(Harry G. Schaeffer) NASTRAN Primer (B-Ok - Xyz) PDF

NASTRAN Primer
Static and Normal Modes Analysis
By
Harry G. Schaeffer
NASTRAN Primer
Static and Normal Modes Analysis
by Harry G. Schaeffer
Copyright © 2015 by Harry G. Schaeffer
All rights reserved. No part of this book shall be reproduced, transmitted by any means without the express written consent
of the Copyright owner. No patent liability is assumed with respect to the information contained herein. Although every
precaution has been exercised by the author of this book, the author assumes no responsibility for errors or emissions.
Neither is any liability assumed for any damages resulting from the use of the information contained herein.
Trademarks
Al items in this book that are known to be trademarks or service marks are Capitalized. In particular, MSC.Nastran is a
trademark of MSC Software and NX NASTRAN is a trademark of Siemens Corp. The author cannot attest to accuracy
of this information. Use of a term in this book should not be regarded as affecting the viability of and trademark or
serviceman.
Warning and Disclaimer

Every effort has been made to make this book as complete and accurate as possible but no warranty or fitness is implied.
The information provided is on an “as is” basis. The author shall have neither liability or responsibility to any person or
entity with respect to any loss or damages arising from use of the information in this book.
 Harry G. Schaeffer, 2015. All Rights Reserved. No part of this book may be reproduced, stored in a retrieval system,
or transmitted in any form or by any means without prior written permission of Harry G. Schaeffer.
Chapter 1 Getting Started
1.1 Historical Introduction ................................................ 1
1.1.1 The Finite Element Method .................................................. 1
1.1.2 Solving NASA’s Simulation Needs ..................................... 2
1.1.3 NASA NASTRAN.................................................................. 3
1.1.4 MSC NASTRAN..................................................................... 3
1.1.5 NX NASTRAN ....................................................................... 4
1.2 NASTRAN® Architecture ........................................... 4
1.2.1 Independent Modules ........................................................... 4
1.2.2 Database .................................................................................. 4
1.2.3 Executive System ................................................................... 4
1.3 NASTRAN Technology ............................................... 5
1.3.1 Engineering Mechanics ......................................................... 5
1.3.2 Numerical Analysis ............................................................... 5
1.3.3 Computer Science .................................................................. 5
1.3.4 Computer Graphics ............................................................... 6
1.4 Design of the Finite Element Model........................... 6
1.4.1 Basic Modeling objects .......................................................... 6
1.4.2 The Design Procedure ........................................................... 7
1.5 Processing the Finite Element Model ........................ 8
1.5.1 Marshall the Model Data ...................................................... 8
1.5.2 Generation of Element Stiffness Matrices .......................... 8
1.5.3 Assemble the Unconstrained Stiffness Matrix................... 8
1.5.4 Detect and remove essential singularities .......................... 8
1.5.5 Apply Constraints.................................................................. 8
1.5.6 Generate Loads....................................................................... 8
1.5.7 Solve for Constrained Displacements ................................. 8
1.5.8 Recover Unconstrained Displacements .............................. 9
1.5.9 Recover Element Behavior.................................................... 9
1.5.10 Stress Averaging at the Grid Points .................................... 9
1.5.11 Strain Energy Density............................................................ 9
1.5.12 Results Files ............................................................................ 9
1.5.13 Post Processing....................................................................... 9
1.6 User Program Interface................................................ 9
1.6.1 Input File ................................................................................. 9
1.6.2 Types of Solutions.................................................................. 9
1.7 The Input File .............................................................. 10
1.7.1 NASTRAN Statement.......................................................... 10
1.7.2 Problem Definition Section................................................. 11
1.7.3 Getting Started with MSC.Nastran ................................... 11
Nastran Primer i
1.8 Specifying the Finite Element Model - Bulk Data.. 13
1.8.1 Free Field Bulk Data ............................................................ 14
1.8.2 Replication and Field Generation ...................................... 14
1.8.3 Summary of Structurally Oriented Bulk Data ................. 15
1.8.4 Data Selection ....................................................................... 20
1.8.5 Load Selection....................................................................... 20
1.8.6 Example: Load Specification .............................................. 20
1.8.7 Thermal Field Selection....................................................... 21
1.8.8 Example: Thermal Field Specification .............................. 21
1.8.9 Example: Specification of Constraints .............................. 21
1.8.10 Specification of Eigenvalue Method.................................. 21
1.8.11 Output Control ..................................................................... 22
1.8.12 Titling Directives.................................................................. 22
1.8.13 Output Line Control ............................................................ 22
1.8.14 Bulk Data Echo ..................................................................... 22
1.8.15 Set Specification.................................................................... 22
1.8.16 Element Output Requests ................................................... 22
1.8.17 Grid and Scalar Point Requests.......................................... 23
1.8.18 Subcases................................................................................. 23
1.8.19 Case Control Examples ....................................................... 26
1.9 Executive Control ....................................................... 28
1.9.1 Solution Sequence ................................................................ 28
1.9.2 Specification of Execution Time......................................... 29
1.9.3 Diagnostic Print Requests................................................... 29
1.9.4 Termination of Executive Control ..................................... 29
1.9.5 Executive Control Examples .............................................. 29
1.10 TRUSS EXAMPLE ...................................................... 30
1.10.1 Components of the File. ...................................................... 30
1.10.2 The Input File........................................................................ 31
1.10.3 Truss Schematic Diagram ................................................... 33
1.10.4 Running the Truss Example ............................................... 33
1.10.5 Examining the OUT File...................................................... 33
1.11 Problems ...................................................................... 35
1.12 References .................................................................... 35
Chapter 2 Basic Relationships from Engineering Mechanics
2.1 Theoretical Foundations ............................................ 37
2.2 Continuum Mechanics ............................................... 37
2.3 Kinematic Relations.................................................... 38
2.4 Strains ........................................................................... 38
2.5 Stress Vector on a Surface.......................................... 40
ii Nastran Primer
2.6 Components of Stress................................................. 42
2.7 Constitutive Behavior ................................................ 44
2.8 Equilibrium Conditions - Newtonian Mechanics .. 45
2.9 Principle of Virtual Work .......................................... 45
2.10 Solutions from the Theory of Elasticity ................... 46
2.10.1 Uniform Stress ...................................................................... 46
2.10.2 Stretching of a Prismatic Bar by Its Own Weight............ 47
2.11 Strategies for Solving Real Problems ....................... 47
2.12 Two Dimensional Theory of Elasticity .................... 48
2.12.1 Plane Stress ........................................................................... 48
2.12.2 Plane Strain ........................................................................... 48
2.12.3 Equilibrium Equations ........................................................ 48
2.12.4 Strain Displacement Relations ........................................... 49
2.12.5 Constitutive Relations ......................................................... 49
2.12.6 Stress Functions.................................................................... 49
2.12.7 Solutions based on Stress Functions ................................. 50
2.13 Beam Theory................................................................ 52
2.13.1 Stress Resultants................................................................... 52
2.13.2 Stresses due to Extension and Bending ............................ 53
2.13.3 Equilibrium Equations ........................................................ 54
2.13.4 Solution to Beam Bending Equation ................................. 55
2.14 Plate theory.................................................................. 55
2.14.1 Kinematic Relationships ..................................................... 56
2.14.2 Kirchhoff Hypothesis .......................................................... 57
2.14.3 Moment and Force Resultants ........................................... 58
2.14.4 Constitutive Relations ......................................................... 59
2.14.5 Equilibrium Equation .......................................................... 60
2.15 Problems ...................................................................... 60
2.16 References .................................................................... 61
Chapter 3 Variation Principles and Approximation Theory
3.1 Principle of Virtual Work .......................................... 63
3.2 Complementary Virtual Work.................................. 64
3.3 Minimum Potential Energy....................................... 64
3.4 Minimum Total Complementary Potential Energy65
3.5 Generalized Principle................................................. 66
3.6 Hellinger-Reissner Principle ..................................... 66
3.7 Approximation Theory .............................................. 67
3.8 Rayleigh-Ritz Soution of a Beam .............................. 68
Nastran Primer iii

3.9 References .................................................................... 69
Chapter 4 Finite Element Formulations
4.1 Finite Element Formulation ...................................... 71
4.2 Rod Element ................................................................ 71
4.3 Steps in Determining the Stiffness Equation .......... 73
4.4 Stiffness Matrix for a Beam Bending Element ........ 74
4.5 Basis Functions for Finite Elements ......................... 77
4.6 Isoparametric Transformation .................................. 79
4.6.1 Basis Functions and the Jacobian of Transformation...... 80
4.6.2 Example - Calculating the Area of a Square .................... 81
4.7 Numerical Integration................................................ 82
4.7.1 Integration for the Membrane Problem............................ 83
4.8 Plate and Shell Elements............................................ 83
4.8.1 Curvature Approximations ................................................ 84
4.8.2 Transverse Shear Strain Approximation .......................... 84
4.8.3 Curved Shell Elements ........................................................ 85
4.8.4 Normal Rotations................................................................. 86
4.8.5 Non-Planar Nodes for a Flat Shell Element ..................... 86
4.9 References .................................................................... 88
Chapter 5 Structural Elements in NASTRAN
5.1 Introduction................................................................. 89
5.2 Defining Element Objects .......................................... 90
5.2.1 Defining Element Connectivity.......................................... 91
5.2.2 Defining Element Properties .............................................. 91
5.2.3 Standard Notation ............................................................... 92
5.3 Scalar Elastic Elements............................................... 92
5.3.1 Spring Connecting Grid Point Degrees of Freedom ....... 93
5.3.2 Spring Connecting Scalar Points........................................ 93
5.3.3 Specification of Connected Degrees of Freedom............. 93
5.3.4 Properties .............................................................................. 93
5.3.5 Stress Recovery..................................................................... 94
5.3.6 Modeling with the ELAS Element..................................... 94
5.4 Rod and Truss Elements ............................................ 95
5.4.1 Description of Input Data for Truss .................................. 96
5.4.2 Stiffness Matrix..................................................................... 96
5.4.3 Stress Recovery..................................................................... 97
5.5 The BAR Beam Bending Element ............................. 99
5.5.1 Description of BAR Input Data ........................................ 100
iv Nastran Primer
5.5.2 BAR Connectivity .............................................................. 101
5.5.3 BAR Element Coordinate System .................................... 101
5.5.4 Defining BAR Offset .......................................................... 103
5.5.5 Pinned Connections........................................................... 103
5.5.6 Defining BAR Properties................................................... 104
5.5.7 Area Properties................................................................... 107
5.5.8 Nonstructural Mass ........................................................... 108
5.5.9 Stress Recovery................................................................... 109
5.5.10 Element Forces - Use of CBARAO................................... 109
5.5.11 Data Recovery Points - CBARAO Form 1 ...................... 110
5.5.12 Data Recovery Points - CBARAO Form 2 ...................... 111
5.5.13 BAR-Element Examples .................................................... 111
5.5.14 Modeling Considerations ................................................. 116
5.6 The BEAM Bending Element .................................. 117
5.6.1 Degrees of Freedom........................................................... 118
5.6.2 BEAM Description ............................................................. 118
5.6.3 CBEAM Data Statement.................................................... 119
5.6.4 Local Coordinate System for the BEAM......................... 119
5.6.5 PBEAM Data Statement .................................................... 120
5.6.6 BEAM Examples ................................................................ 122
5.7 Curved Beam Element ............................................. 127
5.7.1 Defining The BEND Element ........................................... 128
5.7.2 Bend Element Connectivity and Geometry ................... 130
5.7.3 BEND Element Coordinate System................................. 130
5.7.4 Specifying the Arc of Geometric Centrolds ................... 132
5.7.5 BEND Element Properties ................................................ 132
5.7.6 Element Forces and Stresses............................................. 136
5.8 Shear Panels............................................................... 136
5.8.1 SHEAR Definition.............................................................. 137
5.8.2 Description of Shear Panel Input..................................... 137
5.8.3 Element Properties............................................................. 137
5.8.4 Recovery of Forces and Stresses ...................................... 138
5.9 Shell Elements ........................................................... 138
5.9.1 Flat Shell Elements - OUAD4 and TRIA3....................... 139
5.9.2 Defining Connectivity and Properites ............................ 140
5.9.3 Element Connectivity ........................................................ 140
5.9.4 Element Coordinate System ............................................. 141
5.9.5 Material Coordinate........................................................... 141
5.9.6 Reference Surface Offset ................................................... 142
5.9.7 Element Properties - PSHELL .......................................... 142
5.9.8 Specifying Element Behavior ........................................... 142
Nastran Primer v
5.9.9 Element Constitutive Relations........................................ 143
5.9.10 Solid Homogeneous Symmetric Cross-Section ............. 143
5.9.11 Sandwich Type Cross-Section.......................................... 144
5.9.12 Effect of Warping ............................................................... 146
5.9.13 Elastic Stiffness Matrices................................................... 146
5.9.14 Mass Matrix ........................................................................ 146
5.9.15 Stress, Strain and Element Force Recovery .................... 147
5.10 Curved Shell Elements............................................. 148
5.10.1 TRIA6 and QUAD8 Connectivity.................................... 149
5.10.3 Element Coordinate Systems ........................................... 149
5.10.4 Material Orientation .......................................................... 149
5.10.5 Element Thickness ............................................................. 150
5.10.6 Specifying Element Behavior ........................................... 150
5.10.7 Element Stiffness Matrices................................................ 150
5.10.8 Element Mass Matrices ..................................................... 150
5.10.9 Stress, Strain and Element Force Recovery .................... 150
5.11 Solid Elements........................................................... 150
5.11.1 Defining Solid Elements.................................................... 152
5.11.3 The TETRA Element .......................................................... 153
5.11.4 The PENTA Element ......................................................... 153
5.11.5 HEXA Element ................................................................... 153
5.11.6 Properties of Solid Element .............................................. 153
5.11.7 Hexa Element Coordinate System................................... 155
5.11.8 PENTA Element Coordinate System .............................. 156
5.11.9 TETRA Element Coordinate Syatem............................... 157
5.11.10Elastic Stiffness Matrix ...................................................... 157
5.11.11Mass ..................................................................................... 158
5.11.12Stress and Strain Recovery ............................................... 158
5.12 Congruent Elements................................................. 158
5.13 References .................................................................. 159
Chapter 6 Global Analysis Procedures
6.1 The Global Stiffness Matrix ..................................... 161
6.2 Local and Global Coordinate Systems................... 164
6.3 Transformation of Element Stiffness Matrices ..... 164
6.4 Specifying Structural Degrees of Freedom ........... 165
6.5 Defining Scalar Degrees of Freedom - SPOINT ... 167
6.6 Coordinate Systems - CORD................................... 167
6.6.1 Subsidiary Coordinate System......................................... 168
vi Nastran Primer
6.7 Grid Points -GRID .................................................... 171
6.7.1 Grid Identification Number.............................................. 172
6.7.2 Geometric Coordinates ..................................................... 172
6.7.3 Displacement Coordinates................................................ 172
6.7.4 Permanent Constraints...................................................... 173
6.7.5 Default Values - GRDSET ................................................. 173
6.8 External and Internal Degrees of Freedom ........... 173
6.9 Displacement Sets..................................................... 174
6.9.1 Merged Data Sets ............................................................... 175
6.9.2 Mutually Independent Data Sets ..................................... 175
6.10 Multipoint Constraints - MPC ................................ 175
6.10.1 Uses of MPC........................................................................ 175
6.10.2 Reduction to the n-Set ....................................................... 177
6.10.3 MPC Data Entity ................................................................ 180
6.10.4 Combining MPC Sets - MPCADD................................... 180
6.10.5 Specifying the m-set Degrees of Freedom ...................... 181
6.10.6 Recovering and Printing MPC Forces............................. 181
6.10.7 Defining the Constraint Equation.................................... 181
6.11 Single Point Constraints - SPC................................ 184
6.11.1 Reduction to the f-Set ........................................................ 184
6.11.2 Single Point Constraint Forces ......................................... 184
6.11.3 SPC Data Entity .................................................................. 185
6.11.4 Selecting SPC Sets in Case Control.................................. 186
6.11.5 Specifying s-set Degrees of Freedom and Value ........... 186
6.11.6 Purging Degrees of Freedom ........................................... 186
6.11.7 AUTOSPC - Automatic Purging...................................... 187
6.11.8 Example: Specification of Single Point Constraints ...... 188
6.12 Static Condensation -OMIT and ASET.................. 189
6.12.1 Reduction to the a-Set........................................................ 189
6.12.2 Recovery of Omitted Degrees of Freedom ..................... 190
6.12.3 Physical Interpretation of [Goa]....................................... 190
6.12.4 Why Use Static Condensation? ........................................ 190
6.12.5 The OMIT and ASET Data Entities ................................. 191
6.12.6 Specifying Degrees of Freedom ....................................... 192
6.12.7 Example: Static Condensation.......................................... 193
6.13 Support for Free Bodies - SUPORT ........................ 193
6.13.1 Reduction to the l-Set ........................................................ 194
6.13.2 Rigid Body Transformation Matrix ................................. 194
6.13.3 Rigid Body Stiffness Matrix.............................................. 195
6.13.4 SUPORT Bulk Data ............................................................ 196
6.13.5 Specifying Degrees of Freedom ....................................... 196
Nastran Primer vii

6.14 Flexibility to Stiffness Transformation .................. 196
6.14.1 GENEL Bulk Data .............................................................. 200
6.14.2 Example:Beam Stiffness Using GENEL .......................... 201
6.14.3 Direct Specification of Element Stiffness ........................ 202
6.15 References .................................................................. 202
6.16 Problems .................................................................... 203
Chapter 7 Accuracy and Performance of Finite Elements
7.1 Element Performance ............................................... 205
7.2 Interpolation Failure................................................. 205
7.3 Effect of Element Shape ........................................... 206
7.3.1 Two Dimensional Patch Test............................................ 208
7.3.2 Three Dimensional Patch Test.......................................... 210
7.3.3 Patch Test Results .............................................................. 211
7.3.4 MacNeal Beam Tests.......................................................... 212
7.3.5 Rectangular Plate Tests ..................................................... 215
7.3.6 Cylindrical Shell Tests ....................................................... 217
7.3.7 Pinched Sphere................................................................... 218
7.3.8 Incompressible Material.................................................... 219
7.4 Summary of Test Results ......................................... 220
7.5 Problems .................................................................... 222
7.6 References .................................................................. 223
Chapter 8 Rigid and Constraint Elements
8.1 Rigid and Constraint Elements............................... 225
8.2 RROD - Extensional Constraint .............................. 226
8.3 RBAR - Rigid Link Connecting Two Grids........... 227
8.3.1 RBAR Example: Rigid Link Connecting Two Points.... 228
8.4 RTRPLT - Rigid Triangular Constraint Element.. 228
8.5 Rigid Constraint Element - RBE1 and RBE2 ......... 229
8.5.1 Example: Rigid Inclusion.................................................. 231
8.6 Elastic Constraint Element - RSPLINE .................. 231
8.7 Weighted Average Constraint Element - RBE3.... 234
8.7.1 Example -"Beaming" Loads and Masses......................... 235
8.8 References .................................................................. 236
Chapter 9 Material Properties
9.1 Isotropic Material...................................................... 240
9.1.1 One-Dimensional Elements.............................................. 241
9.1.2 Two-Dimensional Elements - Plane Stress..................... 241
9.1.3 Two-Dimensional Elements - Plane Strain..................... 242
viii Nastran Primer

9.1.4 Three- Dimensional Elements .......................................... 242
9.2 Anisotropic Material for Surface Elements........... 243
9.2.1 Orthotropic Material Properties ...................................... 243
9.2.2 Example: Orthotropic Material ........................................ 244
9.3 Orthotropic Material for Surface Elements........... 245
9.4 Anisotropic Material for Solid Elements ............... 246
9.4.1 Orthotropic Constants for Solid Elements ..................... 247
9.5 Transformation of Elastic Constants...................... 248
9.6 Temperature- Dependent Materials....................... 249
9.7 Material Property Table ........................................... 250
9.7.1 Example: Temperature- Dependent Elasticity............... 252
9.8 Modeling Composite Materials .............................. 253
9.8.1 Lamina Properties.............................................................. 254
9.8.2 Laminate Properties........................................................... 255
9.8.3 Defining a Composite Laminate Properties................... 257
9.8.4 Lamina Failure Criteria ..................................................... 259
9.9 Reference.................................................................... 260
Chapter 10 External Loads
10.1 Concentrated External Forces at Grid Points ....... 264
10.2 Force Vector Defined by Components................... 265
10.3 Force Direction Defined by Two Points ................ 266
10.4 Vector Normal to a Surface ..................................... 267
10.5 Load Applied to Scalar Point - SLOAD................. 268
10.6 Distributed Load on Beam - PLOAD1................... 269
10.6.1 Load Type and Direction .................................................. 269
10.6.2 Load and SCALE Fields .................................................... 269
10.7 Uniform Normal Pressure - PLOAD2 ................... 271
10.7.1 Sign Convention for Positive Pressure ........................... 271
10.8 Nonuniform Surface Tractions - PLOAD4............ 272
10.8.1 Direction of Surface Traction............................................ 272
10.8.2 Specifying Load Intensity ................................................. 272
10.8.3 Specifying the Loaded Face of Solid Elements .............. 273
10.9 Gravity Loads - GRAV............................................. 274
10.10 Rotation About an Axis- RFORCE ......................... 275
10.10.1Remarks............................................................................... 277
10.11 Combined Loading - LOAD.................................... 277
10.12 Enforced Displacements - SPCD ........................... 278
Nastran Primer ix
10.13 Enforced Deformation - DEFORM........................ 279
10.14 Thermal Loading ...................................................... 279
10.15 Grid Point Temperatures - TEMP and TEMPD ... 281
10.16 . Thermal Field for Axial Elements - TEMPRB..... 281
10.17 Thermal Field for Surface Elements....................... 282
Chapter 11 Static Analysis
11.1 System Matrices ........................................................ 285
11.1.1 Direct Specification of System Matrices.......................... 285
11.2 Constraint and Static Condensation ...................... 289
11.3 Static Loads................................................................ 289
11.4 Inertia Relief .............................................................. 290
11.5 Data Recovery ........................................................... 291
11.6 Input Specifications .................................................. 291
11.6.1 Executive Control Section................................................. 291
11.6.2 Case Control Section.......................................................... 292
11.6.3 Parameters........................................................................... 293
11.7 Solution Sequence Output....................................... 293
11.7.1 Automatic Output.............................................................. 293
11.7.2 Output of System Response Variables............................ 294
11.8 Fatal Errors ................................................................ 294
11.9 Grid Point Singularity Processing .......................... 295
11.9.1 Legacy NASTRAN............................................................. 296
11.9.2 MSC and NX NASTRAN.................................................. 296
11.9.3 Displacement Set Membership ........................................ 298
11.10 Grid Point Weight Generator.................................. 300
11.10.1Rigid Body Transformation Matrix ................................. 300
11.10.2Rigid Body Mass Matrix ................................................... 300
11.10.3Principal Mass Axes........................................................... 300
11.10.4Centroid............................................................................... 301
11.10.5Moments of Inertia............................................................. 301
11.11 Example Problems .................................................... 301
11.12 Cantilever Beam with Uniform Load .................... 302
11.12.1Element Properties............................................................. 303
11.12.2Distributed Line Load ....................................................... 303
11.12.3The Input File...................................................................... 304
11.12.4Running NASTRAN .......................................................... 306
11.12.5The Output File .................................................................. 306
11.12.6Discussion of Results......................................................... 310
x Nastran Primer
11.13 Simply-Supported Rib-Stiffened Plate................... 312
11.14 Problems .................................................................... 312
11.15 References .................................................................. 313
Chapter 12 Normal Modes Analysis
12.1 Dynamic Motion ....................................................... 315
12.2 The Eigenvalue Problem.......................................... 316
12.3 Standard Form of Eigenvalue Problem ................. 319
12.4 Specification of Element Inertia Properties........... 320
12.4.1 Structural Mass................................................................... 320
12.4.2 Nonstructural Mass ........................................................... 320
12.4.3 Consistent and Lumped Formulations ........................... 320
12.4.4 Mass Elements .................................................................... 321
12.4.5 Scalar Mass - CMASS ........................................................ 321
12.4.6 Concentrated Grid Point Mass - CONM1 ...................... 322
12.4.7 Offset Concentrated Grid Point Mass - CONM2........... 322
12.5 Real Eigenvalue Extraction Techniques ................ 323
12.6 The Inverse Power Method ..................................... 325
12.6.1 Specification of EIGR for Inverse Power with Shifts .... 325
12.6.2 Summary for the Inverse Iteration Method ................... 326
12.6.3 Finding Lowest Eigenvalue with Inverse Method........ 328
12.7 The Lanczos Method ................................................ 329
12.7.1 Description of the Lanczos Procedure ............................ 329
12.7.2 The Specification of the Lanczos Procedure - EIGRL ... 331
12.8 Tridiagonal Methods................................................ 332
12.8.1 EIGR for the Triadiagonalization Methods.................... 333
12.9 Dynamic Degrees of Freedom ................................ 334
12.10 Reducing Dynamic Degrees of Freedom .............. 335
12.10.1Guyan Reduction ............................................................... 335
12.11 Removing Matrix Singularities............................... 336
12.11.1Detect Singularities and Exit - ASING = -1..................... 337
12.11.2Remove Massless Degrees of Freedom ASING = 0....... 337
12.12 Input File for Normal Modes Analysis.................. 338
12.12.1Executive Control Section................................................. 338
12.12.2Case Control Section.......................................................... 338
12.13 Parameters ................................................................. 339
12.14 Example Problems .................................................... 339
12.14.1Cantilever Beam ................................................................. 339
12.14.2Simply Supported Plate .................................................... 340
Nastran Primer xi
12.15 Problems .................................................................... 341
12.16 References .................................................................. 342
Chapter A Use of Parameters
xii Nastran Primer

1 Getting Started
1.1 Historical Introduction

Designers and builders have sought guidance from the sciences throughout recorded
history. Before the development of science as we know it, builders used intuition, rules of
thumb and design rules that became trade secrets passed from generation to generation.
However to design modern structures for the industrial revolution engineers first had to
develop the concepts of physics and mathematics that lead to understand the factors associated
with the design of simple structures.
With the ability to predict the behavior of built-up structures comprised of rods, beams,
and plates, engineers can design bridges and steam engines that altered trade routes and
shortening the time required to transport products to ever-widening markets.
Early research and development activities focused on improving the analytical sciences
providing predictive tools that furthered technological advances. In hindsight, it is remarkable
that the engineering technology of the 1920s and 1930s was adequate for designing systems
such as jet engines and subsonic aircraft. However, predictive tools that supported World War
II aerospace technology were not adequate for the design challenge of high-performance
aircraft with swept wings. It was this challenge that lead to the revolution in predictive analysis
which was developed at Boeing by Turner and his colleagues.
This technology that Ray Clough termed “finite elements,” together with the parallel
development of digital and analog computers, provided designers with new insight on how
loads are carried in a swept wing. This understanding, in turn, lead to the detailed design of
weight efficient structures for transonic and supersonic flight.
Finite element technology was the result of a need and of a focused research program to
develop the enabling technology for high performance aircraft. Other technology sectors also
drove the development of predictive technology used to design jet engines, automobiles,
nuclear power plants, and safe toys. These industries benefited from the predictive power of
the finite element method.
1.1.1 The Finite Element Method

The finite element method synthesizes complicated structural systems as a connected
collection of objects, called finite elements, that embody local physical laws. General purpose
analysis systems containing a library of finite elements can then be developed. Engineers can
then used these modeling elements to create a mathematical model of complicated structural
assemblies. The computer then solves the resulting set of linear algebraic equations to
determine the predicted behavior.
In a very short time after the appearance of Turner’s seminal paper[1] every aerospace
company in the United States, if not in the world, developed its own proprietary finite element
program. These programs had strange-sounding names like SAMUS, ASKA, MAGIC, and
STRUDEL. Each of these programs was promoted by its developers but none was really
available as a commercial product.
1 Nastran Primer
Getting Started
1
Historical Introduction
1.1.2 Solving NASA’s Simulation Needs

In the early 1960s, the National Aeronautics and Space Administration (NASA)
determined that a large scale finite element-based program was an enabling technology for
designing the space system for lunar exploration. NASA’s goal was to develop a structural
analysis system to solve large problems that would employ the finite element method, use the
best currently available algorithms for solving matrix equations and eigenvalue analysis and
run on all major computers of that era. To satisfy this need, a NASA team headed by a
visionary named Tom Butler developed a set of specifications for an innovative general
purpose analysis system.
In hindsight NASA was asking for the moon, literally speaking. Finite element
technology was in its infancy. Fortran, the only high level language of that era, was not mature.
The solution algorithms for equation solving and eigenvalue analysis were very inefficient.
Computers were slow, had small memories, did not have disk storage, and were expensive. In
other words, the development of a system to satisfy NASA’s needs was quite a challenge. The
analysis system was called the NASA Structural Analysis System, which was of course called
NASTRAN®, a name that NASA registered as a trademark
The development team which included the Computer Science Corporation and The
MacNeal-Schwendler Corporation (MSC)1 was working in uncharted territory and made
significant contributions to computer science, numerical analysis, and finite element
approximation theory. When a linker was required for one of the target computers, they
designed and built one. When they needed a technique for allocation of memory at run time,
they developed the “open core” concept. When they needed new finite elements, they designed
them. And they designed a complete operating system with file allocation tables and I/O
utilities, a modular concept for organizing data structures and algorithms, which is now called
object oriented programming, and an internal language called DMAP for binding objects into
an executable code.
The original release of NASTRAN was enthusiastically received by the user community,
which at the time consisted of only the large aerospace companies and government laboratories
that could afford the multi-million dollar computers supported by MSC.Nastran. In years
immediately following the release of the program, NASA hosted a series of NASTRAN
Colloquia that provided a forum for a lively interchange of ideas between the user community
and MSC, which developed the program and then provided maintenance and development
services.
The NASTRAN system was released to NASA contractors in 1968, and users were
invited by NASA to share their experience in a series of NASA-sponsored Colloquia, starting
in 1971. The resulting colloqium proceedings are of historical interest since they chronicle:
1.The diffusion of finite element technology from the aerospace industry to other
industries,including automotive, chemicals, and electronics.
2.The growth of user understanding of the technology and its innovative application
to design.
3.The explosion of problem size and the collateral need for database procedures for
automated substructural analysis.
4.The development of computer graphics and the collateral development of pre- and
post processors.
5.The establishment of the finite element method on a firm variational foundation on
which the modern element library was built.
1.In 1999, The MacNeal-Schwendler Corp. changed its name to The MSC Software Cor-
poration. The company is still commonly referred to as MSC.
2 Nastran Primer
Getting Started
Historical Introduction
The early adapters of finite element analysis were engineering design analysts well versed
in that day’s analysis technology, which included design handbooks and specialty analysis
programs based on strength of materials approximations. These users were skeptical of finite
elements and accepted programs like NASTRAN, only after correlating MSC.Nastran
solutions with those given by Roark’s Handbook and solutions from Timoshenko’s Plates and
Shells. These early adapters found that the use of finite elements lead to reasonable results, but
accurate results depended on:
1.Understanding the physics of the problem.
2.Understanding the behavior of the elements.
3.Selecting the correct element, the number of elements. and their distribution.
4.Critically evaluating the results and making modification in the conceptual model
to improve the accuracy.
To a large extent, these same factors should influence today’s user in creating an analysis
model. In the time since the introduction of commercial finite element programs in the late
1960s and early 1970s, element behavior has improved and automated procedures for
improving accuracy, using either a refined mesh or higher ordered polynomial functions, have
been developed and implemented. But today’s engineer should be just as skeptical of the
results as the early adapters since the finite element method is simply a tool that can be misused
and abused.
Today we approach the use of finite elements from a practical rather than a theoretical
point of view. Our attitude is to ask the program to explain itself using computer aided
empiricalism. That is, let’s find out by doing or at least looking at the results from experiments
that give guidance in producing acceptable finite element analyses.
1.1.3 NASA NASTRAN

NASA continued to maintain the public domain version until 1995. After that date the
program was available by requesting a copy from the NASA Technology Transfer Office at
Langley Research Center. This version was recently released as open source software at the
NASA/NASTRAN GitHub site.
This version can be downloaded and built on your computer using a fortran computer. I
have successfully built using Visual Studio Fortran 6.5 under Windows XP; and GNU gfortran
using MinGW and MSYS on a Windows 8.1 computer. More on that later.
The public domain version clearly represents the state of the art in 1995. It needs new
element technology, the addition of sparse technology for solving matrix equations and for
eigenvalue extraction to name a few.
1.1.4 MSC NASTRAN

In the late 1960s, MSC started to market and support it’s own version of NASTRAN,
called MSC.Nastran. At first, it was similar to the public domain version. Over the years, MSC
has made a large investment in research and development. Dr. MacNeal, the founder of MSC
was responsible for a great deal of this technology which included: designing and adding
robust elements; embracing p-element technology, incorporating best of class algorithms for
sparse matrix equation solvers and Lanczos eigenvalue methods; incorporating multi-physics
capability; developing superelement technology; and, improving the basic architecture of the
program by using modern database technology.
MSC also consolidated the NASTRAN marketplace by purchasing two competitors,
Computerized Structural Analysis and Research Corporation (CSAR) and Universal
Analytics, Inc. (UAI).
Nastran Primer 3
Getting Started
1
NASTRAN® Architecture
1.1.5 NX NASTRAN
The FTC soon took notice of the reduced competition in the NASTRAN submarket which
FTC identified as programs which included the DMAP language. Their action caused MSC to
sell a copy of MSC NASTRAN to UGS Corporation that was subsequently acquired by
Siemens. Siemens markets this version as NX NASTRAN.
1.2 NASTRAN® Architecture
1.2.1 Independent Modules

Conceptually NASTRAN is a collection of independent program united called Modules.
These Modules were further categorized as:
• Structural Modules - Performed tasks associated with: element generation and
assembly; and, applying constraints and loads.
• Matrix Modules - Performed matrix addition, multiplication solution, eigen-extraction,
etc.
• Utility Modules - Performed task involving interaction with the database
1.2.2 Database
The database evolved from a collection of tape reels in the early versions of NASTRAN
to a robust relational database in NASA NASTRAN and in MSC.Nastran which are
implemented using high speed disks. The need for a persistent database in the early versions
of NASTRAN was supported by a feature called Checkpoint and Restart.
Modules communicate through the database. The program input is parsed and validated
and then written to the database as appropriate tables such as the Basic Grid Point Definition
Table (BGPDT), the Grid point Element Connection Table (GPECT) and the Equivalence
between Internal and External points (EQEXIN) to name a few.
1.2.3 Executive System

The Executive System employs anevent loop to run the modules appropriate for the
solution. It is driven by a table called the Operation Sequence Array (OSCAR) that defines
order of modules to be executed and the resources required to perform the operation. The
OSCAR, in turn is created from user input that defines the process steps to be performed using
a NASTRAN problem orient language called Direct Matrix Abstraction Programming
(DMAP). It is interesting to note that the FTC identified DMAP as a distinguishing feature of
the NASTRAN genre.
The Exec extracts the resource requirement, allocates input, output and scratch files, loads
parameter values and executes the module. The next OSCAR entry is read and the process
repeats until the OSCAR is terminated.
4 Nastran Primer
Getting Started
NASTRAN Technology
1.3 NASTRAN Technology
1.3.1 Engineering Mechanics

Although NASTRAN can perform multi-physics simulations, we consider only its
capability for determining the behavior of solid elastic bodies. Solid elastic bodies are a class
of continuous media whose behavior is governed by the a set of partial differential equations
collectively called the theory of elasticity. The foundations for this theory were established in
the late 19th century. However, it is one thing to have a theory and another matter to solve the
field equations for meaningful design problems.
Since designers needed insight about the behavior of structural elements, simplifications
were introduced into the theory of elasticity to make solutions tractable. One avenue of
research at that time was to consider the stress or strain field to be either two- or one-
dimensional.
One dimensional stress theory lead to the rod elements Two dimensional theory lead to
plane stress and plane strain solutions. The two-dimensional theory lead to an understanding
of important engineering design problems. Perhaps most notably are the stress concentration
about a hole in a rectangular membrane and with that an understanding of fatigue failure in
early jet transports; and the exact solution for a cantilever beam. One-dimensional theory lead
to the rod element widely used in bridge design.
However, many structural components were thin in either one or two dimensions. Using
appropriate assumptions for the variation of in plane displacement through the thickness for
resulted in beam, plate, and shells theories.
Each of these theories has its place in structural design. For example, large bridges and
skyscrapers were designed using rod, beam and plate elements, and the DC-3 and Boeing B-
29 were designed using rod and shear panel approximations. However, the need for low-
weight, high-performance structures, especially in the aerospace industry, required better
analysis support to design. So a talented group of engineers at Boeing invented a new
procedure for representing and solving complicated structures using an assemblage of simple
small structural components now called the finite element method.
1.3.2 Numerical Analysis

Engineering mechanics and continuum mechanics together with approximation theory
led to a plethora of finite elements types. Numerical analysis provides the algorithms for
solving systems of equations, the eigenvalue or single value problems
The algorithms for solving has improved significantly from early releases of NASTRAN.
I recall seeing a comment in one of the real eigenvalues routines called Inverse Power with
Shifts that said, “OK Lets eat up the budget”: Computers in those days were slow and
expensive. Fortunately great strides have been made reducing the time required to perform
solutions and the cost of computers is a small fraction of it was when only large mainfromas
could process simulation taks such as NASTRAN.
Sparse matrix techniques changed the economincs of solving large sets of equations. Now
all compertive FEA programs include sparse solvers and eigenvalue routines.
1.3.3 Computer Science

Computer Science is responsible for the basic arcitecture, and the data structures
associated with a particular computational genre. The data structures used by NASTRAN are
meant to optimize the data processing requirements, These same data structures might not
work at all for annother application.
Nastran Primer 5
Getting Started
1
Design of the Finite Element Model
Computer science also develops efficient algoritms for soring and finding things. These
algorithms are always improving so that a cutting edge program must continue to implement
the curent state of the art.
1.3.4 Computer Graphics

When NASTRAN was started the use of computers to create graphic images did not
compute (pun intended). There were no products for Computer Graphics. There was no CAD.
Computervision and Autocad were not on anyone’s radar screen. But there was plotting.
NASTRAN included a capability for creating structural plots and for XY plots.
The use of a graphic data base had to wait until someone invented Computer Aided
Design Drafting (CADD). This had the unintended consequenc of creating a geometric data
base; and soon someone with a tectronix scope figured out how to use the CADD data base to
create a finite element model. This led to the development of programs, independent of the
FEA program, called pre and post processors. PATRAN was developed by PDA Engineering,
since purchased by MSC; and, about the same time Desktop was developed by ANSYS, Inc.
Another program started about the same time was FEMAP, now a Siemens product.
1.4 Design of the Finite Element Model

The finite element model is a conceptual model of the part or assembly that embodies the
physics associated with its behavior and the displacement and force boundary conditions. This
model leads to an approximate solution for the partial differential equations governing the
behavior of the part. The physics of the behavior is, in turn, encapsulated by the finite element
objects. The design of the finite element model thus begins by selecting the finite element
object that incorporates the desired behavior.
Since the finite element object approximates the local behavior, the level of
approximation for solving the governing partial differential equations must be selected by
distributing a finite number of elements over the geometry of the part. This requires that the
user understand how element objects perform as a function of element shape and boundary
conditions. The finite element objects generate a set of linear equations representing the
differential equations. The particular solution of these equations is dependent on the
displacement and force boundary conditions.
1.4.1 Basic Modeling objects

The finite element analysis program generates and solves a set of equations based on
modeling objects that we define. The modeling objects include:
1..A material point object, called a GRID point, that specifies a point in the
undeformed structure that displaces with the structure. The grid point has both a
geometric attribute and a physics attribute. The geometric attribute defines the
location of the point in the undeformed media, and the physics attribute is the
displacement vector that defines where the point is in the deformed media.
2.An element object defining the local physics in a volume defined by connecting two
or more GRID points.
3.A material object that specifies the material properties for an element object.
4.A load object that specifies a load either with reference to a GRID or an element
object.
5.A constraint object that specifies displacement boundary conditions at a GRID
point, or a linear relationship among displacements at GRID points.
6 Nastran Primer
Getting Started
Design of the Finite Element Model
The design objective of the finite element model is to represent the physics of a
component or a structural system and to obtain its behavior to the accuracy required by the
design team. To accomplish this task, finite element analysts must understand the design tools
and how structures behave.
1.4.2 The Design Procedure

The design procedure uses the basic principles of mathematical physics. The accuracy of
the solution is a function of:
1.The number and location of the discrete material points called grid points.
2.The design of the finite element that represents the average behavior in a region.
The conceptual analysis model is designed by:
1.Understanding the strength of materials assumptions appropriate for the analysis.
2.Creating the analysis geometry that reflects the strength of materials assumptions.
3.Choosing the finite elements used to represent the behavior.
4.Choosing an adequate mesh refinement to represent the critical behavior of interest.
5.Creating a field of grid points on the analysis geometry that will lead to the desired
mesh density.
6.Connecting the grid points with the appropriate elements.
7.Describing the material properties.
8.Describing the loads over the analysis geometry that will generate the appropriate
loads at the grid points.
9.Describing the constraints, with respect to the analysis geometry, that will generate
the appropriate set of constraints at the grid points.
10.Describing special relationships that must exist between loads and/or
displacements using a set of linear constraint equations.
Except for the simplest of cases the creation of the finite element model is a major task
and should not be undertaken without the aid of graphic-oriented pre-processors.
Creating a finite element model is far different than creating a geometrically faithful
rendering of the geometry. In some cases the structural component can be much simpler than
the geometry model of a component. For example, a finite element model of a beam having a
very complicated geometric cross section can be modeled with a simple beam element. The
complicated geometry is captured by the cross sectional area and the moments of inertia. The
resulting solution provides the analyst with element forces that can be used to evaluate the
bending stresses at critical points in the cross section. The beam could also be modeled by solid
10 node tetrahedral elements. The beam is the better solution, if it is appropriate, for the
following reasons:
1.For a static simulation, one beam element will give the exact result as long as beam
theory is adequate.
2.The solid elements represent local behavior approximately. A large number of
Tet10 elements may be required to accurately predict the behavior.
3.The density of elements must be sufficient to represent local strain and stress
gradients.
The bottom line is that geometric fidelity is not necessarily the objective of the finite
element model. The objective is behavioral fidelity, and the creation of a physics model
requires an understanding of physics.
Nastran Primer 7
Getting Started
1
Processing the Finite Element Model
1.5 Processing the Finite Element Model

The finite element model is represented by a set of modeling objects stored on a file. The
finite element analysis program processes these objects and generates a solution. The principal
steps performed for a linear static analysis are described below.
1.5.1 Marshall the Model Data

The NASTRAN preface reads and verifies the modeling objects in the input file, and
creates data structures used in later stages of the solution process.
1.5.2 Generation of Element Stiffness Matrices

Element matrices are generated by the EMG Module using the connectivity defined by
the element modeling object, the geometry defined by the grid objects, the material defined by
the material objects, and properties defined by property objects. The element stiffness is
always generated in a local element coordinate system and then transformed to the
displacement coordinate system. NASTRAN allows the reference coordinate system for the
displacement vector for each grid object to be defined by the user.
1.5.3 Assemble the Unconstrained Stiffness Matrix

The element matrix assembler (EMA) nodule generates an assembled stiffness matrix by
appropriately adding the stiffness contributions of each element object connected at a grid
point. The assembler also looks for zeros on the diagonal of the stiffness matrix which, if not
removed by later processing, would lead to a singular matrix.
1.5.4 Detect and remove essential singularities

The assembled stiffness matrix is singular for two reasons. First, the displacement
functions for all elements include rigid body motion so the assembled stiffness contains rigid
body motion. Second, the element objects used in the model might not connect all
displacement degrees of freedom at the grid objects. The rigid body motion must be removed
by constraint objects. However, the removal of unconnected degrees of freedom is essentially
a data processing tasks included in most commercial finite element programs. This procedure
is called Autospc in NASTRAN.
1.5.5 Apply Constraints

After removing the unconnected degrees of freedom from the analysis, the constraints,
which are defined by constraint objects, are processed. The stiffness matrix and load vectors
are reduced by the resulting transformation matrices.
1.5.6 Generate Loads

The loads associated with the grid point displacements are generated using the load
modeling objects. The resulting load vector(s) are transformed to incorporate the effect of the
constraints.
1.5.7 Solve for Constrained Displacements

The solution is performed using Gauss elimination. The procedure has two phases. The
first phase is called decomposition since the stiffness matrix is transformed to lower and upper
triangular factors. This is followed the second phase called forward backward substitution. The
end result is the displacement vector for the constrained set of equations.
8 Nastran Primer
Getting Started
User Program Interface
1.5.8 Recover Unconstrained Displacements

The unconstrained displacements are recovered using the transformations generated by
the constraints.
1.5.9 Recover Element Behavior

MSC.Nastran recovers element strains sing the strain-displacement relations. After
determining the strains, the stresses are recovered and, if appropriate for the element, the
element forces are recovered.
1.5.10 Stress Averaging at the Grid Points

The stresses are evaluated at internal points in the element called the Barlow points, at
which the stress calculations are most accurate. The components of the stress tensors must be
averaged in some way and presented as either an average element stress or as grid point
stresses. The stresses at the grid points are not continuous, so grid point stresses must be further
processed to provide useful stress values to the analyst.
1.5.11 Strain Energy Density

The variation strain energy density provides the analyst with important insight about the
distribution of the element objects. It is generally calculated and plotted as part of the model
verification procedure.
1.5.12 Results Files

After generating all user requested behavioral variables NASTRAN generates several
files, including the output file whose file extension. is “out”. (This file has a “f06” extension
in MSC and NX NASTRAN).
1.5.13 Post Processing

The output file containing all the requested output may be very large. A graphic-oriented
post-processor provides the analyst with a visual display of the results.
1.6 User Program Interface
1.6.1 Input File

Using the text editor utility you can create a model file containing modeling directives.
An input text file is always required to execute the program. Later in this chapter a simple
example is presented to demonstrate the format of the input file.
1.6.2 Types of Solutions.

MSC.Nastran includes a number of solution sequences for performing various types of
analyses, as shown in Table 1-1.
Solution Number Description of Solution

1 Static analysis
2 Static Analysis with Inertia Relief
3 Normal modes analysis
5 Buckling
Nastran Primer 9
Getting Started
1
The Input File

7 Direct complex eigenvalue analysis
8 Direct frequency response
9 Direct transient analysis
10 Modal complex eigenvalue analysis
11 Modal frequency response
12 Modal transient analysis
14* Cyclic symmetry static analysis
15* Cyclic symmetry normal modes analysis
18 Cyclic symmetry frequency response
29 Nonlinear transient analysis
44 Static aeroelastic response
45 Aerodynamic flutter
46 Aeroelastic response
53 Static structural/steady state heat transfer
59 Transient structural/steady state heat transfer
Table 1-1 Solution Sequences in MSC.Nastran

Many of the algorithms in MSC.Nastran are beyond the scope of this text. Only solutions
for linear static, normal modes, and buckling, identified with an asterisk, are described.
These solution sequences are written in DMAP. NASTRAN performs a simulation using
any of the included solution sequences. However, an understanding of DMAP programming
can increase the scope of the problems that can be solved. In addition, unnecessary steps in a
solution sequence can be bypassed and the solution can be selectively stopped and started.
DMAP is described in Appendix E.
The DMAP language allows you to specify symbolic names for matrices in just like any
higher level programming language. DMAP takes care of all of the internal interface to the data
base.
1.7 The Input File

The input file begins with an optional NASTRAN statement followed by)
1.A required Executive Control section
2.A required Case Control section
3.A required Bulk Data section
1.7.1 NASTRAN Statement

The NASTRAN statement is used to specify values for certain system parameters which
are called system cells. The default values for the systems cells are usually appropriate so the
NASTRAN statement is only used in exceptional cases. The format of the NASTRAN
statement is
NASTRAN cellname1=value1, ..., callnamen = valuen
or
NASTRAN system(i)=value1, ..., system(n) = valuen
A complete list of parameters which can be set by the NASTRAN statement is presented
in the NASTRAN User Manual Vol. I.
10 Nastran Primer
Getting Started
The Input File
1.7.2 Problem Definition Section

The the problem definition is completely controlled by user-specified commands and data
statements in the input file. The problem is specified by
1.Defining the physical problems or the system of equations to be solved.
2.Providing user control over input and output.
3.Providing user control over executive functions.
The remaining sections of the input file performs one of the functions described above.
The problem definition section consists of
1.The EXECUTIVE CONTROL section performs function 3 above and is the first
section of the input file.
2.The SUBSTRUCTURE CONTROL section.
3.The CASE CONTROL section performs function 2 above and is the second section
of the input file.
4.The BULK DATA section performs function 1 above and is physically the third
section of the input file.
The subsection sections of the problem definition will be discussed in what appears to be
reverse order since the user typically defines the finite element model first in the Bulk Data
section, then considers Case Control, and then defines the Executive Control necessary to
accomplish a specific task.
The various sections of the problem definition section are separated by delimiter
statements described in Table 1-2. These delimiters are required in every MSC.Nastran input
data file even if the corresponding section is an empty set. These delimiters are
Delimiter Name Purpose

CEND Terminates the Executive Control section.
BEGIN BULK Terminates the Case Control section
ENDDATA Terminates the Bulk Data section
Table 1-2 Delimiter Statements
1.7.3 Getting Started with MSC.Nastran

Let us suppose that you, the user, have a background in matrix structural analysis and the
finite element method and perhaps have used another general purpose finite element program.
The question then isn't “what does the program do?” but “how do I pose my problem so that
MSC.Nastran will understand it?”.
A reasonable approach is to first define the physical problem then worry about the rest of
the MSC.Nastran input deck. It thus seems to make sense to think first about that portion of the
file called Bulk Data in which the finite element models us described by specifying
• The locations of the mesh points
• The element connectivity
• The element properties
• The material properties
• The constraints
Nastran Primer 11
Getting Started
1
The Input File
• The loads
• Eigenvalue extraction routines, if appropriate
• etc.
After defining the problem in the Bulk Data section it is reasonable to think about
selecting data items that are defined in Bulk Data that are also to be included in the present
analysis and, in addition, about specifying the behavioral variables to be generated. The
concept of sets of input data items is important since this allows the user to define any number
of logical sets of loads or constraints in the Bulk Data section and then, by using appropriate
Case Control directives, to specify which of the logical sets are to be included when the
program is executed.
Loads and constraints are given a numerical tag, called a set identification number, which
is specified by an appropriate Case Control directive. For example, suppose that a set of forces
has been given the set identification number 101. Then this set of loads would be specified in
static analysis by the following Case Control directive
LOAD = 101
In addition to selecting input data sets from Bulk Data the Case Control can be used to
• Print or not print the Bulk Data
• Generate and write element behavioral variables such as element forces, stresses, and
strain energy to the output file
• Write displacements, velocities, accelerations, forces of constraint, etc. to the output
file.
• Define Subcases
• Add the results of subcases
Finally, after completely defining the problem the type of analysis must be specified. This
is done by means of Executive Control directives in the Executive Control section which
• Specifies the solution sequence by referring to one of the solution sequence numbers
from Table 1-1.
• Specifies the maximum time for MSC.Nastran execution
• Requests that certain executive tables such as displacement set membership be written
to the output file.
The Primer includes a description of all of the MSC.Nastran capability associated with
statics, normal modes and buckling analyses. The organization of the Primer is as follows:
• Description of the physical format of Bulk Data in Appendix A
• Summary of all structurally-oriented Bulk Data by function with section references to
detailed description in the text in Sec. 1.10
• Description of Case Control in Sec. 1.11
• Description of Executive Control in Sec. 1.12
Examples of static, normal modes and buckling solutions using MSC.Nastran are
included in Chapters 12, 15 and 16, respectively. In addition the MSC.Nastran Demonstration
Manual includes examples of the use of all the solution sequences and most, if not all, of the
MSC.Nastran modeling capability.
12 Nastran Primer
Getting Started
Specifying the Finite Element Model - Bulk Data
After preparing the complete input file the user may still have a dilemma … how to
interface with the computer in order to execute MSC.Nastran. The execution of MSC.Nastran
on any one of the computer systems on which it is installed is controlled by machine-dependent
operating system commands. The computer command structure required for a particular
installation is described in the MSC.Nastran Application Manual for that particular computer
but the best course of action for a new user is to ask a colleague who is familiar with the system.
The new user should run small example problems, the examples in the Primer for
instance, to make sure that the program actually runs and gives the right answers. Then it is
time to think about the problems associated with defining the model of a real-life structure.
After mastering the MSC.Nastran capability described in the Primer the user is to move
an to non linear analysis. At that time it is appropriate to become familiar with the full set of
MSC.Nastran documentation. A selected menu of the documentation that will be of interest is
as follows:
MSC.Nastran User's Manual Section
Complete description of MSC.Nastran input file 2.0
Description of all Solution sequences 3.0
Description of the MSC.Nastran plotting 4.0
Description of DMAP 5.0
1.8 Specifying the Finite Element Model - Bulk Data

The finite element model is specified in the Bulk Data section of the input data file. The
finite element model is then processed to generate the matrix equations that represent the
behavior of the model.
A large proportion of the Bulk Data section is applicable to all problems in structural
mechanics, independent of whether a linear static or non linear dynamic analysis is to be
employed. All the Bulk Data which are independent of analysis type as well as applicable to
linear static, normal modes and buckling analysis will be considered in this book.
The Bulk Data describes the structural model, specifies sets of constraints and/or loads,
and the values of parameters used in the solution sequences. The concept of data sets is of
importance because it allows several load or constraint conditions to be included in Bulk Data.
The particular set of loads and constraints to be utilized at execution time is specified by a Case
Control Directive in the Case Control section.
Each logical Bulk Data statement may consist of one or more physical statements. Each
data statement must be either prepared according to formats described in Appendix A or by
using a free field format described below. Bulk Data statements have the following important
attributes or restrictions:
1.Each statement has 80 characters and is separated into ten fields
2.Each logical statement is given a tag which defines its attributes. This tag is a
mnemonic that must appear in the first field. The use of a mnemonic field allows
data statements to be inserted in Bulk Data in any order. The Bulk Data is then
sorted alpha numerically before interpreting the data fields.
3.Each type of data statement is described in alphanumerical order in the
MSC.Nastran User's Manual and the Quick Reference Guide. The Bulk Data
appropriate for linear static, normal modes and buckling are summarized by
function in Summary of Structurally Oriented Bulk Data (p. 15) of the present text,
and a detailed description and the associated capability is presented in later
chapters. The user should note that there are restrictions on the format (i.e., real or
integer) and range that the data elements must satisfy.
Nastran Primer 13
Getting Started
1
4.The datum may be placed anywhere within the appropriate fields. The input
interpreter assumes that blanks or a commas are a termination characters; thus,
imbedded blanks within the field are not allowed.
5.A data statement may consist of one or more physical lines so that continuation is
allowed. A continuation is effected by a unique continuation mnemonic in the last
field of the “parent” statement and the same mnemonic is preceded by a
continuation character in the first field of the “child”. The construction of
continuations is discussed in more detail in Appendix A, but it should be
emphasized at this point that the continuation mnemonic must be unique among all
Bulk Data mnemonics in the entire input file. When Bulk Data is sorted, the
existence of a special continuation character in the mnemonic field implies that a
“chiId”/"parent" relation exists, and MSC.Nastran then searches for a "parent" with
the same continuation mnemonic as that found on the continuation card.
1.8.1 Free Field Bulk Data

MSC.Nastran provides the user with an alternative to the Bulk Data card format which is
described in Appendix A. The alternate form allows the user to separate field entries for small
field cards by a comma (,) thereby greatly simplifying the task of preparing Bulk Data input.
The form of the free field Bulk Data card is
f1, f2, f3, f4, f5, f6, f7, f8, f9, f10
where the datum, fi, are separated by commas. Successive commas define blank fields so that
the free field Bulk Data statement
TABLED1,,,,,,,,,+T1
would be interpreted as TABLED1 in field one, blanks in fields two through nine and ‘+T1’ in
field ten.
1.8.2 Replication and Field Generation

A simple, but useful, set of duplication and generation commands can be imbedded in
Bulk Data. These commands are
1.Duplication of field from previous statement is defined by an equal sign, (=), in the
associated field.
2.Duplication of all remaining fields is defined by two equal signs, in the first of the
fields to be repeated.
3.Generation of an incremented value of a field from a statement is defined by the
construct, *(w), in the associated field where w is the real or integer value of the
increment.
4.Repeated duplication and incrementation is defined by including a duplication
control statement after the last statement definition having the form =(n), where n
is the number of additional statements to generate.
The rules which govern replication features are as follows:
1.Only small-field data statements can be generated
2.Continuations cannot be generated
3.Data items must be separated with one or more blanks or a comma
14 Nastran Primer
Getting Started
1.8.3 Summary of Structurally Oriented Bulk Data

The Bulk Data statements of interest for static and normal modes analyses of structures
are grouped by function and are described below
Data Statement Description Reference
1.8.3.1 Scalar Degrees of freedom
SPOINT Defines scalar point Defining Scalar

Degrees of
Freedom -
SPOINT
1.8.3.2 Coordinate Systems

CORD1C Cylindrical Coordinate System defined by point Subsidiary
coordinates Coordinate
System
CORD1R Rectangular Coordinate System defined by point Subsidiary
System
CORD1S Spherical Coordinate System defined by point Subsidiary
System
CORD2C Cylindrical Coordinate System defined by grid points Subsidiary
Coordinate
System
CORD2R Rectangular Coordinate System defined by grid points Subsidiary
Coordinate
System
CORD2S Spherical Coordinate System defined by grid points Subsidiary
Coordinate
System
1.8.3.3 Grid Points

GRDSET Default values for Grid data fields Grid Points
-GRID
GRID Defines Grid point Grid Points
-GRID
1.8.3.4 Constraint and Partitioning

MPC Specify linear relationship between degrees of freedom Multipoint
Constraints - MPC
MPCADD Forms the union of MPC sets Multipoint
Constraints - MPC
SPC Specify constraints to displacement degrees of freedom Single Point
Constraints - SPC
SPC1 Specify constraints to displacement degrees of freedom Single Point
Constraints - SPC
SPCADD Forms the union of SPC sets Single Point
Constraints - SPC
Nastran Primer 15
Getting Started
1
OMIT Specify degrees of freedom to be removed from the Static

analysis set by static condensation Condensation
-OMIT and ASET
OMIT1 Specify degrees of freedom to be removed from the Static
analysis set by static condensation Condensation
-OMIT and ASET
ASET Specify degrees of freedom to be retained in analysis set Static
Condensation
-OMIT and ASET
ASET1 Specify degrees of freedom to be retained in analysis set Static
Condensation
-OMIT and ASET
SUPORT Specify statically determinate set of degrees of freedom Support for Free
to remove rigid body motion Bodies - SUPORT
1.8.3.5 General Element

