYSICAL
MODAL
STATSUB[,DIFFK]
n =
Main Index
215
CHAPTER 5
Transient Response Analysis
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
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)
Describer Meaning
Main Index
MD Nastran Dynamic Analysis User’s Guide
Initial Condition Usage in Linear Transient Analysis
216
Initial Condition Specification for Enforced Motion Usage via
SPC/SPCD
Enforced acceleration or enforced velocity usage in transient analysis via SPC/SPCD specification
requires integration to compute the corresponding enforced velocities and/or displacements. This
integration involves the use of initial conditions. The user can specify initial displacements for enforced
DOFs in the case of enforced velocity usage via SPC/SPCD and can specify initial displacements as well
as initial velocities for enforced DOFs in the case of enforced acceleration usage via SPC/SPCD. The
initial displacement and velocity values are specified via corresponding factors in two new fields that
have been added to the TLOAD1 and TLOAD2 Bulk Data entries. Details will be clear from the description
of these expanded entries in the MD Nastran Quick Reference Guide. This capability will greatly help
users in performing enforced motion studies with a variety of scenarios.
It should be noted here that the initial conditions for the enforced DOFs mentioned here are distinct from,
and may be used in conjunction with, the initial conditions for independent DOFs that may be specified
by a TIC Bulk Data entry.
Main Index
217
CHAPTER 5
Transient Response Analysis
Dynamic Data Recovery in Modal Transient Response
Analysis
In modal transient response analysis, two options are available for recovering displacements and
stresses: mode displacement method and matrix method. Both methods give the same answers, although
with cost differences.
The mode displacement method computes the total physical displacements for each time step from the
modal displacements and then computes element stresses from the total physical displacements. The
number of operations is proportional to the number of time steps.
The matrix method computes displacements per mode and element stresses per mode, and then computes
physical displacements and element stresses as the summation of modal displacements and element
stresses. Costly operations are proportional to the number of modes.
Since the number of modes is usually much less that the number of time steps, the matrix method is
usually more efficient and is the default. The mode displacement method can be selected by using
PARAM,DDRMM,1 in the Bulk Data. The mode displacement method is required when “time frozen”
deformed structure plots are requested (see Results Processing (Ch. 15)).
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modal Versus Direct Transient Response
218
Modal Versus Direct Transient Response
Some general guidelines can be used in selecting modal transient response analysis versus direct transient
response analysis. These guidelines are summarized in Table 52.
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 highfrequency excitation. To capture high frequency
response in a modal solution, less accurate highfrequency 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 52 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.
Table 52 Modal Versus Direct Transient Response
Modal Direct
Small Model X
Large Model X
Few Time Steps X
Many Time Steps X
High Frequency Excitation X
Normal Damping X
Higher Accuracy X
Initial Conditions X X
Main Index
219
CHAPTER 5
Transient Response Analysis
Transient Excitation Definition
An important aspect of a transient response analysis is the definition of the loading function. In a
transient response analysis, the force must be defined as a function of time. Forces are defined in the
same manner whether the direct or modal method is used.
The following Bulk Data entries are used for the transient load 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.
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.
TimeDependent Loads  TLOAD1 Entry
The TLOAD1 Bulk Data entry is the most general form in which to define a timedependent load. It
defines a dynamic loading of the form
(525)
The coefficients of the force are defined in tabular format. You need not explicitly define a force at every
instant in time for which the transient solution is evaluated. Only those values which describe the
character of the loading are required. MD Nastran interpolates linearly for intermediate values.
TLOAD1 Tabular input
TLOAD2 Analytical function
DAREA Spatial distribution of dynamic load
TABLEDi Tabular values versus time
DLOAD Combines dynamic load sets
DELAY Time delay
1 2 3 4 5 6 7 8 9 10
TLOAD1
SID
EXCITEID
DELAYI/
DELAYR
TYPE TID US0 VS0
P t ( ) { } A F t t – ( ) · { } =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Transient Excitation Definition
220
Example:
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:
TimeDependent Loads – TLOAD2 Entry
The TLOAD2 entry is a general analytical form with which to define a timedependent load. The value
of the force at a particular instant in time is determined by evaluating the analytic function at the specific
time. You enter the appropriate constants in the function.
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 . 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 degreesoffreedom 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 . (Integer > 0)
US0 Factor for initial displacements of the enforced degreesoffreedom. See Remarks 10.
and 12. (Real; Default = 0.0)
VS0 Factor for initial velocities of the enforced degreesoffreedom. See Remarks 11. and
12. (Real; Default = 0.0)
A { }
t
t
F t ( )
TYPE TYPE of Dynamic Excitation
0, L, LO, LOA or LOAD Applied load (force or moment) (Default)
1, D, DI, DIS, or DISP Enforced displacement using large mass or SPC/SPCD data
2, V, VE, VEL or VELO Enforced velocity using large mass or SPC/SPCD data
3, A, AC, ACC or ACCE Enforced acceleration using large mass or SPC/SPCD data
Main Index
221
CHAPTER 5
Transient Response Analysis
The TLOAD2 Bulk Data entry defines dynamic excitation in the form:
(526)
where: .
.
Example:
1 2 3 4 5 6 7 8 9 10
TLOAD2
SID
EXCITEID
DELAYI/
DELAYR
TYPE T1 T2 F P
C B US0 VS0
TLOAD2 4 10 5.0 2.1 4.7 12.0
2.0
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 . 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 degreesoffreedom 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 degreesoffreedom. See Remarks 10.
and 12. (Real; Default = 0.0)
VSO Factor for initial velocities of the enforced degreesoffreedom. See Remarks 11. and
12. (Real; Default = 0.0)
P t ( ) { }
0 , t T1 t + ( ) or t T2 t + ( ) > <
At
˜
B
e
Ct
˜
2tFt
˜
P + ( ) cos , T1 t + ( ) t T2 t + ( ) s s
¹
¦
´
¦
¦
=
t
˜
t T1 t – ( ) – =
A { }
t
t
Main Index
MD Nastran Dynamic Analysis User’s Guide
Transient Excitation Definition
222
Spatial Distribution of Loading – DAREA Entry
The DAREA Bulk Data entry defines the degreesoffreedom where the dynamic load is to be applied and
a scale factor to be applied to the loading. The DAREA entry provides the basic spatial distribution of
the dynamic loading.
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.
Time Delay  DELAY Entry
The DELAY Bulk Data entry defines the time delay in an applied load.
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.
Dynamic Load Tabular Function – TABLEDi Entries
The TABLEDi Bulk Data entries (i = 1 through 4) each define a tabular function for use in generating
frequencydependent dynamic loads. The form of each TABLEDi entry varies slightly, depending on
the value of i, as does the algorithm for y(x). The x values need not be evenly spaced.
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.
1 2 3 4 5 6 7 8 9 10
DELAY
SID P1 C1 P2 C2
Field Contents
SID Set ID specified by TLOADi entry.
Pi Grid, extra, or scalar point ID.
Ci Component number.
Time delay for Pi, Ci.
t
t
1
t
2
t
i
Main Index
223
CHAPTER 5
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 53. 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.
Figure 53 Interpolation and Extrapolation for TABLED1, TABLED2, and TABLED3 Entries
The TABLED1 Bulk Data entry has the following format:
The TABLED1 entry uses the algorithm
(527)
The algorithms used for interpolation and extrapolation are as follows:
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 xaxis.
(Character: “LINEAR” or “LOG”; Default = “LINEAR”)
YAXIS Specifies a linear or logarithmic interpolation for the yaxis.
(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
x
Linear Extrapolation of Segment x
2
– x
1
Linear Extrapolation of Segment x
6
– x
7
Linear Interpolation Between Endpoints
x
1
x
2
x
3
x
4
x
5
x
6
x
7
y y
T
x ( ) =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Transient Excitation Definition
224
The TABLED2 Bulk Data entry has the following format:
The TABLED2 entry uses the algorithm
(528)
The TABLED3 Bulk Data entry has the following format:
XAXIS YAXIS
LINEAR LINEAR
LOG LINEAR
LINEAR LOG
LOG LOG
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.
1 2 3 4 5 6 7 8 9 10
TABLED3
TID X1 X2
x1 y1 x2 y2 x3 y3 etc. ENDT
y
T
x ( )
xj x –
xj xi –
 yi
x xi –
xj xi –
 yj +
xj x ( ) ln
xj xi ( ) ln
 yi
x xi ( ) ln
xj xi ( ) ln
 yj +
xj x –
xj xi –
 yi ln
x xi –
xj xi –
 yj ln + exp
xj x ( ) ln
xj xi ( ) ln
 yi ln
x xi ( ) ln
xj xi ( ) ln
 yj ln + exp
y y
T
x X1 – ( ) =
Main Index
225
CHAPTER 5
Transient Response Analysis
The TABLED3 entry uses the algorithm
(529)
The TABLED4 entry has the following format:
The TABLED4 entry uses the algorithm
(530)
is the degree of the power series. When , is used for ; when , is used for .
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
Field Contents
TID Table identification number.
X1, X2 Table parameters.
xi, yi Tabular values.
ENDT Ends the table input.
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. .
Ai Coefficients.
y y
T
x X1 –
X2

\ .
 
=
X2 0.0; X3 X4 < = ( )
y Ai
x X1 –
X2

\ .
 
i
i 0 =
N
¯
=
N x X3 < X3 x x X4 > X4 x
Main Index
MD Nastran Dynamic Analysis User’s Guide
Transient Excitation Definition
226
in addition to the following entries in the Bulk Data Section:
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 54 shows the TABLED1 time history and the applied load (scaled by the DAREA entry and time
shifted by the DELAY entry).
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
Main Index
227
CHAPTER 5
Transient Response Analysis
Figure 54 Time History from the TABLED1 Entry (Top) and Applied Load (Bottom)
Static Load Sets
MD Nastran does not have specific data entries for many types of dynamic loads. Only concentrated
forces and moments can be specified directly using DAREA entries. To accommodate more complicated
loadings conveniently, the EXCITEID on the TLOADi Bulk Data entry is used to define static load
entries that define the spatial distribution of dynamic loads. The EXCITEID may refer to one or more
static load entries (FORCE, PLOADi, GRAV, etc.). All static loads with the Set ID referenced on the
EXCITEID entry define the spatial distribution of the dynamic loading. MD Nastran converts this
information to equivalent dynamic loading.
Figure 55 demonstrates the relationships of these entries. To activate a load set defined in this manner,
the DLOAD Case Control command refers to the Set ID of the selected DLOAD or TLOADi entry. The
EXCITEID entry points to the static loading entries that are used to define dynamic loadings and/or
DAREA Set ID. Together this relationship defines a complete dynamic loading. To apply dynamic
loadings in this manner, the DLOAD Case Control command and the TLOADi Bulk Data entry must be
defined.
0.5 1.0 1.5 2.0 2.5
0.5 1.0 1.5 2.0 2.5
1.0
Time (sec)
Amplitude
on
TABLED1
4.0
Time (sec)
Force
Amplitude
0
0
Main Index
MD Nastran Dynamic Analysis User’s Guide
Transient Excitation Definition
228
Figure 55 Relationship of Dynamic and Static Load Entries
Dynamic Load Set Combination  DLOAD
One of the requirements of transient loads is that all TLOAD1s and TLOAD2s must have unique SIDs.
If they are to be applied in the same analysis, they must be combined using the DLOAD Bulk Data entry.
The total applied load is constructed from a combination of component load sets as follows:
(531)
where:
The DLOAD Bulk Data entry has the following format:
= overall scale factor
= scale factor for ith load set
= ith set of loads
= total applied load
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.
DLOAD
TLOADi
Static Load Entries Dynamic Load
(Temporal
Distribution)
(Spatial
Distribution)
Case Control
Bulk Data
P { } S S
i
P
i
{ }
i
¯
=
S
S
i
P
i
P
Main Index
229
CHAPTER 5
Transient Response Analysis
As an example, in the following DLOAD entry:
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.
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
Field Contents
Main Index
MD Nastran Dynamic Analysis User’s Guide
Integration Time Step
230
Integration Time Step
The TSTEP Bulk Data entry is used to select the integration time step for direct and modal transient
response analysis. This entry also controls the duration of the solution and which time steps are to be
output. The TSTEP Bulk Data entry is selected by the Set ID referenced on the TSTEP Case Control
command.
The integration time step must be small enough to represent accurately the variation in the loading. The
integration time step must also be small enough to represent the maximum frequency of interest. The
maximum frequency of interest is often called the cutoff frequency. It is recommended to use at least
ten solution time steps per period of response for the cutoff frequency. For a given integration time step,
integration errors increase with increasing natural frequency because there is an upper limit to the
frequency that can be represented by a given time step. Also, integration errors accumulate with total
time.
In both direct and modal transient analysis, the cost of integration is directly proportional to the number
of time steps. For example, doubling the load duration doubles the integration effort.
In specifying the duration of the analysis on the TSTEP entry, it is important to use an adequate length
of time to properly capture long period (low frequency) response. In many cases, the peak dynamic
response does not occur at the peak value of load nor necessarily during the duration of the loading
function. A good rule is: always solve for at least one cycle of response for the lowest frequency mode
after the peak excitation.
You may change 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 NO1
N2 NO2
etc.
Field Contents
SID Set ID specified by a TSTEP Case Control command.
Ni Number of time steps of value .
Integration time step.
NOi Output every NOith time step.
At
At
1
At
2
At
i
At
i
Main Index
231
CHAPTER 5
Transient Response Analysis
Transient Excitation Considerations
A number of important considerations must be remembered when applying transient loads. The
averaging of applied loads (Eq. (53)) in the integration smooths the force and decreases the apparent
frequency content. Very sharp spikes in a loading function induce a highfrequency transient response.
If the highfrequency transient response is of primary importance in an analysis, a very small integration
time step must be used.
It is also important to avoid defining discontinuous forcing functions when describing applied loads. The
numerical integration of discontinuous forcing functions may cause different results for the same
analysis run on different computers because of slight numerical differences on different computer types.
If the analysis calls for loadings with sharp impulses, it is best to smooth the impulse over at least one
integration time increment.
The loading function must accurately describe the spatial and temporal distribution of the dynamic load.
Simplifying assumptions must not change the character of the load in magnitude, location, or frequency
content.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Solution Control for Transient Response Analysis
232
Solution Control for Transient Response Analysis
The following tables summarize the data entries that can be used to control a transient response analysis.
Certain data entries are required, some data entries are optional, while others are user selectable.
In the Executive Control Section of the MD Nastran input file, a solution must be selected using the SOLi
statement where i is an integer value chosen from Table 53.
We have applied these solutions in the examples that follow.
In the Case Control Section of the MD Nastran input file, you must select the solution parameters
associated with the current analysis (i.e., time steps, loads, and boundary conditions) and also the output
quantities required from the analysis. The Case Control commands directly related to transient response
analysis are listed in Table 54. They can be combined in the standard fashion with the more generic
commands, such as SPC, MPC, etc.
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 55 and Table 56.
Table 53 Transient Response Solutions in MD Nastran
Structured Solution Sequences
Direct 109
Modal 112
Table 54 Transient Response Case Control Commands
Case Control
Commands
Direct
or
Modal Description
Required/
Optional
DLOAD Both Select the dynamic load set from the Bulk Data Required*
*Not required when using initial conditions.
TSTEP Both Select the TSTEP entry from the Bulk Data Required
METHOD Modal Select the eigenvalue extraction parameters Required
LOADSET Both Select the LSEQ set from the Bulk Data Optional
SDAMPING Modal Select the modal damping table from the Bulk Data Optional
IC Direct Select TIC entries for initial conditions from the Bulk Data Optional
OTIME Both Select the times for output (default = all) Optional
Main Index
233
CHAPTER 5
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 57.
Table 55 Grid Point Output from a Transient Response Analysis
Case Control
Command Description
ACCELERATION Grid point acceleration time history for a set of grid points
DISPLACEMENT
(or VECTOR)
Grid point displacement time history for a set of grid points
GPSTRESS Grid point stress time history (requires SURFACE/VOLUME definition in the
OUTPUT(POST) section of the Case Control)
OLOAD Requests applied load table to be output for a set of grid points
SACCELERATION Requests solution set acceleration output: dset in direct solutions and modal
variables in modal solutions
SDISPLACEMENT Requests solution set displacement output: dset in direct solutions and modal
variables in modal solutions
SVECTOR Requests real eigenvector output for the aset in modal solutions
SVELOCITY Requests solution set velocity output: dset in direct solutions and modal
variables in modal solutions
SPCFORCES Requests forces of singlepoint constraint for a set of grid points
VELOCITY Grid point velocity time history for a set of grid points
Table 56 Element Output from a Transient Analysis
Case Control Command Description
ELSTRESS (or STRESS) Element stress time history for a set of elements
ELFORCE (or FORCE) Element force time history for a set of elements
STRAIN Element strain time history for a set of elements
Main Index
MD Nastran Dynamic Analysis User’s Guide
Solution Control for Transient Response Analysis
234
Table 57 Bulk Data Entries for Transient Response Analysis
Bulk Data Entry
Direct or
Modal Description
Required/
Optional
TTEMP  MD Only Both Integration time step and solution control Required
TLOAD1i Both Dynamic loading Required*
*Not required for initial conditions.
EIGR or EIGRL Modal Eigenvalue analysis parameters Required
LSEQ Both Dynamic loading from static loads Optional
TABLED1i Both Timedependent tables for TLOADi Optional*
TIC Direct Initial conditions on grid, scalar, and extra points Optional
DAREA Both Load component and scale factor Optional
DELAY Both Time delay on dynamic load Optional
DLOAD Both Dynamic load combination, required if TLOAD1
and TLOAD2 are used
Optional
TABDMP1 Modal Modal damping table Optional
Main Index
235
CHAPTER 5
Transient Response Analysis
Restarts in Linear Transient Analysis
Restarts in Linear Transient may be used to continue time step integration from a previous run without
repeating the earlier computations. This is accomplished by extending the usage of the STIME parameter
to SOLs 109 and 112 (while retaining its current usage in SOLs 129 and 159).
In order to use this feature in SOLs 109 and 112, ensure that the model and the constraints, as well as the
subcase setup in the restart run, are the same as those in the previous run. The user may, however, specify
different TSTEP and DLOAD requests in the Case Control and also different TSTEP and dynamic
loading entries in the Bulk Data compared to the previous run.
PARAM,STIME,x is used to specify the starting time of the restart run. If STIME exceeds the last output
time of the previous run, the starting time is assumed to be the last output time. Otherwise, the starting
time is assumed to be the output time of the previous run (not necessarily the last output time) that is
closest to STIME. In other words, the starting time of the restart run need not be the last output time of
the previous run, but may be any earlier output time of the previous run. MD Nastran informs the user
that it is a restart run and indicates the starting time (determined as above) that is used for the restart run.
The loading and the results output from the continued restart run will start from the new starting time.
It is important to note that the use of the continue feature requires that the model and the constraints, as
well as the subcase setup in the restart run be the same as those in the previous run. It also requires that
the databases from the previous run be available. It is the user’s responsibility to ensure that this
condition is satisfied. If this condition is not met, the program may terminate the execution with a fatal
error or give erroneous results.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
236
Examples
This section provides several examples showing the input and output. These examples are
These examples are described in the sections that follow.
TwoDOF Model
Consider the twoDOF system shown in Figure 56. 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
of 0.01 second. Damping is neglected in the analysis. Listing 51 shows part of the input file. See
Install_dir/mdxxxx/doc/dynamics/db05two.dat.
Figure 56 TwoDOF Model
Model
Transient Response Bulk
Data Entries Output
bd05two TSTEP, TIC XY plots
bd05bar EIGRL, TSTEP,
TABDMP1, DLOAD,
TLOAD2, DAREA,
DELAY
XY plots
bd05bkt EIGRL, TSTEP,
TABDMP1, TLOAD1,
LSEQ, TABLED1,
PLOAD4
XY plot
At
y
x
Grid Point 1
Grid Point 2
m
1
k
1
m
2
k
2
Main Index
237
CHAPTER 5
Transient Response Analysis
Listing 51 Input File (Abridged) for the TwoDOF Example
$ FILE bd05two.dat
$
$ TWODOF SYSTEM
$ CHAPTER 5, TRANSIENT RESPONSE
$
TIME 5
SOL 109 $ DIRECT TRANSIENT RESPONSE
CEND
TITLE = TWODOF SYSTEM
SUBTITLE = DIRECT FREQUENCY RESPONSE
LABEL = INITIAL DISPL. AT GRID 2
$
$ SPECIFY SPC
SPC = 996
$
$ SPECIFY DYNAMIC INPUT
TSTEP = 888
IC = 777
$
$ SELECT OUTPUT
SET 11 = 1,2
DISPLACEMENT(PLOT) = 11
$
$ XYPLOTS
$
... XY plot commands ...
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10.$
$
$ ENTRIES FOR TRANSIENT RESPONSE
$
$ INITIAL CONDITION
$TIC SID G C U0 V0
TIC 777 2 2 0.1
$
$ TIME STEP
$TSTEP SID N1 DT1 NO1
TSTEP 888 1000 0.01 1
$
... basic model ...
$
ENDDATA
Table 58 shows the relationship between the Case Control commands and the Bulk Data entries. This
example represents the simplest form of dynamic response input. The only required entries are those that
define the time step and the initial conditions. Note that the unspecified initial conditions are assumed
to be zero. Note, too, that the initial conditions are available only for direct transient response analysis.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
238
Figure 57 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 closelyspaced 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:
(532)
where:
In this example, is 5.04 Hz and is 0.25 Hz. The lower frequency is called the beat frequency
and is the frequency at which energy transfer occurs.
Table 58 Relationship Between Case Control Commands and Bulk Data Entries for the
TwoDOF Model
Case Control Bulk Data
TSTEP TSTEP
IC TIC
f1 = lower of the closelyspaced mode frequencies
f2 = higher of the closelyspaced mode frequencies
f
hi gher
f1 f 2 +
2
 =
f
l ower
f2 f1 –
2
 =
f
hi gher
f
l ower
Main Index
239
CHAPTER 5
Transient Response Analysis
Figure 57 Displacements of Grid Points 1 and 2
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
240
Cantilever Beam Model
Consider the cantilever beam shown in Figure 58. This beam model is the same as in Examples (Ch. 4).
Modal transient response (SOL 112) is run with loads applied to grid points 6 and 11 as shown in
Figure 59. The analysis is run for a duration of 2 seconds with a of 0.001 second. Modal damping
of 5% critical damping is used for all modes. Modes up to 3000 Hz are computed using the Lanczos
method. Listing 52 shows part of the input file. See Install_dir/mdxxxx/doc/dynamics/db05bar.dat.
Figure 58 Cantilever Beam Model with Applied Loads
Figure 59 Applied Loads for the Beam Model
Listing 52 Input File (Abridged) for the Beam Example
$ FILE bd05bar.dat
$
$ CANTILEVER BEAM MODEL
At
y
x
6
p
6
(t)
11
p
11
(t)
Time
(sec)
6
6
Force 11 (N)
0.5
Time
(sec)
3
3
Force 6 (N)
0.1 0.5
Main Index
241
CHAPTER 5
Transient Response Analysis
$ CHAPTER 5, TRANSIENT RESPONSE
$
SOL 112 $ MODAL TRANSIENT RESPONSE
TIME 10
CEND
TITLE = CANTILEVER BEAM
SUBTITLE = MODAL TRANSIENT RESPONSE
$
SPC = 21
DLOAD = 22
TSTEP = 27
SDAMPING = 25
$
METHOD = 10
$
$ 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
$
$ XYPLOTS
$
... XY plot commands ...
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10..
...$
$
$EIGRL SID V1 V2 MSGLVL
EIGRL 10 0.1 3000. 0
$
$TSTEP SID N1 DT1 NO1
TSTEP 27 2000 0.001 1
$
$ MODAL DAMPING OF 5% IN ALL MODES
$TABDMP1 TID TYPE +TABD
$+TABD F1 G1 F2 G2 ETC.
TABDMP1 25 CRIT +TABD
+TABD 0. 0.05 1000. 0.05 ENDT
$
$ DYNAMIC LOADING
$DLOAD SID S S1 L1 S2 L2
DLOAD 22 1.0 1.0 231 1.0 232
$TLOAD2 SID DAREA DELAY TYPE T1 T2 F P +TL1
$+TL1 C B
TLOAD2 231 241 0 0.0 0.5 2.0 90.
TLOAD2 232 242 262 0 0.0 0.5 4.0 90.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
242
$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 59 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.
Plotted output is shown in the following figures. Figure 510 shows the applied loads at grid points 6 and
11. Figure 511 shows the plots of the displacements for grid points 6 and 11. Figure 512 shows the
accelerations for grid points 6 and 11. Figure 513 shows the bending moment at end A in plane 1 for
element 6. Figure 514 shows the modal displacements for modes 1 and 2.
Table 59 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
DAREA
TLOAD2 231
DAREA
DELAY
TLOAD2 232
Main Index
243
CHAPTER 5
Transient Response Analysis
Figure 510 Applied Loads at Grid Points 6 and 11
3.0
3.0
0.0
6.0
6.0
0.0
0.0 2.0
Time (sec)
F
o
r
c
e
1
1
(
N
)
F
o
r
c
e
6
(
N
)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
244
Figure 511 Displacements at Grid Points 6 and 11
0.03
0.0
0.0
0.08
0.06
0.0
0.0 2.0
Time (sec)
D
i
s
p
l
a
c
e
m
e
n
t
6
(
m
)
D
i
s
p
l
a
c
e
m
e
n
t
1
1
(
m
)
Main Index
245
CHAPTER 5
Transient Response Analysis
Figure 512 Accelerations at Grid Points 6 and 11
4.0
6.0
0.0
10.0
15.0
0.0
0.0 2.0
Time (sec)
A
c
c
e
l
e
r
a
t
i
o
n
6
m
2
s
e
c
(
)
A
c
c
e
l
e
r
a
t
i
o
n
1
1
m
2
s
e
c
(
)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
246
Figure 513 Bending Moment A1 for Element 6
25.0
20.0
0.0
0.0 2.0
Time (sec)
B
e
n
d
i
n
g
M
o
m
e
n
t
A
1
(
N
m
)
Main Index
247
CHAPTER 5
Transient Response Analysis
Figure 514 Modal Displacements for Modes 1 and 2
Bracket Model
Consider the bracket model shown in Figure 515. A pressure load of 3 psi is applied to the elements in
the top face in the zdirection with the time history shown in Figure 516. 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.
0.08
0.10
0.0
0.0015
0.0010
0.0
0.0 2.0
Time (sec)
S
D
i
s
p
l
a
c
e
m
e
n
t
1
S
D
i
s
p
l
a
c
e
m
e
n
t
2
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
248
Figure 515 Bracket Model
Figure 516 Time Variation for Applied Load
Listing 53 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 510 shows the
relationship between the Case Control commands and the Bulk Data entries.
Listing 53 Input File (Abridged) for the Bracket Model
$ FILE bd05bkt.dat
$
$ BRACKET MODEL
$ CHAPTER 5, TRANSIENT RESPONSE
$
SOL 112 $ MODAL TRANSIENT RESPONSE
TIME 100
CEND
3 psi
0 0.10 0.15 5
Time (sec)
Pressure
Main Index
249
CHAPTER 5
Transient Response Analysis
TITLE = BRACKET MODEL
SUBTITLE = MODAL TRANSIENT RESPONSE ANALYSIS
$
SPC = 1
$
METHOD = 777
$
DLOAD = 2
LOADSET = 3
SDAMPING = 4
TSTEP = 5
$
$ OUTPUT REQUEST
SET 123 = 999
DISPLACEMENT(PLOT)=123
$
$ XYPLOTS
$
... XY plot commands ...
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10..
....
$
$ NORMAL MODES TO 3000 HZ
$EIGRL SID V1 V2
EIGRL 777 0.1 3000.
$
$ 4 SECONDS OF RESPONSE
$TSTEP SID N1 DT1 NO1
TSTEP 5 800 0.005 1
$
$ MODAL DAMPING OF 2% CRITICAL
$TABDMP1 TID TYPE +TABD1
$+TABD1 F1 G1 F2 G2 ETC.
TABDMP1 4 CRIT +TABD1
+TABD1 0.0 0.02 3000.0 0.02 ENDT
$
$ LOAD DEFINITION
$
$TLOAD1 SID DAREA DELAY TYPE TID
TLOAD1 2 999 22
$
$LSEQ SID DAREA LID TID
LSEQ 3 999 1
$
$ TIME HISTORY
$TABLED1 TID +TABL1
$+TABL1 X1 Y1 X2 Y2 ETC.
TABLED1 22 +TABL1
+TABL1 0.0 0.0 0.1 0.0 0.15 1.0 5.0 1.0 +TABL2
+TABL2 ENDT
$
$ PRESSURE LOAD OF 3 PSI PER ELEMENT
$PLOAD4 SID EID P1
PLOAD4 1 171 3.
PLOAD4 1 172 3.
PLOAD4 1 160 3.
etc.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
250
$
... basic model ...
$
ENDDATA
Figure 517 shows a plot of the zdisplacement of grid point 999, which is the concentrated mass at the
center of the cutout.
Table 510 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
Main Index
251
CHAPTER 5
Transient Response Analysis
Figure 517 Displacement Time History for Grid Point 999
=
D
i
s
p
l
a
c
e
m
e
n
t
(
m
)
0
0.04
0.08
0 4 Time (sec)
Main Index
MD Nastran Dynamic Analysis User’s Guide
MAX/MIN Output (New Form)
252
MAX/MIN Output (New Form)
The primary use of the MAXMIN operation is to allow for rapid identification of the grids and elements
with high response values. The flexible syntax of the MAXMIN(DEF) Case Control entry allows
grouping of operations to further reduce the number of components monitored. The commands have been
developed to avoid the standard data recovery operations unless downstream operations require the
information for intermediate durations. The FULL operation on the MAXMIN(DEF) entry, recovers the
complete grid or element entry which remains in the final response table.
A major change in data recovery strategy is the introduction of a “sort2” processing sequence. Data
recovery is normally performed using only “sort1” recovery techniques. The output tables produced are
then transposed to form the “sort2” order, that is, for time or frequency domain recovery of grids or
element responses. This transpose operation, although efficient, consumed computer resources, most
notably, was the disk storage requirements. During the “sort2” technique in Modal Transient (SOL 112)
the matrices are transposed in modal space before data recovery is performed. This eliminates the “sort1”
step. However, when BAR elements with intermediate station data recovery is needed, the transpose to
“sort1” is still required. Be aware that the “op2” tapes used in passing data to the post processor may also
require transpose operations due to their current interface limitations. The param,post,0 operation, which
is the default option for MSC.Patran, is sort independent and does not require the intermediate step.
The Nastran Data Definition Language (NDDL) is needed to interpret the table entries. Because the
output data block, generically named OMM, is mostly definable by the NDDL, it can be operated upon
by the standard OUTPUT2 formatting function.
The output is sequenced within the print file before the standard data recovery of grid displacements and
elements. The standard output heading is:
M A X / M I N S U M M A R Y ( xxxxxx )
where xxxxx describes the class being monitored, such as displacements for grids or element names for
stresses, forces or strains.
Using the FULL action keyword causes additional grid or element information to be accumulated in the
OMM data block for the retained entries meeting the monitoring criteria.
All directives that select MAXMIN operations are located within the Case Control Section.
The Case Control command MAXMIN(DEF) must appear before encountering any SUBCASE entry.
Within a SUBCASE definition, additional entries of the form MAXMIN(GRID), MAXMIN(ELEM), or
MAXMIN(BOTH) activate the operation for the SUBCASE during data recovery. Both entries are
necessary for any operation to occur.
The MAXMIN(DEF) entry can be subdivided into four definitional areas. Multiple MAXMIN(DEF)
entries can be present in the Case Control Section. It is recommended that only one be supplied for a
class, although more than one class can be supplied on one entry. Multiple class action selection could
result in formulation of an accumulative output domain.
MAXMIN(DEF) class(es) <element type(s)> component(s)
action(s)
Main Index
253
CHAPTER 5
Transient Response Analysis
The grid classes are: DISPLACMENT, VELOCITY, ACCELERATION, OLOAD, SPCF (Single
Point Constraint Forces), MPCF (MultiPoint 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
MD Nastran Dynamic Analysis User’s Guide
MAX/MIN Output (New Form)
254
MAXMIN(ELEM,[{PRINT/NOPRINT},PUNCH],[VONMISES/SHEAR],[STRCUR/FIB
ER],[CENTER/BILINetc]) = {ALL,SET,NONE)
MAXMIN(BOTH,[{PRINT/NOPRINT},PUNCH],[VONMISES/SHEAR],[STRCUR/FIB
ER],[CENTER/BILINetc]) = {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
255
CHAPTER 5
Transient Response Analysis
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.
Guidelines and Limitations
The MAXMIN capability was fundamentally designed for the Modal Transient solution sequence
employing a sort2 processing methodology. This placed a certain perspective upon the global/local
orientation of grids and elements during the monitoring operations. Because element or grid orientation
does not change within the time domain, global directions requiring no reorientation were perceived for
the standard processing methodology. The grids were enhanced to support other directions and are
necessary when the sort1 processing requirement was established. The current element orientation
technology is limited in sort1, especially considering the plate elements. Considering the Transient
Solution again, the placement and number of the MAXMIN(DEF) are geared to time domain processing.
When processing Direct Transient (SOL 109), the sort1 path is used. The primary consideration here is
the computational costs for performing the transpose of the “displacement” matrix at the solution set.
Since the MAXMIN operations are performed internally to the SDR2 module, effects of BAR distributed
loads are currently ignored and lump end recovery is currently available. This restriction is also true for
M A X / M I N S U M M A R Y ( ACCELERATION )
POINT CID OPERATION COMPONENT TIME VALUE
1011 0 MIN 1 R2 9.999999E02 8.637509E01
1011 0 2 9.000000E02 5.546898E01
1011 0 3 8.000000E02 2.548527E01
1011 0 4 7.000000E02 2.930702E02
1011 0 5 6.000000E02 2.917157E01
1011 0 ABS 1 R2 0.000000E+00 1.140825E+00
1011 0 2 1.000000E02 1.103539E+00
1011 0 3 2.000000E02 1.021089E+00
1011 0 4 3.000000E02 8.954600E01
1011 0 5 9.999999E02 8.637509E01
M A X / M I N S U M M A R Y ( BAR )
ELEMENT CID OPERATION COMPONENT TIME VALUE RMS
1001 1 MAX 1 SX1B 9.999999E02 2.695622E02 1.362836E02
1001 1 2 SX1B 9.000000E02 2.279591E02
1001 1 3 SX1B 8.000000E02 1.871261E02
1001 1 4 SX1B 7.000000E02 1.481509E02
1001 1 5 SX1B 6.000000E02 1.120520E02
1001 1 MIN 1 SX1A 9.999999E02 6.621534E02 1.362836E02
1001 1 2 SX1A 9.000000E02 5.536992E02
1001 1 3 SX1A 8.000000E02 4.501026E02
1001 1 4 SX1A 7.000000E02 3.533644E02
1001 1 5 SX1A 6.000000E02 2.653448E02
Main Index
MD Nastran Dynamic Analysis User’s Guide
MAX/MIN Output (New Form)
256
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
257
CHAPTER 5
Transient Response Analysis
Example 2: Linear Static Model of a Quarter Plate
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
SUBCASE 1
M A X / M I N S U M M A R Y ( HEXA )
ELEMENT CID OPERATION COMPONENT TIME VALUE GRID
1 0 MAX 1 P1 6.000000E04 3.211954E+02 11
1 0 2 P1 9.999999E04 2.747574E+02 23
1 0 3 P1 7.000000E04 2.726083E+02 11
1 0 4 P1 5.000000E04 2.719700E+02 11
1 0 5 P1 4.000000E04 1.485749E+02 11
1 0 6 P1 8.000000E04 1.052457E+02 11
1 0 7 P1 9.000000E04 9.128755E+01 13
1 0 8 P1 3.000000E04 2.646395E+01 11
1 0 9 P1 2.000000E04 1.693706E+01 13
1 0 10 P1 1.000000E04 4.520476E+00 21
1 0 MIN 1 P2 6.000000E04 3.211504E+02 23
1 0 2 P2 9.999999E04 2.750641E+02 11
1 0 3 P2 7.000000E04 2.724407E+02 23
1 0 4 P2 5.000000E04 2.722952E+02 23
1 0 5 P2 4.000000E04 1.487349E+02 23
1 0 6 P2 8.000000E04 1.050558E+02 23
1 0 7 P2 9.000000E04 9.140302E+01 21
1 0 8 P2 3.000000E04 2.612224E+01 23
1 0 9 P2 2.000000E04 1.695657E+01 21
1 0 10 P2 1.000000E04 4.504962E+00 23
1 0 ABS 1 P1 6.000000E04 3.211954E+02 11
1 0 2 P2 9.999999E04 2.750641E+02 11
1 0 3 P1 7.000000E04 2.726083E+02 11
1 0 4 P2 5.000000E04 2.722952E+02 23
1 0 5 P2 4.000000E04 1.487349E+02 23
1 0 6 P1 8.000000E04 1.052457E+02 11
1 0 7 P2 9.000000E04 9.140302E+01 21
1 0 8 P1 3.000000E04 2.646395E+01 11
1 0 9 P2 2.000000E04 1.695657E+01 21
1 0 10 P1 1.000000E04 4.520476E+00 21
Main Index
MD Nastran Dynamic Analysis User’s Guide
MAX/MIN Output (New Form)
258
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.
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.
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.
M A X / M I N S U M M A R Y ( DISPLACEMENTS )
POINT CID OPERATION COMPONENT SUBCASE VALUE
14 0 MAX 1 T1 1001 1.993413E03
13 0 2 1001 1.532535E03
12 0 3 1001 1.186146E03
11 0 4 1001 1.104745E03
10 0 5 1001 1.092352E03
10 0 MIN 1 T1 1001 1.092352E03
11 0 2 1001 1.104745E03
12 0 3 1001 1.186146E03
13 0 4 1001 1.532535E03
14 0 5 1001 1.993413E03
14 0 MAX 1 T2 1001 1.716866E05
11 0 2 1001 1.911883E05
12 0 3 1001 2.190853E05
10 0 4 1001 3.123875E05
13 0 5 1001 4.483166E05
13 0 MIN 1 T2 1001 4.483166E05
10 0 2 1001 3.123875E05
12 0 3 1001 2.190853E05
11 0 4 1001 1.911883E05
14 0 5 1001 1.716866E05
SUBCASE 1001
D I S P L A C E M E N T V E C T O R
POINT ID. TYPE T1 T2 T3 R1 R2 R3
10 G 1.092352E03 3.123875E05 0.0 0.0 0.0 0.0
11 G 1.104745E03 1.911883E05 0.0 0.0 0.0 0.0
12 G 1.186146E03 2.190853E05 0.0 0.0 0.0 0.0
13 G 1.532535E03 4.483166E05 0.0 0.0 0.0 0.0
14 G 1.993413E03 1.716866E05 0.0 0.0 0.0 0.0
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
MD Nastran Dynamic Analysis User’s Guide
Overview
260
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
261
CHAPTER 6
Damping
Modeling Damping Effects
The four types of damping in MD Nastran are viscous, structural, modal, and nonlinear. Many of the
basic input formats and applications for the damping coefficients in MD Nastran are explained in Finite
Element Input Data (Ch. 2). The following discussions are directed to applications and the special
problems of damping. Damping specifications in MD Nastran are cummulative. Thus viscous, structural,
and modal damping will be applied if specified and supported in the solution sequence
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, . 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 61. Shown is the
force due to a stiffness matrix term, , plotted against the displacement of the point. As the
displacements cycle sinusoidally, the imaginary stiffness, , 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.
G
K
G
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modeling Damping Effects
262
Figure 61 Equivalent Hysteresis Path for Imaginary Stiffness Damping
The basic definition for the steadystate actual displacements, in terms of the complex displacements is
(61)
where is the actual real displacement, is normally a complex variable that we will set to a real
number, is the steady state frequency, and . If a structural damping coefficient,
, is added to the stiffness matrix, the resulting complex forces are in the following form:
(62)
The real part of the forces are
(63)
The incremental work done over a period of time is
(64)
Real
Force
Real
Displacement
u
0
u 0° =
Ku
0
GKu
0
u 90° =
u 180° =
u
r
t ( ) Re u
0
e ( )e
i et
( ) =
u
r
u
0
e e
i x
x ( ) i x ( ) sin + cos =
G
F e ( ) 1 iG + ( )Ku
0
e
i et
=
F
r
Re F ( ) Ku
0
et G et sin – cos ( ) = =
dW F
r
du
r
=
Main Index
263
CHAPTER 6
Damping
where, from Eq. (61)
(65)
Combining Eq. (63) through Eq. (65) and integrating over a full cycle, results in the work
(66)
Evaluating the integral, we obtain work loss per cycle
(67)
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 61 is equivalent to the damping energy
loss, .
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 modebymode basis. These damping factors may be converted to a frequencydependent table,
TABDMP. If the MD Nastran normal mode frequencies are close and the damping factors are small.
The third reasonwhen damping factors are specified by a third partyis 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
(68)
In this equation, is the generalized coordinate of the ith mode.
du
r
u
0
et sin ( )edt – =
W Ku
0
2
et et G et sin – cos ( )e sin – t d
0
2t e
í
=
W tGKu
0
2
=
W
m
i
c
··
i
b
i
c
·
i
k
i
c
i
+ + P
i
=
c
i
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modeling Damping Effects
264
In accordance with modal definitions, the modal viscous damping coefficient may be expressed as:
(69)
where is the undamped vibration frequency (equal to ), expressed in radians per unit time. The
dimensionless input coefficient, , is equal to twice the critical damping ratio of the mode. It is
evaluated by linear interpolation of a userspecified function of frequency, .
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, , 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 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 ( ) or amplification quality factor ( ). 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 pairs on the TABDMP1 entry.
The complete damping matrix for modal transient analysis, , is
(610)
where is a diagonal matrix whose elements are given by Eq. (69), is the matrix of eigenvectors,
and is the matrix of nonmodal damping terms. Note that the damping effects which enter may
well be duplicated by the effects included in so that, in general, the user should be careful when using
both forms simultaneously.
Also, the matrix in Eq. (610), or the equivalent complex stiffness matrix, is generally coupled
so that the efficient uncoupled methods of analysis cannot be used when 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. (68) transitions from
b
i
b
i
g
i
e
i
m
i
=
e
i
k
i
m
i
g
i
g f ( )
b
f g .
C C
c
Q
f, g
B  
B   b
i
  o  
T
B
v
  o   + =
b
i
  o  
B
v
  B
v
 