GENEL Flexibility or stiffness for a general element Support for Free
Bodies - SUPORT
1.8.3.6 Scalar Elements
CELAS1,2 Springs connecting Grid Point degrees of freedom Scalar Elastic

Elements
PELAS Property for elastic spring Scalar Elastic
Elements
CELAS3,4 Springs connecting scalar degrees of freedom Scalar Elastic
Elements
1.8.3.7 One Dimensional Elements
CROD Axial rod element Rod and Truss

Elements
PROD Property for the ROD element Rod and Truss
Elements
CTUBE An axial rod having a cross section of a thin walled tube Rod and Truss
Elements
PTUBE Property for the TUBE element Rod and Truss
Elements
CBAR A beam bending element having a uniform cross section The BAR Beam
Bending Element
PBAR Property for the BAR element The BAR Beam
Bending Element
BEAM A beam bending element having non uniform cross The BEAM
section Bending Element
PBEAM Property for the BEAM element The BEAM
Bending Element
CBEND A curved beam element Curved Beam
Element
PBEND Property for the BEND element Curved Beam
Element
16 Nastran Primer
Getting Started
1.8.3.8 Two-Dimensional Shear Element

CSHEAR Quadrilateral Shear Panel Shear Panels
PSHEAR Property for the SHEAR element Shear Panels
1.8.3.9 Two-Dimensional Shell Elements

CQUAD4 Flat isoparametric quadrilateral shell connection Shell Elements
CQUAD8 Curved isoparametric quadrilateral connection Shell Elements
CTRIA3 Flat isoparametric triangular shell connection Shell Elements
CTRIA6 Curved isoparametric triangular shell connection Shell Elements
PSHELL Common property card for QUAD4, QUAD8, TRIA3, Shell Elements
andTRIA6
PCOMP Layered composite element property PCOMP Layered
Composite
1.8.3.10 Three Dimensional Elements

CHEXA Six-sided isoparametric element with variable number Defining Solid
of node points Elements
CPENTA Five-sided isoparametric element with variable number Defining Solid
CTETRA Four sided isoparametric element with variable number Defining Solid
PSOLID Common properties for HEXA and PENTA Defining Solid
Elements
1.8.3.11 Constraint Elements

RROD Extensional constraint in direction of line segment RROD -
between two points Extensional
Constraint
RBAR Rigid link between two points RBAR - Rigid
Link Connecting
Two Grids
RTRPLT Rigid connection between three points RTRPLT - Rigid
Triangular
Constraint
Element
RBE1 Rigid connection between an arbitrary Rigid Constraint
Element - RBE1
and RBE2
RBE2 number of points Rigid Constraint
Element - RBE1
and RBE2
RSPLINE Elastic spline-fit constraint element Elastic Constraint
Element -
RSPLINE
RBE3 Constraint element with dependent degrees of freedom Weighted Average
equal to a weighted average of other degrees of freedom Constraint
Element - RBE3
1.8.3.12 Material Specification

MAT1 Isotropic material coefficients Isotropic Material
Nastran Primer 17
Getting Started
1
MAT2 Anisotropic coefficients for two-dimensional elements Anisotropic

Material for
Surface Elements
MAT8 Orthotropic material coefficients Orthotropic
Material for
Surface Elements
MAT9 Anisotropic material for three-dimensional stress state Anisotropic
Material for Solid
Elements
MATT1 Specifies temperature dependence of material Temperature-
coefficients on MAT1 Dependent
Materials
MATT2 Specifies temperature dependence of coefficients Temperature-
specified on MAT2 Dependent
Materials
MATT9 Specifies temperature dependence of material Temperature-
coefficients on MAT9 Dependent
Materials
TABLEM1,2,3,4 Material Property Tables Material Property
Table
1.8.3.13 Specification of Static Loads

FORCE Concentrated force at geometric grid point defined by Concentrated
vector components External Forces at
Grid Points
FORCE1 Concentrated force at geometric grid point defined two Concentrated
grid points External Forces at
Grid Points
FORCE2 Concentrated force at geometric grid point defined cross Concentrated
product External Forces at
Grid Points
MOMENT Concentrated moment at geometric grid point defined Concentrated
by vector components External Forces at
Grid Points
MOMENT1 Concentrated moment at geometric grid point defined Concentrated
two grid points External Forces at
Grid Points
MOMENT2 Concentrated moment at geometric grid point defined Concentrated
cross product External Forces at
Grid Points
SLOAD Load applied to scalar point Load Applied to
Scalar Point -
SLOAD
PLOAD1 Defines load distribution along length of BAR or Distributed Load
BEAM elements on Beam -
PLOAD1
PLOAD2 Defines a normal pressure load on two-dimensional Uniform Normal
elements Pressure -
PLOAD2
PLOAD4 Defines a surface load on two and three dimensional Nonuniform
isoparametric elements Surface Tractions
- PLOAD4
18 Nastran Primer
Getting Started
GRAV Specifies gravity vector for entire structure Gravity Loads -

GRAV
RFORCE Specifies rotational velocity vector for calculating Rotation About an
centrifugal forces Axis- RFORCE
LOAD Load combination Combined
Loading - LOAD
SPCD Specify an enforced displacement for static analysis Enforced
which is treated as a load rather than as a constraint Displacements -
SPCD
DEFORM Initial misfit of lineal elements Enforced
Deformation -
DEFORM
TEMP Grid point temperature Grid Point
Temperatures -
TEMP and
TEMPD
TEMPD Default grid point temperature Grid Point
Temperatures -
TEMP and
TEMPD
TEMPRB Specify temperature distribution for truss and beam . Thermal Field
elements for Axial
Elements -
TEMPRB
TEMPP1 Specify temperature distribution in plate and membrane Thermal Field for
elements Surface Elements
1.8.3.14 Mass Elements

CMASS1 Scalar mass connection Scalar Mass -
CMASS
CMASS
CMASS
CMASS
PMASS Scalar mass property Scalar Mass -
CMASS
CONM1 Specifies a 6 x 6 symmetric mass matrix partition on the Concentrated Grid
diagonal of the system mass matrix Point Mass -
CONM1
CONM2 Concentrated mass at (or optionally offset from) GRID Offset
point Concentrated Grid
Point Mass -
CONM2
1.8.3.15 Eigenvalue Extraction

EIGR Eigenvalue extraction method, Inverse Power The Inverse Power
Method
EIGR Eigenvalue extraction method, Tridiagonal Methods EIGR for the
Triadiagonalizatio
n Methods
Nastran Primer 19
Getting Started
1
EIGL Lanczos method of eigenvalue extraction The Specification

of the Lanczos
Procedure -
EIGRL
1.8.3.16 Miscellaneous Data Statements

DMIG Direct input of matrix coefficients Direct
Specification of
System Matrices
PARAM Specifies parametric input data Bulk Data Image
A-1
The Case Control Directives perform the following functions:

1.Specify set of Bulk Data input that are to be included in the analysis at execution
time.
2.Select solution techniques as appropriate.
3.Control the calculation and display of derived quantities.
4.Control subcases.
The Case Control Directives are free-field. The directive is separated from the
specification by one or more blanks or by an equal sign.
1.8.4 Data Selection

The concept of data sets in Bulk Data allows the user to define any number of different
load and constraint sets. The particular set to be used at execution time is then specified by an
appropriate Case Control Directive. The form of the data selection Case Control directive is
<name> = SID
where <name> indicates the particular type of data to be included in the analysis and SID is
the set identification number associated with the data in Bulk Data. In this book, the following
convention is to be used for variable names or quantities to be filled in by the user: The name
or quantity will be lowercase, preceded by ‘<‘, and succeeded by ‘>’. The angle brackets(<
and>) are not part of the name or quantity to be added by the user. All Case Control directives
may be shortened to the first four characters provided they are unique.
1.8.5 Load Selection

The following Case Control Directives are used to select sets of load-type Bulk Data:
DEFORM Selects initial element deformations specified by a set of

DEFORM Bulk Data.
LOAD Selects a set of static loads defined by LOAD, PLOAD, PLOAD1,
PLOAD2, PLOAD4, FORCE, MOMENT, FORCE1, MOMENT1,
FORCE2, MOMENT2, GRAV, RFORCE, or SPCD data statements.
1.8.6 Example: Load Specification

The Case Control Directives
DEFORM = 2
LOAD = 13
20 Nastran Primer
Getting Started
specify that the set of initial deformations defined by DEFORM Bulk Data with set
identification number 12 and the set of static loads defined by appropriate Bulk Data having
set identification number 13 are to be combined to define the set of grid point loads.
1.8.7 Thermal Field Selection

The following Case Control Directives are used to specify thermal fields defined by
temperature-type Bulk Data:
TEMP(LOAD) Selects a temperature set defined by TEMP, TEMPD,

TEMPRB, TEMPP1 Bulk Data to be used for calculating
equivalent thermal loads.
TEMP(MATERIAL) Selects a temperature set defined in Bulk Data to be used only
for determining temperature-dependent material! properties.
TEMPERATURE Selects a temperature set from Bulk Data for calculation of
thermal loads and temperature-dependent properties.
1.8.8 Example: Thermal Field Specification

The Case Control Directive
TEMP = 25
specifies that the temperatures defined by suitable Bulk Data with set identification 25 are to
be used to calculate both thermal loads and temperature-dependent properties.
Selection of Constraints
MPC Selects a multipoint constraint set defined by MPC or

MPCADD Bulk Data.
SPC Selects a single point constraint set defined by SPC, SPC1, or
SPCADD Bulk Data.
1.8.9 Example: Specification of Constraints

The directives
MPC = 10
SPC = 10
specify that multipoint constraints defined by MPC Bulk Data with set identification number
10 and single point constraints defined by SPC Bulk Data with set identification number 10 are
to be used to define constraints.
1.8.10 Specification of Eigenvalue Method

The user specifies an eigenvalue extraction algorithm for normal modes analysis with the
following Case Control Directive:
METHOD Selects a real eigenvalue extraction method defined by an

EIGR, EIGRL or EIGB Bulk Data.
Example: Specification of Eigenvalue Routine

The Case Control Directive
METHOD = 10
specifies that the EIGR Bulk Data with set identification number 10 defines the real eigenvalue
extraction algorithm to be used.
Nastran Primer 21
Getting Started
1
1.8.11 Output Control

The Case Control Directives that are used for output control are listed below in functional
groups.
1.8.12 Titling Directives

TITLE Defines text to be printed on first line of each page of output.
SUBTITLE Defines text to be printed on second line of each page of output.
LABEL Defines text to be printed on third line of each page of output.
1.8.13 Output Line Control

LINE Defines the number of data lines per printed page, default is 50
lines for 11-inch paper.
MAXLINES Defines the maximum number of output lines, default is
100,000.
1.8.14 Bulk Data Echo

ECHO options for Bulk Data
SORT Prints only sorted Bulk Data (default)
UNSORT Prints only the unsorted Bulk Data
BOTH Prints both unsorted and sorted Bulk Data
NONE Prints neither
PUNCH Writes the sorted Bulk Data on the system punch file
NOSORT Prints only Bulk Data cards which were changed on a
"check pointed" run (see Sec. 1.9.3)
SORT(t1, t2) Prints sorted echo of only card types t1, t2,...along with
their line numbers.
The user can selectively control the echo of selected portions of the unsorted Bulk Data deck
by requesting
ECHO = UNSORT
and including the desired Bulk Data between pairs of ECHOON and ECHOOFF statements in
Bulk Data.
1.8.15 Set Specification

SET Defines set of point numbers, element numbers, times, or
frequencies for use in point and element output directives.
1.8.16 Element Output Requests

ELFORCE Requests forces in a set of structural elements.
STRESS Requests the stresses in a set of structural elements.
ESE Requests strain energy for set of elements (see associated
PARAMeter "TINY" in Appendix B).
22 Nastran Primer
Getting Started
1.8.17 Grid and Scalar Point Requests

SPCFORCES Requests single point constraint forces at a set of points.
OLOAD Selects applied loads at a set of grid or scalar points.
ACCELERATION Requests accelerations for a selected set of grid or scalar points.
DISPLACEMENT Requests displacements for a selected set of grid or scalar points.
VELOCITY Requests velocities for a selected set of grid or scalar points.
VECTOR Equivalent to DISPLACEMENT.
GPFORCE Requests grid point force balance for a selected set of physical points.
MPCFORCE Requests multi-point constraint forces at a set of points
The form of the output directives is

<request> = N
where <request> is one of the output directives and N is the number of a SET that defines the
elements or grid points for which the calculated quantity is to be displayed. If N is ALL, then
the calculated quantity will be displayed for ALL elements or ALL points. A subset of ALL
must be defined before it can be referenced by an output control directive. For example, the
specification
ELFORCE = ALL
specifies that the element forces in all the elements in the finite element model are to be printed.
The specification
SET 5 = 5, 6, 10
ELFORCE = 5
specifies that element forces are to be printed for set five which consists of elements five, six,
and ten. The specification
SET 10 = 1 THRU 25, EXCEPT 10
DISP = 10
specifies that the displacements be printed at grid or scalar point 1 through 25, except for point
10.
1.8.18 Subcases
In general, a separate subcase may be defined for each loading condition. In static
analysis, separate subcases are also allowed for each set of constraints. Subcases may be used
in connection with output directives, such as requesting different output for each mode in a real
eigenvalue solution.
Case Control is structured so that a minimum amount of repetition is required when using
subcases. Only one level of subcase definition is provided and all items placed above the
subcase level (ahead of the first subcase) will be used for all following subcases, unless
overridden within the individual subcase.
In static problems, provision has been made for the combination of the results of several
subcases. This capability is convenient for combinations of individual loading conditions and
for superposition of solutions for symmetrical and antisymmetrical boundary conditions.
Typical examples of subcase definition are given, following a brief description of the
subcase Case Control Directives.
SUBCASE Defines the beginning of a subcase that is terminated by the next
SUBCASE Case Directive.
SUBCOM Defines a combination of two or more immediately preceding
subcases in statics problems. Output requests above the subcase
level are used.
Nastran Primer 23
Getting Started
1
SUBSEQ Must appear in a subcase defined by SUBCOM to give

coefficients for making the linear combination of the preceding
subcases.
SYM Defines a subcase in statics problems for which only output
requests within the subcase will be honored. Primarily for use
with symmetry problems where the individual parts of the solution
may not be of interest.
SYMCOM Defines a combination of two or more immediately preceding
SYM subcases in static problems. Output requests above the
subcase level are used.
SYMSEQ May appear in a subcase defined by SYMCOM to give the
coefficient for making the linear combination of the preceding
SYM subcases. A default value of 1.0 is used if no SYMSEQ card
appears.
REPCASE Defines a subcase in statics problems that is used to make additional output
requests for the previously run subcases. This capability is required because
multiple output requests for the same item are not permitted in the same
subcase. Output requests above the subcase level are still used.
MODES Repeats the subcase in which it appears MODES times for eigenvalue
problems. Used to repeat the same output request for several consecutive
modes.
The following examples of the use of subcases in the Case Control deck have been
included:
1.Static Analysis and Multiple Loads
DISPLACEMENT = ALL
MPC = 3
SUBCASE 1
SPC = 2
TEMP(LOAD) = 101
LOAD = 11
SUBCASE2
SPC = 2
DEFORM = 52
LOAD = 12
SUBCASE3
SPC = 4
LOAD = 12
SUBCASE4
MPC 4
SPC 4
Four subcases are defined by this example. The output request for displacements will be
honored for all subcases and MPC set three will be used for all subcases except subcase four
where MPC = 4 overrides the specification above the subcase level.
Since subcase one and two have the same constraint sets the solutions will be performed
simultaneously. The thermal load set 101 and external load set 11 will be combined for subcase
one, and the external load set 12 will be combined with deformation set 52 for subcase two.
Since there is no load set specified in subcase four, the SPC set four must include enforced grid
point displacements.
2.Linear Combination of Subcases
SPC = 2
OUTPUT
SET 1 1 THRU 10, 20,30
DISP ALL
24 Nastran Primer
Getting Started
STRESS =1
SUBCASE1
LOAD = 101
OLOAD = ALL
SUBCASE2
LOAD = 201
OLOAD ALL
SUBCOM 51
SUBSEQ 1.0, 1.0
SUBCOM 52
SUBSEQ 2.5, 1.5
Two static load cases are defined. SPC 2 is used for each subcase because the
specification is made above the subcase level. The subcase combination SUBCOM = 51 is a
linear combination of one times the result of subcase one and one times the result of subcase
two. The subcase combination SUBCOM = 52 consists of 2.5 times subcase 1 plus 1.5 times
the result of subcase 2. The displacements at all grid points and the stresses in the elements
defined by set one will be printed.
3.Statics Problem with One Plane of Symmetry
SET 1 = 1, 11, 21, 31, 41
SET 2 = 1 THRU 50
DISP = 1
ELFORCE = 2
SYM 1
SPC = 11
LOAD = 21
OLOAD = ALL
SYM 2
SPC = 12
LOAD = 22
SYMCOM 3
SYMSEQ 1., 1.
SYMCOM 4
SYMSEQ 1., - 1.
Two symmetry subcases are defined by subcases one and two. The symmetric subcase
combination SYMCOM 3 defines the sum of the previous subcases while SYMCOM 4 define
the difference. The nonzero components of static load will be printed for subcase one; and the
displacements associated with set one and the element forces for set two will be printed for
symmetric combination three and four. Note that only output requests defined within the
symmetric subcase are honored while output requests above the subcase level are honored by
SYMCOM.
4.Use of REPCASE in Statics:
SET 1 = 1 THRU 10
SET 2 = 11 THRU 20
SET 3 = 21THRU30
SUBCASE1
LOAD = 100
SPC 101
DISP ALL
SPCF = ALL
ELFO = 1
REPCASE 2
ELFO = 2
REPCASE 3
ELFO = 3
Nastran Primer 25
Getting Started
1
One subcase is requested for solution and with two additional subcases defined for output.
The displacements and SPC forces at all grid points and the element forces in SET 1 will be
displayed for subcase one. The element forces in SET 2 will be printed for REPCASE 2 and
those in SET 3 will be printed for REPCASE 3.
5.Use of MODES in Eigenvalue Problem
METHOD = 12
SUBCASE 1
MODES 5
STRESS = ALL
SUBCASE 6
DISPLACEMENT = ALL
The METHOD Case Control Directive points to an EIGR data statement that defines the real
eigenvalue method to be used and associated parameters. The MODES Case Control Directive
causes the results for each eigenvalue to be considered as a separate subcase starting with the
subcase number containing the MODES request. The stress will be printed in all elements for
the first five modes, and the displacements will then be printed for all additional modes.
If the MODES request is not included in a subcase, the output requests will apply to all
eigenvalues starting with the subcase number. The following output request could then be used
to print the displacements associated with the first three modes.
METHOD = 12
SUBCASE1
MODES = 3
DISP = ALL
SUBCASE 4
DISP = NONE
1.8.19 Case Control Examples

The following examples illustrate the use of Case Control Directives for static and normal
modes analysis.
1.8.19.1 Static Analysis

The following tasks
1.Print both sorted and unsorted bulk data.
2.Specify maximum number of output lines to be 75000.
3.Specify 35 lines of output per page.
4.Two sets of loads have been defined; a temperature set five, and a force set six. Print
the results for each and then combine the results by multiplying thermal loads by a
factor of one and the external loads by two.
5.Multipoint constraints for both load sets are defined by MPC set 14.
6.Single point constraints for both load sets are defined by SPC set 12.
7.Print external loads, displacements, and forces of reaction at all grid points.
8.Print element type output for elements 100 through 200, except for element 151.
could be performed by the following Case Control directives:
TITLE = STATIC ANALYSIS OF TRUSS
SUBTITLE = CASE CONTROL EXAMPLE
ECHO =BOTH
MAXLINES = 75000
LINES =35
SET 100 = 100 THRU 200, EXCEPT 151
MPC 14
26 Nastran Primer
Getting Started
SPC 12
SPCFORCE = ALL
OLOAD = ALL
DISPLACEMENT = ALL
STRESS = 100
ELFORCE = 100
SUBCASE 1
LABEL = THERMAL STRESSES
TEMP(LOAD) = 5
SUBCASE 2
LABEL STRESSES DUE TO EXTERNAL LOADS
LOAD 6
OLOAD ALL
SUBCOM 12
LABEL COMBINED OUTPUT
SUBSEQ 1.0, 2.0
TEMP(LOAD) = 5
BEGIN BULK
It should be noted that the TEMP(LOAD) = 5 must be included under the SUBCOM Case
Control directive in order to properly account for thermal strains when element stresses and
element forces are calculated for the combined loading condition.
1.8.19.2 Normal Modes Analysis

The following tasks
1.Print out sorted and unsorted Bulk Data.
2.Specify maximum number of output lines to be 125,000.
3.Specify 35 lines of output per page.
4.Print nodal displacements for all degrees of freedom for all nodes. (10 eigenmodes
are expected)
5.Print stress output for elements 2, 5, and 10 for eigenmodes four and five.
could be performed by the following Case Control directives:
TITLE = NORMAL MODE ANALYSIS OF TRUSS
SUBTITLE = CASE CONTROL EXAMPLE
ECHO = BOTH
METHOD = 12
MAXILINES = 125000
LINES = 35
SET 5 = 2, 5, 10
DISP = ALL
SUBCASE 1
LABEL = OUTPUT FOR MODES 1 THRU 3
MODES = 3
SUBCASE 4
LABEL = OUTPUT FOR MODES 4 AND 5
STRESS = 5
MODES = 2
SUBCASE 6
LABEL= OUTPUT FOR MODES 6 SUBCASE 11THRU
10
MODES = 5
SUBCASE 11
DISP = NONE
VELO = NONE
ACCEL = NONE
BEGIN BULK
Nastran Primer 27
Getting Started
1
Executive Control
1.9 Executive Control

The Executive Control section is the first of the four subsections that comprise the input
file. The Executive Control Directives perform the following functions:
1.Provides user control over the computer operating system by an optional
NASTRAN data statement.
2.Specifies the solution sequence number from Table 1-1 using the SOL directive.
3.Provides a means of modifying the DMAP for a altering solution sequences by
means of ALTER directive.
4.Provides a means of printing internal data sets.
The Executive Control Directives that accomplish these functions are freeform, which
means that the directive may start at any place on the data card. The directive is separated from
the specification by one or more blanks.
1.9.1 Solution Sequence

The solution sequence is specified by the SOL Executive Control statement
SOL K
where the allowable solution numbers are defined in Table 1-1 which is repeated here for
convenience.

101* Static analysis
103* Normal modes analysis
105* Buckling
106 Non linear static analysis
107 Direct complex eigenvalue analysis
108 Direct frequency response
109 Direct transient analysis
110 Modal complex eigenvalue analysis
111 Modal frequency response
112 Modal transient analysis
114* Cyclic symmetry static analysis
115* Cyclic symmetry normal modes analysis
118 Cyclic symmetry frequency response
129 Non linear transient analysis
144 Static aeroelastic response
145 Aerodynamic flutter
146 Aeroelastic response
153 Static structural/steady state heat transfer
159 Transient structural/steady state heat transfer
28 Nastran Primer
Getting Started
Executive Control
1.9.2 Specification of Execution Time

The time specified in Executive Control is MSC.Nastran time and is used in conjunction
with timing algorithms for each of the computer models. The system estimates the time
required for a matrix operation and then checks to see if there is sufficient time remaining,
based on the current time used and the MSC.Nastran time estimated. If sufficient time remains,
MSC.Nastran will terminate normally. The Executive Control directive is
TIME K
where K is the maximum allowable cpu execution time in minutes. The default is 60 minutes.
1.9.3 Diagnostic Print Requests

The DIAG Executive request provides the user with a means of requesting printout of
executive tables or for requesting override of defaults. The form of the request is
DIAG <K>
The DIAGs of general interest to the structural analyst are described below. Complete
tabulation of all DIAGs can be found in Section 3 of the MSC.Nastran Quick Reference Guide.
DIAGs which are of general interest are.
DIAG Meaning
Number
4 Print cross reference table which is produced during DMAP compilation.
5, 6 Print BEGIN and END time for each DMAP module
8 Print matrix summary data, called matrix trailers, as matrices are generated.
Matrix trailers contain information that may be helpful in determining the
cause of fatal errors.
14 Print the DMAP source compilation.
17 Writes the compiled DMAP sequence to the system PUNCH file.
31 Writes the Link Specification Table and Module Property Table to the output
file
Multiple options may be selected by using multiple integers separated by commas. Other
options and other rules associated with the DIAG directive that primarily concern the
programmer can be found in the NASTRAN User’s Manual Vol. I.
1.9.4 Termination of Executive Control

CEND (Required) Indicates end of Executive Control deck.
1.9.5 Executive Control Examples
1.9.5.1 Static Analysis

TIME 3
SOL 101
CEND
SOL 101 specifies that the static analysis is to be used. TIME specifies that the total
MSC.Nastran execution time is expected to be three minutes or less.
1.9.5.2 Static Analysis, Optional Output

TIME 3
SOL 101
DIAG 14
Nastran Primer 29
Getting Started
1
TRUSS EXAMPLE
CEND
The Executive Control is the same as the previous case except for the inclusion of the DIAG
directive. As a result of this statement the DMAP listing for the SOL 101 will be printed.
Recommended practice if DMAP ALTERS or user-supplied DMAP program is used.
1.9.5.3 Normal Modes Analysis

The previous examples are applicable to the normal modes analysis as well as static
analysis. The only change is that the SOL directive is changed to specify normal modes:
SOL 103
1.10 TRUSS EXAMPLE

Consider the following input file:
Line Entry
1 SOL 1
2 CEND
3 TITLE = STATIC ANALYSIS OF A PLANE TRUSS
4 LOAD = 101
5 SPC = 201
6 DISPLACEMENTS = ALL
7 ELFORCE = ALL
8 STRESS = ALL
9 BEGIN BULK
10 $...1...*...2...*...3...*...4...*...5...*...6...*...7....
11 GRID 1 0 0. 0. 0. 0
12 GRID 2 0 0. 180. 0. 0
13 GRID 3 0 0. 360. 0. 0
14 GRID 4 0 240. 0. 0. 0
15 GRID 5 0 240. 180. 0. 0
16 CROD 1 8 1 2
17 CROD 2 4 2 3
18 CROD 3 6 1 4
19 CROD 4 4 2 4
20 CROD 5 6 2 5
21 CROD 6 4 3 5
22 CROD 7 4 4 5
23 PROD 4 4. 30
24 PROD 6 6. 30
25 PROD 8 8. 30
26 MAT1 30 30.E6 0.3
27 FORCE1 101 3 60000. 4 1
38 SPC1 201 2 1
29 SPC1 201 12 4
30 PARAM AUTOSPC YES
31 ENDDATA
1.10.1 Components of the File.

The input file contains the three sections of the Problem Definition Section described
above. The Bulk Data Section completely describes the simulation model and it is by far the
largest part of the input file. The bulk data statements used in the above examples are described
in the following chapters of the Primer:
Chapter Modeling Component
6 Coordinate systems
6 Grid and scalar points
5 The element library
9 Constraints and partitioning
30 Nastran Primer
Getting Started
TRUSS EXAMPLE
9 Material properties
10 Loads
1.10.2 The Input File

The following section describes each line of the input file:
Statement
Description
Number
1 SOL 1
This statement specifies the solution algorithm that is to be run. You may prefer to use
the name of the algorithm, such as STATICS, rather than a solution number.
2 CEND
A required termination for the "Executive Control" section of the input file. The next
section, which starts after CEND, is the "Case Control Section."
3 TITLE
An optional Case Control directive specifying a title that is to be printed at the top of
every page of output.
4 LOAD
A required Case Control directive that selects a set of loads (set 101 in this case, from
the Bulk Data Section).
5 SPC
A Case Control directive that selects a set of single point constraints (set 201 in this
case).
6 DISPLACEMENTS
A Case Control directive that requests the recovery and output of displacements at
specified grid points (ALL) in the model.
7 ELFORCES
A Case Control directive requesting that element forces be calculated and output for
specified elements (ALL) in the model.
8 STRESS
A Case Control directive requesting that the calculation and output of element stress for
specified elements (ALL) in the model.
9 BEGIN BULK
A required Case Control directive terminating the Case Control Section.
10 Comment
Comment lines always begin with a dollar sign, “$”; everything to the right of the $-sign
is a comment. This comment helps align data in the Bulk Data fields.
11-15 The GRID Bulk Data described in Chapter 5 specifies the location of material points in
the analysis. The location of the grid points is shown by Table 1-4
16-22 CROD
The CROD, which is described in Chapter 6, specifies the use of the ROD element,
identifies the connected grid points G1 and G2, and points to a PROD property as
described by Table 1-5:
23-25 PROD
The PROD described in Chapter 6 defines the properties for each ROD element and
points to a material set as described Table 1-6
26 MAT1
The MAT1 described in Chapter 8 defines the physical properties for an isotropic
material having a MID = 30, E = 3.e7 and ν = 0.3
Nastran Primer 31
Getting Started
1
TRUSS EXAMPLE
Statement
Description
Number
27 FORCE1
The FORCE1 described in Chapter 10 specifies a concentrated force at grid point 3 with
a magnitude of 60,000 and in the direction of the line drawn from grid point 4 to grid
point 1 which is the negative X-direction.
28-29 SPC1
The SPC1 described in Chapter 7 specifies the displacement constraints to uy at Grid 1
and to ux and uy at Grid 4:
30 The PARAM data entity defines a scalar parameter named AUTOSPC whose value is
YES in the solution algorithm. This parameter controls the action of a processor which
detects singularities in the assembled stiffness matrix as described in Chapter 12.
31 ENDDATA
Required terminator of the Bulk Data Section.
Table 1-1 Grid Point Locations

Grid ID Reference X Y Z
Coordinate
System
1 Basic 0. 0. 0.
2 Basic 0. 180. 0.
3 Basic 0. 360. 0.
4 Basic 240. 0. 0.
5 Basic 240. 180. 0.
Table 1-2 ROD Elements

ROD ID PID G1 G2
1 8 1 2
2 4 2 3
3 6 1 4
4 4 2 4
5 6 2 5
6 4 3 5
7 4 4 5
Table 1-3 PROD Property definitions

PID Material ID Area
4 30 4.
6 30 6.
8 30 8.
32 Nastran Primer
Getting Started
TRUSS EXAMPLE
1.10.3 Truss Schematic Diagram

With the interpretation provided above for each Bulk Data statement, the physical model
can now sketched, including grid point locations, element connections, loads, constraints, etc.,
as shown in Figure 1-1.
y
60,000 lbs.
2 6
5
2 5
4
7
1
4
1 x
3
Figure 1-1 Truss Example.
1.10.4 Running the Truss Example

NASTRAN can be executed from a command window. The actual form of the command
is dependent on the version of NASTRAN. That being said, MSC Nastran is executed from the
command line by the following command:
Nastran filename [keyword1=value1 keyword2=value2 ... ]
where filename is the name of an input file having a “.dat” extension and where keyword is
one or more optional keywords which are described in Chapter 1 of the MSC.Nastran Quick
Reference Guide. Assuming you have saved the truss example input in file ‘truss.dat’, the
example would be executed by
Nastran TRUSS.DAT
The output file is given the name of the data file with an ".OUT" extension. The truss
output which will appear in the same directory containing TRUSS.DAT. After the program is
finished executing you can now use the editor to examine the output file.
1.10.5 Examining the OUT File

As you scroll through the output file using your editor, you will find the following
sections:
1.10.5.1 The Title Page:

Tells you the current version number and the creation date.
1.10.5.2 Executive Control Section Echo

This section is an echo of the Executive Control submitted for this example.
Nastran Primer 33
Getting Started
1
TRUSS EXAMPLE
1.10.5.3 The Case Control Section Echo

This section is an echo of the Case Control submitted for this example.
1.10.5.4 The Unsorted Bulk Data Echo

The Bulk Data echoed exactly as it appears in the input file. Note that there is a format
bar on the top of the page that helps justify the data fields of the Bulk Data.
1.10.5.5 The Sorted Bulk Data Echo

After reading the data file the Bulk Data statements are first sorted and then interpreted.
The interpretation phase makes sure that MSC.Nastran knows about the Bulk Data statement
name, such as GRID. The remaining datum must satisfy type and range criteria. A type is
associated with each data item, i.e., real, integer, or literal string and several datum have
specific ranges of values.
Each Bulk Data item has a line number in the sorted bulk data that is used to reference
detected errors. No errors exist in this example, so MSC.Nastran will go to the next section.
1.10.5.6 Grid Point Singularity Table

This table is automatically printed if there are any singularizes. The presence of the
AUTOSPC parameter with a value YES tells MSC.Nastran to remove any detected similarities
using either single or multipoint constraints as appropriate and to proceed with the analysis.
1.10.5.7 Solver message - Epsilon

The solver will print diagnostic information that you should look at for every analysis.
The information includes the value of normalized virtual work done by the calculated set of
displacements. If the displacements are the true equilibrium set, the virtual work should be
zero. Generally, the value will be small, usually 10.E-8 or less. A warning message will appear
if the value is greater than 10.E-5.
The displacements at the grid points.

Printed because of the DISP = ALL Case Control directive. You should see the following
nonzero values.
DISPLACEMENT VECTOR
POINTTYPE T1 T2 T3 R1 R2 R3
1 G 0.0 0.0 0.0 0.0 0.0 0.0

2 G 0.0 -6.75E-2-2.850E-1 0.0 0.0 0.0
3 G 0.0 -1.35E-1-7.512E-1 0.0 0.0 0.0
4 G 0.0 0.0 0.0 0.0 0.0 0.0
5 G 0.0 6.75E-2-3.650E-1 0.0 0.0 0.0
1.10.5.8 The forces of constraint.

Printed because of the SPCFORCE Case Control directive. You should see the following
non zero values.
FORCES OF SINGLE-POINT CONSTRAINT
POINTTYPE T1 T2 T3 R1 R2 R3
1 G 0.0 9.00E+4 0.0 0.0 0.0 0.0

4 G 0.0 -9.00E+4 6.00E+4 0.0 0.0 0.0
34 Nastran Primer
Getting Started
Problems
1.10.5.9 END OF JOB.

An End of Job statement is printed at the end of program execution.
1.11 Problems
[1] Run the truss example and compare the results in the output file, truss.f06, with those
presented in Chapter 1.
[2] For the truss shown by Figure 1-1, determine the exact node point displacements and
element forces and compare with the calculated results.
1.12 References
[1] M.Turner, R. Clough, H. Martin and L. Topp, “Stiffness and deflection analysis of
complex structures”, J. Aero. Sci., Vol 23, No. 9, Sept. 1956, pp. 805-823
Nastran Primer 35
Getting Started
1
References
36 Nastran Primer
2 Basic Relationships
from Engineering Mechanics
2.1 Theoretical Foundations

The primary focus of this book is understanding the theory and use of finite elements for
simulating the linear behavior of structures in support of engineering design. The theoretical
foundation for this area of engineering mechanics is called the linear theory of elasticity. The
complete nonlinear theory would include the effects of nonlinear geometry including large
displacements and non linear strain-displacement relations and non linear material behavior.
In the present study we will consider only linear material behavior but we will include non
linear strain-displacement relations which are required to present the theory of buckling.
In this chapter we will state the relevant principles that will allow us to study the linear
theory of finite elements. In later chapters we will present general concepts for matrix
structural analysis and the solution of linear sets of equations, the inclusion of mass effects and
the study of structural dynamics and the inclusion of first order non linear effects in the element
formulation and the study of linear buckling.
2.2 Continuum Mechanics

The structure is considered to be a continuum wherein the material is distributed without
gaps and empty spaces. Continuum mechanics is then employes to study the behavior of the
solid continuum under the effect of external forces. The continuum is assumed to have certain
material properties including the mass density per unit volume, ρ, and certain other properties
which will be identified and described in subsequent sections
37 Nastran Primer
Basic Relationships from Engineering Mechanics
2
Kinematic Relations
2.3 Kinematic Relations

In the continuum mechanics the deformed state is defined by determining how each
material point moves under the action of external forces. The position of a material point, A, is
identified by a position vector, R which is also represented in the text by the notation {R}=
{Rx , Ry , Rz} as shown by Figure 2-1 where the components are understood to be defined with
reference to the cartesian reference system.
x3
A u
r
A*
x3 R
y3
x2 x1
x1
y2 x2
y1
Figure 2-1 Displacement Vector
As a result of applied loads and enforced boundary displacements, the material point
moves to a nearby point, A* having a position vector, {R}, in the deformed configuration. The
displacement of the point from {r} to {R} is represented by the displacement vector, {u}
having components u, v, and w in the x, y and z directions, respectively. The displacement at a
point P then defines as:
{u} = {R} -{r} (Eq. 2-1)
2.4 Strains
As the continuum moves as a result of displacements, an elemental volume in the
neighborhood of the point A translates and its neighborhood rotates as a rigid body, and strains.
The internal strain energy of the body is associated with strains which cause volumetric
changes. The linear components strain at a point are given by:
{ε} = [L]{u} (Eq. 2-2)
Where [L] is the linear differential operator:
38 Nastran Primer
Strains
∂
0 0
∂x
∂
0 0
∂y
∂
0 0
[L] = ∂z (Eq. 2-3)
∂ ∂
0
∂y ∂x
∂ ∂
0
∂z ∂y
∂ ∂
0
∂z ∂x
The strains produced by [L] are called the engineering strains. We should note that the six
independent components of strain are represented as the following pseudo vector:
 ε xx 
 
 ε yy 
 
 ε zz 
ε =   (Eq. 2-4)
 γ xy 
 
 γ yz 
 
 γ zx 
The components ε xx, ε yy, ε zz are called the extensional strains and the
components γ xy, γ yz, γ zx are called the shear strains. The extensional strains are changes in
length per unit of original length in the coordinate directions; and, the shear strains are changes
of the ninety degree angle formed by the two coordinate axes associated with its idices. It
should be noted that the order of the strain components is rather arbitrary and different authors
prefer alternate orders. The same order will be used for the components of stress.
The engineering strains are related to, but are not equal to, the components of a second
order strain tensor which we will designate in indicial notation as lij where the indices have a
range of i, j = x, y, z. It is convenient to use the tensor strain components in tensor
transformation equations. The two sets of strains are related as follows:
l xx = ε xx l yy = ε yy l yy = ε zz
and
(Eq. 2-5)
γ xy γ yz γ zx
l xy = ------
- l yz = ------ l zx = ------
2 2 2
The components of the tensor are symmetric under interchange of the subscripts as are the
components of the engineering strains. Combining equations (Eq. 2-4) and (Eq. 2-5) gives the
following strain-displacement relations:
Nastran Primer 39
2
Stress Vector on a Surface
 
 ∂u 
 ∂x 
 
 ∂v 
 
 ∂y 
 
 ∂w 
 ∂z 
{ε} =   (Eq. 2-6)
 ∂u ∂v 
 + 
 ∂y ∂x 
 ∂v + ∂w 
 
 ∂z ∂y 
 
 ∂w + ∂u 
 ∂x ∂z 
 
The set of three displacements and six strain components represent the kinematic
behavior of the element. In order to complete the definition of the behavior of a deformable
solid we need to describe the internal forces and the associated equilibrium conditions.
2.5 Stress Vector on a Surface

The intensity of a force, {∆F}, acting at a point P on a surface with having an outward
unit normal, {n}, is represented by the stress vector,
 
 τx 
 
{ τ } =  τy  (Eq. 2-7)
 
 τz 
 
where:
{τ} = lim {-------------

∆F }- (Eq. 2-8)
∆A ⇒ 0 ∆A
40 Nastran Primer
Stress Vector on a Surface
x3
Deformed Body
n
P ∆F τ
∆A
x2
x1
Figure 2-2 Stress Vector
The stress vector is a function of position, time and the orientation of the unit normal. The
stress vector can be resolved into a normal components in the direction of the normal and two
orthogonal shear components acting in the surface.
Nastran Primer 41
2
Components of Stress
2.6 Components of Stress

We now consider a differential volume at an arbitrary point P in the continuum where
stress vectors act an each face as shown by Figure 2-3 The stress vector acting on faces having
positive normals in the x, y and z directions are {τ}x, {τ}y, and {τ}z, respectively where the
subscript on the vector defines the face on which the vector acts by denoting the coordinate
associated with the normal to the face.
{ τ }z
i3
i2
i1
y
{ τ }y
x
{ τ }x
Figure 2-3 Stress vectors acting on the faces

σ zz
τ zy
τ zx
τ yz
τ xz
τ yx σ yy
τ xy
σ xx
Figure 2-4 Components of Stress
42 Nastran Primer
Components of Stress
Each stress vector acting on a faces can be resolved into components in the coordinate
directions. The set of all components of the stress vectors acting on all faces is called the
components of stress and is represented by a three by three matrix [σ]. Each row of the matrix
represents components of the stress vector acting on a face.
σ xx σ xy σ xz = σ xx τ xy τ xz = τx τy τz
x
σ yx σ yy σ yz = τ yx σ yy τ yz = τx τy τz
y
σ zx σ zy σ zz = τ zx τ zy σ zz = τx τy τz
z
Where the shear stress components are represented in the following by either the τij or σij
since they are equivalent. The first subscript of the general matrix element, σij, defines the face
on which the stress vector acts and the second defines the direction. The set [σ] is called the
components of stress and transforms as a second order tensor as described in Appendix D.
The components of stress acting normal to a surface and are called the normal stresses and
are denoted by σxx, σyy, and σxx. The other components of stress which act tangent to the
surface are called the shear stresses and denoted by τxy, τyz and τzx.
The components of stress are actually components of a second order tensor that has nine
components. Under most conditions the tensor is symmetric and there are then only six
independent components of stress. The six independent components of stress are represented
by the pseudo vector, {σ}.
 σ xx 
 
 σ yy 
 
 σ zz 
{σ} =   (Eq. 2-9)
 τ xy 
 
 τ yz 
 
 τ zx 
It should be noted that the order of the components of stress in the pseudo vector is rather
arbitrary, and different authors prefer different orders. The components of stress are positive
in the positive coordinate directions when acting on a face having a positive normal; and, are
positive in the negative coordinate directions when acting on a face having a negative normal.
In Figure 2-4, we only show the stress components on faces with positive normals.
It is not hard to show that the stress vector acting on a face with a normal, {n}, is:
   
 τx   σ xx n x + σ yx n y + σ zx n z 
   
τ
 y  =  σ xy n x + σ yy n y + σ zy n z  (Eq. 2-10)
   
 τz   σ xz n x + σ yz n y + σ zz n z 
   
Nastran Primer 43
2
Constitutive Behavior
2.7 Constitutive Behavior

The components of stress and components of strain are related by a material law.
Empirical evidence has shown that, for a wide range of materials having engineering
significance, a linear equation relating the six components of stress to the six components of
strain compares well with test data. The most general linear relation is of the form:
{ σ } = [ E ] ( { ε } – { ε0 } ) + { σ0 } (Eq. 2-11)
where {ε0} are thermal strains and {σ0}are initial stresses.

The matrix of elastic coefficients, [E], can be shown to be symmetric and thus has at most
21 independent elastic coefficients. Using various material symmetry conditions the number
of independent coefficients can be reduced. An orthotropic material that has reflective
symmetry about three mutually orthogonal planes has nine independent material coefficients;
and, an isotropic material, for which material properties are independent of direction or
orientation of axes at a point, has two independent material coefficients. The two independent
coefficients are generally taken to be the so-called engineering constants the modulus of
elasticity, or Young’s modulus, E, and Poisson’s ratio, ν . These two constants are found from
tension test data. Young’s modulus is the slope of the load deflection curve; and, Poisson’s
ration is the strain normal to the direction of loading measured as a fraction of the direct strain.
Poisson’s effect is such that the strains in the transverse direction are the negative of those in
the direction of the loads.
Based on test data the extensional strains are related to the stresses as follows for an
isotropic material:
1
ε xx = --- [ σ xx – ν ( σ yy + σ zz ) ]
E
1
ε yy = --- [ σ yy – ν ( σ zz + σ xx ) ] (Eq. 2-12)
E
1
ε zz = --- [ σ zz – ν ( σ xx + σ yy ) ]
E
The shear strains are then given by:
1
γ xy = ---- τ xy
G
1
γ yz = ---- τ yz (Eq. 2-13)
G
1
γ zx = ---- τ zx
G
where G = E ⁄ ( 1 + 2ν ) .
It is rather straight-forward, for an isotropic material, to obtain the inverse relation:
44 Nastran Primer
Equilibrium Conditions - Newtonian Mechanics
( 1 – ν )-
2------------------- 2ν - 2ν -
 -------------- -------------- 0 0 0 
σ xx  1 – 2ν 1 – 2ν 1 – 2ν ε xx 
   
 σ yy  2ν 2( 1 – ν) 2ν  ε yy 
  --------------- -------------------- --------------- 0 0 0  
1 – 2ν 1 – 2ν 1 – 2ν
 σ zz  E -  ε zz 
  = 2------------------- 2ν - 2ν - ( 1 – ν -)
2-------------------   (Eq. 2-14)
 τ xy  (1 + ν) -------------- -------------- 0 0 0  γ xy 
1 – 2ν 1 – 2ν 1 – 2ν
   
 τ yz  0 0 0 1 0 0  γ yz 
   
 τ zy  0 0 0 0 1 0  γ zy 
0 0 0 0 0 1
2.8 Equilibrium Conditions - Newtonian Mechanics

The stress components must satisfy equilibrium. Employing Newtonian mechanics we
can write the equilibrium conditions as:
T
[L] {σ} – {X} = 0 (Eq. 2-15)
Where [L] is the same operator that was used in the strain-displacement equations and where
{X} are the external body forces. The equilibrium conditions in expanded form are:
∂σ xx ∂τ yx ∂τ zx
+ + – Xx = 0
∂x ∂y ∂z
∂τ xy ∂σ yy ∂τ zy
+ + – Xy = 0 (Eq. 2-16)
∂x ∂y ∂z
∂τ xz ∂τ yz ∂σ zz
+ + – Xz = 0
∂x ∂y ∂z
2.9 Principle of Virtual Work

The principle of virtual work is an alternative statement of equilibrium. The virtual work
can be expressed as the sum of the virtual work of the internal and external forces. Since the
virtual work of the internal forces is the negative of the virtual change of the internal strain
energy, U, the virtual work performed by all of the forces acting on a body moving through a
virtual displacement is expressed as:
δW = – δU + δW ex (Eq. 2-17)
where:
∫ {σ}
T
δU = { δε } dΩ
Ω
(Eq. 2-18)
∫ {τ} { δu } dΓ + ∫ { X } { δu } dΩ
T T
δW ex =
Γ Ω
where Ω is the volume of the continuum and Γ is its surface.

The principle of virtual work states that:
Nastran Primer 45
2
Solutions from the Theory of Elasticity
If the virtual work is equal to zero for any arbitrary kinematically admissible
virtual displacement then the system is in equilibrium.
This is a powerful statement that can be used to obtain the equilibrium equations
presented in the previous section directly by substituting the strain displacement relations,
which are kinematic constraints on the displacements, into the virtual work expression. The
equilibrium equations, which result after employing integration by parts and noting the
arbitrary nature of the virtual displacements, can then be interpreted as a set of Lagrange
operators which are the forces of constraint required to enforce the kinematic conditions.
The principle of virtual work and other energy functionals are a very important foundation
for developing approximate solutions since equilibrium is ‘weakly’ enforced in some
integrated sense rather than being enforced point-wise which is the case if we insist on
satisfying the equilibrium equations everywhere in the continuum. The principle of virtual
work and energy functionals will be used in the next Chapter to obtain the solutions for certain
continuum problems using a Rayleigh Ritz approach; and, the equilibrium equations for finite
elements.
2.10 Solutions from the Theory of Elasticity

The theory of elasticity problem is to find a set of three displacements, six strains and six
stress components that satisfy the strain displacement equations, the equilibrium equations and
the equation of state on the boundary and in the interior of the deformable body and satisfy
specified boundary conditions. Not too surprisingly, there a very few solutions to the three
dimensional field equations that are of practical engineering importance. However, those
solutions which are available are of interest to finite element practitioners since they provide
exact solutions which can be used to test elements for which the behavior is three dimensional.
We provide two examples in the following sections.
2.10.1 Uniform Stress

We consider a uniform tension force uniformly distributed over the end of a prismatic bar
so that the stress vector at the ends at x = 0 and x = L of the solid is a vector in the x - direction.
The magnitude of the stress vector is P/A where P is the applies load and A is the cross sectional
area. The only non zero components of stress on the end face is σxx, whose magnitude is P/A.
The equilibrium equations and stress boundary conditions are satisfies by taking:
σxx= P/A, σyy = σxx = τxy = τyz= τzx = 0 (Eq. 2-19)
For an isotropic material the strains are given by:
1
ε xx = --- σ xx
E
–ν
ε yy = ------ σ xx (Eq. 2-20)
E
–-----ν-
ε zz = σ
E xx
With all shear strains equal to zero. The strains are constant and therefore satisfy the
compatibility equations. The displacements can be determined by integrating the strain
displacement relations.Taking the origin of the coordinate at the left face and the x - axis to be
at the center of the solid as shown hives the following displacements:
46 Nastran Primer
Strategies for Solving Real Problems
σ xx Px
u = -------- x = -------
E AE
νσ xx
v = – ----------- y = – νPy
---------- (Eq. 2-21)
E AE
νσ xx
w = – ----------- z = – νPz
---------
E AE
2.10.2 Stretching of a Prismatic Bar by Its Own Weight

Letting ρ g be the weight per unit volume of the bar the body forces are:
pz = –ρ g ;p x = 0 ;p y = 0 (Eq. 2-22)
It is shown in [2] that the components of stress which satisfy the equilibrium equations
and the boundary conditions are:
σ zz = ρ gz , and σ xx = σ yy = τ xy = τ yz = τ zx = 0 (Eq. 2-23)
2.11 Strategies for Solving Real Problems

The three dimensional field equations were formulated early in the 19th century but were
of little use in supporting the design of structures such as bridges, pressure vessels and steam
ships. The needs to support design with simulation lead engineers to seek approximate
solutions based on simplifying assumptions. Two major classes of approximations were used.
One class, called plane stress and plane strain approximations, assumed strains or stresses were
functions of two rather than three dimensions and lead to two dimensional theory of elasticity.
The other major class is based on assumed displacements for plate and beam bending.
Both classes of approximations lead to solutions which were of great use to design
engineers. For example:
1.Determining the stress concentration due to hole in a stretched plate,
2.The fact that a concentrated load led to infinite local stresses,
3.Determining the exact solution for beam bending using two dimensional theory of
elasicity
4.Determining strength of material solutions for beam bending.
5.Stresses due to torsion using two dimensional theory of elasticity
6.Plate bending for classical boundary conditions and general loading using strength
of materials approximations
7.Stresses in cylindrical and spherical shells
However, the needs of modern technology required more accurate analysis support to
design. For example, stress analysis techniques bases on simple design formulas that had been
adequate for the aircraft from the Wright Flyer through the Boeing B-29 were no longer
adequate for the new swept wing B-47. New techniques were required and, at this time a great
deal of research was devoted to the behavior of skewed plates which lead to a better
understanding of the behavior of the bending of non rectangular plates. However, more
accurate results analysis were required to design an efficient swept wing.
Nastran Primer 47
2
Two Dimensional Theory of Elasticity
Fortunately, a group of talented engineers at Boeing, developed a new approximation

technique that lead to a much better representation of the structure. Their original work [1] was
widely recognized as a revolutionary way of modeling and analyzing structures and lead, in a
very short time, to an industry-wide development of general purpose analysis engines based on
this new technology which Clough called the finite element method.
The early development of two and three dimensional elements was creative, to say the
least. These early developments were based on direct formulations using the field equations
for one, two and three dimensional elasticity; and plate, shell and beam theory. Only the beam
bending element performed well by today’s standards but the early two and three dimensional
elements gave tremendous insight to load paths in the complex structures being developed in
the late 1950s and early 1960s.
The field equations for two dimensional theory of elasticity, plates and beams will be
summarized in succeeding sections. The development of the finite element method will be
discussed in Chapter 4.
2.12 Two Dimensional Theory of Elasticity
2.12.1 Plane Stress

The basic assumption associated with two dimensional theory is that the behavioral
variables do not vary in one dimension. In addition, if a thin plate is loaded only by forces at
the boundaries parallel to the plane of the plate all components of stress in the z-direction are
assumed to vanish with those in the x and y-directions then the only non-zero components of
stress are:
 
 σ xx 
 
{ σ } =  σ yy  (Eq. 2-24)
 
 σ xy 
 
2.12.2 Plane Strain

Consider bodies for which one dimension is large compared to the other two, and that the
loading is perpendicular to and does not vary in the long dimension. In this case the strains in
the z-direction are assumed to vanish so that the only non zero components of strain are:
 
 ε xx 
 
{ ε } =  ε yy  (Eq. 2-25)
 
 ε xy 
 
2.12.3 Equilibrium Equations

For either plane stress or plane strain the equilibrium equations are:
48 Nastran Primer
∂σ xx ∂τ xy
----------- + ---------- + X = 0
∂x ∂y
(Eq. 2-26)
∂σ yy ∂τ xy
----------- + ---------- + Y = 0
∂y ∂x
2.12.4 Strain Displacement Relations

For plane stress and plane strain, the strain displacement relations reduce to:
∂u
ε xx = -----
∂x
ε yy = ∂----v- (Eq. 2-27)

∂y
∂u ∂v
γ xy = ----- + -----
∂y ∂x
2.12.5 Constitutive Relations

For plane stress the constitutive relations for an isotropic material are:
   
 ε xx  1 –ν 0  σ xx 
  1  
 ε yy  = --E- – ν 1 0  σ yy  (Eq. 2-28)
   
 γ xy  0 0 2(1 + ν)  τ xy 
   
For plane strain we have:
σ zz = ν ( σ xx + σ yy ) (Eq. 2-29)
Hook’s law for this case is then:
   
 ε xx  2
( 1 – ν ) –ν ( 1 + ν ) 0  σ xx 
  1-  
 ε yy  = E –ν ( 1 + ν ) ( 1 – ν2 )
--
0  σ yy  (Eq. 2-30)
   
 γ xy  0 0 2(1 + ν)  τ xy 
   
2.12.6 Stress Functions

If the body forces are derivable from a potential, V as:
X = – ∂-----
V-
∂x
(Eq. 2-31)
Y = – ∂-----V-
∂y
Then, the stresses and body forces are derivable from a stress function, φ(x,y), as follows:
Nastran Primer 49
2
2
∂ϕ
σ xx – V =
2
∂y
2
σ yy – V = ∂ ϕ (Eq. 2-32)
2
∂x
2
τ xy = – ∂ φ
∂ x ∂y
Substituting (Eq. 2-32) into (Eq. 2-26) then leads to:
4 4 4 2 2
∂ φ + 2∂ φ  
+ ∂ φ = – ( 1 – ν )  ∂ V + ∂ V (Eq. 2-33)
∂x
4 2 2
∂ x ∂y ∂y
4 ∂x 2
∂y 
2
Where the right hand side is zero for the case of no body forces. The solution to two
dimensional is thus reduced to solving for a stress function which satisfies (Eq. 2-30) and an
appropriate set of boundary conditions.
2.12.7 Solutions based on Stress Functions

Using the so-called semi-inverse method several interesting problems can be solved by
noting that second order polynomial functions automatically satisfy (Eq. 2-30). Using this fact
the several problems of interest can be solved. These solutions, taken from Timoshenko[2] are
summarized below.
Bending of a Cantilever Loaded at One End

The cantilever is shown by Figure 2-5
l l
c A
x
c
P
y
Figure 2-5 Cantilever Beam
The stresses are given by:
Pxy
σ xx = – ---------
I
2
P y- (Eq. 2-34)
τ xy = – ----- 1 – ----
2I c
2
σ yy = 0
Where I is the area moment of inertia of the cross section.

It is interesting to note that the axial stress at a given position x varies linearly over the
cross section; and, that the transverse shear stress varies quadratically over the cross section.
The displacements are given by:
50 Nastran Primer
2 3 3 2 2
Py
Px y- – ν --------
u = – ----------- Py - +  --------
- + --------- Pc - y
Pl - – ---------
2EI 6EI 6GI  2EI 2GI
(Eq. 2-35)
2
Py x Px3 Pl 2 x Pl 3
v = ν ------------ + --------
- – ----------- + ---------
2EI 6EI 2EI 3EI
Evaluating v(x,y) at y = 0 gives the deflection curve:
3 2 3
Px - Pl x + --------
v = -------- – ----------- Pl - (Eq. 2-36)
6EI 2EI 3EI
Further, evaluating (Eq. 2-36) at x = 0, the loaded end gives the maximum tip displacement:
3
Pl
v max = --------- (Eq. 2-37)
3EI
Effect of a Circular Hole On Stress Distribution

A rectangular plate with a circular hole is subject to a uniform tension of magnitude S in
the x-direction as shown by Figure 2-6.
m1
2a
m
b
p q x
n θ
n1 S
y
Figure 2-6 Hole in a Rectangular Plate
In terms of radial coordinates, the components of stress are found to be:
S a  S---  1 3a 4a 
2 4 2
σ rr = ---  1 – ----
- +  + -------- – -------- cos 2θ
2 r  2
2
r
4
r 
2
S a - – S---  1 + 3a
2 4
σ θθ = ---  1 + ----   -------- cos 2θ (Eq. 2-38)
2 r  2
2
r 
4
S 4 2
3a + 2a
τ rθ = ---  1 – -------- -------- sin 2 θ
2 r
4
r 
2
It can be seen that σtt is greatest at θ = π/2 or θ = 3π/2. At these points:
( σ θθ ) max = 3S (Eq. 2-39)
Nastran Primer 51
2
Beam Theory
2.13 Beam Theory
2.13.1 Stress Resultants

In beam theory it is convenient to use the force and moment resultants shown by
Figure 2-7 rather than stress components.
z
y
y
B σ xy
σ xx
σ xz z A
x
O
x dx
z
Vz
My
y
Vy
P O x
Mx
x Mz
Figure 2-7 Stress components and Stress Resultants on a Beam Cross Section
The forces and moments stress resultants acting on the cross section are obtained by
integrating the stress components over the cross section as follows: It should be noted that the
surface has a negative normal in the x-direction and that the sign of the bending moment
resultants is chosen so that a positive moment will produce a positive curvature of the elastic
curve associated with the beam reference axis:
P = ∫ σxx dA V y = – ∫ τ xy dA V z = – ∫ τ xz dA (Eq. 2-40)

A A A
Mt = ∫ ( τxz y – τxy z ) dA (Eq. 2-41)

A
M y = – ∫ σ xx z dA M z = – ∫ σ xx y dA (Eq. 2-42)
A A
where A is the area of the cross section.
52 Nastran Primer
Beam Theory
If the beam is statically determinate then the stress resultants can be found using only
equilibrium considerations. Otherwise it is necessary to consider the deformations.
2.13.2 Stresses due to Extension and Bending

Classical beam theory is based on the observation that for a long prismatic bar loaded with
transverse loads the displacement in the axial direction varies in accordance with the Bernoulli-
Euler theory of bending. The displacement at point B in Figure 2-7 is assumed to be:
uB = u ( x ) – θy z – θz y (Eq. 2-43)
where O is a point on a reference axis, and where u(x) is the axial displacement along the x-
axis and θy and θz are the angles by which the plane of the cross section rotates about the y and
z axes, respectively. Using the first of (Eq. 2-3), the extensional strain in the axial direction is
then given by:
du dθ z dθ y
ε xx = – y– z = C1 + C2 y + C2 z (Eq. 2-44)
dx dx dx
The Bernoulli-Euler displacement assumption is equivalent to postulating a linear strain

distribution over the cross section and that the shear stress components are small compared to
the axial stress, σxx. The axial stress is then given by the one dimensional stress strain relation:
σ xx = E ( ε xx – ε 0 ) (Eq. 2-45)
The substitution of the stress strain relation, (Eq. 2-45) and the strain displacement
relation, (Eq. 2-44) into the expressions for stress resultants then gives:
P
--- = AC 1 + yAC 2 + zAC 3
E
M
– ------y- = zAC 1 + I yz C 2 + I yy C 3 (Eq. 2-46)
E
M
– ------y- = yAC 1 + I zz C 2 + I yz C 3
E
where
A = ∫ dA
A
1 1
y = --- ∫ y dA z = --- ∫ z dA (Eq. 2-47)
A A
A A
1 2 1 2 1
I yy = --- ∫ z dA I zz = --- ∫ y dA I yz = --- ∫ yz dA
A A A
A A A
The solution for the coefficients in (Eq. 2-36) is simplified if the y and z axes are chosen
so that y = z = 0 . The x-axis is therefore taken as the axis of centroids of the cross section.
The coefficients are then found to be:
Nastran Primer 53
2
Beam Theory
P-
C 1 = ------
EA
( M z I yy – M y I yz ) ( M y I zz – M z I yz ) (Eq. 2-48)
C 2 = – --1- --------------------------------------
- C 3 = – --1- -------------------------------------
-
E ( I I – I2 ) E ( I I – I2 )
yy zz yz yy zz yz
where Iyy, Izz and Iyz are the area moments of inertia of the cross section. Using these values
the stress becomes
P  M z I yy – M y I yz-  --------------------------------
M y I zz – M z I yz
σ xx = --- –  ---------------------------------  y – - z (Eq. 2-49)
A  I I – I2   I I – I2 
yy zz yz yy zz yz
If the axes coincide with principle axes then the stress is:
P Mz My
σ xx = --- –  ------- y –  ------- z (Eq. 2-50)
A  I zz   I yy 
Comparing (Eq. 2-49) and (Eq. 2-50) it would appear that the use of principal axes results
in a great simplification of the stress. However, for those cases where the principle axes are not
apparent, the calculation of principal axes, principal moments of inertial, and then the
transformation of loads and stress recovery points to the principal axes takes much more time
than the use of (Eq. 2-49)directly. And of course there is always the possibility of making
computational errors in process.
2.13.3 Equilibrium Equations

We assume the beam is subjected to an applied force, {p} and moment, {m}, per unit
length as shown by Figure 2-8
dM z
M z + ---------- ds
ds
dP dM t
P + ------- ds M t + --------- ds
ds mz ds
Vz Pz A′
A′ dV y
Py V y + --------- ds my
ds dM y
Px My mt M y + ---------- ds
O′ O′ ds
Vy dV z
V z + --------- ds
ds Mt
P
Mz
(a) (b)
Figure 2-8
Using standard procedures described by Rivello[3] the equilibrium equations are
dP dV y dV z
= –px = py = pz
dx dx dx
(Eq. 2-51)
dM t dM y dM z
= –mt = Vz + my = Vy – mz
dx dx dx
54 Nastran Primer
Plate theory
where ϕ is the rotation about the x axis.

For a beam loaded by a line load in the z-direction and using principal axes for the
moment of inertia we obtain:
2
d My
= pz
2
dx
(Eq. 2-52)
2
d  EI d w
2
 yy 2  = p z
dx  dx 
2
2.13.4 Solution to Beam Bending Equation

Integrating (Eq. 2-52) for the case where the load is zero results in:
3 2
w ( x ) = a3 x + a2 x + a1 x + a0 (Eq. 2-53)
The coefficients can then be found using specific boundary conditions on displacement
or load. The solution for the cantilever beam which is fixed at the end x = 0 and for which the
boundary conditions are: w(0) = w'(0) = 0; Vz(L) = P; My(L) = 0 is:
P 3 2
w ( x ) = -------------- ( x – 3L x )
6EI yy
The maximum displacement at x = L is
3
PL -
w max = ------------- (Eq. 2-54)
3EI yy
which agrees with the exact solution from two dimensional theory given by (Eq. 2-37).
2.14 Plate theory

Plate theory is obtained from the three-dimensional theory of elasticity for a continuum
which is planar and which has one dimension, t, which small compared to a characteristic
length of the plate as shown by Figure 2-9
z,w
y,v
x,u,β
Nastran Primer 55
2
Plate theory
Figure 2-9
The plate is embedded in a reference coordinate system as shown below such that the x-
y plane coincides with the reference surface of the plate and the z-axis is in the direction of the
thickness.
2.14.1 Kinematic Relationships

The three dimensional plate is replace by a two dimensional reference surface to which
we will ascribe properties associated with the thickness. We denote the displacement at any
point in the plate by (u, v, w) which are defined in terms of displacements and rotations of the
reference surface. We will represent the displacement at any point in terms of the
displacements, (u0, v0, w0)and mean rotations, (α, β) of the reference surface where, following
MacNeal’s notation[4]:
α = –θy β = θx (Eq. 2-55)
where the reference surface displacements and rotations are functions of x and y only. The in-
plane displacements are then assumed to vary linearly over the thickness and so the
displacements at any point in the cross section is represented in terms of reference surface
displacements and rotations as follows:
u = u 0 ( x, y ) – zα ( x, y )
v = v 0 ( x, y ) – zβ ( x, y ) (Eq. 2-56)
w = w0
In addition, the normal stress in the direction normal to the plane of the plate is assumed
equal to the mean surface pressure:
σ zz = p z ( x, y ) (Eq. 2-57)
where pz is the symmetrical part of the transverse load applied to the top and bottom surfaces.
The strains at any point in the cross section can then be expressed in terms of reference
surface displacements and rotations as:
∂u ∂u 0 ∂α 0
ε xx = = –z = ε xx – zχ x
∂x ∂x ∂x
∂v ∂v ∂β 0
ε yy = = 0 – z = ε yy – zχ y (Eq. 2-58)
∂x ∂x ∂x
ε zz = ∂w = 0
∂z
and
∂w – α
γ xz =
∂x
γ yz = ∂w – β (Eq. 2-59)
∂y
γ xy = γ xy – z  ∂α + ∂β = γ xy – zχ xy
0 0
 ∂ y ∂ x
56 Nastran Primer
Plate theory
where the membrane strains are denoted by a superscript and the curvatures of the
reverence surface are:
χ x = ∂α χ y = ∂β χ xy = ∂α + ∂β (Eq. 2-60)
∂x ∂y ∂y ∂x
2.14.2 Kirchhoff Hypothesis

The Kirchhoff hypothesis is that the transverse shear components γxz and γyz are zero. The
curvatures are then related to the transverse displacements of the reference surface as follows:
2 2 2
∂w
χx = χy = ∂ w χ xy = ∂ w (Eq. 2-61)
∂x
2
∂y
2 ∂ x ∂y
Using this assumptions Timoshenko[5] shows that the behavior of the plate is governed by the
following fourth order differential equation:
4 4 4
∂ w + 2∂ w + ∂ w = -----z
p
2 2
(Eq. 2-62)
4 4 D
∂x ∂ x ∂y ∂y
where D for a solid cross section of thickness, t, having an isotropic material is:
3
Et
D = ------------------------- (Eq. 2-63)
2
12 ( 1 – ν )
Solutions to (Eq. 2-62) can be obtained for rectangular plates with various boundary
conditions and loads as described by Timoshenko. For the case of a simply supported boundary
loaded by a sine pressure distribution:
p z = p 0 sin πx
------ sin πy
------
2a 2b
where a and b are the length and width of the plate, respectively. A displacement function of
the form:
πx- sin -----

w ( x, y ) = w 0 sin ----- πy-
2a 2b
satisfies the simple support boundary conditions. The coefficient w0 can then be determined
by substituting the expression for w(x, y) into (Eq. 2-62):
p0
w 0 = ---------------------------------
-
4 1 1-
π D ----- + ----
 2 2
a b
Nastran Primer 57
2
Plate theory
Several solutions obtained by Timoshenko are of particular interest for validating the
behavior of plate-type finite elements.The following table presents the central displacement for
a square plate having a Poisson’s ratio, ν = 0.3, is presented for clamped and simply supported
edges for both a uniform pressure load and a concentrated load at the center.
Table 2-1 Central Deflection of a Square Plate
Uniform Load Concentrated Load

Simple Supports 4 3
0.00416 ( ( p z a ) ⁄ D ) 0.01 ( ( Pa ) ⁄ D )
Clamped 4 3
0.00140 ( ( p z a ) ⁄ D ) 0.00560 ( ( Pa ) ⁄ D )
Early plate-type finite elements based on the Kirchhoff hypothesis were found to give
poor performance. A large literature has been devoted to improving the performance of both
Kirchhoff plate finite elements as well as those based on Mindlin plate theory.
2.14.3 Moment and Force Resultants

In plate and shell theory the stress components are integrated through the thickness to
obtain a set of transverse shear force and moment resultants shown by Figure 2-10.
z Qy
y
my Qx
x
m xy mx
m xy
m xy m xy
mx
my
Qx
Qy
Figure 2-10 Shear and Moment Resultants
The membrane force resultants, { N } ,(not shown), are related to the symmetric part of the
stress which is called the membrane stress and is designated by {σ}m:
   
 Nx  --t-
 σ xx 
   
∫– --t-  σyy
2
{ N } =  Ny  =  dz (Eq. 2-64)
  
 N xy  2
 τ xy 
   m
The transverse shear resultants, {Q}, are:
58 Nastran Primer
Plate theory
  --t-
 
 Qx   τ xz 
∫– --t-  τ
2
{Q} =   = dz (Eq. 2-65)
 Qy  yz 
  2
 
The moment resultants, {M}, are the integrated moment of the bending stress which is the
antisymmetrical part of the stress components:
   
 Mx  t  σ xx 
  2 
---

{ M } =  My  = – ∫ t  σ yy  zdz (Eq. 2-66)
  – ---  
 M xy  2
 τ xy 
   b
2.14.4 Constitutive Relations

The plate cross section is assumed to be represented by n layers of Anisotropic material.
This implies that the integration presented in the previous section is to be replaced by a
piecewise integration over each individual layer having a thickness, ti. Substituting the strain
displacement relation into(Eq. 2-64) and integrating over the thickness then leads to:
{N}=[A]{ε0}+[B]{χ} (Eq. 2-67)
where Jones[6] shows the matrices [A] and [B] are obtained as follows for a layered composite
material:
n ti + 1
[A] = ∑ ∫ [ E m ] i dz (Eq. 2-68)

i=1 t
i
and
ti + 1
n
[B] = ∑ ∫ z [ Eb ] i dz (Eq. 2-69)

i=1 t
i
Similarly, the moment resultant is:
{M}=[D]{c}+[B]{ε0} (Eq. 2-70)
where [D] is:
t
n i+1
∑ ∫
2
[D] = z [ E b ] i dz (Eq. 2-71)
i=1 t
i
In the same manner, the transverse shear, {Q}, is related to the shear strains as follows:
{ Q } = k[ G ]{ γ} (Eq. 2-72)
where the parameter, k, is a transverse shear factor for the cross section and where [G] is:
Nastran Primer 59
2
Problems
n ti + 1
[G] = ∑ ∫ [ E s ] i dz (Eq. 2-73)

i=1 t
i
2.14.5 Equilibrium Equation

The moment and resultants are shown by Figure 2-10. The equilibrium equations for a flat
pate are then found in the usual way to be:
∂M x ∂M xy
+ + Qx + mx = 0
∂x ∂y
∂M y ∂M xy
+ + Qy + my = 0 (Eq. 2-74)
∂y ∂x
∂Q x ∂Q y
+ + pz = 0
∂x ∂y
where mx and my are applied moments.
2.15 Problems
[1] Considering (Eq. 2-14), what are the limiting values of Poisson’s ratio for an isotropic
material
[2] For an isotropic material, verify the matrix of elastic coefficients , [ E ] , given by
(Eq. 2-30)
[3] Given the displacement functions
–3
u = 10 ( x + y ⁄ 2 )
–3
v = 10 ( y + x ⁄ 2 )
w = 0
Determine the components of strain for for the case of plane stress.
[4] Using the components of strain detrmined from problem 3, determine the components of
stress for an isotropic material
[5] Given the following relations for transverse displacment and mean rotations for plate
bending, determine the extensional and shear strains.
–3
10 2 2
u = ----------- ( x + xy + y )
2
∂w –3
θx = = 10 ( x + y ⁄ 2 )
∂y
θ y = – ∂w = – 10 ( y + x ⁄ 2 )
–3
∂x
[6] Given the following relations for displacements, determine the components of strain.
60 Nastran Primer
References
–3
u = 10 ( x + y ⁄ 2 + z ⁄ 2 )
–3
v = 10 ( x ⁄ 2 + y + z ⁄ 2 )
–3
w = 10 ( x ⁄ 2 + y ⁄ 2 + z )
[7] Given the components of strain from Problem 6, determine the components of stress for
an isotropic material
[8] Given the transverse displacement for a uniform beam
Pz 3 2 3
w ( x ) = – -------------- ( x – 3L x + 2L )
6EI yy
Determine the moment, transverse shear and distributed load at any point along the length of
the beam.
2.16 References
[1] M.Turner, R. Clough, H. Martin and L. Topp, “Stiffness and deflection analysis of
complex structures”, J. Aero. Sci., Vol 23, No. 9, Sept. 1956, pp. 805-823
[2] S. Timoshenko and J.N. Goodier, Theory of Elasticity, McGraw-Hill, 1951.
[3] R.M. Rivello, Theory of Flight Structures, McGraw-Hill, 1969, p.156.
[4] R.H. MacNeal, Finite Elements: Their Design and Performance, Marcel Decker, New
York, 1994, p.371.
[5] S. Timoshenko and J.N. Goodier, Theory of Elasticity, McGraw-Hill, 1951, p.82.
[6] R.M. Jones, Mechanics of Composite Materials, McGraw-Hill, New York, 1975, pp. 152-
156.
Nastran Primer 61
2
References
62 Nastran Primer
3 Variation Principles and Approximation
Theory
Early finite elements developed using direct formulations served as a useful bridge to
modern computational analysis. However, truly useful finite element formulations for two and
three dimensional continuum elements are based on variational principles and approximation
theory.
Variational principles are based on work and energy considerations, and the concept of
admissible behavioral functions. Admissible displacements, for example, are those that satisfy
the kinematic constraints, and admissible stresses are those that satisfy the equilibrium
conditions in the interior and the stress boundary conditions on that part of the boundary having
prescribed stresses. These two sets of behavioral variables are related by constitutive relations.
Variational methods allow us to determine the equations associated with the theory of
elasticity and, by employing Lagrange multipliers, yield the Hellinger-Reissner principle and
the principle of minimum complementary energy. We will see that the principle of minimum
potential is associated with the displacement formulation of finite element while the Hellinger-
Reissner and Principle of Minimum Complementary potential are useful for mixed and stress
formulations.
When the problem of elasticity cannot be solved exactly, and that is generally the case,
the use of a variational principle will yield an approximate solutions. However, as we will see
in a simple example, care must be taken when applying the Raleigh-Ritz method since the
stress distribution is not nearly as accurate as the displacement approximation
3.1 Principle of Virtual Work

The principle of virtual work states that if the virtual work is zero for any admissible
displacement then the system is in equilibrium. The virtual work is the sum of the virtual work
of the internal forces plus the virtual work of the external forces. Since the virtual work of the
internal forces is the negative of the variation of the internal energy we can write:
– δU + δW ext = 0 (Eq. 3-1)
where δU is the variation of the internal strain energy and δWext is the virtual work of all
external forces. The variation of the internal strain energy for a virtual strain variation is:
δU = ∫ ( σxx δεxx + σyy δεyy + σzz δε zz + τxy δγxy + τyz δγyz + τzx δγzx ) dΩ (Eq. 3-2)
Ω
Using the components of stress and strain as pseudo vectors we can write:
∫ {σ}
T
δU = { δε } dΩ (Eq. 3-3)
Ω
The virtual work is then:
63Nastran Primer
Variation Principles and Approximation Theory
3
Complementary Virtual Work
– ∫ { σ } { δε } dΩ + ∫ { X } { δu } dΩ + ∫ { t } { δu } dΓ = 0
T T T
(Eq. 3-4)
Ω Ω Γ
The equilibrium equations and the boundary conditions can then be obtained by using the
strain displacement relations and using integration by parts.
3.2 Complementary Virtual Work

The principle of complementary virtual work is obtained by considering a body in
equilibrium. We then consider a variation of stress, {δσ}such that the equilibrium equations
are satisfied. During this process the strains are given by the strain displacement relations so
that the virtual work done by the virtual change of stress is:
∫ ({ε} – [L]{u}) ∫ ({u} – {u})

T T
{ δσ } dΩ + { t }dS = 0 (Eq. 3-5)
Ω Γ2
Where Γ2 is that portion of the boundary on which the displacement is prescribed. Using
integration by parts and noting that the stress components satisfy equilibrium in the interior and
that the variation of the stress vector must vanish on the part of the boundary on which the
stress is prescribed leads to the following equation which is called the principle of
complementary virtual work:
∫ {ε} ∫ {u}
T T
{ δσ }dΩ – { δt } dΓ = 0 (Eq. 3-6)
Ω Γ
2
3.3 Minimum Potential Energy

In stating the principle of virtual work we expressed the virtual change of the work of the
internal forces in terms of the variation of the internal strain energy. In developing the principle
of minimum potential energy we assume the existence of a positive definite strain energy
density function, A, of the following form:
1 T
A = --- { ε } [ E ] { ε } (Eq. 3-7)
2
Taking the variation of A with respect to the strains and using the constitutive relations
for a linearly elastic body then gives:
T
δA = { σ } { δε } (Eq. 3-8)
Assuming the body and surface forces are derivable from potential functions Φ(x,y,z) and
Ψ(x,y,z), respectively we can transform the principle of virtual work to:
– δU + δΦ + δΨ = δΠ = 0 (Eq. 3-9)
where
64 Nastran Primer
Minimum Total Complementary Potential Energy
δU = ∫ δ AdΩ
Ω
∫ {X}
T
δΦ = { δu } dΩ (Eq. 3-10)
Ω
∫ {t}
T
δΨ = { δu } dΓ
Γ
1
and,
∫ ( A – Φ ) dΩ – ∫ { t }
T
Π = { u } dΓ (Eq. 3-11)
Ω Γ1
The principle of minimum total potential energy states:

Among all states of strain which satisfy the kinematic constraints within and on
the boundary of the body, that which makes the total potential an absolute
minimum also satisfies equilibrium.
It is noted that the principle of total strain energy is restricted to linearly elastic bodies
whereas the principle of virtual work is valid for any constitutive relation.
3.4 Minimum Total Complementary Potential Energy

The principle of minimum total complementary energy can be derived from the principle
of virtual complementary work by expressing the equation of state in terms of a stress rather
than a strain as:
1 T
B = --- { σ } [ S ] { σ } (Eq. 3-12)
2
where [S] = [E]-1. The principle of complementary work can then be transformed to:
∫ δB dΩ – ∫ { t }
T
{ δu } dΓ = δΠ c (Eq. 3-13)
Ω Γ2
where
∫ ( A – Φ ) dΩ – ∫ { t }
T
Π = { u } dΓ (Eq. 3-14)
Ω Γ
1
The principle of minimum total complementary energy states:

Among all states of stress satisfying equilibrium, that state which makes the total
complementary potential energy a minimum satisfies the kinematic constraints in
the interior and on the boundary.
Nastran Primer 65
3
Generalized Principle
3.5 Generalized Principle

Modern element formulations use multi-field approximations; i.e., displacements are
assumed on the boundary and either an assumed stress or strain field is assumed in the interior.
Other forms of a minimum principle are thus desirable. A very general form is presented by
Washizu[1], as follows, which is termed a generalized principle.
∫ [A – {X} { u } ] dΩ – ∫ ( { ε } – [ L ] { u } ) { σ } dΩ –
T T
ΠI = (Eq. 3-15)
Ω Ω
∫ {t} ∫ ({u} – {u})

T T
{ u } dΓ – { p } dΓ
Γ Γ
1 2
In this generalized principle {σ} and {p} are Lagrange multipliers, which are to be
determined, which are required to hold the constraints of the strain-displacement relations in
the interior and the imposed displacement on Γ2, respectively. The actual solution is found by
determining the stationary conditions for the functional, ΠΙ. Taking the variation with respect
to the displacement, {u}, the strains, {ε} and the Lagrange multipliers, {σ} and {p} leads to:
∫ [([E]{ε } – {σ} )
T T
δΠ I = { δε } – ( { ε } – [ L ] { u } ) { δσ } (Eq. 3-16)
Ω
T T
∫ [({X} – {X})
T
– ( [ L ] { σ } + { X } ) { δu } ] dΩ + { δu } ] dΓ
Γ1
∫ [({u} – {u}) ∫ [({X} – {p}

T T
– { δp } ] dΓ + ) ] { δu } dΓ +
Γ2 Γ2
with the following stationary conditions:
{ σ } = [ E ] { ε } in Ω (Eq. 3-17)
{ ε } = [ L ] { u } in Ω (Eq. 3-18)
T
[ L ] { σ } + { X } = 0 in Ω (Eq. 3-19)
{ t } = { t } on Γ 1 (Eq. 3-20)
{ u } = { u } on Γ 2 (Eq. 3-21)
{ p } = { t } on Γ 2 (Eq. 3-22)
3.6 Hellinger-Reissner Principle

The Hellinger-Reissner principle is a special case of the generalized principle, (Eq. 3-14),
for which the strains are no longer independent. The coefficients of the variations of strain,
{δε}, in (Eq. 3-16) must vanish so that the strains must be determined from the stress strain
relations. The Hellinger-Reissner principle is then expressed as:
66 Nastran Primer
Approximation Theory
∫ (– [B + {X}
T T
ΠR = { u } ] + { σ } [ L ] { u } ) dΩ – (Eq. 3-23)
Ω
∫ {t} ∫ ({u} – {u})

T T
{ u } dΓ – { p } dΓ
Γ Γ
1 2
where B is:
T
B = {σ} {ε} – A (Eq. 3-24)
Elimination of the strain components using the stress strain relations then gives the
following relation for the variation of B:
T
δB = { ε } { δσ } (Eq. 3-25)
The independent quantities open to variation are the stress components, {σ} and the stress
vector, {p}, and the displacement, {u}. Taking variation with respect to these quantities leads
to the following stationary conditions
[ L ] { u } = [ E ] { σ } in Ω (Eq. 3-26)
together with (Eq. 3-19) through (Eq. 3-22).

Another form of the functional (Eq. 3-23) obtained by using integration by parts is:
T
∫ [B + ([L] ∫ ({t} – {t}) ∫ {t}
* T T T
–ΠR = { σ } ) { u } ] dΩ – { u } dΓ – { u } dΓ (Eq. 3-27)
Ω Γ Γ
1 2
where (Eq. 3-22) has been used to eliminate {p}. The independent quantities open to variation
are then the components of stress, {σ}, and the displacements, {u}.
3.7 Approximation Theory

Approximation theory lets us choose a reasonable approximation to the solution defined
in terms of a one or more generalized coordinates and then employ one of the variation
principles to determine the generalized coordinate. The resulting approximation will then
satisfy the subsidiary relations in some mean square sense.
In the principle of virtual work the displacements must satisfy the strain displacement
relations in Ω and the displacement boundary conditions on Γ2, that portion of the boundary
on which displacements are prescribed. We take the displacements to be:
u ( x, y, z ) = u 0 ( x, y, z ) + ∑ a n u n ( x, y, z )
n
N
v ( x, y, z ) = v 0 ( x, y, z ) + ∑ b n v n ( x, y, z ) (Eq. 3-28)
n
N
w ( x, y, z ) = w 0 ( x, y, z ) + ∑ c n w n ( x, y, z )
n
Nastran Primer 67
3
Rayleigh-Ritz Soution of a Beam
where u0, v0 and w0 are chosen such that they satisfy the enforced displacements on Γ2; and
where un, vn, and wn are chosen to vanish on Γ2 for all indices. The displacements are now
functions of the undermined parameters so the variation of w, for example, is:
δw ( x, y, z ) = w 0 ( x, y, z ) + ∑ δc n w n ( x, y, z )
n
After substituting the assumed displacements and into the strain displacement relations
the virtual work is seen to be a function of the variations of the independent coefficients. Since
the variations of an, bn and cn are arbitrary it follows that the coefficients of each variation must
vanish independently resulting in 3N equations for the 3N independent coefficients.
3.8 Rayleigh-Ritz Soution of a Beam

As an example, consider the case of a simply supported beam of length, L, a cross section
area moment of inertia, I, and modulus of elasticity, E, with a concentrated load at the center.
We consider an approximate solution of the form:
w ( x ) = c 1 sin πx
------ (Eq. 3-29)
L
which satisfies the displacement boundary conditions at x = 0, L, where w(x) is the transverse
displacement in the z-direction.
It can be shown that the principle of virtual work for the beam bending in the x-z plane is:
– ∫ w ,xx EIδw ,xx dx + Pδw  --- = 0

L
(Eq. 3-30)
2
0
where the second term accounts for the virtual work of the applied load.
Taking the variation of the assumed displacement function,(Eq. 3-29) and substituting
into (Eq. 3-30)then gives:
 L 4 
πx 2
δc 1  – ∫ -------------4- c 1  sin ------ dx + P = 0
EIπ
(Eq. 3-31)
2L
 0 ( 2L ) 
Since the variation of the parameter, δc 1 , is arbitrary, the coefficient must be equal to
zero. Evaluating of the integrals and solving for c 1 then gives:
3
c 1 = 2PL
-------------
4
π EI
(Eq. 3-32)
3
2PL - πx
w = ------------
4
sin ------
π EI L
The difference between the approximate and exact displacement at the center of the beam
is less than 3%. However, when the moment is evaluated using the moment curvature
relationship the error is found to be as great as 25%.
68 Nastran Primer
References
This example shows the general result found for assumed approximate displacements:
The displacements are reasonably accurate but the force quantities obtained by differentiating
the displacement function may vary widely from the exact solution. And, of course it is
generally the stress type behavior that is used as a failure criterion in design.
In the next Chapter we use approximation theory to obtain the equilibrium equations in a
finite region of the continuum. The result will be the equations for the finite element method.
3.9 References
[1] K. Washizu, Variational Methods in Elasticity and Plasticity, Second Edition, Pergamon
Press, Oxford, 1974, p.32.
Nastran Primer 69
3
References
70 Nastran Primer
4 Finite Element Formulations
Successful finite element formulations for two and two dimensional elements are based
on variational principles. The use of the variational principle allows us to satisfy equilibrium
in some average sense and is termed a weak formulation.
The objectives of this chapter are to:
• Develop a procedure for determining the equilibrium equations for finite elements by
considering the axial rod and beam bending elements
• Introduce basis functions for interpolating behavioral variables in a region
• Introduce the concept of mapping from metric space to parameter space
• Develop the finite element equations for two dimensional elasticity for a displacement
and mixed formulations
• Develop the finite element equations for the dynamic response of a quadrilateral shell
element
4.1 Finite Element Formulation

In the formulation of the approximate equilibrium equations for an element the
displacement field within the element is approximated in a Rayleigh-Ritz sense. The number
of generalized coordinates in the approximation must be the same number as the number of
node point displacements for the element so that the generalizes coordinates can be related to
the node point degrees of freedom.
4.2 Rod Element

In order to describe the procedure let us consider the uniform rod shown by Figure 4-1 .
u2,F2 u1,F1 x
1 2
71 Nastran Primer
Finite Element Formulations
4
Rod Element
Figure 4-1 Axial Rod Element
The axial displacements and forces at the node points are:
T
{u} = u1 u2
(Eq. 4-1)
T
{F} = F1 F2
The only non zero stress component is σ xx so the principle of virtual work becomes:
δW = – ∫ { σ xx } { δε xx } dΩ + { F } { δu } = 0
T T
(Eq. 4-2)
Ω
The axial displacement is assumed to vary linearly along the axis of the rod so that the
following polynomial is assumed to represent the displacement
u = a + bx (Eq. 4-3)
The coefficients a and b are now related to the node point displacements by noting that
u ( 0 ) = u1
(Eq. 4-4)
u ( L ) = u2
Solving for a and b gives
a = u1
u2 – u1 (Eq. 4-5)
b = ----------------
-
L
The assumed displacement function, written in terms of the node point displacements as
generalized coordinates, is
u ( x ) =  1 – --x- u 1 +  --- u 2 = N 1 u 1 + N 2 u 2
x
 L  L
(Eq. 4-6)
N 1 =  1 – --x- N 2 =  ---
x
 L  L
or
u( x) = N {u} (Eq. 4-7)
The strain is then
 
ε xx =
d u( x ) dN 1 dN 2  u 1 
=   (Eq. 4-8)
dx d x d x  u2 
 
The node point displacements are to be determined such that the principle of virtual work
is satisfied. Taking the variation of the displacement gives
72 Nastran Primer
Steps in Determining the Stiffness Equation
 
 δu 1 
δu ( x ) = N 1 δu 1 + N 2 δu 2 = N1 N2   (Eq. 4-9)
 δu 2 
 
Using Hooke’s Law, the principle of virtual work becomes:
T
– ∫ {u}
T dN 1 dN 2 T
E dN 1 dN 2 { δu } dΩ + { F } { δu } = 0 (Eq. 4-10)
Ω dx dx dx dx
Since the variations, { δu } , are arbitrary it follows that
[k]{u} = {F} (Eq. 4-11)
where [k] is
∫
[k] = dN 1 dN 2 E dN 1 dN 2 dΩ
Ω dx dx dx dx
(Eq. 4-12)
AE
[ k ] = ------- 1 – 1
L –1 1
The equilibrium equation for a rod element is then
   
AE 1 – 1  u1   F1 
L – 1 1  u  =  
------- (Eq. 4-13)
2   F2 
   
It is important to note that the stiffness equation for the element is singular since the
displacement field includes rigid body motion. As a consequence, the equilibrium equations
cannot be solved until sufficient constraints are applied to remove the rigid body motion.
Setting the displacement u 1 = 0 then gives
F2 L
u 2 = --------- (Eq. 4-14)
AE
which is the well known expression for the displacement of the end of an axial rod subject to
and end force.
4.3 Steps in Determining the Stiffness Equation

The procedure for determining the stiffness equation for any finite element is essentially
the same as that employed for determining the stiffness matrix for the axial rod element:
Nastran Primer 73
4
Stiffness Matrix for a Beam Bending Element
1.Represent the displacement field in terms of a set of basis functions where the
number of generalized parameters must be the same as the number of node point
degrees of freedom. The displacement function must include all rigid body modes
for the element. In the case of the rod, there is only one rigid body mode: translation
of the rod in the axial direction. However, in the case of a three dimensional
element, the displacement field must include three rigid body translations and three
rigid body rotations.
2.Relate the generalized parameters in the displacement approximation to the node
point displacements
3.Determine the strains using the strain-displacement relations
4.Use Hooke’s law to express the stress in terms of node point displacements
5.Substitute the expressions for stress and virtual strains into the principle of virtual
work
6.Evaluate the stiffness matrix.
4.4 Stiffness Matrix for a Beam Bending Element

We now apply this procedure to obtain the equilibrium equations for a beam subject to a
distributed line load as shown by Figure 4-2 .
p(x)
v1,P1 v2,P2
2 x
1
θ1,M1 θ2,M2
b
L
z
Figure 4-2 Beam Bending in the x-y Plane
The behavior of the beam is represented by the displacement, v , and the rotation, θ , at
each end of the beam as shown in the figure. The set displacements and associated force and
moment resultants at the node points is represented by
74 Nastran Primer
   
 v1   P1 
   
 θ1   M1 
{u} =   {Q} =   (Eq. 4-15)
 v2   P2 
   
 θ2   M2 
   
In the displacement approach we now represent the displacement using a cubic

polynomial which has four generalized parameters as
2 3
v ( x ) = a + bξ + cξ + dξ (Eq. 4-16)
where ξ = x ⁄ L . We note that the displacement function includes rigid body translation and
rotation.
In order to determine the generalized parameters in terms of the node point displacements
we need the derivative:
d v(x ) b c d 2
= --- + 2 --- ξ + 3 --- ξ (Eq. 4-17)
dx L L L
The generalized parameters can now be determined in terms of the node point
displacements from the following relations
v ( 0 ) = v1 v' ( 0 ) = θ 1 v ( 1 ) = v2 v' ( 1 ) = θ 2 (Eq. 4-18)
After determining the generalized parameters the displacement can be expressed as:
v ( x ) = N1 v1 + N2 θ1 + N3 v2 + N4 θ2
(Eq. 4-19)
v(x) = [N ]{ u}
where
2 3
N 1 = 1 – 3ξ + 2ξ
2
N2 = x ( 1 – ξ )
(Eq. 4-20)
2 3
N 3 = 3ξ – 2ξ
2
N4 = x ( ξ – ξ )
The functions, [ N ] are called shape functions. It is easy to verify that N 1 and N 3 are
equal to unity when evaluated at the position of the node point displacement and equal to zero
otherwise; and, that N 2 and N 4 have the same property when evaluating the derivative of the
displacement function.
The virtual work expression for the beam, (Eq. 3-30) in Chapter 1, modified to account
for bending in the x-y plane and the virtual work of the applied line load and concentrated node
point forces is:
L L
– ∫ v ,xx EIδv ,xx dx + ∫ p ( x )δv dx + { Q } { δu } = 0

T
(Eq. 4-21)
0 0
Nastran Primer 75
4
Using (Eq. 4-19) to represent the displacement and taking the indicated derivatives then
leads to the following virtual work expression:
L 2 T 2 L
dN dN
–∫ { u } { δu } dx + ∫ p ( x ) [ N ] { δu } dx + { Q } { δu } = 0
T T
2
EI zz 2
(Eq. 4-22)
dx dx
0 0
Since the virtual displacement, { δu } , is arbitrary it follows that the equilibrium equation
for the beam is
[ k ] { u } = { P } = { P we } + { Q } (Eq. 4-23)
where { P we } are called work equivalent forces since they perform the same amount of virtual
work as the distributed loads and the stiffness matrix for the beam is
L 2 T 2
dN dN
[k] = ∫ dx
2
EI zz
dx
2
dx (Eq. 4-24)
0
Noting that [ N ] = N1 N2 N3 N4 and that the shape functions are given by

(Eq. 4-20) then leads to
2
( N'' 1 ) ( N'' 1 ) ( N'' 2 ) ( N'' 1 ) ( N'' 3 ) ( N'' 1 ) ( N'' 4 )
L 2
( N'' 2 ) ( N'' 1 ) ( N'' 2 ) ( N'' 2 ) ( N'' 3 ) ( N'' 2 ) ( N'' 4 )
[ k ] = EI zz L ∫ dx (Eq. 4-25)
2
0 ( N'' 3 ) ( N'' 1 ) ( N'' 3 ) ( N'' 2 ) ( N'' 3 ) ( N'' 3 ) ( N'' 4 )
2
( N'' 4 ) ( N'' 1 ) ( N'' 4 ) ( N'' 2 ) ( N'' 4 ) ( N'' 3 ) ( N'' 4 )
where
2
d Ni
N'' i ≡ 2
i = 1, 2, 3 ,4
dx
Evaluating the derivatives and integrating then leads to the stiffness matrix for the beam
12 6L – 12 6L
EI zz 6L 4L 2 – 6L 2L 2
[ k ] = ---------
3
- (Eq. 4-26)
L – 12 – 6L 12 – 6L
2 2
6L 2L – 6L 4L
The determination of the work equivalent loads is left as an exercise. However, it should
be noted that the use of work equivalent loads leads to the exact solution of the beam equations
for the case of a distributed line load. The use of ad hoc load lumping does not lead to the exact
result.
76 Nastran Primer
Basis Functions for Finite Elements
4.5 Basis Functions for Finite Elements

The one dimensional elements considered in the previous sections are not finite elements
in the true sense of the word since the results are exact within the restrictions of one
dimensional theory of elasticity and beam theory. In this section we consider the formulation
of elements which represent two and three dimensional states of stress and which lead to an
approximate representation of the behavior except for states of uniform stress.
There are two element formulations called the h- and p-element formulations,
respectively. In the h-formulation the behavior within a finite volume is represented using
basis functions that are polynomial functions such that we can express a displacement
component, u, for example as:
u = ∑ Ni ui (Eq. 4-27)
i=1
where the functions, Ni, are the basis functions and ui are the displacements at the node points.
As we found for the cases of the rod and beam elements, the Ni have the property that they are
equal to 1 when evaluated at the node point associated with ui and are equal to zero at all other
node points.
Suitable basis functions for a two dimensional rectangular membrane shown by
Figure 4-3, referred to the coordinates of metric space are:
N 1 = ---  1 – ---  1 – ---

1 x y
4 a  b
N 2 = ---  1 + ---  1 – ---

1 x y
4 a b
(Eq. 4-28)
N 3 = ---  1 + ---  1 + ---
1 x y
4 a  b
N 4 = ---  1 – --x-  1 + --y-

1
4 a  b
2a
y
2b x
Figure 4-3 Rectangular Membrane Element
A similar expression for the displacement, v, can be written using the same basis functions
so the vector of displacement components can be written as:
Nastran Primer 77
4
Basis Functions for Finite Elements
 u1 
 
 v1 
 
 u2 
 
 u  N1 0 N2 0 N3 0 N4 0  v2 
{u} =   =   = [ Ne ] { ue } (Eq. 4-29)
 v  0 N1 0 N2 0 N3 0 N4  u3 
 
 v3 
 
 u4 
 
 v4 
Basis functions for other geometric regions can also be determined using polynomial
functions as will be shown later. At this point, let us proceed to determine the equation of
motion using the principle of virtual work.
The virtual work expression is:
∫ {σ} { δε } dΩ – ∫ ( { X } – ρ { u·· } ) { δu }t dA – { P Ce } { δu e } = 0
T T T
(Eq. 4-30)
Ω A
where ρ is the mass density and the body forces, { X } , are augmented with D’Alembert forces
representing the inertia forces associated with dynamic motion of the element.
The stresses and strains are represented in terms of the shape functions by substituting
(Eq. 4-29) into (Eq. 2-2) in Chapter 2 and ((Eq. 2-11) in Chapter 2, respectively
{ ε } = [ L ] { u } = [ L ] [ N e ] { δu e } = [ B ] { δu e }
(Eq. 4-31)
{ σ } = [ E ] { ε } = [ E ] [ L ] [ Ne ] { ue } = [ E ] [ B ] { ue }
Substituting (Eq. 4-31) into (Eq. 4-30) then gives:
∫ [N] ∫ [B]
T T T T
{ u·· e } ρ [ N ] { δu e }t dA + { u e } [ E ] [ B ] { δu e }t dA
A A
– ∫ { X } [ N e ] { δu e }t dA – { P Ce } { δu e } = 0
T T
(Eq. 4-32)
A
or
T T T T
( { u·· e } [ m ee ] + { u e } [ k ee ] – { P Xe } – { P Ce } ) { δu e } = 0
Where:
∫ [N]
T
[ m ee ] = ρ [ N ]t dA
A
∫ [B]
T
[ k ee ] = [ E ] [ B ]t dA (Eq. 4-33)
A
∫ {X}
T T
{ P Xe } = [ N e ]t dA
A
where [ m ee ] is thhe consistent mass matrix for the element and { P Xe } are the work equivalent
forces associated with the distributed loads.
78 Nastran Primer
Isoparametric Transformation
The discrete equilibrium expression for the element is determined from (Eq. 4-32) by
noting that, since the variation of the generalized coordinates is arbitrary, the coefficient matrix
must be zero, so that:
[ m ee ] { u·· e } + [ k ee ] { u e } – { P Xe } – { P Ce } = 0
(Eq. 4-34)
[ m ee ] { u·· e } + [ k ee ] { u e } = { P Xe } + { P Ce } = { P e }
For a rectangular shaped element the integration is straight forward but tedious. The
resulting element performs well as long as the elements are indeed rectangular. But accuracy
drops off significantly for even small deviations from rectangular shape. For general modeling
a more geometrically robust formulation is required where the element shape is a general
quadrilateral.
4.6 Isoparametric Transformation

The basis functions described in the previous section are appropriate for a rectangular or
square region. We therefore define a mapping from metric space to a parametric space that is
bi-square as shown by Figure 4-4:
(-1, 1) (1, 1)
4 3
4 η
y 3
ξ
x
1 2 1 2
(-1, -1) (1, -1)
Figure 4-4 Mapping from Metric to Parameter Space
Since the variational formulation is a scalar equation it applies to any coordinate system
so that the stiffness matrix is still given by the second of (Eq. 4-32) but where the integration
is to be taken with respect to the parametric coordinates. The stiffness matrix is then given by:
∫ [B]
T
[ k ee ] = [ E ] [ B ]t dA (Eq. 4-35)
A′
where A' is the area in parametric space. The differential area in metric space is related to the
differential area in parametric space by:
dA = Jdξdη (Eq. 4-36)
where J is the determinant of the Jacobian of transformation which will be described below.
The stiffness matrix in parameter space is then given by:
1 1
∫ ∫ [B]
T
[ k ee ] = [ E ] [ B ]tJ dξ dη (Eq. 4-37)
–1 –1
Nastran Primer 79
4
4.6.1 Basis Functions and the Jacobian of Transformation

The basis functions in parameter space are:
1
N 1 = --- ( 1 – ξ ) ( 1 – η )
4
1
N 2 = --- ( 1 + ξ ) ( 1 – η )
4
(Eq. 4-38)
1
N 3 = --- ( 1 + ξ ) ( 1 + η )
4
1
N 4 = --- ( 1 – ξ ) ( 1 + η )
4
As we showed in the previous section, the strains are then determined using(Eq. 4-31).
However, since we are integrating with respect to the parametric coordinates and the
differential operators in the matrix, [L], are with respect to metric space. We therefore need to
determine the relation between the differential operators in metric and parametric space.
Assuming three dimensional metric and parametric spaces we have, using the chain rule:
 ∂  ∂
 
 ∂x  ξ ,x η ,x ζ ,x ∂ ξ
 ∂  ∂
  = ξ ,y η ,y ζ ,y (Eq. 4-39)
 ∂y  ∂η
 ∂  ξ ,z η ,z ζ ,z ∂
 
 ∂z  ∂ζ
where a comma indicates partial differentiation with respect to the parameter following the
comma as indicated by (Eq. 4-41).
In order to evaluate the partial derivatives of metric coordinates with respect to the
parametric coordinates we need a transformation between metric and parameter space. Using
the same basis functions used to represent displacements leads to an isoparametric
transformation. For three dimensions in both metric and parametric space we have:
x = ∑ Ni xi
i=1
N
y = ∑ Ni yi (Eq. 4-40)
i=1
N
z = ∑ Ni zi
i=1
where xi, yi, zi are the coordinates of the nodes. The partial of x with respect to the parametric
coordinates then gives:
N
∂x ∂N i
x ,ξ =
∂ξ
= ∑ ∂ ξ xi (Eq. 4-41)
i=1
80 Nastran Primer
which are the reciprocal of the partial derivatives required for the transformation of the partial
derivatives with respect to metric space. We therefore express the partial with respect to
parameter space and invert:
 ∂   ∂   ∂ 
     
 ∂ξ  x ,ξ y ,ξ z ξ  ∂ x   ∂x 
 ∂   ∂   ∂ 
  = x ,η y ,η z ,η   = [J]  (Eq. 4-42)
 ∂η   ∂ y   ∂y 
 ∂  x ,ζ ,ζ ,ζ  ∂ 
y z  ∂ 
     
 ∂ζ   ∂x   ∂x 
where [J] is the Jacobian of the transformation. The inverse relationship is:
 ∂  –1  ∂ 
∂{x}  = [J] ∂{ξ}  (Eq. 4-43)
   
Comparing (Eq. 4-39) and (Eq. 4-42) shows that:
ξ ,x η ,x ζ ,x
–1
[J] = ξ ,y η ,y ζ ,y (Eq. 4-44)
ξ ,z η ,z ζ ,z
The inverse of the Jacobian for two dimensional space is:
1 y – y ,η
= --- ,ξ
–1
[J]
J –x x (Eq. 4-45)
,ξ ,η
J = det [ J ] = x ,ξ y ,η – y ,ξ x ,η
4.6.2 Example - Calculating the Area of a Square

It may seem trivial but it is instructive to calculate the area of a unit square where:
T
{ x e } = x 1 x 2 x 3 x 4 = – 1--- 1--- 1--- – 1---
2 2 2 2
T
{ y e } = y 1 y 2 y 3 y 4 = – 1--- – 1--- 1--- 1---
2 2 2 2
Then
x = N1 x1 + N2 x2 + N3 x3 + N4 x4
y = N1 y1 + N2 y2 + N3 y3 + N4 y4
The partial derivatives are:
Nastran Primer 81
4
Numerical Integration
∂x = – 1--- ( 1 – η )x + 1--- ( 1 – η )x + 1--- ( 1 + η )x – 1--- ( 1 + η )x

∂ξ 4 1 4 2 4 3
4 4
∂x 1 1
= – --- ( 1 – ξ )x 1 – 1--- ( 1 + ξ )x 2 + --- ( 1 + ξ )x 3 + --- ( 1 – ξ )x 4
1
∂η 4 4 4 4
∂y 1 1
= – --- ( 1 – η )y 1 + --- ( 1 – η )y 2 + --- ( 1 + η )y 3 – 1--- ( 1 + η )y 4
1
∂ξ 4 4 4 4
∂y 1 1
= – --- ( 1 – ξ )y 1 – 1--- ( 1 + ξ )y 2 + --- ( 1 + ξ )y 3 + --- ( 1 – ξ )y 4
1
∂η 4 4 4 4
Substituting the positions of the vertices in metric space gives, J = 1/4. Integrating over the
parametric coordinates the gives A = 1. You will note the Jacobian is a constant for a
rectangular element. However, J is generally a function of the parametric coordinates making
the strains an irrational function of the parametric criminates.
4.7 Numerical Integration

Now that we have transformed [B] to parametric coordinates the stiffness and other
element matrices can be determined for arbitrary quadrilateral geometry, provided metric
space transforms to parameter space. As we noted in the previous section, the Jacobian will be
a function of the parametric coordinates for the general case so that it will be impractical to
integrate in closed form. Instead we use Gauss integration since it requires fewer function
evaluations that other numerical integration procedures such as Euler or trapezoidal
integration. Gauss integration evaluates an integral by summing the values of the function
determined at specific points, called Gauss points, that are multiplied by weighting factors.
Using Gauss integration the evaluation of the stiffness matrix is:
[ k ee ] = ∑ f ( ξg, ηg )wg (Eq. 4-46)

g=1
where G is the number of Gauss points, ( ξ g, η g ) , is the location of the Gauss point and w g is
the weighting factor.
The accuracy of the numerical integration is a function of the number of Gauss points.
Since both the interpolation point and weighting factor are known, a linear polynomial can be
2
integrated exactly using a single Gauss point so the error is O ( ξ ) . Similarly two point Gauss
4
can integrate a cubic exactly so the order of error is O ( ξ ) . The following table shows the
location of the Gauss points, the weighting factors and order of error for Gauss integration
through Gauss integration of order 3.
Table 4-1 One Dimensional Gauss Integration
Integration Order ξg wg Order of Error

1 0 2 2
O(ξ )
2 1
(± 3) O(ξ )
4
3
( ± 0.6, 0 )  5---, 8--- O(ξ )
6
 9 9
82 Nastran Primer
Plate and Shell Elements
4.7.1 Integration for the Membrane Problem

The integration of the stiffness and other system matrices requires that we know the
number of Gauss points and their location. The number of Gauss points is determined by the
polynomial order of the integrand. For the stiffness matrix the polynomial order is:
2
Pk = ( PB ) PE PJ (Eq. 4-47)
The typical term in [B] is
P B = X ,ξ ξ ,x
(Eq. 4-48)
ξ ,x = y ,η ⁄ P J
The polynomial order of [B] is therefore
P B = X ,ξ y ,η ⁄ P J (Eq. 4-49)
And, the polynomial order of the element stiffness matrix [kee] is:
2
P k = ( X ,ξ y ,η ⁄ P J ) P E P J (Eq. 4-50)
For the case when the J is a constant P k = 4 , where we take the elastic matrix to be
constant in the element. This will require a two by two Gauss integration where the Gauss
points and weighting factors are shown by Figure 4-5.
1-
------
3
w2
Figure 4-5 Two by Two Gauss Integration
It is interesting to note that the integration of the mass matrix requires 3x3 Gauss
integration for the general case; and, that the work equivalent forces required 2x2 Gauss
integration.
4.8 Plate and Shell Elements

The field equations for a flat plate were developed in Plate theory for both Mindlin and
Kirchhoff theories. It was noted in that section that plate elements based on Kirchhoff
hypothesis did not perform well in practice. Later studies revealed that the problem was related
to the need to represent the second derivative terms such as w,xx at the nodes. To do so the
displacement functions must be complete through quadratic terms. This is a much more
difficult requirement that requiring only continuous displacements for elements based on two
and three dimensional theory of elasticity.
Robust plate bending elements are based on Mindlin theory which retains α and β as well
w to represent the displacements u and v.
Nastran Primer 83
4
4.8.1 Curvature Approximations

The shape functions, (Eq. 4-28) are used to represent the rotations { θ } , so that the
curvatures, (Eq. 2-55) in Chapter 2 are:
∂
  0
 χx  ∂x
  ∂  α 
 χy  = 0  
  ∂ y  β 
 χ xy  ∂ ∂
  (Eq. 4-51)
∂y ∂x
 
 α  0 1  θx 
  =   = [T ]{ θ }
 β  –1 0  θy 
 
Representing the rotations in terms of the shape functions than gives:
{χ} = [L]([T][N]){θ} = [B]{θ}

(Eq. 4-52)
[B] = [L]([T][N] )
4.8.2 Transverse Shear Strain Approximation

The transverse shear strains are:
∂u ∂w b
γ xz = + = w ,x – α
∂z ∂x
(Eq. 4-53)
= ∂v + ∂w = w ,y – β
b
γ yz
∂z ∂y
so that, if the same order interpolents are used for both w and the rotations, the rotations will
contain higher ordered terms than those present in w,x and w,y. This leads to very large shear
stiffness and to an undesirable phenomenon called transverse shear locking.
In order to eliminate transverse shear locking MacNeal[4]used a multi-field approach
where the transverse shears were represented using the following approximation:
 
 a1 
   
 γ xz  1 y 0 0  a2 
  =   (Eq. 4-54)
 γ yz  0 01 x  a3 
   
 a4 
 
The coefficients, {a}, are evaluated by evaluating the transverse shears in parametric
coordinates at the center of the edges of the element. The shear components in parametric
coordinates are related to those in metric coordinates by:
γ sz = γ xz cos δ e + γ yz sin δ e (Eq. 4-55)
84 Nastran Primer
The shear strains are then evaluated at the Gauss points. The performance of this and other
elements in NASTRAN and those presented in the literature is discussed and compared in
Chapter 7.
4.8.3 Curved Shell Elements

There is an extensive literature on shell theory and curved shell elements [5]-[6]. Shell
equations are based on modifying the field equations for three dimensional theory of elasticity
for structures which are both thin and one or more principle radii of curvature. Shells may be
of any geometric shape, such as the Opera House in Sydney Australia. However a large number
of shells using in industry are shells of revolution for which the revolved line segment can be
a general shape. If the revolved line segment is a straight line the shell has the shape of either
a cone or a cylinder. If the revolved shape is a semicircle then the shell is a sphere.
Curved shell elements are complicated by the fact that the surface of the element is curved
rather than flat. The curvature terms in shell theory couple the in-plane and bending behavior,
much the same way that the material [B] matrix couples membrane and bending in flat
elements which include both in-plane and bending behavior. MSC.Nastran includes the Tria6
and Quad8 shell elements whose formulations are based on shell theory. In order to represent
the curvature of the shell these elements include nodes on the edges of the shell element and
are termed higher order elements. The shell equations for these elements is beyond the scope
of the primer and the interested reader is referred to References [5]-[6].
4.8.3.1 Modeling Shells with Flat Elements

Shells can be represented to very good accuracy by an assemblage of flat elements as long
as the elements include both in-plane and bending behavior. This is a two dimensional analogy
to representing a curved arch with straight beam sections. Let’s see what is happening by
considering two such elements connecting a common node as shown byFigure 4-1.
m1
N1 θ
Q1
l1
N 1 = N 2 cos θ N2
Q 1 = – N 2 cos θ
m 1 = – N 2 l 1 cos θ
Figure 4-1 Connecting Flat Shell Elements
The element stiffness matrices are generated in element coordinates. After generating the
stiffness it is transformed from element to the Basic coordinate system at each connected node
and the stiffness coefficients are associated with the g-set degrees of freedom at the node
points. As we show later in Transformation of Element Stiffness Matrices, the element
stiffness is given by:
e T
[ k gg ] = [ T eg ] [ k ee ] [ T eg ] (Eq. 4-1)
Nastran Primer 85
4
where [ T eg ] is the set of transformation matrices for each node point. The effect of the
transformation is to couple the in-plane and bending modes due to the relative angles between
the surfaces of the elements.
The Quad4 element in MSC.Nastran includes both in-plane and bending modes and is
widely used to model shells in model modeling situations. As we will see in Chapter 7, the flat
Quad4 element generally gives results which are as accurate as those given by the higher order
element which include curvature.
4.8.4 Normal Rotations

In plate and shell theory the rotation about the normal to the shell reference surface
vanishes. For flat shell elements this means that there are only five degrees of freedom at a grid
point that are elastically coupled by the stiffness matrix. As we note in Chapter 6, this leads to
problems in the solution phase since a matrix having a row or columns of zeros cannot be
solved. We mention this here since the transformation, (Eq. 4-1) will result in a stiffness about
the normal which will, of course, become vanishingly small as the angle between element
surfaces becomes very small. NASTRAN incorporates a processor controlled by the
AUTOSPC Parameter that detects small stiffness and corrects the numerical problem
automatically.
However, there is a test problem called the Raasch challenge, Knight [7], that shows the
inadequacy of the Autospc processor for certain geometries. This problem shows that a
Autospc is not appropriate to remove the singularities in the problem and that another
technique is required. The technique is to generate the set of linear constraint equations at each
node point that will set the stiffness for the normal rotation to zero. MSC has included the
modifications at the element level which remove the difficulty identified by Raasch.
4.8.5 Non-Planar Nodes for a Flat Shell Element

When using the flat shell element to model curved surfaces the node points might not lie
on a flat plane. In this case the element experiences unbalanced forces and moments. The
stiffness calculated for the flat surface must be transformed to the actual nodal geometry.
MacNeal [8] shows the Transformation is of the form:
T
[ k ] nodes = [ R ] [ k ] plane [ R ] (Eq. 4-2)
T
where [ R ] is transformation matrix which relates forces at the nodes to forces in the
plane:
T
{ F } nodes = [ R ] { F } plane (Eq. 4-3)
86 Nastran Primer
The desires transformation for the forces is obtained by considering the geometry
associated with non-planar nodes as shown by Figure 4-2
3
h
F 21
F 41 z
l 14 1 h
h
h 2
4
F 12 F 14
Figure 4-2 Non-Planar Nodes for Flat Shell Element
The corrective force at node 1 and 4 for example is a vertical force
h
∆F 1z = – ∆F 4z = ------ ( F 14 – F 41 ) (Eq. 4-4)
l 14
In addition to the force correction MacNeal [8] also shows a moment correction is
required as shown by Figure 4-3
Mz
B α
M1
M2
Figure 4-3 Moment due to Twisted Element
The moment M1 must be equilibrated by a force directed along the node line:
sin α
F 1 = – F 2 = ----------- M 1 (Eq. 4-5)
l 12
Using these modifications the matrix can be constructed and used to transform the
stiffness matrix. It is interesting to note that the result of not putting the element into proper
equilibrium can result in results that are orders of magnitude too large.
Nastran Primer 87
4
References
4.9 References
[1] T.H.H. Pian, “Derivation of element stiffness matrices by assumed stress distribution”,
AIAA J., 2, 1333-1336, 1964
[2] T. Pian and S.W. Lee, “Improvement of plate shell finite elements by mixed formulation”,
AIAA J. 16(1), 29-34, 1978
[3] T.H.H. Pian and K. Sumihara, “Rational approach for assumed stress finite elements”,
Int. J. Numer. Methods Engrg., 20(9), 1547-1569, 1984
[4] R.H. MacNeal, “Derivation of element stiffness matrices by assumed strain
distributions”, Nuc. Engr. Design, 70, 3-12, 1982.
[5] R.H. MacNeal, “Specifications for the QUAD8 quadrilateral curved shell element’,
MacNeal-Schwendler Corp. Memo RHM-46B, 1980
[6] R.H. MacNeal, Finite Elements: Their Design and Performance, Marcel Dekker, New
York, p 419-481, 1994.
[7] N.F. Knight, “Raasch challenge for shell elements”, AIAA J. 35(2), 375-381, 1997
[8] R.H. MacNeal, Finite Elements: Their Design and Performance, Marcel Dekker, New
York, p 438-440, 1994.
88 Nastran Primer
5 Structural Elements in NASTRAN
5.1 Introduction
The element libraries of Legacy and MSC Nastran share some common elements, noted
by “Both” in the Version column in the following table. However, except for the addition of
QUAD4 and TRIA3, Legacy Nastran’s element library for shell and solid elements is obsolete
or, at best in the case of the isoparametric IHEXi elements, not supported by modern pre
processor programs.
I have. therefore not included details for elements whose Version is Legacy. For those
elements, Legacy Nastran users are referred to the Nastran User’s Manual, Vol I (NUM) that
is include in the GitHub download site for Legacy Nastran. You will note that the page
reference for legacy elements is given in the form “NUM <page No.>.
The joint NASTRAN library includes the following types of elements:
Element
Descriptions Version Reference
Type
BAR A constant section beam bending element Both Bulk Data Image 1-3
An elastic spring connecting two displacement
ELAS Both Bulk Data Image 1-1
degrees of freedom
A flat quadrilateral shell element connecting four
QUAD4 Both Bulk Data Image 1-7
vertex nodes
ROD A one-dimensional theory of elasticity element Both Bulk Data Image 1-2
SHEAR A for node shear panel Both Bulk Data Image 1-6
A flat triangular element connecting three vertex
TRIA3 Both Bulk Data Image 1-7
nodes
ELBOW Curved beam or elebow element Legacy NUM 2.4-44
Six sided element having 8 vertex nodes formed
HEXA1 Legacy NUM 2.4-105
by 5 Legacy TETRA elements
Six sided element having 8 vertex nodes formed
HEXA2 Legacy NUM 2.4-105
by 10 overlapping Legacy TETRA elements
Eight node isoparametric solid. Uses same
IHEX1 Legacy NUM 2.4-56
numbering scheme as the HEXA element
IHEX2 Twento node isoparametric solid. Legacy NUM 2.4-57
IHEX3 32 node isoparametric solid Legacy NUM 2.4-58
Four node membrane quad formed by four
QDMEM Legacy NUM 2.4-82
overlapping TRMEM elements
QDMEM1 Four node isoparametric quad Legacy NUM 2.4-83
Four node membrane quad formed by four
QDMEM2 Legacy NUM 2.4-84
overlapping TRMEM elements
89 Nastran Primer
Structural Elements in NASTRAN
5
Defining Element Objects
Element
Descriptions Version Reference
Type
QDPLT Four node bending element Legacy NUM 2.4-85
QUAD1 Four node membrane and bending Legacy NUM 2.4-86
QUAD2 Four node membrane and bending Legacy NUM 2.4-87
A tetrahedral element connecting four vertex
TETRA Legacy NUM 2.4-96
nodes and six midsides nodes
TRIA1 Three node triangular membrane and bending Legacy NUM 2.4-101
TRIA2 Three node triangular membrane and bending Legacy NUM 2.4-102
TRIAM6 Six node lineat strain membrane Legacy NUM 2.4-105
TRMEM Three node triangular membrane Legacy NUM 2.4-106
TRPLT Three node triangular plate Legacy NUM 2.4-107
TRPLT1 Six node lineat strain membrane and bending Legacy NUM 2.4-108
TRSHL Six node shell Legacy NUM 2.4-109
Six node element havinf three quadralateral and
WEDGE two triangular faces. Same numbering as the Legacy NUM 2.4-113
PENTA element
A beam bending element with variable cross
BEAM MSC Bulk Data Image 1-4
section
BEND A curved beam element MSC Bulk Data Image 1-5
A hexahedral element connecting eight vertex
HEXA MSC Bulk Data Image 1-9
nodes and twelve midsides nodes
A pentahedral element connecting six vertex
PENTA MSC Bulk Data Image 1-9
nodes and nine midsides nodes
A curved shell element connecting four vertex
QUAD8 MSC Bulk Data Image 1-8
and four midside nodes
A tetrahedral element connecting four vertex
TETRA MSC Bulk Data Image 1-9
nodes and six midsides nodes
A curved shell element connecting three vertex
TRIA6 MSC Bulk Data Image 1-8
and three midside nodes
The elements are described in this chapter by describing:

• The appropriate Bulk Data statements
• The degrees of freedom for each element
• Special modeling features
• Stress recovery features
5.2 Defining Element Objects

The evaluation of the element stiffness and mass matrices requires: a description of the
local geometry; element properties for all but solid elements; and, the material properties.
Since the data base required to define a finite element model may include thousands of
elements it is important that a minimum set of data be specified for the element definition.
Typical structural systems are generally composed of relatively few materials. Since
several parameters may be required to define the properties of a material it makes sense to
attach a set number, i.e., a numerical flag, to the data associated with each individual material.
Then, instead of repeating the material parameters for each element, the appropriate material
90 Nastran Primer
Defining Element Objects
can be specified by pointing to the correct material set number. Furthermore, area properties
such as plate thickness and beam cross-sectional area moments of inertia tend to be the same
over several elements. So once again the concept of attaching a numerical flag to a related set
of data is useful in defining a minimum data set.
NASTRAN uses the set concept for the definition of the element data base by defining
• Connectivity information on a connection data statement that includes a pointer to a set
of element properties.
• Element properties on a property data statement that, in turn, includes a pointer to a set
of material properties.
• Material properties on a material data statement.
The data statements associated with the element connectivity and properties are described
in this chapter. The Bulk Data associated with the definition of material properties are
described in Chapter 10.
The form of the connection and property statements is similar for all the elements. The
general characteristics of these data statements and standard notation that will be used
throughout this chapter are described in subsequent sections.
5.2.1 Defining Element Connectivity

A connection statement specifies the region of the continuum represented by the object
and, based on the gird point locations, the geometry of the object. Members of the finite
element library are characterized by a name such as ROD, BAR, or HEXA for truss, beam, and
solid hexagonal elements, respectively. The mnemonic for a connection statement is formed
by prefixing a 'C' to the element name. For example, CROD, CBAR, and CHEXA are the data
statement names for specifying connectivity data for the truss, beam, and solid elements.
The physics in a local region is defined by the connection statement name, and a list of
grid points on the connection statement appropriate for the element.
The connection statements have the following general characteristics:
1.The type of element is specified by the mnemonic in the first field of the data
statement.
2.The second field is the Element Identification Number (EID). The EID is the
external tag for the element and must be a unique integer number for each element
in the input data file.
3.The third field is a Property Identification Number (PID). The PID is an integer
number that must be unique for a specific property data statement.
4.The number and the format of additional fields are dependent on the specific data
statement. The format rules for each data statement are defined in an internal table
called the IFP table that is used to interpret each Bulk Data statement. Data
statements that violate the format rules are flagged in the output and an error flag is
set to terminate the program prior to execution.
5.2.2 Defining Element Properties

The property statement always includes a pointer to a material statement. The mnemonic
name of a property data statement is formed by a ‘P’ preceding the element mnemonic or, for
solids and shell elements by generic properties statements beginning with a P. For example,
PROD, PBAR, and PBEAM are associated with the ROD, BEAM and BAR elements.
PSOLID and PSHELL are generic property statements for solid and shell elements,
Nastran Primer 91
5
Scalar Elastic Elements
respectively. The property statement includes all the geometric properties appropriate for a
given type of element. A single property statement may be referenced by all connection
statements which define elements with the same property. The use of the pointer system thus
eliminates repetition of data. Property statements have the following general characteristics:
• The type of property is specified by the mnemonic in the first field.
• The second field is the Property Identification, which must be an integer number. The
PID must be unique among the property numbers for a specific type of element.
• The third field is a Material Identification (MID), which is an integer number of the
MID of a material data statement.
• The remainder of the property data statement is dependent on the specific property data
statement.
5.2.3 Standard Notation

NASTRAN input data statements must satisfy specific restrictions on the type of number
(real or integer) and permissible range of values. In the remainder of this chapter the
following standard notation will be used:
EID Element identification number, (integer, > 0).