b
i
 
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.
o  
T
B
v
  o  
B
v
 
Main Index
265
CHAPTER 6
Damping
a damped sinusoid to a nonoscillating, decaying exponential. A solution for the modal equation shows
the value of critical damping, , to be
(611)
The fraction of critical damping, , is calculated from the equation
(612)
Another form of damping specification is the amplification quality factor, , with the definition
(613)
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 62, 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.
Figure 62 Damping Coefficients Versus Modal Frequency
C
c
C
c
2 k
i
m
i
=
.
i
.
i
b
i
C
c
 ÷
b
i
2 k
i
m
i
 =
Q
i
Q
i
1
2.
i

1
g
i
 = =
Modal Frequency
Modal Damping
Constant Viscous Damping
Structural Damping
b e
i
b Ge
i
2
~ ( )
b m
i
g
i
e
i
= ( )
e
i
e
i
b c = ( )
Main Index
MD Nastran Dynamic Analysis User’s Guide
Modeling Damping Effects
266
Note that structural damping will overdamp the system modes above and underdamp system modes
below . The time domain solutions require the specification of 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.
e
i
e
i
e
i
Main Index
267
CHAPTER 6
Damping
Damping Effect on Complex Eigenvalues
Damping characteristics can be demonstrated with a single degree of freedom system using complex
eigenvalue analysis. Two types of damping are considered in this exercise, structural and viscous. The
viscous damping causes the damped natural frequency to be lower than the undamped natural frequency.
The structural damping coefficient (complex stiffness matrix) causes the damped natural frequency to be
higher than the undamped natural frequency. The damping characteristic of the equation of motion is
illustrated with both damping types in the equation. The basic complex eigenvalue equation of motion
for a single degree of freedom is expressed as follows:
(614)
where:
Dividing the above equation by and making some substitutions, the equation becomes:
(615)
where:
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.
(616)
Putting the complex number inside the radical into polar format, yields:
(617)
= mass
= viscous damping
= structural damping coefficient
= stiffness
= operator notation for
=
= displacement
=
=
mp
2
bp 1 i g + ( ) + +  u 0 =
m
b
g
k
p d dt
i 1 –
u
m
p
2
2¸ep 1 ig + ( )e
2
+ +  u 0 =
e
2
k m
2¸e b m
¸
p e ¸ ¸
2
1 ig + ( ) – ± –   =
p e ¸ – i 1 ¸
2
– ( )
2
g
2
e
i u
+
4
± =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Damping Effect on Complex Eigenvalues
268
where
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:
(618)
Using the above equation one can compute the complex eigenvalues for any value of , and to
obtain the results shown in Figure 63 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.
Figure 63 Complex Eigenvalues Comparison for Theory and MD Nastran
If one looks at the two cases as separate solutions, i. e., either set then the first equation becomes:
(619)
and the roots are complex conjugate pairs. With , the equation becomes:
u tan
1 – g
1 ¸
2
–

\ .
 
=
p e ¸ – 1 ¸
2
– ( )
2
g
2
+
4
u 2 ( ) sin i u 2 ( ) cos –   ±
¹ )
´ `
¦ ¹
=
g ¸ e
g 0 =
mp
2
bp k + +  u 0 =
¸ 0 =
Main Index
269
CHAPTER 6
Damping
(620)
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 and vary the viscous damping value, the
damping comparison is made in Figure 64. Again the lines represent the theoretical solution and the
solid symbols are results from MD Nastran.
Figure 64 Effect of Mixed Structural and Viscous Damping
mp
2
1 ig + ( )k +  u 0 =
g 0.02 =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Structural Damping for Shell and Bush Elements
270
Structural Damping for Shell and Bush Elements
Introduction
The structural damping applied to shell elements is supported by the structural damping coefficients, on
the MID1, MID2, MID3, and MID4 fields of the PSHELL entry.
The CBUSH element supports separate frequencydependent structural damping coefficients for all six
degrees of freedom.
The CBUSH element is a generalized springdamper element. In frequency response, the structural
damping is associated with the imaginary portion of the complex stiffness matrix while viscous damping
is associated with the portion.
Matrices
PSHELL
If PARAM,SHLDAMP,DIFF is present in the run, a structural damping matrix is formed for each
shell element based on a relationship of the following form:
(621)
CBUSH
If both stiffness K and structural damping are specified, each with its own frequency dependent
tables, matrix terms of the following form are created:
(622)
If stiffness K with frequency dependent tables and non frequency dependent structural damping are
specified, matrix terms of the following form are created:
(623)
If nonfrequency dependent stiffness K with frequency dependent structural damping are specified,
matrix terms of the following form are created:
(624)
ie B ·
K
4
K
4
ge
mi d1
membrane stiffness · ×
+ ge
mi d2
bending stiffness
+ ge
mi d3
transverse stiffness
+ ge
mi d4
combined stiffness · ×
· ×
· ×
=
ge
K
j
f
i ge
j
f
K
j
f
· · +
ge
K
j
f
i ge
j
0
K
j
f
· · +
ge
K
j
0
i ge
j
f
K
j
0
· · +
Main Index
271
CHAPTER 6
Damping
In the above three expressions, the superscript denotes frequency dependent and the superscript 0
denotes a nominal value. The subscript implies the jth degree of freedom of the CBUSH element. The
real term goes into the element stiffness matrix and the imaginary term goes into the element 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 matrices will have appropriate terms accounting for additional structural damping
coefficients
Guidelines and Limitations
The following suggestions are associated with this capability:
1. Multiple values on the PBUSH or the use of PARAM,SHLDAMP,DIFF causes additional
element stiffness matrix generations which will increase run time.
2. With the PARAM,SHLDAMP,DIFF option, relatively large values for the structural damping
coefficient associated with MID4 can cause structural instability in transient dynamics.
3. For nonlinear shell elements, PARAM,SHLDAMP,DIFF is ignored and the MID1 associated
value of structural damping coefficient is used.
f
j
K
4
K
4
ge
i
Main Index
MD Nastran Dynamic Analysis User’s Guide
Geometric and Material Nonlinear Spring and Damper Element
272
Geometric and Material Nonlinear Spring and Damper
Element
Introduction
The element CBUSH1D is a one dimensional version of the CBUSH element (without the rigid
offsets). The element supports large displacements.
The CBUSH1D element has axial stiffness and axial damping. The element includes the effects of
large deformation. The elastic forces and the damping forces follow the deformation of the element
axis if there is no element coordinate system defined. The forces stay fixed in the xdirection of the
element coordinate system if the user defines such a system. Arbitrary nonlinear force
displacement and forcevelocity functions are defined with tables and equations. A special input
format is provided to model shock absorbers.
An element where damping follows large deformation. Arbitrary force deflection functions can
now be modeled conveniently. When the same components of two grid points must be connected,
we recommend using forcedeflection functions with the CBUSH1D element instead of using
NOLINi entries. The BUSH1D element produces tangent stiffness and tangent damping matrices,
whereas the NOLINi entries do not produce tangent matrices. Therefore, CBUSH1D elements are
expected to converge better than NOLINi forces.
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
Defines the connectivity of a onedimensional spring and viscous damper element.
Format
1 2 3 4 5 6 7 8 9 10
CBUSH1D
EID PID GA GB CID
Field Contents Default Values
EID Element identification number. (0 < Integer < 100,000,000) Required
PID Property identification number of a PBUSH1D entry. (Integer
> 0)
EID
GA Grid point id of first grid. Required
Main Index
273
CHAPTER 6
Damping
PBUSH1D
Defines linear and nonlinear properties of a onedimensional spring and damper element.
Format
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.
Guidelines and Limitations
The element is available in all solution sequences. In static and normal modes solution sequences, the
damping is ignored.
GB Grid point id of second grid blank
CID Coordinate system id. (Integer > 0) blank
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
Field Contents Default
PID Property identification number. (Integer > 0). Required
K Stiffness. (Real > 0). See Remark 1.
C Viscous damping. (Real > 0). See Remarks 1. and 2.
M Total mass of the element. (Real > 0). Blank
SA Stress recovery coefficient [1/area]. (Real > 0). Blank
SE Strain recovery coefficient [1/length]. (Real > 0). Blank
Field Contents Default Values
Main Index
MD Nastran Dynamic Analysis User’s Guide
Geometric and Material Nonlinear Spring and Damper Element
274
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 forcedisplacement and forcevelocity 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, roundoff 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 65. 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 61, see Install_dir/mdxxxx/doc/dynamics/ar29src1.dat for the complete file.
Main Index
275
CHAPTER 6
Damping
Figure 65 Swinging Pendulum
Listing 61 MD Nastran Input File for the Swinging Pendulum
$
$ file: ar29src1.dat
$
$ A swinging pendulum is modeled. The pendulum is loaded with
$ gravity load. The analysis goes through one and a half cycles,
$ starting in horizontal position. Large deformation is turned on.
$
$ The pendulum is modeled with two separate systems.
CROD and CVISC, CROD rotates, CVISC stays in horizontal
$orientation.
CBUSH1D with SPRING and DAMPER, spring and damper rotate.
$
TIME 10 $
SOL 129 $
CEND
$
TITLE= ar29src1 swinging elastic pendulum
SUBTITLE= demonstrate rotating damping
. . . $
BEGIN BULK
$
param, lgdisp, 1
$
$ GEOMETRY
$ . . .
$ CBUSH1D spring and rotating $
$CBUSH1D ,EID ,GA,GB,CID cbush1d, 108, 208, 41, 42 $PBUSH1D,PID,K,C
pbush1d, 208, 1.e+7, 1000.
$
$$$$
$ SOLUTION STRATEGY
$
. . . $ ENDDATA
Main Index
MD Nastran Dynamic Analysis User’s Guide
Geometric and Material Nonlinear Spring and Damper Element
276
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 zdisplacements of the free end are shown in Figure 66.
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 xdirection during the whole motion.
Main Index
277
CHAPTER 6
Damping
Figure 66 Displacement of Swinging Pendulum, Model with Rod/Visc Element
Versus Model with BUSH1D Element
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 66.
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 67. 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 62 see Install_dir/mdxxxx/doc/dynamics/ar29be01.dat for the complete file. The
beam responds with a rotation of about 450 before the motion is damped out. The load and the z
displacement at the tip of the overhang are shown in Figure 34. The example demonstrates that
the CBUSH1D element can undergo large deformation and can simulate complex nonlinear force
deflection laws.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Geometric and Material Nonlinear Spring and Damper Element
278
Figure 67 Simple Model of an Automotive Suspension
Listing 62 MD Nastran Input File for the Automotive Suspension
$
$** * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
$Test of nonlinear shock absorber
$under large deformation
$** * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* * * * *
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
279
CHAPTER 6
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
MD Nastran Dynamic Analysis User’s Guide
Geometric and Material Nonlinear Spring and Damper Element
280
Figure 68 Load and Displacement at Tip C for the Automotive Suspension Model
Displacement at Tip C in the ZDirection
Time [sec]
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
MD Nastran Dynamic Analysis User’s Guide
Overview
282
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 roundoff errors and pseudo rigid body modes. There is also the Lagrange Multiplier technique,
but it is very manual and requires DMAP.
Main Index
283
CHAPTER 7
Enforced Motion
Direct Enforced Motion in Dynamic Analysis
The method called the SPC/SPCD method is regarded as the method of choice for all enforced motion
applications in dynamic analysis. This method of specifying enforced motion permits direct specification
of displacements, velocities, or accelerations via SPC/SPCD data, thus eliminating the need to employ
large mass or Lagrange Multipliers. MD Nastran directly utilizes this enforced motion information in
the equations of motion, partitioning and integrating them (in the case of transient analysis) in
accordance with the type of motion specified.
The enforced motion capability is implemented in direct and modal frequency analysis (Solutions 108
and 111), direct and modal transient analysis (Solutions 109 and 112), aeroelastic response (SOL 146)
and design optimization (Solution 200).
A user parameter called ENFMOTN is used with the SPC/SPCD method in SOLs 108, 109, 111, 112,
146 and 200. This parameter controls how the results of the analysis are output.
The total solution of a dynamic enforced motion analysis using SPC/SPCD can be regarded as a
combination of a static enforced motion solution (similar to what is done in SOL 101) and a dynamic
enforced motion solution that is relative to this staticbased solution. If the value of the ENFMOTN
parameter is specified as TOTAL or ABS, then the output results of the analysis represent the TOTAL
solution of the dynamic analysis which includes both the staticbased solution and the dynamic solution
that is relative to the staticbased solution. If the value of ENFMOTN is specified as REL, then the
output results represent the dynamic solution RELATIVE to the staticbased solution. In the case of
modal dynamic analysis, this latter scenario is equivalent to employing the large mass approach and
excluding the low frequency mode(s) resulting from the large mass usage. The default value for the
ENFMOTN parameter is TOTAL (or ABS).
It should be noted that, in general, the staticbased solution mentioned above may not be a stressfree
solution. Therefore, in general, the usage of PARAM,ENFMOTN,TOTAL (or ABS) and
PARAM,ENFMOTN,REL in an SPC/SPCD enforced motion analysis may give different results not only
for displacements, but also for stresses.
In the special case when the only constraints are those due to the SPC/SPCD enforced motion, the static
based solution will represent a stress free solution and may be regarded as the base motion of the model.
In this case, the usage of PARAM,ENFMOTN,TOTAL (or ABS) and PARAM,ENFMOTN,REL will
give different displacements, but the same stresses. Further, in modal dynamic analysis for this case, the
results using PARAM,ENFMOTN,TOTAL (or ABS) will give the same results as those obtained by
using the large mass approach with all rigid body modes included and the results using
PARAM,ENFMOTN,REL will give the same results as those obtained by using the large mass approach
with all rigid body modes excluded.
Frequency Response Analysis
The fundamental assumption in frequency response analysis is that transient effects are negligible, and
the timedependent nature of the loads can be expressed purely in terms of harmonic forcing functions.
The forced response of the structure to these harmonic loads thus occurs at the same frequency, and in
proportion to the magnitude of the applied loads.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Direct Enforced Motion in Dynamic Analysis
284
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:
will necessarily give rise to the inphase displacement
with the corresponding velocity and acceleration
and
yielding the familiar equations of frequency response. Here they are written after multipoint constraint
partitioning operations have been performed leaving just the free (fset) and constrained (sset) degrees
offreedom
where are the external loads applied to the sset, and are the corresponding forces of constraint.
If the constraints specify zero motion , the solution for the free degreesoffreedom may be
obtained directly from the upper part of the above equation as:
(71)
and the corresponding constraint forces from the lower part as:
(72)
If enforced displacements, velocities, or accelerations are applied, and the free degreesof
freedom, again from are:
(73)
P t ( ) P e ( ) e
i et
· =
u t ( ) U e ( ) e
i et
· =
u
·
t ( ) i = eU e ( ) e
i et
·
u
··
t ( ) e
2
– U e ( ) = e
i et
·
e
2
M
f f
M
f s
M
sf
M
ss
– ie
B
f f
B
f s
B
sf
B
ss
K
f f
K
f s
K
sf
K
ss
+ +
\ .