PID Property identification number, (integer, > 0).
MID Material identification number, (integer, > 0).
G Grid or scalar point number, (integer, > 0).
C Degree of freedom (integer, > 0 or blank) code at grid point, equal to zero or
blank for a scalar point.
CID Coordinate identification number, (integer, > 0).
NSM Non structural mass defined per unit length for lineal elements and per unit
area for two-dimensional elements, (real).
5.3 Scalar Elastic Elements

A scalar element having a single elastic constant that relates the displacements of two
degrees of freedom.
z
k
G2
G1
y
x
Figure 1-1 Scalar Elastic Element
92 Nastran Primer
Scalar elements may be defined between two grid point degrees of freedom by the ELAS1
and ELAS2 elements and between two scalar points by the ELAS3 and ELAS4 elements as
shown byBulk Data Image 1-1.
5.3.1 Spring Connecting Grid Point Degrees of Freedom
1 2 3 4 5 6 7 8 9 10
CELAS1 EID PID G1 C1 G2 C2
PELAS PID K GE S1
CELAS2 EID K G1 C1 G2 C2 Ge S
5.3.2 Spring Connecting Scalar Points
1 2 3 4 5 6 7 8 9 10
CELAS3 EID PID S1 S2
CELAS4 EID K S1 S2
Bulk Data Image 1-1 Elastic Spring, CELAS
5.3.3 Specification of Connected Degrees of Freedom

The connected degrees of freedom at a grid point are defined on the ELAS1 and ELAS2
data statements by specifying the external degree of freedom code using the pair G, C where
G is a grid or scalar point
and
C is a degree of freedom code that may be zero or blank if G is a

scalar point. If G is a grid point, then 1 < C< 6, where the degree
of freedom codes are interpreted in terms of the displacement
coordinate system of G as specified on the GRID data
statement.
If the second entry for the connected degree of freedom is left completely blank, then the
displacement of the spring is connected to ground and the displacement at the second end is
zero.
If the element is to connect only scalar points, then the CELAS3 and CELAS4 data
statements can be used where the degree of freedom is specified by the scalar point number. It
is interesting to note that the scalar elements imply the existence of the scalar points so that
they do not need to be explicitly defined using the SPOINT data statement.
5.3.4 Properties
The properties associated with the scalar spring may be defined directly on the connection
data statement, if CELAS2 or CELAS4 are used, or on a separate PELAS property data
statement. The properties in either case are defined as
K The spring rate (real)
Nastran Primer 93
5
Ge Damping coefficient (real)

S Stress recovery coefficient (real)
5.3.5 Stress Recovery

The stress recovery factor defined on the PELAS or CELAS data statement is used to
calculate the stress in the scalar element by
σ = P
--- (5-1)
S
where P is the force in the element and S is the stress recovery coefficient. The stress is printed
by including an appropriate STRESS data statement in Case Control.
5.3.6 Modeling with the ELAS Element

The ELAS element can be used to
• Define elastic constraint for generalized coordinates.
• Represent boundary flexibilities.
• Define elastic connections between degrees of freedom.
This general capability could be used to model an axial elastic element between two grid
points as shown by Figure 1-2. However, the ELAS element is not recommended as a
substitute for the ROD element, which is described in Rod and Truss Elements (p. 95), because
even a slight misalignment between the spring displacement and the coordinate axis can lead
to serious spurious internal constraints.
In spite of the potential problems involved with using the ELAS element to represent
ROD behavior we will consider modeling the elastic spring described below to illustrate the
modeling procedures involved in Its specification.
The spring has a spring rate of 5000 lbf inch and connects the points G1 and G2 that have
the following coordinates
G1 = (5, 3, 4) inch
G2 = (2, 5, 3) inch
The most straightforward way of specifying the spring is to define and connect
displacement degrees of freedom which are in the direction of the line segment which connects
the two grid points. This can be accomplished by:
• Defining a new coordinate system which has an axis parallel to the line between the two
points.
• Referring the displacement degrees of freedom at the two grid points to the new
coordinate system.
This modeling procedure has the advantage that the element and coordinate axes are aligned
so that no spurious constraints will be produced. The spring can be represented by the
following data statements.
1 2 3 4 5 6 7 8 9 10
CORD1R 100 1 2 3
94 Nastran Primer
Rod and Truss Elements
1 2 3 4 5 6 7 8 9 10
GRID 1 0 5. 3. 4. 100
GRID 2 0 2. 5. 3. 100
GRID 3 0 0. 0. 0.
CELAS1 1 5 1 3 2 3
PELAS 5 5.E3
where:
1.Coordinate system CID = 100 is a right-handed system whose origin is at point 1,
whose X3-axis is along the line joining points 1 and 2.
2.Specification of CID = 100 on the GRID data statements means that grid point
displacements and forces at these points will be interpreted relative to the CID =
100 coordinate system.
3.The elastic spring defined by the CELAS1data statement defines a scalar spring in
the X3-direction of the CID = 100 coordinate system.
5.4 Rod and Truss Elements

A truss element is a lineal structural element whose properties and behavioral variables
are continuously distributed along a line that joins two grid points and that provides resistance
to axial displacement and torsion, as shown in Figure 1-2.
θ x1
z
u x1
G1
θ x2
y A,J
G2
ux2
x
Figure 1-2 Truss Element
The element has a uniform cross-sectional area, A, and torsional constant, J, and connects
the two grid points G1 and G2, as shown. The element x-axis is defined by the directed line
drawn from G1 to G2.
Nastran Primer 95
5
The element may be defined either by CONROD, which includes connective as well as
property data on one data statement, or by the connectivity, property pairs CROD/PROD or
CTUBE/PTUBE, as shown in Bulk Data Image 1-2 The difference between the TUBE and
ROD elements is that the cross-section of the tube is assumed to be a circular cylinder of
diameter DIAM and wall thickness, THICK, while the rod has a cross-sectional area, AREA,
and torsional constant,
1 2 3 4 5 6 7 8 9 10
CONROD EID G1 G2 MID A J C NSM
CROD EID PID G1 G2

PROD PID MID A J C NSM
Bulk Data Image 1-2 Specification of Truss Element
1 2 3 4 5 6 7 8 9 10
CTUBE EID PID G1 G2
PTUBE PID MID D T NSM
5.4.1 Description of Input Data for Truss

The input data items included on the connect and property data statements are:
Field Description
Element identification number, which must be unique among all elements
EID
defined in Bulk Data. (integer, > 0)
G1, G2 Geometric grid point joined by the element. (integer, >0, G1 G2)
Material identification number that points to a set of material properties.
MID
(integer, > 0)
AREA Cross-sectional area. (real)
Torsional constant. The torsional constant is equal to the polar moment of
inertia only for circular cross-section. In general, J must be determined
J
experimentally, from a handbook such as the American Institute for Steel
Construction, or by means of torsion theory. (real)
D Outside diameter of the tube. (real)
T Thickness of the tube. (real, THICK < 1/2DIAM)
Stress recovery coefficient, which is the distance from the centroidal axis
C
to the point at which the shear stress is desired. (real)
NSM Non structural mass per unit length of the element.
5.4.2 Stiffness Matrix

The stiffness matrix for the truss element is based on the assumption that the axial
displacement and the rotation about the local x-axis vary linearly along the length of the beam.
The displacement and rotation are then taken to be:
96 Nastran Primer
 
 u x1 
 
 u  N1 N2 0 0  u x2 
  =   (Eq. 5-2)
 θ  0 0 N1 N2  θ x1 
 
 θ x2 
 
where:
N 1 =  1 – --x- N 2 = --x- (Eq. 5-3)

 L L
and L is the length of the element between grid points G1 and G2. The resulting stiffness
equation for the element is given by:
AE
------- – AE------- 0 0
L L    
 u x1   F x1 
– AE
AE
------- ------- 0 0    
L L  u x2   F x2 
  =   (Eq. 5-4)
JG
------- – JG  θ x1   M x1 
0 0 -------    
L L θ x2 
  M x2 
JG JG    
0 0 – ------- -------
L L
where A is the cross sectional area, J is the torsional constant, E is the Modulus of Elasticity
and G is the shear modulus.

The axial and shear stress are calculated as:
σ xx = P---
A
(Eq. 5-5)
TC
σ xy = -------
J
The stress recovery coefficient, C, is specified directly on the PROD data statement and is
taken to be D/2 for the TUBE. If A or J is not specified the associated stress recovery is
bypassed. Stress recovery is requested by the STRESS Case Control Directive.
Nastran Primer 97
5
5.4.3.1 Example: Plane Truss

Consider the truss shown in Figure 1-3. The Bulk Data statements required to specify the
grid point locations, the element connections, purge of the unconnected degree of freedom, and
the specification of geometric constraints using permanent single point constraints are shown
below.
y Note:
15 ft. 1.Element numbers are shown in circles.
1.The areas for all elements is 3in2.
1.Only loads in the x, y-plane are considered.
4
2 1.Simply supported at nodes 1 and 2.
15 ft.
3
1
3
4 10 ft.
1 5 2 x
Figure 1-3 Plane Truss
1 2 3 4 5 6 7 8 9 10
GRDSET 0 3456
GRID 1 0 0. 0. 0. 0 123456
GRID 2 0 180. 0. 0. 0 23456
GRID 3 0 180. 120. 0. 0
GRID 4 0 0. 180. 0. 0
CROD 1 1 1 4
CROD 2 1 4 3
CROD 3 1 1 3
CROD 4 1 3 2
CROD 5 1 1 2
PROD 1 100 3.
MAT1 100 3.E7
where
1.A GRDSET data statement specifies a default for the permanent single point
constraint field of the Grid data statement to purge degrees of freedom 3, 4, 5, and
6. This allows each grid point to move only in the x- and y-directions.
2.The coordinates of the grid points and the area on the PROD data statement are
expressed in consistent units (inches).
3.The fields associated with the torsional constant, J, the torsional stress recovery
coefficient, C, and the non structural mass, NSM, have been left blank. The values
of these fields are set equal to zero.
4.The PROD points to a MAT1 material data statement whose MID is 100. Since only
the modulus of elasticity is used to calculate the stiffness for the rod element, the
entries for other material coefficients have been omitted.
98 Nastran Primer
The BAR Beam Bending Element
5.The structure is restrained against rigid body motion by setting the x and y
components of displacement at Grid point one and the y component at Grid point
two equal to zero by means of the PSPC field. Since any entry in the PSPC field of
the GRID data statement overrides the default value on GRDSET, it is necessary to
include all the degrees of freedom as shown.
5.5 The BAR Beam Bending Element

A beam is a lineal element whose properties and behavioral variables are continuously
distributed along a line segment that joins two grid points. The BAR can represent bending
behavior about two perpendicular planes, axial behavior and torsion about the axis as shown
in Figure 1-4.
u6
z u3 u5
u4
u2 P6
u1
A
P3 P5
y
Za
P2
B
P1
GA
Zb P4
GB
x
Figure 1-4 The BAR Element
Legacy NASTRAN element library includes the BAR while MSC and NX include two
straight beam-type elements, called the BAR and the BEAM, as well as a curved beam and pipe
element called the BEND. The BEAM allows an offset shear center and axially varying area
properties while the BAR represents a subclass of the BEAM element capability without these
features. The BAR element is described in this section while the additional capabilities of the
BEAM are described in The BEAM Bending Element (p. 117). The BEND element is
described in Curved Beam Element (p. 127).
Referring to Figure 1-4, the BAR element may be offset from the connected grid points
to allow the modeling of eccentrically stiffened plates and shells. The element connects the
offset ends A and B and the element x-axis is taken as the directed line segment from offset
end A to offset end B of the beam.
The centroidal axis of the element is assumed to coincide with the local x-axis. This
assumption precludes the use of the BAR element to model a beam whose centroidal axis and
shear center do not coincide. The BEAM element described in Sec. 6.6 must be used in this
case.
Nastran Primer 99
5
The BAR element has the following modeling features and limitations:
1.The element represents bending behavior in two perpendicular planes in addition to
axial and torsional behavior.
2.The centroidal axis may be offset from the grid points.
3.The centroidal axis and axis of shear centers coincide.
4.Transverse shear flexibility may be included.
5.Pinned connections may be defined.
6.The area properties are constant.
7.The principal axes of inertia need not coincide with the element coordinate axes.
8.Stress can be recovered at up to four points on the cross-section at each end.
5.5.1 Description of BAR Input Data

The Bulk Data for describing the BAR element include the CBAR, CBARAO, PBAR,
and BAROR data statements shown on Bulk Data Image 1-3
1 2 3 4 5 6 7 8 9 10
BAROR PID V1,G0 V2 V3
CBAR EID PID GA GB V1,G0 V2 V3

PA PB Z1A Z2A Z3A Z1B Z2B Z3B
PBAR PID MID A Izz Iyy J NSM

Cy Cz Dy Dz Ey Ez Fy Fz
Ky Kz Iyz
CBARAO EID SCALE X1 X2 X3 X4 X5 X6

or
CBARAO EID SCALE NPTS X1 ∆X
where:
CBAR Defines the connectivity, the beam offset, the element coordinate system, and
pinned connections.
CBARAO Defines a series of points along the BAR element at which stress and/or
internal element forces may be recovered. (MSC only)
PBAR Defines the area properties relative to the element coordinate system, the non
structural mass, the points on the cross section at which stresses are to be
recovered, and the shear flexibility factors.
BAROR Provides default values for the PID and the orientation of the element.
Bulk Data Image 1-3 BAR Beam Bending Element
The fields of the CBAR and PBAR data statements are described in later sections.
100 Nastran Primer

5.5.2 BAR Connectivity

The CBAR Bulk Data statement defines connectivity for the BAR element, including the
definition of the offsets at grid points GA and GB; and, the element coordinate system. The
fields of the CBAR data statement are described by Table 5-1.
Table 5-1 Fields of CBAR
Field Description
Grid point identification numbers of connection points. (integer, > 0,
GA, GB
GA ≠ GB )
Components of a free vector at end A defined with respect to the
V1, V2, V3 displacement (i.e. CD) coordinate system at grid point GA. The vector is
used to determine the orientation of the element coordinate system. (real)
Identification number of third grid point to optionally define the free
G0
vector. (integer, >0)
Pin flags for bar ends A and B. Used to specify forceless degrees of
freedom corresponding to the pin flag number at the appropriate end of
PA, PB
the bar. (Up to 5 of the unique digits 1-6 can be put anywhere in the field
with no imbedded blanks, integer, > 0).
(Z1A,Z2A,Z3A) Components of offset vectors defined with respect to the displacement

coordinate systems at grid points GA and GB that define the position of
(Z1B,Z2B,Z3B)
the centroidal axis relative to the grid points. (real)
5.5.3 BAR Element Coordinate System

The features of offset centroidal axis and the three-dimensional behavior of the element
complicate the specification of the element coordinate system. Points A and B are points at the
ends of the centroidal axis of the BAR. They coincide with the grid points only if there are no
offsets. The element coordinate system is shown by Figure 1-5
Nastran Primer 101

5
z
y V
Z
A x
Y B
Za
Zb
GB
GA
Figure 1-5 Local Coordinates for BAR Element
where
1.The element x-axis is coincident with the line segment between points A and B,
positive from A to B. (If these is no offset then the x axis is the line segment drawn
from Grid GA to Grid GB.)
2.The element xy-plane is defined by the element x-axis and a user-defined vector,
V , having its origin at end A. The vector, V , may not be coincident with x-axis.
The user may define the vector in either of two ways
1.By specifying the components (V1,V2, V3) of a vector, V , that are defined in terms
of the displacement coordinate system at Grid point, GA, and that has its origin at
end A. (Note: The displacement coordinate system is specified by the CD
coordinate system field of the Grid data statement which specifies Grid point GA.)
2.By specifying a third Grid point, G0. The vector is taken as the directed line
segment from GA to G0 even if offsets are present.
If a third grid point, G0, is defined only for the purpose of specifying the vector, then the
degrees of freedom at G0 are not connected to the structure and will be removed automatically.
The element coordinate system is then defined using the element x-axis and V
z element = x element × V element
and
y element = z element × x element
102 Nastran Primer

5.5.4 Defining BAR Offset

The BAR element may be offset from the connected Grid points by specifying offset
vectors Z a and Z b as shown in Figure 1-5. The position of points A and B is then given by
ra = rG + Za
a
rb = rG + Zb
b
where
1.The offset vectors Z a and Z b are defined by components (Z1a, Z2a, Z3a) and (Z1b,
Z2B, Z3b), respectively, on the CBAR data statement. These are components
referred to the displacement coordinate system for the respective points. Thus, if the
displacement coordinate systems associated with GA and GB are rectangular and
cylindrical, respectively, the components of Z a and Z b will be interpreted in the
sense of unit vectors in the (X, Y, Z) and (R, θ, Z) directions, respectively.
2.The offset is treated as a rigid link between the Grid points and the associated ends
of the BAR.
3.The element coordinate system is defined with respect to the offset centroidal axis
of the element.
5.5.5 Pinned Connections

The pin flags PA and PB are used to specify one or more degrees of freedom, in the element
coordinate system, that does not transmit a force from the connected grid point to the
element. The pin flag field is a packed set of degree of freedom code numbers The pin flag
degree of freedom codes are:
Pin Flag Meaning
1 No force in element x-direction
2 No force in element y-direction
3 No force in element z-direction
4 No moment about element x-axis
5 No moment about element y-axis
6 No moment about element z-axis
The specification of a pinned connection results in an additional degree of freedom at a

grid point since the element and grid point displacements are not compatible as indicated in
Figure 1-6, which shows two BAR elements that are attached at grid point two.
z
a
θ y2
θ y1 b
θ y2
1 (a) 2 (b) 3 x
θ y3
u z1 u z2 u z3
Nastran Primer 103

5
Figure 1-6 Independent Rotations at a Hinge
A momentless hinge associated with the rotation about the z-axis at Grid point two is
specified using either PB = 5 for element (a) or PA = 5 for element (b). Since the pin flag defines
a forceless degree of freedom, the order of the stiffness matrix for the element in which the pin
flag is specified can be reduced by noting that the stiffness equation for the pinned degree of
freedom is equal to zero and that the associated row defines a constraint equation. For example,
the stiffness equation for element (a) in Figure 1-6 is given by:
   
k 11 k 12 k 13 k 14  u 1z   F 1z 
   
k 21 k 22 k 23 k 24  θ 1y   M 1y 
  =   (Eq. 5-1)
k 31 k 32 k 33 k 34  u 2z   F 2z 
   
k 41 k 42 k 43 k 44  θ 2y   M 2y 
   
Setting the moment M2y equal to zero and solving for θ2y gives:
1
θ 2y = – ------- k 41 u 1z + k 42 θ 1y + k 43 u 2z (Eq. 5-2)
k 44
This equation is of the form of linear constraint relation for the pinned degree of freedom
in terms of the remaining degrees of freedom that is used to reduce the order of the stiffness
matrix for the element in which the pin connection is specified. The stiffness coefficient for the
pinned degree of freedom must therefore be nonzero.
Because the pinned degree of freedom is reduces the stiffness matrix for the element we
see that the hinge may be specified in either element but not in both. If the same degree of
freedom were specified in both elements it would be completely uncoupled (i.e. it would have
no stiffness).
The displacement degree of freedom associated with the pin are not recovered. This fact
may influence the user's choice of the element in which the pin is defined.
5.5.6 Defining BAR Properties

The properties of the BAR element are specified on the PBAR data statement, whose
fields are described below
Field DESCRIPTION
AREA Area of bar cross-section. (real)
Area moments of inertia about the element y- and z-axes and the
Iyy, Izz, Iyz
associated cross product of inertia, respectively. (real)
J Torsional constant. (real)
NSM Non structural mass per unit length. (real)
Transverse shear factors in the element y- and z-directions, respectively.
Ky, Kz
The shear factor is dependent on the beam cross-section.
Element y and z-coordinates for four points on beam cross-section at
(Cy, Cz) which stresses are to be recovered at each end of the BAR. (real). (These
points are identified in the output as points 1 through 4, respectively.)
Table 5-2 Description of PBAR Fields
104 Nastran Primer

The area properties specified on the PBAR data statement are calculated with respect to
the element coordinate system. In order to specify the area properties we thus consider the
cross-section of a beam element as shown in Figure 1-7.
z′
z
dA
y′ y
P
z′
θ
y′
O
Figure 1-7 BAR cross section
where
O is a point which lies on the centroidal axis
P is a general point on the cross section with an area dA
The second area moments of inertial relative to an arbitrary set of axes, (y',z'), are given by:
∫ ( z' )
2
I y'y' = dA (Eq. 5-3)
∫ ( y' )
2
I z'z' = dA (Eq. 5-4)
I y'z' = ∫ y'z' dA (Eq. 5-5)
A = ∫ dA (Eq. 5-6)
(y,z) is the set of local coordinates which is shown by Figure 1-5 and which is to be
defined on CBAR data statement. The moments of inertia of the cross section
relative to (y, z) are defined in terms of those found with respect to (y', z') as
follows
The area moments of inertia can be transformed to a set of axes, (y, z) which are related
to the (y', z') coordinates by a rotation, θ, as follows:
2 2
I yy = I y'y' cos θ + I z'z' sin θ – I y'z' sin 2θ (Eq. 5-7)
Nastran Primer 105

5
2 2
I zz = I z'z' cos θ + I y'y' sin θ – I y'z' sin 2θ (Eq. 5-8)
2 1
I yz = I y'z' cos θ + --- ( I y'y' – I z'z' ) sin 2θ (Eq. 5-9)
2
Since the moments of inertia for the cross-section are defined relative to a specific
coordinate system the analyst must specify the moments of inertia on the PBAR that are
calculated with respect to the element coordinates specified on the CBAR.
The other definitions of cross-sectional properties, J, the torsional constant and the
transverse shear coefficients, Ky and Kz, are not as straight forward since both depend on the
distribution of area in a more complex way than the bending moments of inertia and area.
The torsional constant, J, relates the angle of twist, θx, to the torsional moment, Mx, the
length, L, and the shear constant, G, as follows
TL
θ x = ------- (Eq. 5-10)
JG
The torsional constant is equal to the polar moment of inertia only for the case of a circular
cross-section. For other cross-sections the analyst should consult handbooks such as [1].
The shear flexibility coefficients, Ky and Kz, define the shear-effectiveness of the beam.
These shear coefficients define the shear displacements vys and vzs, respectively. The total
displacement of the reference axis is then given by
vy= vyb + vys (Eq. 5-11)
and
vz = vzb +vzs (Eq. 5-12)
where vyb and vzb are bending displacements.

The total shear displacement of a beam of length L in the z-direction is related to the shear
strain γxz as follows:
vzs = γxzL (Eq. 5-13)
where the shear strain and shear stress are related by:
σ xz
γ xz = -------
- (Eq. 5-14)
G
Then, since σxz = Vz/A where Vz is the transverse shear force, we have:
Vz Q
σ xz = ---------
- (Eq. 5-15)
I zz b
where Vz is the transverse shear force, Q, is the moment of the area beyond that value of z for
which the shear stress is to be determined, and b is the width of the cross section at that point.
Since Q/Izzb has the units of area we can represent(Eq. 5-15) as:
Vz
σ xz = ---------
- (Eq. 5-16)
Kz A
106 Nastran Primer

where Kz is a factor which depends on the cross-section and accounts for the of shear
distribution over the cross-section. The substitution of (Eq. 5-16) and ((Eq. 5-14) into
(Eq. 5-13) then leads to the following relationship for the shear displacement
Vz L
v zs = -------------
- (Eq. 5-17)
K z AG
where the factor 1/KAG is called the shear flexibility. The shear factor, K, which is the
reciprocal of that defined in [5] is: K = 5/6 for rectangular cross-sections, K = 9/10 for circular
cross-sections, K ≅ A f ⁄ ( 1.2A ) and K ≅ A w ⁄ A for bending of wide flange I-beam about
minor and major axes of inertia, respectively, where Af is the flange area and Aw is the web
area. The analyst should consult handbooks such as [5] for additional material.
5.5.7 Area Properties

The BAR element relates the six grid point displacement degrees of freedom at each of
its ends to the associated stress and moment resultants as indicated by P1 through P6 on Figure
1-4. The components of this force vector are defined as
P1 Axial force in direction of the element x-axis
P2 Axial force in direction of the element y-axis
P3 Axial force in direction of the element z-axis
P4 Moment about the element x-axis
P5 Moment about the element y-axis
P6 Moment about the element z-axis
The following area coefficients relate the element displacement degree of freedom {ue},
and force resultants, {Pe}
A Cross-sectional area defines resistance to an axial load P1.

Ky Shear area coefficient, resists the transverse load P2.
Kz Shear area coefficient, resists the transverse load P3.
J Torsional constant, resists the torsional moment P4.
Iyy Area moment of inertia about the element y-axis, resists the bending moment P5
and the transverse load P3.
Izz Area moment of inertia about the element z-axis, resists the bending moment P6
and the transverse load P2.
Iyz Cross product of inertia, nonzero unless the element y- and z-axes are principal
axes of the cross section; for non-principal axes, resists all loads except P1 and
P 3.
The area coefficients generate non zero terms in the stiffness matrix associated with the
element degrees of freedom as shown by Table 5-3
Connected Displacements
Area Coefficient Description
u1 u2 u3 u4 u5 u6
A Area X
Ky Shear Coefficient X
Nastran Primer 107

5
Connected Displacements
Area Coefficient Description
u1 u2 u3 u4 u5 u6
Kz Shear Coefficient X
J Torsional Constant X
Iyy Area moment of inertia X X
about y-axis
Izz Area moment of inertia X X
about z-axis
Iyz Cross product of inertia X X X X
Table 5-3 Connected Displacements

The absence of one of the area coefficients on the PBAR data statement will result in zero
coefficients in the rows and columns of the element stiffness matrix for the associated degrees
of freedom. For example, if Iyy is the only nonzero area coefficient specified on the PBAR
statement, then the stiffness matrix will provide resistance to only the u3 and u5 degrees of
freedom (the displacement in the z direction and the rotation about the y-axis) in the element
coordinate system.
The shear coefficients Ky and Kz can be specified only if the element coordinate system
coincides with the principal axis for the cross-sectional inertia. If Iyz ≠ 0,the BAR element is
assumed to be infinitely rigid to shear deformations.
5.5.8 Nonstructural Mass

The mass of the beam per unit length may be entered by the parameter NSM in field eight
of the PBAR data statement. The calculation of mass properties is discussed in Chapter 12,
Normal Modes Analysis.
108 Nastran Primer


The location of up to four points labeled C, D, E and F at which stress is recovered on the
cross-section at each end of the bar element can be specified relative to the centroidal axis as
shown by Figure 1-8. The location of these points is specified by element y- and z-coordinates
at from one to four points on the PBAR data statement.
y
Fz
C
F
Fy
D E
Figure 1-8 Stress Recovery Points on the BAR
The following stress output recovered at ends A and B of the BAR element by STRESS
Case Control Directives:
1.The bending stress at points C, D, E and F on the cross-section
2.The axial stress
3.The maximum stress
4.The margins of safety based on stress limits prescribed on a MAT1 material data
statement.
5.5.10 Element Forces - Use of CBARAO

The CBARAO data statement is a feature that is available in MSC Nastran that allows the
user to recover element forces and stresses at intermediate points along the length of the BAR
element. The CBARAO data statement is typically used in conjunction with a load distribution
defined by a PLOAD1 data statement.
The inclusion of internal data recovery points may lead to some confusion concerning the
sign convention used to interpret the element forces since either of the following sign
conventions can be used.
• Beam Convention
• Coordinate Convention
Nastran Primer 109

5
The beam sign convention which is shown by Figure 1-9 is that positive forces and
moments cause positive curvature of the beam element. The coordinate sign convention, on the
other hand, is that positive forces and moments are in the direction of positive displacements
and rotations, respectively.
y
z
Vyb
Mza Mzb Mya
Myb
P P
P P
Vza
Vya
x x
a) Plans 1 (xy) b) Plane 2 (xz)
Figure 1-9 Beam Forces and Moments
The sign convention used for element forces for the BAR element is:
• Coordinate convention if PLOAD1 and/or CBARAO data statements exist
• Beam sign convention otherwise
For other NASTRAN beam-type elements, i.e., the BEAM and BEND elements the element
forces are always interpreted using the coordinate convention.
There are two forms of the CBARAO Bulk Data data statement shown on Bulk Data
Image 1-3 that provide the user with two different ways of prescribing internal data recovery
points.
Form 1- Allows the user to prescribe up to six unique positions on the BAR
Form 2- Allows the user to specify positions by means of a length increment and
number of increments.
5.5.11 Data Recovery Points - CBARAO Form 1

The data fields on the first form of the CBARAO data statement allow the user to define
internal points by at up to six interior points along the length of the bar. The data fields shown
on Bulk Data Image 1-3 are described as follows.
EID Element identification number of a specific BAR element (integer > 0)
SCALE A literal parameter that defines how the axial coordinates of interior points
are to be scaled as follows
LE - The values xi are actual distances.
FR - The values xi are normalized distances where element length

is the normalizing factor.
xi The positions of up to six interior points along the element axis. The end
points are not to be included since the forces and stresses are normally
recovered at these points. (Real > 0.0)
If the literal value 'LE' is supplied for the SCALE field then a PLOAD1 Bulk Data data
statement for the element must be present.
110 Nastran Primer

5.5.12 Data Recovery Points - CBARAO Form 2

The second form allows the user to specify a number of equally spaced internal points
with a minimum input. The specific form of the data statement is implied by the type of number
in the fourth field as follows
Real -Form 1
Integer -Form 2
The fields of the second form which are different from the first form are as follows
NPTS Number of interior data recovery points. (Integer 0 < NPTS <
19)
xi Position of first point. (Real > 0.0)
∆x Incremental distance along element axis. (Real > 0.0)
The series interior points

x1 = xi - 1 + ∆x (i = 1, 2 ....NPTS)
is then generated.
5.5.13 BAR-Element Examples
5.5.13.1 Specifying the Element Coordinate System

Consider the structure shown in Figure 1-10, which consists of two beams at right angles.
The structure is fully constrained at grid point one, and may be loaded with arbitrary forces and
moments at grid points two and three. Each of the beams has the same cross-sectional inertia
properties, but the moments of inertia about the two orthogonal axes are not equal. Assuming
that the element coordinates are principal axes, we wish to orient the axes of the two beams
such that each beam provides maximum resistance to bending in the Basic X, Y-plane.
y
200 in.
2
50 in. 2 2 3
1 1
2 I12 = 0
1
I22 = 420 in4
100 in. I11 = 1170in4
1 A = 30.3 in2
G0
K1 = K2 = 0
50 in. J = 10 in4
E = 30 X 106 psi
x
Figure 1-10 Example of Specifying Local Axes
Nastran Primer 111

5
The orientation vector, V , and the element x-axis, define the element x-y plane. The
vector can be specified either by a third geometric grid point or by its components. The same
vector can be used for both elements if it is not colinear with the element x-axis for either
element. The vector can then be specified, in terms of components, by the following BAROR
data statement.
1 2 3 4 5 6 7 8 9 10
BAROR 1. 1. 0.
where the components define a vector in the Basic X-Y plane at an angle of 45 degrees to the
Basic X-axis. The element x,y-plane is then coincident with the Basic X,Y-plane, since V lies
in the Basic X,Y-plane and, by definition, also lies in the element x,y-plane.
Alternately, grid points three and one could be used to specify G0 for elements one and
two, respectively, or an additional grid point G0, as shown, could be defined and be referenced
by both elements.
Using the BAROR specification of the orientation vector, the element coordinate system
is known and the element properties can be specified using the PBAR statement. The data set
required to define the grid points, the element properties, and the constraints is:
1 2 3 4 5 6 7 8 9 10
GRID 1 0 50. 50. 0. 0 123456
GRID 2 0 50. 100. 0. 0
GRID 3 0 200. 100. 0. 0
BAROR 20 1. 1. 0.
CBAR 1 1 2
CBAR 2 2 3
PBAR 20 30 30.3 1170. 420. 10.
MAT1 30 3.E7 .3
where
1.The constraints at grid point 1 have been specified by means of permanent single
point constraints on the Grid statement
2.(The unconnected degrees of freedom will be removed during the analysis by the
AUTOSPC procedure described in Chapter 11.
3.A consistent set of units is used in specifying grid point coordinates and element
properties.
4.The BAROR data statement has been used to define the components of V for each
element. Note that the vector lies in the Basic X,Y-plane. The element x,y-plane for
each element is thus coplanar with the Basic X,Y-plane.
5.The same PBAR statement is referenced by both CBAR data statements
6.The PBAR statement references a MAT1 Bulk Data statement described in Chapter
9 having a material set 30.
112 Nastran Primer

5.5.13.2 Use of Pin Flags

Consider the structural system shown in Figure 1-11, which consists of two cantilevered
beams attached with a momentless hinge at point A, loaded in the Basic Y-direction.
y P a
4
I zz = 7cm
x
A
2.5m 2.5m
Figure 1-11 Example of Forceless Hinge
The system is modeled using two BAR elements as shown:

y
1 2
x
1 2 3
The condition that no moment Mz is transmitted from element one to element two is
specified by specifying the pin f lag to be 6 in element two at grid point two. This modeling
specification means that the rotation of element one will be different than the rotation of
element two at grid point two. This degree of freedom is not recovered by the output modules.
The displacement recovered at grid point two will be the rotation of element one and the
common displacement u2y.
The, data statements describing the system are
1 2 3 4 5 6 7 8 9 10
GRID 1 0 0. 0. 0. 0 123456
GRID 2 0 250. 0. 0. 0 1345
GRID 3 0 500. 0. 0. 0 123456
CBAR 1 15 1 2 0. 1. 0.
CBAR 2 15 2 3 0. 1. 0. +C1
PBAR 15 10 7.
MAT1 10 1.E8
+C1 6
where
1.A consistent set of metric units (cgs) has been used so that the load would be
specified in a Newtons
2.The constraints are specified on the grid statement.
3.The element x,y-axes of the BARs have been defined to be coincident with the
Basic X,Y-axes.
Nastran Primer 113

5
4.The pin flag is specified in element two. The pin flag could be specified in either
element but not in both. The rotational degree of freedom at grid point two in
element two is not recovered since the displacement degree of freedom associated
with the pin flag is removed from the analysis set of displacements
5.The pin flag is specified in a continuation whose parent is the CBAR statement for
element two. The continuation, C1, has been placed after the MAT1 statement to
emphasize that the Bulk Data statements may be placed in any order.
5.5.13.3 Use of Offset Vector

Consider the case of the eccentrically stiffened plate shown in Figure 1-12 with geometric
parameters for the plate and beam given as follows
L = 10 cm Iyy = 1 cm4
a = b = 20 cm Izz = 0.25 cm4
h = 3 cm J = 0.5 cm4
t = 0.5 cm A = 0.75 cm2
h
x
b y
a
t
Figure 1-12 Eccentrically Stiffened Plate
114 Nastran Primer

The load is assumed to be symmetric so that the system can be modeled by using
one-quarter of the structure and appropriate symmetry boundary conditions. The finite element
model for the eccentrically stiffened plate is shown in Figure 1-13, where the quarter panel has
been discretized by using nine Quad4 shell elements and three BAR elements as shown. The
BAR centroidal axis is to be offset from the centroidal surface of the plate through a distance
of (h+t/2).
Z4
Beam centroidal
Z3 4
Z2 3 8
Z1 2 7
12
1 6
x 11
5 16
10 y
15
9
14
13
Figure 1-13 Finite Element Model of Stiffened Plate
The Quad4 shell elements, which are described in a later section, must include both
bending and membrane action. Only the input for modeling the offset beam is included in the
data set shown below:
1 2 3 4 5 6 7 8 9 10
CBAR 1 1 1 2 0. 1. 0. +C1
CBAR 2 1 2 3 0. 1. 0. +C2
CBAR 3 1 3 4 0. 1. 0. +C3
+C1 0. 0. 1.75 0. 0. 1.75
+C2 0. 0. 1.75 0. 0. 1.75
+C3 0. 0. 1.75 0. 0. 1.75
PBAR 1 1 .3725 .125 .5 .25
In this Bulk Data we have defined three BAR elements that include the offset vectors.
1.The offset vector Z is the same at both ends of each BAR. The vector is the Basic
Z-direction and has a magnitude of (t + h)/2.
2.The area properties are one-half those for the physical beam because the other half
of the system is accounted for by symmetry conditions.
3.The oreintation vector, V , is taken as a unit vector in the y-direction for each BAR.
This vector has its origin at the offset points lying on the centroidal axis. The vector
thus lies in a plane z = 1.75.
4.Consistent metric units have been used.
Nastran Primer 115

5
5.5.14 Modeling Considerations

The stiffness matrix for the BAR is exact, within the context of beam theory, for the case
where the beam cross-section is constant and the element is loaded only by concentrated forces
and moments at the ends. More often than not the structural analysis problem involves
distributed loads and nonuniform cross-sections, and the question of modeling errors naturally
arises.
5.5.14.1 Distributed Loads

Some indication of the number of elements required to model adequately a distributed
load acting on a beam can be obtained by considering the cantilevered beam subject to a
uniform load as shown in Figure 1-14.
z
po
EI y
Figure 1-14 Uniformly Loaded Cantilever
The beam is represented as a collection of beam elements where the lumped load at the
ends of each element is given by
1 L
p i = --- p o ----
2 N
where N is the number of elements used to discretize the beam.

The approximate tip displacement, normalized with respect to the exact solution, is
presented as a function of the number of elements in Figure 1-15. For this particular case of
load distribution and boundary conditions it can be seen that four elements are required to give
a reasonable solution for the tip displacement when load lumping is used. One would expect
that even more elements would be required to provide adequate resolution to the stress
distribution, especially in the area of the base where there are high curvature gradients.
The correct procedure for distributed loads is using work equivalent loads which leads to
the exact solution by using one element. This load, which is is available in MSC Nastran called
PLOAD1 is described in Distributed Load on Beam - PLOAD1 (p. 269).
δ tip
-------------
-
δ exact 1
1 2 3 4 5 6
Figure 1-15 Discretization Error in Modeling Uniform Load
116 Nastran Primer

The BEAM Bending Element
5.5.14.2 Nonuniform Cross-Section

In order to investigate the effect of nonuniform cross-section, a cantilever beam is
considered with a variable moment of inertia loaded by a concentrated load in Fig. Figure 1-16.
y
P
I ( x ) = I o  1 – ---
x
L
EI(x) x
Figure 1-16 Variable Cross-Section Beam
The nondimensional tip displacement is presented as a function of the number of elements

in Figure 1-17. The cross-sectional moment of inertia for each element is taken to be that at the
midpoint of the element. It can be seen that for this particular case of element geometry and
external loads, approximately four elements are required to obtain reasonable convergence to
the exact solution. In NASTRAN a variable cross-section bending element is more properly
defined by using the BEAM element which is described in 5.6.
Both these examples indicate that the knowledgeable user will perform simple analyses
of this kind to evaluate the effect of parametric changes on the number and distribution of
structural elements.
δ tip
-------------
- 1
δ exact
1 2 3 4 5
Figure 1-17 Normalized Tip Displacement for Variable Cross-Section Beam
5.6 The BEAM Bending Element

The BEAM element is a generalization of the BAR element that is availble in MSC
Nastran that includes all of the modeling capability of the BAR and the following additional
features.
1.The centroidal axis, the axis of shear centers, and the axis of centers of nonstructural
mass may all be different.
2.Section area properties and nonstructural mass may vary arbitrarily along the beam.
Nastran Primer 117

5
3.Contribution of cross-section warping to torsional stiffness may be included.

4.Distribution of mass polar moment of inertia can be specified.
5.Shear relief due to taper can be defined.
5.6.1 Degrees of Freedom

The degrees of freedom associated with the BEAM element are the three displacements
and three rotations, as shown in Figure 1-4 for the BAR element and one additional degree of
freedom at each end, the twist φ. The twist is related to the rotation about the axis of the beam
as follows:
dθ x
ϕ = (Eq. 5-1)
dx
5.6.2 BEAM Description

The BEAM element connects two offset points A and B as shown by Figure 1-18.
z
Z
M yA A′′
A′ Axis of Nonstructural Mass
M zA
y Centers of Gravity
N yA N zA
A Centroidal Axis
v
ZA
M yB
B′′
GA N yB
B′ M zB
N zB
Axis of Shear Centers
B Y
ZB x
X GB
Figure 1-18 BEAM Tapered Beam Bending Element
The geometry of the BEAM is similar to that for the BAR element, but the element x-axis,
which is the line segment drawn between A and B, is now assumed to be coincident with the
axis of shear centers rather than the centroidal axis.
The BEAM is specified by the CBEAM and PBEAM data statements shown by Bulk Data
Image 1-4. The connectivity for the BEAM is defined on a CBEAM data statement which is
similar to the CBAR data statement except for the additional scalar degrees of freedom at each
end that represent the twist.
118 Nastran Primer

The BEAM properties are specified by the PBEAM data statement. The PBEAM
statement is similar to the PBAR statement but, since variable area properties and internal
stress recovery points can be specified, the form of the PBEAM must support these options.
Because the BEAM represents an extension of the BAR modeling capability only those BEAM
features that are not supported by the BAR element will be described. Note that the
continuation mnemonics, 001, 002 etc. are referred to subsequently when describing the input
data fields.
1 2 3 4 5 6 7 8 9 10
CBEAM EID PID GA GB V1 V2 V3
PA PB Z1A Z2A Z3A Z1B Z2B Z3B
SA SB
PBEAM PID MID AA IzzA IzzA IyzA JA NSMA

CyA CzA DyA DzA EyA EzA FyA FzA
SO X/XB A IZZ IYY IYZ J NSM
Cy Cz Dy Dz Ey Ez Fy Fz
Ky Kz Sy Sz NSIA NSIb CWA CWb
MyA MzA MyB MzB NyA NzA NyB NzB
Bulk Data Image 1-4 Specification of BEAM Element
5.6.3 CBEAM Data Statement

The CBEAM Bulk Data statement includes all of the data described on Bulk Data Image
1-3 for the BAR element and an optional second continuation (+CB2) containing the fields SA
and SB where
SA, SB Scalar or grid point identification numbers for the warping variables at ends
A and B, respectively. (integer > 0 or blank)
The warping degrees of freedom SA and SB must be defined using either SPOINT or GRID
data statements. If GRID data statements are used then the first degree of freedom is
considered to be the warpage.
5.6.4 Local Coordinate System for the BEAM

The definition of the local coordinate system for the BEAM element is complicated by
the ability to define noncoincident centroidal axis, axis of shear centers, and axis of centers of
nonstructural mass in addition to defining the orientation of the local beam axes with respect
to the global coordinates.
The element coordinate system for the BEAM is shown by Figure 1-18 where
1.The element x-axis is taken to be along the line from end A to end B.
2.The axis of shear centers coincides with the element x-axis.
3.Ends A and B may be offset from the geometric grid points GA and GB by vectors
Z A and Z B as described for the BAR element.
4.The element x,y-plane is defined by the vector, V , as shown by Fig. 6-5 where the
vector can be specified in either of two ways.
Nastran Primer 119

5
• Components of a free vector (V1, V2, V3) taken from end A

• The vector from end A to a third grid point GO
5.The centroidal axis is along the line from points A' to B' which are defined relative
to A and B by specifying the components of the offset vectors, N A and N B ,
respectively. The components of these vectors (NyA, NzA) and (NyB, NzB) are
specified in element coordinates on the PBEAM statement (+PB5)
6.The axis of nonstructural mass centers of gravity is along the line from points A" to
B" which are defined relative to A and B by the components of the offset vectors
M A and M B respectively. The components of these vectors (MyA, MzA) and (MyB,
MzB) are specified in the element coordinates on the PBEAM (+PB5).
5.6.5 PBEAM Data Statement

The properties of the BEAM element are specified on the PBEAM data statement shown
on Bulk Data Image 1-4. The PBEAM statement is designed to allow the specification of a
significant amount of descriptive data for the most general case but for the case where the
modeling capability required coincides with the BAR capability the PBEAM data fields are
closely related to the PBAR fields. The user may thus consider using the BEAM exclusively
since it contains the BAR capability as a subset.
The PBEAM together with all of the continuation statements looks formidable but the
detail is necessary to allow the analyst to use all of the BEAM capability. The PBEAM data
fields are described as follows, where the reference is made to the sequence number in field 10
of the data statements in Bulk Data Image 1-4.
The PBEAM is described by numbers 004 through 008. These statements are:
Statement Name Description

PBEAM The first statement of PBEAM is almost identical to the PBAR and is the
only required input. The only essential difference is the inclusion of the
Iyz field in the parent data statement.
004 An optional continuation that specifies the position of up to four stress
recovery points on the cross section at end A. This continuation statement
can be omitted even if additional PBEAM continuations are required.
005 - 006 Optional continuations that allow the user to specify area properties at
specific location as along the length of the BEAM. Up to nine internal
points can be defined using pairs of statements of this form.
007 An optional continuation statement defining shear factors, (Ky, Kz), Shear
relief for taper, (Sy, Sz), nonstructural mass moment of inertia, NSI, and
warping coefficients, CW.
008 An optional continuation on which offset axes of non structural mass and
offset centroidal axes are defined
PBEAM Includes a pointer to a MAT1-type Bulk Data set and specifies the cross-sectional
area properties and the non structural mass where the fields are defined as follows
AA Cross-sectional area at end A (real > 0., no default)
(IyyA, IzzA) Cross-sectional moment of inertias about axes parallel to y and z
axes which pass through centroidal axis (real > 0., no default)
IyzA Cross product of inertia at end A. (real or blank, default is 0.)
JA Torsional constant at end A. (real or blank, default is 0.)
120 Nastran Primer

NSMA Nonstructural mass per unit length at end A. (real or blank, default is
0.)
Statement An optional continuation statement that specifies the location of four stress recovery points
004 (C, D, E, and F) on the cross section at end A. It may be omitted if stress recovery is not
required and if data fields on subsequent continuations do not require its use.
Sequence Successive packets of statements having the form of statements 005 and 006 that specify
005 area properties and nonstructural mass and/or stress recovery points at internal positions,
x/xB along the length of the BEAM. The second field, SOPT, of 005 is a character string
and
which does two things: it identifies the continuation as the specification of a internal
Sequence position; and, it controls the calculation of stress and internal forces at intermediate
006 points depending on the value of the parameter as follows
YES Calculate stresses at the points defined on the immediately
following 0006-type continuation
YESA Calculate stresses at the same points defined at end A by a
004-type statement which is then required if this option is specified
NO No intermediate stress output is desired.
The 005-type statement is allowed only if a 005-type statement has a value, YES, in field
2.
The normalized position along the beam is defined by the x/xB-field on a 005-type
statement followed by fields defining the area coefficients and non-structural mass at that
position. Up to nine points can be defined but one of them must have a value of 1.0 for
x/xB to define the properties of end B.
Statement
An optional continuation statement defining
006
(Ky, Kz) Shear stiffness factors, K, which appears in the relation, KAG, in
the y and z directions. (real or blank, note that the default is 1,
whereas the default is zero for the PBAR)
(Sy, Sz) Shear relief coefficient due to taper in the y and z directions. (real
or blank, default is 0.)
NSIA, NSIB) Nonstructural mass moment of inertia per unit length about the
nonstructural mass center of gravity at ends A and B. (real or
blank, default for NSIA is 0., that for NSIB is the value of NSIA)
(CWA, CWB) Warping coefficients at ends A and B. (real or blank, default for
CWA is 0., that for CWB is the value of CWA)
Statement An optional continuation statement defining the components of the offset vectors
007
(MyA, MzA) Offset of axis of nonstructural mass center of gravity from the axis
(MyB, MzB) of shear centers at end A and B. (real or blank, default for My is 0.,
that for Mz is the value of My)
NyA, NzA) Offset of centroidal axis from the axis of shear centers at ends A
(NyB, NzB) and B. (real or blank, default is Ny is 0., that for Nz is the value of
Ny)
Nastran Primer 121

5
5.6.6 BEAM Examples
5.6.6.1 Specifying Noncoincident Shear and Centroidal Axes.

Consider the channel beam shown in Figure 1-19.
L
A
z Ny
Section Properties Relative to Centroidal Axis
n e
2
t 3b - = 0.469in.
e = ---------------
6b + h
3
t 4
J = ---- ( 2b + h ) = 0.009in
3
y h 3 2
C w = -------------  ------------------- = 0.301in
N S tb h 3b + 2h 6
12 6b + h
2
A = 0.986in
2
b I yy = 1.58in
4
I zz = 0.172in
t = 0. 17 in. N y = e + n = 0.766in
b = 1.315 in.
h = 3.17 in.
E = 30 x lb/in.2
I = 20.0 in.
Figure 1-19 Channel Beam
The BEAM modeling capability provides a great deal of freedom in defining the shear
and centroidal axes. By using the various offset vectors we can define either the shear and
centroidal axes to be coincident with the line segment between two geometric grid points. We
will proceed to define the BEAM both ways, but first we need to obtain the geometric
properties of the cross-section.
The position of the shear center and the centroidal axes must be calculated based on the
cross-sectional geometric parameters. The position of the shear center relative to the center of
the web is given by e. The position of the shear center can be calculated using the techniques
outlined in [4] or can be looked up in handbooks such as [5]. The centroidal axis offset from
the center of the web, n, can also be determined from the condition that the first area moment
about the centroidal axis must be zero. The offset of the centroidal axis from the shear center
as well as other cross-sectional properties for the channel can then be found and are shown on
Figure 1-19.
122 Nastran Primer

5.6.6.2 Shear Axis on Line Joining Grid Points

The axis of shear centers lies on the line joining offset points A and B as shown by Figure
1-18. If the offset vectors of each end of the beam are null then the axis of shear centers lies on
the line joining the geometric grid points. The channel could be modeled with a single BEAM
element which connects the structural degrees of freedom associated with geometric grid
points 1 and 2 at each end and two scalar degrees of freedom, points 3 and 4, which represent
the twist at the left and right ends, respectively, by the following Bulk Data
1 2 3 4 5 6 7 8 9 10
GRID 1 0 0. 0. 0. 0 0001
GRID 2 0 20. 0. 0. 0 0002
SPOINT 3 4 0003
CBEAM 1 2 1 2 0. 1. 0. 0004
+003 0005
+004 3 4 0006
PBEAM 2 101 0.986 0.172 1.58 0. 0.009 0007
+007 0008
+008 0. 0. 0.301 0.301 0009
+009 0.766 0. 0.766 0. 0010
where sequence 0001 and 0002 in field 10 define grid points along the X-axis of the Basic
coordinate system and sequence 0003 defines two scalar points which are associated with
warping. The degrees of freedom associated with these points are connected by the CBEAM,
sequence 0004 through 0006 where:
1.Grid points 1 and 2 are associated with ends A and B, respectively
2.The element coordinate system is defined by vector components (0., 1., 0.) on 0004
in terms of the Basic coordinate system so that the coordinates of the BEAM
element coincide with the coordinate system shown in Fig. 6-19
3.Sequence 0005 is required even though the fields are null because the twist degrees
CBEAM.
4.The continuation, sequence 0006, specifies scalar degrees of freedom 3 and 4 are
associated with the warpage at ends A and B, respectively.
5.6.6.3 Non Coincident Shear and Centroidal Axes

The properties are defined on the PBEAM, sequence 0007 and its continuations 0008, 0009
and 0010 where:
1.The specification of moments of inertia is appropriate for the element coordinate
system defined by the connection statement, where Izz has been transferred to the
shear axis.
2.The PBEAM continuation 0008 is required in this case even though stress recovery
is not desired because data fields in subsequent continuations must be defined.
3.The continuation 0009 specifies the warping coefficients.
4.The continuation 0010 specifies the component of the offset vector from the shear
center to the centroidal axis. The vector components (0.766, 0.) at each end are
interpreted in element coordinates so that the centroidal axis is offset 0.766 inch
along the element y-axis at each end.
Nastran Primer 123

5
5.6.6.4 Centroidal Axis on Line Joining Grid Points

The centroidal axis can be made coincident with the line joining grid points GA and GB
if points A' and B' as shown by Figure 1-18 coincide with GA and GB. This can be
accomplished by defining Z A = – N A and Z B = – N B . The channel could then be modeled
using the following Bulk Data
GRID 1 0 0. 0. 0. 0 0001
GRID 2 0 20. 0. 0. 0 0002
SPOINT 3 4 0003
CBEAM 1 2 1 2 0. 1. 0. 0004
+0004 -0.766 0. 0. -0.766 0. 0. 0005
+0005 3 4 0006
PBEAM 2 101 0.986 0.172 1.58 0. 0.009 0. 0007
+007 0008
+008 0.301 0.301 0009
+009 0.766 0. 0.766 0. 0010
Sequence 0005 specifies the components of the offset vectors Z A and Z B . The components
(-0.766, 0., 0.) define points A and B to be offset from the geometric grid points GA and GB by
0.766 inch in the negative Basic Y-direction. Since the element and basic coordinate systems
are coincident, the offset defined by 0010 makes points A' and B' coincident with GA and GB
as desired.
5.6.6.5 Tapered Beam with Shear Relief

Consider the tapered beam with heavy flange shown by Figure 1-20.
Figure 1-20 Tapered Beam
The beam is to be modeled using five Beam elements connecting the grid points as shown. The
element properties are presented by Table 5-1 where:
1.The local element x-axis is parallel to the Basic X axis
2.The cross sectional area is equal to the web area
124 Nastran Primer

3.The bending moment of inertia about the y-axis is calculated using only the flange
area
4.The web depth varies linearly along the length of the beam
5.The bending moment of inertia about the z-axis is equal to 1.
Table 5-1 Area Properties for Tapered Beam
Element No. Area Properties

AA (in2) IyyA (in4) IzzA (in4)
1 2. 240. 1.
2 3. 540. 1.
3 4. 960. 1.
4 5. 1500. 1.
5 6. 2160. 1.
The shear factors which account for the effect of beam taper can be determined by
considering the idealization of the Beam element shown byFigure 1-21
Pz1 P1
z
Px1
MA MB
A B
x hB
hA
V
V
α
Px2
Pz2 P2
L
Figure 1-21 Tapered Beam Element
The shear, QB, at cross section B is found to be
QB = V + sin α (P1 - P2) (Eq. 5-1)
where P1 and P2 are the forces in upper and lower flanges, respectively. The flange forces are
related to the moment, MB, by the equilibrium of moments so that
2M B
P 2 – P 1 = ------------------
- (Eq. 5-2)
h B cos α
The substitution of (Eq. 5-2) into(Eq. 5-1) then gives
Nastran Primer 125

5
2M B tan α
Q B = V – ----------------------- (Eq. 5-3)
hB
The transverse shear in the element y and z axes can then be written as
Sy
Q y = V y – ----- M z
l
(Eq. 5-4)
Sz
Q z = V z – ----- M y
l
where Sy and Sz are the average shear relief coefficients for taper. The shear coefficient, Sz, is
given by:
2 ( hA – hB )
S z = -------------------------
- (Eq. 5-5)
( hA + hB )
The values for the stress relief factors for the tapered beam shown by Figure 1-20, calculated
using (Eq. 5-5), are presented by Table 5-2 which can be generated in the Patran model using
the Element Properties form shown below
.
Table 5-2 Shear Coefficients for Taped Beam Example
Web Depth Shear Relief

Element Number
hA hB Coefficient Sz
1 20. 10. 0.667

2 30. 20. 0.400
3 40. 30. 0.286
4 50. 40. 0.222
5 60. 50. 0.182
The tapered beam presented by Figure 1-20 could be modeled using the following data
statements
1 2 3 4 5 6 7 8 9 10
GRID 1 0 100. 0. 0. 0 0001
GRID 2 0 80. 0. 0. 0 0002
GRID 3 0 60. 0. 0. 0 0003
GRID 4 0 40. 0. 0. 0 0004
GRID 5 0 20. 0. 0. 0 0005
GRID 6 0 0. 0. 0. 0 0006
CBEAM 1 1 2 1 0. 1. 0. 0007
CBEAM 2 2 3 2 0. 1. 0. 0008
CBEAM 3 3 4 3 0. 1. 0. 0009
CBEAM 4 4 5 4 0. 1. 0. 0010
CBEAM 5 5 6 5 0. 1. 0. 0011
PBEAM 1 101 2. 240. 1. 0012
126 Nastran Primer

Curved Beam Element
1 2 3 4 5 6 7 8 9 10
+012 NO 1. 1. 60. 1. 0013
+013 0. 0. 0. 0.667 0. 0014
PBEAM 2 101 3. 540. 1. 0015
+015 NO 1. 2. 240. 1. 0016
+016 0. 0. 0. 0.400 0. 0017
PBEAM 3 101 4. 960. 1. 0018
+018 NO 1. 3. 540. 1. 0019
+019 0. 0. 0. 0.286 0. 0020
PBEAM 4 101 5. 1500. 1. 0021
+021 NO 1. 4. 960. 1. 0022
+022 0. 0. 0. 0.222 0. 0023
PBEAM 5 101 6. 2160. 1. 0024
+024 NO 1. 5. 1500. 1. 0025
+025 0. 0. 0. 0.182 0. 0026
where
1.The grid point sequence and orientation vector defined on CBEAM Bulk Data
statements define a element coordinate system in same sense as the Basic
coordinate system shown byFigure 1-20.
2.Since the properties for each element are different, each element connection must
reference a different PBEAM.
3.The second optional continuation as shown by Bulk Data Image 1-4 is not required
because SOPT=NO
5.7 Curved Beam Element

The cureved beam element in MSC Nastran is called the BEND and is decribed below.
The comparable element is Legacy Nastran is called the ELBOW and is described in [10].
Nastran Primer 127

5
Curved Beam Element
A curved beam element, which is convenient for modeling piping systems, is shown by
Figure 1-22.
Arc of the Neutral Axis
Arc of Geometric Centers
uz B
∆N
uθ
ur
θz
θθ
θr A
GB
ZC Center of
GA Curvature
RC
RB
O′
Figure 1-22 Curved Pipe Element - CBEND
The NASTRAN BEND element has the following characteristics

1.The area properties and radius of curvature are constant.
2.The element coordinate axes (r, z) must coincide with the principal axes of inertia
for the cross section.
3.Flexibility and stress intensification factors can be specified in several ways.
4.Bending about and transverse shear in the direction of two perpendicular axes as
well as extension and torsion.
5.Only the consistent mass matrix can be calculated for structural or nonstructural
mass. The arc of geometric centers can be offset from the arc joining the geometric
grid points. The offset is the same at both ends.
5.7.1 Defining The BEND Element

The Bulk Data for the connective and properties of the BEND element are shown by Bulk
Data Image 1-5.
1 2 3 4 5 6 7 8 9 10
CBEND EID PID GA GB V1, G0 V2 V3 GEOM
128 Nastran Primer

Curved Beam Element
1 2 3 4 5 6 7 8 9 10
PBEND PID MID A Izz Irr J RB θB
Cr Cz Dr Dz Er Ez Fr Fz
Kr Kz NSM RC ZC ∆N
Alternate form of PBEND for Elbows and Curved Pipe
1 2 3 4 5 6 7 8 9 10
PBEND PID MID FSI RM T P RB θB
NSM RC ZC
where
Field Description
EID Element identification number, integer
PID Property identification number of a PBEND data statements, integer
GA, GB Connected grid identification numbers, integer
Identification number of third grid point used to define the element local
G0
coordinate system as described below, integer.
Components of the oreintation vector used to define the element local
V1, V2, V3
coordinate system as described below, real.
Flag used to select the specify the option used for the element local
GEOM
coordinate system as described below, integer.
MID Material identification number, integer.
A Cross sectional area of element, real.
Izz Area moment of inertia about the element local z axis.
Irr Area moment of inertia about the element local r axis.
J Torsional constant, real.
RB Bend radius of the line of centroids, real.
θB Optional arc angle, in degrees, of the element, real
The r,z locations, in the element local coordinate system, of up to four
Ci, Di, Ei, Fi, i = r,z
points at which stresses will be calculated, real.
Kr, Kz Shear factors in the r and z directions, real.
NSM Nonstructural mass per unit length, real
Radial offset of the geometric centroid from the connected grid points,
Rc
real.
Offset of the geometric Centroid in the direction perpendicular to the
Zc
plane of the connected points and the orientation vector, real.
Radial offset of the neutral axis from the geometric centroid, positive
∆N
toward the center of curvature, real.
Flag selecting the flexibility and stress intensification factors,
FSI
integer=1,2,3.
RM Mean cross sectional radius of the curved pipe, real.
T Wall thickness of the pipe, real
P Internal pipe pressure, real.
Nastran Primer 129

5
Curved Beam Element
Bulk Data Image 1-5 Specification of BEND element
5.7.2 Bend Element Connectivity and Geometry

A curved beam or pipe element is specified by
1.Defining the connected grid points, GA and GB.
2.Defining the plane containing the arc from the two connected grid points.
3.Defining the curvature and center of curvature of the BEND element.
4.Defining the position of the arc of geometric centers and the arc of neutral axes
relative to the arc connecting the grid points.
In general the geometry of the BEND element cannot be completely defined by the data
on the CBEND statement. CBEND is sufficient only if the following conditions are satisfied.
1.The element is an elbow or curved pipe and is described by the alternate form of the
PBEND data statement. In this case the offset of the arc of the neutral axis is
calculated.
2.The arc of geometric centers coincide with the arc joining the grid points.
3.The radius and centers of curvatures are defined using the GEOM field by
• Specifying a line which contains the center of curvature, (GEOM = 1).
• Specifying a line which is tangent to the arc of the element, (GEOM = 2).
For all other cases the additional data required to specify the geometry of the BEND element
are defined on the PBEND as follows:
RB The radius of the arc of the geometric axis. (real)

θB The arc angle of the element. (real > 0.)
Rc The radial offset of the arc of geometric centers from the arc joining the
geometric grid points. (real)
Zc Offset of the axis of geometric centroids in a direction perpendicular to plane
containing points GA, GB and the vector, V . (real)
∆N Radial offset of the arc of the neutral axis from the geometric centroid, positive
toward the center of curvature. (real)
5.7.3 BEND Element Coordinate System

The element coordinate system for the BEND element that is shown by Figure 1-23 is
defined by one of four methods using the GEOM field on the CBEND statement. The GEOM
options define:
1.How the vector, V , is to be used in determining the orientation element coordinate
system.
130 Nastran Primer

Curved Beam Element
2.Whether radius of curvature or arc angle for the element are specified on an
associated PBEND statement.
center of
curvature
B
arc of geometric
GB centroids
z
θ
ZC
GA
A
RB RC r
Figure 1-23 Local Coordinate System for the BEND Element
The geometric configurations associated with the GEOM-options are shown by Figure 1-
24
GO
V
GA GA
GB GB
a) Specification of Center of Curvature GO

b) Specification of Tangent to Arc at
(GEOM = 1)
GA (GEOM = 2)
GA RB GB θB
A B GA
GB
RC
V
V
c) Specification of Radius of d) Specification of Arc Angle

Curvature GEOM = 3) (GEOM = 4)
Figure 1-24 Specification of Curvature for BEND Element
Nastran Primer 131

5
Curved Beam Element
where
GEOM Meaning
(integer > 0)
1 The center of curvature lies on a line which is coincident with the vector,
V , or the line connecting GA and G0.
2 The tangent to the arc of the element at GA is coincident with the vector,
V , or the line connecting GA and G0.
3 The BEND radius, RB, is specified on PBEND. The center of curvature
lies in a plane which is parallel to that containing the line connecting GA
and GB and the vector, V , and lies on the opposite side of the line AB
from GO or the vector, V .
4 The arc angle is specified on PBEND. The center of curvature is then
located as described above for GEOM = 3.
The positive sense of the element curvilinear coordinates system (r, θ, z) shown on Figure
1-24 depends on the GEOM option chosen by the analyst. The element z-axis is taken in the
sense of a vector, z , which is defined by the cross product AB × V for GEOM = 1 and
V × AB for all other GEOM options, where AB is the vector from GA to GB.
The components of the vector, V , can be defined in two ways on the CBEND statement
1.Three components of a free vector (V1, V2, V3) originating at GA and defined
relative to the displacement coordinate system at GA. (real)
2.The directed line segment from GA to a third grid point, G0. (integer)
5.7.4 Specifying the Arc of Geometric Centrolds

The offset of the arc of the axis of geometric centroids is specified with respect to the
element coordinate system by the dimensions Rc and Zc, using either form of the PBEND
statement as shown by Bulk Data Image 1-5. The element (r, θ, z) coordinate system is defined
in the offset plane as shown by Figure 1-22 where
1.The r-θ plane is parallel to the plane which contains the vector, V , and the line
joining the geometric grid points GA and GB.
2.The element angular coordinate, θ, has its origin at end A of the curved beam or
pipe element.
5.7.5 BEND Element Properties

The BEND element properties are defined using one of the two forms for the PBEND
statement shown by Figure 1-22. The first of these two alternate forms can be used to define
the properties of a curved beam element having an arbitrary cross section. The second can be
used to define elbow or pipe elements having circular cross sections. For either form, fields
two and three of the PBEND statement are
PID Property identification number (integer > 0)
MID Identification number of an isotropic material (MAT1) (integer > 0)
5.7.5.1 Geometric Parameters for Curved Beam

A BEND element having an arbitrary cross section can be described by the first form of
the PBEND statement shown on Bulk Data Image 1-5 where the fields are defined as:
A Cross Sectional area (Real > 0.0)
132 Nastran Primer

Curved Beam Element
Izz, Irr Area moments about the element z and r axes, respectively (Real > 0.)
J Torsional constant (Real > 0.0)
The offset of the arc of the neutral axis from the arc of the axis of geometric centroids can
be specified by the parameter, ∆N (real or blank) which is positive towards the center of
curvature. If the field associated with ∆N is left blank then an approximate offset will be
calculated using the relation
I zz
∆N = ---------- (Eq. 5-1)
AR B
where the parameters in (Eq. 5-1) are defined on PBEND statement. The default value is
recommended if (R2B A)/Izz>15 in which case the calculated value is within five percent of the
exact expression for an element having either a circular or square cross section. Alternatively,
the user may choose to calculate the exact value for the offset using the following analytical
expression
RB
∆N = -------------------
- (Eq. 5-2)
2
RB A
1 + ----------
Z
where
2
r -
Z = ∫ 1---------------
+ r-
-----
dA (Eq. 5-3)
A
RB
and where the integration is taken over the cross section, and r is the radius to a point on the
cross section of the element.
Nastran Primer 133

5
Curved Beam Element
5.7.5.2 Circular Cross Section

An elbow or a pipe element having a circular cross section as shown by Figure 1-25
z
Center of
curvature
R
θ r
t
Zc
GB
θB GA
Rc
RB
Figure 1-25 Curved Pipe Element
can be described by using the alternate or second form of the PBEND data statement as shown
by Bulk Data Image 1-5. The following properties are then calculated:
1.Cross sectional area, area moments and torsional constant
2.Neutral axis offset
3.Shear factors
The cross section of the elbow or pipe element is completely defined by the following
fields on the alternate form of the PBEND data statement
R Mean radius of elbow or pipe. (Real > 0.)
T Wall thickness. (Real > 0.; R + T/2 < RB)
If T=0. then a solid cross-section of radius R is assumed.
5.7.5.3 Intensification Factors

The flexibility and stress intensification factors are defined by an integer value for the FSI
field on the alternate form of the PBEND as follows
FSI Meaning
134 Nastran Primer

Curved Beam Element
1 The flexibility factors which multiply the bending terms of the flexibility matrix
are set to unity. The stress intensification factor for bending about the z-axis is
set to unity and that associated with bending about the r-axis is set equal to Si
where
I zz 1 R B – ∆N
S i = ---------- --- + -----------------------------------
- (5-1)
AR B r i ∆N + ( R B + r i )
and ri is the radius of the ith stress recovery point relative to the element
coordinate system.
2 ASME Code Section III, NB-3687.2., NB-3685.2.,1977.
3 Empirical factors from Welding Research Council Bulletin 179 by Dodge and
Moore.
Finally, an internal pipe pressure can be specified by the P (Real) field on the alternate form
of the PBEND statement.
5.7.5.4 Shear Factors

The shear factors are automatically calculated if the alternate form of the PBEND data
statement is used. If the first form is used to describe a curved beam then the shear factors are
specified by the fields Kr and Kz (Real) which are terms in the relation KAG. A blank or zero
value specifies that shear flexibility is to be set to zero where the factors are calculated
automatically for the alternate form.
5.7.5.5 Stress Recovery Points

The position of four stress recovery points (C, D, E and F) can be specified by defining
the position of the four points in element coordinates. For example (Cr, Cz) define r and z
coordinates of point C relative to the θ-axis. If the alternate form of the PBEND is used then
the stress recovery points are automatically located as shown by Figure 1-26.
z
C
E r
Figure 1-26 Stress Recovery Points for Elbow and Pipe Element
Nastran Primer 135

5
Shear Panels
5.7.6 Element Forces and Stresses

The element forces for the BEND element are interpreted as internal forces at each end of
the beam. The positive sign convention for internal forces in the bend element are shown by
Figure 1-27 at an arbitrary angular position, θ.
Center of curvature VZ
Vr
MZ
A Mr
Fθ B
Mθ
GA
θ GB
RC
RB
Figure 1-27 Positive Element Forces for BEND Element
The following element forces are requested by an 'ELFORCE' Case Control Directive
1.Bending moments Mz and Mr, (identified as M1 and M2, respectively, in the output
file, and Shear forces Vr and Vz. (identified as V1 and V2, respectively, in the
printout)
2.The average axial force, Fθ.
3.The torque about the axis of geometric centroids, Mθ.
The following stresses are requested by a STRESS Case Control Directive.
1.Longitudinal stress at the four stress recovery points at each end.
2.Maximum and minimum longitudinal stress.
3.Margins of safety in tension and compression if stress limits are defined on the
MAT1 Bulk Data data statement.
The stresses are modified if the alternate form of the PBEND is used to account for internal
pressure and stress intensification due to curvature as indicated by the FSI field. Tensile stress
is considered positive.
5.8 Shear Panels

The shear panel is an essential element for modeling aerospace as well as other structures
characterized by very thin elastic sheets and stiffeners. This class of structures has been
traditionally modeled by a combination of ROD and SHEAR elements where the RODs
account for the extensional behavior of the stiffeners and the SHEARs account for the load
carried by the thin elastic sheets.
136 Nastran Primer

Shear Panels
5.8.1 SHEAR Definition

The shear panel is shown in Figure 1-28.
k4
F41 F43
k3
4
q3
k2 F32
k1
q4
q2 3 F34
F12
F21
1 q1 2
F23
F14
Figure 1-28 Shear Panel Showing Corner Forces and Shear Flows
where the points 1, 2, 3, and 4 refer to the order in which the connected grid points are specified
using the CSHEAR data statement shown by Bulk Data Image 1-6
1 2 3 4 5 6 7 8 9 10
CSHEAR EID PID G1 G2 G3 G4
PSHEAR PID MID T NSM
or
PSHEAR PID MID T NSM F1 F2
Bulk Data Image 1-6 Shear Panel
5.8.2 Description of Shear Panel Input

The Bulk Data for the shear panel description consists of a CSHEAR statement having a
unique element identification number that references an appropriate PSHEAR. The shear panel
connects four grid points which must be ordered consecutively around the perimeter of the
element, as shown byFigure 1-28.
5.8.3 Element Properties

The element properties are defined on the PSHEAR Bulk Data data statement. The
material property for the element is defined by reference to a material specification, which
must be a MAT1-type data statement. The other element properties are
T Shear panel thickness. (real, ≠ 0.)

NSM Nonstructural mass per unit area. (real)
Nastran Primer 137

5
Shell Elements
The alternate form of the PSHEAR allows the user to define equivalent axial RODs on
the periphery of the element by means of the fields F1 and F2 which are area effectiveness
parameters. Field F1 defines RODs along edges 1-2 and 3-4 while F2 is associated with edges
1-4 and 2-3. Real entries in these fields will cause the generation of an effective extensional
area as follows
Fi < 1.01 Set effective ROD areas equal to 1/2Fi (Twi) where w, is the effective width
in the i direction and T is the panel thickness.
Fi > 1.01 Set effective ROD areas equal to 1/2 FiT2.
5.8.4 Recovery of Forces and Stresses

The element forces may be output by the Case Control Directive
ELFORCE = <set>
where <set> is a set of elements defined in Case Control. The shear panel forces consist of the
force applied to the element at the node points, kick forces at the corners in a direction normal
to the plane formed by the two adjacent sides, and shear flows (force per unit length) along the
four edges as shown by Figure 1-28
The element stresses consist of the average of the shear stress calculated at the corners
and the maximum stress. The shear stresses as well as margin of safety may be output by the
Case Control Directive
STRESS = <set>
5.9 Shell Elements

The NASTRAN element library includes both flat and curved shell elements which
support membrane and bending behavior. The form of the element specification allows the
analyst to control the generation of stiffness coefficients for
• Membrane-only behavior
• Bending-only behavior
• Bending and membrane behavior
• Materially-coupled bending-membrane behavior
The modern shell elements in the NASTRAN include
TRIA3 A three node isoparametric flat element
QUAD4 A four node isoparametric flat element
TRIA6 A six node triangular isoparametric curved thin shell element (MSC and NX
only)
QUAD8 An eight node isoparametric quadrilateral curved thin shell element (MSC
and NX only)
where the four grid points associated with a QUAD4 shell element need not lie in a plane.
138 Nastran Primer

Shell Elements
5.9.1 Flat Shell Elements - OUAD4 and TRIA3

The triangular and quadrilateral flat shell elements are shown by Figure 1-29a and Figure
1-29b, respectively, and are defined using the data statements shown on Bulk Data Image 6-7.
z
z
y
G4
uz
θy G1 y uz
uy
G1 G3 O
β θy uy
θx θm
θm
ux α G3
xm θx
γ α
xm ux
G2
G2 x
x
a) Triangular Element - TRIA3 b) Quadrilateral Element - QUAD4
Figure 1-29 Isoparametric Flat Thin Shell Elements
Nastran Primer 139

5
Shell Elements
5.9.2 Defining Connectivity and Properites
5.9.2.1 Triangular Element Connectivity -- TRIA3
1 2 3 4 5 6 7 8 9 10
θm or
CTRIA3 EID PID G1 G2 G3 ZOFFS
MCID
T1 T2 T3
5.9.2.2 Quadrilateral Element - Connectivity- QUAD4
1 2 3 4 5 6 7 8 9 10
θm or
CQUAD4 EID PID G1 G2 G3 G4 ZOOFS
MCID
T1 T2 T3 T4
5.9.2.3 Common Propery - PSHEL

The mechanical behavior of the shell elements is defined by a common PSHELL property
Bulk Data statement that is shown below. In addition, material properties for composiste lay-
ups can be defined by PCOMP Layered Composite (p. 258), as described in Laminate
Properties (p. 255) in Chapter 9.
1 2 3 4 5 6 7 8 9 10
PSHELL PID MID1 T MID2 RI MID3 RT NSM
Z1 Z2 MID4
Bulk Data Image 1-7 Specification of Flat Shell Elements - QUAD4 and TRIA3
The general capabilities of the TRIA3 and QUAD4 are

1.Thickness may vary isoparametrically over the surface of the element.
2.Either coupled and lumped mass formulations can be requested; however, the
coupled mass matrix does not include inertias associated with rotational degrees of
freedom.
3.The anisotropic elastic coefficients associated with in-plane, bending, transverse
shear, and coupling between in-plane and bending behavior may be defined
independently. These elements can thus be used to model unbalanced composite
materials or plates offset from node points.
5.9.3 Element Connectivity

The triangular and quadrilateral elements connect three and four grid points, respectively,
as shown by Figure 1-29a and Figure 1-29b. The grid point numbers associated with the fields
G1, G2, G3, (and G4) must be unique and must be ordered consecutively around the perimeter.
All interior angles for the QUAD4 must be less than 180 degrees.
The flat shell elements connect the five degrees of freedom in element coordinates
including the two in-plane displacements, ux and uy, the normal displacement, uz, and the two
bending rotations, θx and θy. There is no rotational degree of freedom about the normal for
these elements so that there will be a row and column of zeros in the six-by-six stiffness matrix
140 Nastran Primer

Shell Elements
associated with each connected grid point. The assembled system stiffness matrix, [Kgg], will
thus be singular unless the normal rotation is connected by another type of element such as a
BAR or if the flat plates joining at the grid point do not lie in a plane so that there is a
geometry-related stiffness.
5.9.4 Element Coordinate System

The definition of the local element coordinate system is of interest to the analyst in
determining the direction and positive sense of the normal to the surface. The positive normal
direction then defines the sense of positive normal pressure on the PLOAD2 and PLOAD4
Bulk Data and the positive element face for stress recovery. The coordinate systems for the
triangular and quadrilateral flat shell elements are shown by Figure 1-29a and Figure 1-29b,
respectively.
The element coordinate system for the TRIA3 is defined such that
1.The x-axis coincides with the line drawn from grid point G1 to G2.
2.The three grid points define the x-y plane
3.The z-direction is normal to the plane containing the three grid points and is positive
in the sense of the right hand rule when applied to the order of grid point
specification.
4.The y-direction is found by: y = z × x
The local coordinate system for the QUAD4 element, shown byFigure 1-29b, has the
desirable properties of
1.Being uniquely defined, independent of the starting grid point number within even
multiples of 90 degrees.
2.Being continuous with respect to changes in element shape.
3.Being parallel to the edges for a rectangular element.
4.Having the x-axis nearly parallel to the longest bisector of opposite sides in the
general case.
The local z-axis is perpendicular to the element x-y plane and is positive in the sense of
the right hand rule applied to the grid point numbering scheme.
5.9.5 Material Coordinate

The shell elements support anisotropic materials. Since nonisotropic material properties
are defined relative to a set of material coordinates the analyst must specify the orientation of
one of the material axes relative to a reference axis in the element.
The material coordinate axis can be specified by either a material angle, θm, or by the a
material coordinate system which is identified by the MCID.
The field θm (Real), on the connection statements is the orientation angle, in degrees,
between the reference line between G1 and G2 and the material axis. If MCID is specifies then
the x-axis of the material coordinate system is determined by projecting the x-axis of the MCID
coordinate system on the surface of the element.
Nastran Primer 141

5
Shell Elements
5.9.6 Reference Surface Offset

Elements can be offset from the surface of the connected grid points by the ZOFFS field
of the connection statement. Other parameters such as material matrices and stress fiber
distances are given relative to the offset reference plane. A positive value of ZOFFS implies
that the element reference surface is offset a distance along the positive normal of the element.
If the reference surface is offset then bending and inplane forces are coupled and both MID1
and MID2 must be specified.
5.9.7 Element Properties - PSHELL

The element properties are defined using the PSHELL statement for all thin shell
elements. The fields on the property data statement for isoparametric shell elements are
defined as follows
MID1 Material identification number for membrane behavior (integer >0 or blank)
T Membrane Thickness (Real)
MID2 Material identification number for bending behavior (Integer > 0 or blank)
RI Normalized bending inertia per unit length. RI = I/Io where Io = T3/12 and I is
the actual bending moment of inertia per unit length of the cross-section. (Real
or blank, default = 1.)
MID3 Material identification number for transverse shear behavior. (integer > 0 or
blank)
RT Normalized shear thickness, R T = T S ⁄ T , where TS is the shear thickness and
T is the membrane thickness. (Real or blank, default = 0.833333)
NSM Nonstructural mass per unit area (Real)
Z1,Z2 Stress recovery distances for bending (Real, default Z1 = Z2 = T/2 thickness)
MID4 Material identification number for membrane-bending coupling (integer > 0 or
blank)
5.9.8 Specifying Element Behavior

The shell elements include in-plane, bending and transverse shear behavior. The presence
or absence of the corresponding entry of a material identification number in the appropriate
field then specifies whether of not the associated behavior is to be included in the element
formulation. If an entry is present, its value is the material set to be used in determining the
material properties associated with the behavior. The behavior associated with each of the MID
fields is:
Material Field Type of Behavior
MID1 Membrane (MAT1 or MAT2)
MID2 Bending (MAT1 or MAT2)
MID3 Transverse shear (MAT1 or MAT2)
MID4 Bending-Membrane Coupling (MAT2 only)
Material fields are not exclusive so entries for MID1 and MID2, and MID4 would specify
that coupled membrane-bending behavior without transverse shear is desired, while entries in
only MID1 would specify membrane-only behavior.
142 Nastran Primer

Shell Elements
5.9.9 Element Constitutive Relations

In order to clarify what is meant by the various types of behavior and their associated
material identification numbers it is worthwhile to consider the constitutive relations that are
supported by the NASTRAN shell elements. The general form of the constitutive relation is
2  o o 
 {N}  T [ G1 ] T [ G4 ] 0  { ε } – { εI } 
   
 { M }  = T2 [ G ]T I [ G ] 0  { χo } – { χo }  (Eq. 5-1)
  4 2  I 
 {Q}  0 0 TS [ G3 ]  o 
 {γ } 
o
where {N} and{M} are the in-plane force and moment resultants, respectively, { ε }}are
o o o
reference surface strains, { χ } are reference surface curvatures, { ε I } , { χ I } are initial strains
o
and curvatures, and { Q } and { γ } are defined as
 
 Qx 
{Q} =   (Eq. 5-2)
 Qy 
 
 o 
o  γ 
{ γ } =  xz  (Eq. 5-3)
 γo 
 yz 
o o
The quantities Qx and Qy are the transverse shear forces and γ xz , are γ yz the
corresponding transverse shear strains of the reference surface. The cross section properties T,
I, and Ts are the membrane thickness, second area moment for the cross section and shear
thickness respectively, referred to the reference surface. The matrices [G1], [G2], [G3], and
[G4] are elastic coefficients defined by appropriate material statements having material IDs
MID1, MID2, MID3, and MID4, respectively.
The form of the relation (6-33) allows for the coupling of bending and membrane action
through the material coupling terms T2 [G4]. The coupling terms do not exist for symmetric
cross-sections when the middle surface is taken as the reference surface. However, they have
been included in the element formulation to allow modeling composite materials having
unsymmetrical layups and offset plates. In the usual case of a solid homogeneous plate the field
associated with material coupling, MID4, would be blank and the cross sectional properties
would be evaluated with respect to the middle surface.
5.9.10 Solid Homogeneous Symmetric Cross-Section

The desired behavior is specified by entries in the material identification fields, as
appropriate. We can define, for example, uncoupled membrane or bending behavior and
include transverse shear effects by using the MID1, MID2, and MID3 fields. Since there is no
material coupling of bending and membrane behavior for a solid homogeneous symmetric
cross section if the middle surface is used as the reference surface, it follows that MID4 must
be blank.
The membrane thickness can vary isoparametrically over the surface of the element. The
fields T1, T2, T3, (and T4) on the CTRIA3 and CQUAD4 statements define the membrane
thickness at the associated grid points. The T-field on the PSHELL defines the default value
for membrane thickness.
Nastran Primer 143

5
Shell Elements
The bending moment of inertia per unit length of the edge and the transverse shear
thickness are defined in a manner such that the default values are correct for solid cross
sections. The bending moment of inertia is specified as the ratio, RI, which is equal to 1.0 for
a solid homogeneous cross section. If bending behavior is requested (i.e. an entry in the
MID2-field which points to a material data statement) and if the plate has a solid homogeneous
cross section then the RI-field can be left blank since the default is 1.0.
Transverse shear behavior is specified by an entry in the MID3 field whose value is the
material ID of a material data statement defining the material shear coefficients. If the ratio RT
is blank then the default value, RT = 0.833333, is taken. It a MAT2 data statement is used to
define transverse shear material coefficient only the G33 is used.
For example, a membrane having a uniform thickness of 0.5 in. could be defined by the
following property statement
1 2 3 4 5 6 7 8 9 10
PSHELL 1 10 0.5
where the material membrane behavior is defined by material set number 10. Similarly, an
element having membrane, bending and transverse shear behavior could be defined by the
following
1 2 3 4 5 6 7 8 9 10
PSHELL 1 10 0.5 10 10
where material set 10, which we suppose is a MAT1 statement, has been specified in the fields
associated with membrane, bending, and shear behavior and where the default values for the
fields Rl and RT have been taken.
5.9.11 Sandwich Type Cross-Section

A sandwich plate is constructed of a core and two face sheets as shown by Figure 1-30.
The core is typically a light-weight material such as metal or paper honeycomb, or wood,
whose purpose is to separate the facesheets in order to obtain a desired bending moment of
inertia. The core is assumed to exhibit only shear behavior while the face sheets provide
membrane and bending behavior. Generally the shear modulus of the core is an
order-of-magnitude smaller than the elastic constants for the face sheets so that appreciable
shear deformation can be expected.
reference surface
t1
z
h t2
zo
t3
Figure 1-30 Sandwich Plate
The constitutive relations for the sandwich plate are then formulated by noting that the
moment and force resultants are defined by (2-61) and (2-59), respectively. The components
of stress in these equations can then be written in terms of strains and curvatures of a reference
surface located at a distance zo from the bottom of the cross section as follows:
144 Nastran Primer

Shell Elements
{σ } = [E]({ε} – z{χ}) (Eq. 5-4)
where z is the distance from the reference surface. The substitution of (Eq. 5-4) into (Eq. 2-58)
and (Eq. 2-60)then gives the following expressions for moment and force resultants.
{ M } = – ∫ z [ E ] ( { ε } – z { χ } ) dz
o o
(Eq. 5-5)
∫ [E]({ε
o o
{N} = } – z { χ } ) dz (Eq. 5-6)
where the integration is taken through the thickness of the plate.

In general, the elastic modulae are piecewise constant functions of the shell depth. In the
present case the elastic coefficients are constant over the core depth, t2, and may be different
but constant over the cover plate thicknesses t1 and t3 The force resultants can then be found
o o
by piecewise integration provided that we are careful to define { ε } and { χ } to be the strains
and curvatures of the reference surface, which is defined at a distance z0 from the bottom of
the cross section. The expression for the force resultants for the sandwich plate shown by
Figure 1-30 then becomes
h–z
o
∫ [ E1 ] ( { ε
o o
{N} = } + z { χ } ) dz (Eq. 5-7)
h – zo – t1
h – zo – t1
∫ [ E2 ] ( { ε
o o
+ } + z { χ } ) dz
h – zo – t1 – t2
h – zo – t1 – t2
∫ [ E3 ] ( { ε
o o
+ } + z { χ } ) dz
h – zo – t1 – t2 – t3
For a symmetric cross section we have: t1 = t3 = t; [E1] = [E3] = [Ef]; t2 = c; and [E2] =
[Ec] so that the integration of (Eq. 5-7) leads to the following expression.
o 1 o
{ N } = { ε } [ 2t [ E f ] + c [ E c ] ] – --- { χ } ( 2th [ E f ] + ch [ E c ] ) ( 1 – 2a ) (Eq. 5-8)
2
where zo = ah and where a is a parameter that can be chosen by the analyst.

Inspection of the coefficient of the curvature term shows that the in-plane forces can be
made independent of the curvature if the parameter, a, is set equal to one-half. The strains and
curvatures are then associated with the reference surface located at h/2 which is the middle
surface for a symmetric layup. However, the NASTRAN shell elements allow for the coupling
terms so that the analyst has the freedom to choose any convenient reference surface.
The force and moment resultants associated with a reference surface at the middle of a
symmetric cross section can now be determined by using (Eq. 5-5) and (Eq. 5-6) and by noting
that the strains and curvatures are defined at the middle surface. The coupling term in the
moment relation then disappears so that the moment and force resultants are the moment given
by:
Nastran Primer 145

5
Shell Elements
o o
{ N } = 2t [ E f ] { ε } = T [ G 1 ] { ε } (Eq. 5-9)
3
{ M } = [ E f ]  c t + ct + t---- { χ } = I [ G 2 ] { χ }
2 2 o o
(Eq. 5-10)
 3
where we have assumed that the core carries no membrane forces.

As a specific example consider the case where t = 0.25 in. and c = 0.50 in. The face sheets
are isotropic aluminum and the core has a shear modulus of 300,000 lb./in.2. The element
properties could then be defined using the following data statements:
1 2 3 4 5 6 7 8 9 10
PSHELL 1 101 0.5 101 9.5 102 1.0
MAT1 101 10.e6 0.3
MAT1 102 3.+5 0.3
where the bending moment of inertia and shear thickness have been scaled appropriately by
the membrane thickness.
The subject of unsymmetrical sandwich construction and composite materials is beyond
the scope of the present text. The interested reader should consult Refs. 5-8 and 5-9 for more
details.
5.9.12 Effect of Warping

The QUAD4 element connects four grid points that need not lie in a plane. The element
matrix for a flat element is modified in this case by appropriate pre- and post-multiplications
in order to satisfy rigid body properties. The procedure which is used is described in Non-
Planar Nodes for a Flat Shell Element (p. 86) and is only satisfactory for small deviations from
flatness and accounts for the unbalanced nodal forces and moments.
5.9.13 Elastic Stiffness Matrices

Both the TRIA3 and the QUAD4 use the same shape functions used to interpolate
displacements to transform the element geometry from physical to parametric space. The basic
theory is straight-forward and follows the isoparametric development outlined in Chapter 4.
The formulation of the shear stiffness for the QUAD4 element is handled quite differently
than the extensional stiffness. The extensional strains are calculated at the usual Gauss points
but the shear strains are evaluated at special points as described by MacNeal [4-8] for the
QUAD4 element. A reduced integration scheme is then employed for the shear strains that
generally results in improved element performance over a broad of element geometry and
loads.
5.9.14 Mass Matrix

Lumped and consistent mass matrices are calculated at user option. The lumped mass is
the default; the consistent mass is requested using the COUPMASS parameter which is
described in Appendix B. Mass coefficients are calculated for only translational degrees of
freedom for both the lumped and consistent formulations.
146 Nastran Primer

Shell Elements
The structural mass is calculated using the density associated with the membrane material
data statement. Therefore, even if bending-only behavior is desired for dynamics, the
membrane properties must be specified in order to generate the structural mass matrix. The
nonstructural mass is calculated using the value for NSM on the PSHELL. If bending-only
behavior is desired the mass matrix for transverse displacements can be calculated by using
NSM without specifying membrane action.
5.9.15 Stress, Strain and Element Force Recovery

Components of stress, strain and element forces can be recovered using STRESS,
STRAIN and ELFORCE Case Control directives. Strain and stress components are computed
at distances z1 and z2 from the reference plane along a line perpendicular to the reference plane
at the centroid of the element. The stress is given by
i z o
{ σ } = – ---i { M } + { σ } + { ∆σ } (Eq. 5-11)
I
where the i takes the values 1 and 2 for the bottom and top surface of the element, respectively
and where the membrane stress is given by:
o o o o o
{ σ } = [ G1 ] ( { ε } – { ε1 } ) + T [ G4 ] ( { χ } – { χ1 } ) (Eq. 5-12)
The thermal stress, { ∆σ } , results from the difference between the outer fiber temperature and
a linear fit between the temperatures at the plate surfaces; and the strains and curvatures of the
reference surface are determined using the grid point displacements.
Stresses and strains are tensor quantities which are calculated in the element local
coordinate system. We recall from Chapter 3 that the stress components for the two
dimensional flat shell element are:
T
{ σ } = σ xx σ xx τ xy
In addition to these quantities, the following quantities are calculated for the element:
1.Principal stresses and principal stress directions
2.Von Mises stress
The interpretation of stress (and strain) quantities is complicated since
1.The calculated stress and strain components are not continuous between elements
2.The stress and strain components are most accurate at the Gauss points used for
calculating the strain matrix, [B], but are actually determine at the element centroid.
Furthermore, the orientation of the element coordinate system changes from element to
element. For simple structures, such as a flat rectangular plate, it is possible to align the
element coordinate systems with the Basic coordinate system by using a consistent ordering
procedure for the element connection list. But, for an arbitrary component which may be
curved, the element coordinate system for each element will have a different orientation
relative to a Basic coordinate system.
In order to interpret the components of stress at the grid points they must be averaged in
some way. However, since the stress components for each element connected to grid point is
defined in the local element coordinate system, the components of stress must first be
transformed to a common coordinate system prior to stress averaging.
Nastran Primer 147

5
Curved Shell Elements
In addition to the stresses, the force resultants {M}, {N}, and {Q} are calculated at the
Grid points. The user may request stress and force output by the STRESS and ELFORCE Case
Control Directives, respectively.
5.10 Curved Shell Elements

The QUAD8 and TRIA6 curved shell elements have all of the general modeling
capability associated with the flat shell elements. including:
1.lsoparametric thickness variation over the element
2.Element thermal field
3.Anisotropictemperature-dependent material properties
4.Elastic coupling of membrane and bending behavior
The major differences between the curved and the flat shell elements are
1.The shape functions used for the QUAD8 and TRIA6 for displacements and
geometry are of a higher order than those used for the QUAD4 and TRIA3.
2.The QUAD8 and TRIA6 may connect up to eight and six node points, respectively.
The TRIA6 and QUAD8 elements are shown by Figure 1-31a and Figure 1-31b,
respectively, where it can be seen that the elements connect all degrees of freedom except the
normal rotation in element coordinates at the grid point.
z z
G3
G3 G7
G6 zl
ym xm G4
xm
G6 θ G5 ym θ η = const. yl
η = const. G2 θ
zl yl
G8
θ xl
ξ = const. yl xl
θ yl G5
G2
G1
G4 θ xl G1
y
xl y
a) Six-Node Triangle x b) Eight-Node Quadrilateral
x
Figure 1-31 Isoparametric Curved Thin Shell Elements
148 Nastran Primer

Curved Shell Elements
5.10.1 TRIA6 and QUAD8 Connectivity

The connectivity for the elements is defined by the CQUAD8 and the CTRIA6 Bulk Data
which are shown in Bulk Data Image 1-8. The PSHELL property data statement which is
shown on Bulk Data Image 1-7 and which was described in the previous section is used to
define the element properties for the QUAD8 and TRIA6 as well as the TRIA3 and QUAD4.
In addition layered compoite properties can be defined using PCOMP-type Bulk Data as
decribed by Bulk Data Image 9-7
1 2 3 4 5 6 7 8 9 10
CTRIA6 EID PID G1 G2 G3 G4 G5 G6
θm or
ZOFFS T1 T2 T3
MCID
CQUAD8 EID PID G1 G2 G3 G4 G5 G6

θm or
G7 G8 T1 T2 T3 T4 ZOFFS
MCID
Bulk Data Image 1-8 TRIA6 and QUAD8 Thin Shell Elements

The TRIA6 and QUAD8 element connect up to six and eight nodes as shown by Figure
1-31a and Figure 1-31b, respectively. The mid side nodes may be selectively deleted so that
the TRIA6 can connect from 3 to 6 nodes and the QUAD8 can connect from 4 to 8 nodes.
The connected node points are defined by G1 through G6 on the CTRIA6 and G1 through
G8 on the CQUAD8. The grid point entries must be ordered as shown.
All of the corner node points must be defined. The edge nodes should be at or near the
center of the edges; however, any of the edge nodes can be deleted by leaving the appropriate
field blank. For example, a blank in the G6-field of the CTRIA6 data statement would imply
that there is no point on the edge connecting nodes G1 and G3.
5.10.3 Element Coordinate Systems

The coordinate system for the curved shell elements is taken to be an orthogonal
curvilinear system in geometric space as shown by Figure 1-31a and Figure 1-31b for the
TRIA6 and QUAD8 elements, respectively. The normal coordinate is taken to be
perpendicular to the surface with a positive sense determined by applying the right hand rule
to the corner node number sequence.
5.10.4 Material Orientation

The orientation of the material coordinate system relative to the element coordinates is
defined by specifying either θm or the MCID. The angle θ (real) in degrees is taken to be
constant over the element and is used at each Gauss integration point to define the orientation
of the xm material axis as a line in the plane tangent to the surface of the element and at an angle
θ to the intrinsic element coordinate η = constant as shown. If MCID is specified the x-axis of
the material coordinate system is determined by projecting the x-axis of the MCID coordinate
system onto the surface of the element.
Nastran Primer 149

5
Solid Elements
5.10.5 Element Thickness

The membrane thickness can be defined as a constant over the element using the T-field
on the PSHELL data statement; or to vary over the surface by defining the thickness at the
corner node points on the appropriate connection statement using the fields T1, T2, T3, (and
T4) (Real). If the thickness fields on the connection statement are left blank the thickness on
the PSHELL is used.
5.10.6 Specifying Element Behavior

The TRIA6 and QUAD8 are capable of selectively representing membrane, bending, or
coupled membrane-bending behavior depending on the entries in the MID-type fields of the
PSHELL, Bulk Data Image 1-7, and by PCOMP, PCOMP1 and PCOMP2 Bulk Data Image
9-7. The resulting constitutive equations are then defined by (Eq. 5-1)
5.10.7 Element Stiffness Matrices

The TRIA6 and QUAD8 are isoparametric elements using quadratic interpolation for
displacement as well as mapping the geometry to parametric space. The stiffness and mass
matrices are then formulated in parametric space using the techniques described in Chapter 4.
Both the TRIA6 and QUAD8 use reduced integration of the shear energy. As noted in
Chapter 4, reduced integration generally results in a more flexible element having superior
convergence characteristics compared the standard approach of evaluating all strain
components at the appropriate Gauss integration points. The resulting element is non
conformable in the sense that the displacement fields do not match along common element
edges and the behavior does not converge monotonically to theoretical solution, but it does
converge.
5.10.8 Element Mass Matrices

The lumped mass formulation is the default for both the TRIA6 and QUAD8 but the
analyst can request the consistent mass by means of the COUPMASS parameter described in
Chapter 12. Mass coefficients for rotational degrees of freedom are not calculated in either case
so that the elemental mass matrix associated with bending behavior is singular.
5.10.9 Stress, Strain and Element Force Recovery

The element strains and curvatures, and therefore the stress components, are calculated in
element coordinates as described for the flat shell elements in the previous section. The
components of stress are computed using (Eq. 5-11) at distances z1 and z2 from the reference
surface which are defined on the PSHELL statement. The stresses are recovered at the vertices
and the centroid for each element..
5.11 Solid Elements

The NASTRAN solid elements include: a four-sided element called the TETRA; a
five-sided element called the PENTA; and, a six sided element called the HEXA. These
elements are shown by Figure 1-32, Figure 1-33 and Figure 1-34, respectively. Each of these
elements can include optional edge nodes. However, the 4-node TETRA and the 6-node
PENTA result in very stiff elements and should not be used for modeling. The 10-node
TETRA and the 15-node PENTA perform quite well, especially compared to their lower order
counterparts. The 8-node HEXA performs well and can be used for general modeling
situations calling for elements capable of representing a three dimensional stress field.
These elements were developed to
150 Nastran Primer

Solid Elements
1.Support Anisotropic material

2.Improve stress output
3.Improve performance in representation of curved shell behavior
4.Improve treatment of differential stiffness
G4
G8
G10
G9
G3
G7
G1
G6
G5
G2
Figure 1-32 Four Sided TETRA Element
6
15
4
14
13
5 12
10
11
3
9
1
8
7
2
Nastran Primer 151

5
Solid Elements
Figure 1-33 Five-Sided Isoparametric Solid Element - PENTA
8
20
19
5
18 7
16
17
6 15
13
14 4
12 11
1 3
10
9
Figure 1-34 Six-Sided Isoparametric Element - HEXA
5.11.1 Defining Solid Elements

The Bulk data statements for defining solid elements representing three dimensional
stress states are as follows:
1 2 3 4 5 6 7 8 9 10
CTETRA EID PID G1 G2 G3 G4 G5 G6
G7 G8 G9 G10
1 2 3 4 5 6 7 8 9 10
CPENTA EID PID G1 G2 G3 G4 G5 G6
G7 G8 G9 G10 G11 G12 G13 G14
G15
1 2 3 4 5 6 7 8 9 10
CHEXA EID PID G1 G2 G3 G4 G5 G6
G7 G8 G9 G10 G11 G12 G13 G14
G15 G16 G17 G18 G19 G20 G13 G14
1 2 3 4 5 6 7 8 9 10
PSOLID PID MID CORDM IN STRESS ISOP FCTN
Bulk Data Image 1-9 Isoparametric Solids; PENTA and HEXA
152 Nastran Primer

Solid Elements

The TETRA, PENTA and HEXA elements connect the three displacement degrees of
freedom at each node point on the element. The node points to be connected are identified by:
G1 through G10 on the CTETRA; G1 through G15 on the CPENTA; and G1 through G20 on the
CHEXA data statements shown by Bulk Data Image 1-9. The property field, PID, refers to a
common PSOLID data statement.
5.11.3 The TETRA Element

The TETRA element four vertex nodes and six edge nodes. The node number sequence
shown on Figure 1-32 must be preserved, i.e. nodes G1 through G4 must define the vertices.
The edge nodes, G5 through G10, are optional and positional. That is: G5 is the node on the
edge between G1 and G2, G6 is the node on the edge between G2 and G3, etc. If the element is
to be modeled without the edge node, G5, then the field associated with that node is left blank.
The continuation is not required if all of the edge nodes are omitted.
The 4-node TETRA element performs poorly in almost all modeling situations and is not
recommended. However, the 10-node TETRA performs well and is widely used.
5.11.4 The PENTA Element

The PENTA element has six vertex nodes and nine edge nodes. The node number
sequence shown on Figure 1-33 must be preserved, i.e. nodes G1 through G3 must define the
vertices of a triangular face; and, nodes G4 through G6 must then define the opposite triangular
face with G4, G5, and G6 on the edges connecting G1, G2, and G3, respectively.
The edge nodes G7 through G15 are optional and positional. That is: G7 is the node on the
edge between G1 and G2, G8 is the node on the edge between G2 and G3, etc. If the element is
to be modeled without an edge node, G7 for example, then the field associated with that node,
field 2 on the first continuation of the CPENTA statement, contains no entry and is thus blank.
Continuations are not required if all edge nodes are omitted.
5.11.5 HEXA Element

The HEXA element has eight vertex nodes and twelve edge nodes. The node number
sequence shown on Figure 1-34 must be preserved. Nodes G1 through G4 must define one face
and nodes G5 through G8 must then define the opposite face with G5 lying on the edge
connecting G1 etc.
The edge nodes G9 through G20 are optional and positional; and, any or all may be
defined. As in the case of the TETRA and PENTA elements, the edge nodes are positionally
defined with node G9 on the line segment connecting G1 and G2 and G17 on the line segment
between G5 and G6, for example. The absence of an entry for G9, field 4 of the first
continuation of the CHEXA, implies there is no node on the edge connecting G1 and G2. If no
edge nodes are defined then the second continuation is not required.
5.11.6 Properties of Solid Element

The properties of the TETRA, PENTA and HEXA solid elements are defined by a
common property data statement, the PSOLID, as shown on Bulk Data Image 6-11. The fields
of PSOLID are defined as:
Field Description
PID The property ID, integer > 0.
MID The material set ID, integer > 0.
Nastran Primer 153

5
Solid Elements
Field Description
The set ID of a coordinate system to be used as the reference system for a
CORDM
non-isotropic material as described below, integer, blank or > 0.
Specifies the integration procedure to be used, integer or character as
IN
described below.
Specifies location for stress output, integer or character as described
STRESS
below.
ISOP Integration scheme, integer, character or blank as described below.
FCTN Fluid element flag, blank for structural solution.
5.11.6.1 Material Coordinate System

The material identification refers to either an isotropic or an anisotropic material defined
by a MAT1 or MAT9 material statement, respectively. If the material is anisotropic then the
orientation of the material axis relative to element coordinates must be defined by an entry in
the CORDM field as follows
CORDM Meaning
Blank or - 1 Align material coordinates with element coordinates.
0 Material axes are aligned with Basic coordinates.
>0 Material axes are aligned with coordinates defined by coordinate
identification number appearing in the field.
5.11.6.2 Gauss Integration Procedure

The IN and ISOP fields provide control of
• Gauss integration order
• Use of Bubble functions
• Use of reduced integration
Two sets of integration points are provided for each element. The user can specify either
of the sets by including an integer or character value for the IN field as shown by Table 5-1.
Table 5-1 Specifying Gauss Integration Order
Element Type
IN
HEXA PENTA Tetra
2 or TWO 8 points 6 points 1 point
3 or THREE 27 points 21 points 5 point
If the elements have no midside nodes, IN can also be set equal to BUBBLE and ISOP =
REDUCED, which are the defaults, which minimize locking due to shear and Poisson’s ratio
effects. The defaults are recommended.
154 Nastran Primer

Solid Elements
5.11.6.3 Stress Locations

Stress components are calculated at the vertex nodes by default. Alternatively, the stress
components can be calculated at the Gauss integration points for elements having no edge
nodes by setting STRESS to either ‘GAUSS’ or 1. The locations of the Gauss points must, of
course, be known in order to process the stress components. These locations are described in
Section 15.3 of the NASTRAN Reference Manual.
5.11.7 Hexa Element Coordinate System

The orientation of the element coordinate system is of importance to the analyst if
anisotropic material axes are to be aligned with the element axes. The element coordinates for
the HEXA element are chosen using the following procedure.
1.Let a 1 , a 2 , and a 3 be the directed line segments joining the centroids of faces 1
and 2, faces 3 and 4, and faces 5 and 6, respectively where the faces contain nodes
as follows
Face Node1 Node2 Node3 Node4

1 1 2 3 4
2 5 6 7 8
3 1 2 6 5
4 4 3 7 8
5 1 5 8 4
6 2 6 7 4
2.Choose the unit vector, a i such that | a i | (i = 1, 2, 3) is a maximum.