 
U
f
U
s
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
P
f
P
s
q
s
+
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
P
s
q
s
U
s
0 { } = ( )
e
2
M
f f
– ieB
f f
K
f f
+ + ( )U
f
P
f
=
q
s
P
s
– e
2
M
sf
– ieB
sf
K
sf
+ + ( )U
f
+ =
U
s
0 { } =
e
2
M
f f
– ieB
f f
K
f f
+ + ( )U
f
P
f
e
2
M
f s
– ieB
f s
K
f s
+ + ( )U
s
– =
Main Index
285
CHAPTER 7
Enforced Motion
With corresponding constraint forces:
(74)
Comparing Eqs. (73) and (74) with Eqs. (71) and (72) shows that the enforced motion modifies the
forces applied to the fset degreesoffreedom. Note also that, were enforced velocities to be applied
instead, the resultant applied displacements would differ by a factor of , and by in the
case of enforced accelerations.
Transient Response Analysis
Because it involves displacement and its higherorder time derivatives, the solution of the transient
dynamic equations of motion must be performed in a stepwise integral fashion for every time step, t. As
with frequency response, the implications for constrained motion are that any enforced displacement,
velocity or acceleration uniquely defines the other two quantities for that degreeoffreedom, with the
necessary forces of constraint determined by solution of the governing equations of motion at the
particular time step(s) of interest. Unlike frequency response, however, higherorder displacement
derivatives for the enforced degreesoffreedom must be determined by finite difference, with lower
order quantities (for example, displacements and velocities for applied accelerations) determined by
numerical integration.
After multipoint constraint partitioning, the transient dynamic equations of motion, expressed in terms
of the free (fset) and constrained (sset) degreesoffreedom are:
(75)
In the case of zero constrained motion, and the solution for the free degreesof
freedom is available directly from Eq. (75) as:
(76)
with corresponding forces of constraint from:
(77)
In the case of enforced motion, , , and the solution for the free degreesoffreedom, again
from Eq. (75) is then:
(78)
with constraint forces,
q
s
P
s
– = e
2
M
sf
– ieB
sf
K
sf
+ + ( )U
f
e
2
M
ss
– ieB
ss
K
ss
+ + ( )U
s
+ +
1 ie ( ) ( ) 1 e
2
( ) –
M
f f
M
f s
M
sf
M
ss
u
··
f
u
··
s
¹ )
´ `
¦ ¹ B
f f
B
f s
B
sf
B
ss
u
·
f
u
·
s
¹ )
´ `
¦ ¹ K
f f
K
f s
K
sf
K
ss
u
f
u
s
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
+ +
P
f
t ( )
P
s
t ( ) q
s
t ( ) +
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
u
s
u
·
s
u
··
s
0 { } = = =
M
f f
u
··
f
B
f f
u
·
f
K
f f
u
f
+ + P
f
t ( ) =
q
s
t ( ) P
s
t ( ) – = M
sf
u
··
f
B
sf
u
·
f
K
sf
u
f
+ + ( ) +
u
s
u
·
s
u
··
s
0 { } =
M
f f
u
··
f
B
f f
u
·
f
K
f f
u
f
+ + P
f
t ( ) = M
sf
u
··
s
B
sf
u
·
s
K
f s
u
s
+ + ( ) –
Main Index
MD Nastran Dynamic Analysis User’s Guide
Direct Enforced Motion in Dynamic Analysis
286
(79)
As with frequency response, note that the effect of enforced motion is to modify the loads on the fset,
and the sset forces of constraint.
User Interface and Specification of Enforced Motion
Enforced acceleration or enforced velocity usage in transient analysis via SPC/SPCD specification
requires integration to compute the corresponding enforced velocities and/or displacements. This
integration involves the use of initial conditions. The user can specify initial displacements for enforced
DOFs in the case of enforced velocity usage and initial displacements as well as initial velocities for
enforced DOFs in the case of enforced acceleration usage. The initial displacement and velocity values
are specified via corresponding factors in the US0 and VS0 fields on the TLOAD1 and TLOAD2 Bulk Data
entries.
It should be noted here that the initial conditions for the enforced DOFs mentioned here are distinct from,
and may be used in conjunction with, the initial conditions for independent DOFs that may be specified
by a TIC Bulk Data entry.
Enforced motion in early versions of MD Nastran used a nonzero value in the TYPE field (field 5) on the
TLOAD1 and TLOAD2 entries to specify whether an enforced displacement, velocity, or acceleration
were to be applied using the large mass method. Although these integer values are still allowed, the field
has now been generalized to permit character input, and a similar TYPE field has been included in field
8 of the RLOAD1 and RLOAD2 entries to support enforced motion in frequency response analysis. The
new TYPE field for these dynamic loading/enforced motion entries now reads as follows:
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 SPCtype 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 SPCtype data in the case of enforced
Value Excitation Function
0, or LOAD Applied load (Default)
1, or DISP Enforced displacement using large mass or SPC/SPCD data
2, or VELO Enforced velocity using large mass or SPC/SPCD data
3, or ACCE Enforced acceleration using large mass or SPC/SPCD data
q
s
t ( ) P
s
t ( ) – M
sf
M
ss
 
u
··
f
u
··
s
¹ )
´ `
¦ ¹
+ = B
sf
B
ss
 
u
·
f
u
·
s
¹ )
´ `
¦ ¹
K
sf
K
ss
 
u
f
u
s
¹ )
´ `
¦ ¹
+ +
Main Index
287
CHAPTER 7
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.
Residual Vector (Modal Augmentation)
When enforced motion is employed in modal dynamic response analysis, the use of Modal
Augmentation vectors is highly recommended in order to improve solution accuracy. Without these
vectors, the obtained results may be unacceptable. Qualitatively, the enforced motion shapes, or
“modes”, may not necessarily be represented by the component modes of the structure, leading to a
truncation of the enforced motion shapes during projection to the modal space. These shapes can be
regained by the use of Modal Augmentation Vectors, and can be triggered simply enough by specifying
PARAM,RESVEC,YES in the Bulk Data Section also see Residual Vector (Ch. 12). The necessary static
loads are automatically derived by MD Nastran from the enforced motion data, and no additional loads
specification is necessary.
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 SPCtype 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.
Procedure for Using the Enhanced Enforced Motion Feature
In summary, the procedure for applying enforced motion in a dynamic analysis:
1. Specifies the appropriate type of enforced motion to be applied via the TYPE field in
RLOAD1/RLOAD2 (field 8) or TLOAD1/TLOAD2 (field 5) Bulk Data entry, as appropriate.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Direct Enforced Motion in Dynamic Analysis
288
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.
Direct Enforced Motion Usage in Frequency Response Analysis
.
.
.
SOL 108
CEND
TITLE = SPECIFICATION OF ENFORCED DISPLACEMENT
$ THE FOLLOWING REQUEST SELECTS SPC1 ENTRY 1000
SPC = 1000
$ THE FOLLOWING REQUEST SELECTS RLOAD1 ENTRY 100
DLOAD = 100
.
BEGIN BULK
$ THE FOLLOWING SPC1 ENTRY IS SELECTED
$ VIA CASE CONTROL
SPC1,1000,1,70,80,90
$ THE TYPE FIELD IN THE FOLLOWING RLOAD1 ENTRY
$ SPECIFIES ENFORCED DISPLACEMENT. ACCORDINGLY,
$ EXCITEID ID 200 HERE REFERS TO SPCD DATA.
$RLOAD1,SID,EXCITEID,DELAY,DPHASE,TC,TD,TYPE
RLOAD1,100,200, , , 300, ,DISP
$ THE FOLLOWING SPCD ENTRY IS SELECTED BY THE
$ RLOAD1 DATA ABOVE
$ NOTE THAT GRID POINT 80  COMPONENT 1 SPECIFIED
$ HERE ALSO APPEARS IN THE SELECTED SPC1 DATA.
$ NOTE ALSO THAT THE VALUE OF 1.1 SPECIFIED HERE
$ REPRESENTS DISPLACEMENT
$SPCD,SID,G,C,VALUE
SPCD,200,80,1,1.1
.
.
.
Direct Enforced Motion Usage in Transient Response Analysis
.
.
.
SOL 109
CEND
TITLE = SPECIFICATION OF ENFORCED ACCELERATION
$ THE FOLLOWING REQUEST SELECTS SPC1 ENTRY 1000
SPC = 1000
$ THE FOLLOWING REQUEST SELECTS TLOAD1 ENTRY 100
DLOAD = 100
Main Index
289
CHAPTER 7
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 71) is subjected to a 1G (9.81m/sec
2
)
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 71 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/sec
2
in the xdirection of grid point 1. The degreeoffreedom with enforced
motion specified must also be constrained, once again xdirection 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
MD Nastran Dynamic Analysis User’s Guide
Direct Enforced Motion in Dynamic Analysis
290
Figure 71 Tower for Enforced Motion.
Listing 71 Partial Input for Enforced Motion
$
$ FILE bd07enf.dat
$
INIT MASTER(S)
SOL 111
DIAG 8
CEND
SET 99=1,547
DISPLACEMENT(SORT1,PHASE) = 99
SPCFORCE(SORT1,PHASE) = 99
ACCELERATION(SORT1,PHASE) = 99
DLOAD = 101
FREQUENCY = 3
METHOD = 1
SDAMPING = 2
SPC=30
BEGIN BULK
param,post,0
param,resvec,yes
1
547
Main Index
291
CHAPTER 7
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 72 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/sec
2
, which is what one would expect, since that is the actual enforced acceleration
value.
Listing 72 Partial Output from Enforced Motion
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)
POINT ID. TYPE T1 T2 T3
R1 R2 R3
0 1 G 9.810000E+00 0.0 0.0
0.0 0.0 0.0
0.0 0.0 0.0
0.0 0.0 0.0
0 547 G 3.791660E+01 8.261231E02 2.580789E+00 7.377076E
03 1.751959E+00 1.085166E01
343.6485 326.0110 338.0020
348.9515 337.9252 338.4973
Main Index
MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
292
Enforced Motion with Loads
Three distinct methods are used in MD Nastran for dynamic enforced motions: the Large Mass/Spring
approach, the Inertial Loads method, and the Lagrange Multiplier (LMT) method). The LMT will only
discuss the theory.
Large Mass/Spring Method
The Large Mass/Spring approach is actually a modeling technique in which the user places an element
with a large mass or stiffness at the points of known acceleration or displacement. In effect, this large
element acts as a constraint on the connected point. The user then supplies a corresponding large force
via RLOADi or TLOADi inputs to produce the desired motion. If the added element is sufficiently stiff
or heavy, the reaction forces from the actual structure will not affect the input motions. Applications of
both large mass and large spring techniques are discussed below.
Using a large mass to enforce boundary motions is a standard practice in MD Nastran. It works well on
a simply supported structure when a single grid point is excited by a well defined acceleration input. A
typical example is the earthquake analysis of a tall building where a single base input is assumed.
However, the method may be abused when other types of boundary conditions are attempted.
Some additional considerations for the analyst are listed below:
1. If enforced motion is applied to a redundant set of boundary points, a danger exists that the large
masses (or springs) may create fictitious forces and stresses in the structure. This occurs when
the enforced displacements inputs are not precisely synchronized. Also, in a modal formulation,
the extra masses will cause fictitious low frequency modes to occur. The recommended
procedure is to connect the redundant points with RBE (rigid) elements to prevent their
independent motions. The Lagrange Multiplier method is much better, but will not correct for
errors in the loading functions and is hard to use.
2. Small errors in the loading history may cause large errors in the structural response. When using
enforced accelerations in a transient solution, a small bias in the inputs (from instrumentation or
processing) may cause a large spurious drift where the structure displaces a large amount as a rigid
body. Solutions to remove the drift are as follows:
• Supply a corrective load function, obtained from an initial run, to cancel the measured drift;
• Add dampers and springs in parallel with mass and tuned to filter the input signal; or
• Use the modal method and drop the zerofrequency modes. (See item 4 below.)
An example of a mechanical filter is shown in the sketch below. A large mass, , is attached
to the base of a structure to allow enforced accelerations. For control of the spurious
displacements and velocities, attach a scalar spring, , and damper element, , between the
mass and ground for each direction. If the first modal frequency is then set for
a filter frequency approximately one tenth of the first vibration mode frequency, . Set the
damping coefficient to a value near critical to eliminate spurious oscillations.
M
0
K B
f
1
K 0.4 Mf
1
2
=
f
1
Main Index
293
CHAPTER 7
Enforced Motion
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 degreesoffreedom. 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 degreesoffreedom 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
Structure
Seismic Mass
B K
Filter
M
0
Main Index
MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
294
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.
The Inertial Loads Approach
By changing to a moving coordinate system, the reference accelerations may be converted to inertial
loads. If the dynamic problem involves a free body connected to a single rigid base, the GRAV and
dynamic load inputs may be used to supply forces proportional to the base acceleration. The solution
will be defined by the displacements measured relative to the fixed base.
Approach
The basic approach is shown in the sketch below:
We assume that the motions of the rigid base are known and the structure is fixed only to the base. Define
the displacement vector, , as a sum of the known base motion, and the relative displacements,
by the equation
(710)
where 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
(711)
Structure
Base
ou
u
u
0
u u
0
ou
u { } ou { } D   u
0
{ } + =
D  
K   D   0 =
and B   D   0 =
Main Index
295
CHAPTER 7
Enforced Motion
where is the stiffness matrix and 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
(712)
Substituting Eqs. (710) and (711) into Eq. (712), we obtain
(713)
The solution matrices have not changed, but the solution, , must be zero at the base attachment points.
The righthand term in Eq. (713) 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 timedependent gravity loads. By
definition, the GRAV input generates load vectors of the form
(714)
where is the gravity vector. Then if were time dependent, it could be used to replace in Eq. (713)
if
(715)
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.
K   B  
M   u
··
{ } B   u
·
{ } K   u { } + + P { } =
M   ou
··
{ } B   ou
·
{ } K   ou { } + + P { } M   D   u
··
0
{ } – =
ou
P
g
{ } M   D   g { } =
g g u
0
g t ( ) { } u
··
0
t ( )   – =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
296
3. The method is compatible with superelement and nonlinear solution sequences.
4. No special sets or SUPORT data is necessary.
5. The main drawback is that the output accelerations are also relative and will not correlate to
measured accelerations.
Lagrange Multiplier Technique
The Lagrange Multiplier Technique (LMT) is a special option for processing constraints in matrix
solutions. It can be used for all constraints and reduction methods such as SPC, MPC, and rigid elements.
LMT can also be used for enforced motion analysis in dynamic analysis. Unlike the approximate large
mass approach, the LMT is an exact method. The LMT requires adding extra degreesoffreedom to the
matrix solution that are used as force variables for the constraint functions. Coefficients are added to the
matrices for the equations that couple the constrained displacement variables to the points at which
enforced motion are applied.
The LMT produces indefinite system matrices that require special resequencing of variables for
numerical stability by either the user or by a special solver. The sparse matrix includes an automatic row
and column interchange capability that allows reliable and efficient solution of indefinite matrices
without the need for the user to sequence variables for numerical stability. This solver is available in the
transient and frequency response solver modules.
However, the implementation requires the use of DMAP, making it hard to use, so only the theoretical
discussion will be included in this guide.
Compare Condensation and Augmentation Solution Methods
A small model is used here to contrast the conventional method of constraint elimination and the LMT
method of constraint augmentation. Consider the statics model shown:
Given this model, static equations of motion are as follows:
(716)
k
u
1
u
2
enforced motion Y
1
applied
force P
2
applied
m
1
m
2
K   u { } P { } q { } + =
Main Index
297
CHAPTER 7
Enforced Motion
(717)
contains the unknown constraint forces required to cause the enforced motion and react to the
applied force. is the known enforced displacement, and 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 is moved to the left side,
(718)
This row and a corresponding column are added to the system matrix to produce
(719)
In the equation , 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 . These null terms are an indication of an indefinite matrix.
For a conventional modal solution, the unconstrained equation for eigensolution is
(720)
When the constraint reduction process is used, the first row and column are eliminated. The resulting
equation for the eigenvalue is then
(721)
Alternatively, using the LMT process, the system is augmented with the constraint variables, with the
resulting determinant
(722)
u
1
Y
1
=
q { }
Y
1
P
2
q
1
0 1 0
q
1
–
u
1
u
2
Y
1
=
K  
0 1 0
1 k k –
0 k – k
, P { }
Y
1
0
P
2 ¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
, u { }
q
1
–
u
1
u
2
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
= = =
K   u { } P { } =
K  
k ìm
1
– k –
k – k ìm
2
–
u
1
u
2
¹ )
´ `
¦ ¹
0 { } =
ì
ì k m
2
=
0 1 0
1 k ìm
1
– k –
0 k – k ìm
2
–
0 =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
298
When this determinant is evaluated, the only nonnull product results from the (12)(21)(33) terms,
resulting in the same expression, .
Three counterintuitive consequences are as follows:
1. does not enter into the determinant at all, even though it is present in the matrix. In a constraint
reduction process, by contrast, any mass term that appears in the matrix used for eigensolution
affects the eigenvalues.
2. Although the matrix has three rows and columns and two mass terms on its diagonal, only one
value of is defined. One might expect that this system would have two or three values for ,
but it does not. Other values of beyond this single value are neither zero nor infinity nor any
other value. They are simply undefined.
3. The results from a condensation solution and an augmentation solution are identical. This is a
general rule.
The resulting eigenvector is
(723)
Where is the constrained eigenvector for the displacements and is the reaction force. Note the scale
factor of the whole eigenvector may be affected by 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 is an
enforced displacement, the generalized modal forces are
(724)
The result of applying an enforced displacement on the LMT coordinates, , will be a modal force in
the opposite of the direction of motion. The results of a modal solution will be relative to . See below
for a solution that provides total displacements.
Matrix Theory for Dynamic Response
Notation
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 , , and
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 bset, and have the
following meanings:
New Unique Sets for Enforced Motion
ì k m
2
=
m
1
ì ì
ì
o { }
r –
o
c1
o
c2 ¹ )
¦ ¦
´ `
¦ ¦
¦ ¹ k
0
1.0
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
= =
o
c
r
r
Y
1
P
c
oc
T
P
g
r
T
Y
1
– P
2 k –
Y
1
= =
Y
1
Y
1
u
1
u
2
u
3
Main Index
299
CHAPTER 7
Enforced Motion
The Unique Sets
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
The statement of the frequency response problem is
(725)
where:
(726)
, , and 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 sset constraints are eliminated
along with the oset. The set is then the only constraint set in the dset.
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 frequencyvarying enforced
motions. The value of the SETNAME entry on the USET entries determines the type of input. A value
of indicates an enforced displacement type. A value of indicates an enforced velocity type, and a
value of indicates an enforced acceleration type. These dynamic quasiloads are transferred to the
enforced motion variables by the transformation matrix as follows:
(727)
is the load vector output by the FRLG module. In this context it may contain both applied forces
and enforced motions. has a unit term in each row at the column corresponding to the set point
where enforced motion is applied. The enforced motions are removed from the displacement variables
by the transformation matrix ,
Enforced displacement set
Enforced velocity set
Enforced acceleration set
u
1
u
2
u
3
A
dd
e ( )   u
d
{ } P
d
e ( ) { } =
A
dd
e ( )   K ieB e
2
M – +  
dd
=
K   B   M   e
e
u
3
u
1
u
2
u
3
T
4p
 
P
4
e ( )   T
4p
  P
p
e ( )   =
P
p
 
T
4p
  u
i
T
pp
x
 
Main Index
MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
300
(728)
where is an identity matrix except for the rows associated with enforced motion terms, where it has
a zero row. The set and set components are merged to form .
The stiffness and constraint components are separated into and , respectively, where
. This is done in the GMA subDMAP to ensure that the constraints are not
multiplied by the damping factor iG.
The equations for transient response analysis are similar to those for frequency response analysis. The
subDMAPs for the two types of analysis closely parallel each other.
A new set of partitions not actually used in the DMAP implementation is introduced here. This set aids
in understanding the effects of the enforced motion variables. Let
(729)
and
(730)
The 3set contains the constraint force and displacement variables associated with enforced motion.
is the vector of constraint forces and the vector of constraint displacements. Both are partitions of
the vector and is their complement. is the vector of enforced displacements,
additional forces placed on the enforced motion points, and their complement. The enforced
motions of any LMT variables contained in are zero by definition. contains both displacement
and constraint force unknown variables, while contains both known input displacements and
applied forces.
Consider the case of enforced displacement. then has the partitions
(731)
is a subset of the constraint matrix . By inspection, , an identity matrix, and
is null. The form of the equation of motion is then
P
p
e ( )   T
pp
x
  P
p
e ( )   =
T
pp
x
 
p u
4
P
p
e ( )  
K
aa
g
  R
aa
 
K
aad
  K
aa
g
  R
aa
  + =
u
d
{ }
q
3
u
3
u
c
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
P
d
{ }
Y
3
P
3
P
c
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
q
3
{ }
u
3
{ }
u
d
{ } u
c
{ } Y
3
{ } P
3
{ }
P
c
{ }
P
c
{ } u
c
{ }
P
c
{ }
K  
K  
0 R
33
R
3c
K
33
K
3c
Sym. K
cc
=
R
33
  R
4g
  R
33
  I
33
  = R
3c
 
Main Index
301
CHAPTER 7
Enforced Motion
(732)
For this case and have no terms for the set variables. Their partitions are not shown.
Evaluating the first row of this equation, we obtain
(733)
Note that is the force of constraint with a sign change.
Enforced Accelerations
Now consider the case of enforced acceleration. The equation of motion is then
(734)
Again, from the first row,
The input quantity is an acceleration vector because the first partition of 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 . 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, is the force of constraint.
Modal Equations in the Frequency and Time Domains
All constraint equations are initially placed in the stiffness matrix to allow the solution of the eigenvalue
problem when rigid body modes are present. The system eigensolution is obtained from the equation
(735)
where is a natural frequency, is the corresponding eigenvector, and is the number of
eigenvectors calculated. After the equations are reduced to modal variables, the constraints for enforced
motion are moved to the damping or mass matrix for enforced velocity or enforced acceleration,
respectively, using the same user interface as the direct solutions.
M
dd
  u
··
d
{ } B
dd
  u
·
d
{ }
0 I
33
0
I
33
K
33
K
3c
0 K
c3
K
cc
q
3
u
3
u
c
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
+ +
Y
3
P
3
P
c
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
M
dd
  B
dd
  u
3
u
3
{ } Y
3
{ } =
q
3
0 I
33
0
I
33
M
33
M
3c
0 M
c3
M
cc
q
··
3
u
··
3
u
··
c
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
B
dd
  u
·
d
{ } K
dd
{ } u
d
{ } + +
Y
··
3
P
3
P
c
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
u
··
3
{ } Y
··
3
{ } =
Y
··
3
{ } P
d
{ }
Y
3
{ }
q
··
3
K
aa
e
2
M
aa
2
–   o
az
{ } 0 =
e o
az
{ } z
Main Index
MD Nastran Dynamic Analysis User’s Guide
Enforced Motion with Loads
302
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 , where is defined in Eq. (729). is the transformation between the modal
and physical dynamic variables. , the unknown variable vector for modal analysis, has the
components
(736)
is the vector of generalized displacements for the flexible modes. and are the unknown
constraint force variables and displacement variables for the constraint modes. The corresponding
known variables are
(737)
is the modal force vector and and the known enforced motion variables and applied
forces of the enforced motion points.
is obtained from the equation
(738)
where is defined in Eq. (730).
The equation relating and is used to replace in Eq. (725). The resulting equation is
premultiplied by to form the reduced basis equations of motion as follows:
(739)
When the multiplication of Eq. (739) is carried out, the components of the modal force vector expressed
in the partitions of Eq. (737) become:
(740)
(741)
(742)
u
d
{ } T
dh
  u
h
{ } = u
d
{ } T
dh
 
u
h
{ }
u
z
q
3
u
3
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
u
h
{ } =
u
z
{ } q
3
{ } u
3
{ }
P
h
{ }
P
z
Y
3
P
3
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
P
z
{ } Y
3
{ } P
3
{ }
P
h
{ }
P
h
{ } T
dh
 