3.Among the remaining directed line segments choose the unit vector v in the
direction of the next longest directed line segment.
4.Find the unit vector z perpendicular to the plane containing ( i ,v ) by forming the
cross product k = i × v .
5.Define the unit vector j to be perpendicular to the plane containing ( i ,k ) by forming
the cross product j = k × i .
6.Permute the directions ( i ,j ,k ) so that i' is approximately parallel to the line segment
from G1 to G2 and j' is approximately parallel to the line segment joining G1 and G4.
Nastran Primer 155

5
Solid Elements
Consider the example shown by Figure 1-35
8
7
a1
a2
v
i
5 4
6
a 3 ,k 3
2
Figure 1-35 Orientation of Local Coordinates for HEXA Element
Apparently, a 2 is the longest segment so that the unit vector, i , is in that direction. The
line segment a 1 is the next longest segment so that v is a unit vector in this direction. The unit
vector k (assumed to be coincident with a ) is then perpendicular to the plane containing i and
v while j (here assumed to be coincident with v ) is perpendicular to the plane containing ( i
and k ). The direction k is seen to be approximately parallel to the edge 1-2 while i is
approximately parallel to the edge 1-4. The local coordinate system is then taken to be x = k ,
v = i , and z = j
5.11.8 PENTA Element Coordinate System

The local coordinate system for the PENTA element is formed as shown by Figure 1-36
which shows only the vertices.
z
6
y v
5
3
1
2 x
Figure 1-36 PENTA Local Coordinate System
156 Nastran Primer

Solid Elements
The local x-axis is taken to be the line segment that joins the centroids of the straight line
segments which join vertices 1-4 and 2-5, respectively. A vector, v , is then defined by the
directed line segment that joins the centroids of the straight line segments which join vertices
1-4 and 3-6. The vector, v , is then taken to be in the local x-y plane so that z = x × v
and y = z × x .
5.11.9 TETRA Element Coordinate Syatem

The element coordinate system for the TETRA element is generated using the three
vectors, R , S and T shown by Figure 1-37.
G4 S
G3
G1
T
G2
Figure 1-37 TETRA Element Coordinate System
The element coordinate system is obtained from the three vectors, R , S and T which join
the midpoints of the opposite edges as shown by Figure 1-37. The element coordinate system
is chosen as close as possible to these vectors and points in the same general direction.
5.11.10 Elastic Stiffness Matrix

HEXA and PENTA and TETRA elements may have any or all of the edge nodes deleted.
The calculation of the stiffness matrix is then formulated by taking the deleted nodes into
account.
5.11.10.1 The TETRA Element

The element is transformed to parametric space to a right isosceles tetrahedron. The
stiffness matrix is then calculated using Gaussian integration with either of two integration
schemes which correspond to 4- and 10-node elements.
The 4-noded element is the three dimensional analogy of the two dimensional linar
triangle element for which one Gauss point is appropriate. Elements having edge nodes use the
higher order integration scheme.
5.11.10.2 The HEXA Element

The element is transformed to parametric space as described in Chapter 4. The stiffness
matrix is then calculated using Gaussian integration with either of two integration schemes
which correspond to the 8- and 20-noded elements.
Nastran Primer 157

5
Congruent Elements
The 8-noded element is the three dimensional analogy of the two dimensional bilinear
quadratic element. The same type of shear correction is thus used in the three dimensional
element that is used to improve the membrane shear behavior of the QUAD4 element. Thus,
for the 8-noded element the extensional strains are evaluated at a 2 × 2 × 2 array of Gauss
points but the shear strain is evaluated at a reduced number of special points.
The second integration scheme corresponds to elements having one or more undeleted
edge nodes. The extensional strains are evaluated at a 3 × 3 × 3 array of Gauss points with
reduced integration at special points for shear strains.
5.11.10.3 The PENTA Element

The PENTA element is transformed to a right isosceles prism in parameter space by using
appropriate shape functions. Two integration schemes are provided that use six or nine
integration points. The six point scheme is the default when all edge nodes are deleted. In all
other cases the nine node scheme is the default. Reduced integration is used for shear strains
in either integration scheme.
5.11.11 Mass
The solid elements support both the coupled and lumped mass formulations. The lumped
formulation is default only for the four point TETRA, the six node PENTA and the eight node
HEXA elements. If the elements have edge nodes the coupled formulation is used. The coupled
formulation can be specified independent of the default by including a COUPMASS parameter
statement as described in Chapter 12.
5.11.12 Stress and Strain Recovery

The components of stress and strain are calculated in the material coordinate system; i.e.,
the system specified by the CORDM field on the associated PSOLID statement. The recovery
of stress and strain quantities is controlled by the STRESS and STRAIN Case Control
directives. The components of stress (strain) include:
1.The six components of stress
2.The three principal stresses and principal stress directions
3.The von Mises stress
The components of stress are recovered at the centroid of each element and at the vertex
nodes. The post processing procedures used to transform and average the stress components
by Patran are described in Chapter 12.
5.12 Congruent Elements

Congruent elements can be defined to designate secondary elements that are identical to
a primary element. The congruent elements are defined using the CNGRNT data statement,
which is shown in Bulk Data Image 1-10
1 2 3 4 5 6 7 8 9 10
CNGRNT PRID SECID1 SECID2 SECID3 SECID4 SECID5 SECID6 SECID7 0001
+001 SECID8 SECID9 etc. SECIDN
or
CNGRNT PRID SECID1 THRU
Bulk Data Image 1-10 Specification of Identical Elements
158 Nastran Primer

References
where the fields are defined as

PRID Identification number of the primary element for which elemental
properties will be calculated
SECIDi Element identification numbers of secondary elements whose elemental
properties will be identical to the primary element
5.13 References
[1] R.J. Roark, Formulas for Stress and Strain, McGraw-Hill, Third Edition 1954, pp.
119-121.
[2] J.S. Przemieniecki, Theory of Matrix Structural Analysis, McGraw-Hill,1965, pp. 79.
[3] NASTRAN Programmers Manual, NASA SP-223(01) pp. 4.87-21 through -23
[4] R.M. Rivello, Theory and Analysis of Flight Structures, McGraw-Hill, 1969, pp.
241-246.
[5] R.J. Roark, Formulas for Stress and Strain, McGraw-Hill, Third Edition 1954, pp.
129-130.
[6] NASTRAN Application Manual, MacNeal-Schwendler Corp., 1977, p. 2.6-19.
[7] J.E. Ashton, J.C. Halpin, and P.H. Petit, Primer on Composite Materials: Analysis,
Technomic, 1969, pp. 37-45.
[8] R.M. Jones, Mechanics of Composite Material, McGraw-Hill, 1975, pp. 147-173.
[9] R.H. MacNeal, "A Simple Quadrilateral Shell Element", Computers & Structures, Vol. 8,
pp. 175-183, 1978.
[10] NASTRAN USER’S MANUAL, NASA SP-222(08) Vol. I, pp. 2.4-44.
Nastran Primer 159

5
References
160 Nastran Primer

6 Global Analysis Procedures
A structural system is composed of a collection of structural components that are attached

to each other in some defined way. The purpose of this chapter is to consider those operations
that lead to the global stiffness matrix, together with all operations that are performed on the
global stiffness matrix. These operations include the specification of constraints, partitioning,
and flexibility-stiffness transformations.
6.1 The Global Stiffness Matrix

The equilibrium equation for each element obtained using the finite element method
relates element displacements and element forces as follows
[ k ee ] { u e } = { P e }
where {ue} represents the vector of discrete node point displacements and rotations, {Pe}
represents the vector of external forces acting on the element and [kee] is the element stiffness
matrix. The notation used in this text is to represent matrices using a subscript to indicate a
specific set and, implicitly, to specify the size and form of the matrix. The single subscript on
{ue} and {Pe} defines column vectors having e rows as appropriate for a specific element. The
double subscript on [kee] indicates it is a square matrix having e rows and e columns where e
is the number of degrees of freedom in the element.
The set of element displacements is a subset of the set of all the Grid point degrees of
freedom in the model which is designated by {ug}. An element, ‘n’, connects a subset of the
{ug}degrees of freedom. The element degrees of freedom {ue} for element n can be expressed
in terms of {ug} using a Boolean transformation as:
n n
{ u e } = [ B eg ] { u g } (Eq. 6-1)
n
where [ B eg ] is an e-by-g Boolean matrix whose elements are either one or zero and where it
is also assumed that the stiffness matrix has been transformed from element coordinates to the
global set of coordinates as will be described later in this section. The result of representing all
element displacements in terms of the global set of displacements is to enforce displacement
compatibility at all Grid points in the model.
161 Nastran Primer

Global Analysis Procedures
6
The Global Stiffness Matrix
As an example of the construction of the Boolean transformation, consider the structure

shown by Figure 6-1
6 5
6 4
1 7 4
1 3
2
2 3
Figure 6-1 Idealized Structure
where the structural system is composed of six triangular elements that connect the seven node
points. Assuming, for purposes of explanation, that there is a single degree of freedom of each
node, the g-set is given by:
T
{ ug } = u1 u2 u3 u4 u5 u6 u7 (Eq. 6-2)
The set of displacement degrees of freedom connected by element one includes

displacements at Grids one, two, and seven so that the set of displacements for element one is:
 
 u1  1 0 0 0 0 00
1  
{ ue } =  u2  = 0 1 0 0 0 0 0 { ug } (Eq. 6-3)
 
 u3  0 0 0 0 0 01
 
The global stiffness may now be defined using either an energy or a direct approach. Both
methods will be described in following sections.
After the element stiffness matrices have been transformed to the global set of
displacements, the procedure for assembling the global stiffness matrix can be deduced by
noting that the total strain energy for the structure is the sum of the strain energies for the
individual elements. The virtual change of the internal strain energy is thus:
δU = ∑ δUn (Eq. 6-4)

n=1
where the scalar quantity, δUn is the virtual change in the internal strain energy for element ‘n’
undergoing a virtual displacement, {δue}, and δU is the virtual strain energy for entire system
due to a virtual change in the g-set displacements, {δug}. The virtual strain energy can be
expressed in terms of the appropriate element stiffness so that (Eq. 6-4) can be written as:
162 Nastran Primer

The Global Stiffness Matrix
∑ { δue }
T T
δU = { δu g } [ K gg ] { u g } = [ k ee ] { u e } (Eq. 6-5)
e=1
Since the element displacements, after transformation from element to global coordinates, are
subsets of { u g } set, the virtual strain energy becomes:
N
T
∑ { δug }
T T n n n
δU = { δu g } [ K gg ] { u g } = [ B eg ] [ k ee ] [ B eg ] { u g } (Eq. 6-6)
n=1
where the superscript on element-related matrices is the element number; and, the sum is taken
over all N elements in the model. The global stiffness matrix, by inspection:
N
T
∑ [ Beg ]
n n n
[ K gg ] = [ k ee ] [ B eg ] (Eq. 6-7)
n=1
The effect of the transformation, (Eq. 6-7), is to embed each of the [kee] matrices in a
g-size matrix. The stiffness coefficients are not changed in value as a result of this
transformation, they are placed in appropriate rows and columns of g-sized matrix. The
resulting g-size elemental stiffness matrices are then added to obtain the global stiffness matrix
[Kgg]. The procedure is similar to the process of adding springs in parallel.
For example, the transformation of element two of the structural system shown in
Figure 6-1 leads to:
0 0 0 000 0
0 k 222 k 223 0 0 0 k 227
T 0 k 232 k 233 0 0 0 k 237

n n
[ B eg ] [ k 2ee ] [ B eg ] = (Eq. 6-8)
0 0 0 000 0
0 0 0 000 0
0 0 0 000 0
0 k 272 k 273 0 0 0 k 277
where a superscript on [kee] and its coefficients is the element number.

The explicit Boolean transformation, (Eq. 6-7), is not actually used in a finite element
program. Instead, logic is used in the assembly phase to generate the global stiffness
coefficients without formally performing the matrix algebra. The procedure is:
n
1.Generate the stiffness coefficients for all elements [ k ee ] (n = 1, 2, ... , N) and a data
structure that relates element vertices to external grid points. This data structure is
essentially the connectivity list for an element. This phase is performed by the
Element Matrix Generation (EMG) module.
2.Relate element degrees of freedom to global degrees of freedom using internal
tables called the Equivalence Table (EQEXIN) and the Scalar Index List (SIL). The
EQEXIN relates internal grid IDs to external grid IDs; and the SIL relates grid point
degrees of freedom to the degrees of freedom the g-set.
3.Add all elemental stiffness coefficients with the same row and column designations.
This phase is accomplished by the Element Matrix Assembly (EMA) module.
Nastran Primer 163

6
Local and Global Coordinate Systems
6.2 Local and Global Coordinate Systems

The assembly of the system stiffness matrix in the previous section requires that all the
Grid point degrees of freedom for an element be transformed to the displacement coordinate
system which is specified at each Grid point. This is required since the effect of the Boolean
transformation from element to global coordinates is to enforce displacement compatibility for
all elements connected to a Grid point. NASTRAN allows the user to define a different
displacement coordinate system at each grid point. The procedure used in the EMG module is
to:
• Generate a local coordinate system for the element based on the order in which the grid
points are defined. The procedure for defining the local coordinate system is element-
dependent and is described as a part of the element descriptions in the previous chapter.
• Generate the element stiffness relative to the element local coordinate system.
• Determine the set of direction cosines between the element and displacement
coordinate system for each Grid point as defined by the CD field on the Grid Bulk Data
entity.
• Transform the displacement and rotation stiffness partitions of the stiffness matrix
associated with the Grid points to the displacement coordinate system.
6.3 Transformation of Element Stiffness Matrices

The global set of coordinates is the set of all coordinate systems defined by Cord - type
data statements. The coordinate transformation for the global set of coordinates is available in
the Coordinate System Transformation Matrices (CSTM) table in NASTRAN which is
available to the EMG module.
Let { u e' } be the set of element displacements defined relative to the element coordinate
system, ( x', y', z' ) ,and let [ k e'e' ] be the element stiffness generated in the element coordinate
system. Then, calculate the transformation matrix from global to the element coordinate
system, [ λ e'e ] so that':
{ u e' } = [ λ e'e ] { u e } (Eq. 6-9)
The virtual work for the element is a scalar quantity which is independent of coordinate
system where the virtual work of the element in element coordinates is:
T
δW' = { δu e' } [ k e'e' ] { u e' } – { δu e' } { P e' } (Eq. 6-10)
Substituting(Eq. 6-9) into (Eq. 6-10) leads to:
T T T T
δW = δW' = { δu e' } { λ e'e } [ k e'e' ] { λ e'e } { u e' } – { δu e' } { λ e'e } { P e' } (Eq. 6-11)
The element equilibrium equation in global coordinates is:
[ k ee ] { u e } = { P e } (Eq. 6-12)
where the transformed element stiffness and load are defined to be:
T
[ k ee ] = [ λ e'e ] [ k e'e' ] [ λ e'e ] (Eq. 6-13)
164 Nastran Primer

Specifying Structural Degrees of Freedom
and
T
{ P e } = [ λ e'e ] { P e' } (Eq. 6-14)
The transformation of displacements, loads, and stiffness from local element to global
coordinates is completely automated in finite element programs such as NASTRAN. Generally
the user is not concerned with the local element coordinate system; however, there are cases
that the user must understand the how the local element coordinate system is defined relative
to the global set of coordinates. These are
1.The specification of the local element axes for the three-dimensional beam
2.The specification of a normal pressure on plate bending elements
3.The interpretation of stress output.
6.4 Specifying Structural Degrees of Freedom

We saw in earlier chapters that the finite element method represents a structural system
as an assemblage of elements that are connected at discrete points in the structure. The
behavior of the structure is characterized by the value of the generalized displacements at these
discrete points. The accuracy of the solution is therefore dependent on the number of node
points used to describe the system as well as the distribution of the node points throughout the
structure. Suitable criteria for choosing the number and the distribution of node points is
discussed in later chapters of this text.
Node points serve four distinct purposes:
1.The specification of structural topology
2.The specification of degrees of freedom
3.Accurately locate points at which displacements are constrained or at which loads
are applied
4.Define locations where behavioral variables are to be calculated
It is interesting to note the structural topology, which is the most important characteristic
of a node point from a design point of view, is used only during the calculation of the stiffness
matrix. From the analysis point of view the displacement degrees of freedom at the node is the
most important characteristic since they are associated with the behavior of the structure.
For structurally oriented finite element models six degrees of freedom are defined at each
node, or Grid, point defined on the structure. These degrees of freedom at the geometric point
are shown by Figure 6-2 which show the associated forces as well.
Nastran Primer 165

6
Specifying Structural Degrees of Freedom
z
z
u zi P zi
θ zi M zi
u yi P yi
i θ yi M yi
θ xi M xi i
u xi P xi
y y
x x
a) Grid Point Degrees of Freedom b) Grid Point Forces
Figure 6-2 Displacements and Forces at a Grid Point
The displacements are represented by the set
T
{ ũ i } = u1 u1 u3 θ1 θ2 θ3 (6-15)
where { ũ i } is vector of degrees of freedom at node i. Grid point degree of freedom are
identified in NASTRAN by a pair of integer numbers, (G,C) where
G Identification number of a geometric grid point

C A component number having a value from one to six that
identifies a degree of freedom using the following correspondence
table
Table 6-1 Degree of Freedom Codes
Physical Degree of Degree of Freedom

Freedom Code
u1 1
u2 2
u3 3
θ1 4
θ2 5
θ3 6
where the subscripts on the physical degree of freedom variables indicates the coordinate sense
of the displacement, ui, or the axis associated with the rotation, θi.
Geometric grid points are defined using the Grid data statement described in Grid Points
-GRID (p. 171). The coordinate systems used to define the location of the geometric grid
points are defined using Cord-type data statements as described in Coordinate Systems -
CORD (p. 167).
166 Nastran Primer

Defining Scalar Degrees of Freedom - SPOINT
Degrees of freedom may be associated with non structural behavior such as control
system variables, temperatures, and scalar spring mass systems. Scalar degrees of freedom
may be defined directly in NASTRAN using Spoint as described in Defining Scalar Degrees
of Freedom - SPOINT (p. 167).
6.5 Defining Scalar Degrees of Freedom - SPOINT

Scalar degrees of freedom are useful for representing non structural degrees of freedom
such as the average of several displacements. However, it is also used to specify the cross
sectional warping for the Beam element. Scalar degrees of freedom maybe defined in two
different ways:
1.Implicitly by reference on a scalar element data statement (i.e. CELAS, CMASS,
etc.)
2.Explicitly by an SPOINT Bulk Data statement
In either case, a single degree of freedom in defines in the g-set for each scalar point. Scalar
elements are defined explicitly by using the SPOINT Bulk Data entity shown by Bulk Data
Image 6-1.
1 2 3 4 5 6 7 8 9 10
SPOINT SPID1 SPID2 SPID3 SPID4 SPID5 etc.
*alternatively
SPOINT SPID1 THRU SPID2
Bulk Data Image 6-1 Scalar Point - SPOINT
where
Field Description
Scalar point identification number that must be unique among all GRID of
SPID
SPOINT in Bulk Data. (Integer > 0)
6.6 Coordinate Systems - CORD

Coordinate systems define reference systems for geometry and for vector quantities such
as displacements and forces. We know from experience that different types of coordinate
systems are convenient for different geometric topologies. Therefore, we would expect a finite
element analysis program to provide the user with several types of coordinate systems.
The default coordinate system, which exists and needs no definition, is called the Basic
Coordinate system. An unlimited number of number of rectangular, cylindrical and spherical
coordinate systems can be defined. These subsidiary coordinate systems are defined relative to
the Basic Coordinate system or another defined coordinate system by Cord-type data
statements as described below.
Coordinate systems are identified by a coordinate identification (CID) number, which
must be a positive integer. The coordinate identification CID= 0 refers to the Basic Coordinate
system which is a rectangular Cartesian coordinate system.
The set of all coordinate systems defined for the analysis is called the Global Set of
Coordinates. It must be noted that several authors have used the term “global coordinate
system” to define what is termed the Basic Coordinate System in NASTRAN.
Nastran Primer 167

6
Coordinate Systems - CORD
A subsidiary coordinate system (X1', X2', X3') is shown by Figure 6-3.
X' 3
X
3
B
C X' 2
V
A
X
1
X
2
X' 1
Figure 6-3 Subsidiary Coordinate System
The new coordinate system is defined in terms of a reference coordinate system (X1, X2,
X3) by specifying the position of three points A, B, and C as follows:
1.Point A is the origin of the new coordinate system
2.The directed line segment between point A and B defines the X' 1 -axis.
3.Point C lies in the X' 1 , X' 3 -plane.
The X' 2 -axis is thus given by
i′ 2 = i′ 3 × V (6-16)
and
i′ 1 = i′ 2 × i′ 3 (6-17)
where V is the vector from point A to C.
6.6.1 Subsidiary Coordinate System

The location of the three points can be defined two ways: The coordinates of points A, B,
and C can be specified directly by the CORD2-type statements, which are shown in Bulk Data
Image 6-2; or, three geometric Grid points, which correspond to points A, B, and C as shown
by Bulk Data Image 6-3.
168 Nastran Primer

Cylindrical.
1 2 3 4 5 6 7 8 9 10
CORD2C CID RID A1 A2 A3 B1 B2 B3 +1
+1 C1 C2 C3
Rectangular.
1 2 3 4 5 6 7 8 9 10
CORD2R CID RID A1 A2 A3 B1 B2 B3 +1
+1 C1 C2 C3
Spherical.
1 2 3 4 5 6 7 8 9 10
CORD2
CID RID A1 A2 A3 B1 B2 B3 +1
S
+1 C1 C2 C3
Bulk Data Image 6-2 Coordinate Systems Defined by Coordinates of Three Points
The logical CORD2-type statements described by Bulk Data Image 7-2 each require two
lines of input. The coordinate definitions for each type of coordinate system are identical
except for the mnemonic name in the first field of the input statement where CORD2C,
CORD2R, and CORD2S specify cylindrical, rectangular, and spherical coordinates,
respectively. The other fields of the CORD2-type data statements are
Field Description
CID Coordinate system identification number (integer > 0)
The CID of the reference coordinate system. A blank or zero entry in this
field indicates that the Basic Coordinate system is the reference. However,
RID any coordinate system defined in the Bulk Data section may be used. The
coordinates of the points A, B, and C are then interpreted for the various
allowable types of coordinate systems as shown by Table 7-1 (integer > 0)
(A1, A2, A3) Coordinates of point A that defines the origin of coordinate system (real)
(B1, B2, B3) Coordinates of point B that lies on the X3'-axis (real)
(C1, C2, C3) Coordinates of point C that lies in the X3', X1'-plane (real)
Cylindrical.
1 2 3 4 5 6 7 8 9 10
CORD1C CID GA GB GC CID GA GB GC
Rectangular.
1 2 3 4 5 6 7 8 9 10
CORD1R CID GA GB GC CID GA GB GC
Nastran Primer 169

6
Spherical.
1 2 3 4 5 6 7 8 9 10
CORD1S CID GA GB GC CID GA GB GC
Bulk Data Image 6-3 Coordinate System Defined by Three Grid Points
The coordinate definition data statements shown on Bulk Data Image 6-3 are identical for
each type of coordinate system except for the data statement mnemonic in the first field where
CORD1C, CORD1R, and CORD2S specify cylindrical, rectangular, and spherical
coordinates, respectively. The other field designations are
Field Description
CID Coordinate system identification number (integer, > 0)
GA,GB,GC Identification numbers of three GRID points (integer, > 0) where:
GA defines Grid point A at the origin
GB defines Grid point B lying on the X3'-axis, and
GC defines Grid point C lying in the X3'- X1'-plane
The cylindrical, rectangular, and spherical coordinate systems defined by the CORD-type
Bulk Data are shown by Figure 6-4a, Figure 6-4b, and Figure 6-4c, respectively. The
components of the points A, B, and C specified on the CORD2-type statements are interpreted
in terms of the coordinate system defined by coordinate set identification number RID. For
example, if a spherical coordinate system is associated with this set identification number, then
the components would be the R, θ, and φ coordinates of the point as shown by Figure 6-4c.
z
u3
P u2
u1
zp y
xp
yp
x
a) Rectangular
z z
u3
u1
u2 θ P
P u3
R
u1 y
z y
R u2
φ
θ
x b) Cylindrical x c) Spherical
Figure 6-4 Coordinate Definitions and Displacement Components
170 Nastran Primer

Grid Points -GRID
The components of displacement referred to rectangular, cylindrical, and spherical

coordinates are also shown by Figure 6-4a,Figure 6-4b, andFigure 6-4c. NASTRAN identifies
the Grid point degrees of freedom in the output by labeling the components of displacement
T1, T2, and T3 and the components of the rotation R1, R2, and R3. These components are
interpreted as components of displacement in the displacement coordinate system specified on
the Grid data statement, which is defined in the next section, associated with the point, using
Table 6-4. The displacement coordinate system at each geometric grid point can be different
from the system used to define the location of the point.
Coordinate Coordinate Component

Type 1 2 3
Rectangular X Y Z
Cylindrical R θ(degrees) Z
Spherical R φ(degrees) θ(degrees)
Table 6-4 Definition of Coordinate Components
6.7 Grid Points -GRID

A Grid point data statement shown by Bulk Data Image 6-5 defines.
1.The coordinates of a point in space relative to a specified coordinate system
2.six degrees of freedom in the g-set of displacements
3.The reference coordinate system for displacements and forces
4.Degrees of freedom which are to be either constrained or removed from the
analysis.
1 2 3 4 5 6 7 8 9 10
GRDSET CP CD PSPC
GRID GID CP X1 X2 X3 CD PSPC SEID
Bulk Data Image 6-5 Geometric Grid Point Definition - GRID
where
Field Description
GID Grid point identification number (integer, >0).
Identification number of the coordinate system used as reference for the
CP
coordinates (integer, > 0).
Coordinates of the grid point with respect to the CP-coordinate system
(X1, X2, X3) (real). See Table 7-2 in the previous section for definition of coordinate
directions in non rectangular coordinate systems.
Identification number of the displacement coordinate system (integer, >
CD
0).
PSPC Purged permanent single point constraints (integer, > or blank)
The Grid point is a material point at which a displacement vector is defined. The
displacement vector represents the six displacement degrees of freedom at the point: three
components of the displacement vector ( u 1, u 2, u 3 ) and three components of the rotation
vector ( θ 1, θ 2, θ 3 ) . Six degrees of freedom are added to the g-set for each GRID point.
Nastran Primer 171

6
Grid Points -GRID
Since the number of rows and columns in the system matrices is determined by the
number of degrees of freedom in the g-set it may turn out that one or more of the degrees of
freedom are not connected by elastic elements.This would result in a singular stiffness matrix.
For example, suppose the model is to represent in-plane stretching of a plate represented using
membrane elements.
The membrane element only provides connectivity for in-plane displacements so the out
of plane displacement and all rotations have no connectivity. This, in turn, means that the
associated rows and columns are null and that the stiffness matrix is singular. NASTRAN
incorporates a process called AUTOSPC for detecting and removing singular degrees of
freedom. However, they are more properly removed by using the PSPC field since the size of
the assembled stiffness matrix is greatly reduced for some models.
6.7.1 Grid Identification Number

Each Grid point must have a Grid identification (GID) number that is unique among all
Grid and Spoint identification numbers. The GID is called the external Grid point number. The
external GID are converted to internal numbers to simplify internal calculations. Two tables
are maintained that allow conversion from one form to another. Results are always identified
using the external GID.
The GIDs need not form a consecutive numerical sequence. If a structural model were to
be represented by three geometric grid points, for example, the sequences (100, 200, 300), (1,
2, 3), or (7, 12, 17) would produce the same internal sequence of equations and would result
in system matrices with the same matrix topology.
6.7.2 Geometric Coordinates

The location of the Grid point is specified by the coordinates (X1, X2, X3) defined relative
to the reference coordinate system specified by the CP field. The value of the entry in the CP
field is the coordinate system identification (CID) of the reference system. If the CP field is
blank the Basic Coordinate system is used. If a non blank entry exists in the CP field the
coordinate system associated with the entry must be defined or else the system will generate a
fatal error message during execution. If CP is zero, then (X1, X2, X3) are taken to be the
coordinates of the point referenced in the Basic Coordinate system. If CP is the number of a
user-defined coordinate system, then the coordinates are interpreted according to Table 6-4.
6.7.3 Displacement Coordinates

The CD field specifies the reference coordinate system for the displacement and rotation
vectors associated with the Grid point. If CD is zero, then the Basic Coordinate system is used.
If the displacement and rotation vectors are to be defined in some other coordinate system then
the CD field is the CID of a user-defined coordinate system. The components of the
displacement and rotation vectors are interpreted for the various coordinate systems using
Figure 6-4. The rotations (θ1, θ2, θ3) are components of the rotation vector in the directions
using the right-hand rule.
6.7.4 Permanent Constraints

The function of the PSPC field is to specify displacement degrees of freedom for which
permanent constraints are to be prescribed. If no stiffness is associated with degrees of freedom
identified by the PSPC field the resulting action is to remove the them from the analysis during
the assembly of the g-sized stiffness matrix. Since there is no stiffness the associated constraint
force will be zero.
172 Nastran Primer

External and Internal Degrees of Freedom
PSPC can also be used to specify displacement constraints for connected degrees of
freedom. However, it is highly recommended that displacement boundary conditions be
applied using SPC data entities as described in a later section. Constraints applied using PSPC
have a constraint value of zero.
The entry in the PSPC field is a packed set of degree of freedom codes which specify
degrees of freedom included in the permanent constraint set. These degrees of freedom are
interpreted as displacement components in the coordinate system specified by the CD field on
the GRID entity. The codes for the six displacement at the geometric grid point are given by
Figure 6-1 presented in a previous section.
The use and specification of constraints is described in succeeding sections of this

chapter. It is worth repeating that the constraints specified on Grid data entities are applied
during the assembly of the stiffness matrix and cannot be changed. In addition, the intended
use of PSPC is to remove one or more of the six degrees of freedom that are automatically
generated but that do not connected by elastic elements from the analysis.
6.7.5 Default Values - GRDSET

The GRDSET, which is also shown by Bulk Data Image 6-5, is used to define default
values for the CP, CD, and PSPC fields. It is a user convenience for eliminating repetitious
data. However, a word of caution is in order regarding the use of GRDSET: The values CP,
CD, and PSPC on the GRDSET are default values and are used only if the corresponding field
of the Grid is blank. The default value is not used if an integer, including the integer zero, exists
in the associated field of the GRID entity. A common mistake is assuming that permanent
single point constraints placed in the GRID statement are added to those specified in the
GRDSET. On the contrary, a PSPC specified on GRID overrides the default value on
GRDSET.
Only one GRDSET may appear in Bulk Data; and, at least one of the fields, CID, CD, or
PSPC must be different from zero. The inclusion of a GRDSET statement with only zero
entries will result in a fatal error.
6.8 External and Internal Degrees of Freedom

The program logic is simplified if the degrees of freedom are identified by a set of
consecutive numbers starting at one. All user references to degrees of freedom are in terms of
the external Grid numbers and degree of freedom codes and internal computations use internal
codes. The conversion from external to internal degrees of freedom is automatically performed
by NASTRAN as follows:
• The lowest point number is found. If it is a SPOINT then the first internal degree of
freedom corresponds to the SPOINT; if it is a GRID then the first six internal degrees
of freedom correspond to the GRID point.
• The next highest point number is found and one internal degree of freedom is generated
if it is an SPOINT and six internal degrees of freedom if it is a GRID, etc.
The set of all GRID point and SPOINT scalar degrees of freedom is called the g-set and
is denoted by { u g } , The total number of degrees of freedom in the g-set is
Ng = 6N + S
where N is the number of GRID points and S is the number of SPOINTs. The parameter called
LUSET (length of the USET) is given the value of Ng in the solution sequences.
Nastran Primer 173

6
Displacement Sets
6.9 Displacement Sets

It is convenient to consider the displacement degrees of freedom as the elements of a
displacement set. The set of all grid point degrees of freedom including scalar degrees of
freedom is called the g-set and is of order Ng as described above.
The stiffness matrix associated with the g-set is denoted by [ K gg ] and is normally
generated from a finite element model. The system matrix associated with the g-set of degrees
of freedom may be singular, i.e., the rank of [ K gg ] is less than Ng, either because the internal
degree of freedom does not physically exist, or because the elements of the g-set are linearly
dependent. As noted in the description of the GRID entity, six internal degrees of freedom are
defined at each Grid point implying six rows and six columns in the global stiffness matrix that
are initially zero.
If the finite element model does not define elements in the stiffness matrix associated with
a degree of freedom, then the degree of freedom is not connected and must be purged from the
g-set. After purging all unconnected degrees of freedom, the resulting subset of { u g } may still
include rigid body motion that must be restrained to make [ K gg ] non singular.
NASTRAN allows the user to define a number of independent subsets of { u g } associated
with the following types of constraints:
1.Specification of values for specific degrees of freedom, which is termed a single
point constraint (SPC)
2.Specification of linear relation between two or more degrees of freedom, which is
termed a multi point constraint (MPC)
3.Specification of reactionless constraints, which are used to model unsupported
structures (SUPORT)
In addition, static condensation can be specified using ASET and OMIT.
The data sets that are removed by a constraint or reduction operation are mutually
exclusive. A degree of freedom can only exist in one of the mutually exclusive sets so that a
degree of freedom specified in any one of the sets cannot be specified in another removed data
set.
Each of these constraint and reduction operations acts on a merged set of displacement
degrees of freedom to produce partitioned data sets, one of which is retained in the analysis
and one of which is removed. These operations are performed in the sequence shown by the
following table that identifies the combined set, the data set removed, and the data set retained
by the operation.
Constraint Operates Partitioned Data Sets

Operation On Merged Set Set Retained Set Removed
MPC {ug} {un} {um}
SPC {un} {uf} {us}
OMIT
or {uf} {ua} {uo}
ASET
SUPORT {ua} {ul} {ur}
Table 6-6 Displacement Data Sets for Static Analysis
174 Nastran Primer

Multipoint Constraints - MPC
6.9.1 Merged Data Sets

The merged data sets are
{ug} Set of all grid point degrees of freedom including extra points.
{un} Set of all degrees of freedom not eliminated by multi point constraints.
{uf} Set of free degrees of freedom that remain after specification of single point
constraints.
{ua} Analysis set is solution set for real eigenvalue analysis.
6.9.2 Mutually Independent Data Sets

The merged sets are partitioned and one of the sets is removed from the analysis. The set
removed is called a mutually independent set since degrees of freedom cannot be members of
another removed set. The mutually independent displacement sets identified as removed data
sets in Table 6-6 are
{um} Set removed by multi point constraints
{us} Set removed by single point constraints
{uo} Set omitted by static condensation
{ur} Set of forceless degrees of freedom, which remove rigid body
motion and
{ul} Set left over, which is the solution set in static analysis
6.10 Multipoint Constraints - MPC

A multi point constraint is a linear equation relating displacement degrees of freedom.
Relations of this type can be used to define ad hoc connectivity between degrees of freedom.
Using MPCs we can model rigid links between Grid points or to specify constraints that are a
linear combination of Grid point displacements.
6.10.1 Uses of MPC

There are a number of modeling situations that require the use of ad hoc relations between
degrees-of-freedom including:
6.10.1.1 Representing very stiff structural members.

Let's assume that you are modeling a structural assembly that includes one or more
components which are extremely stiff compared with the rest of the structure. Examples would
be a vehicle model that includes the engine, or a spacecraft model that includes a very stiff
interface flange between booster stages. In each of these cases the stiff component defines the
load path through that part of the structure, and in each case the stiff component deforms very
little compared to the deformations in the more flexible components of the structure. They have
the property of being "Rigid" compared to the flexible components.
Before the need for “Rigid” elements was discovered, stiff elements in a finite element
model were represented by stiff elastic elements. After all, what's wrong with faithfully
modeling reality? Unfortunately, early practitioners found the solutions were invalid because
of something called matrix ill conditioning.
Nastran Primer 175

6
Without going into the mathematics, stiff elastic components lead to large differences in
magnitude between the off-diagonal coefficients of the stiffness matrix. These large
differences, in turn, lead to numerical problems in the solution phase. Research into the reasons
for wrong answers to, what appeared to be, well posed finite elements models lead to the
recognition that stiff members must be represented by using proper constraint equations rather
than very stiff elastic elements. All of this discussion leads to two conclusions:
1..Indicators of the accuracy of the solution process are required. NASTRAN
provides several indicators including the approximate conditioning numbers and a
norm called epsilon whose value tells us how many significant figures in the result
are valid.
2.Finite element programs must support linear constraint equations. Linear
constraints must be incorporated in finite element programs. NASTRAN
implements MPC Bulk Data and a number of rigid elements.
6.10.1.2 Load transfer between dissimilar elements.

Let's suppose that you wanted to model one portion of the structure using three
dimensional continuum elements and another portion using a strength of materials shell
approximation. The model looks great, and it contains no apparent singularities. However the
solution fails in decomposition because the reduced stiffness matrix is singular. Why? Because
the model contains a “mechanism”. There is no way to transfer a moment from the shell
elements to the solids because the solids have no rotational degrees-of-freedom.
In order to eliminate the singularity an ad hoc relation must be defined that will, in effect,
create a couple in the solid to resist the moment in the shell. This is an interesting exercise
which will be discussed later.
6.10.1.3 Constraints in an arbitrary direction.

You will recall that the displacement coordinate system, which is defined by the CD field
on the GRID entity, is used for all references to degrees-of-freedom in constraint equations.
Now, how would you describe the constraint equation for a grid point which is constrained to
move parallel to a boundary, where the boundary in the x- direction is not aligned with a
coordinate direction?
A set of linear relations that describes kinematic constraints between degrees of freedom
can be specified directly by Multi Point Constraints (MPC). The prescribed value for a single
degree of freedom is called a Single Point Constraint (SPC) and is discussed in the following
section. The combined capability of MPCs, SPCs and scalar degrees of freedom is sufficient
to specify any linear relationship between degrees of freedom.
176 Nastran Primer

6.10.1.4 4. Rigid Links

For example, consider the rigid link that connects grid points A and B as shown by
Figure 6-1
x AB
u 2B
y
u 6B
u 1B
B
y AB
A
u 1A
u 6A
u 2A
x
Figure 6-1 Planar Rigid Link
Since the link between points A and B is rigid, the following kinematic relations exist
between displacement degrees of freedom at A and B
u 1B = u 1A – y AB u 6A
u 2B = u 2A + x AB u 6A (Eq. 6-1)
u 6B = u 6A
where xAB and yAB represent the distances from point A to B in the x and y directions,
respectively.
6.10.2 Reduction to the n-Set

The general form of the MPC relation is:
[ G mg ] { u g } = 0 (Eq. 6-2)
(Eq. 6-2) represents a set of m constraint equations where [Gmg] is a matrix of coefficients
specified by the user having m rows and g columns. The solution set {ug} must, therefore,
satisfy(Eq. 6-2) in addition to the following set of equilibrium equations for the g-set:
[ K gg ] { u g } = { P g } + { c g } (Eq. 6-3)
where { P g }are external forces applied to the g-set degrees of freedom and { c g } are forces of
constraint.
There are two procedures for incorporating the constraint equations into the solution: a
reduction procedure or augmenting the stiffness to incorporate the constraints using Lagrange
multipliers.
Nastran Primer 177

6
NASTRAN uses the elimination procedure, and there’s a story there. The g-set is first
partitioned into two complementary sets; the m-set which contains exactly m degrees of
freedom where m is the number of constraint equations, and the remaining degrees of freedom
which is called the n-set. The question is how to determine the m-set degrees of freedom, since
once the m-set is known the process for determining the transformation from the ug set to the
n-set is straight forward as we will show below.
For a simple set of constraint equations the choice of the m-set degrees of freedom can be
fairly straight forward. But for a general set of equations there must be a better way: and there
is. It’s all about the rank of the matrix Gmg. If the rank is less than m the equations are not
linearly independent. If the equations are linearly dependent than we need to find the m best
pivot columns. It turns out that is not to hard to do; and MSC has included the automated
determination of the m-set degrees of freedom. Legacy NASTRAN doesn’t include it yet but
hopefully it is on the to-do list as a future enhancement.
So, once the m-set is known, then the g-set is represented in partitioned form as:
{ um }
{ ug } = (Eq. 6-4)
{ un }
The constraint equations are then used to defines a transformation between the m- and n-
set that is used to reduce(Eq. 6-3)to the n-set. In order to reduce the set of equations to the n-set,
we proceed by writing the set of constraint equations (Eq. 6-2) in terms of the m- and n-
partitions as follows:
[ R mm ] { u m } + [ R mn ] { u n } = 0 (Eq. 6-5)
The partitioned constraint equation can now be used to define a transformation between
the mutually independent set, { u m } , and the merged set, { u n } , as follows:
{ u m } = [ G mn ] { u n } (Eq. 6-6)
where
–1
[ G mn ] = – [ R mm ] [ R mn ] (Eq. 6-7)
The value of the mutually independent set, { u m } , is seen to be dependent on the value of
the retained set of displacements, { u n } . Since (Eq. 6-6) shows that { u m } is dependent on the
{ u n } displacements the term “mutually independent” can lead to some confusion. However,
the term means that { u m } is eliminated from and is therefore independent of other eliminated
sets of displacements including the { u s } , { u o } and { u a } sets as we will see later.
We can now use the constraint relation (Eq. 6-6)to obtain a transformation between { u g }
and { u n } by writing (Eq. 6-4) as follows:
 [G ] 
{ ug } =  mn  {u } = [G ]{u } (Eq. 6-8)
 [I ]  n gn n
 nn 
where [ I nn ] is an identity matrix of n-size.

The reduced stiffness matrix is then found by noting that the constraint forces { c g }
perform no work during a virtual displacement, { δu g }, so that the virtual work expression
associated with(Eq. 6-3)is
178 Nastran Primer

T
{ δu g } ( [ K gg ] { u g } + { P g } ) = 0 (Eq. 6-9)
Admissible virtual displacements must satisfy the constraints on the displacement function
(Eq. 6-8), so (Eq. 6-9) becomes
T T
{ δu n } [ G gn ] ( [ K gg ] [ G gn ] { u n } + { P g } ) = 0 (Eq. 6-10)
The virtual displacements,{ δu g }, are arbitrary the satisfaction of (Eq. 6-10) requires that
T T
[ G gn ] [ K gg ] [ G gn ] { u n } – [ G gn ] { P g } = 0 (Eq. 6-11)
or
[ K nn ] { u n } = { P n } (Eq. 6-12)
where the reduced stiffness and load are given by
T
[ K nn ] = [ G gn ] [ K gg ] [ G gn ] (Eq. 6-13)
T
{ P n } = [ G gn ] { P g } (Eq. 6-14)
The multi point constraint specification must, therefore, perform two tasks:
1.Define the matrix of coefficients [ G mg ]
2.Specify the degrees of freedom in the m-set
The form of the MPC Bulk Data thus requires you to:
1.Define each equation in the set of m constraint equations with a single logical MPC
data statement. This means that if there are six constraint equations there will be six
logical MPC Bulk Data, all with the same MPC Identification, MPCID.
2.define an m-set degree of freedom. This degree-of-freedom is then partitioned into
the m-set and thus must be unique among all MPC definitions. Since m MPC Bulk
Data are required to represent the m equations of constraint, this means that the m-
set contains exactly m unique degrees-of-freedom.
3.Assure that [ R mm ] is not singular so that [ G mn ] can be formed.
Nastran Primer 179

6
6.10.3 MPC Data Entity

The MPC data entity is designed to accomplish both tasks. Each constraint equation is
defined by one logical MPC entity. In order to specify ‘m’ constraint equations ‘m’ logical
MPC entities are required. The MPC entity shown by Bulk Data Image 6-1defines a degree of
freedom to be included in the m-set as well as nonzero coefficients of the constraint.
1 2 3 4 5 6 7 8 9 10
MPC MPCSID GM CM Am GN1 CN1 AN1
GN2 CN2 AN2 -etc.-
MPCADD MPCSID S1 S2 S3 S4 S5 S6 S7
S8 S9 -etc.-
Bulk Data Image 6-1 Multi point Constraint Specification
wherethe data fields are described as follows:
Field Name Description

The MPC set identification number that is referenced by an MPC Case
MPCID
Control directive to select a set of constraint equations from Bulk Data
A Grid point number and degree of freedom code for a single degree of
GM, CM freedom that is partitioned from the g-set into the m-set
(GM, integer > 0, 0 < CM < 6)
The constraint coefficient of degree of freedom (GMi, CMi) in the current
AM
constraint equation. (real)
A Grid point number and degree of freedom code defining a single g-set
GNj, CNj degree of freedom in the current constraint equation which has a non zero
coefficient, integer > 0
A Grid point number and degree of freedom code defining a single g-set
GNj, CNj degree of freedom in the current constraint equation which has a non zero
coefficient, integer > 0
The constraint coefficient of degree of freedom (GNj, CNj) in the current
ANj
constraint equation. (real)
MPC set identification number of MPC sets being unioned by MPCADD
Sk
Bulk Data
The set identification number in field two of the MPC data statement allows several
logical MPC sets to be included in Bulk Data. The particular MPC set which is to be used at
execution time is specified by the MPC Case Control Directive as shown below
MPC = <MPCSID>
where MPCSID is the integer number assigned to an MPC set in Bulk Data.
6.10.4 Combining MPC Sets - MPCADD

Multiple constraint equations can be combined into a single logical set by the MPCADD
entity. The MPC sets specified by sets S1, S2,...,SN are combined and given a unique MPCSID
on the MPCADD entity that is then specified by an MPC Case Control Directive. All the MPC
and MPCADD set identification numbers must be unique.
180 Nastran Primer

6.10.5 Specifying the m-set Degrees of Freedom

One logical MPC data statement is required for each degree of freedom to be included in
the m-set. The degrees of freedom are identified by the external degree of freedom number,
which consists of the pair (GM, CM) where
GM is the identification number of a GRID or SPOINT
and
CM is a degree of freedom code at G. If G is a scalar point, then C is
either blank or equal to integer zero.
6.10.6 Recovering and Printing MPC Forces

The forces of associated with the MPC constraints are often of great interest. If MPCs are
used to model a spot weld, for example, the associated MPC forces are those that the weld must
transmit. The forces associated with MPCs are requested using the MPCFORCES case control
directive as described in Chapter 1. All MPC forces could be requested using the following
Case Control directive:
MPCFORCES = ALL
6.10.7 Defining the Constraint Equation

The first degree of freedom, identified by (GM, CM) in fields three and four, is a g-set
degree of freedom which will be partitioned into the m-set. The coefficient Am is the associated
coefficient of [ G mg ] . All subsequent degrees of freedom defined by point number and degree
of freedom code pairs, (GNj, CNj) specify a column in the current constraint equation having
a non zero coefficient Aj in the [ G mg ] matrix. The Grid point degrees of freedom specified on
the MPC data statement are defined with respect to the displacement coordinate system
specified by the CD field on the associated GRID entity.
6.10.7.1 General Rigid Link

It is common to have relatively stiff structure connecting flexible components. One may
be tempted to use a stiff elastic element to connect such points, but experience has shown an
ill conditioned stiffness matrix may result due to orders-of-magnitude-differences in the off-
diagonal stiffness terms. A kinematic constraint relating the displacements at the two points
eliminates the ill conditioning problem. The kinematic constraint is modeled as a rigid link
between the points connecting the stiff component.
Nastran Primer 181

6
Consider two grid points A and B, as shown by Figure 6-1 with a rigid link connecting the
two points.
A
x
Figure 6-1 Rigid Connection of Two Grid Points
The rotations of the two points are θ A and θ B , respectively; and the displacements are
u A and u B . The displacement and rotation of B can then be expressed in terms of the
displacement at A as
u B = u A + θ A × r AB (Eq. 6-1)
where r AB is the vector from point A to point B and, since the rotation of all points on the rigid
link are the same:
θA = θB (Eq. 6-2)
. These vector equations can be written in scalar form as
u1B = u1A - ryu6A + rzu5A

u2B = u2A - rzu4A + rxu6A
u3B = u3A - rxu5A +ryu4A
(Eq. 6-3)
u4B = u4A
u5B = u5A
u6B = u6A
where rx, ry, rz are the components of r AB in the x-, y-, and z-directions.
As a numerical example, let the coordinates of points A and B be given as
A => (1, 2, 3)
B => (5, 3, 1)
which are identified as grid points 363 and 132, respectively. The components of r AB can then
be found to be rx = 4; ry = 1; and rz = 2
The constraint (7-34) can then be written as in the form of (7-19) as follows
182 Nastran Primer

 u 1A 
 
 u 2A 
 
 u 3A 
 
 u 4A 
1 0 0 0 0 0 –1 0 0 0 –2 1  
 uA 
0 1 0 0 0 0 0 –1 0 2 0 –4  
0 0 1 0 0 0 0 0 –1 –1 4 0  u 6A 
 = 0 (Eq. 6-4)
0 0 0 1 0 0 0 0 0 –1 0 0  u 1B 
 
0 0 0 0 1 0 0 0 0 0 –1 0  u 2B 
0 0 0 0 0 1 0 0 0 0 0 –1  
 u 3B 
 
 u 4B 
 
 u 5B 
 
 u 6B 
There are six constraint equation, therefore six MPC data statements are required. Looking at
the first equation, there are four non zero coefficients. The m-set degree of freedom must have
a non zero coefficient so the displacements at B will be arbitrarily chosen to be members of the
m-set. Letting the Grid ID of A and B be 383 and 132, respectively, the MPC entities defining
the constraint equation are:
1 2 3 4 5 6 7 8 9 10
MPC 101 132 1 1. 363 1 -1.
+ M1 363 5 -2. 363 6 1.
MPC 101 132 2 1. 363 2 -1. + M1
+ M2 363 4 2. 363 6 -4.
MPC 101 132 3 1. 363 3 -1. +M2
+M3 363 4 -1. 363 5 4.
MPC 101 132 4 1. 363 4 -1. +M3
MPC 101 132 5 1. 363 5 -1.
MPC 101 132 6 1. 363 6 -1.
where the MPC set identification MPCSID = 101 has been used for each MPC data statement.
The user will note that nine lines of input are required to define the simple kinematic constraint
defined by the rigid link.
The specification of even simple rigid link connecting two points is a laborious task. Since
all of the geometric data required to define the constraint equations is included on the GRID it
would seem logical to let the program generate the associated MPC set programmatically.
NASTRAN does this by including a library of constraint elements. These elements defined the
topology of the rigid element and generate the constraint equations. The example described
above could have been defined by using a single RBAR constraint element The rigid and
constraint elements are described in the chapter on elements
Nastran Primer 183

6
Single Point Constraints - SPC
6.11 Single Point Constraints - SPC

Displacement boundary conditions are used to specify degrees having known values.
Generally the value of these degrees of freedom will be zero but this need not be the case. The
specification of displacement boundary conditions is the most common use of SPC. It is used,
for example to define the “built-in”, or fully constrained, conditions at one end of a cantilever
beam. The statement “built-in” means the displacements associated with the fixed degree of
freedom are not unknowns. They are known and the fact they are zero for a beam results in a
unique solution for the cantilever beam under a certain loading condition.
Single point constraints and multi-point constraints are also used to specify symmetric
boundary conditions. The use of symmetry conditions, if present, can lead to a great reduction
in the resources required to obtain a solution. For example, a square plate with symmetric
boundary conditions and isotropic material can be modeled using one eighth of the structural
model. Symmetry boundary conditions are then used to represent the rest of the plate. Without
getting into the details of how to model symmetry conditions at this time it should be rather
obvious that the size of the structural model is greatly reduced.
The specification of displacement boundary conditions are applied to degrees of freedom
in the s-set. Degrees of freedom specified on SPC entities have an explicit or implicit constraint
value and are partitioned into the s-set. SPC thus defines the partition of the n-set as follows:
{ us }
{ un } = (Eq. 6-5)
{ uf }
where { u s } is prescribed as:
{ us } = { Ys } (Eq. 6-6)
Where { Y s } are known values specified on SPC Bulk Data statements.
6.11.1 Reduction to the f-Set

The equations associated with the n-set are reduced to the f-set by partitioning the n-set
equations,(Eq. 6-12), using (Eq. 6-5) which gives:
   
[ K ss ] [ K sf ]  { u s }   { Ps } 
  =   (Eq. 6-7)
[ K fs ] [ K ff ]  { u f }   { P̃ f } 
   
The { u s }set is prescribed as indicated by (Eq. 6-6) so that the reduced set of equations for the
free or f-set becomes
[ K ff ] { u f } = { P f } (Eq. 6-8)
where
ˆ
{ P f } = { P f } – [ K fs ] { Y s } (Eq. 6-9)
6.11.2 Single Point Constraint Forces

The internal forces are given by
184 Nastran Primer

     
 { cs }  [ K ss ] [ K sf ]  { Ys }   { Ps } 
  =  –  (Eq. 6-10)
 { cf }  [ K fs ] [ K ff ]  { uf }   { Pf } 
     
The forces of reaction associated with {us} degrees of freedom are found from the first of
equations (Eq. 6-10) to be:
{ c s } = [ K ss ] { Y s } + [ K sf ] { u f } – { P s } (Eq. 6-11)
where { c f } = 0 since there are no constraints on the f-set. The single point constraint forces
can be recovered by the Case Control Directive.
SPCFORCE = <sets defined by a SET Case Control

directive and specifies the points at which the forces of
constraint are desired.
6.11.3 SPC Data Entity

Single Point Constraint (SPC) and is defined by SPC, SPC1, and SPCADD data statements as
shown in Bulk Data Image 6-1
1 2 3 4 5 6 7 8 9 10
SPC SID GID1 C1 VALUE1 GID2 C2 VALUE2
SPCADD SID SID1 SID2 SID3 SID4 -etc.-
*Alternate Forms
1 2 3 4 5 6 7 8 9 10
SPC1 SID C GID1 GID2 GID3 GID4 GID5 GID6
GID7 GID8 GID9 -etc.-
or
SPC1 SID C GID1 THRU GID2
Bulk Data Image 6-1 Single Point Constraint Specification
where
Field Description
SID SPC set identification number (integer, > 0)
Pair defining external degree of freedom number. GID is the grid or scalar
point identification number (integer, > 0) and C is the degree of freedom
(GIDi, Ci)
code. C may be a packed integer number 0 < C < 6 to specify all degrees
of freedom at G that have the value given by D.
VALUEi Value of the prescribed displacement (real).
SID1, SID2, ... SPC Set identification numbers
Nastran Primer 185

6
6.11.4 Selecting SPC Sets in Case Control

Several sets of single point constraints can be defined in Bulk Data where each set is given
a unique set identification number. The SPC set that is to be used for a given Subcase is then
specified by the Case Control directive
SPC = <SID>
where SID is the integer number assigned to a SPC set in Bulk Data.
SPC sets can be combined into a single logical set by SPCADD. The SPC sets specified
by set identification numbers SID1, SID2, ... ,etc. are combined and are given a unique SID on
the SPCADD, which is then selected by the SPC Case Control directive.
6.11.5 Specifying s-set Degrees of Freedom and Value

The SPC Bulk Data perform two independent functions. These are
1.Purging nonexistent degrees of freedom.
2.Setting displacement boundary conditions.
The form of the SPC data statement supports these two functions by
1.Defining degrees of freedom to be included in the s-set.
2.Specifying a value for each degree of freedom either implicitly or explicitly.
The SPC data statement is used to explicitly set the degree(s) of freedom defined by the
pair (GID, C) to the value, VALUE, where
GID is a grid or scalar point number.

C is a packed set of one to six degrees of freedom codes at point GID.
All of the C degrees of freedom at grid point GID then are set equal to the value VALUE
The constraint value of a degree of freedom specified by SPC1 is implicitly zero. The
degrees of freedom specified by the packed code C are set equal to zero at all the grid points
listed on the remainder of the SPC1 entity. SPC1 can be continued whereas SPC cannot. A set
of points can be defined using the second form of SPC1 using of the literal string THRU. The
total set of degrees of freedom included in the s-set is taken to be the union of all degrees of
freedom defined on SPC-type Bulk Data and those defined as permanent single point
constraints (PSPC) on GRID Bulk Data.
6.11.6 Purging Degrees of Freedom

It is frequently necessary to purge degrees of freedom from the solution set. Typically
these are degrees of freedom defined for a GRID point but which are not connected by the
finite element model. Examples would include out-of-plane behavior for plane problems, the
rotational degrees of freedom for membranes and solids, and the normal rotation for plates and
shells.
Purging can be accomplished by simply including all unconnected degrees of freedom in
the s-set. Suppose, for example, that only unconnected degrees of freedom are included in the
s-set. Then, since there is no stiffness associated with these degrees of freedom, all of the
matrix partitions in (7-41) associated with the s-set are null. The reduced set of equations is
then given by (7-39) where { P f } independent of the prescribed values { Y s } , since [ K fs ] is null.
186 Nastran Primer

6.11.7 AUTOSPC - Automatic Purging

In large analysis problems the task identifying degrees of freedom which are unconnected
in the system stiffness matrix can be difficult. NASTRAN thus includes the capability of
automatically detecting singular degrees of freedom which can then be removed under user
control. The process, which takes place after the unconstrained stiffness matrix [ K gg ] has been
generated, is controlled by the AUTOSPC parameter. Both Legacy and MSC NASTRAN and
MSC include a capability for detecting essential singularities but the procedure differs as does
the form of the AUTOSPC parameter.
The procedures described below do not deal with the finding linear dependences that are
due to rigid body motion or one or more components of the structure. These can only be
determined during the matrix solve operation in static analysis.
6.11.7.1 Legacy NASTRAN

Legacy NASTRAN find the rank of the 3x3 matrices associated with displacement and
rotational degrees of freedom on the diagonal of the [ K gg ] . The AUTOSPC parameter then
controls the action taken if singularities that are not removed by existing SPC or MPC Bulk
Data. The AUTOSPC parameter has the following form
1 2 3
PARAM AUTOSPC value
where value is an integer number whose default value is zero. If singularities are found
and no AUTOSPC parameter having a non zero value is present then a list of singularities is
printed in the output file and program execution proceeds and will fail in the matrix solve phase
due to one or more singular columns, The best advice is to include
1 2 3
PARAM AUTOSPC 1
in the Bulk Data section.
Nastran Primer 187

6
6.11.7.2 MSC NASTRAN

The mathematical procedure for detecting potential singularities looks at the eigenvalues
of the 3x3 matrices for displacement and rotational degrees of freedom on the diagonal of
[ K gg ] . In essence the inclusion of a PARAM as shown below with the name AUTOSPC and
a value of YES will cause NASTRAN to remove unconnected degrees of freedom. The default
value of the parameter is “YES” in the MSC solution sequences so no user action is required
to remove essential singularities
1 2 3
PARAM AUTOSPC YES
6.11.8 Example: Specification of Single Point Constraints

Consider the cantilever beam in Figure 6-1.
y
u y2 = 2 in.
1 EI 2
x
L
Figure 6-1 Enforced Displacement of a Beam
The beam is built-in at the left end can deflect only in the xy-plane. An enforced
displacement uy2 = 2 in. is to be specified at the right end of the beam. The constraints are
specified by the following SPC:
SPC1 1 1345 1 2
SPC1 2 26 1
SPC 3 2 2 .02
SPCADD 4 1 2 3
where:
1.The SPC1having SID = 1 retains only displacements in the xy-plane,
2.The SPC1having SID = 2 constrains ux and θz at Grid 1,
3.The SPC having SID = 3 specifies the enforced displacement at Grid 2,
4.The SPCADD having SID = 4 combines the previous sets of constraints
The combined set of constraints would be selected by a SPC = 4 Case Control directive.
It should be noted that the out-of-plane displacements could also be removed by using a PSPC
having a value of 1345 at all Grid points.
188 Nastran Primer

Static Condensation -OMIT and ASET
6.12 Static Condensation -OMIT and ASET

Static condensation is a basic transformation used for retaining a boundary set in
substructures and as an effective transformation for dynamics. Static condensation in
substructures is used to represent the boundary displacements of a component in terms of
displacements internal to the component. The internal displacements are partitioned into the o-
set and removed from the analysis set.
The intent in dynamics is entirely different. In dynamics, static condensation results in a
transformation that is useful for reducing the number of dynamic degrees of freedom by
redistributing mass and stiffness coefficients to a smaller number of degrees of freedom. The
mathematical procedure is the same for both applications: the f-set is partitioned into the o-set
which is to be removed from the analysis; and, the a-set which is retained. In normal modes
analysis the a-set is the solution set in which eigenvalues and eigenvectors are extracted as
described a a later chapter. In statics, the a-set may be further partitioned by SUPORT as
described in the next section.
6.12.1 Reduction to the a-Set

The first step in static condensation is to partition the f-set into two complementary sets:
the o-set that will be omitted from the analysis, and the a-set hat will be retained in the analysis.
Partitioning { u f } gives:
 
 { uo } 
{ uf } =   (Eq. 6-1)
 { ua } 
 
The reduced stiffness matrix associated with the { u f } set is written in partitioned form as
   
[ K oo ] [ K oa ]  { u o }   { Po } 
  =   (Eq. 6-2)
[ K ao ] [ K aa ]  { u a }   { Pa } 
   
The transformation between the { u o } and { u a } displacement sets, which is used in the
reduction process, is determined by solving for { u o } from the first of the two matrix equations
in (Eq. 6-2):
{ u o } = [ G oa ] { u a } (Eq. 6-3)
where
T
[ G oa ] = – [ K oo ] [ K oa ] (Eq. 6-4)
T
where the term – [ K oo ] { P o } which is associated with the loads on the o-set will be added
during the recovery of the o-set displacements as shown by(Eq. 6-10).
The {uf} set can be expressed entirely in terms of {ua}by substituting (Eq. 6-3) into
(Eq. 6-1) giving:
[ G oa ]
{ uf } = { ua } (Eq. 6-5)
[ I aa ]
Nastran Primer 189

6
or
{ u f } = [ G fa ] { u a } (Eq. 6-6)
where [ I aa ] is an identity matrix of a-size. The reduced set of equations for the a-set is then
given by:
[ K aa ] { u a } = { P a } (Eq. 6-7)
where
T
[ K aa ] = [ G fa ] [ K ff ] [ G fa ] (Eq. 6-8)
and
T
{ P a } = [ G fa ] { P f } (Eq. 6-9)
6.12.2 Recovery of Omitted Degrees of Freedom

After the displacements in the a-set are determined, the o-set is recovered by augmenting
the displacements calculated by means of(Eq. 6-3) with the displacements produced by the
loads { P o } . The o-set displacements are given by:
o
{ u o } = [ G oa ] { u a } + { u o } (Eq. 6-10)
where
o –1
{ u o } = [ K oo ] { P o } (Eq. 6-11)
6.12.3 Physical Interpretation of [Goa]

The transformation matrix [ G oa ] (Eq. 6-4), that relates the omitted degrees of freedom to
those retained in the analysis has a useful physical interpretation. The columns of are static
displacement modes of the structure. That is each column represents the static displacement of
the entire o-set for a unit displacement of the a-set degree of freedom associated with that
column, with all other a-set degrees of freedom set equal to zero.
6.12.4 Why Use Static Condensation?