T
P
d
{ } =
P
d
{ }
u
d
{ } u
h
{ } u
d
{ }
T
dh
T
 
T
dh
 
T
Z
dd
e ( )   T
dh
  u
h
{ } T
dh
 
T
P
d
{ } P
h
{ } = =
P
z
{ } q
3z
 
T
Y
3
{ } o
3z
 
T
P
3
{ } o
cz
 
T
P
c
{ } + + =
u
3
{ } u
3
{ } Y
3
{ } = =
P
3
{ } q
33
 
T
Y
3
{ } P
3
{ } u
c3
 
T
P
c
{ } + + =
Main Index
303
CHAPTER 7
Enforced Motion
The modal forces have components from all three partitions of . is identically .
includes loads applied directly to 3set points , as well as the constraint forces needed to
maintain the motion of the unknown input point variables at the prescribed input motion . (In
this implementation the forces are moved to .)
If there are enforced motion terms, the rows and columns of the constraint terms, and , are
moved to the corresponding stiffness, damping or mass matrices. The load vector 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 hset matrices, requires the use of
the coupled solution algorithms for frequency response and transient analysis. The matrices are sparse,
however, with offdiagonal terms occurring only in the columns associated with enforced motion so that
the effect on solution cost is small.
P
z
{ } P
d
{ } u
3
{ } u
3
{ }
P
3
{ } P
3
{ }
u
3
{ } Y
3
{ }
P
3
{ } Y
3
{ }
u
3
{ } q
3
{ }
P
d
{ }
Main Index
MD Nastran Dynamic Analysis User’s Guide
Large Mass Method  Direct
304
Large Mass Method  Direct
If a very large mass , which is several orders of magnitude larger than the mass of the entire structure,
is connected to a degreeoffreedom and a dynamic load is applied to the same degreeoffreedom, then
the acceleration of the degreeoffreedom, to a close approximation, is as follows:
(743)
In other words, the load that produces a desired acceleration is approximately
(744)
The accuracy of this approximation increases as is made larger in comparison to the mass of the
structure. The only limit for the size of is numeric overflow in the computer. MSC recommends that
the value of be approximately times the mass of the entire structure for an enforced translational
degreeoffreedom and times the mass moment of inertia of the entire structure for a rotational DOF.
The factor 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 on all enforced degreesoffreedom and supplying applied dynamic loads specified by
Eq. (744); that is, the function is input on entries normally used for the input of loads, and the scale
factor 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 (744) is not directly helpful if enforced displacement or enforced velocity is specified rather
than enforced acceleration. However, Eq. (744) can be made serviceable in frequency response analysis
by noting that
(745)
so that
(746)
The added factor ( or ) 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 degreesoffreedom.
m
o
p
u
··
1
m
o
 p =
u
··
p m
o
u
··
=
m
o
m
o
m
o
10
6
10
6
10
6
m
o
u
··
m
o
u
··
ieu
·
e
2
u – = =
p m
o
u
··
iem
o
u
·
e
2
m
o
u – = = =
ie e
2
–
Main Index
305
CHAPTER 7
Enforced Motion
• Apply large masses with CMASSi or CONMi Bulk Data entries to the DOFs where the
motion is enforced. The magnitude of should be approximately times the entire mass of
the structure (or approximately 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. (746)
to each enforced degreeoffreedom.
• 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 degreeoffreedom equal to , , or ,
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
(747)
Enforced velocity and enforced displacement are changed likewise.
You may well ask whether a stiff spring may be used instead of a large mass. In that case the applied
load is
(748)
where 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 rigidbody drift when applying enforced
motion on statically determinate support points. (Rigidbody 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.)
m
o
m
o
10
6
10
6
m
o
u m
o
u
·
m
o
u
··
u
··
p
m
o
WTMASS ·
 =
p k
o
u =
k
o
Main Index
MD Nastran Dynamic Analysis User’s Guide
Large Mass Method  Modal
306
Large Mass Method  Modal
The steps described in the previous section must also be followed when a modal method of response
analysis is used. Also, if the enforced degrees of freedom are not sufficient to suppress all rigidbody
motions, which may be the case for an airplane in flight, additional DOFs that describe the remaining
rigidbody motions can also be entered on the SUPORT entry. Use of the SUPORT entry is discussed in
Real Eigenvalue Analysis (Ch. 3). Note that the use of the SUPORT entry is optional.
The rigidbody 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 rigidbody
modes are included because the rigidbody modes do not contribute to them. Rigidbody modes can be
discarded to remove rigidbody drift.
Rigidbody 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 lowfrequency modes
occur but they are not all rigidbody modes; some are modes that represent the motion of one large mass
relative to the others. These very lowfrequency, nonrigidbody 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 rigidbody modes, then r must be set
below the frequency of the first relative motion mode.
Consider the twodimensional clamped bar in Figure 72. Assume that each end of the bar is subjected
to the same enforced acceleration time history in the ydirection. One way to model the bar is to use two
large masses (one at each end), which are unconstrained in the ydirection. This model provides two very
lowfrequency modes: one that is a rigidbody 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 rigidbody mode, which can be safely discarded if only the
answers relative to the structure are desired.
Main Index
307
CHAPTER 7
Enforced Motion
Figure 72 ClampedClamped Bar Undergoing Enforced Acceleration
Basic Model
Two Large Masses at Each End
One Large Mass and One RBE2 Element
y
x
u
··
t ( )
u
··
t ( )
p t ( )
p t ( )
p t ( )
R
B
E
2 R
B
E
2
Main Index
MD Nastran Dynamic Analysis User’s Guide
User Interface for the Large Mass Method
308
User Interface for the Large Mass Method
There is no special user interface for the large mass method other than to specify a large mass at excitation
DOFs and to specify the large force. For transient response, the type of enforced motion (displacement,
velocity, or acceleration) must be specified on the TLOAD1 entries. The remainder of the input is
identical to that of frequency response (Frequency Response Analysis, 133) or transient response
(Transient Response Analysis, 201) analysis.
The force applied at a point is the product of terms from the DLOAD, DAREA, and TABLEDi entries.
The scaling of the large force is arbitrary; it can be on any one or more of these entries. These entries
follow:
DLOAD Format:
DAREA Format:
TABLED1 Format:
TABLED2 Format:
TABLED3 Format:
TABLED4 Format:
The TABLED4 entry defines a power series and is convenient in frequency response for enforced
constant velocity or displacement.
1 2 3 4 5 6 7 8 9 10
DLOAD
SID S S1 L1 S2 L2 etc.
DAREA
SID P1 C1 A1 P2 C2 A2
TABLED1
TID
x1 y1 x2 y2 x3 y3 etc. ENDT
TABLED2
TID X1
x1 y1 x2 y2 x3 y3 etc.
TABLED3
TID X1 X2
x1 y1 x2 y2 x3 y3 etc.
TABLED4
TID X1 X2 X3 X4
A0 A1 A2 A3 A4 A5 etc.
Main Index
309
CHAPTER 7
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
(749)
where and are input on the DLOAD Bulk Data entry, is input on the DAREA entry, and
and are input on the TABLEDi entries. Note that the “i ” nonsubscript term in the expression
is .
If a DLOAD entry is used to scale RLOAD2 input, the applied force magnitude in terms of MD Nastran
input is
(750)
where and are input on the DLOAD Bulk Data entry, is input on the DAREA entry, and 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:
(751)
Enforced Velocity
Enforced velocity requires a conversion factor
(752)
For constant velocity , it may be easiest to use the RLOAD1 and TABLED4 entries because the
imaginary term of Eq. (749) and the frequencydependent term can be specified directly.
Enforced Displacement
Enforced displacement also requires a conversion factor
(753)
P
i
f ( ) S S
i
A
i
C
i
f ( ) i D
i
f ( ) · +   · ( )
i
¯
· =
S Si A
i
C
i
f ( )
D
i
f ( )
i D
i
f ( ) · 1 –
P
i
f ( ) S S
i
A
i
B
i
f ( ) · · ( )
i
¯
· =
S S
i
A
i
B
i
f ( )
p e ( ) m
o
u
··
e ( ) =
p e ( ) iem
o
u
·
i 2tf ( )m
o
u
·
e ( ) = =
u
·
e ( )
i D f ( ) · 2tf
p e ( ) e
2
m
o
u – 2tf ( ) –
2
m
o
u e ( ) = =
Main Index
MD Nastran Dynamic Analysis User’s Guide
User Interface for the Large Mass Method
310
For constant displacement , it may be easiest to use the TABLED4 entry because the
frequencydependent term 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
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.
1 2 3 4 5 6 7 8 9 10
TLOAD1
SID DAREA DELAY TYPE TID
TLOAD2
SID DAREA DELAY TYPE T1 T2 F P
C B
TYPE 0 (or blank) = Applied force (default)
TYPE 1 = Enforced displacement
TYPE 2 = Enforced velocity
TYPE 3 = Enforced acceleration
u e ( )
2tf ( )
2
–
Main Index
311
CHAPTER 7
Enforced Motion
Examples
This section provides several examples showing the input and output. These examples are:
These examples are described in the sections that follow.
TwoDOF Model
Consider the twoDOF model first introduced in Real Eigenvalue Analysis, 43 and shown below in
Figure 73. For this example, apply a constant magnitude base acceleration of 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 73. See Install_dir/mdxxxx/doc/dynamics/bd07two.dat.
Figure 73 TwoDOF Model with Large Mass
Model Analysis Type Enforced Motion
bd07two Frequency Response Constant Acceleration
bd07bar1 Transient Response Ramp, Acceleration
bd07bar2 Transient Response Ramp, Displacement
bd07bar3 Transient Response Ramp, Displacement,
Discard RigidBody Mode
1.0 m/sec
2
Large
Mass
Grid Point 1
Grid Point 2
Grid Point 3
y
x
m
1
0.1 kg =
m
2
10.0 kg =
m
3
10
7
kg =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
312
Listing 73 Input File for Enforced Constant Acceleration
$ FILE bd07two.dat
$
$ TWODOF SYSTEM
$ CHAPTER 7, ENFORCED MOTION
$
TIME 5
SOL 111
CEND
TITLE = TWODOF SYSTEM
SUBTITLE = MODAL FREQUENCY RESPONSE
LABEL = ENFORCED CONSTANT ACCELERATION MAGNITUDE
$
$ SPECIFY MODAL EXTRACTION
METHOD = 10
$
$ SPECIFY DYNAMIC INPUT
DLOAD = 999
FREQ = 888
SDAMPING = 777
$
$ SELECT OUTPUT
DISPLACEMENT(PHASE,PLOT) = ALL
ACCELERATION(PHASE,PLOT) = ALL
$
$ XYPLOTS
$
... XY plot commands ...
$
BEGIN BULK
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10..
...$
$
$ LARGE MASS AT BASE GRID POINT
CONM2 999 3 1.0E7
$
$ LOAD DEFINITION (INCLUDES SCALE FACTORS FOR ENFORCED ACCELERATION)
$DLOAD SID S S1 RLOAD1
DLOAD 999 1.0E7 1.0 998
$RLOAD1 SID DAREA TC
RLOAD1 998 997 901
$DAREA SID P1 C1 A1
DAREA 997 3 2 1.0
$TABLED4 TID X1 X2 X3 X4 +TAB4
TABLED4 901 0. 1. 0. 100. +TAB901
$+TAB4 A0 A1 A2 A3 A4 A5
+TAB901 1.0 ENDT
$
$ MODAL EXTRACTION
$EIGRL SID V1 V2 ND MSGLVL
EIGRL 10 1. 30. 0
$
$ FREQUENCY RANGE 210 HZ
$FREQ1 SID F1 DF NDF
FREQ1 888 2. 0.05 160
$
$ MODAL DAMPING OF 5% CRITICAL
$TABDMP1 TID TYPE
Main Index
313
CHAPTER 7
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 74 shows the XY 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 ; there is a very slight variation between 0.9999 and 1.0000 due to the large mass
approximation.
1.0 m/sec
2
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
314
Figure 74 Displacements and Accelerations for the TwoDOF Model
Main Index
315
CHAPTER 7
Enforced Motion
Figure 75 Displacements and Accelerations for the TwoDOF Model (continued)
This model was analyzed with several values of large mass. Table 71 shows the results. Note that the
model with the 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 is a good value to use for this
model.
10
6
10
6
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
316
This model can also be changed to apply constant velocity or constant displacement at its base.
Listing 74 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.
Listing 74 Bulk Data Entries for Enforced Constant Motion
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......10..
...$
$
$ ENTRIES FOR ENFORCED MOTION
$
$ LARGE MASS AT BASE GRID POINT
CONM2 999 3 1.0E7
$
$ LOAD DEFINITION (INCLUDES SCALE FACTORS FOR ENFORCED ACCELERATION)
$
$DLOAD SID S S1 RLOAD1
DLOAD 999 1.0E7 1.0 998
$DAREA SID P1 C1 A1
DAREA 997 3 2 1.0
$
$ UNIQUE ENTRIES FOR ENFORCED CONSTANT ACCELERATION MAGNITUDE
$
$RLOAD1 SID DAREA TC
$RLOAD1 998 997 901
Table 71 Models with Different Large Mass Ratios
Ratio of Large
Mass to
Structure
Natural
Frequencies*
(Hz)
*Resonant frequencies for the constrained model are 4.7876 and 5.2909 Hz.
Response Peaks (m/sec
2
)
10
2
10
9
4.8011
5.3025
52.0552 6.5531 1.0335 0.9524
10
4
10
10
4.7877
5.2910
52.2823 6.7641 1.0003 0.9995
10
6
0.0
4.7876
5.2909
52.2836 6.7661 1.0000 0.9999
10
8
0.0
4.7876
5.2909
52.2836 6.7662 1.0000 1.0000
X
··
1
max
X
··
2
max
X
··
3
max
X
··
3
min
Main Index
317
CHAPTER 7
Enforced Motion
$TABLED4 TID X1 X2 X3 X4 +TAB4
$+TAB4 A0 A1 A2 A3 A4 A5
TABLED4 901 0. 1. 0. 100. +TAB901
+TAB901 1.0 ENDT
$
$ UNIQUE ENTRIES FOR ENFORCED CONSTANT VELOCITY MAGNITUDE
$
$
$RLOAD1 SID DAREA TD
$RLOAD1 998 997 902
$TABLED4 TID X1 X2 X3 X4 +TAB4
TABLED4 902 0. 1. 0. 100. +TAB902
$+TAB4 A0 A1 A2 A3 A4 A5
+TAB902 0.0 6.283185ENDT
$
$ UNIQUE ENTRIES FOR ENFORCED CONSTANT DISPLACEMENT MAGNITUDE
$
$
$RLOAD1 SID DAREA TC
RLOAD1 998 997 903
$TABLED4 TID X1 X2 X3 X4 +TAB4
TABLED4 903 0. 1. 0. 100. +TAB903
$+TAB4 A0 A1 A2 A3 A4 A5
+TAB903 0.0 0.0 39.4784ENDT
Each input utilizes the TABLED4 entry. The TABLED4 entry uses the algorithm
(754)
where is input to the table, is returned, and is the degree of the power series. When ,
is used for ; when , is used for . 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
(755)
where:
Therefore, these terms define a constant (1.0 in this case).
= 1.0
= 0.0
= 1.0
Y A
i
x X1 –
X2

\ .
 
i
i 0 =
N
¯
=
x Y N x X3 < X3
x x X4 > X4 x
A0 A1
x X1 –
X2

\ .
 
+
A0
X1
X2
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
318
Constant velocity involves a scale factor that is directly proportional to circular frequency ( ). The
power series for this case becomes
(756)
where:
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
with a sign change. The power series for this case becomes
(757)
where:
Table 72 summarizes the coefficients for the power series.
= 0.0
=
= 0.0
= 1.0
= 0.0
= 0.0
=
= 0.0
= 1.0
Table 72 Coefficients for the Power Series
Type of
Excitation A0 A1 A2
Enforced u 0.0 0.0
Enforced 0.0 –
Enforced 1.0 – –
2tf
A0 A1
x X1 –
X2

\ .
 
+
A0
A1
2t 6.283185 =
X1
X2
2tf ( )
2
A0 A1
x X1 –
X2

\ .
 
A2
x X1 –
X2

\ .
 
2
+ +
A0
A1
A2
2t ( )
2
– 39.4784 – =
X1
X2
2t ( )
2
–
u
·
2t
u
··
Main Index
319
CHAPTER 7
Enforced Motion
Cantilever Beam Model
Consider the cantilever beam first introduced in Frequency Response Analysis, 133 and shown in
Figure 76. In this case the planar model is analyzed for bending; therefore, only three DOFs per grid
point are considered: T1 (xtranslation), T2 (ytranslation), and R3 (zrotation). An acceleration ramp
function in the ydirection is enforced at the base (grid point 1) by applying a large mass and a force. T1
and R3 are constrained at grid point 1 since the enforced motion is in only the T2direction. Modal
transient response analysis (SOL 112) is run with 5% critical damping used for all modes. Modes up to
3000 Hz are computed with the Lanczos method. Figure 77 shows the idealized ramp function and the
MD Nastran implementation. The excitation is not cut off abruptly; instead it is cut off over two time
steps. A time step of 0.001 second is used, and the analysis is run for 1.0 second. Listing 75 shows the
abridged input file. See Install_dir/mdxxx/doc/dynamics/bd07bar1.dat.
Figure 76 Beam Model with Large Mass
Large Mass
1.0E9 kg
y
x
Grid Point 1
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
320
Figure 77 Idealized Ramp Function Versus MD Nastran Ramp Function
Listing 75 Abridged Input File for Enforced Acceleration
$ FILE bd07bar1.dat
$
$ CANTILEVER BEAM MODEL
$ CHAPTER 7, ENFORCED MOTION
$
SOL 112
TIME 10
CEND
TITLE = CANTILEVER BEAM
SUBTITLE = MODAL TRANSIENT RESPONSE
LABEL = ENFORCED ACCELERATION
$
SPC = 21
DLOAD = 22
TSTEP = 27
SDAMPING = 25
$
METHOD = 10
$
$ PHYSICAL OUTPUT REQUEST
SET 11 = 1,11
Idealized Input
0.15
0 0.05 Time (sec)
0.15
0.05
0.052
Time (sec)
MD Nastran Input
Y
··
(m/sec
2
)
Y
··
(m/sec
2
)
Main Index
321
CHAPTER 7
Enforced Motion
DISPLACEMENT(PLOT) = 11
ACCELERATION(PLOT) = 11
$
$ XYPLOTS
$
... XY 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 T2direction. The load is scaled to give a peak input acceleration of
. 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. (747)). 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 78 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.
0.15 m/sec
2
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
322
Figure 78 Response for Enforced Acceleration
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 79 shows the idealized input displacement
time history. Figure 710 shows the displacement and acceleration response at grid points 1 and 11.
Main Index
323
CHAPTER 7
Enforced Motion
Figure 79 Input Displacement Time History
Figure 710 Response for Enforced Displacement (With the RigidBody Mode)
Now, consider a change to the enforced displacement run. In this case, remove the rigidbody mode's
contribution either by not computing the rigidbody mode (by setting V1 to a small positive value, such
as 0.01 Hz) or by neglecting the rigidbody 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 711 shows the resulting displacement and acceleration responses at grid points 1 and 11. Note
Idealized Input
0.15
0 0.05 Time (sec)
Y(m)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
324
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., ) as a result of the sufficiently large mass.
Figure 711 Response for Enforced Displacement (Without the RigidBody Mode)
10
10 –
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 8: Random Analysis
8
Random Analysis
Random Analysis with Coupled Excitations
Main Index
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
326
Random Analysis with Coupled Excitations
In MD Nastran, random response analysis is treated as a data reduction procedure that is applied to the
results of a frequency response analysis. First, the frequency response analysis is performed for
sinusoidal loading conditions, , for each separate subcase , at a sequence of frequencies .
Normal data reduction procedures are then applied to the output of the frequency response analysis
module, resulting in a set of output quantities , corresponding to an output and subcase . The
calculations of power spectral densities and autocorrelation functions for the output quantities are
performed in the random analysis module.
Each loading condition subcase represents a unique random load source, which may be applied to many
grid points. Typically, these loads are chosen to be unit loads such as unit “g” loads or unit pressures.
Since the input is unity, the output response is also known as Transfer Function. The probabilistic
magnitude of each load source is defined by spectral density functions on RANDPS input data. If the
load subcases are correlated, the coupling spectral density is also defined on one or more RANDPS
functions. An example of coupled spectral density would be the forces on four wheels of a vehicle
traveling over a rough road.
Figure 81: Flow Diagram for Random Analysis Module is a simplified flow diagram for the random
analysis module. The inputs to the module are the frequency responses, , of quantities to
loading conditions at frequencies , and the auto and crossspectral densities of the loading
conditions and . The response quantities, , may be displacements, velocities, accelerations,
internal forces, or stresses. The power spectral densities of the response quantities are calculated by
different procedures depending on whether the loading conditions are correlated or uncorrelated. The
spectral densities due to all sources, considered independent, will be combined into one set of outputs.
P
a
{ } a e
i
u
j a
e
i
( ) j a
H
j a
e
i
( ) u
j
P
a
{ } e
i
S
a
S
ab
S
j
Main Index
327
CHAPTER 8
Random Analysis
Figure 81 Flow Diagram for Random Analysis Module
Enter
Combined
Loading
?
Compute Autocorrelation
Functions
Cross
Spectral
Density
?
Exit
Compute rms
value
Yes No
Yes
No
Yes
No
Autocorrelation?
S
j
S
j a ¯
=
S
j
H
j a
H
j b

S
ab
b
¯
a
¯
= S
j a
H
j a
2
S
a
=
Main Index
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
328
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 , of a physical variable, , which is defined by
(81)
Note that is the time average value of , which is an important quantity in the analysis of structural
failure. The onesided power spectral density of is defined by
(82)
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
(83)
from the meansquare theorem, the rms (root mean squared) magnitude, , is
(84)
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, , can be found from the
power spectral density;
(85)
The mean frequency, , 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 is the frequency response of any physical variable, ,
due to an excitation source, , which may be a point force, a loading condition or some other form of
excitation, i.e., if
R
j
t ( ) u
j
R
j
t ( ) T ·
1
T
 u
j
t ( )u
j
t t – ( ) t d
0
T
í
÷ =
lim
R
j
0 ( ) u
j
2
S
j
e ( ) u
j
S
j
e ( ) T ·
2
T
 e
i et –
u
j
t ( ) t d
0
T
í
2
÷ =
lim
R
j
t ( )
1
2t
 S
j
e ( ) et ( ) cos e d
0
·
í
=
u
j
u
j
2
R
j
0 ( )
1
2t
 S
j
e ( ) e d
0
·
í
= =
N
0
N
0
2
e 2t ( )
2
S
j
e ( ) e d
0
·
í
S
j
e ( ) e d
0
·
í
 =
N
0
H
j a
e ( ) u
j
Q
a
Main Index
329
CHAPTER 8
Random Analysis
(86)
where and are the Fourier transforms of and , then the power spectral density of the
response , is related to the power spectral density of the source, , by
(87)
Eq. (87) is an important result because it allows the statistical properties (e.g., the Autocorrelation
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 , , , etc., are statistically
independent, i.e., if the crosscorrelation function between any pair of sources
(88)
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
(89)
If the sources are statistically correlated, the degree of correlation can be expressed by a crossspectral
density, , and the spectral density of the response may be evaluated from
(810)
where is the complex conjugate of .
CrossPower Spectral Density and CrossCorrelation Functions
Theory
As shown in Eq. (88), for a pair of response quantities, and , the crosscorrelation function
between them is defined as
(811)
where and can be displacement, velocity or singlepoint constraint force responses at the same
or different grid points; or stress, strain, and force components in the same or different elements. The
crosscorrelation functions have the following relation
(812)
u
j
e ( ) H
j a
e ( ) Q
a
e ( ) · =
u
j
e ( ) Q
a
e ( ) u
j
Q
a
S
j
e ( ) S
a
e ( )
S
j
e ( ) H
j a
e ( )
2
S
a
e ( ) · =
Q
1
Q
2
Q
3
R
ab
t ( ) T ·
1
T
 q
a
t ( )q
b
t t – ( ) t d
0
T
í
÷ =
lim
S
j
e ( ) S
j a
e ( )
a
¯
H
j a
e ( )
2
S
a
e ( )
a
¯
= =
S
ab
S
j
H
j a
H
j b

S
ab
b
¯
a
¯
=
H
j b

H
j b
u
a
t ( ) u
b
t ( )
R
ab
t ( )
1
T

T · ÷
lim u
a
t ( )u
b
t t – ( ) t d
0
T
í
=
u
a
t ( ) u
b
t ( )
R
ab
t – ( ) R
ba
t ( ) =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
330
The crosspower spectral density (crossPSD) of and is defined as Eq. (813).
(813)
and is related to by the relation
(814)
It should be mentioned that the crossPSD is a complexvalued function of angular frequency .
The following relations hold for the crossPSDs
(815)
where is the complex conjugate of . If we denote by its real and imaginary parts as
then from Eq. (815), we can see that is an even function and is an odd function of , i.e.,
(816)
(817)
Substituting Eqs. (816) and (817) into Eq. (814), we have
(818)
If the system is subject to multiple sources of excitation, the crossPSD of responses and is
related to the PSDs of excitation sources by the frequency response functions as
(819)
u
a
t ( ) u
b
t ( )
S
ab
e ( )
2
T
 u
a
t ( )e
i et –
t d
0
T
í
\ .


 
u
a
t ( )e
i et
t d
0
T
í
\ .


 
T · ÷
lim =
2 R
ab
t ( )e
i et –
t d
· –
· +
í
=
R
ab
t ( ) S
ab
e ( )
R
ab
t ( )
1
4t
 S
ab
e ( )e
i et
e d
· –
· +
í
=
S
ab
e ( ) e
S
ab
*
e ( ) S
ab
e – ( ) S
ba
e ( ) = =
S
ab
*
e ( ) S
ab
e ( ) S
ab
e ( )
S
ab
e ( ) S
ab
R ( )
e ( ) = iS
ab
I ( )
e ( ) +
S
ab
R ( )
e ( ) S
ab
I ( )
e ( ) e
S
ab
R ( )
e – ( ) S
ab
R ( )
e ( ) =
S
ab
I ( )
e – ( ) S –
ab
I ( )
e ( ) =
R
ab
t ( )
1
2t
 S
ab
R ( )
e ( ) et ( ) cos S
ab
I ( )
e ( ) et ( ) sin –   e d
0
·
í
=
u
a
t ( ) u
b
t ( )
S
ab
e ( ) H
aj
e ( )H
bk
*
e ( )S
j k
e ( )
k
¯
j
¯
=
Main Index
331
CHAPTER 8
Random Analysis
where is the frequency response function of corresponding to the excitation source ,
the complex conjugate of , and the crossPSD of excitation sources when .
If the two different excitation sources and are not correlated, we have .
The aforementioned equations of crossPSDs and crosscorrelation functions become the ones for auto
PSDs and autocorrelation functions when .
The calculation of the crosscorrelation function from Eq. (818) is carried out by numerical integration
based on the trapezoidal approximation.
Numerical Integration Using LogLog Approximation
Theory
The Root Mean Square (RMS) of a random response quantity is defined as
(820)
where is the PSDF of and the angular frequency.
In numerical calculation, the integral in Eq. (820) is computed for a specified frequency range
(821)
where ( i = 1, 2, …, N ) are a set of selected frequencies.
In a LogLog scaled coordinate system, denote
where log( .) is the logarithmic function to the base 10. The linear interpolation in a subdomain
in this logarithmic coordinate system yields the following relation
(822)
From Eq. (822), the approximation of takes the form as
(823)
H
aj
e ( ) u
a
t ( ) Q
j
t ( )
H
bk
*
e ( ) H
bk
e ( ) S
j k
e ( ) j k =
Q
j
Q
k
S
j k
e ( ) 0 ÷
u
a
t ( ) u
b
t ( ) =
u
j
t ( )
u
j
1
2t
 S
j
e ( ) e d
0
·
í
1
2

=
S
j
e ( ) u
j
t ( ) e
S
j
e ( ) e d
0
·
í
S
j
e ( ) e d
e
i
e
i 1 +
í
i 1 =
N 1 –
¯
=
e
i
Y logS
j
e ( ) and X e ( ) log = =
e
i
e
i 1 +
.  
Y Y
i
–
Y
i 1 +
Y
i
–
X
i 1 +
X
i
–
 X X
i
– ( ) =
S
j
e ( )
S
j
e ( )
S
j
e
i
( )
e
i
o
 e
o
e
i
e e
i 1 +
s s . =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
332
with
(824)
The integral in subdomain is calculated as follows
(825)
is defined as
(826)
where
(827)
Using the LogLog interpolation for the approximation, we have
(828)
This logarithmic approach yields more accurate results if the PSDF curve is a straight line in an
integration subdomain in a LogLog scaled coordinate system. When the PSDF curve is close to a
straight line in a LogLog scaled coordinate system, it can be expected that the LogLog 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, LogLog selects the LogLog approach for calculating RMS, N0 and CRMS.
o
S
j
e
i 1 +
( ) S
j
e
i
( ) ( ) log
e
i 1 +
e
i
( ) log
 =
e
i
e
i 1 +
.  
S
j
e ( ) e d
e
i
e
i 1 +
í
e
i
S
j
e
i
( )ln e
i 1 +
e
i
( ) o 1 – = .
S
j
e
i
( )
1 o + ( )e
i
o
 e
i 1 +
o 1 +
e
i
o 1 +
– ( ) o 1 – = .
¹
¦
´
¦
¦
=
N
0
N
0
r
j
u
j
 =
r
j
1
2t ( )
3
 e
2
S
j
e ( ) e d
e
i
e
i 1 +
í
i 1 =
N 1 –
¯
1
2

=
e
2
S
j
e ( ) e d
e
i
e
i 1 +
í
e
i
3
S
j
e
i
( )ln e
i 1 +
e
i
( ) o 3 – = .
S
j
e
i
( )
3 o + ( )e
i
o
 e
i 1 +
o 3 +
e
i
o 3 +
– ( ) o 3 – = .
¹
¦
¦
´
¦
¦
¦
=
Main Index
333
CHAPTER 8
Random Analysis
Cumulative Root Mean Square (CRMS)
Theory
The cumulative RMS function of a random response quantity at a set of selected N frequency
points is defined as
(829)
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 realworld 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.
Example 1: Cantilevered Plate  Subjected to Base Input in 3 Directions
Problem Definition: Plate Dimension 10 inch x 1 Inch, Thickness = 0.1 inch.
Material: E = 1.E07, Poison’s Ratio = 0.25, Density = 0.1 lbs/inch
3
.
Damping = 0.04% (Critical)
Boundary Condition: Fixed at one shorter edge.
Random Input:
PSD Input X, Y Direction PSD Input Z Direction
Frequency G^2/Hz (X/Y) Frequency G^2/Hz (Z)
10.00 0.180 10.00 0.150
15.00 0.180 34.90 0.150
15.10 0.120 35.00 0.250
u
j
e ( ) u
j
t ( )
u
j
e
i
( )
1
2t
 S
j
e ( ) e d
e
1
e
i
í
1
2

= i 1 2 . N . . . = .
Main Index
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
334
Input:
See Listing 81 for a partial input file, the complete input file is
nstall_dir/mdxxxx/doc/dynamics//random_xyz.dat
Listing 81 PLATE with PSDF Output
SOL 111 $ Modal Frequency Response
CEND
$
TITLE = CANTILEVERED PLATE Subjected to 3 simulataneous Random Input
$
echo=None
set 907=105,111
Set 908 = 1,5
$
METHOD = 219
SPC = 666
RANDOM = 99
25.00 0.120 35.00 0.250
25.10 0.220 55.10 0.150
110.00 0.220 310.00 0.150
110.10 0.120 310.10 0.075
1000.00 0.120 1000.00 0.075
PSD Input X, Y Direction PSD Input Z Direction
Frequency G^2/Hz (X/Y) Frequency G^2/Hz (Z)
Main Index
335
CHAPTER 8
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 itemcode
$ 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
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
336
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
337
CHAPTER 8
Random Analysis
Typical Printed Output:
POINTID = 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 2.120942E03 7.021265E+00 6.942450E04
1.096478E+01 G 2.687511E+04 2.689941E+04 2.469965E+04 3.224830E03 1.061186E+01 1.003899E03
1.202264E+01 G 2.687514E+04 2.690436E+04 2.525660E+04 4.960394E03 1.620609E+01 1.451786E03
1.318257E+01 G 2.687521E+04 2.691035E+04 2.597651E+04 7.742561E03 2.507842E+01 2.099709E03
1.445440E+01 G 2.687526E+04 2.691752E+04 2.692406E+04 1.231372E02 3.947413E+01 3.037153E03
1.500000E+01 G 2.687528E+04 2.692080E+04 2.739068E+04 1.493719E02 4.765870E+01 3.523486E03
…
…
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
POINTID = 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 5.453022E18 1.056328E17 3.099549E+02 5.773942E01 1.416458E+02 6.138918E19
1.015469E+01 G 5.800392E18 1.123618E17 3.296992E+02 6.139689E01 1.506693E+02 6.529981E19
1.031177E+01 G 6.169956E18 1.195208E17 3.507051E+02 6.528609E01 1.602694E+02 6.946032E19
1.047129E+01 G 6.563146E18 1.271375E17 3.730536E+02 6.942172E01 1.704832E+02 7.388679E19
1.063326E+01 G 6.981471E18 1.352410E17 3.968308E+02 7.381930E01 1.813499E+02 7.859621E19
1.079775E+01 G 7.426557E18 1.438629E17 4.221290E+02 7.849557E01 1.929119E+02 8.360691E19
1.096478E+01 G 7.900112E18 1.530364E17 4.490453E+02 8.346807E01 2.052135E+02 8.893813E19
...
...
9.261187E+02 G 9.289492E14 1.807436E13 2.098790E+08 6.027814E+02 5.506023E+05 1.036910E14
9.404448E+02 G 9.372058E14 1.824654E13 1.759028E+08 3.949018E+02 6.941084E+05 1.045697E14
9.549926E+02 G 9.456252E14 1.842211E13 1.505420E+08 2.619290E+02 8.747291E+05 1.054565E14
9.697653E+02 G 9.542219E14 1.860153E13 1.310669E+08 1.819211E+02 1.074708E+06 1.063531E14
9.847666E+02 G 9.630063E14 1.878519E13 1.157493E+08 1.403549E+02 1.285253E+06 1.072605E14
1.000000E+03 G 9.719865E14 1.897341E13 1.034542E+08 1.277149E+02 1.502181E+06 1.081792E14
Main Index
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
338
A C C E L E R A T I O N V E C T O R
( ROOT MEAN SQUARE )
POINT ID. TYPE T1 T2 T3 R1 R2 R3
105 G 4.496616E+03 5.020162E+03 6.371774E+03 2.940515E+02 2.655310E+03 1.715237E+03
111 G 4.606414E+03 1.628293E+04 1.115256E+04 4.965731E+01 6.367835E+03 2.439676E+03
…
…
A C C E L E R A T I O N V E C T O R
( NUMBER OF ZERO CROSSINGS )
POINT ID. TYPE T1 T2 T3 R1 R2 R3
105 G 5.598356E+02 4.084951E+02 3.535132E+02 6.539027E+02 5.911788E+02 3.902388E+02
111 G 5.601279E+02 3.423357E+02 3.342666E+02 8.198106E+02 5.421155E+02 4.417164E+02
ELEMENTID = 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 )
FIBER  STRESSES IN ELEMENT COORDINATE SYSTEM 
FREQUENCY DISTANCE NORMALX NORMALY SHEARXY
1.000000E+01 5.000000E02 1.342324E+04 1.358092E+02 1.628790E01
5.000000E02 1.342324E+04 1.358092E+02 1.628790E01
1.096478E+01 5.000000E02 1.399410E+04 1.414498E+02 1.629220E01
5.000000E02 1.399410E+04 1.414498E+02 1.629220E01
1.202264E+01 5.000000E02 1.473256E+04 1.487424E+02 1.629737E01
5.000000E02 1.473256E+04 1.487424E+02 1.629737E01
1.318257E+01 5.000000E02 1.570457E+04 1.583351E+02 1.630361E01
5.000000E02 1.570457E+04 1.583351E+02 1.630361E01
…
…
9.120108E+02 5.000000E02 2.387684E01 1.323565E02 1.282505E02
5.000000E02 2.387684E01 1.323565E02 1.282505E02
9.536902E+02 5.000000E02 6.290167E01 3.275067E02 1.407314E02
5.000000E02 6.290167E01 3.275067E02 1.407314E02
9.733126E+02 5.000000E02 1.064134E+00 5.111575E02 1.467233E02
5.000000E02 1.064134E+00 5.111575E02 1.467233E02
1.000000E+03 5.000000E02 2.290897E+00 9.822020E02 1.550554E02
5.000000E02 2.290897E+00 9.822020E02 1.550554E02
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 )
ELEMENT FIBER  STRESSES IN ELEMENT COORDINATE SYSTEM 
ID. DISTANCE NORMALX NORMALY SHEARXY
1 5.000000E02 2.436790E+03 2.411689E+02 1.719076E+01
5.000000E02 2.436790E+03 2.411689E+02 1.719076E+01
5 5.000000E02 1.110793E+03 1.885808E+00 1.456949E+01
5.000000E02 1.110793E+03 1.885808E+00 1.456949E+01
…
…
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 )
ELEMENT FIBER  STRESSES IN ELEMENT COORDINATE SYSTEM 
ID. DISTANCE NORMALX NORMALY SHEARXY
1 5.000000E02 4.962503E+01 5.812841E+01 3.034584E+02
5.000000E02 4.962503E+01 5.812841E+01 3.034584E+02
5 5.000000E02 7.719165E+01 3.581298E+02 3.069451E+02
5.000000E02 7.719165E+01 3.581298E+02 3.069451E+02
Main Index
339
CHAPTER 8
Random Analysis
Figure 82 Power Spectral Density, Acceleration, Node 111(Tz)
Example 2: Plate  subjected to Pressure and point load with CrossSpectrum 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/inch
3
.
Damping = 0.03% (Critical)
Boundary Condition: Fixed at one shorter edge.
Random Input:
AutoSpectra  Pressure
Load AutoSpectra  Corner Force
CrossSpectrum of Pressure and
Corner Load
Frequency PSI^2/Hz Frequency lb^2/Hz Frequency Real Part
Imaginary
Part
10.0 0.10 10.0 0.50 10.0 0.099619 0.007816
20.0 0.10 20.0 0.50 20.0 0.099619 0.043579
30.0 1.00 30.0 2.50 100.0 0.498097 0.043579
100.0 1.00 100.0 2.50 500.0 0.070711 0.070711
500.0 0.10 500.0 2.50 1000.0 0.000001 0.000001
1000.0 0.10 1000.0 1.00E06   
Main Index
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
340
Figure 83 FEM Model
Loading:
Main Index
341
CHAPTER 8
Random Analysis
Input
CrossSpectrum 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 nonzero and imaginary (Y) component to zero. The TID on this RANDPS points to table
TABRND1 with real component of crosspsd input. In the second RANDPS entry, set the real (X)
component to zero and imaginary (y) component to nonzero with TID points to TABRND1 table with
imaginary component of crosspsd input (see Listing 82). 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.
Listing 82 PLATE with CrossPSD
SOL 111
$
CEND
$
TITLE = Plate subjected to pressure and concentrated load.
ECHO = NONE
$
Random = 101
line = 99999
$
set 5 = 28
SDAMPING = 1
METHOD = 1
FREQUENCY = 1
SPC = 1
ACCELERATION(plot,rprint,psdf)=5
SUBCASE 1
SUBTITLE=Applied Pressure
DLOAD = 11
SUBCASE 2
SUBTITLE=Applied Load at Tip
DLOAD = 12
Output (XYOUT)
xgrid = yes
ygrid = yes
xlog = no
ylog = no
xtitle = Time (sec)
ytitle = Acceleration Auto Corrolation Node 55(Tz)
xyplot xyprint acce auto / 55(T3)
Xlog = yes
Ylog = yes
Xmin = 10.
Xmax = 1000.
$
xtitle = Freq (Hz)
Ytitle = Acceleration PSDF Node 28(Tz)
XYplot Xyprint ACCE PSDF / 28(T3)
$
BEGIN BULK
PARAM POST 0
PARAM WTMASS .002589
PARAM GRDPNT 0
PARAM,NOCOMPS,1
Main Index
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
342
PARAM PRTMAXIM YES
FREQ2 1 10. 1000. 300
FREQ4 1 10. 1000. .1 3
$
$ Capture all break points frequencies of input PSD profile
$
Freq, 1, 20., 30., 100., 500., 1000.
TABDMP1 1 CRIT
.1 .03 2000. .03 ENDT
EIGRL 1 .1 2000. 0
$
$ Unit Pressure Load (Subcase 1)
$
RLOAD1 11 4 1
$
$ Unit Pressure
$
PLOAD4 4 1 1. THRU 40
$
$ Unit Point Load (Subcase 2)
RLOAD1 12 5 1
$
$ Unit Nodal Force
$
FORCE 5 11 0 1. 0. 0. 1.
$
$ Constant Load Table
$
TABLED1 1
0. 1. 1000. 1. ENDT
$
$ Random Input
$
$ Auto Pressure Load
RANDPS,101,1,1,1.,0.,200022
$
$ Auto Point Load
RANDPS,101,2,2,1.,0.,300022
$
$ Cross – Real Part between Pressure and Point Load
RANDPS,101,1,2,1.,0.,400011
$
$ Cross – Imaginary Part between Pressure and Point Load
RANDPS,101,1,2,0.,1.,500011
$
$ PSD autopress
$
TABRND1, 200022,log,log,,,,,,+
+, 10.00, 0.100, 20.00, 0.100, 30.00, 1.000, 100.0, 1.000,
+, 500.0, 0.100, 1000., 0.100, endt
$
$ PSD autoforce
$
TABRND1, 300022,log,log,,,,,,+
+, 10.00, 0.500, 20.00, 0.500, 30.00, 2.500, 100.0, 2.500,
+, 500.0, 2.500, 1000.,1.006, endt
$
$ Cross PSD  Real part
$
TABRND1, 400011,linear,linear,,,,,,+
Main Index
343
CHAPTER 8
Random Analysis
+, 10.0, 0.09962, 20.0,0.09962, 100.,0.49810, 500., 0.07071,
+, 1000., 1.06, endt
$
$ Cross PSD  Imaginary part
$
TABRND1, 500011,linear,linear,,,,,,+
+, 10.0, 0.00782, 20.0, 0.04358, 100., 0.04358, 500.,0.07071,
+, 1000., 1.06, endt
$
$ Autocorrelation Function Time Lag
$
Randt1, 101, 2000, 0., 0.10
$
SPC1 1 12345 1 12 23 34 45
$
Typical Printed Output
POINTID = 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 5.453022E18 1.056328E17 3.099549E+02 5.773942E01 1.416458E+02 6.138918E19
1.015469E+01 G 5.800392E18 1.123618E17 3.296992E+02 6.139689E01 1.506693E+02 6.529981E19
1.031177E+01 G 6.169956E18 1.195208E17 3.507051E+02 6.528609E01 1.602694E+02 6.946032E19
1.047129E+01 G 6.563146E18 1.271375E17 3.730536E+02 6.942172E01 1.704832E+02 7.388679E19
1.063326E+01 G 6.981471E18 1.352410E17 3.968308E+02 7.381930E01 1.813499E+02 7.859621E19
1.079775E+01 G 7.426557E18 1.438629E17 4.221290E+02 7.849557E01 1.929119E+02 8.360691E19
1.096478E+01 G 7.900112E18 1.530364E17 4.490453E+02 8.346807E01 2.052135E+02 8.893813E19
...
...
9.261187E+02 G 9.289492E14 1.807436E13 2.098790E+08 6.027814E+02 5.506023E+05 1.036910E14
9.404448E+02 G 9.372058E14 1.824654E13 1.759028E+08 3.949018E+02 6.941084E+05 1.045697E14
9.549926E+02 G 9.456252E14 1.842211E13 1.505420E+08 2.619290E+02 8.747291E+05 1.054565E14
9.697653E+02 G 9.542219E14 1.860153E13 1.310669E+08 1.819211E+02 1.074708E+06 1.063531E14
9.847666E+02 G 9.630063E14 1.878519E13 1.157493E+08 1.403549E+02 1.285253E+06 1.072605E14
1.000000E+03 G 9.719865E14 1.897341E13 1.034542E+08 1.277149E+02 1.502181E+06 1.081792E14
Main Index
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
344
A C C E L E R A T I O N V E C T O R
( ROOT MEAN SQUARE )
POINT ID. TYPE T1 T2 T3 R1 R2 R3
28 G 5.976811E05 8.319094E05 6.028776E+05 1.329319E+05 3.110180E+05 2.005465E05
A C C E L E R A T I O N V E C T O R
( NUMBER OF ZERO CROSSINGS )
POINT ID. TYPE T1 T2 T3 R1 R2 R3
28 G 1.737170E+02 1.738071E+02 5.728496E+02 5.317244E+02 2.119516E+02 1.736413E+02
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 1.000E01 1.630E+12 3.150E03 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
ACCELERATION CURVE ID = 55 COMPONENT = 5 WHOLE FRAME
PRINT NUMBER XVALUE YVALUE CARD NUMBER
1 0.000000E+00 2.145029E+12
2 5.000000E05 2.131748E+12
3 1.000000E04 2.092620E+12
4 1.500000E04 2.029721E+12
5 2.000000E04 1.946369E+12
...
...
1996 9.975000E02 3.013784E+10
1997 9.980001E02 3.508981E+10
1998 9.985001E02 3.997675E+10
1999 9.990001E02 4.479540E+10
2000 9.995001E02 4.954452E+10
2001 1.000000E01 5.421790E+10
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
ACCELERATION CURVE ID = 28 COMPONENT = 5 WHOLE FRAME
PRINT NUMBER XVALUE YVALUE CARD NUMBER
1 1.000000E+01 3.099549E+02
2 1.015469E+01 3.296992E+02
3 1.031177E+01 3.507051E+02
4 1.047129E+01 3.730536E+02
5 1.063326E+01 3.968308E+02
...
...
309 9.404448E+02 1.759028E+08
310 9.549926E+02 1.505420E+08
311 9.697653E+02 1.310669E+08
312 9.847666E+02 1.157493E+08
313 1.000000E+03 1.034542E+08
Main Index
345
CHAPTER 8
Random Analysis
Figure 84 Autocorrelation, Acceleration of Grid 55 (Tz)
Figure 85 Power Spectral Density, Acceleration of Grid 28 (Tz)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Random Analysis with Coupled Excitations
346
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 9: Response Spectrum Analysis
9
Response Spectrum Analysis
Shock and Response Spectrum Analysis
Response Spectrum Examples
Main Index
MD Nastran Dynamic Analysis User’s Guide
Shock and Response Spectrum Analysis
348
Shock and Response Spectrum Analysis
Shock spectra analysis and response spectrum analysis are methods used by many engineers to estimate
the maximum dynamic response of a structure. Most applications involve complicated timedependent
loads or accelerations that excite the base of a structure, such as an earthquake ground motion on a
building or an explosive shock on a small component in a ship. (Note that the only difference between
shock and response spectra is whether output displacements are measured in a fixed frame of reference
or relative to the base motion.)
The advantage of these methods over a conventional transient analysis is economy and simplicity. The
only major calculation step is obtaining a sufficient number of normal modes to represent the entire
frequency range of the input excitation and resulting response. The disadvantage of the method is that
the accuracy may be questionable and the requirement of special input data in the MD Nastran solution
sequences. In many cases, a direct transient analysis with the actual excitation load will be more
accurate, easier to use, and faster.
The procedure involves two stages. First the applied loads or base excitations are converted in a direct
transient solution (SOL 109) into a spectrum table consisting of peak response magnitudes for a set of
single degreeoffreedom oscillators. Each oscillator is a scalar spring/mass/damper having a different
natural frequency and damping ratio. This stage is optional since the shock spectrum data is frequently
given in the contractual design specifications or, in the case of earthquakes, is available through
governmental agencies.
The second stage of the analysis consists of a modal analysis of the structure, data recovery, and the
response calculation that combines the modal properties of the analysis model with the spectrum data of
the applied loads. This stage is performed in a modal analysis solution sequence (SOL 103). If a database
was saved from the first stage, a restart will provide the spectrum data automatically. Otherwise, the
spectrum data must be supplied in a direct tabular input (response versus natural frequency for several
damping ratios).
Theoretical Background
Starting with a modal transient analysis, the general approximation for a response quantity, , is
(91)
where and are the modal outputs and generalized displacements. The actual modal equations are
(92)
where is the vector of loading functions. For loading due to base accelerations, the equivalent inertial
loads are
(93)
u
k
u
k
t ( ) o
i k
c
i
t ( )
i
¯
=
o c
c
··
i
g
i
e
i
c
·
i
e
i
2
c
i
+ + o
i
T
P t ( ) { } =
P
P t ( ) { } M
aa
  D
ar
  u
··
r
t ( ) { } – =
Main Index
349
CHAPTER 9
Response Spectrum Analysis
where the columns of represent vectors of rigid body motions of the whole structure and the
accelerations correspond to the base motions. Substituting Eq. (93) into Eq. (92) 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
(94)
where is a response function in direction , and is a function of the variables , , and . The peak
values of , obtained over a range of frequencies and damping factors is called the response spectrum
for the excitation, .
Next, from the normal mode analysis, we define the participation factors , for mode and direction
, as
(95)
Then, from Eq. (91), the actual transient response at a physical point is
(96)
The peak magnitudes of in Eq. (96) are usually dominated by the peak values of occurring at
the natural frequencies. In spectrum analysis the peak values of
are approximated by combining
functions of the peak values, , in the approximation
(97)
ABS Option
Equations (96) and (97) 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
(98)
where the average peak modal magnitude, is
D
ar
 