The purpose in going through this rather straight forward derivation is to show you how
simple the idea of static condensation really is. It is a partitioning operation that is exact for
statics that leads to a smaller solution set. The partitioning is defined by defining either the o-
set or the a-set. The question is why it might be useful to define a reduced set of analysis
degrees-of-freedom.
In your mind's eye, think about a large structure, such as a large commercial airliner. Let's
say that your task is to analyze the wing due to a variety of flight loads. How do you support
the wing? You could choose to fix the points where the wing connects to the fuselage. But that
support condition doesn't properly represent the elastic compliance of the fuselage structure.
190 Nastran Primer

The use of a model of both the fuselage and wing will have the correct compliance but
may result in a very large analysis model. However, if static condensation was used on the
fuselage model to reduce the stiffness to the boundary displacements between the two
structures, the compliance would be correct and the model sized reduced. This is the basic idea
in the using components in NASTRAN.
After completing the stress analysis of a component, such as a wing, the next task might
be to determine the ten lowest modes and associated natural frequencies.
Here the problem is one of problem size and computer resources. The static stress analysis
model for the wing might have 500,000 to 1,000,000 degrees-of-freedom. Based on the fact
that more than one normal mode is required, one of several eigenvalue routines such as Givens,
Modified Givens or Lanczos might be employed. Assuming the model will be used for design
trade-off studies a reduced solution set size is desirable. The use a static condensation to reduce
the size of the analysis set is one option.
The transformation represented by (Eq. 6-3) might be of interest. It describes how a
“dependent” set of displacements, the o-set, responds, statically, to displacements in the a-set,
where the transformation is based solely on the stiffness of the system. As a matter of fact,
think about the a-set as a displacement vector which is all zeros except for a single component
which has a unit value. This little thought experiment tells us that the columns of the
transformation [ G oa ] are the “static displacement modes” of the o-set relative to the a-set.
Guyan, [1], recognized that the "static displacement modes", if properly chosen, could be
used in dynamics to reduce the number of dynamic degrees-of-freedom. It was such a good
idea that the technique is now a standard tool in the dynamic analyst’s bag of tricks and is
called Guyan reduction. Guyan reduction can lead to accurate dynamic response provided the
o-set is properly selected as described in the chapter on normal modes.
6.12.5 The OMIT and ASET Data Entities

Static condensation is specified by the inclusion of either OMIT or ASET entities,
described below by Bulk Data Image 6-1 and Bulk Data Image 6-2, respectively:.
1 2 3 4 5 6 7 8 9 10
OMIT G C G C G C G C
*Alternate Forms
OMIT1 C G1 G2 G3 G4 G5 G6 G7 + OM1
+OM1 G8 G8 -etc.-
or
OMIT1 C GI THRU GJ
where
Gj Grid or scalar point identification number (integer, > 0)

C Degree of freedom code, an integer if G is a geometric point or zero or blank if G
is a scalar point.
Bulk Data Image 6-1 Matrix Reduction - OMIT
Nastran Primer 191

6
OMIT is used to specify degrees of freedom to be partitioned into the removed o-set and
the ASET specifies degrees of freedom to be retained in the a-set. Either OMIT or ASET can
be used.
1 2 3 4 5 6 7 8 9 10
ASET G C G C G C G C
*Alternate Forms
ASET1 C G1 G2 G3 G4 G5 G6 G7 + A1
+A1 G8 G8 -etc.-
or
ASET1 C GI THRU GJ
Bulk Data Image 6-2 Matrix Partitioning - ASET
6.12.6 Specifying Degrees of Freedom

There are alternate forms of the OMIT or ASET called OMIT1 or ASET1s. The OMIT
and ASET data statements allow the user to specify a degree of freedom by the external degree
of freedom number, which consists of the pair (G, C) where
G is the grid or scalar point number
and
C is one or more degree of freedom codes at point G.
Degrees of freedom specified on OMIT are partitioned into the o-set while those specified
on ASET data statements are partitioned into the a-set. Note that neither the OMIT nor ASET
data statements can be continued.
The OMIT1 and ASET1 allow the user to define degrees of freedom to be partitioned into
the o- or a-sets, respectively, by specifying the degrees of freedom using the degree of freedom
codes, C, followed by all the Grid points at which the degree(s) of freedom are to be omitted
or retained. This form of the OMIT1 and ASET1 data statements can have continuations.
Finally, both the OMIT1 and ASET1 data a statements have an alternative form that
allows the user to specify an inclusive set of grid and scalar points by use of the literal string
THRU in the field separating the first and last grid or scalar point.
192 Nastran Primer

Support for Free Bodies - SUPORT
6.12.7 Example: Static Condensation

Consider the finite element model shown inFigure 6-2, in which we wish to place all
degrees of freedom associated with the interior grid points in the o-set.
z
21 22 23 24 25 y
16 17 18 19 20
11 12 13 14 15
6 7 8 9 10
1 2 3 4 5
x
Figure 6-2 Elimination of Internal Degrees of Freedom using OMIT
The structure is that of a plate, and it is assumed that the rotation about the normal has
been purged by use of single point constraints. The following Bulk Data then reduce the
degrees of freedom in the analysis set to those associated with the boundary.
1 2 3 4 5 6 7 8 9 10
OMIT1 12345 7 8 9 12 13 14 17 + O1
+ O1 18 19
where the number in the second field specifies the degrees of freedom that are to be included
in the o-set for each of the grid points listed. Conversely, the following ASET1 data statements
could be used
1 2 3 4 5 6 7 8 9 10
ASET1 12345 1 THRU 5
ASET1 12345 6 10 11 15 16 20
ASET1 12345 21 THRU 25
For this example the o-set is smaller than the a-set, and it makes sense to use the OMIT type
data statement. One could visualize the case where the converse would be true and then the
ASET data statement should be used.
6.13 Support for Free Bodies - SUPORT

An elastic body that is capable of undergoing rigid body motion is termed a free body.
Free elastic bodies are capable of motion that produces no internal forces in the body. The
stiffness matrix for a free body is singular and the structure must be supported to remove the
rigid body motion for static analysis.
The required constraints could be specified by single point or multi point constraints, or
by means of free body supports. The advantage in using free body supports is that the rigid
body characteristics are calculated and the supports are checked for sufficiency.
Nastran Primer 193

6
The singularity can be removed by restraining sufficient degrees of freedom {ur}to

eliminate rigid body motion of the structure without introducing redundant reactions. The set
of forces associated with the r-set thus must be statically determinate.
6.13.1 Reduction to the l-Set

The stiffness equation for the a-set(Eq. 6-7) can be written in partitioned form as
[ K̃ rr ] [ K rl ] { ur } { Pr }  { Cr } 
= +  (Eq. 6-12)
[ K lr ] [ K ll ] { ul } { Pl }  0 
where{ur}is a set of displacements that if constrained would remove rigid body motion; {ul}
is the set left over after all constraint and partition operations and is the solution set for static
analysis; and, { C r } are constraint forces associated with the r-set and where the a-set is
partitioned as follows
{ ur }
{ ua } = (Eq. 6-13)
{ ul }
The r-set is specified on the SUPORT data statement described by Bulk Data Image 6-3. If the
displacements of the r-set and are set equal to zero we have, from the second of (Eq. 6-12),
[ K ll ] { u l } = { P l } (Eq. 6-14)
The stiffness matrix [ K ll ] is non singular so that (Eq. 6-14) can be solved for the {ul} set.
6.13.2 Rigid Body Transformation Matrix

Although the SUPORTed degrees of freedom, {ur}, are set equal to zero in reducing to
the solution set, {ul}, this operation differs from that associated with the use of SPCs in a very
fundamental way. The forces of constraint associated with the r-set are equal to zero. Thus,
from the first of (Eq. 6-12) we find
[ C r ] = 0 = – { P r } + [ K rl ] { u l } (Eq. 6-15)
The forces applied to the SUPORTed degrees of freedom, {Pr}, can now be expressed in
terms of the loads applied to the solution set, {Pl}, by noting that {ul}can be found from
(Eq. 6-14) as
–1
{ u l } = [ K ll ] { P l } (Eq. 6-16)
so that
{ P r } = – [ G rl ] { P l } (Eq. 6-17)
where [ G rl ] is called the rigid body transformation matrix and is given by
–1
[ G rl ] = – [ K rl ] [ K ll ] (Eq. 6-18)
194 Nastran Primer

It is left as an exercise to show that, in the absence of external loads, the motion of the
f-set is related to motion of the r-set through the following relation
T
{ u l } = [ G lr ] { u r } = [ G rl ] { u r } (Eq. 6-19)
6.13.3 Rigid Body Stiffness Matrix

The SUPORTed degrees of freedom, { u r } , define the rigid body motion of the structure.
Therefore, the reduced stiffness matrix for the r-set should be null. The reduced matrix is found
by noting that rigid body motion of the { u l } is related to r-set displacements by(Eq. 6-19) so
we can write (Eq. 6-13) for rigid body motion as
{ u a } = [ G ar ] { u r } (Eq. 6-20)
where the rigid body transformation matrix is
[ I rr ]
[ G ar ] = (Eq. 6-21)
[ G lr ]
Then, the reduced stiffness matrix [ K rr ] is given by
T
[ K rr ] = [ G ar ] [ K aa ] [ G ar ] (Eq. 6-22)
Representing [Kaa] in partitioned form, the substitution of (Eq. 6-21) into (Eq. 6-22) then gives
[ K rr ] = [ K˜rr ] + [ K rl ] [ G lr ] (Eq. 6-23)
where [ G rl ] is given by (Eq. 6-18). The reduced matrix [ K rr ] should be completely null at this
point. NASTRAN calculates the rigid body error ratio
K rr
ε = ------------ (Eq. 6-24)
K̃ rr
and the strain energy associated with giving the r-set degrees of freedom unit displacements or
rotations where the notation K rr means the Euclidean norm of the matrix. These quantities
are automatically printed after they are calculated. The rigid body error ratio and the strain
energy should be zero, neglecting round-off error, if a compatible set of statically determinate
supports have been chosen by the user. These quantities may be nonzero for any of the
following reasons.
1.Round off error accumulation.
2.The { u r } set is over-determined leading to redundant supports (high strain energy).
3.The { u r } set is under-specified leading to a singular reduced stiffness matrix (high
rigid body error ratio).
4.The multipoint constraints are incompatible (high strain energy and high rigid body
error ratio).
5.There are too many single point constraints (high strain energy and high rigid body
error ratio).
Nastran Primer 195

6
Flexibility to Stiffness Transformation
6. [ K rr ] is null (unit value for rigid body error but low strain energy). This is an
acceptable condition and may occur when generalized dynamic reduction is used.
NASTRAN can not determine the modeling error. The user must review the prescribed
constraints to determine possible changes in the finite element model.
The structure could be restrained against rigid body motion by single point constraints,
but there are advantages to using the SUPORT option. These are
1.The [ G rl ] -matrix is used to calculate the rigid body mass matrix for dynamic
analysis and static analysis with inertia relief.
2.The NASTRAN program automatically calculates and prints diagnostic data
concerning sufficiency of constraints.
6.13.4 SUPORT Bulk Data

The SUPORT data statement is used to specify the degrees of freedom {ur}that will
remove rigid body motion.SUPORT is shown by Bulk Data Image 6-3
1 2 3 4 5 6 7 8 9 10
SUPORT GID1 C1 GID2 C2 GID3 C3 GID4 C4
where
GIDi Grid or scalar point identification (integer, > 0)
Ci Component number integer 0 < C < 6 or blank if G is a scalar point.
Bulk Data Image 6-3 Specification of Statically Determinate Supports.
6.13.5 Specifying Degrees of Freedom

The user specifies the degrees of freedom that will restrain rigid body motion by the
external degrees of freedom numbers, which consist of the pair (G, C) where G is the ID of a
grid or scalar point and C is a packed set of one or more degree od freedom codes.
6.14 Flexibility to Stiffness Transformation

The stiffness equation for an element was found to be of the form
[ K ee ] { u e } = { P e } (Eq. 6-25)
There is a complementary approach based on flexibility whereby we find a set of relations in

the form
[F ]{ P} = { u} (Eq. 6-26)
where [ F ] is called the flexibility matrix. The stiffness coefficients kij are the forces at degrees
of freedom, i, due to unit displacement of degree of freedom, j, holding all other displacements
equal to zero. It is tempting to use an analogous approach to define the flexibility coefficients,
however a force cannot be specified for an unconstrained element. Since there are no support
forces the equilibrium conditions for the element would not be satisfied so that flexibility
coefficients cannot be defined for an unsupported element.
196 Nastran Primer

As an alternative, consider an element supported in such a manner that the support forces
can be expressed in terms of the forces at the unsupported degrees of freedom. Structures that
are supported in this manner are called statically determinate because the support forces can
be evaluated solely by static equilibrium considerations.
The flexibility equations for a statically determinate element are
[ F 11 ] [ F 12 ] { P1 } { u1 }
= (Eq. 6-27)
[ F 12 ] [ F 22 ] { P2 } { u2 }
The flexibility coefficients are then seen to be the displacements of degree of freedom i
due to a unit force at and in the direction of degree of freedom j with all other forces equal to
zero. It is easy to show that the flexibility matrices for the simply supported and cantilever
beam which are shown by Figure 6-3 and Figure 6-4 are given by (Eq. 6-28) and (Eq. 6-29),
respectively.
y, u 1
u3 P 4, u 4
1
x
P 2, u 2
   
 u2  l - 2 –1  P2 
  =
6EI – 1 2  P 
-------- (Eq. 6-28)
 u4  4
   
Figure 6-3 Flexibility equation for simple supported beam
y, u 1 u 3, P 3
u2
1 2 u 4, P 4
x
   
 u3  l - 2l 2 3l  P 3 
  = --------
6EI   (Eq. 6-29)
 u4  3l 6  P 4 
   
Figure 6-4 Flexibility equation for cantilever beam
Nastran Primer 197

6
The flexibility coefficients are much easier to obtain by a laboratory test than the stiffness
coefficients. All that is required for the flexibility approach is mounting of the structure in a
statically determinate manner. Loads are then applied at each of the selected node points, and
the displacements at all node point degrees of freedom are measured using appropriate
instrumentation. These measurements are then scaled by the magnitude of the applied load to
give the flexibility coefficients
While the analysis problem can be formulated using a flexibility approach, the stiffness
approach has found almost universal acceptance and is the only approach implemented in
NASTRAN. A method is therefore required to transform element flexibility coefficients into
an equivalent stiffness matrix.
We proceed by first making the observation that
1.The stiffness matrix is singular because the node point displacements include rigid
body motion.
2.The flexibility matrix is formed for a structural element that is statically
determinate.
These two statements are really equivalent and the transformation of the flexibility coefficients
to a set of stiffness coefficients for the elements must include the rigid body degrees of
freedom.
Let us consider the set of all grid point displacements for the element as the union of two
subsets; the set of degrees of freedom {ud}and {ui}where
{ud} are a set of degree of freedoms that would eliminate rigid body motion if
constrained.
and
{ ui } are unconstrained degrees of freedom that may displace relative to the d-set
The flexibility equations relate the forces and displacements of the i-set so that
[ F ii ] { P i } = { u i } (Eq. 6-30)
The stiffness equations relate the forces and displacements for both the d- and i-sets. The
stiffness equations can thus be represented in partitioned form as
[ K ii ] [ K id ] { ui } { Pi }
= (Eq. 6-31)
[ K di ] [ K dd ] { ud } { Pd }
Setting the set of displacements { u d } equal to zero then gives
[ K ii ] { u i } = { P i } (Eq. 6-32)
A comparison of (Eq. 6-32) with (Eq. 6-30) shows that the [ K ii ] partition of the stiffness
matrix is given by
–1
[ K ii ] = [ F ii ] (Eq. 6-33)
198 Nastran Primer

In order to determine the other partitions of the stiffness matrix we must include
information about the relation between the { u i } and { u d } displacement sets, since all the other
partitions involve the { u d } set. In the absence of forces { P i } , the element may displace as a
rigid body so that a relation between the { u i } and the { u d } degrees of freedom must exist in the
form
r
{ u i } = [ S id ] { u d } (Eq. 6-34)
If the element is subjected to a set of forces { P i } and the { u d } set is not equal to zero, then
the total displacement of the { u i } set is the sum of the displacements given by
f r
{ ui } = { ui } + { ui } (Eq. 6-35)
f r
where { u i } are the displacements due to structural flexibility and { u i } represent rigid body
motion relative to the d-set. The substitution of equations (Eq. 6-30) and (Eq. 6-34)
into(Eq. 6-35) gives
{ u i } = [ F ii ] { P i } + [ S id ] { u d } (Eq. 6-36)
which can be put into the form
–1 –1
[ F ii ] { u i } – [ F ii ] [ S id ] { u d } = { P i } (Eq. 6-37)
Now, comparing the first of (Eq. 6-31) with (Eq. 6-37) we see that
–1
[ K id ] = – [ F ii ] [ S id ] (Eq. 6-38)
and, since the stiffness matrix and flexibility matrices are symmetric, we have
T –1
[ K di ] = – [ S id ] [ F ii ] (Eq. 6-39)
The final part of the flexibility to stiffness transformation is the determination of the
partition, [ K dd ] , which relates the { u d } displacements to the forces, { P d } . This can be done
by noting that the flexibility relations for an element are obtained by specifying that { u d } are
constrained so that the element is statically determinate. This means that the forces, { P d } , can
be determined from { P i } by a relation of the form
{ P d } = [ R di ] { P i } (Eq. 6-40)
The substitution of(Eq. 6-37) into (Eq. 6-40) then gives
–1 –1
{ P d } = [ R di ] [ F ii ] { u i } – [ R di ] [ F ii ] [ S id ] { u d } (Eq. 6-41)
The [ R di ] and [ S id ] matrices are related. The relation is implicit if we note that [ K di ] given
by (7-81) as a result of symmetry and is also obtained from (Eq. 6-41) as
–1
[ K di ] = [ R di ] [ F ii ] (Eq. 6-42)
A comparison of these two expressions shows that
Nastran Primer 199

6
T
[ R di ] = – [ S id ] (Eq. 6-43)
where we have used (Eq. 6-39) to relate [ R di ] and [ S id ] . A more direct way of showing that the
relation holds is to note that the virtual work of the external forces must vanish if we allow only
rigid body motion.
T T
{ P i } { δu i } + { P d } { δu d } = 0 (Eq. 6-44)
Then we have from (Eq. 6-40))
T T T
{ P d } = { P i } [ R di ] (Eq. 6-45)
and, from (Eq. 6-34),
{ δu i } = [ S id ] { δu d } (Eq. 6-46)
The substitution of (Eq. 6-45)and (Eq. 6-46) into (Eq. 6-44)then gives
T
[ R di ] = – [ S id ]
Finally, the coefficients of { u d } are the [ K dd ] partition and are given by
T –1
[ K dd ] = [ S id ] [ F ii ] [ S id ] (Eq. 6-47)
The element stiffness matrix can thus be formed from the element flexibility, [ F ] , and
either [ S ] or [ R ] as follows
[ K ii ] [ K id ] [F]
–1 –1
–[ F ] [ S ]
[K] = = (Eq. 6-48)
T
[ K id ] [ K dd ] T
–[ S ] [ F ]
–1 T
[S ] [ F][ S]
The stiffness equations for an element can be generated using the flexibility and rigid
body transformation matrices specified using GENEL Bulk Data as described in the next
section.
6.14.1 GENEL Bulk Data

The GENEL contains four logical sections that
1.Specify flexible degrees of freedom that comprise the i-set. These degrees of
freedom must be defined using GRID or SPOINT Bulk Data.
2.Specify the restrained degrees of freedom.
3.List the coefficients of the flexibility matrix.
4.List the coefficients of the rigid body transformation matrix.
*Define degrees of freedom in the i-set
1 2 3 4 5 6 7 8 9 10
GENEL EID GI1 CI1 GI2 CI2 GI3 CI3 +1
+1 GI4 CI4 etc. +2
200 Nastran Primer

*Define degrees of freedom in the d-set
+2 UD GD1 CD1 GD2 CD2 GD3 CD3 +3

+3 GD4 CD4 etc. +4
*Define elements of flexibility matrix by columns
+4 Z Z11 Z21 Z31 etc. Z22 Z32 Z42 +5

+5 etc. Z33 Z43 Z53 etc. Z44 Z54 etc. +6
+6 etc. +7
*Define Rigid Body Transformation Matrix by Rows
+7 S S11 S12 S13 etc. S21 S22 S23 +B

+8 etc. S33 S32 etc.
where
Field Description
EID Unique element number (integer, > 0).
Pairs defining the degrees of freedom that comprise the i-set of
GIk,CIk
displacements for the element (integer).
Characters in field two, which defines the beginning of the list of d-set
UD
degrees of freedom.
GDk, CDk Pairs defining the degrees of freedom that comprise the d-set.
Z Character in field two that defines the beginning of flexibility coefficients.
Elements of the flexibility matrix by columns starting with the diagonal in
Zmn
the order of the list for the i-set.
Character in field two, which defines the beginning of rigid body
S
coefficients.
Elements of the rigid body transformation matrix by rows in the order of
Smn
the list for the d-set.
Bulk Data Image 6-4 Flexibility to Stiffness Transformation - GENEL
6.14.2 Example:Beam Stiffness Using GENEL

The stiffness properties for the beam element can be defined by using either (Eq. 6-28)
or(Eq. 6-29) together with the appropriate rigid body transformation matrix. We therefore
consider the problem of determining the response of a beam supported as a cantilever using the
flexibility coefficients for the simple supported beam as given by (Eq. 6-28).
Let the beam properties be given as
EI = 1
L = 6
Then the flexibility coefficients (Eq. 6-26) become
Nastran Primer 201

6
References
[F] = 2 –1
–1 2
The rigid body transformation matrix is obtained by nothing that
( u3 – u1 )
u 2 = u 3 = ---------------------
L
so that [ S id ] is given by
1
[ S id ] = --- – 1 1
L –1 1
We will define the degrees of freedom connected by the beam by SPOINT rather than by
Grid points using the displacement numbers defined on Figure 6-4. The bulk data required to
define the beam element and the constraints and apply a unit force at the tip are:
1 2 3 4 5 6 7 8 9 10
SPOINT 1 THRU 4
GENEL 2 4 +G1
+G1 UD 1 3 +G2
+G2 Z 2. -1. 2. +G3
+G3 S -.1667 1. -.1667 1.
SPC 100 1 0.
SPC 100 2 0.
SLOAD 100 3 1.
6.14.3 Direct Specification of Element Stiffness

The GENEL can also be used in an alternative form for specifying the stiffness
coefficients associated with the partitioned equations
   
 { Fi }  [ K ii ] – [ K ii ] [ S id ]  { ui } 
  =   (6-49)
 { Fd }  T T
– [ S id ] [ K ii ] [ S id ] [ K ii ] [ S id ]  { ud } 
   
The only change to the GENEL data statement is the replacement of the character ‘Z’ by ‘K’
on continuation statement ‘+4’ shown in Bulk Data Image 6-4. The coefficients are then taken
to be the elements of the stiffness partition [ K ii ] rather than the flexibility matrix.
The GENEL element thus provides the user with a means of defining elements of the
global stiffness matrix directly without using a finite element model. The coefficients can be
defined either in terms of stiffness or flexibility.
6.15 References
[1] R. Guyan, “Reduction of stiffness and mass matrices”, AIAA J., 3(2), p 380, 1965
202 Nastran Primer

Problems
6.16 Problems
7.1 The Lagrange multiplier approach to the incorporation of linear constraint equation involves the
definition of a set of workless constraint forces { λ m } . Using the fact that the constraint forces
produce no virtual work show that the augmented stiffness equations can be written as
[ K gg ] [ G mg ]  { u }   { Pg } 
T
  =  
[ G mg ] [ 0 ]  { λ m }   O 
7.2 Given the augmented set of equations from problem 7.1 show that { λ m } can be found as
–1
{ λm } = [ G ] [ K ]–1 [ G ]T  [G ] [K ] {P } 
mg gg mg  mg gg g 
7.3 Suppose that a rigid link is to connect node points 1 and 2 as shown below where there is no
restraint to rotation θy2,
u y2
u x2
2
y u y1 θ z2
u x1 Rigid Element
1
θ z1
Assuming that the nodes are located the basic coordinate system at X1 = (1.,1.,0.) and X2 =
(3.,5.,0.), write the set of constraint relations which define the motion of point 2 in terms of the
displacements at point 1.
7.4 Given the system of axial members shown below show that the augmented set of equations for
the constraint condition u2 - u3 = 0 are as follows
Nastran Primer 203

6
Problems
2k o – k o 1  u 2   P2 
   
– k o 2k o – 1  u 3  =  P 3 
   
1 –1 0  λ   0 
where k o = AE
-------
L
1 1 2 2 3 3 4
P2 P3
L L L
AE = constant
7.5 Show that the Lagrange multiplier for problem 7.4 is λ = 1/2(P2-P3). What is the physical
interpretation of λ?
7.6 Using the stiffness to flexibility relation (Eq. 7-90) show that the correct element stiffness matrix
for the beam is obtained when starting from the flexibility for either the simply supported beam
(Eq. 7-70) or the cantilever beam (Eq. 7-71).
7.7 The stiffness matrix for the beam element is singular because the node point displacements
include rigid body motion. Show that the reduced stiffness matrix associated with including the
transverse displacements at each end of the beam in the r-set is null.
7.8 Verify that (Eq. 7-82) relates the rigid body displacements of the f-set to rigid body motion of the
r-set displacement.
204 Nastran Primer

7 Accuracy and Performance of Finite
Elements
7.1 Element Performance

MacNeal[1] describes a simple test called the patch test, first proposed by Irons [2], for
studying the performance of finite elements. The idea behind the test is that under simple
loading conditions the finite element should be capable of representing exact behavior. For
example, if a two dimensional region is subject to a uniform boundary traction, an arbitrary
element should represent the correct value of strain and therefore stress. MacNeal [3] shows
that the patch test imposes the following set of sufficient conditions on the element
formulation:
1.Strains and stress computed from nodal point displacements are correct to the
desired order
2.The integral relating the generalized forces to stresses
{ Fi } = ∫ [ Bi ] { σ } dΩ (Eq. 7-1)
Ω
is exact for the desired order of {σ}.
3.Equilibrium at internal nodes is satisfied because either

a) the elements are conforming, or
b) the nonconforming part of the strain-displacement matrix of each element

produces zero grid point forces, or
c) the shape functions associated with interior nodes satisfy the interelement
traction boundary conditions
Failure to satisfy these three conditions is due to interpolation failure, integration failure
or equilibrium failure.
7.2 Interpolation Failure

Interpolation failure is due to the incorrect interpolation of the node point generalized
coordinates. This type of failure is one of the consequences of parametric mapping in
isoparametric elements.
Interpolation failure also causes a phenomenon called locking that is caused by excessive
stiffness for a particular deformation mode of the element. In hindsight locking was
responsible for the poor performance of early lower-order elements and gave impetus to
developing higher-order elements.
205 Nastran Primer

Accuracy and Performance of Finite Elements
7
Effect of Element Shape
The locking phenomenon is still not well understood by finite element developers and
certainly is somewhat of a mystery to the users, especially those who expect analysis fidelity
to be equivalent to geometric fidelity. MacNeal suggests that interpolation failure must be
present for locking to exist; but, this contention is still the subject of future research on the
subject.
Locking is caused by a number of element parameters including loading, element
geometry, part geometry and material properties. Incompressible materials are an obvious
problem since the bulk modulus becomes infinite for an isotropic material when Poisson’s ratio
becomes 0.5. This can be overcome by using a reformulated element or by taking Poisson’s
ratio close to, but not equal to 0.5.
7.3 Effect of Element Shape

Element testing is a vital part of due diligence that must be performed by both the
developer and the user. After-the-fact testing of the original NASTRAN membrane element,
the QUAD2, showed that it gave good results only for rectangular geometry and then only for
a restricted range of element aspect ratio. For non rectangular regions, and for aspect ratios
greater than 2:1 for rectangular regions, the element produced spectacularly poor results.
Due to a concern about element performance in general and the elements included in
MSC.Nastran in particular, MacNeal and Harder [4] proposed a set of standard tests for
element performance that included all of the parameters identified in the proceeding section.
The loading should induce all behavior modes in the element including inplane extension,
bending in two plane and twist of two dimensional elements, for example.
Elements have standard shapes. A quadrilateral element formulation is based on a bi-
square in parameter space and preforms best when the Jacobian of transformation is a unit
matrix, or at least a diagonal matrix. The test geometry must therefore include all types of
geometric anomalies such as those shown by Figure 7-1for a quadrilateral element.
206 Nastran Primer

Aspect Ratio b a⁄b

a
Skew δ
Taper d---
d b
(2 Directions) b
h---
Warp h a
Figure 7-1 Types of Geometric Distortion of a Quadrilateral
In addition to element shape the part geometry also affects the accuracy of the finite
element model and of course dictates the type of element to be employed. Curved shells can
only be approximated by flat surfaces; and, shells with one and two radii of curvature are
chosen for the tests. Even the boundary geometry of a flat assemblage can have on effect which
could be tested using a skewed plate, a swept wing and a rhombic plate.
The test problems described below are called the MacNeal-Harder tests which are
summarized by Table 7-1
Table 7-1 Summary of Tests
Elements appropriate for the Test
Test problem Plate
Beam Plate Bending Shell Solid
Membrane
Patch test X X X
Straight slender X X X X
cantilever beam
Curved Beam X X X X
Twisted Beam X X X X X
Nastran Primer 207

7
Table 7-1 Summary of Tests

Elements appropriate for the Test
Test problem Plate
Beam Plate Bending Shell Solid
Membrane
Rectangular Plate X X X
Scordelis-Lo Cylindrical X X X X
Section
Spherical Shell X X X
Thick-walled cylinder X
7.3.1 Two Dimensional Patch Test

The geometry for the two dimensional patch test is shown by Figure 7-2 where the node
points are defined by Table 7-2.
y
b
4 3
2
1
x
Figure 7-2 Two Dimensional Patch Test Geometry
The dimensions of the rectangular region and material properties are a = 0.12, b = 0.24, t
= 0.001; E = 1.0 x106; ν = 0.25.
Table 7-2 Vertices for 2D Patch
Node x y
1 0.04 0.02
2 0.18 0.03
3 0.16 0.08
4 0.08 0.08
7.3.1.1 Membrane Boundary Conditions

The loads are applied using the following displacement boundary conditions
–3
u = 10 ( x + y ⁄ 2 )
–3
v = 10 ( y + x ⁄ 2 )
208 Nastran Primer

7.3.1.2 Membrane Theoretical Solution

The membrane strain and stress components are
–3
ε xx = ε yy = γ xy = 10
σ xx = σ yy = 1333
τ xy = 400
7.3.1.3 Bending Boundary Conditions

The loads are applied by the following boundary displacements
–3
10 2 2
w = ----------- ( x + xy + y )
2
∂w –3
θx = = 10 ( x + y ⁄ 2 )
∂y
θ y = – ∂w = – 10 ( y + x ⁄ 2 )
–3
∂x
7.3.1.4 Bending Theoretical Solution

The bending moments and surface stresses are
–7
m x = m y = 1.111x10
–7
m xy = 10
σ xx = σ yy = −+ 0.667
τ = −
xy + 0.200
Nastran Primer 209

7
7.3.2 Three Dimensional Patch Test

The geometry for the two dimensional patch test is shown by Figure 7-1 where the node
points are defined by Table 7-1.
8 3
7
1
2
5 6
Figure 7-1 Three Dimensional Patch Test Geometry
The region is a unit cube; E = 1.0 x106; ν = 0.25.

Table 7-1 Vertex Locations for 3D Patch
Node x y z
1 00.249 0.342 0.192
2 0.826 0.288 0.288
3 0.850 0.649 0.263
4 0.273 0.750 0.230
5 0.320 0.186 0.643
6 0.677 0.305 0.683
7 0.788 0.693 0.644
8 0.165 0.745 0.702
The region is subject to the following displacement boundary conditions
–3
u = 10 ( 2x + y + z ) ⁄ 2
–3
v = 10 ( x + 2y + z ) ⁄ 2
–3
w = 10 ( x + y + 2z ) ⁄ 2
210 Nastran Primer

The strain and stress components for the theoretical solution are
–3
ε xx = ε yy = ε zz = 10
–3
γ xy = γ yz = γ zx = 10
σ xx = σ yy = σ zz = 2000
τ xy = τ yz = τ zx = 400
7.3.3 Patch Test Results

The results of the patch test for MSC.Nastran elements is presented by Table 7-2
Table 7-2 Patch Test Results
Element Error in Stress, percent
Constant stress Constant Curvature

Loading Loading
QUAD2 0.0 30.7
QUAD4 0.0 0.0
QUAD8 18.0 51.6
HEXA(8) 0.0 N/A
HEXA(20) 0.0 N/A
where the HEXA(8) and the HEXA(20) are results for the HEXA element with 8 and 20 nodes,
respectively. The QUAD2 element was one of the original elements in MSC.Nastran. Due to
its poor performance it was replaced by the QUAD4 element.
Nastran Primer 211

7
7.3.4 MacNeal Beam Tests

Three types of beams were used by MacNeal: the straight beam, the curved beams and the
twisted beam shown by Figure 7-2 through Figure 7-4 respectively with exact results for four
loading conditions shown by Table 7-3.
(a)
45° 45°
(b)
45°
(c)
Figure 7-2 MacNeal’s Slender Straight Beam.
212 Nastran Primer

90°
FIXED
Figure 7-3 MacNeal’s Curved Beam.
FIXED
END
Figure 7-4 MacNeal’s Twisted Beam.
Table 7-3 Theoretical Solutions for Beam Problems

Tip load direction Displacement in load direction
Straight beam Curved beam Twisted beam
Extension 3.0x10-5
In-plane shear 0.1081 0.08734 0.005424
Out-of-plane shear 0.4321 0.5022 0.001754
Twist 0.03206
Nastran Primer 213

7
7.3.4.1 Results for Slender Straight Beam

The results for the slender straight beam test for MSC.Nastran two and three dimensional
elements by Table 7-1.
Table 7-1 Results for Straight Beam
Tip Load Direction Normalized tip displacement in load direction
QUAD2 QUAD4 QUAD8 HEXA8 HEXA20
(a) Rectangular elements
Extension 0.992 0.995 0.999 0.988 0.994
In-plane shear 0.032 0.904 0.987 0.981 0.970
Out-of-plane shear 0.971 0.986 0.991 0.981 0.961
Twist 0.566 0.041 0.950 0.910 0.904
(b) Trapezoidal elements
Extension 0.992 0.996 0.999 0.989 0.994
In-plane shear 0.016 0.071 0.946 0.069 0.886
Twist 0.616 0.951 0.943 0.906 0.904
(c) Parallelogram elements
Extension 0.992 0.996 0.999 0.989 0.994
In-plane shear 0.014 0.080 0.995 0.080 0.967
Twist 0.615 0.945 0.965 0.910 0.904
7.3.4.2 Results for Curved Beam

The results for the curved beam are shown by Table 7-1.
Table 7-1 Results for Curved Beam
Tip Loading Direction Normalized tip displacement in load direction
In-plane (vertical) 0.025 0.833 1.007 0.880 0.875
7.3.4.3 Results for Twisted Beam

The results for the twisted beam are shown by Table 7-1.
Table 7-1 Results for Twisted Beam
Tip Loading Direction Normalized tip displacement in load direction
In-plane (vertical) 100.4 0.993 0.998 0.983 0.991
214 Nastran Primer

7.3.5 Rectangular Plate Tests

The transversely loaded rectangular plate is shown by Figure 7-1
a sym
sym
Figure 7-1 Rectangular Plate
The plate dimensions and material properties are; a = 2, b = 2 or 10, E = 1.7472e7, ν = 0.3, t =
0.0001 for plates and 0.001 for solids.
The plate is subject to two symmetric loading conditions and two sets of symmetric
boundary conditions. The loads are a uniform normal pressure or a central concentrated force.
The boundary conditions are: all edges simply supported or all edges clamped. Only one
quarter of the plate is modeled using symmetry conditions, and the mesh density can be varied
to study convergence. The theoretical solutions for all cases of loads, boundary conditions and
aspect ratios are presented by Table 7-2
Table 7-2 Theoretical Center Displacement
Displacement at Center of Plate

Boundary Supports Aspect Ratio b/a
Uniform Pressure Concentrated Force
Simple Support 1.0 4.062 11.60
Simple Support 5.0 12.97 16.96
Clamped 1.0 1.26 5.60
Clamped 5.0 2.56 7.23
The results for the simple support cases with uniform load and clamped cases with
concentrated load are shown by Table 7-3 throughTable 7-6.
Table 7-3 Simple Supports, Uniform Load, a/b = 1
Number of Elements per Edge

Element Type
2 4 6 8
QUAD2 0.971 0.995 0.998 0.999
QUAD4 0.981 1.004 1.003 1.002
QUAD8 0.927 0.996 0.999 1.000
HEXA(8) 0.989 0.998 0.999 1.000
HEXA(20) 0.023 0.738 0.967 0.991
Nastran Primer 215

7
Table 7-4 Simple Supports, Uniform Load, a/b = 5

Element Type
2 4 6 8
QUAD2 0.773 0.968 0.993 0.998
QUAD4 1.052 0.991 0.997 0.998
QUAD8 1.223 1.003 1.000 1.000
HEXA(8) 0.955 0.978 0.990 0.995
HEXA(20) 0.028 0.693 1.066 1.026
Table 7-5 Clamped Edges, Concentrated Load, a/b = 1

Element Type
2 4 6 8
QUAD2 0.979 1.008 1.006 1.005
QUAD4 0.934 1.010 1.012 1.010
QUAD8 1.076 0.969 0.992 0.997
HEXA(8) 0.885 0.972 0.988 0.994
HEXA(20) 0.002 0.072 0.552 0.821
Table 7-6 Clamped Edges, Concentrated Load, a/b = 5

Element Type
2 4 6 8
QUAD2 0.333 0.512 0.638 0.723
QUAD4 0.519 0.863 0.940 0.972
QUAD8 0.542 0.754 0.932 0.975
HEXA(8) 0.321 0.850 0.927 0.957
HEXA(20) 0.001 0.041 0.220 0.374
216 Nastran Primer

7.3.6 Cylindrical Shell Tests

The Scordelis-Lo roof [5] is used to test the ability of elements to represent a shell with
one radius of curvature. The shell is shown by Figure 7-2
z
y
SY
M
M
SY
EE
FR L
40°
R
Figure 7-2 Scordelis-Lo Roof
The dimensions and material properties of the roof are; R = 25, L = 50, t = 0.25, E =
4.32e8, ν= 0.0. The shell is loaded by a uniform load of 90 units of force per unit area in the
negative Z direction where the units are consistent among all physical quantities. The boundary
conditions on the curved edges are ux = uz = 0.
The shell is modeled using an N by N mesh of the shaded area. Since the loads, geometry,
material are symmetric, only one quarter of the structure is modeled using appropriate
symmetry conditions.
The theoretical solution for the vertical displacement at the center of the straight edge is
given by [5] to be 0.3086. MacNeal[4] suggest using a slightly different value, 0.3024.
The results of as a function of mesh refinement is presented by Table 7-7.
Table 7-7 Results for Scordelis-Lo Roof

Element Type
2 4 6 8 10
QUAD2 0.784 0.665 0.781 0.854 0.987
QUAD4 1.376 1.050 1.018 1.008 1.004
QUAD8 1.021 0.984 1.002 0.997 0.996
HEXA(8) 1.320 1.028 1.012 1.005
HEXA(20) 0.092 0.258 0.589 0.812
Nastran Primer 217

7
7.3.7 Pinched Sphere

The spherical shell us used to test the ability of elements to represent shell structures
having two radii of curvature. The shell is shown by Figure 7-3.
z 18°
FREE
SY
SYM M
y
F = 2.0
(on quadrant)
FREE
x F = 2.0 (on quadrant)
Figure 7-3 Spherical Shell with Opening
The dimensions and material properties of the shell are; R = 10.0, t = 0.04, E = 6.825e7,
ν = 0.3. The shell is open on the top as shown. The boundaries are free with equal and opposite
concentrated loads applied to quadrants as shown. The geometry and loads are symmetric so
only one quarter of the shell is modeled by an N by N mesh using appropriate symmetry
conditions .
There is no closed form solution for the test case. MacNeal [4] recommends a value of
0.0940 for the radial displacement at the point of load application. The normalize results are
presented as a function of mesh refinement by Table 7-8.
Table 7-8 Spherical Shell with a Central hole

Element Type
2 4 6 8 10 12
QUAD2 0.928 0.990 0.990 0.986 0.984 0.982
QUAD4 0.972 1.024 1.013 1.005 1.001 0.998
218 Nastran Primer

Table 7-8 Spherical Shell with a Central hole

Element Type
2 4 6 8 10 12
QUAD8 0.025 0.121 0.494 0.823 0.955 0.992
HEXA(8) 0.039 0.730 0.955
HEXA(20) 0.001 0.021 0.097
7.3.8 Incompressible Material

A thick cylindrical shell shown by Figure 7-4 is used to test the effect of Poisson’s ratio.
Radius
9.00
6.75
5.20
SYM
4.20
3.50
3.00
FREE
10°
p
SYM
Figure 7-4 Thick Walled Cylinder
The Cylinder has an inner radius, Ri = 3.0, an outer radius, Ro = 9.0, the elastic modulus
is E = 1000 and Poisson’s ratio is varied. The cylinder is loaded with a unit internal pressure.
Plane strain is assumed and the 5 by 1 mesh shown is used for ν = 0.49, 0.499 and 0.4999.
The theoretical result for the radial displacement is given by
2 2
( 1 + ν )pR i R o
- ------ + ( 1 – 2ν )r
u r = -------------------------- (Eq. 7-1)
2 2
E ( Ro – Ri ) r
The radial displacement at the inner surface is presented by Table 7-9.

Table 7-9 Radial Displacement at Inner Surface
Poisson’s Ratio Radial Displacement

0.49 5.0399e-3
0.499 5.0602e-3
0.4999 5.0623e-3
Nastran Primer 219

7
Summary of Test Results
The results for the MSC.Nastran elements are shown by Table 7-10.
Table 7-10 Test Results for Radial Displacement
Poisson’s Ratio
Element Type
0.49 0.499 0.4999
QUAD2 0.643 0.156 0.018
QUAD4 0.846 0.359 0.053
QUAD8 1.0000 0.997 0.967
HEXA(8) 0.986 0.986 0.986
HEXA(20) 0.999 0.986 0.879
7.4 Summary of Test Results

The tests used in this study are by no means complete and should be augmented by others
presented in the literature several of which are described by Table 7-11
Table 7-11 Additional Tests for Finite Elements
Test Name Description

The pinched ring test presented by Sze[7] can be used to test an
Pinched Ring assemblage elements whose stiffness is coupled by the structural
geometry
The swept plate test presented by Cook [16] can be used to test an
Swept Plate
assemblage of elements having variable trapezoidal geometry
The short beam test was proposed by Cheung [10] and is used to test an
Short Beam
assemblage of elements having both skewed and trapezoidal geometry.
Short Beam for Three The short beam test generalized to test three dimensional stress elements
Dimensional Elements by Cheung [11]]
The circular plate has been used by several authors, including[11] to test
Circular Plate
the effect of meshing strategies in best representing circular plates
The skew plate test presented by [17]can be used to test an assemblage of
Skew Plate
skewed elements and to compare the results to theory [15]
The NAFEMS organization was established in Great Briton to develop
NAFEMS
standard tests for elements [13].
These tests for static loading can be used to determine the relative performance of
elements. These tests must be augmented by tests for other types of analyses including normal
modes, transient dynamics, buckling and nonlinear analyses. There are many factors that affect
the ability of the model to simulate the behavior of a virtual part or assemblage. As we rely
more heavily on simulation in design as an alternative to prototype test it becomes more and
more important to use a part of the analysis budget to design and verify the virtual prototype.
The results of the tests presented in this Chapter are encouraging. They are restricted to
only those elements included in NASTRAN since those are the elements we have at our
disposal. The QUAD2 was included to show how poorly early elements performed and to
provide a baseline for showing the improvements in elements that resulted from this type of
testing.
220 Nastran Primer

Summary of Test Results
In [4] the elements are graded based on the error in achieving the exact or target result.
The test results are summarized for two and three dimensional elements by Table 7-12 and
Table 7-13, respectively using the following grading scale:
Grade Rule
A 2% ≥ error
B 10% ≥ error < 2%
C 20% ≥ error < 10%
D 50% ≥ error < 20%
F error > 50%
Table 7-12 Summary of Results for Shell Elements
Element Loading Element Element Type

Test
In-plane Out of plane Shape QUAD2 QUAD4 QUAD8
Patch test X Irregular A A C
Patch test X Irregular D A D
Straight beam, extension X All A A A
Straight beam, bending X Regular F B A
Straight beam, bending X Irregular F F B
Straight beam, bending X Regular B A A
Straight beam, bending X Irregular B B A
Straight beam, twist All D B B
Curved beam X Regular F C A
Curved beam X Regular D B B
Twisted beam X X Regular F A A
Rectangular plate, N = 4 X Regular C B B
Scordelis-Lo X X Regular D B A
Spherical Shell X X Regular A A C
Incompressible Material X F F B
Regular
(ν = 0.4999)
Number of Failures (D’s and F’s) 9 2 1
Table 7-13 Summary of Results for Solid Elements
Element Element Type

Test
Shape Hexa(8) Hexa(20)
Patch test Irregular A A
Straight beam, extension All A A
Straight beam, bending Regular A B
Straight beam, bending, inplane Irregular F B
Straight beam, bending, out of F B
Irregular
plane
Nastran Primer 221

7
Problems
Table 7-13 Summary of Results for Solid Elements
Element Element Type

Test
Shape Hexa(8) Hexa(20)
Straight beam, twist All B B
Curved beam, in plane loading Regular C C
Curved beam, out of plane loading Regular C B
Twisted beam Regular B B
Rectangular plate, N = 4 Regular B F
Scordelis-Lo, N = 4 Regular B F
Spherical Shell, N = 8 Regular D F
Incompressible Material A C
Regular
(ν = 0.4999)
Number of Failures (D’s and F’s) 3 3
The QUAD4 element, performs well for all tests except the curved beam with out of plane
loading. The failure of this test is due to locking. The QUAD8 performs well for all tests except
for the patch test.
The HEXA(8) passes the patch test but performs badly in bending. However, it handles
incompressible materials well. The HEXA(20) does well for beam bending but performs badly
for plate and shell problems.
Elements performance is continuously evaluated by the element designers, and new or
revised formulation are implemented as appropriate to improve element performance.
Certainly, the implementation of p-element technology into MSC/NASTAN is evidence of the
commitment of program developers to provide state of the art performance. There are some
exciting developments in the area of hybrid formulations, such as the Pian-Sumihara
membrane element[6], and significant contributions by Sze[7],[8], To [9] and Cheung [10],
[11] and [14], which will undoubtedly be evaluated and perhaps find their way into commercial
programs in the future.
7.5 Problems
[1] Verify the results for the two dimensional patch test using the geometric and material
parameters given in Sec. 8.3.1 for the membrane and bending displacements given. Normalize
the results using the exact values.
[2] For the 2-D patch, use the geometry given in Sec. 8.3.1 with material properties,
6
E = 30 ×10 psi and ν = 0.3 and determine the stress components normaized by the exact
values.
[3] Verify the results for the three dimensional patch test using the geometric and material
parameters given in Sec. 8.3.2 for the Hexa element with 8 nodes. Normalize the results using
the exact values.
[4] Repeat Problem 3 using the Hexa element with 20 nodes.
[5] Repeat Problem 3 using the Tetra element with 4 nodes.
[6] Repeat Problem 3 using the Tetra element with 10 nodes.
[7] Verify the results for the straight beam test uning the geometry shown in Figure 7-2 for
an extensional load, normalize the results using the exact values.
222 Nastran Primer

References
[8] Repeat Problem 7 for a unit inplane shear at the free end
[9] Repeat Problem 7 for a unit out of plane shear at the free end.
[10] Verify the results for the rectangular plate tests given by Table 7-3 using the Quad4
element
[11] Repeat Problem 10 using the Quad8 with 8 nodes
[12] Repeat Problem 10 using the Tria3 element.
[13] Repeat Problem 10 using the Tria6 with 6 nodes.
[14] Verify the results for Scordelis-Lo given by Table 7-7 for the Hexa with 20 nodes.
7.6 References
[1] R.H. MacNeal, Finite Elements: Their Design and Performance, Marcel Decker, Inc.,
New York, NY, p. 179, 1994.
[2] B.W. Irons, “Numerical integration applied to finite element methods”, Conf. on Use of
Digital Computers in Structural Engineering, Univ. of Newcastle, 1966.
[3] R.H. MacNeal, Finite Elements: Their Design and Performance, Marcel Decker, Inc.,
New York, NY, pp. 193-194, 1994.
[4] R.H. MacNeal and R.L Harder, “A proposed set of standard problems to test finite
element accuracy”, Finite Elements in Analysis and Design, 1, 3-20, 1993.
[5] A.C. Scordelis and K.S. Lo, “Computer analysis of shells”, J. Amer. Concrete Inst. 61,
539-561. 1969.
[6] T.H.H. Pian and K. Sumihara, “Rational approach for assumed stress finite elements,”.
Int. J. Num. Meth. Engr.,20, 1685-95, 1984.
[7] K.Y. Sze, “Hybrid plane quadrilateral element with corner rotations,”. Journal of
Structural Engineering, 119(9), 2552-2572, 1993.
[8] K.Y Sze, “Hybrid hexahedral element for solids, plates, shells and beams by selective
scaling”, Inter. J. Num. Meth. in Engr., 36, 677-693, 1993.
[9] C.W.S. To, “Hybrid strain-based three node flat triangular laminated composite shell
elements”, Finite elements in Analysis and Design, 28(3), 177-208, 1998.
[10] Wanji Chen and Y.K. Cheung, “A new approach for the hybrid element method”,. Inter.
J. Num. Meth. in Engr., 24, 1697-1709, 1987.
[11] Y.K. Cheung and Chen Wanji, “Isoparametric hybrid hexahedral elements for three
dimensional stress analysis,”. Inter. J. Num. Meth. in Engr., 26, 677-693, 1988.
[12] Y.F. Dong, C.C. Wu and J.A. Teixeira de Freitas, “The hybrid model for Mindlin-
Reissner plates based on a stress optimization condition,”, Comp. and Struct., 46(5) 877-897,
1993.
[13] anon., The Finite Element Primer, National Agency for Finite Elements, Department of
Trade and Industry, Glasgow, pp 85-115, 1986.
[14] Y.K. Cheung and Chen Wanji, “Hybrid quadrilateral element based on Mindlin/Reissner
plate theory,”. Comp. and Struct., 32(2), 327-329.
[15] L.S.D. Morley, Skew Plates and Structures, Pergamon Press, Oxford, 1963.
[16] R.D. Cook, “Improved two-dimensional finite element”, J. Struct. Div, ASCE, ST9, 1851-
1863, 1974.
Nastran Primer 223

7
References
[17] R.L. Taylor and F. Auriccho, Linked interpolation for Reissner-Mindlin plate elements:
Part II - A simple triangle”, Inter. J. Num. Meth. in Engr., 36,3057-3066, 1993.
224 Nastran Primer

8 Rigid and Constraint Elements
8.1 Rigid and Constraint Elements

The rigid and constraint elements automate the generation of linear constraint equations.
Linear constraint equations can always be defined using the MPC data statement as described
in Chapter 6. However, when the constraint equations are associated with rigid body motion
or other specific types of constraints such as weighted averages or spline fits to Grid point
degrees of freedom the generation of the constraint equations can be automated.
NASTRAN incorporates a number of rigid and constraint elements that generate linear
constraint equations between degrees of freedom for:
• Rigid Elements
• Elastic Spline Fit
• Weighted Average
The NASTRAN constraint elements are summarized by Table 8-1 where the element
names such as RROD are the mnemonics for the associated data statements.
The NASTRAN rigid elements represent kinematic constraints between degrees of
freedom based on rigid body motion between connected degrees of freedom. They include the
RROD, RBAR, RTRPLT, RBE1 and RBE2. The other two elements, the RSPLINE and the
RBE3, are more correctly termed constraint elements because the relationship between degrees
of freedom is based on assumptions other than rigid body motion. The RSPLINE assumes an
cubic spline fit and the RBE3 assumes a weighted average between specified degrees of
freedom.
Constraint Number of equations

Description
Element Name m
An extensional constraint in the direction of a line
RROD 1
segment between two grid points.
RBAR A rigid link between two grid points 1≤m≤6
RTRPLT A rigid link between three grid points 1 ≤ m ≤ 12
A rigid connection between an arbitrary number of grid
RBE1 points where both independent and dependent degrees of m ≥ 1
freedom are specified by the user.
A rigid connection between an arbitrary number of grid
RBE2 points where independent degrees of freedom are at a m≥1
specified reference point.
RBE3 A weighted average constraint 1≤m≤6
RSPLINE A spline element m≥1
Table 8-1 Rigid and Constraint Elements
225 Nastran Primer

Rigid and Constraint Elements
8
RROD - Extensional Constraint
The specification of constraint elements requires an understanding of the form and

function of the linear constraint equations defined in NASTRAN. As was noted in Chapter 7,
a set of m linear constraints is incorporated into the solution set by eliminating m degrees of
freedom. The MPC operation partitions the g-set into two subsets, the m-set and the n-set that
are related by a set of m linear constraint equations. Using the constraint equations, the m-set
is related to the n-set and is removed from the analysis.
The purpose of the NASTRAN constraint elements is to provide an efficient means of
generating constraint equations in the form of (Eq. 6-2). The data statements associated
constraint element must therefore provide a mechanism for specifying the set membership of
the connected degrees of freedom. Furthermore, since the constraint elements generate MPCs,
all of the rules associated with MPCs must be followed including:
1.A degree of freedom specified in the m-set cannot be included in any of the other
removed sets. This means that it cannot specified on SPC, OMIT or SUPORT data
statements. Doing so will result in a fatal error message.
2.A degree of freedom can be specified in the m-set only once. A fatal error will be
issued if the same degree of freedom is designated as dependent on two constraint
elements.
3.The degrees of freedom in the n-set for all rigid elements except the RROD must
be sufficient to represent rigid body motion.
8.2 RROD - Extensional Constraint

The RROD constraint element is shown by Figure 8-1 and is defined by the RROD Bulk
Data shown on Bulk Data Image 8-1.
z
GB
GA
x
Figure 8-1 RROD Constraint Element
1 2 3 4 5 6 7 8 9 10
RROD EID GA GB CMA CMB
where
GA,GB are grid point identification numbers (Integer >0)
226 .Nastran Primer

RBAR - Rigid Link Connecting Two Grids
CMA, CMB component number of one translational degree of freedom at either end
A or at end B that is to be included in the m-set. One of the two fields
contains the integer code for the m-set degree of freedom while the
other field is blank.
Bulk Data Image 8-1 Extensional Constraint Element - RROD
The extensional constraint condition implies that the relative displacement in the
direction of the line segment joining the connected geometric grid points must be zero. Thus
we have
{ d }A = { d }B (Eq. 8-1)
where { d } is the displacement in the direction of the line segment from grid point A to grid
point B. Expressing { d } in terms of displacement components at A and B gives
u xA α x + u yA α y + u zA α z = u xB α x + u yB α y + u zB α z (Eq. 8-2)
where (αx, αy, αz) are direction cosines between the directed line segment and the coordinate
axes.
The RROD thus specifies a single equation which can be cast in the form of the MPC,
Multipoint Constraints - MPC (p. 175), provided that the dependent degree of freedom, i.e. the
m-set degree of freedom, is specified. This specification is made by identifying one of the three
displacement degrees of freedom in (Eq. 8-2) as an m-set degree of freedom and entering the
associated degree of freedom code in field 5 or field 6 of the RROD depending on whether the
m-set degree of freedom is at end A or B, respectively. Finally, with reference to (Eq. 8-2), it
is apparent that the direction cosine associated with the m-set degree of freedom must be non
zero. This implies that the m-set degree of freedom cannot be perpendicular to the line segment
joining A and B.
8.3 RBAR - Rigid Link Connecting Two Grids

The RBAR element provides the capability of modeling the rigid link shown by
Figure 6-1 by using the single data statement shown on Bulk Data Image 8-2 rather than a set
of MPC data statements necessary to specify the associated constraint equations (Eq. 6-3).
1 2 3 4 5 6 7 8 9 10
RBAR EID GA GB CNA CNB CMA CMB
where
CNA, CNB are codes for n-set degrees of freedom at ends A and B, respectively
(integer > 0 or blank)
CMA, CMB are codes for m-set degrees of freedom at ends A and B, respectively
(integer > 0 or blank)
Bulk Data Image 8-2 RBAR Data Statement
Nastran Primer 227

8
RTRPLT - Rigid Triangular Constraint Element
The RBAR element defines a rigid connection for up to six degrees of freedom at the
connected grid points. The constraint equations conditions for rigid body motion can then be
expressed by (Eq. 6-3). The RBAR, and the other rigid elements, allow the user to specify
which of the constraint equations is to be generated. The first requirement is to specify exactly
six degrees of freedom for the rigid element that represent rigid body motion using the CN
fields. Then, the specific constraint equation(s) to be generated are specified by the CM fields
which serves the dual purpose of assigning the associated degrees of freedom to the m-set.
The genius of the formulation of the rigid elements is that the CM fields allows the user
to specify which of the six constraint equations are to be generated. If the RBAR is to generate
only constraint equations associated with displacements then the CM field would include only
displacement degree of freedom codes and no constraints would be generated for the rotational
degrees of freedom.
8.3.1 RBAR Example: Rigid Link Connecting Two Points

Suppose that two grid points, 363 and 132, are connected by a member which is extremely
stiff compared to other elements in the structure. The correct modeling procedure in this case
is to define a set of constraint equations using a RBAR rather than using an extremely stiff
elastic beam element.
The set of MPC Bulk Data required to model this rigid body constraint equations is given
in Defining the Constraint Equation (p. 181). The RBAR can be used as an alternative in which
case the rigid link is defined by the following single RBAR statement.
1 2 3 4 5 6 7 8 9 10
RBAR 153 363 132 123456
where the six degrees of freedom at point 363 have been included in the n-set implying that the
six degrees of freedom at point 132 are in the m-set in agreement with the MPC Specification.
8.4 RTRPLT - Rigid Triangular Constraint Element

A constraint element which connects three geometric grid points is shown by Figure 8-2
and is described using the RTRPLT data statement shown on Bulk Data Image 8-3.
GC
GA
y
x
GB
Figure 8-2 Triangular Constraint Element
228 .Nastran Primer

Rigid Constraint Element - RBE1 and RBE2
1 2 3 4 5 6 7 8 9 10
RTRPLT EID GA GB GC CNA CNB CNC
CMA CMB CMC
where
CNA, CNB, CNC are codes for n-set degrees of freedom at points A, B, and C,
respectively (integer > 0 or blank)
CMA, CMB, CMC are codes for m-set degrees of freedom at A, B, and C,
respectively. (Integer > 0 or blank)
Bulk Data Image 8-3 RTRPLT Data Statement
The six rigid body degrees of freedom for the rigid element are represented by the n-set
degrees of freedom which must be specified at the connected grid points by the CNA, CNB,
and CNC fields. By default, if the CM fields are left blank, all of the remaining degrees of
freedom are includes in the m-set and twelve constraint equations will be generated. As in the
case of the RBAR, the CM fields can be used to generate specific constraint equations at the
grid points and include the associate degrees of freedom in the m-set.
8.5 Rigid Constraint Element - RBE1 and RBE2

The RBE1 and RBE2 constraint elements are generalizations of the RBAR and RTRPLT
elements that allow connection of an arbitrary number of geometric grid points. Six degrees of
freedom representing rigid body motion must be assigned to the n-set. RBE1 and RBE2 differ
only in way the n-set is specified. The form of the RBE1 and RBE2 rigid elements is shown
on Bulk Data Image 8-4.
1 2 3 4 5 6 7 8 9 10
RBE1 EID GN1 CN1 GN2 CN2 GN3 CN3
GN4 CN4 GN5 CN5 GN6 CN6
“UM” GM1 CM1 GM2 CM2 GM3 CM3
GM4 CM4 -etc.-
1 2 3 4 5 6 7 8 9 10
RBE2 EID GN CM GM1 GM2 GM3 GM4 GM5
GM6 GM7 etc.
Nastran Primer 229