x
··
r
gex
·
r
e
2
x
r
+ + u
··
r
t ( ) =
x
r
r e g t
x
r
u
··
r
v
i r
i
r
v
i r
o
i
T
M
aa
  D
ar
{ } – =
u
k
t ( ) o
i k
v
i r
x
r
e
i
, g
i
, t ( )
r
¯
i
¯
=
u
k
x t ( )
u
k
x
ri
e
i
, g
i
( ) max x
ri
e
i
, g
i
, t ( ) =
u
k
t ( ) o
i k
v
i r
x
ri
e
i
, g
i
( )
r
¯
i
¯
~
u
k
o
i k
c
i
( )
2
i
¯
~
c
i
Main Index
MD Nastran Dynamic Analysis User’s Guide
Shock and Response Spectrum Analysis
350
(99)
SRSS Method
This approach is known as the SRSS (square root of sumsquared) 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 . 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
where the jth mode is the mode that produces the largest magnitude in the product . The peak modal
magnitudes, , are calculated with Eq. (99).
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:
The value for CLOSE is set by PARAM,CLOSE (the default is 1.0).
The modal summation option is set via PARAM,OPTION (the ABS method is the default). Both
PARAM,OPTION and PARAM,CLOSE may be set in any subcase, allowing summation by several
conventions in a single run. See SCRSPEC (p. 907) in the MD Nastran Quick Reference Guide.
c
i
v
i r
x
r
e
i
, g
i
i
( ) ( )
2
r
¯
=
u
k
u
k
o
j k
c
j
o
i k
c
i
( )
2
i j =
¯
+ ~
o
j k
c
j
o
j k
c
j
f
i 1 +
CLOSE f
i
· <
Main Index
351
CHAPTER 9
Response Spectrum Analysis
Generating Response Spectra Curves
A response spectrum is generated as follows:
1. Assume that there is a series of small, single degreeoffreedom (SDOF) oscillators each attached
to the same location at the connection to the base structure. (In the examples in the introduction
to this section, the base structures are the building and the spacecraft). These oscillators each
have a different resonant frequency, and all have the same modal damping ratio (2%, for
example).
2. Apply a transient excitation to the base structure. Use the base structure’s damping when
computing the base structure’s transient response at the location of the SDOF oscillators.
3. Use the base structure transient response as input to each SDOF oscillator. Compute the
magnitude of peak response of each SDOF oscillator, and plot the peak response versus oscillator
resonant frequency.
4. Change the modal damping ratio for the oscillators (to 5%, for example) and repeat Steps 2 and
3 for the range of expected damping.
The response spectrum, therefore, depicts the maximum response magnitude of an SDOF system as a
function of resonant frequency and damping. Figure 91 depicts the generation of a spectrum.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Shock and Response Spectrum Analysis
352
Figure 91 Response Spectrum Generation
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
(910)
(a)
(b)
(c)
Peak
Response
Resonant Frequency
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.
Resonant Frequency (f)
(d)
= 5% critical damping
= 2% critical damping
= 0% critical damping
Peak
Response
(e)
Series of Oscillators
Transient
Response
Base
Structure Transient
Excitation
f
1
f
2
f
3
. . . f
max
f
1
f
2
f
3
. . . f
max
f
n
X
·
r
eX
r
~
X
··
e
2
X
r
~
Main Index
353
CHAPTER 9
Response Spectrum Analysis
Design spectra are most often defined in terms of , , and .
Note that for very low oscillator frequencies ( ),
(911)
where is the motion of the base of the oscillator.
Similarly, for very high oscillator frequencies ( ),
(912)
The approximate relationships between , , and , shown in Eq. (910), 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.
Peak Response Calculation
Response spectra are applied to a structure for subsequent analysis. Note that the structure in this case
is not the base structure for which the spectra were computed but, rather, a smaller structure such as a
piece of equipment.
Response spectrum analysis approximates the peak structural response (typically stresses and
displacements, though not limited to those quantities). Approximations are made by assuming:
1. Only the peak response is computed for the oscillators (see the preceding section).
2. There is no phase information or sign computed; only the magnitude of peak response is
computed.
3. The oscillator mass is small relative to the base structure’s mass.
4. The displacements, velocities, and accelerations are related by the approximate relationships in
Eq. (910).
5. The peak modal responses are combined to form the overall response via various combination
methods (see below).
The spectra themselves are often approximate. For example, design spectra have been developed for
seismic analysis, and these have conservatisms built into them by the fact that safety factors are
incorporated by either increasing the spectra values and/or by decreasing the damping values.
These approximations make response spectrum analysis a tool that is useful for design and efficiency.
The effects of these approximations are further described in the remainder of this section.
X
r
X
·
r
X
··
e 0 ÷
X 0 ÷
X
r
U
b
÷
U
b
e · ÷
X U
b
÷
X
r
0 ÷
X
r
X
··
r
X
··
Main Index
MD Nastran Dynamic Analysis User’s Guide
Shock and Response Spectrum Analysis
354
Modeling and Analysis Considerations
Response spectrum application is a postprocessing function of normal modes analysis. It is run in the
normal modes solution sequences, so the modeling and analysis considerations that apply for normal
modes analysis also apply for response spectrum application. The additional considerations also need to
be followed:
1. The structure is run as an unrestrained model in the direction(s) of the load spectrum application.
2. A large mass, on the order of to times the mass of the structure, must be used at the
structure’s base grid points where the input occurs. One way to do this is to use RBEs or MPCs
to connect the base points to a separate grid point, and apply the large mass to that separate grid
point. This separate grid point is where the spectrum is applied.
3. A SUPORT Bulk Data entry is required at the spectrum input location. See SUPORT Entry
(Ch. 3) for an explanation.
4. The modes must be mass normalized (which is the default).
The spectra that MD Nastran can apply are absolute acceleration, relative displacement, and relative
velocity spectra. You specify A, D, or V for acceleration displacement or velocity, respectively, to
specify the spectrum type.
Use all modes within the frequency range specified by the spectrum, but do not use modes outside of the
spectrum range. Usually, spectra to apply are considered to have zero values outside of their range of
definition; for example, an absolute acceleration spectrum defined from 0 to 30 Hz is assumed to be zero
beyond 30 Hz. However, MD Nastran extrapolates spectral values for modes beyond the spectral range,
which may lead to unexpected answers. You can limit the number of modes used in the spectrum
application by limiting the number of computed modes (via the EIGRL or EIGR entry) or by using
PARAM,HFREQ, HFREQFL,f (where f is the highest frequency of structural mode to use) or
PARAM,LFREQ, LFREQFL,f (where f is the lowest frequency of structural mode to use).
Consider the entire response spectrum process—generation and application—as a twostep process. Step
1 is generation of the response spectra and Step 2 is the application of the response spectra. For a given
input, transient applied to the base structure (Step 1), the same stresses occur (Step 2) regardless of
whether acceleration or displacement spectra were computed in Step 1. However, displacements and
accelerations are different, because answers computed by using the absolute acceleration spectrum
contain the rigid body contribution, whereas answers computed by using the relative displacement
spectrum do not contain the rigid body contribution. Displacement and acceleration responses can be
made equal regardless of which spectra was used by using PARAM,LFREQ, LFREQFL,0.01 (or some
other small number) to remove the rigid body mode contribution from the answers.
Stresses and other element quantities are unaffected by the contribution of any rigid body modes. The
same situation applies to relative velocity spectra as to relative displacement spectra. However, because
the relationships in Eq. (910) are approximate, all answers (including stresses) will be slightly different
depending on whether displacement, velocity, or acceleration spectra were used.
10
3
10
6
Main Index
355
CHAPTER 9
Response Spectrum Analysis
Dynamic Response Predictions
Once a spectrum is computed, it can be used for the dynamic response analysis of an MD Nastran model
of the component. For example, the spectrum generated for a floor in a building that is subjected to an
earthquake can then be applied to a complex model of a piece of equipment attached to that floor. The
peak response of each mode of the equipment model is obtained from the spectrum, and these peak
responses are combined to create the overall response.
Because response spectrum generation involves transient response, similar modeling and analysis
considerations apply. In addition, the time step (field 4 on the TSTEP Bulk Data entry, DT) should not
be changed during the run, because MD Nastran uses only the initial DT specification for the entire
response spectrum generation run and therefore wrong answers could occur.
The time step, DT, and time duration, (where 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.
User Interface for Response Spectra Generation
Response spectra are generated in the transient response solution sequences (SOL 109 for direct and
SOL 112 for modal). Transient response input is required to apply the transient excitation to the base
structure. Additional input required to generate response spectra are described in Table 91 and
Table 92.
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).
Table 91 Case Control Input for Response Spectrum Generation
Case Control Command Description
XYPLOT SPECTRAL Compute spectra.
XYPUNCH SPECTRAL Punch spectra for subsequent use.
DT N · N
Main Index
MD Nastran Dynamic Analysis User’s Guide
Shock and Response Spectrum Analysis
356
Examples of these commands include:
$ Plot absolute acceleration spectra for grid point 85, T3 component
XYPLOT ACCE SPECTRAL 1 /85(T3RM)
$ Punch relative displacement spectra for grid point 3, T1 component
XYPUNCH DISP SPECTRAL 1 /3(T1IP)
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.
User Interface for Spectrum Application
Response spectrum application is done in the normal modes solution sequences (SOL 103, for example).
In addition to the input for computing normal modes, input is required for applying the spectra, as shown
in Table 93 and Table 94.
Table 92 Bulk Data Input for Response Spectrum Generation
Bulk Data Entry Description
PARAM,RSPECTRA,0 Requests calculation of spectra.
DTI,SPSEL,0
DTI,SPSEL,1
Header for DTI.
Selects oscillator frequencies, oscillator damping
values, and grid points at which spectra will be
computed.
FREQi Specifies oscillator damping values.
FREQi Specifies oscillator frequencies.
Table 93 Input for Response Spectrum Application
Case Control Command Description
METHOD Selects eigenvalue extraction method.
SDAMPING Selects the TABDMP1 Bulk Data entry.
DLOAD Selects the DLOAD Bulk Data entry.
Main Index
357
CHAPTER 9
Response Spectrum Analysis
All input listed in the table is required with the exception of PARAM,OPTION and PARAM,CLOSE.
Table 94 Input for Response Spectrum Application
Bulk Data Entry Description
PARAM,SCRSPEC,0 Requests response spectrum application.
EIGR or EIGRL Eigenvalue extraction method.
TABDMP1 Specifies damping for the structure.
DLOAD Defines spectrum multipliers.
DTI,SPECSEL,0 Header for DTI.
DTI,SPECSEL,1 Specifies type of spectrum (A, V, or D) and selects damping.
A = absolute acceleration spectrum.
V = relative velocity spectrum.
D = relative displacement spectrum.
TABLED1 Specifies input spectrum values.
SUPORT Specifies input spectrum grid points.
CONM2,CMASS1, etc. Defines large mass used for the input spectrum.
PARAM,OPTION,a Specifies modal combination method (a = ABS [default], SRSS,
or NRL).
PARAM,CLOSE,r Specifies closeness parameter for modal combinations (the
default is 1.0).
Main Index
MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
358
Response Spectrum Examples
This section provides an example of response spectra generation and an example of response spectrum
application. The base structure for which response spectra are computed is a model of a chimney, excited
by an earthquake time history. The spectra are calculated at base and tip of the chimney. The resulting
spectra at base is applied to base of of chimney and results are compared with transient analysis.
Response Spectrum Generation
The grid points at the base of the chimney are tied to a separate grid point (9999) via RBE2 entries.
Earthquake excitation is applied to this grid point, in the x direction (T1). The absolute acceleration
response spectrum is computed for grid point 9999(base) and 2593(tip). SOL 112, modal transient
response, is used to compute the spectrum. Damping of 2% critical is used for the building and damping
of 2% critical is used for the generated response spectrum. Listing 91 shows a portion of the input file
for this model. See Install_dir/mdxxxx/doc/dynamics/createshock.dat for a complete model.
Figure 92 Chimney Model (24 inch radius, 700 inch long and 0.25 inch thick)
Main Index
359
CHAPTER 9
Response Spectrum Analysis
Listing 91 Abridged Input File for Response Spectrum Generation
sol 112
$c
cend
echo=none
line=99999
Subcase 11
subtitle = Transient Analysis
method = 99
spc = 77
dload = 66
sdamp = 88
tstep = 123
set 55 = 9999, 2593
acce(plot) = all
disp(plot) = 55
velo(plot) = 55
force(plot) = all
stress(plot) = all
output(xyout)
xaxis = yes
xgrid = yes
yaxis = yes
ygrid = yes
xmax = 15.0
xtitle = Acceleration T.H. at base
xyplot,acce,resp/9999(t1)
xtitle = Acceleration T.H. at tip
xyplot,acce,resp/2593(t1)
xlog = yes
xmax = 100.
xmin = 1.
ylog = yes
$ ymin = 100.
xtitle = Shock Spectrum at Base  Damping = 0.02
xyplot,xyprint,xypunch acce,spectral/9999(t1rm)
xtitle = Shock Spectrum at Tip  Damping = 0.02
xyplot,xyprint,acce,spectral/2593(t1rm)
$
begin bulk
$
param,wtmass,.002589
param,grdpnt,0
param,rspectra,0
param,rsprint,1
param,post,0
$
dti,spsel,0
dti,spsel,1,91, 92, 9999,2593
$
$ set 91 selects damping of oscilator and set 92 selects the
$ frequencies at which spectra will be calcuated.
$
freq, 91, .02
freq2, 92, 1.0, 40.0, 200
freq4, 92, 1.0, 40.0, 0.2, 5
$
eigrl, 99, , 40.0,
tstep, 123, 3000, 0.005, 1
Main Index
MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
360
$
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: 07May10 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 Xdirection 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 93.
Main Index
361
CHAPTER 9
Response Spectrum Analysis
.
Figure 93 Absolute Acceleration Response Spectrum Plots for Grid 9999 (base) and 2593
(tip)
A portion of the printed output file is shown in Listing 92. 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 93; this output is
contained in the punch file.
Listing 92 Printed Output (Abridged)
TRANSIENT ANALYSIS
FRACTION OF CRITICAL DAMPING = .02 ABSOLUTE IN REAL LOCATION, RELATIVE IN IMAG. LOCATION
POINTID = 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
MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
362
Listing 93 Punched Response Spectrum
Response Spectrum Application
The response spectra computed at base in first example will be applied at the base of the structure and
results are compared with transient analysis. A portion of the input file is shown in Listing 94. See
.Install_dir/mdxxxx/doc/dynamics/applyshock.dat for the complete file.
Listing 94 Abridged Input File for Response Spectrum Application
$ Response Spectra Application
$
sol 103
$c
cend
echo=none
line=99999
subcase 1
subtitle = Shock Spectrum Analysis
$SUBCASE 1 1
$ACCE 3 9999 3 1 2
$ 2.000000E02
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
Note: Only PARAM,POST,1 (op2) option is supported for postprocessing results.
Main Index
363
CHAPTER 9
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
$ 2.000000E02
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
Response spectrum application is invoked via PARAM,SCRSPEC,0. The DTI,SPECSEL entry
specifies that the input spectrum is acceleration (denoted by the A in field 5). The TABDMP1 entry
Main Index
MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
364
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 95 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 95. These quantities are computed using the SRSS method.
Listing 95 Printed Output (Abridged)
Main Index
365
CHAPTER 9
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
MATRIX FN (GINO NAME 101 ) IS A DB PREC 1 COLUMN X 21 ROW RECTANG MATRIX.
COLUMN 1 ROWS 2 THRU 21 
ROW
2) 3.5650D+00 3.5650D+00 1.0981D+01 1.0981D+01 1.3047D+01 1.3047D+01 2.1430D+01 2.1430D+01 2.2396D+01 2.2396D+01
12) 3.0897D+01 3.0897D+01 3.1140D+01 3.1140D+01 3.2443D+01 3.2443D+01 3.6118D+01 3.6118D+01 3.9044D+01 3.9044D+01
THE NUMBER OF NONZERO TERMS IN THE DENSEST COLUMN = 20
THE DENSITY OF THIS MATRIX IS 95.24 PERCENT.
PSIT
POINT VALUE POINT VALUE POINT VALUE POINT VALUE POINT VALUE
COLUMN 1
9999 T1 5.08829E+02
COLUMN 2
9999 T1 4.13628E06
COLUMN 3
9999 T1 2.07668E+00
COLUMN 4
9999 T1 1.29937E10
COLUMN 5
9999 T1 8.55402E11
...
...
COLUMN 18
9999 T1 2.22644E12
COLUMN 19
9999 T1 4.85724E12
COLUMN 20
9999 T1 3.55792E13
COLUMN 21
9999 T1 7.24552E13
Main Index
MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
366
Following fringe plots shows the comparison between transient analysis and shock spectrum analysis in
Figure 94 and Figure 95. The maximum (envelope) response across all time steps is plotted using
MATRIX UHVR (GINO NAME 101 ) IS A REAL 3 COLUMN X 21 ROW RECTANG MATRIX.
COLUMN 1 ROWS 2 THRU 21 
ROW
2) 2.5135E06 1.2620E+00 5.3064E12 3.4933E12 1.4455E12 3.2655E13 2.8290E10 8.8593E03 1.8225E13 2.2641E13
12) 2.5064E14 2.8510E14 5.5913E14 3.4243E14 8.9867E15 2.8723E14 5.2496E15 1.1452E14 6.8395E16 1.3928E15
COLUMN 2 ROWS 2 THRU 21 
ROW
2) 5.6301E05 2.8268E+01 3.6612E10 2.4102E10 1.1850E10 2.6770E11 3.8092E08 1.1929E+00 2.5645E11 3.1860E11
12) 4.8658E12 5.5346E12 1.0940E11 6.6998E12 1.8319E12 5.8552E12 1.1913E12 2.5990E12 1.6779E13 3.4169E13
COLUMN 3 ROWS 2 THRU 21 
ROW
2) 1.2611E03 6.3320E+02 2.5260E08 1.6629E08 9.7146E09 2.1946E09 5.1290E06 1.6062E+02 3.6087E09 4.4832E09
12) 9.4460E10 1.0745E09 2.1404E09 1.3109E09 3.7343E10 1.1936E09 2.7035E10 5.8980E10 4.1162E11 8.3825E11
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 NORMALX NORMALY SHEARXY ANGLE MAJOR MINOR VON MISES
1 1.250000E01 1.484016E+03 1.375312E+02 6.375009E+00 0.2713 1.484047E+03 1.375010E+02 1.420297E+03
1.250000E01 1.526726E+03 1.457677E+02 5.816203E+00 0.2413 1.526751E+03 1.457432E+02 1.459347E+03
2 1.250000E01 1.438925E+03 1.333524E+02 1.893104E+01 0.8306 1.439199E+03 1.330779E+02 1.377490E+03
1.250000E01 1.480337E+03 1.413386E+02 1.727163E+01 0.7389 1.480560E+03 1.411158E+02 1.415288E+03
3 1.250000E01 1.350112E+03 1.251217E+02 3.091186E+01 1.4446 1.350892E+03 1.243421E+02 1.293212E+03
1.250000E01 1.388968E+03 1.326150E+02 2.820226E+01 1.2853 1.389601E+03 1.319822E+02 1.328536E+03
4 1.250000E01 1.220277E+03 1.130892E+02 4.195344E+01 2.1669 1.221865E+03 1.115018E+02 1.170105E+03
1.250000E01 1.255396E+03 1.198619E+02 3.827598E+01 1.9284 1.256685E+03 1.185732E+02 1.201794E+03
...
...
2589 1.250000E01 2.738482E01 2.450364E+00 7.298062E01 73.0767 2.672420E+00 5.179233E02 2.646904E+00
1.250000E01 2.116490E01 2.492312E+00 7.265424E01 73.7487 2.704097E+00 1.363183E04 2.704165E+00
2590 1.250000E01 3.029849E01 2.711076E+00 5.377296E01 77.9672 2.825696E+00 1.883649E01 2.736380E+00
1.250000E01 2.341678E01 2.757487E+00 5.353248E01 78.5042 2.866359E+00 1.252956E01 2.805810E+00
2591 1.250000E01 3.229155E01 2.889412E+00 3.293143E01 82.8035 2.930994E+00 2.813340E01 2.800944E+00
1.250000E01 2.495715E01 2.938877E+00 3.278416E01 83.1490 2.978266E+00 2.101827E01 2.878934E+00
2592 1.250000E01 3.330345E01 2.979956E+00 1.108930E01 87.6052 2.984594E+00 3.283968E01 2.834698E+00
1.250000E01 2.573921E01 3.030970E+00 1.103970E01 87.7243 3.035357E+00 2.530049E01 2.917095E+00
Main Index
367
CHAPTER 9
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.
Figure 94 Maximum Acceleration (near top of the chimney)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
368
Figure 95 Maximum vonMises Stress (near base of chimney)
Example: Structure subjected to multiple spectra
The input is usually defined as response spectrum curves. It is not unusual to have an excitation defined
in terms of a set of response spectrum curves. Each curve of the set is associated with a value of damping
(% critical damping). Based on modal damping defined by TABDMP1 table (selected by SDAMP= in
case control section), Nastran will interpolate the excitation input value at each modal frequency from
the Spectrum Input curves. In addition, the input could be in one direction only or any two or all three
directions simultaneously.
.Given Spectrum input in (x, y and z) direction
Input spectrum could be Acceleration, Velocity or Displacement.
Table 95 Input in XDirection
2% Damping 5% Damping 7% Damping
Freq Accel(G) Freq Accel(G) Freq Accel(G)
10.0 2.0 12.0 1.5 5.0 1.0
20.0 2.0 15.0 1.5 25.0 1.0
Main Index
369
CHAPTER 9
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 
100.0 8.0 90.0 4.0 110.0 3.0
900.0 8.0 1200.0 4.0 850.0 3.0
1100.0 1.0 1300.0 0.7 1200.0 0.5
2000.0 1.0 2500.0 0.7 3000.0 0.5
Table 96 Input in YDirection
1% Damping 4% Damping 8% Damping
Freq Accel(G) Freq Accel(G) Freq Accel(G)
10.0 1.0 12.0 0.8 5.0 0.4
20.0 1.0 15.0 0.8 25.0 0.4
100.0 6.0 90.0 5.0 110.0 3.0
900.0 6.0 1200.0 5.0 850.0 3.0
1100.0 1.5 1300.0 0.9 1200.0 0.5
2000.0 1.5 2500.0 0.9 3000.0 0.5
Table 97 Input to ZDirection
1% Damping 3% Damping 7% Damping
Freq Accel(G) Freq Accel(G) Freq Accel(G)
10.0 1.5 12.0 1.0 5.0 0.6
20.0 1.5 15.0 1.0 25.0 0.6
100.0 7.0 90.0 5.0 110.0 3.5
900.0 7.0 1200.0 5.0 850.0 3.5
1100.0 1.8 1300.0 1.4 1200.0 0.7
2000.0 1.8 2500.0 1.4 3000.0 0.7
Table 95 Input in XDirection
2% Damping 5% Damping 7% Damping
Freq Accel(G) Freq Accel(G) Freq Accel(G)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
370
$  \          
$ +>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 set ID (or Record ID) of 72 for Y 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
X direction, 2 % damping Table ID 101
X direction, 5 % damping Table ID 102
X direction, 7 % damping Table ID 103
Y direction, 1 % damping 201
Y direction, 4 % damping 202
Y direction, 8 % damping 203
Z direction, 1 % damping 301
Z direction, 3 % damping 302
Z direction, 7 % damping 303
Main Index
371
CHAPTER 9
Response Spectrum Analysis
+, 1200.0, 0.5, 3000.0, 0.5, endt
And so on for SPECSEL 72 and corresponding Tabled1 201, 202 and 203
…….…… for SPECSEL 73 and corresponding Tabled1 301, 302 and 303
Current Nastran implementation support only Large Mass Method.
The Large Mass Method is same as Large Mass Method for Base Excitation in Frequency Response
Analysis or Transient Analysis.
Assigned Large Mass at point of application(grid)
Use SUPORT entry to define all DOF subjected to Response Spectrum Input.
If weight unit is used for density, make sure we use conversion factor (1/G) using Param,Wtmass,
Conv_Factor to convert Weight to Mass unit.
Param, Wtmass, .002589 (1/386.4).
Conm2,199,99,,1.e8
Suport,99,123
Consider three cases:
Subcase 1: Spectrum input in direction Y only.
Subcase 2: Spectrum input in direction X and Z
Subcase 3: Spectrum input in direction X, Y and Z
Use DLOAD to define the Spectrum Input. Note Factor of 386.4 since spectrum input curves are defined
in G units.
Format for DLOAD:
DLOAD, SID, Factor, S1, L1, S2, L2, S3, L3
Note that on Dload and Si, Li entries the DLOAD entry in response spectrum analysis is used to select
DTI entry with name SPECSEL. The DLOAD entry MUST CONTAIN ‘r’ pair of Si, Li entries where
‘r’ is number of DOF listed on SUPORT entry. In addition, the Li, Si pairs (i=1,2,…,6) correspond to
the components(in ascending order) entered on the SUPORT entry. For example, for SUPORT,99,236
then L1,S1 correspond to DOF 2, L2,S2 correspond to DOF 3 and S3,L3 correspond to DOF 6.
For subcase 1 (input in Y direction only) we will have:
DLOAD, 701, 386.4, 0.0, 71, 1.0, 72, 0.0, 73
The SID 701 will be selected by DLOAD=701 in Case Control Section under subcase 1
For subcase 2 (input in direction X + Z)
DLOAD, 702, 386.4, 1.0, 71, 0.0, 72, 1.0, 73
Note: Each SPECSEL may have different number of Tabled1 for different damping values.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
372
And for subcase 3 (Input in X + Y + Z)
DLOAD, 703, 386.4, 1.0, 71, 1.0, 72, 1.0, 73
Also eigenvalue extraction method and number of modes etc need to be specified. ONLY MASS
Normalization is valid in Response Spectrum Analysis.
EIGRL, 66, 0., 5000.
Specify Modal Damping (selected by SDAMP=88 in case control)
TABDMP1, 88, Crit
+, 0.0, .03, 100.0, .05, 1500., 0.05, 2000., .02
+, 5000., 0.02, Endt
Some parameters are needed to complete the analysis.
Param,SCRSPEC,0 Tells MSC.Nastran to perform shock spectrum analysis.
Param, Option will select the method of combining modal responses.
There are three different options available:
Param, Option, ABS (add absolute modal response)
Param, Option, SRSS (add using Square Root of Sum of the Square)
Param Option, NRL (add based on NAVSEA0908LP0003010 Specification)
With Param, (SRSS and NRL), another parameter may be defined. Param,CLOSE,Fraction
(Fraction=0.0 to 1.0  Default 1.0)
PARAM, CLOSE, 0.1
With NRL/SRSS option, close natural frequencies will be summed by the ABS convention, where close
frequencies meet the inequality F(i+1) < CLOSE * Fi and then combined with other modal response in
accordance with the param,option.
Param,Option and Param,Close may be set differently in any subcase, allowing summation by different
method in a single run.
Putting it all together, Listing 96 shows the complete input file. See
Install_dir/mdxxxx/doc/dynamics/shock3xyz.dat for a copy.
Listing 96 Multiple Spectra Example
$ Nastran Input Data Deck
ID Mohan, Barbela
TIME 100
$
$ Must be SOL 103
$ =================
$
SOL 103 $ Modal Frequency Response
$
CEND
$
Main Index
373
CHAPTER 9
Response Spectrum Analysis
TITLE = CANTILEVERED BEAM MADE OF PLATES
$
echo = unsort
$
stress = all
force = all
acceleration = all
disp = all
$
spc = 77
METHOD = 66
SDAMP = 88
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
MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
374
$
$
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
375
CHAPTER 9
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.e6 0.
$
enddata
Main Index
MD Nastran Dynamic Analysis User’s Guide
Response Spectrum Examples
376
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 10: Restarts In Dynamic Analysis
10
Restarts In Dynamic Analysis
Overview
Examples
Remarks
Main Index
MD Nastran Dynamic Analysis User’s Guide
Overview
378
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 “solutiontype” 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., nonsuperelement), 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
379
CHAPTER 10
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 101 summarizes this series of nine runs along with a
brief description. Listings of the ten runs are also included (Listing 101 through Listing 1010). 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.
Table 101 Typical Series of Restart Runs
Run
Sequence
Number
Name of
Input File
Solution
Sequence
Number Description of Runs
Version
Created
Version
Deleted
1 bd10bar1.dat 103 Perform a normal modes cold start analysis
and save the database.
1 None
2 bd10bar2.dat 103 Restart from run number 1 and request
eigenvector output.
2 None
3 bd10bar3.dat 103 The first two modes of the structure are very
close to one of the forcing frequencies. The
structure is modified in order to stay away
from resonance. This run restarts from run
number 2 to delete the old PBAR entry and
replace it with the modified PBAR entry. The
modes are recalculated, and the eigenvector
output is requested in this run.
3 2
4 bd10bar4.dat N/A This run deletes the data blocks in the database
previously occupied by Version 1. This space
can then be reused by future restarts. Note that
the statement DBCLEAN does not reduce the
database size. It merely removes some of its
contents so that this space can be reused. A
new version is not created as a result of this
run. This is an optional run especially if disk
space is of no concern to you.
None 1
Main Index
MD Nastran Dynamic Analysis User’s Guide
Examples
380
5 bd10bar5.dat 112 This is a transient restart run from the modes
saved in Version 3. The applied load is a unit
step function. The modes calculated in run
number 3 are also saved at the end of this run.
Since the calculation of the modes is the most
expensive operation in a dynamic analysis, it
is probably a good idea to save Version 3 once
you have confidence in the results. This way
you can always restart from this version.
Partial output is shown at the top of
Figure 101. A 1% critical damping value is
applied to the structure.
4 None
6 bd10bar6.dat 112 This is another transient restart run using the
solution from run number 5. The purpose of
this run is to request additional output. Partial
output is shown.
5 4
7 bd10bar7.dat 112 This is another restart run from Version 6 with
a different load condition (triangular pulse).
Partial output is shown in Figure 102. In this
case, you can just as easily restart from
Version 3.
6 None
8 bd10bar8.dat 111 This is a frequency response restart run. Note
that this restart is from Version 3 using a read
only restart saved in the database from the
modes run. A new database called bd10bar8 is
created. Partial output is shown in Figure 103.
A 2% critical damping value is applied to the
structure.
None None
9 bd10bar9.dat N/A This is a database directory printout run. As
shown in Figure 103, there are six versions in
the database. Run numbers 4, 8 and 9 did not
create any new versions. Versions 3, 5 and 6
are restartable. This is an optional run.
None None
10 bd10bar10.da
t
112 This is a restart run from Version 5 with an
extension of the time steps without having to
recompute output the earlier times, partial
output is shown at the bottom of Figure 101.
Note times start at .6.
7 5
Table 101 Typical Series of Restart Runs (continued)
Run
Sequence
Number
Name of
Input File
Solution
Sequence
Number Description of Runs
Version
Created
Version
Deleted
Main Index
381
CHAPTER 10
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 nonsuperelement 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
nonsuperelement 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.
Listing 101 Input File for Normal Modes Run
$ FILE  bd10bar1.dat
$
$ NORMAL MODES RUN
$
ID CANT BEAM
SOL 103
CEND
TITLE = CANTILEVER BEAM  NORMAL MODES  COLD START RUN
SPC = 1
METHOD = 10
$
BEGIN BULK
$
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.
EIGRL 10 0.1 50.
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.
MAT1 1 7.1+10 .33 2.65+4
PARAM AUTOSPC YES
PARAM WTMASS .102
PBAR 1 1 6.1584 3.8 3.8 6.8 2.414
SPC1 1 123456 1
Main Index
MD Nastran Dynamic Analysis User’s Guide
Remarks
382
$
ENDDATA
Listing 102 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 103 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,1,1,6.1584,2.98,3.18,6.8,2.414
$
ENDDATA
Main Index
383
CHAPTER 10
Restarts In Dynamic Analysis
Listing 104 Input File for Cleaning a Database
$ FILE  bd10bar4.dat
$
assign master=’bd10bar1.MASTER’
dbclean version=1
endjob
Listing 105 Input File for Transient Response
$ FILE  bd10bar5.dat
$
$ THIS IS A TRANSIENT RESTART RUN FROM THE MODES
$ CALCULATED BY THE RUN “bd10bar3.dat”
$
RESTART VERSION=3,KEEP
ASSIGN MASTER=’bd10bar1.MASTER’
ID CANT BEAM
SOL 112
CEND
TITLE = TRANSIENT RESTART  UNIT STEP FUNCTION INPUT
SUBTITLE = REQUEST DISPLACEMENT TIME HISTORY AT GRID POINT 11
SPC = 1
METHOD = 10
SET 1 = 11
DISP = 1
SUBCASE 1
SDAMP = 100
TSTEP = 100
DLOAD = 100
$
BEGIN BULK
$
$ ADDITIONAL ENTRIES FOR DYNAMIC LOADS
$ FOR UNIT STEP FUNCTION
$
$ SID DAREA DELAY TYPE TID
TLOAD1 100 101 102
$
DAREA,101,11,3,1.0
$
TABLED1,102,,,,,,,,+TBL1
+TBL1,0.0,0.0,.001,1.0,10.0,1.0,ENDT
$
$ TRANSIENT TIME STEPS
$
$ SID N(1) DT(1) NO(1)
TSTEP 100 600 .001 5
$
$ MODAL DAMPING TABLE
$
TABDMP1,100,CRIT,,,,,,,+TDAMP
+TDAMP,0.,.01,200.,.01,ENDT
$
ENDDATA
Main Index
MD Nastran Dynamic Analysis User’s Guide
Remarks
384
Listing 106 Input File for an Additional Output Request
$ FILE  bd10bar6.dat
$
$ THIS IS ANOTHER TRANSIENT RESTART RUN. THE PURPOSE
$ OF THIS RUN IS TO REQUEST ADDITIONAL OUTPUT.
$
RESTART
ASSIGN MASTER=’bd10bar1.MASTER’
ID CANT BEAM
SOL 112
CEND
TITLE = T R A N S I E N T R E S T A R T
SUBTITLE = U N I T S T E P F U N C T I O N I N P U T
SPC = 1
METHOD = 10
SET 1 = 11
SET 2 = 10
ACCE = 2
SUBCASE 1
SDAMP = 100
TSTEP = 100
DLOAD = 100
$
$
$ PLOT RESULTS
$
$...XY plot commands ...
$
BEGIN BULK
$
$
ENDDATA
Listing 107 Input File for an Additional Transient Load
$ FILE  bd10bar7.dat
$
$ THIS IS ANOTHER TRANSIENT RESTART RUN USING
$ A DIFFERENT LOAD CONDITION. NOTE THAT SINCE
$ THERE ARE NO MODEL CHANGES, THE SAME MODES
$ WERE USED FROM THE DATABASE FOR THE RESPONSE
$ CALCULATIONS.
$
RESTART KEEP
ASSIGN MASTER=’bd10bar1.MASTER’
ID CANT BEAM
SOL 112
CEND
$
$ NOTE THAT TITLE CHANGES HAVE NO EFFECT
$ ON SOLUTION PROCESS, THEY ONLY CHANGE THE
$ PRINTOUT TITLE
$
TITLE = T R A N S I E N T R E S T A R T
SUBTITLE = TRIANGLE PULSE  1.0 AT T=0 AND 0.0 AFTER .2 SEC
SPC = 1
METHOD = 10
SET 1 = 11
DISP = 1
Main Index
385
CHAPTER 10
Restarts In Dynamic Analysis
SUBCASE 1
SDAMP = 100
TSTEP = 100
DLOAD = 300
$
$ PLOT RESULTS
$
$...XY 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 108 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
MD Nastran Dynamic Analysis User’s Guide
Remarks
386
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
Listing 109 Input File to Print the Database Dictionary
$ FILE  bd10bar9.dat
$
assign master=’bd10bar1.MASTER’
dbdir
endjob
Main Index
387
CHAPTER 10
Restarts In Dynamic Analysis
Listing 1010 Input File to Extend Time Steps
$
$ FILE  bd10bar10.dat
$
$ THIS IS ANOTHER TRANSIENT RESTART RUN. THE PURPOSE
$ OF THIS RUN IS TO REQUEST ADDITIONAL OUTPUT.
$
RESTART Version=5
ASSIGN MASTER='bd10bar1.MASTER'
ID CANT BEAM
SOL 112
CEND
TITLE = T R A N S I E N T R E S T A R T
SUBTITLE = E X T E N D T S T E P U S I N G S T I M E
SPC = 1
METHOD = 10
SET 1 = 11
SET 2 = 10
ACCE = 2
SUBCASE 1
SDAMP = 100
TSTEP = 700
DLOAD = 100
$
$
$ PLOT RESULTS
$
OUTPUT(XYPLOT)
XAXIS = YES
YAXIS = YES
XGRID = YES
YGRID = YES
XTITLE = T I M E ( S E C )
YTITLE = Z D I S P O F G R I D 11 (M)
XYPLOT DISP /11(T3)
YTITLE = Z ACCELERATION OF GRID 10 ( M / SEC**2 )
XYPLOT ACCE /10(T3)
$
BEGIN BULK
$
PARAM,STIME,.6
TSTEP,700,700,.001,5
$
ENDDATA
Main Index
MD Nastran Dynamic Analysis User’s Guide
Remarks
388
Figure 101 Partial Output from Transient Analysis with Unit Step Function Input (bd10bar5
and bd10bar10)
Figure 102 Partial Output from Transient Analysis with a Triangular Pulse (bd10bar9)
POINTID = 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
5.000000E03 G 1.194303E16 3.936567E17 2.504727E05 0.0 6.120401E05 4.001821E16
1.000000E02 G 2.737714E16 6.817522E17 8.184779E05 0.0 1.331090E04 8.292225E16
2.300000E01 G 3.588051E14 1.043014E14 7.835704E03 0.0 3.673377E03 1.088909E13
2.350000E01 G 3.580078E14 1.044822E14 7.866833E03 0.0 3.710257E03 1.086752E13
2.400000E01 G 3.609918E14 1.051499E14 7.883687E03 0.0 3.728710E03 1.095553E13
2.450000E01 G 3.662778E14 1.059462E14 7.889598E03 0.0 3.744088E03 1.111338E13
2.500000E01 G 3.699655E14 1.062077E14 7.889201E03 0.0 3.778521E03 1.121871E13
2.550000E01 G 3.697565E14 1.058640E14 7.877129E03 0.0 3.814069E03 1.120912E13
5.950000E01 G 1.870177E14 4.673678E15 4.031476E03 0.0 2.030503E03 5.616841E14
6.000000E01 G 1.959096E14 4.961804E15 4.248531E03 0.0 2.099409E03 5.888898E14
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./ XMINFRAME/ XMAXFRAME/ YMINFRAME/ X FOR YMAXFRAME/ X FOR
ID TYPE NO. PANEL : GRID ID ALL DATA ALL DATA ALL DATA YMIN ALL DATA YMAX
1 DISP 1 11( 5) 6.000000E01 1.300000E+00 6.125921E04 9.450001E01 7.731026E03 7.200000E01
6.000000E01 1.300000E+00 6.125921E04 9.450001E01 7.731026E03 7.200000E01
1 ACCE 2 10( 5) 6.000000E01 1.300000E+00 7.251530E01 7.300000E01 6.462884E01 9.200000E01
6.000000E01 1.300000E+00 7.251530E01 7.300000E01 6.462884E01 9.200000E01
POINTID = 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
5.000000E03 G 1.187877E16 3.927214E17 2.489746E05 0.0 6.077061E05 3.977257E16
1.000000E02 G 2.680967E16 6.663936E17 8.054515E05 0.0 1.304303E04 8.105531E16
1.650000E01 G 1.905787E14 5.570934E15 4.101126E03 0.0 1.843881E03 5.796270E14
1.700000E01 G 1.956794E14 5.692628E15 4.179738E03 0.0 1.906047E03 5.953207E14
1.750000E01 G 1.976381E14 5.703682E15 4.256193E03 0.0 2.003543E03 6.003546E14
1.800000E01 G 1.997031E14 5.702407E15 4.310292E03 0.0 2.079599E03 6.055504E14
1.850000E01 G 2.037484E14 5.747392E15 4.327041E03 0.0 2.091791E03 6.172204E14
1.900000E01 G 2.075757E14 5.765805E15 4.314581E03 0.0 2.086080E03 6.278964E14
1.950000E01 G 2.070708E14 5.747701E15 4.273104E03 0.0 2.050172E03 6.267379E14
5.950000E01 G 1.196185E14 3.004771E15 2.607365E03 0.0 1.203624E03 3.601540E14
6.000000E01 G 1.261992E14 3.250992E15 2.775512E03 0.0 1.249155E03 3.805073E14
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./ XMINFRAME/ XMAXFRAME/ YMINFRAME/ X FOR YMAXFRAME/ 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 6.000000E01 4.039584E03 4.200000E01 4.327041E03 1.850000E01
0.000000E+00 6.000000E01 4.039584E03 4.200000E01 4.327041E03 1.850000E01
Main Index
389
CHAPTER 10
Restarts In Dynamic Analysis
Figure 103 Partial Output from Frequency Response Analysis (bd10bar8)
Figure 104 Partial Output from a Database Directory Run (bd10bar9)
POINTID = 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 1.906914E15 5.344224E16 4.089051E04 0.0 2.044525E04 5.763767E15
180.0000 180.0000 0.0 0.0 180.0000 180.0000
5.000000E01 G 2.017782E15 5.676515E16 4.327695E04 0.0 2.154499E04 6.100925E15
179.4355 179.4089 359.4349 0.0 179.4725 179.4328
2.073000E+00 G 3.901148E14 8.436090E15 8.402392E03 0.0 3.875251E03 1.167835E13
123.5868 98.5582 303.5746 0.0 123.7787 122.7733
2.087000E+00 G 4.417046E14 9.296475E15 9.514233E03 0.0 4.382797E03 1.323076E13
108.1114 89.8882 288.0990 0.0 108.3053 107.5244
2.100632E+00 G 4.612975E14 9.812769E15 9.936970E03 0.0 4.572159E03 1.383304E13
90.0803 76.7938 270.0676 0.0 90.2761 89.6427
2.224000E+00 G 1.434210E14 3.595956E15 3.091593E03 0.0 1.406759E03 4.331211E14
19.3916 17.3197 199.3766 0.0 19.6055 19.3040
2.347000E+00 G 7.249022E15 1.926738E15 1.563781E03 0.0 7.030737E04 2.195193E14
10.2994 9.4297 190.2818 0.0 10.5329 10.2551
9.950000E+01 G 4.810872E18 4.064172E19 4.350609E07 0.0 6.019894E07 5.027250E18
181.2857 293.2593 187.6541 0.0 144.7264 333.6623
1.000000E+02 G 4.755274E18 4.680135E19 4.165692E07 0.0 6.826717E07 5.581692E18
181.2771 302.8268 188.1974 0.0 148.2450 335.0572
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./ XMINFRAME/ XMAXFRAME/ YMINFRAME/ X FOR YMAXFRAME/ X FOR
ID TYPE NO. PANEL : GRID ID ALL DATA ALL DATA ALL DATA YMIN ALL DATA YMAX
1 DISP 1 11( 5,) 1.000000E01 1.000000E+02 2.536850E07 6.250000E+01 9.936970E03 2.100632E+00
0.000000E+00 1.000000E+02 2.536850E07 6.250000E+01 9.936970E03 2.100632E+00
1 DISP 1 11(, 11) 1.000000E01 1.000000E+02 1.815511E+02 5.500000E+00 3.594349E+02 5.000000E01
0.000000E+00 1.000000E+02 0.000000E+00 0.000000E+00 3.594349E+02 5.000000E01
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
PROJECT_ID ASSIGNED INT. VALUE VERSION_ID CREATION TIME
   
" B L A N K " 1 ** 1 5/28/10 14:14.35
** 2 5/28/10 14:14.47
3 5/28/10 14:14.57
** 4 5/28/10 14:15.16
5 5/28/10 14:15.26
6 5/28/10 14:15.34
Main Index
MD Nastran Dynamic Analysis User’s Guide
Remarks
390
Main Index
MD Nastran Dynamic Analysis User’s Guide
Chapter 11: Coupled FluidStructure Analysis
11
Coupled FluidStructure Analysis
Overview
Axisymmetric Hydroelastic Analysis
Virtual Fluid Mass
Coupled and Uncoupled Acoustic Analysis
Exterior Acoustics
BW Method of FluidStructure 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
MD Nastran Dynamic Analysis User’s Guide
Overview
392
Overview
Four major methods are available in MD Nastran to model fluid effects. These methods are described
below.
Axisymmetric Hydroelastic Analysis
Small motions of compressible or incompressible fluids coupled to a structure may be analyzed with this
option. The fluid is modeled with axisymmetric hydroelastic elements (CFLUIDi), which may be
connected to an arbitrary structure modeled with an axisymmetric wetted surface. Each fluid point
(RINGFL) on a cross section defines the scalar pressure, which is expanded to a Fourier series around
the circumference. Normal modes, complex modes and frequency response solutions are available for the
coupled fluidstructure problems. All solutions may include gravity effects (i.e., sloshing) on a free
surface. This capability was developed specifically to analyze liquidfueled booster rockets but may also
be useful for problems involving other types of axisymmetric storage tanks.
Virtual Fluid Mass
Small motions of incompressible fluids may be coupled directly to the structure with this option. Fluids
may be coupled to the interior and exterior surfaces (with infinite fluid boundaries). There is no explicit
fluid model; only the wetted structural elements (ELIST) have to be defined. Although free surfaces are
allowed, no gravity effects are included directly. Since the fluid is represented by a coupled mass matrix
attached directly to the structural points, this capability is allowed in all dynamic solution sequences.
This capability may be used to model a wide variety of fluidstructure interaction problems. Some
examples are fuel tanks, nuclear fluid containers, drilling platforms, and underwater devices.
Coupled and Uncoupled Acoustics
You may analyze the dynamics of compressible fluids contained within arbitrarily shaped structures with
the coupled fluidstructure method. You would model a threedimensional fluid with conventional solid
elements (CHEXA, etc.) using acoustic property and material data. Each grid point in the fluid defines
the pressure at its location as its degree of freedom. The fluid is automatically connected to the structure
via the geometry and ACMODL Bulk Data inputs. You can connect acoustic absorber elements
(CHACAB) to the structural surfaces to simulate soundproofing material. In addition, the CAABSF
element is now available. This element acts as a thin layer of acoustic absorbing material along the
fluidstructure interface. Acoustics has introduced several features, which are parallel to those available
for structural analysis, such as direct damping, modal damping and the ability to control the modes in a
response analysis through the use of parameter. You can define panels to provide integrated response
data. Effects of gravity, large motions, and static pressures are ignored. Complex eigenvalues, frequency
response, and transient response are the available solution sequences. Design sensitivity and
optimization processes may reference the acoustic outputs as responses with appropriate design
constraints. Applications for the coupled fluidstructure option are automotive and truck interiors,
aircraft cabins, and acoustic devices, such as loudspeakers and microphones.
Main Index
393
CHAPTER 11
Coupled FluidStructure 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 twodimensional 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 threedimensional 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
MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
394
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 degreesoffreedom 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 rz 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 fluidstructure 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 smallmotion pressure formulation and also produces unsymmetric
matrices. For more generality, the acoustic fluid model is defined by threedimensional 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
395
CHAPTER 11
Coupled FluidStructure 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
largeorder 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 rz 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 and (i.e., harmonic pressures), describing the pressure function, , in the
equation
where the set of harmonics , and 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
degreesoffreedom 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 structurefluid 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.
• XY 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 XY 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.
p
n
p
n*
P o ( )
P o ( ) P
0
p
n
no cos
n 1 =
N
¯
p
n*
no 0 N 100 < < sin
n 1 =
N
¯
+ + =
0 n n*
o
o
Main Index
MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
396
• 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 fluidrelated entries are present. The data references a fluidrelated 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
397
CHAPTER 11
Coupled FluidStructure 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 zaxis upward and the raxis to the right. Plot the locations of the fluid points on the
righthand 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 zaxis 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 111 illustrates a typical application of the free surface and structural boundary lists.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
398
Figure 111 Examples of Boundary Lists
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 and the second plane of
symmetry is assumed at where 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.
FLUID
Structur
Boundary
List #2
Free Surface List
Axis of
Symmetry
Structure
Boundary
List #1
FSLIST: 36, 37, 38, 39, 40
BDYLIST #1: AXIS, 22, 23, 29, 36
BDYLIST #2: 40, 35, 28, 21, 14, 7, 6, 5, 4, 3, 2, 1, AXIS
1 2 3 4 5 6
22
29
36
23
37 38 39
40
35
28
21
14
7
o 0.0 =
o 360° M = M
Main Index
399
CHAPTER 11
Coupled FluidStructure Analysis
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 degreesoffreedom. 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 fluidstructure interface is mathematically described by a set of unsymmetric matrices. Since
most solution sequences are restricted to the use of symmetric matrices, the fluidstructure
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 structurefluid 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.
Series Plane 1
Plane 2
S A
Cosine S
A
0,2,4,...
none
1,3,5,...
none
Sine
(*)
S
A
none
1,3,5,...
none
2,4,6,...
Main Index
MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
400
SOL 108 or 200 – Direct Frequency and Random Response
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 fluidstructure 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 and ) 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 fluidstructure 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.
Hydroelastic Data Processing
The fluidrelated Bulk Data entries are converted by the program into equivalent grid point, scalar point,
element connection, and constraint data entry images. Each specified harmonic, , of the Fourier series
solution produces a complete set of special grid point and connection entry images. In order to retain
unique identification numbers, the user identification numbers are encoded by the algorithm below:
RlNGFL Points
where:
CFLUIDi Connection Entries
= cosine series
= sine series
e
3
g
N
Grid point ID User ring ID 1,000,000 I
N
× + =
I
N
N 1 +
I
N
N 1 2 +
Element ID User ring ID 1,000 I
N
× + =
Main Index
401
CHAPTER 11
Coupled FluidStructure Analysis
where is defined above for each harmonic .
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:
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, XY
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 zerogravity field, the pressures are
constrained to zero. For this case, the internally generated set of singlepoint constraints are internally
combined with any userdefined structural constraints and will always be automatically selected.
If pressures at points in the fluid (PRESPT) or gravitydependent 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 userdefined 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.
Sample Hydroelastic Model
Table 111 contains a list of the input data for a sample hydroelastic problem. See
Install_dir/mdxxxx/doc/dynamics/bd11d107d2.dat. Figure 111 describes the problem and lists the
parameters. The relatively small number of grid points were chosen for purposes of simplicity and not
accuracy. The symbols for the fields in the hydroelastic data entries are placed above each group.
Structural data entries are included in their standard forms. The explanations for the data are given in
the following notes:
Harmonic ID
0 1,000,037
1* 1,500,037
1 2,000,037
2* 2,500,037
2 3,000,037
I
N
N
Main Index
MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
402
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 degreesoffreedom.
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 fluidstructure 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 onequarter symmetry was used for the structure requires the use of the FLSYM
entry. Symmetricantisymmetric boundaries indicate that only the cosine terms for the odd
harmonics interact with the structure. If symmetricsymmetric 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.
Table 111 Sample Hydroelastic Problem
$ FLUID MASS DEMO
SOL 107
TIME 2
CEND
1. TITLE = SAMPLE HYDROELASTlC PROBLEM
SUBTITLE = EIGENVALUE ANALYSIS WITH FLEXIBLE BOUNDARY
AXISYM = FLUID
SPC = 3
CMETHOD = 1
OUTPUT
2.
PRESS = ALL
HARMONICS = ALL
ELFORCE = ALL
BEGIN BULK
BULK DATA FIELD
p
Main Index
403
CHAPTER 11
Coupled FluidStructure Analysis
3. $AXIF,CID,G,DRHO,DB,NOSYM
AXIF,2,32.2,0.03,,NO
$,N
1
,N
2
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
Table 111 Sample Hydroelastic Problem (continued)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Axisymmetric Hydroelastic Analysis
404
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
Table 111 Sample Hydroelastic Problem (continued)
Main Index
405
CHAPTER 11
Coupled FluidStructure Analysis
Figure 112 Sample Hydroelastic Problem
7 8
9 10 11 12
6 5 4 3 2 1
17 16 15 14
13
17
16
15
14
9 3
10 4
11 5
12 6
Symmetric
Anti
symmetric
Free Surface
Fluid
Points
Structure
Points
Fixed
r
4.0 4.0
5.0
5.0
Fluid:Density p = 0.03
Bulk Modulus B = 1
Gravity g = 32.2
Structure:Thickness t = 0.5
Density r = 0.05
Main Index
MD Nastran Dynamic Analysis User’s Guide
Virtual Fluid Mass
406
Virtual Fluid Mass
A virtual fluid volume produces a mass matrix which represents the fluid coupled to a boundary
consisting of structural elements and other effects, such as free surfaces, planes of symmetry, and infinite
fluids. The incompressible fluid produces a mass matrix defined with full coupling between
accelerations and pressures on the flexible structural interfaces. For each requested volume, the
boundaries may be combinations of the following:
1. Structural finite element faces, with one or two wetted sides.
2. Free surfaces with zero pressure, but no sloshing effects.
3. Planes of symmetry with symmetric or antisymmetric motion.
4. Infinite boundaries, by default, on nonclosed volumes.
The virtual mass method is well suited for the following problem types:
1. An infinite fluid surrounding part of a structure.
2. A fluid with a free surface contained within a flexible structure.
3. Multiple fluid volumes with combinations of the options above such as a half of a ship floating in
water with several internal tanks, each with free surfaces and symmetric boundaries.
Note that compressibility and surface gravity effects are neglected. It is assumed that the important
frequency range for the structural modes is above the gravity sloshing frequencies and below the
compressible acoustic frequencies. It is further assumed that the density within a volume is constant and
no viscous (rotational flow) or aerodynamic (high velocity) effects are present. In other words, a ship
traveling at a high rate through a thick oil patch would require some extra modeling effort.
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
407
CHAPTER 11
Coupled FluidStructure Analysis
Figure 113 Virtual Mass Terminology
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...
Fluid
Volume 2
Fluid
Volume 1
Free Surfaces
Element
List #2
Element
List #1
Structure
Main Index
MD Nastran Dynamic Analysis User’s Guide
Virtual Fluid Mass
408
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
antisymmetry. 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 xy 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 is the value of a point source of fluid (units are volume flow rate per area) located at location ,
and is assumed acting over an area , the vector velocity at any other point is
(111)
o
j
r
j
A
j
u
·
i
r
i
u
·
i
o
j
e
i j
r
i
r
j
–
2
 A
j
d
A
j
í
j
¯
=
Main Index
409
CHAPTER 11
Coupled FluidStructure Analysis
where is the unit vector in the direction from point to point . Note that the gradient of the vector
is the potential function which satisfies Laplace’s Equation on a term by term basis.
The other set of necessary equations are the pressures, , at any point, , in terms of the density, ,
sources and geometry, namely
(112)
The results of integrating Eqs. (111) and (112) over the finite element surfaces are collected
respectively in two matrices, and where
(113)
and
(114)
where are the forces at the grid points. The matrix is obtained by integrating Eq. (112). An
additional area integration is necessary to convert the pressures to forces. A mass matrix may now be
defined using Eqs. (113) and (114) as
(115)
where the virtual fluid mass matrix, is
(116)
Singularities for Enclosed Volumes
Note that if is singular or nearly singular, the mass matrix will cause problems with the coupled
solution. The singularity is active when a set of nonzero sources can produce zero velocities on the
structural surfaces. This occurs when the fluid is completely enclosed by the structure and with planes
of only symmetric motion. Large pressures could occur within the fluid volume from an infinitesimal
volume change on the incompressible fluid.
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, , is added near the
center of each MFLUID volume. The expanded versions of Eqs. (113) and (114) are
(117)
(118)
e
i j
j i
u
·
i
p
i
i p
p
i
po
·
j
e
i j
r
i
r
j
–
 A
j
d
A
j
í
j
¯
=
;   A  
u
·
{ } ;   o   =
F { } A   o
·
{ } =
F A  
F { } M
f
  u
··
{ } =
M
f
 
M
f
  A   ;  
1 –
=
;  
;  
s
2
u
·
{ } ;   o { } ;
2
{ }s
2
+ =
F { } A   o
·
{ } A
2
{ }s
2
+ =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Virtual Fluid Mass
410
An additional equation provides that the sum of the squares of the ordinary sources in the vector,
is minimized and a set of Lagrange multipliers, are used to enforce Eq. (117) as constraints. The
function, , to be minimized is defined as
(119)
Taking the derivatives of with respect to , , and , we obtain three sets of equations. Combining
them, we obtain the following matrix equation:
(1110)
The vector may be eliminated by solving the second row partition and substituting into the first row
partition of the matrix, with the result:
(1111)
Equation (1111) may be solved for and , which are then used to obtain pressure. To obtain the
pressure, we substitute for in Eq. (118) to obtain the matrix equation
(1112)
Equations (1111) and (1112) may then be combined into a single matrix defining the fluid, namely
(1113)
where:
(1114)
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.
o
1
{ }
ì
U
U
1
2
 o  
T
o { } ì  
T
u
·
{ } ;   o { } ;
2
{ }s
2
– – { } + =
U ì o
1
s
2
0 ; ;
2
;
T
I – 0
;
2
T
0 0
ì
o
s
2
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹ u
·
0
0
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
o { }
;;
T
;
2
;
2
T
0
ì
s
2 ¹ )
´ `
¦ ¹ u
·
0
¹ )
´ `
¦ ¹
=
ì { } s
2
o { }
p { }
A;
T
A
2
ì
·
s
·
2 ¹ )
´ `
¦ ¹
=
F { } M
f
  u
··
{ } =
M
f
 
A;
T
A
2
;;
T
;
2
;
2
T
0
1 –
I
0
=
Main Index
411
CHAPTER 11
Coupled FluidStructure 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
(1115)
where is a boundary grid point and and 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.
Using Phantom Structural Boundaries
In many cases an ELIST boundary is desired where no CQUAD4 or CTRIA3 elements exist. Examples
occur:
• When grid points connected by other element types, such as CHEXA, CSHEAR, CBEAMs,
form the boundary.
• When only a small hole connects two separate fluid volumes (since poor numerical conditioning
(less roundoff) is expected with a single MFLUID).
• If the fluid contains two free surfaces due to entrapped air at a different pressure.
• If the fluid volume is a complex labyrinth, such as a boiler made with many tubes.
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 inplane 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, , calculated from
A
i
n
i
i
¯
u
i
· 0 =
i A
i
u
i
K
i
Main Index
MD Nastran Dynamic Analysis User’s Guide
Virtual Fluid Mass
412
(1116)
where is the area under the point, is the density, and 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 114. The categories are GOOD
= legal configuration, BAD = illegal boundary, and MAYBE = conditionally legal, which could be
permitted if a phantom boundary is used.
K
i
A
i
pg · =
A
i
p g
Main Index
413
CHAPTER 11
Coupled FluidStructure Analysis
Figure 114 Fluid Combinations
Fs
GOOD BAD MAYBE
Air
Air
Fs
Air
Fs Fs
Fs
Fs
Fluid
Fluid
Fs
Air
Fluid
Hole
Infinite
Fluid
Free
Surfaces
Internal
Only
PB = Phantom
Boundary Between 2
Fluids
Ai
Fluid
Fluid Fluid
PB
Fluid
Fs
Fluid
PB
Fluid
Fluid
PB
Fluid
Fluid Fluid
Main Index
MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
414
Coupled and Uncoupled Acoustic Analysis
Acoustic analysis in MD Nastran is used to determine vibrations in structural cavities filled with fluid or
air. The boundaries of the cavities may be defined as rigid, open, damped, or flexible as defined by
constraints and finite elements. Typical applications are automobile or aircraft interiors, musical
instruments, loudspeaker boxes, and solid rocket motors. Any type of problem that involves small, linear
motions of the fluid and structure may be a candidate. The analyst may obtain natural frequencies and
damping factors along with mode shapes for both the structure and the coupled fluid. Other solution
sequences may be used to obtain results for forced response in the frequency or time domain.
The analyst starts with the geometry of the the acoustic cavity boundary. The interior fluid volume is
modeled in three dimensions with conventional MD Nastran solid finite elements connected to a special
set of GRID points. Boundary motions are included if the structure lies on the boundary, or fixed if no
structural elements are connected (by default). The structural interface simply requires that finite
elements lay on the exterior fluid boundaries. Note that coupled interactions between the fluid and
structural model are generated internally using the geometry definitions of the elements of the two
models and require very little user intervention.
Other types of boundaries that are available are free surfaces, acoustic barriers, and acoustic absorbers.
The free boundary is modeled by simply constraining the fluid pressure degreesoffreedom to zero. The
barrier and absorber elements provide a spring/mass/damper system between the fluid and the structure.
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.
Building the Models
The following steps are recommended for avoiding difficulties later.
1. Define the geometry and select the mesh sizes.
The first step in the modeling process is to estimate the overall FE mesh size and the level of detail
needed in the geometric approximations. A set of threedimensional elements and special fluid
GRID points (with CD = –1) are used to define the fluid while normal structural elements and
points define the container. A rule of thumb for fluids is that sixelementsperwave are
acceptable for the highest frequencies exciting the fluid. For acoustics this usually translates into
a corresponding coarse mesh. For incompressible fluids, the only waves occur due to gravity
effects on the free surface. For heavy liquids, the mass effects are usually more important than
with air, and the fluid FE mesh should be modeled to match the structural mesh.
Main Index
415
CHAPTER 11
Coupled FluidStructure Analysis
It is generally easier to build the threedimensional 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 threedimensional 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 degreeoffreedom defining the
pressure at a point at a fixed location. The fluid itself is modeled with existing threedimensional
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).
Userinput 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 higherorder 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 highorder and
loworder 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
MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
416
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 degreesoffreedom 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 standalone 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 userselected 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 fluidstructure 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
417
CHAPTER 11
Coupled FluidStructure 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 3D 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 degreeoffreedom per point to the model a
threedimensional 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 nonboundary 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
MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
418
For moderately large problems, the ASET/OMIT method of reduction is available for both the
fluid and structural degreesoffreedom. However, the accuracy is highly dependent on the
analyst’s choice of retained degreesoffreedom. 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 3D 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 .
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.
Running the Jobs
Several diagnostics will be generated automatically by MD Nastran. Examples are shown in the
following pages.
The following special output is produced by the program when acoustic boundaries are encountered. The
first shows the automatic connections found by matching GRID locations:
2t
Main Index
419
CHAPTER 11
Coupled FluidStructure Analysis
The following message is a sample of a case when the boundaries do not match:
and:
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.
*******************************************************
* *
* MATCHING GRID TABLE *
* *
*******************************************************
* USER GRID IDS. *
* ACOUSTIC GRIDS * STRUCTURAL GRIDS *
* * *
*******************************************************
* * *
* 1 * 1001 *
* * *
* 2 * 1002 *
* * *
* 3 * 1003 *
* * *
* 4 * 1004 *
* * *
* 5 * 1005 *
* * *
* 11 * 1011 *
* * *
ETC
*******************************************************
*** USER WARNING MESSAGE 6151, ACOUSTIC COUPLING
MATRIX WILL NOT BE CREATED FOR FACE = 42 82 83 43
OF ELEMENT ID = 42 ,BECAUSE ITS FLUID GRID POINTS DO NOT HAVE CORRESPONDING
STRUCTURAL GRID POINTS.
USER ACTION: IF COUPLING IS DESIRED, THEN CHECK GRID POINT DATA.
*
*
*
*
ETC....ETC.
^^^ 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.
FL./STR. INTERFACE CHECK
T1 T2 T3 R1 R2 R3
1 2.3554080E18 8.4260993E02 1.4653091E28 6.0271138E10 2.6100844E19 1.8405548E10
Main Index
MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
420
• 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 outputonly
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 freebody 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
degreesoffreedom 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
421
CHAPTER 11
Coupled FluidStructure 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 loadcarrying 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 timedependent 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.
Figure 115 Example of TwoDimensional Automotive Model With Absorbers
Main Index
MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
422
Figure 116 Acoustic Suspension Loudspeaker
Test Problem Description
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 116. 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.
Width: 0.5 M Depth: 0.4 M
Height: 0.6 M Hole Diam. 0.345 M
Box E Modulus: 11.61E9 Box Density: 562 Kg/M**3
Wall thickness: 0.015 M
Main Index
423
CHAPTER 11
Coupled FluidStructure Analysis
Executive and Case Control
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 onedimensional wave solutions. Results for the box were checked
against calculated natural frequencies of simplysupported plates.
Air Wave speed: 344 M/Sec Air Density: 1.11KG/M**3
Cone E Modulus: 3.4E9 Cone Density: 450.0
Cone thickness: 0.1E3 M Cone Depth: 0.04 M
Mode No. Frequency, Hz Type
1 0.0 Air Constant Pressure (fictitious)
2 292 Air 1st z (up/down)
3 297 Box u(y) Back panel
4 351 Air 1st x (Left/Right)
Main Index
MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
424
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
largeorder nonsuperelement 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) foreaft 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
5 359 Box u(x) Sides inPhase
6 379 Box u(x) Sides outofphase
7 450 Box u(z) Top/bottom
8 453 Air 1st y (Fore/Aft)
Mode No. Frequency Type of Dominant Motion
1 0.0 Air Constant Pressure
2 291 Air u(z)
3 295 Coupled u(y) back panel
4 336 Coupled u(x) sides
5 371 Coupled u(y) cone
6 373 Coupled u(x) sides
7 445 Coupled u(z) top/bottom
8 453 Coupled u(y) cone
Mode No. Frequency, Hz Type
Main Index
425
CHAPTER 11
Coupled FluidStructure 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 ROOTTRACKING MESSAGES.
CEND
TITLE=SPEAKER BOX WITH CONE, SIMPLE CORNER SUPPORTS
SUBTITLE = COUPLED BOUNDARY, NONMATCHING 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,10,3,0.13,3,,,0.223
$
$ PROPERTIES OF AIR
MAT10,100,131.94+3,1.115
PSOLID,100,100,,2,,1,PFLUID
Type Solution Solution Size CPU Time (Seconds)
Uncoupled Modes 538 110.2
Direct, No Modes 538 586.3
Struct. Modes Only 114 143.2
All Modes 25 106.9
Main Index
MD Nastran Dynamic Analysis User’s Guide
Coupled and Uncoupled Acoustic Analysis
426
$
$ 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
427
CHAPTER 11
Coupled FluidStructure 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
MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
428
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 nonreflecting 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 socalled 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.
Definition of Infinite Elements
The geometry of an infinite element is defined by the geometry of its base and the location of the pole,
see Figure 117. The base of the infinite element is that part that is in contact with the finite domain. The
geometry of the base is defined by its connectivity and the location of the corner grids. In order to avoid
overlapping of the infinite elements, the surface they are attached to must be convex. However, it is not
necessary that this surface is smooth.
Main Index
429
CHAPTER 11
Coupled FluidStructure Analysis
Figure 117 Geometry of Infinite Element
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
MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
430
Defines an acoustic conjugate infinite element with triangular base
Format
Defines an acoustic conjugate infinite element with quadrilateral base
Format
CACINF3  MD Only
Acoustic Conjugate Infinite Element Base Connection
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)
CACINF4  MD Only
Acoustic Conjugate Infinite Element Base Connection
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
431
CHAPTER 11
Coupled FluidStructure Analysis
Defines the properties of acoustic conjugate infinite elements.
Format
Definition of Field Point Meshes
Acoustic field point meshes are defined in separate sections of the bulk data file. These sections follow
the main bulk data section. Each of the sections begins with
BEGIN BULK AFPM = afpmid
or
BEGIN AFPM = afpmid
where afpmid is the acoustic field point mesh identifier (Integer > 0). Acoustic field points are defined
using the standard GRID Bulk Data entry. Because all grid points defined in this section of the bulk data
file are acoustic field points, it is not necessary to place a 1 into field 7.
The grid points can be connected by any type of elements. However, only CTRIA3 and CQUAD4
elements define a field point mesh that is used to compute normal components of the acoustic intensity
and the power through the field point mesh. Legal property identifiers have to be specified on the
CTRIA3 and CQUAD4 Bulk Data entries. However, the referenced PSHELL entries need not be
defined.
If output to an .op2 file is requested, parameter POST has to be defined within the section of the acoustic
field point mesh. Output of different field point meshes can be sent to different .op2 files using parameter
OUNIT2 within the sections of the acoustic field point meshes.
Example
BEGIN AFPM=100
$
PARAM, POST, 1
$
$ Isolated Field Points along a Line
PACINF  MD Only
Acoustic Conjugate Infinite Element Property
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
MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
432
$
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
Case Control Commands
The following case control commands control postprocessing which is related to exterior acoustics.
Requests output of the power radiated from the wetted surface.
ACPOWER  MD Only
Acoustic Power Output Request
Main Index
433
CHAPTER 11
Coupled FluidStructure Analysis
Format
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.
ACPOWER
SORT1
SORT2
PRINT,PUNCH
PLOT
CSV unit =   . .
\ .

 
ALL
n
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
Main Index
MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
434
Requests output of acoustic intensity on wetted surface.
Format
INTENSITY  MD Only
Acoustic Intensity Output Request
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.
INTENSITY
SORT1
SORT2
PRINT,PUNCH
PLOT
.
\ .

 
ALL
n
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
Main Index
435
CHAPTER 11
Coupled FluidStructure Analysis
Requests output of field point mesh results.
Format
Output
In addition to the standard acoustic results, the following results may be requested:
• Acoustic energy radiated from the wetted surface or from panels
ACFPMRESULT  MD Only
Acoustic Field Point Mesh Results Output Request
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.
ACFPMRESULT
SORT1
SORT2
PRINT,PUNCH
PLOT
VELOCITY
YES
NO
¹ )
´ `
¦ ¹
= . .
\

REAL or IMAG
PHASE
POWER
YES
NO
¹ )
´ `
¦ ¹
=
.


.
ALL
n
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
Main Index
MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
436
• 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 118 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 119.
Main Index
437
CHAPTER 11
Coupled FluidStructure Analysis
Figure 118 Geometry
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 1110.
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 1111 shows the fluid mesh and
one quarter of the field point mesh.
Main Index
MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
438
Figure 119 Structure and Quarter of Fluid Mesh
Figure 1110 Structure and Quarter of Infinite Element Mesh
Main Index
439
CHAPTER 11
Coupled FluidStructure Analysis
Figure 1111 Fluid Mesh and Quarter of Field Point Mesh
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 zaxis 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 111.
Listing 111 Part of Exterior Acoustics Example
$
$ file bd11aie.dat
$
SOL 108
CEND
TITLE = Vibrating Plate Problem
SUBTITLE = Direct Frequency Response
$
ECHO = SORT(EXCEPT, GRID, CHEXA, CQUAD4, CACINF4)
$
DLOAD = 10
FREQ = 20
SMETHOD = 30
SPC = 1
$
SET 1 = 1088, 1440, 1792, 2144, 2496, 2848,
3200, 3728
DISP(PHAS) = 1
ACFPMRESULT(PHAS) = ALL
$
BEGIN BULK
$
PARAM, POST, 1
$
Main Index
MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
440
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 = 1.0E4
$
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. 1.25
.......
$ 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
441
CHAPTER 11
Coupled FluidStructure Analysis
SPC1 1 123 160 THRU 176
$ Referenced Coordinate Frames
CORD2R 1 0. 0. 0. 0. 0. 1.
1. 0. 0.
$
BEGIN AFPM=100
$
$ Isolated Field Points
$
GRID, 10001,, 0., 0., 3.5
GRID, 10002,, 0., 0., 5.
GRID, 10003,, 0., 0., 10.
$
GRID, 11001,, 7., 6., 6.
GRID, 11002,, 7., 6., 6.
GRID, 11003,, 7., 6., 6.
GRID, 11004,, 7., 6., 6.
$
ENDDATA
Results
Figure 1112 shows the acoustic pressure in the Finite Element mesh and Figure 1113 shows some
results printed to the .f06.
Figure 1112 Acoustic Pressure in the Finite Element Mesh
Main Index
MD Nastran Dynamic Analysis User’s Guide
Exterior Acoustics
442
Figure 1113 Partial Results of Exterior Acoustics Example
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)
POINT ID. TYPE T1 T2 T3 R1 R2 R3
1088 S 4.965144E02
355.7659
1440 S 2.997001E02
352.8348
1792 S 2.046974E02
349.4032
2144 S 1.523430E02
345.6827
2496 S 1.202095E02
341.7944
2848 S 9.881609E03
337.8060
3200 S 8.368078E03
333.7535
3728 S 6.789113E03
327.6001
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)
POINT ID. TYPE P P(RMS) DB DB(A)
1088 S 4.965144E02 3.510887E02 2.608136E+01 9.648136E+01
355.7659 355.7659 355.7659 355.7659
1440 S 2.997001E02 2.119200E02 3.046626E+01 1.008663E+02
352.8348 352.8348 352.8348 352.8348
1792 S 2.046974E02 1.447430E02 3.377775E+01 1.041777E+02
349.4032 349.4032 349.4032 349.4032
2144 S 1.523430E02 1.077228E02 3.634355E+01 1.067435E+02
345.6827 345.6827 345.6827 345.6827
2496 S 1.202095E02 8.500094E03 3.840123E+01 1.088012E+02
341.7944 341.7944 341.7944 341.7944
2848 S 9.881609E03 6.987353E03 4.010345E+01 1.105034E+02
337.8060 337.8060 337.8060 337.8060
3200 S 8.368078E03 5.917124E03 4.154749E+01 1.119475E+02
333.7535 333.7535 333.7535 333.7535
3728 S 6.789113E03 4.800628E03 4.336374E+01 1.137637E+02
327.6001 327.6001 327.6001 327.6001
FREQUENCY = 1.000000E+01 ACOUSTIC FIELD POINT MESH = 100
A C O U S T I C F I E L D P O I N T M E S H R E S U L T S
ACOUSTIC PRESSURE INTENSITY COMP. INTENSITY INTENSITY INTENSITY
POINT ID. MAGNITUDE PHASE NORMAL TO FPM X Y Z
10001 5.866256E03 3.224648E+02 0.0 2.931354E12 1.101986E09 4.104157E08
10002 4.148095E03 3.068126E+02 0.0 1.247390E12 3.946541E10 2.070771E08
10003 2.089736E03 2.541933E+02 0.0 1.670165E13 5.247429E11 5.294857E09
11001 1.899346E03 2.438697E+02 0.0 2.784603E09 2.450892E09 2.336591E09
11002 1.899338E03 2.438701E+02 0.0 2.784757E09 2.449285E09 2.337945E09
11003 1.899360E03 2.438697E+02 0.0 2.784794E09 2.450661E09 2.336715E09
11004 1.899329E03 2.438699E+02 0.0 2.784759E09 2.394420E09 2.392732E09
Main Index
443
CHAPTER 11
Coupled FluidStructure Analysis
BW Method of FluidStructure Analysis
Introduction
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:
Figure 1114 Performance
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
MD Nastran Dynamic Analysis User’s Guide
BW Method of FluidStructure Analysis
444
The format for the BW method is:
Example
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:
Example
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
1 2 3 4 5 6 7 8 9 10
ACMODL INTER INFOR FSET SSET NORMAL METHOD SKNEPS DSKNEPS
INTOL ALLSSET SRCHUNIT
ACMODL .25
1 2 3 4 5 6 7 8 9 10
ACMODL INTER INFOR FSET SSET NORMAL METHOD
ACMODL CP
R
i
N
f
S p
i
1 p
j
0 = ; = { } d
S
í
= i 1 N grid/elem . =
Main Index
445
CHAPTER 11
Coupled FluidStructure 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)
d. Using the origin of the free fluid face, determine the center of pressure . The
relationship will be of the form
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:
where is the vector of resulting load distribution at the grids of each of the j structural
elements. is a diagonal weighting matrix. 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.
F
i
{ } N
f
{ }
T
N
f
S p
i
{ } d
S
í
=
X
P
i
Y
P
i
. ( )
X
P
i
F
i
R
i
 X
j
X
0
– ( )
j
grids
¯
=
Y
P
i
F
i
R
i
 Y
j
X
0
– ( )
j
grids
¯
=
F
j
{ } W   R   R  
T
W   R   ( )
1 –
R
i
0
0
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
F
j
{ }
W   R  
Main Index
MD Nastran Dynamic Analysis User’s Guide
BW Method of FluidStructure Analysis
446
Example
The model shown below, see Install_dir/mdxxxx/doc/dynamics/acoustic1.dat, is a square column of fluid
with a round top which interacts with an inner structural plate with a hole in it and an outer structural
plate:
Figure 1115
The inner plate has a hole as shown:
Main Index
447
CHAPTER 11
Coupled FluidStructure Analysis
Figure 1116
The relative location of the fluid face and the structural plates is shown:
Figure 1117
Main Index
MD Nastran Dynamic Analysis User’s Guide
BW Method of FluidStructure Analysis
448
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:
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
$ ********************
$ 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 fluidstructure 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 fluidstructure 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 fluidstructure 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 fluidstructure 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
Main Index
449
CHAPTER 11
Coupled FluidStructure Analysis
which creates a MSC.Patran session file named “acoustic1_flstgrp.ses”, which when run in MSC.Patran
will create two groups:
These two groups are shown below:
Figure 1118
Two things are indicated:
• extra structure elements were selected that do not project directly onto the fluid faces,
• only elements from the inner structural plate were selected.
acoustic1_fluid the fluid interface elements and grids
acoustic1_structure structure interface elements and grids
Main Index
MD Nastran Dynamic Analysis User’s Guide
BW Method of FluidStructure Analysis
450
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.
Figure 1119 Search Box
•  Search box height in positive normal direction from fluid face, where L is
smallest fluid face side length,
•  height in negative normal direction,
•  inplane 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.
INTOL L ·
NORMAL L ·
INTOL L ·
1 SKNEPS + ( ) D ·
Main Index
451
CHAPTER 11
Coupled FluidStructure 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 1120
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
MD Nastran Dynamic Analysis User’s Guide
BW Method of FluidStructure Analysis
452
Figure 1121
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
453
CHAPTER 11
Coupled FluidStructure 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 1122 shows the fluid pressure distribution and the structural deflection.
Figure 1122
Main Index
MD Nastran Dynamic Analysis User’s Guide
BW Method of FluidStructure Analysis
454
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
455
CHAPTER 11
Coupled FluidStructure Analysis
Direct Input of Interface [A] Matrix
You can input the [A] matrix directly using DMIG with the A2GG Case Control command. The format
of the A2GG input is similar to the other Gtype DMIG input (e.g., K2GG). By default, the A2GG fluid
structure coupling matrix will be added to the computed coupling matrix. To replace the computed
coupling matrix with the selected A2GG matrix, set PARAM,ASCOUP,NO.
Examples
• A2GG = ADMIG
• A2GG = ADMIG1, ADMIG2, ADMIG3
• A2GG = 1.5*ADMIG1, .5*ADMIG2
Main Index
MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
456
Acoustic Modal Participation Factors
Modal Participation Theory
The fluid and structure equations of motion are:
(1117)
where:
The participation factors are obtained as follows:
(1118)
(1119)
where are the uncoupled, undamped structural modes and are the uncoupled, undamped, rigid
wall acoustic modes. The vectors and are the modal amplitudes. Substituting these relations
into Eq. (1117) and premultiplying by the modal matrices, we get the equation
(1120)
or
(1121)
= mass
= damping
= stiffness
= displacement
= pressure
= structure
= fluid
= interface matrix
= load
M
s
0
A –
T
M
f
u
··
s
p
··
¹ )
´ `
¦ ¹ B
s
0
0 B
f
u
·
s
p
·
¹ )
´ `
¦ ¹ K
s
A
0 K
f
u
s
p
¹ )
´ `
¦ ¹
+ +
P
s
P
f
¹ )
´ `
¦ ¹
=
M
B
K
u
p
s
f
A
P
u
s
{ } u
s
  c
s
{ } =
p { } u
f
  c
f
{ } =
u
s
  u
f
 
c
s
{ } c
f
{ }
u
s
T
M
s
u
s
0
u
f
T
A
T
u
s
– u
T
M
f
u
f
c
··
s
c
··
f
¹ )
´ `
¦ ¹ u
s
T
B
s
u
s
0
0 u
T
B
f
u
f
c
·
s
c
·
f
¹ )
´ `
¦ ¹ u
s
T
K
s
u
s
u
s
T
Au
f
0 u
f
T
K
f
u
f
c
s
c
f ¹ )
´ `
¦ ¹
+ +
u
s
T
P
s
u
f
T
P
f
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
m
s
0
a
T
– m
f
c
··
s
c
··
f
¹ )
´ `
¦ ¹ b
s
0
0 b
f
c
·
s
c
·
f
¹ )
´ `
¦ ¹ k
s
a
0 k
f
c
s
c
f
¹ )
´ `
¦ ¹
+ +
Q
s
Q
f
¹ )
´ `
¦ ¹
=
Main Index
457
CHAPTER 11
Coupled FluidStructure Analysis
Define the following harmonic solution forcing frequency .
(1122)
(1123)
(1124)
For the bottom equation of Eq. (1121), we get:
(1125)
Define as
(1126)
then
(1127)
The fluid mode participation is defined as
(1128)
where is the diagonalized vector of fluid modal amplitudes per excitation frequency.
The structure mode participation is defined as
(1129)
where is the diagonalized vector of structural modal amplitudes per excitation frequency.
The fluid load participation is defined as
(1130)
The fluidstructure panel participation is defined as
(1131)
The fluidstructure panelgrid participation is defined as
(1132)
e
c { }
c
s
c
f
¹ )
´ `
¦ ¹
e
i et
= =
c
·
{ } iee
i et
=
c
··
{ } e
2
e
i et
– =
e
2
a  
T
c
s
{ } e –
2
m
f
  ie b
f
  k
f
  + +   c
f
{ } + Q
f
{ } =
Z
2
 
Z
2
  e
2
m
f
  – ie b
f
  k
f
  + +  
1 –
=
c
f
{ } e
2
Z
2
  o  
T
c
s
{ } – Z
2
  Q
f
{ } + =
P
f
  u
f
  c
f
{ }   =
c
f
{ }  
P
s
  e
2
u
f
  Z
2
  a  
T
c
s
{ }   – =
c
s
{ }  
P
l
{ } u
f
  Z
2
  Q
f
{ } =
P
p
  e
2
u
f
  Z
2
  u
f
 
T
A  
panel
T
u
s
  c
s
{ }   – =
P
g
 
i
e – =
2
u
f
  Z
2
  u
f
 
T
A
b
T
¹ )
´ `
¦ ¹
u
s
  c
s
 
Main Index
MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
458
where
(1133)
are the columns extracted from the bth boundary panel for panel grid i and are the rows of the
structural modal matrix corresponding to panel grid i.
Participation Factor Results
Introduction
The Case Control commands PFMODE, PFPANEL and PFGRID are used to request modal, panel and
grid participation factors.
The output format is used for all types of modal and panel participation factors. The output format of grid
participation factors is similar to the output format of displacements.
The implementation supports
• panel definition by referencing a set of grids, a set of elements or a set of properties.
• panel and grid participation factors in direct frequency response analysis (SOL 108).
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 fluidstructure interface,
the socalled wetted surface. The pressure due to the acoustic sources in a rigid cavity is called the load
participation factor.
A
b
T
¹ )
´ `
¦ ¹
u
s
Main Index
459
CHAPTER 11
Coupled FluidStructure Analysis
The acceleration of the fluidstructure 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 lowfrequency 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
MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
460
Input
Computation and output of modal participation factors is controlled by the PFMODE Case Control
command:
Examples
SET 20 = 25/T3, 33/T3
PFMODE(STRUCTURE, STRUCTMP=ALL) = 20
Compute structural modal participation factors for ztranslations at grid points 25 and 33
SET 20 = 11217
SET 90 = 25., 30., 35.
PFMODE(FLUID, STRUCTMP=ALL, FLUIDMP=ALL,
SOLUTION = 90) = 20
Compute acoustic structural and fluid modal participation factors for pressure at grid point 11217 at
excitation frequencies 25Hz, 30Hz and 35Hz.
PFMODE
STRUCTURE
FLUID
PRINT, PUNCH
PLOT
REAL or IMAG
PHASE
. . .
\

SORT sorttype =   KEY sortitem =   ITEMS
ALL
itemlist ( )
¹ )
´ `
¦ ¹
= . . .
FLUIDMP
ALL
m
f
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
= STRUCTMP
ALL
m
s
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
= . .
PANELMP
ALL
setp
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
= SOLUTION
ALL
setf
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
= FILTER frati o =   . .
NULL i power =   )
setdof
NONE
¹ )
´ `
¦ ¹
=
.
Main Index
461
CHAPTER 11
Coupled FluidStructure Analysis
Computation and output of panel participation factors is controlled by the PFPANEL Case Control
command:
Example
SET 100 = 10., 12.
SET 200 = 1222, 1223
PFPANEL(SOLUTION=100, SORT=ABSD) = 200
Compute acoustic panel participation factors for pressure at grid points 1222 and 1223 at excitation
frequencies 10Hz and 12Hz. Output is sorted according to descending modal fractions.
Computation and output of grid participation factors is controlled by the PFGRID Case Control
command:
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.
PFPANEL
PRINT, PUNCH
PLOT
REAL or IMAG
PHASE
PANEL
ALL
setp

¹ )
´ `
¦ ¹
= . . .
\

SORT sorttype =   KEY sortitem =   ITEMS
ALL
it emlist ( )
¹ )
´ `
¦ ¹
= . . .
SOLUTION
ALL
setf
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
= FILTER fratio =   NULL ipower =   . .
setdof
NONE
¹ )
´ `
¦ ¹
=
PFGRID
PRINT, PUNCH
PLOT
REAL or IMAG
PHASE
GRIDS
ALL
setg
¹ )
´ `
¦ ¹
= . .
\

SOLUTION
ALL
setf
NONE
¹ )
¦ ¦
´ `
¦ ¦
¦ ¹
=
.




setdof
NONE
¹ )
´ `
¦ ¹
=
Main Index
MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
462
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 fluidstructure control parameters in one
place.
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.
FLSTCNT ACSYM
YES
NO

¹ )
´ `
¦ ¹
= ACOUT
PEAK
RMS

¹ )
´ `
¦ ¹
= .
PREFDB
1.0
prp

¹ )
´ `
¦ ¹
= ASCOUP
YES
NO

¹ )
´ `
¦ ¹
= .
SKINOUT
NONE
PUNCH
PRINT
ALL
¹ )
¦ ¦
¦ ¦
´ `
¦ ¦
¦ ¦
¦ ¹
=
Main Index
463
CHAPTER 11
Coupled FluidStructure Analysis
The data are presented in ten columns:
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.
Column Label Description
1 Mode Id Mode number
2 Natural Freq (Hz) Resonance frequency
34 Modal Response: Real / Imaginary Real and imaginary part of participation factor
56 Modal Response: Magnitude / Phase Magnitude and phase of participation factor
7 Projection Magnitude Projected participation factor: That part of the
participation factor that is in phase with the total
response
8 Relative Phase Phase of participation factor relative to total
response
9 Modal Fraction Real part of normalized participation factor
10 Scaled Response Magnitude Projected participation factor divided by largest
magnitude of all participation factors
Main Index
MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
464
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.
Example Participation Factors
The example shows the acoustic analysis of a cabin. The cabin is excited by four forces at the corners of
the seat. The result of interest is the pressure at the driver’s ear.
The frequency response of the pressure shows a peak at 40Hz. To investigate this peak, acoustic fluid and
structure modal participation factors and panel participation factors are requested at a frequency of 40Hz.
Main Index
465
CHAPTER 11
Coupled FluidStructure 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 112 for a partial
input.
$ Participation Factor Test Problem: Cabin
$
$ Coupled Modal Frequency Response Analysis
$
$ Participation Factor Example  Cabin
Main Index
MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
466
$
$
$ 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
$
$ =======================================================
$
Listing 112 Partial Input File
SOL 111
CEND
TITLE = Cabin Example
SUBTILE = Direct Enforced Motion, Modal Frequency Response
$
ECHO=SORT(EXCEPT,CBEAM,CQUAD4,CHEXA,CPENTA,GRID)
AUTOSPC(NOZERO) = YES
$
METHOD(STRUCTURE)=1
METHOD(FLUID) =2
$
FREQ = 100
DLOAD = 200
SPC = 300
$
SET 1 = 1, 7
SET 20 = 11217
SET 90 = 40.
$
ACCE(PHAS,SORT2) = 1
Main Index
467
CHAPTER 11
Coupled FluidStructure 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, 2.82845
$
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 FluidStructure 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
MD Nastran Dynamic Analysis User’s Guide
Acoustic Modal Participation Factors
468
, 497, THRU, 562
$
PANEL, LEFT, 101, RIGHT , 201, FRONT, 301, REAR, 401
PANEL, TOP , 501, BOTTOM, 601
$
$ Structural Model
$
.......
ENDDATA
Listing 113 Acoustic Structural Mode Participation Factors:
Listing 114 Acoustic Panel Participation Factors:
The acoustic structural mode participation factors show that the largest contribution comes from the 9
th
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.
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) = 9.42534E01 / 1.03865E02, (M/P) = 9.42591E01 / 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 = 1.00000E03
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 6.25017E03 1.13741E+00 359.69 1.13725E+00 179.05 1.20652E+00 9.99864E01
8 8.18779E+01 6.32888E01 4.67901E03 6.32905E01 179.58 6.32798E01 1.05 6.71338E01 5.56350E01
3 6.82472E+01 6.25327E01 6.91886E03 6.25365E01 179.37 6.25213E01 1.27 6.63291E01 5.49681E01
21 1.33089E+02 5.71325E01 1.54503E03 5.71327E01 179.85 5.71273E01 0.79 6.06066E01 5.02258E01
84 2.42589E+03 4.37526E01 1.03197E04 4.37526E01 180.01 4.37501E01 0.62 4.64147E01 3.84647E01
31 1.60995E+02 3.90763E01 2.29250E04 3.90763E01 359.97 3.90737E01 179.34 4.14534E01 3.43532E01
14 1.07144E+02 2.38134E01 1.45832E03 2.38139E01 179.65 2.38104E01 0.98 2.52605E01 2.09338E01
79 4.55425E+02 2.27102E01 8.55791E05 2.27102E01 359.98 2.27088E01 179.35 2.40918E01 1.99653E01
5 7.74596E+01 1.86438E01 1.13864E03 1.86441E01 359.65 1.86414E01 179.02 1.97768E01 1.63894E01
1 3.61553E+01 1.61339E01 1.79486E02 1.62334E01 186.35 1.61527E01 5.72 1.71364E01 1.42013E01
13 1.05870E+02 1.54939E01 5.68473E04 1.54940E01 179.79 1.54923E01 0.84 1.64359E01 1.36207E01
32 1.70901E+02 1.41683E01 1.99820E04 1.41683E01 179.92 1.41672E01 0.71 1.50301E01 1.24557E01
29 1.51451E+02 1.20258E01 1.69764E04 1.20258E01 359.92 1.20249E01 179.29 1.27573E01 1.05722E01
50 2.09592E+02 9.86342E02 1.15611E04 9.86342E02 179.93 9.86269E02 0.70 1.04634E01 8.67119E02
11 9.45225E+01 9.81448E02 3.61858E04 9.81454E02 359.79 9.81348E02 179.16 1.04112E01 8.62792E02
80 4.99581E+02 9.78103E02 5.94714E05 9.78103E02 179.97 9.78037E02 0.67 1.03760E01 8.59881E02
36 1.79440E+02 7.51393E02 7.29040E05 7.51393E02 180.06 7.51355E02 0.58 7.97116E02 6.60584E02
43 2.01720E+02 7.43311E02 5.88513E05 7.43311E02 359.95 7.43259E02 179.32 7.88528E02 6.53467E02
23 1.33661E+02 7.42764E02 1.22552E04 7.42765E02 179.91 7.42705E02 0.73 7.87940E02 6.52980E02
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) = 9.42534E01 / 1.03865E02, (M/P) = 9.42591E01 / 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 8.05866E03 2.63941E+00 179.83 2.63915E+00 0.81 2.79989E+00 9.99901E01
FRONT 1.11163E+00 4.59632E03 1.11164E+00 359.76 1.11152E+00 179.13 1.17921E+00 4.21123E01
TOP 1.09365E+00 3.39035E03 1.09365E+00 359.82 1.09354E+00 179.19 1.16015E+00 4.14314E01
REAR 2.33400E01 2.06344E02 2.34310E01 185.05 2.33613E01 4.42 2.47841E01 8.85095E02
RIGHT 6.01577E02 5.20980E03 6.03829E02 175.05 6.00966E02 5.58 6.37568E02 2.27690E02
LEFT 6.01573E02 5.20977E03 6.03825E02 175.05 6.00963E02 5.58 6.37564E02 2.27688E02
LOAD 3.82035E02 6.22966E06 3.82035E02 179.99 3.82011E02 0.64 4.05278E02 1.44734E02
Main Index
469
CHAPTER 11
Coupled FluidStructure Analysis
Acoustic XY Plots, Random, and Restarts
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
Also, the following “ptype”s:
• MODE
• for given fluid mode plot mode
• participation vs excitation frequency
• O2E=“NO” on FLSPOUT Case Control command (default)
• FREQ
• for given excitation frequency
• plot mode participation vs natural frequency
• O2E=“YES” on FLSPOUT Case Control command
• point plot only
See the following example:
Random Analysis
A “AUTO” and “PSDF” XYPLOT random “ptype”s support the new acoustic modal participation
“yvtype”s:
• FMPF
• SMPF
• PMPF
Main Index
MD Nastran Dynamic Analysis User’s Guide
Acoustic XY Plots, Random, and Restarts
470
• 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 fluidstructure 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 115 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 116 and see Install_dir/mdxxxx/doc/dynamics/bd11acxx.dat for the data file.
Listing 115 Input File For Base Line Run
SOL 111 $ MODAL frequency response
CEND
Main Index
471
CHAPTER 11
Coupled FluidStructure Analysis
TITLE= FLUIDSTRUCTURE INTERACTIONmpf.
$
FLSFSEL LFREQFL=1.0,LFREQ=2.0
FLSTCNT PREFDB=20.6
$
OLOAD=ALL
DISP(SORT2,PHASE)=ALL
STRESS(SORT2,PHASE)=ALL
RANDOM 10
$
SPC =1313
FREQ=200
METHOD(struct)=30
METHOD(fluid)=20
$
SUBCASE 100
DLOAD=100
$
SUBCASE 200
DLOAD=200
$
OUTPUT(XYPLOT)
XGRID=YES
YGRID=YES
XTITLE=frequency #14
YTITLE=psd grid 61
XYPLOT,XYPEAK DISP PSDF /61(T3)
XTITLE=frequency #15
YTITLE=auto grid 61
XYPLOT,XYPEAK DISP AUTO /61(T3)
XTITLE=frequency #16
YTITLE=psd grid 224
XYPLOT,XYPEAK DISP PSDF /224(T1)
XTITLE=frequency #17
YTITLE=auto grid 224
XYPLOT,XYPEAK DISP AUTO /224(T1)
$
BEGIN BULK
$
$ PRODUCE OUTPUT FORTRAN FILES
PARAM,POST,1
$
$ DYNAMIC LOADING
$
$234567890
DLOAD 100 1. 1. 1002 1. 1001
DLOAD 200 1. 1. 2002 1. 2001
RLOAD11001191001
RLOAD1200129 1001
$
DAREA19413125.
DAREA19453125.
DAREA19613125.
DAREA19653125.
$
DAREA29 413125.
DAREA29 453125.
DAREA29 613125.
DAREA29 653125.
$ ON FLUID
Main Index
MD Nastran Dynamic Analysis User’s Guide
Acoustic XY Plots, Random, and Restarts
472
ACSRCE 1002 101 1001 1. 1.
SLOAD 101 120 125.
SLOAD 101 124 125.
SLOAD 101 220 125.
SLOAD 101 224 125.
ACSRCE 2002 102 1001 1. 1.
SLOAD 102 120 250.
SLOAD 102 124 250.
SLOAD 102 220 250.
SLOAD 102 224 250.
$
$ the load is (2*pi)^2 *f^2 so when Q is calculated it becomes a constant
$ for all frequencies
TABLED4 1001 0.0 1.0 0.0 1.E6
0.0 0.0 39.478420.0 ENDT
$
$ FRQUENCY RANGE
FREQ1 200 2.5 2.5 100
$ DAMPING
PARAMG0.05
PARAMGFL0.05
$ METHOD CARD FOR EIGEN VALUE
EIGRL206
EIGRL304
$
$THE STRUCTURAL POINTS
GRID 41 0.0 0.0 0.0
GRID 45 1.1 0.0 0.1
$
GRID 61 0.0 0.20 0.0
GRID 65 1.0 0.20 0.0
$
$ 2 3 4 5 6 7 8 9 0
$ STRUCTURAL POINTS FOR ABSORBER
GRID 261 0.0 0.2 0.001 12456
GRID 265 1.0 0.2 .001 12456
GRID 141 0.0 0.0 0.001 12456
GRID 145 1.1 0.0 .1001 12456
$THE STRUCTURAL ELEMENTS
CQUAD4 100 4444 41 45 65 61
$
CELAS2 1007 .0625 61 3
CELAS2 1008 .0625 65 3
CELAS2 1009 .0625 41 3
CELAS2 1010 .0625 45 3
$ STRUCTURAL PROPERTIES
PSHELL 4444 77 .05 77 1.0
MAT1 77 100. .333 1.000
$
$ FLUID POINTS
$ GRIDS 200 THRU 204 AND 100 THRU 104 ARE CONTACT GRIDS. KEEP THEM IN THE
$ RESIDUAL
GRID 200 0.0 0.2 0.001 1
GRID 204 1.0 0.2 .001 1
GRID 220 0.0 0.2 1.0 1
GRID 224 1.0 0.2 1.0 1
GRID 100 0.0 0.0 0.001 1
GRID 104 1.1 0.0 .1001 1
GRID 120 0.0 0.0 1.0 1
GRID 124 1.0 0.0 1.0 1
Main Index
473
CHAPTER 11
Coupled FluidStructure 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 116 Restart Run
ASSIGN MASTER='bd11acxybL.MASTER'
RESTART VERSION=1 KEEP
SOL 111 $ MODAL frequency response
CEND
TITLE= FLUIDSTRUCTURE 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
MD Nastran Dynamic Analysis User’s Guide
Acoustic XY Plots, Random, and Restarts
474
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
475
CHAPTER 11
Coupled FluidStructure 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 1123
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
MD Nastran Dynamic Analysis User’s Guide
Acoustic XY Plots, Random, and Restarts
476
For PSD curves, the rootmeansquare value across frequencies is calculated and printed in the .f06 file:
Listing 117 PSD RMS Value from Restart
This RMS value, 5.075398E+6, can be used as a measure of the combined relative effect of mode 9 across
all the frequencies.
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
FREQ NO. 0 BNDY : 61
PSDF GRIDP 3 224( 3) 5.075398E+06 8.487302E+01 2.500E+00 2.525E+02 1.942E+06 5.000E+00 4.808E+12 2.250E+01
Main Index
477
CHAPTER 11
Coupled FluidStructure Analysis
Acoustic Source Change
Simple Acoustic Source (ACSRCE) Equation
The complex source strength is defined in Reference 2. in Appendix B as
(1134)
Also, similar to the MaxwellBetti 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 in terms of complex source strength,
which for a pulsating sphere is a real value . The resulting expression as used in MD Nastran is:
(1135)
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. (1135) 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. (1135), and apply it
as needed on the RLOADi and TLOADi Bulk Data entries.
Qe
i et
u
·
s
í
n dS · =
P f ( )
Q
Q
1
2tf

8tcP f ( )
p
 =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Rigid Porous Absorber  MAT10
478
Rigid Porous Absorber  MAT10
Introduction
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 Bazley
1
, methods which
have been widely used to describe sound propagation in fibrous materials.
In the equivalent fluid approach, the equation of motion reads
where is the equivalent density, the equivalent bulk modulus, the complex pressure amplitude
and the circular excitation frequency, in which it can be shown
and
1
p
e
 V
2
P
e
2
B
e
 P + 0 =
p
e
B
e
P
e
1
p
e

1
p
 1 iGE + ( ) =
Main Index
479
CHAPTER 11
Coupled FluidStructure Analysis
if it is assumed that the parameters are not frequency dependent, a reasonable assumption for the study
of frequencies in a narrow band. Here, , and 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 coefficient
2
, to be entered.
Inputs
Defines material properties for fluid elements in coupled fluidstructural analysis.
Format
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
MAT10 Damping Coefficient
MAT10 Bulk Modulus
MAT10 Normalized Admittance Coefficient
MAT10
Fluid Material Property Definition
1 2 3 4 5 6 7 8 9 10
MAT10 MID BULK RHO C GE ALPHA
1
B
e

1
B
 = i
A
e
 –
1
B
 1
io
e
 –
\ .
 
=
p B GE
o
p
p
p
r
2
p
i
2
+
p
r
 =
p
r
p
i
complex density, real part
complex density, imaginary part
GE
GE
p
i
p
r
 – =
B
B
B
r
2
B
i
2
+
B
r
 =
complex bulk modulus, real part
complex bulk modulus, imaginary part
B
r
B
i
o e
B
i
B
r
 =
Main Index
MD Nastran Dynamic Analysis User’s Guide
Rigid Porous Absorber  MAT10
480
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 midrange 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 nonzero 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 1124.
An acoustic source is placed at the location indicated and the acoustic response (pressure) at the centre
of the fluid is monitored.
Figure 1124 Porous medium
o
Porous medium
Response
Acoustic source
Air
Main Index
481
CHAPTER 11
Coupled FluidStructure Analysis
The following properties were determined using experimental methods.
Air
Porous Absorber
From which the following equivalent bulk moduli were obtained
Air
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
MAT10 for the porous absorber material
Notice that the values of bulk modulus, GE damping coefficient and alpha are all negative; this is a
normal characteristic of the implementation.
Density 1.225 + i0
Speed of sound 340.0 + i3.4
Density 3.8663 + i14.2204
Speed of sound 92.7076 + i70.2854
Bulk modulus 141595.8 + i2832.2
Bulk modulus 171190.0 + i102356.3
MID BULK RHO C GE ALPHA
MAT10 1 141652.5 1.225 0.0 31.41907
MID BULK RHO C GE ALPHA
MAT10 2 232389. 56.16948 3.67804 939.196
Main Index
MD Nastran Dynamic Analysis User’s Guide
Rigid Porous Absorber  MAT10
482
Main Index
MD N