8
Rigid Constraint Element - RBE1 and RBE2
where the fields of the RBE1 are
Field Description
GNi are grid points at which n-set degrees of freedom
GNi are grid points at which n-set degrees of freedom
UM is a literal string which terminates the specification of the n-set
GMi are grid points at which m-set degrees of freedom are specified
CMi are codes for degrees of freedom which are to be included in the
m-set.
and the fields of the RBE2 are
Field Description
GN Is the reference point whose six degrees of freedom represent rigid
body motion.
CM Degree of freedom codes for constraint equations to be generated at
the GMi grid points
GMi Grid points at which m-set degrees of freedom are specified and for
which constraint equations are generated.
Bulk Data Image 8-4 RBE1 and RBE2
The RBE1 and RBE2 perform the same function but have the following differences
1.The RBE1 allows the user to define the six degrees of freedom in the n-set that are
capable of representing rigid body motion by specifying exactly six grid point
degrees of freedom. These degrees of freedom may be at one or more Grid points.
If the n-set is defined at three or less grid points then the first continuation is not
required. The end of the n-set is signified by the literal string ‘UM’ on a
continuation statement. The dependent degrees of freedom are then specified after
the UM field by pairs (GMi, CMi) where GMi is the grid point number and CMi are
degree of freedom code numbers.
2.The RBE2 takes the six degrees of freedom at the grid point defined in the GN-field
to be in the n-set. The CM-field then contains the code for up to six degrees of
freedom at the grid points GMi that are included in the m-set and for which
constraint equations are generated
230 .Nastran Primer

Elastic Constraint Element - RSPLINE
8.5.1 Example: Rigid Inclusion

It may be necessary to model a rigid inclusion as a structure such as a tightly fitting bolt
in a bolt hole. The grid points on the circumference of the bolt hole shown by Figure 8-3 can
thus have no relative motion. For this example it assumed that the elastic continuum is modeled
using solid elements so that constraint equations are to be generated for only the displacement
degrees of freedom.
5
4
6 100 3
1
2
Figure 8-3 Rigid Circular Inclusion.
The equations that specify the desired constraint condition can be defined by using either
the RBE1 or RBE2 rigid element, both of which are shown below
1 2 3 4 5 6 7 8 9 10
RBE1 156 100 123456 0001
+001 UM 1 123 2 123 3 123 0002
+002 4 123 5 123 6 123
or
RBE2 157 100 123 1 2 3 4 5 0003
+003 6
In both cases the six degrees of freedom at the center of the hole are included in the n-set
while the three displacements degrees of freedom at each point around the circumference are
included in the m-set.
8.6 Elastic Constraint Element - RSPLINE

The elements described in the previous sections are rigid elements. For those elements the
constraint equations result from the kinematic condition that there is no relative motion
between specified points on the rigid body. The RSPLINE element described in this section
and the RBE3 element described in the next section define relationships between degrees of
freedom based on assumptions other than rigidity. The RSPLINE uses a cubic spline fit and
the RBE3 uses a weighted average to relate degrees of freedom. In either case, since the
Nastran Primer 231

8
concept of rigidity is not employed, it is no longer appropriate to specify six degrees of

freedom which are capable of representing rigid body motion in the n-set. On the contrary, for
both of these elements an arbitrary number of degrees of freedom can be put in the n-set.
However, there must be at least six.
The elastic interpolation element is shown by Figure 8-4 and is defined by the RSPLINE
Bulk Data shown on Bulk Data Image 8-5
1 2 3 4 5 6 7 8 9 10
RSPLINE EID D/L G1 G2 C2 G3 C3 G4

C4 G5 C5 -etc.-
where
D/L is the ratio of elastic tube diameter to total tube length. (Real > 0. Default = 0.1)
Gi are grid point numbers. (Integer > 0)
Ci are degree of freedom codes. (Integer > 0 or blank)
Bulk Data Image 8-5 RSPLINE Data Statement
The elastic interpolation element connects two end points (identified by solid circles) and
user specified interior points (identified as solid triangle).
G1
Figure 8-4 Elastic Interpolation Element - RSPLINE
The RSPLINE cubic spline represents a curve and has the property of matching the
displacement and slope at the end points and at the user-specified 'primary' interior points. The
displacements at user-designated 'secondary' interior points (indicated by the open squares)
can then be determined by using the equation for the elastic curve which matches displacement
at the primary designated points.
232 .Nastran Primer

The values of the displacement at the secondary points are thus dependent on the
displacements of the primary points that determine the equation of the spline. The
displacement degrees of freedom at the secondary points are thus included in the m-set while
the degrees of freedom associated with the end points and primary interior points are
independent and are included in the n-set.
Consider the case shown by Figure 8-5, where we wish to connect the grid points which
lie along the line y = constant using an elastic spline.
θ z3
4’
θ z1 2’
3’
1’ θ z5
u y3
5’
u y1
u y5
1 2 3 4 5
x
Figure 8-5 Elastic Spline Fit Passing Through Displaced Points
The elastic curve is to be determined by the displacement uy and rotation θ, at end points
1 and 5 and a primary interior point 3. The displacements at secondary points 2 and 4 are then
to be evaluated from the equation of the elastic curve which passes through points 1, 3, and 5.
The RSPLINE constraint element in effect uses the elastic beam equation to provide a relation
between specified and interpolated degrees of freedom.
The RSPLINE data statement allows the user to define the elastic interpolation function
as follows
1.The first and last grid points are taken to be the ends of the spline. All of the degrees
of freedom at these points are members of the n-set.
2.An arbitrary number of interior points can be defined which may be either primary
points that define the elastic curve such as point 3 in Figure 8-5 or secondary points
at which interpolated values are to be determined such as points 2 and 4. The
interior degrees of freedom are defined by a pair of fields such as (G2, C2) where
G2 is the grid point number. The presence or absence of the associated degree of
freedom code field then determines whether the points are primary or secondary as
follows.
Value of C-field Meaning

Blank The associated grid point is primary and the degrees of freedom
at the grid point are used to determine the equation for the elastic
curve and are thus to be included in the n-set.
Integer > 0 A secondary grid point. The degrees of freedom associated with
the packed degree of freedom code are to be evaluated from the
equation for the elastic curve and are thus to be included in the
m-set. All other degrees of freedom at the grid point (if any) are
to be used to define elastic behavior and are thus in the n-set.
Nastran Primer 233

8
Weighted Average Constraint Element - RBE3
3.The ratio D/L defines the ratio of bending "stiffness" to torsional "stiffness" for the
elastic tube which connects the independent (primary) degrees of freedom. A value
for this parameter can be specified or the default value D/L = 0.1 will be taken if
the field is blank. The value is of interest only if the RSPLINE connects grid points
which do not lie on a straight line. In that case bending and torsional moments are
coupled and some difference in the behavior of the RSPLINE element will be
observed if D/L is changed. The default is recommended unless the user has a
specific reason to use another value.
The RSPLINE was developed to provide a way of connecting regions of different mesh
density. However, Ransom [1] presented the results of a study done in conjunction with
development of interface elements that show that the results for even simple test cases do not
give good results. We therefore do not recommend the RSPLINE for transitioning.
8.7 Weighted Average Constraint Element - RBE3

The RBE3 constraint element,Bulk Data Image 8-6, provides a means of specifying that
the value of selected degrees of freedom is to be a weighted average of displacement of the
other degrees of freedom
1 2 3 4 5 6 7 8 9 10
RBE3 EID GREF CREF Wi Ci G1i G2i

G3i Wj Cj G1j G2j Wk Ck G1k
-etc.-
“UM” GM1 CM1 GM2 CM2 GM3 CM3
GM4 CM4 GM5 CM5 -etc.-
Bulk Data Image 8-6 RBE3 Data Statement
The form of the RBE3 constraint element allows the user to define the motion of a
reference point by
1.Specifying the reference point using the fields
• GREF - the reference grid point
• CREF - the degrees of freedom at the reference grid point.
2.Specifying weighting factors, Wi.
3.Specifying degrees of freedoms associated with the weighting factors by first
specifying the degree of freedom code, Ci, followed by the grid point identification
numbers of which the degree of freedom, Ci, are to be given the weight Wi.
4.Specifying up to six degrees of freedom which are to be placed in the m-set. These
(the m- degrees of freedom may be specified as follows
a. The default m-set includes the degrees of freedom specified at the reference
point.
b. The m-set can be explicitly defined by placing the literal string UM in field two
of a continuation statement after the definition of the weighting factors and
associated degrees of freedom. The literal string (UM) is then followed by pairs
(GM, CM) which specify one or more degrees of freedom (CM) at the grid point
(GM) that are to be included in the m-set.
234 .Nastran Primer

Weighted Average Constraint Element - RBE3
A set of linear equations is then formed by taking the displacement of the reference
point to be the weighted average of the set of independent displacements. The m-set
is then taken to be the reference set by default or those degrees of freedom explicitly
defined as the m-set. If the m-set is explicitly defined then
a. The total number of degrees of freedom in the m-set must be the same as the
number of degrees of freedom defined at the reference point.
b. The degrees of freedom must be a subset of those at the reference and the
weighted grid points.
c. The coefficient matrix of the m-set, Rmm, must be nonsingular.
8.7.1 Example -"Beaming" Loads and Masses

The user will often have to "transfer" loads and masses from a reference node, which
might not be part of the structural model, to nodes which are part of the model, Consider the
cross-section of a cylindrical shell as shown by Figure 8-6 where the total mass of the adjacent
structure and the associated center of gravity is at node 100 which lies on the axis of the
cylinder, This mass and inertia must be distributed to the structural nodes 1 through 6. The
following RBE3 element will properly distribute the mass equally to the six nodes on the shell.
l
z
5
6
4
100
60o
1 3
2
x
y
Figure 8-6 Circular Cross-Section
1 2 3 4 5 6 7 8 9 10
RBE3 74 100 123456 1. 123 1 2 +RB1
+RB1 3 4 5 6
The mass could be defined by means of a CONM2 mass element that specifies a
concentrated mass at Grid point 100. Note that degrees of freedom 1 through 6 of point 100 are
dependent while all components at nodes 1 through 6 are independent. The displacements of
node 100 represent the weighted average motions at nodes 1 through 6.
Nastran Primer 235

8
References
We might also consider the case where the mass is to be distributed to the nodes on the
cylinder but where 2/3 of the total mass is to be equally distributed to nodes 2 through 4 and
that 1/3 of the total mass is to be equally distributed to points 1, 5, and 6. The associated RBE3
bulk data is then
1 2 3 4 5 6 7 8 9 10
RBE3 75 100 123456 2. 123 2 3 +EF
+EF 4 1. 123 1 5 6
Applied static or dynamic loads can be distributed to structural degrees of freedom by

means of the same technique. For example, if loads are defined at point 100 then the first RBE3
which was defined in this section would equally distribute the loads to points 1 through 6.
The degrees of freedom at the reference point are generally included in the m-set. In some
modeling situations the degrees of freedom at the reference point must be included in the n-set.
In these cases the user can explicitly define the degrees of freedom in the m-set. For example,
the following RBE3 element will evenly distribute loads at the reference point to the structure:
1 2 3 4 5 6 7 8 9 10
RBE3 74 100 123 1.0 123 1 2 +EF
+EF 3 4 5 6 +HI
+HI UM 6 123
In this case the first two statements are similar to the first example discussed above except
that here the RBE3 element is connected to only the three translational degrees of freedom of
the reference grid point. The last data statement has been added to define components 123 of
node 6 as dependent. Now components 123 of node 100 are included in the n-set and can thus
be included in a removed data set or if not removed will be included in the m-set.
Any of the digits 1-6 may be used in the CREF field. Omission of any of these digits
merely indicates that the RBE3 element is not connected to that degree of freedom. The
recommended values for the C fields are the codes 1-3. While the codes 4-6 may be used in
these fields, determination of proper weighting factors is much more difficult as the ratios
between translation and rotational weighting factors are related to the dimensions of the
structure which is connected by the RBE3 element.
8.8 References
[1] J.A. Ransom, private communication, October, 1999.
236 .Nastran Primer

References
Nastran Primer 237

8
References
238 .Nastran Primer

9 Material Properties
The constitutive equations as well as other material properties are described in this
chapter. Constitutive relations may be of the following types in NASTRAN:
1. Isotropic temperature-dependent material.
2. Anisotropic temperature-dependent material for two- and three-dimensional
elements.
Temperature-independent properties are defined using MATi Bulk Data statement.
Temperature dependence of properties is then prescribed by MATTi Bulk Data, which define
pointers to appropriate material tables on which the temperature dependence is defined.
The NASTRAN Bulk Data data statement used to specify material properties are
summarized in Table 9-1.
Table 9-1 Material Specifications for Structural Analysis
Data Name Description

MAT1 Linear, temperature-independent, isotropic material
Linear, temperature-independent, anisotropic material for
MAT2
two-dimensional elements
Linear, temperature-independent, orthotrophic material for
MAT8
two-dimensional elements
Linear, temperature-independent, anisotropic material for
MAT9
three-dimensional elements
Specifies table references for temperature-dependent material
MATT1
properties in conjunction with MAT1
Specifies temperature-dependent material properties in
MATT2
conjunction with MAT2
Specifies temperature-dependent material properties in
MATT9
conjunction with MAT9
The allowable material relations for the recommended NASTRAN elements are
summarized in Table 9-2.
Table 9-2 Allowable Material Relations for Structural

Elements in NASTRAN.
Element Type Material Type
MAT1 MAT2 MAT8 MAT9

BAR X
BEAM X
BEND X
HEXA X X
239 Nastran Primer

Material Properties
9
Isotropic Material
Table 9-2 Allowable Material Relations for Structural

Elements in NASTRAN.
Element Type Material Type
MAT1 MAT2 MAT8 MAT9

PENTA X X
QUAD4 X X X
QUAD8 X X X
ROD X
SHEAR X
TETRA X X
TRIA3 X X X
TRIA6 X X X
TUBE X
9.1 Isotropic Material

Material properties for a linear, temperature-independent isotropic material are specified
on a MAT1 data statement, which is shown in Bulk Data Image 9-1.
1 2 3 4 5 6 7 8 9 10
MAT1 MID E G NU RHO A TREF GE
ST SC SS MCSID
where
Field Description
MID Material identification number (integer, > 0)
E Modulus of elasticity (real, > 0. or blank)
G Shear modulus (real, > 0. or blank)
NU Poisson's ratio (- 1.0< NU < 0.5, real or blank)
RHO Mass density (real)
A LPHA Coefficient of thermal expansion (real)
TREF Reference temperature for thermal expansion (real)
Ge Structural element damping coefficient (real)
Stress limits for tension, compression, and shear, respectively,
ST, SC, SS (real). Used to compute margins of safety in certain elements and
have no effect on computational procedures.
Coordinate System with respect to which material properties are
MATCSID
defined (integer)
Bulk Data Image 9-1 MAT1 - Isotropic Material Property Definition
Features of the MAT1 data statement are:

1. The MID field must be unique among all MATi-type data statements.
240 Nastran Primer

Material Properties
Isotropic Material
2. The value of RHO is used to calculate the structural mass matrix for all elements
with a defined volume. (The weight density may be specified in weight units and is
converted to mass by including the WTMASS parameter having a value 1/g where
g is the acceleration of gravity (see Appendix B for parameters).
3. If E and NU, or G and NU, are both blank, they will be set equal to zero. If two of
the three coefficients are defined, the remaining coefficient is calculated using the
isotropic relation (Eq. 9-4)).
9.1.1 One-Dimensional Elements

The constitutive relations for the one-dimensional element is of the form
σ xx = E ( ε xx – α∆T ) (Eq. 9-1)
where E is the modulus of elasticity, α is the coefficient of thermal expansion, and

∆T = T – T REF . The only elastic constant that needs to be specified for the ROD and BAR
elements is the modulus E. The shear modulus and Poisson's ratio may be left blank.
9.1.2 Two-Dimensional Elements - Plane Stress

The isotropic constitutive relations for the plane stress formulation are of the form:
{ σ } = [ E ] ( { ε } – { α }∆T ) (Eq. 9-2)
where
T
{ σ } ≡ σ xx σ yy τ xy
T
{ ε } ≡ ε xx ε yy γ xy
T
{α} ≡ α α 0
(Eq. 9-3)
E - ------------------ E -
------------------
2
ν 2
0
(1 – ν ) (1 – ν )
[E] = E E -
ν ------------------
2
- ------------------
2
0
(1 – ν ) (1 – ν )
0 0 G
and where
E Modulus of Elasticity
ν Poisson's ratio
G Shear modulus
a Coefficient of thermal expansion
∆T Temperature change, T- TREF
For an isotropic material, only two of the three material coefficients are independent since
E -
G = ------------------- (Eq. 9-4)
2(1 + ν)
Nastran Primer 241

Material Properties
9
Isotropic Material
9.1.3 Two-Dimensional Elements - Plane Strain

The isotropic constitutive relations for the plane strain formulation are of the form
{ σ } = [ E′ ] ( { ε } – { α′ }∆T ) (Eq. 9-5)
where:
(1 – ν) ν 0
E
[ E′ ] = -------------------------------------- ν (1 – ν) 0
( 1 + ν ) ( 1 – 2ν ) (-------------------
1 – 2ν )-
0 0
2
and
T
{ α′ } = ( 1 + ν ) α α 0
The material properties for plane strain can be specified by using the MAT1 Bulk Data
data statement if E, ν, and α are replaced by E', ν', and α' where:
E
E′ = ------------------
-
2
(1 – ν )
ν (Eq. 9-6)
ν′ = ----------------
(1 – ν)
α′ = ( 1 + ν )α
9.1.4 Three- Dimensional Elements

The constitutive relations for the three-dimensional case are:
{ σ } = [ E ] ( { ε } – { α }∆T ) (Eq. 9-7)
where:
T
{ σ } = σ xx σ yy σ zz τ xy τ yz τ zx
T
{ ε } = ε xx ε yy ε zz γ xy γ yz γ zx (Eq. 9-8)
T
{α} = α α α 0 0 0
and
1 ν ν 0 0 0
ν 1 ν 0 0 0
ν ν 1 0 0 0
E (----------------
1 – ν)
[ E ] = ------------------
2
- 0 0 0
2
0 0 (Eq. 9-9)
(1 – ν )
(----------------
1 – ν)
0 0 0 0 0
2
(----------------
1 – ν)
0 0 0 0 0
2
242 Nastran Primer

Material Properties
Anisotropic Material for Surface Elements
9.2 Anisotropic Material for Surface Elements

The material coefficients for a linear temperature-independent anisotropic material are
shown in Bulk Data Image 9-2
1 2 3 4 5 6 7 8 9 10
MAT2 MID G11 G12 G13 G22 G23 G33 RHO

A1 A2 A3 TREF Ge ST SC SS
MCSID
where
Field Description
Material ID number (integer). Must be unique over all material
MID
IDs.
Diagonal and Upper triangular coefficients of elasticity matrix as
G11, G12, G13, G22, etc.
defined by Eq. 10-11 (real).
ALPHA1, ALPHA2,
Coefficients of thermal expansion (real).
ALPHA12
Coordinate ID for coordinate system used for material
MATCID
coefficients (integer).
Bulk Data Image 9-2 MAT2 Anisotropic Material
The Stress strain equation is given by (Eq. 9-2) and the components of stress and strain
are the same as those presented by (Eq. 9-3). The thermal strain vector and elastic matric must
be modified for non isotropic behavior as shown:
T
{ α } ≡ α 1 α 2 α 12
G 11 G 12 G 13 (Eq. 9-10)
[ E ] = G 12 G 22 G 23
G 13 G 23 G 33
where:
1.[E] is a symmetric matrix so that only the upper half is specified by MAT2.
2.The 1 and 2 directions are relative to the MATCID coordinate system.
3.The α12 term can exist as a result of a coordinate transformation.
9.2.1 Orthotropic Material Properties

The orthotropic relations for two-dimensional behavior are given by:
Nastran Primer 243

Material Properties
9
Anisotropic Material for Surface Elements
    
 σ 11  G 11 G 12 0   ε 11   α 1 
     
 σ 22  = G 12 G 22 0   ε 22  – ∆T  α 2  (Eq. 9-11)
     
 σ 12  0 0 G 33   γ 12   0 
    
The orthotropic material coefficients are defined relative to a specific set of axes which
are called the material axes. The elastic coefficients for an orthotropic material are related to
engineering material coefficients as:
E1
G 11 = -----------------------------
-
( 1 – ν 12 ν 21 )
ν 12 E 2 ν 21 E 1
G 12 = -----------------------------
- = -----------------------------
- (Eq. 9-12)
( 1 – ν 12 ν 21 ) ( 1 – ν 12 ν 21 )
E2
G 22 = -----------------------------
-
( 1 – ν 12 ν 21 )
where
E1 Elastic modulus in the direction of the 1- material axis.
E2 Elastic modulus in the direction of the 2- material axis.
ν 12 Poisson's ratio for transverse strain in the 2-direction when stress is applied
in the 1-direction.
ν 21 Poisson's ratio for transverse strain in the 1-direction when stress is applied
in the 2-direction.
G33 Shear modulus
α1 Coefficient of thermal expansion in the 1-direction
α2 Thermal expansion coefficient in the 2-direction
9.2.2 Example: Orthotropic Material

The properties of an E-glass epoxy lamina are
Ex = 7.8 x 106 psi Ey = 2.6 x 106 psi

νxy = 0.25 G33 = 1.25 x 106 psi
α1 = 3.5 x 10-6 in. /in. -oF α2 = 11.4 x 10-6 in./in. -oF
The stress limits for a composite material are generally vastly different in the directions
of the material axes since one axis is aligned parallel to the fiber direction and one axis is
transverse to the fiber direction. Typical values for the lamina strengths in strength of E-glass
are
XT = Xc = 150 x 103 psi
YT = 4x103 psi
Yc = 20 x 103 psi
Ss = 6 x 103 psi
244 Nastran Primer

Material Properties
Orthotropic Material for Surface Elements
where:
(XT, Xc) Tensile and compressive strengths in the fiber direction

(YT, Yc) Tensile and compressive strengths transverse to fiber direction
SS Shear strength
For the purpose of this example, we will assume that the longitudinal stress is of interest, and
we will therefore take the following stress limits:
ST= 150 x 103 psi
SC = 150 x 103 psi
SS = 6 x 103 psi
The elastic constants are calculated by first noting that the missing Poisson's ratio can be
calculated using the second of Eq. 10-15 to give:
ν xy E y
ν yx = -------------
- = 0.083
Ex
We then find:
G11 = 7.965 x 106 psi
G12 = 0.664 x 106 psi
G22 = 2.655 x 106 psi
These material properties can be defined on a MAT2 data statement as follows
1 2 3 4 5 6 7 8 9 10
MAT2 101 7.965+6 0.664+6 0.0 2.655+6 0.0 1.25+6 0.0 +M1
+M1 3.5-6 11.4-6 0.0 72.0 1.5+5 6.0+3
9.3 Orthotropic Material for Surface Elements

The calculation of the coefficients of the elastic matrix using (Eq. 9-12) can be eliminated
if we use the MAT8 data statement as shown by Bulk Data Image 9-3, which allows us to
specify engineering material constants
1 2 3 4 5 6 7 8 9 10
MAT8 MID E1 E2 ν12 G12 G1z G2z RHO

Α1 Α2 TREF Xt Xc Yt Yc Ss
Ge F12 STRN
Bulk Data Image 9-3 MAT8 Data Statement
The MAT8 material data statement is used in conjunction with the PCOMP property data
statement, for thin shell elements to define the lamina properties with respect to a set of
principal material axes (X1, X2). The elastic constants with respect to the material coordinates
are then defined by(Eq. 9-10) and are transformed by using the lamina orientation angle θ,
which is defined on PCOMP statement, by using appropriate transformation equations (see Ref
[1], for example).
Nastran Primer 245

Material Properties
9
Anisotropic Material for Solid Elements
The material coefficients that are specified by MAT8 Bulk Data are described as follows
El, E2 Modulus of Elasticity along and normal to the fiber direction,
respectively (real not ≠ 0)
ν12 Poisson's ratio for loading in the 1-direction. The poisson's ratio ν21 is
related to ν12 by the equation ν12E2 = ν21E1. (real)
G12 In-plane shear modulus. (real ≠ 0.)
G1z,G2z, Transverse shear modulae (real or blank)
RHO Mass Density (real or blank)
α1,α2 Coefficients of thermal expansion in 1 - and 2 - directions. (real or blank)
TREF Reference temperature for calculation of thermal expansion
XT, XC Allowable stress in tension and compression, respectively, in the fiber
direction. (real or blank)
YT, YC Allowable stress in tension and compression, respectively, normal to fiber
direction. (real or blank)
SS Allowable Shear stress (real or blank)
Ge Structural damping coefficient. (real or blank)
F12 Interaction term in tensor polynomial for Tsai-Wu. (real or blank)
STRN If maximum strain failure theory is chosen by the FT field of the
associated PCOMP data statement, then the value of this field indicates
whether the allowables are stress or strain allowables. STRN =1. (default)
for strain, STRN = 0. for stress.
9.4 Anisotropic Material for Solid Elements

The material coefficients for a linear temperature-independent three dimensional material
are defined using the MAT9 data statement shown on Bulk Data Image 9-4 The only elements
that can reference this material formulation are the TETRA, PENTA and HEXA
three-dimensional isoparametric elements.
1 2 3 4 5 6 7 8 9 10
MAT9 MID C11 C12 C13 C14 C15 C16 C22

C23 C24 C25 C26 C33 C34 C35 C36
C44 C45 C46 C55 C56 C66 RHO α1
α2 α3 α4 α5 α6 TREF Ce
where:
Field Description
MID The material ID (integer). Must be unique among all material type
data statements.
C11,C12, …, C66 Coefficients of upper half of symmetrical elastic matrix (real)
RHO Mass density per unit volume.
246 Nastran Primer

Material Properties
Anisotropic Material for Solid Elements
Field Description
α1,α2, …, α6 Coefficients of thermal expansion (real).
TREF Reference temperature for calculation of thermal expansion
(real).
The general form of the elastic relationship for three-dimensional elements in a

three-dimensional stress state is given by (Eq. 9-8) and the coefficients of stress and strain by
the (Eq. 9-9). The thermal strain vector is:
T
{α} = α1 α2 α3 α4 α5 α6 (Eq. 9-13)
and where the elasticity matrix, [E], is a symmetric 6 by 6 matrix whose coefficients are Cij.
The form of the MAT9 data statement thus allows the user to define an anisotropic
material having twenty-one independent material coefficients. The inclusion of the six
components of the thermal expansion vector allows the definition of transformed material
coefficients.
9.4.1 Orthotropic Constants for Solid Elements

Engineering materials having orthotropic properties are finding increased application in
the design of structural systems. An orthotropic material is completely defined by nine
independent elastic constants. The most common elastic constants are the following
1.Elastic modulae, E1, E2, E3 in three orthogonal directions
2.Poisson's ratios, νij, for transverse strain in the j-directions due to stress in
i-direction.
3.Shear moduli, G12, G23, G31 in the 1-2, 2-3, and 3-1 planes, respectively.
The compliance matrix, [S], which is the inverse of the elastic matrix is given by:
1- ν 21 ν 31
----- – -------- – -------- 0 0 0
E1 E2 E3
ν 12 1 ν 32
- ------ – -------
– ------- - 0 0 0
E1 E2 E3
ν 13 ν 23 1
–1
- – -------- ------
– ------- 0 0 0
[S] = [E] = E1 E2 E3 (Eq. 9-14)
1
0 0 0 --------- 0 0
G 12
1
0 0 0 0 --------- 0
G 23
1
0 0 0 0 0 ---------
G 31
The compliance matrix is symmetric so that the following symmetry relations must hold:
Nastran Primer 247

Material Properties
9
Transformation of Elastic Constants
ν 12 E 2 = ν 21 E 1
ν 23 E 3 = ν 32 E 2 (Eq. 9-15)
ν 31 E 1 = ν 13 E 3
The nonzero stiffness coefficients, Cij, are found by inverting the compliance matrix
(Eq. 9-14) and are:
1 – ν 23 ν 32 ν 12 + ν 32 ν 13 ν 13 + ν 12 ν 23
------------------------- ------------------------------ ------------------------------ 0 0 0
CE 2 E 3 CE 1 E 3 CE 1 E 3
ν 12 + ν 32 ν 13 1 – ν 31 ν 13 ν 23 + ν 21 ν 13
------------------------------ ------------------------- ------------------------------ 0 0 0
CE 1 E 3 CE 1 E 3 CE 1 E 2
[ E ] = ν 13 + ν 12 ν 23 ν 23 + ν 21 ν 13 1 – ν 12 ν 21 (Eq. 9-16)
------------------------------ ------------------------------ ------------------------- 0 0 0
CE 1 E 3 CE 1 E 2 CE 1 E 2
0 0 0 G 12 0 0
0 0 0 0 G 23 0
0 0 0 0 0 G 31
where:
1 – ν 12 ν 21 – ν 23 ν 32 – ν 31 ν 13 – 2ν 21 ν 32 ν 13
C = -------------------------------------------------------------------------------------------------------
- (Eq. 9-17)
E1 E2 E3
9.5 Transformation of Elastic Constants

The elastic constants for non-isotropic materials must be defined with respect to a specific
material coordinate system. The user then specifies both a material set identification number
and the coordinate identification number associated with the material coefficients on a
PSOLID data statement as shown by Bulk Data Image 6-10. The ability to specify coordinate
systems using the CORD-type data statement and the reference system used on the PSOLID
relieves the analyst from the burden of transforming elastic constants from one coordinate
system to another.
248 Nastran Primer

Material Properties
Temperature- Dependent Materials
9.6 Temperature- Dependent Materials

Any of the temperature-independent material property fields specified on the MAT1,
MAT2, or MAT9 data statements can be made temperature dependent by means of MATTi
data statements, as shown in Bulk Data Image 9-5.
1 2 3 4 5 6 7 8 9 10
MATT1 MID T(E) T(G) T(NU) T(RHO) T(A) T(Ge)

T(ST) T(SC) T(SS)
MATT2 MID T(G11) T(G12) T(G13) T(G22) T(G23) T(G33) T(RHO)

T(A1) T(A2) T(A3) T(Ge) T(ST) T(SC) T(SS)
MATT8 MID T(E1) T(E2) T(NU12) T(G12) T(G2z) T(G2z) T(RHO)

T(A1) T(A2) T(Xt) T(Xc) T(Yt) T(YC) T(Ss)
T(Ge) T(F12)
MATT9 MID T(C11) T(C12) T(C13) T(C14) T(C15) T(C16) T(C22)

T(C23) T(C24) T(C25) T(C26) T(C33) T(C34) T(C35) T(C36)
T(C44) T(C45) T(C46) T(C55) T(C56) T(C66) T(RHO) T(A1)
T(A2) T(A3) T(A4) T(A5) T(A6) T(Ge)
where:
Field Description
The ID of a MAT-type data statement whose coefficients are to be
MID
temperature-dependent (integer > 0).
The ID of a TABLEM-type tables that specifies the temperature
T(E), T(G),T(NU),etc. dependence of the associated field of the MAT1that is indicated in
parentheses.
The ID of a TABLEM-type tables that specifies the temperature
T(G11), …, T(G33),etc. dependence of the associated field of the MAT2 that is indicated
in parentheses.
The ID of a TABLEM-type table that specifies the temperature
T(E1), …, T(F12),etc. dependence of the associated field of the MAT8 that is indicated
in parentheses.
The ID of a TABLEM-type table that specifies the temperature
T(C11), …,T(C66), etc. dependence of the associated field of the MAT9 that is indicated
in parentheses.
Bulk Data Image 9-5 MATTi Data Statements
Nastran Primer 249

Material Properties
9
Material Property Table
The form of all MATTi data statements is the same. The MID of the MATTi statement
must match the MID on the corresponding MAT-type statement whose properties are to be
made temperature dependent. The temperature-dependence of a specific property is then
defined by reference to an appropriate table in the field of the MATTi statement associated
with the temperature-dependent property. Blank or zero entries in the table reference field of
MATTi are taken to mean that the property on the associated MATi data statement is
temperature independent.
9.7 Material Property Table

The temperature dependence of material properties is defined by means of material
property tables. Each of the four types of TABLEMi data statements that are shown in Bulk
Data Image 9-6 uses a different algorithm evaluate the function.Tables TABLEM1 through
TABLEM3 use linear interpolation between table entries. TABLEM4 uses a power series
expansion. The algorithms are defined in Table 9-3
1 2 3 4 5 6 7 8 9 10
TABLEM1 TID XAXIS YAXIS
T1 P1 T2 P2 T3 P3 -etc.- ENDT
TABLEM2 TID Ta
TABLEM3 TID Ta Tb
TABLEM4 TID Ta Tb TL TU
A0 A1 A2 A3 A4 A5 -etc.- ENDT
where
Field Meaning
TID Table identification number, Integer > 0
XAXIS,YAXIS Specifies the type of interpolation to be applied in interpolating
temperature values. Blank or one of the character strings,
‘LINEAR’ or ‘LOG’. Linear interpolation is used if the field is
blank.
(Ti, Pi) Pairs of temperature and property value, real.
Ta, Tb, TL,TU Table parameters whose use depends on the specific table
algorithm, real.
Ai Coefficients of power series expansion used only by TABLEM4,
real.
ENDT A required character string, ‘ENDT’, which signifies the end of
table entries.
Bulk Data Image 9-6 Temperature- Dependent Material Property Tables
250 Nastran Primer

Material Properties
Table 9-3 Algorithms Used for Material Tables
Table Type Look-up Algorithm
TABLEM1 P = PT ( T )
TABLEM2 P = P T ( T – T1 )
P = P ref P T  ----------------
T – T1
TABLEM3  T2 
N
T – T1 i
TABLEM4 P = P ref ∑ A i  ----------------
 T2 
i=0
Features of the TABLEMi data statements are:

1. At least two entries must exist for tables TABLEMi (i = 1,2,3).
2. At least one entry must exist for TABLEM4.
3. Tables must be terminated with the alphabetic string 'ENDT' in the field following
the last entry.
4. The table temperatures must define either an ascending or descending sequence.
TABLEM1 provides a simple tabular look-up. The pairs (Ti, Pi) are specified where T is
the temperature and P is the associated material property. The table look-up evaluates the
property P at temperature T by using linear interpolation within the table and linear
extrapolation outside the table based on the last two points at the end of the table. A
discontinuous material property may be defined at a specific temperature, as indicated in
Figure 9-1
Figure 9-1 Discontinuity of Property Value
P4
P3
P2
P1
T1 T2 T3
Discontinuities may not be defined at the endpoints of the table and the average value of
the property is taken at the jump. For example, at T = T2
P3 + P2
P ( T 2 ) = ------------------ (Eq. 9-18)
2
Nastran Primer 251

Material Properties
9
TABLEM2 is similar to TABLEM1 except that tabulated values are used to scale the basic
property value Pref defined on the MATi data statement which is referenced on the MATTi data
statement. The inclusion of a reference temperature, Ta, is an added convenience since many
material properties are defined relative to room temperature. TABLEM3 data statement is
similar to the TABLEM2 data statement, except that the temperature difference is normalized
with respect to the parameter Tb.
TABLEM4 allows the analyst to define the temperature-dependent property using a
power series in the normalized independent variable T = (T - Ta)/Tb. As in TABLEM2 and
TABLEM3, the reference value of the, property, Pref, is that which is on the appropriate field
of the MATi data statement. The coefficients of a power series expansion are entered on the
continuation statement. The first coefficient A0 must appear in the second field of the
continuation statement, followed by A1, A2,...,AN. If one of the coefficients is zero, then a zero
entry must be included in the corresponding field. The parameters TL and TU are lower and
upper temperature limits, respectively so that T is bounded as:
 T L if T < T L 
T =  
 T U if T > T U 
9.7.1 Example: Temperature- Dependent Elasticity

Consider the temperature-dependent modulus of elasticity that is shown in Figure 9-2 The
isotropic room temperature properties are
E = 10x 106 psi

ν = 0.3
α = 7 x 10-6 in. /in. -oF
γ = 0. 1 lbl /in.3
Tref = 72o
–6
E × 10 psi
10
T (oF)
72 250 400
Figure 9-2 Temperature- Dependent Elastic Modulus
252 Nastran Primer

Material Properties
Modeling Composite Materials
The following data statements specify this a temperature dependent modulus of Elasticity, E:
1 2 3 4 5 6 7 8 9 10
MAT1 1 1.E7 .3 .1 7.E-6 72.
MATT1 1 37
TABLEM1 37 +TM1
+TM1 72. 1.0E7 250. 0.8E7 400. 2.0E6 ENDT
where
1.The MATT1 data statement has the same material identification number, MID, as
the MAT1 data statement.
2.The TID of the temperature table is entered in the field of MATT1 which
corresponds to the modulus of elasticity, E, and points to a TABLEM1 data
statement whose identification number is 37.
3.The table must be terminated by the alphabetic string 'ENDT' in the field following
the last entry.
The temperature-dependent modulus could have been defined using a TABLEM2 data
statement:
1 2 3 4 5 6 7 8 9 10
TABLEM2 37 72. +TM2
+TM2 0. 1. 178. .8 328. .2 ENDT
where the table temperatures have been reduced by the value of the reference temperature and
the property entries have been normalized with respect to the value on the MAT1 data
statement.
9.8 Modeling Composite Materials

Composite materials are attractive for design of structures since materials having
attractive attributes including high strength-to weight ratio and low coefficients of thermal
expansion can be designed and manufactured. The mechanics of composite materials is well
covered in the literature (Ref. [1] through [4]) and will only summarized here.
Nastran Primer 253

Material Properties
9
9.8.1 Lamina Properties

A laminated composite material is manufactured by bonding lamina in some stacking
sequence as shown by Figure 9-3.
Figure 9-3 Laminate Construction
The individual lamina have orthotropic properties defined relative to a set of lamina material
axes as shown by Figure 9-4.
Figure 9-4 Lamina Material Axes
254 Nastran Primer

Material Properties
The orthotropic lamina strain-stress relations relative to the lamina coordinate system are
given by(Eq. 9-10) and (Eq. 9-12). These lamina equations must be transformed to the
Laminate material coordinate system, which for a plane Lamina is a rotation about the normal
through an angle, θ, as shown by Figure 9-5.
y
2
1
+θ
x
Figure 9-5 Transformation of Lamina Axes
The transformation of the stress strain relations is relatively straight forward since the
stress components are a second order tensor and the engineering strain components can be
transformed to tensor strain components as follow:
{ ε } tensor = [ T ] { ε' } tensor

(Eq. 9-19)
{ σ } = [ T ] { σ' }
The engineering strains can be transformed to tensor components as follows:
   
 ε 11  1 0 0  ε 11 
    (Eq. 9-20)
 ε 22  = 0 1 0  ε 22  = [ R ] { ε } tensor
   
 γ 12  0 0 2  ε 12 
 Engr·  Tensor
Jones[1] shows that the transformed material matrix and thermal strain vector are then
given by:
T
[ G ] = [ T ] [ G' ] [ T ]
(Eq. 9-21)
T
{ α } = [ T ] [ R ] { α' }
where [T] is the transform from the prime to the unprime coordinate system. The Lamina stress
strain relationship referred to the material coordinate system is then:
{ σ } = [ G ] { ε } + ∆T { α } (Eq. 9-22)
9.8.2 Laminate Properties

The laminate is assumed to be constructed by bonding individual lamina. The bond
between lamina is to be infinitesimally thin and have no shear deformations. The strains and
curvatures of the laminate are then defined by plate bending theory as described in Chapter 2.
The stress coefficients are not continuous through the thickness. However, the variation
of strain is a function of the strain at the reference surface and the curvatures as follows:
Nastran Primer 255

Material Properties
9
   o   
 ε 11   ε 11   χ 11 
     
 ε 22  =  ε o22  + z  χ 22  (Eq. 9-23)
     
 γ 12   γo   χ 12 
   12   
where the reference surface strains and curvatures are given by Eqs 2-53 and 2-54. The stress
strain relation for the kth lamina is then found by substituting (Eq. 9-23) into (Eq. 9-22):
   o     
 σ 11  G 11 G 12 G 13   ε 11   χ 11   α1  
       
 σ 22  = G 21 G 22 G 23   εo  + z  χ 22  + ∆T  α 2   (Eq. 9-24)
    22     
 τ 12  G 31 G 32 G 33  o
k γ
  χ 12   α3  
 k  12     k
The laminate stacking order is assume as to be as shown by Figure 9-6.
1
1--- z0 2
2 z1 MIDDLE
z2 SURFACE
zk – 1
z zk
k zN – 1
zN
N
LAYER NUMBER
Figure 9-6 Laminate Stacking Order
The stress and moment resultants described in Chapter 2 are then found by piecewise
integration over the thickness to be:
 
 Nx  N  zk z
k 
   
{ N } =  Ny

 =

∑ [ G ]k  ∫ { ε } dz + ∫ { χ }z dz (Eq. 9-25)
 N xy  k=1  zk – 1 zk – 1 
 
 
 Mx  N  zk zk 
   2 
{ M } =  My

 =

∑ [ G ]k  ∫ { ε }z dz + ∫ { κ }z dz (Eq. 9-26)
 M xy  k=1  zk – 1 zk – 1 
 
Since the reference surface strains and curvatures are not functions of z, the stress and
moment resultants can be written as:
256 Nastran Primer

Material Properties
o
{ N} = [ A]{ε } + [B ]{ χ }
(Eq. 9-27)
o
{M} = [B]{ε } + [D]{χ}
where the elastic stiffness coefficients are:
 N 
A ij =  ∑ ( G ij ) k ( z k – z k – 1 )
 
k = 1 
 N 
B ij = ---  ∑ ( G ij ) k ( z k – z k – 1 )
1 2 2
(Eq. 9-28)
2 
k = 1 
 N 
D ij = ---  ∑ ( G ij ) k ( z k – z k – 1 )
1 3 3
3
k = 1 
The elastic matrix, [A] and [D], are called the extensional and bending stiffness,
respectively and the matrix [B] is the coupling matrix. Since the coupling matrix is an even
function it will be null for a symmetric laminate.
9.8.3 Defining a Composite Laminate Properties

The laminate is defined by the order and orientation of each lamina with respect to a
material coordinate system. The PCOMP data statement defines the properties of an N-ply
laminate as shown by Bulk Data Image 9-7.
Properties of n-ply laminated composite material
1 2 3 4 5 6 7 8 9 10
PCOMP PID Z0 NSM Sb FT Tref Ge LAM
MID1 T1 θ1 SOUT1 MID2 T2 θ2 SOUT2
MID3 T3 θ3 SOUT3 -etc.-
Properties of n-ply laminated composite material where all plies

are of the same material and equal thickness
1 2 3 4 5 6 7 8 9 10
PCOMP1 PID Z0 NSM Sb FT MID TPLY LAM

θ1 θ2 θ3 -etc.-
Properties of n-ply laminated composite material where all plies

are of the same material
1 2 3 4 5 6 7 8 9 10
PCOMP2 PID Z0 NSM Sb FT MID LAM

T1 θ1 T2 θ2 -etc.-
Nastran Primer 257

Material Properties
9
where
Field Description
PID Property identification number. (0 < integer < 1000000)
Distance from the reference plane to the bottom surface, real. The
z0
default is -0.5*T where T is the element thickness.
NSM Non structural mass, real > 0.
Allowable interlaminar shear stress, required if FT is also
Sb
specified. Real.
Failure theory. One of the following character strings:
“HILL” - Hill Theory
FT “HOFF” - Hoffman theory
“TSAI” - Tsai-Wu theory
“STRN” - Maximum strain theory
Tref Reference temperature, real or blank. Default is 0.0.
Ge Damping Coefficient, real or blank
Symmetric laminate option, either the character string, SYM, or
blank. If LAM = SYM then only plies on one side of middle
LAM surface are specified. If an odd number of plies is desired then the
center ply thickness is one half the actual thickness. If blank all
plies must be specified.
MID of a lamina starting with the bottom lamina. Must refer to a
MIDi
MAT1, MAT2 or MAT8 data statement. Integer> 0 or blank.
Ti Ply thickness, starting with the bottom ply. Real
Orientation angle of the longitudinal axis of each ply with the
θi
element material axis, starting with the bottom ply. Real or blank
A character string requesting lamina stress and strain output.
SOUTi SOUT = YES - print stress and strain for the lamina
SOUT = NO - do not print stress and strain output
Default is NO
Bulk Data Image 9-7 PCOMP Layered Composite
9.8.3.1 Remarks
1.The minimum number of plies is one.
2.One of the fields, MIDi, Ti, and θi must exist for each ply. Those not specified
default to the last specified value for the field.
3.The reference non uniform structural damping coefficient, Ge, and temperature
specified on the material data statements is not used. The damping coefficient and
reference temperature specified on PCOMP is used for all plies.
4.The ELFORCE and STRESS Case Control directives must be present if ply stress
or failure indices are desired.
5.PCOMP is processed by the Input File Processor and to generate PSHELL and
MAT2 data statements. Theses generated statements are printed if an ECHO =
SORT request is included in Case Control.
258 Nastran Primer

Material Properties
6.PCOMP is also used for post processing ply stresses so that the equivalent PSHELL
and MAT2 will not result in ply stresses.
7.If the specified value of z0 is not one half the thickness of the laminate and
PARAM,NOCOMPS,-1 is specified then the homogeneous element stresses are
incorrect while the lamina stress and force and strains are correct. For correct
homogeneous stresses use ZOFFS on the shell connection statement.
9.8.4 Lamina Failure Criteria

The general topic of failure of fiber-reinforced orthotropic lamina is presented in detail
by Jones[1]and Gurdal [2]. The failure criteria incorporated in the composite modeling
capability of NASTRAN is summarized in the following sections.
9.8.4.1 Maximum Strain Theory

The maximum strain criterion states that failure occurs when one of the following
inequalities is violated:
t X t Yt
ε 1 < ε 1 = ------t ,ε 2 < ε 2 = ------ for ε 1 ,ε 2 > 0
E1 E2
c X c Y
ε 1 < – ε 1 = – -----c- ,ε 2 < – ε 2 = – ------t for ε 1 ,ε 2 < 0 (Eq. 9-1)
E1 E2
s S
γ 12 < γ 12 = ---------
G 12
where S is the Ultimate in-plane shear strength, X and Y are ultimate strengths along and
transverse to the fiber directions, respectively; and, t and c sub and superscripts refer to tension
and compressive values, respectively.
9.8.4.2 Tsai-Hill Theory

The maximum strain criteria fail to represent the effect of interactions of strain
components on failure. The Von Mises criteria accounts for these interactions an various
modifications are used to represent the failure surface. Those used most frequently are due to
Tsai-Hill, Tsai-WU and Hoffman [5]and [6]].
Hill Theory[3] extends von Mises’ failure criterion for anisotropic materials with equal
strength in tension and compression for a three dimensional stress state. The modified theory
for plane stress is generally termed the Tsai-Hill theory which has the following failure
criterion:
1- 2 -----
1 2 1 1 2
----- σ + 2 σ 22 – -----2- σ 11 σ 22 + ----2- τ 12 < 1
2 11
(Eq. 9-1)
Xt Yt Xt S
9.8.4.3 Hoffman Theory

Hoffman[4] generalized the Tsai-Hill theory to allow for different strengths in tension and
compression. The Hoffman Criterion is:
σ σ +  ---- 1  σ +  ---- 1- σ + ----

1 2 ----------
1 - 2 ---------- 1- 1- – ----- 1 – ---- 1- 2
---------- σ + σ – τ <1 (Eq. 9-1)
X t X c 11 Y t Y c 22 X t X c 11 22  X t X c 1  Y t Y c 2 S 2 12
Nastran Primer 259

Material Properties
9
Reference
9.8.4.4 Tsai-Wu Theory

The Tsai-Wu theory also allows for different strengths in tension and compression. The
Tsai-Wu criterion is:
σ 11 σ 22 +  ----- – ----- σ 1
1 2 --------------------------
1 - 2 ---------- 1 1 1
---------- σ + σ – (Eq. 9-1)
X t X c 11 Y t Y c 22 XX YY X X 
t c t c t c
+  ----
1 1- σ + ----
1- 2
τ <1
 Y – ----
Y  2 2 12
t c S
9.9 Reference
[1] Jones, R. M., Mechanics of Composite Materials, McGraw-Hill, 1975, pp. 147-164.
[2] Z. Gurdal, R.T. Haftka and P. Hajela, Design and Optimization of Laminated Composite
Materials, Wiley Interscience, 1999, pp. 231-245.
[3] R. Hill, “A theory of the yielding and flow of anisotropic materials”, Proceedings of the
Royal Society, A(193), 1948, 281-297.
[4] O. Hoffman, “The brittle strength of orthotropic materials”, Journal of Composite
Materials, 1(2), 1967, 200-29.
[5] S.W. Tsai, “Strength theories of filamentary composites”, In Fundamental Aspects of
Fiber Reinforced Plastic Composites, R.T. Schwartz and H.S. Schwartz (eds.), Wiley
Interscience, New York, 1968, pp. 3-11.
[6] E.M. Wu, “Strength and fracture of composites”. In Composite Materials Vol. 5, pp. 191-
247. L.J. Brountma and R.H. Krock (eds.) Academic Press, New York, 1974.
260 Nastran Primer

Material Properties
Reference
Nastran Primer 261

Material Properties
9
Reference
262 Nastran Primer

10 External Loads
External applied loads may result from a variety of effects, including:

1. Concentrated forces and moments at Grid points
2. Applied element surface forces
3. Distributed element body forces
4. Enforced Grid point displacement
5. Enforced element deformation
6. Element thermal fields
The total load applied to each element is the superposition of each type of loading so that
the total element load, {Pe}, is given by:
{ P e } = { P ex } + { P es } + { P eg } + { P ei } (Eq. 10-1)
where {Peg} are the concentrated loads applied to the grid points and where {Pex}, {PeS}, and
{Pei}, are the work equivalent Grid point forces associated with body forces, surface tractions
and initial strains, respectively.
The data statements for specifying loads are summarized by Table 10-1.
Table 10-1 External Static Loads
Load Type Description Reference
FORCE Concentrated static force or moment vector at a grid Bulk Data
MOMENT point, defined using components. Image 1-1
FORCE1 Concentrated static force or moment vector having a

Bulk Data
magnitude and a direction specified by the line
MOMENT1 Image 1-2
connecting two grid points.
FORCE2 and Concentrated static force or moment having a magnitude

Bulk Data
and a direction determined by the cross product of two
MOMENT2 Image 1-3
vectors.
Bulk Data
SLOAD Load at scalar point
Image 1-4
Distributed load along the length of BAR, BEAM, and Bulk Data
PLOAD1*
BEND elements Image 1-5
Bulk Data
PLOAD2 Uniform normal pressure on two-dimensional elements
Image 1-6
Surface traction for TETRA, HEXA, PENTA, QUAD4, Bulk Data
PLOAD4**
QUAD8, TRIA3, and TRIA6 elements Image 1-7
Gravity load specified by defining gravity vector in any Bulk Data
GRAV
defined coordinate system Image 1-8
Bulk Data
RFORCE Load due to centrifugal force field
Image 1-9
263 Nastran Primer

External Loads
10
Concentrated External Forces at Grid Points
Table 10-1 External Static Loads

Bulk Data
SPCD Load due to enforced grid point displacement
Image 1-9
Enforced axial deformation for one-dimensional Bulk Data
DEFORM
elements Image 1-12
Bulk Data
TEMP Temperature at a grid point
Image 1-13
Bulk Data
TEMPD Default grid point temperature
Image 1-13
Bulk Data
TEMPRB One-dimensional element temperature field
Image 1-14
Bulk Data
TEMPP1 Plate element temperature field
Image 1-15
* MSC Nastran only

**Restricted to surface elements in Legacy NASTRAN
These data statements are used to specify concentrated forces and moments at the grid
points, scalar forces, distributed body forces due to gravity and centrifugal acceleration fields,
enforced grid point displacements, enforced element deformations, and temperature fields.
These various types of structural loading are described in subsequent sections of this chapter.
10.1 Concentrated External Forces at Grid Points

Concentrated force and moment vectors can be defined at geometric grid points in the
following alternative ways:
1. Specification of the Grid point G at which the vector acts and the vector
components, as shown in Figure 1-1a.
2. Specification of the Grid point G at which the vector acts and the sense of the vector
using two geometric grid points, as shown in Figure 1-1 b.
264 Nastran Primer

External Loads
Force Vector Defined by Components
3.Specification of the geometric grid point G at which the vector acts with the vector
direction specified as the cross product of two vectors, as shown in Figure 1-1.
z z
F, M F, M
Fz , Mz
G G
G2
N
Fy , My
G1
x Fx , Mx x
y y
a) Vector Components b) Directed Line Segment Between Two Grid Points
F, M
G2
a G4
b
x
G1 y
G3
c) Vector Direction Using Cross-Product
4.
Figure 1-1 Definition of Concentrated Force and Moments
10.2 Force Vector Defined by Components

The force or moment vector can be defined by specifying the magnitude, A, and the
components of a free vector (N1, N2, N3) that acts in the direction of the force by means of the
FORCE and MOMENT data statements shown in Bulk Data Image 1-1.
1 2 3 4 5 6 7 8 9 10
FORCE SID G CID A N1 N2 N3
MOMENT SID G CID A N1 N2 N3
Nastran Primer 265

External Loads
10
Force Direction Defined by Two Points
where
Field Description
Load set identification number. Load sets must be selected using a
SID
LOAD = SID statement in Case Control. (integer, > 0)
G Grid point at which load is to be applied. (integer, > 0)
A Magnitudes of force or moment. (real ≠ 0.)
Components of a free vector defined with respect to the CID
N1, N2, N3
coordinate system. (real)
CID Coordinate identification number. (integer)
Bulk Data Image 1-1 Static Force and Moment Using Vector Components.
The vector components (N1, N2, N3) define a vector
{ N } = n1 i1 + n2 i2 + n3 i3
where the unit vectors are taken in the sense of the displacement degrees of freedom for the
CID coordinate system, as shown by Fig. 7-4. The resulting vector is then given by
{F } = A{ N} (Eq. 10-2)
so that the magnitude of the force is given by
{F} = A { N} (Eq. 10-3)
The magnitude of the force (or moment) is thus equal to the magnitude A specified on the
FORCE or MOMENT data statement only if {N} is a unit vector.
10.3 Force Direction Defined by Two Points

The force or moment vector can be defined by specifying a magnitude A and a direction
determined by the directed line segment between two grid points by means of the FORCE1 and
MOMENT1 data statements shown in Bulk Data Image 1-2.
1 2 3 4 5 6 7 8 9 10
FORCE1 SID G A G1 G2
MOMENT1 SID G A G1 G2
where
Field Description
SID Load set identification number. (integer, > 0)
G Grid point at which the load is to be applied. (integer, > 0)
A Magnitude of the force or moment. (real)
Two non coincident grid points which define the sense of the load
G1, G2 vector as the unit vector, {N}, along the directed line segment
from G1 to G2. (integers, > 0, G1 ≠ G2)
266 Nastran Primer

External Loads
Vector Normal to a Surface
Bulk Data Image 1-2 Direction of Force Defined by Two Points
The force or moment is defined as
{F } = A{ N} (Eq. 10-4)
where {N} is a unit vector in the direction of the line segment drawn between geometric grid
points G1, G2
10.4 Vector Normal to a Surface

The force or moment vector can be defined by specifying a magnitude A and a direction
determined by the vector product of two vectors by using the FORCE2 and MOMENT2 data
statements shown in Bulk Data Image 1-3
1 2 3 4 5 6 7 8 9 10
FORCE2 SID G A G1 G2 G3 G4
MOMENT2 SID G A G1 G2 G3 G4
where
Field Description
SID Load set identification number. (integer, > 0)
G Grid point at which the load is to be applied. (integer, > 0)
A Magnitude of the force or moment. (real)
G1, G2, G3, G4 Pairs of grid points that define the vectors a and b , respectively
as shown on Fig.11-1c. G1 ≠ G2, G3 ≠ G4. (integers, > 0)
Bulk Data Image 1-3 Direction of Force Defined by Vector Cross Product
The force or moment is defined as { F } = A { N }

where {N} is a unit vector defined by the cross product:
a × b-
N = -------------- (Eq. 10-5)
a×b
and where
a is the vector from grid point G1 to G2
and
b is the vector from grid point G3 to G4
The force acts at grid point G, which may be the same as points G1 and/or G3. This form
of force specification is useful when defining a force vector that is perpendicular to the surface
of an element, such as that shown in Figure 1-2, where a force is to be defined at grid point one
in the direction of the outward normal.
Nastran Primer 267

External Loads
10
Load Applied to Scalar Point - SLOAD
The force can be most conveniently specified by means of the FORCE2 data statement as
shown below.
1 2 3 4 5 6 7 8 9 10
FORCE2 101 1 10. 1 2 1 3
where a is defined from point one to two and b is defined from one to three. Since three points
define a plane, it follows that the cross product defines a vector normal to the plane as desired.
z
F = 10
b
1
Figure 1-2 Force Acting Normal to a Surface.
10.5 Load Applied to Scalar Point - SLOAD

A static load can be applied to a scalar degree of freedom by the SLOAD data statement
shown in Bulk Data Image 1-4.
1 2 3 4 5 6 7 8 9 10
SLOAD SID S F S F S F
where
Field Description
SID Load identification number. (integer, >0)
S Scalar point identification number. (integer, >0)
F Magnitude of the force. (real)
Bulk Data Image 1-4 Specification of Force at Scalar Points
268 Nastran Primer

External Loads
Distributed Load on Beam - PLOAD1
10.6 Distributed Load on Beam - PLOAD1

The PLOAD1 shown by Bulk Data Image 1-5 is used to specify distributed and/or
concentrated loads along the BAR, BEAM, and BEND element. PLOAD1 generates work-
equivalent forces at the grid points associated with the element.
1 2 3 4 5 6 7 8 9 10
PLOAD1 SID EID TYPE SCALE X1 P1 X2 P2
where
Field Description
SID The set identification number assigned to the load. Must be
selected using a LOAD = <SID> Case Control statement
(integer>0)
EID The identification number of the element to be loaded (integer>0)
TYPE Defines the type of load as described below.
SCALE Distance scaling as described below
(Xi, Pi) Load position and value as described below
Bulk Data Image 1-5 Distributed Loads on Beams-Bending Elements - PLOAD1
10.6.1 Load Type and Direction

The type of load, i.e., force or moment, and the vector sense of the load with respect to
either local element or basic coordinates is specified by an appropriate character strings in the
TYPE field. The allowable character strings are:
TYPE Field Description

FX, FY, FZ Force in the Basic X, Y, or Z directions, respectively
MX, MY, MZ Moment about the Basic X, Y, or Z axes, respectively
FXE, FYE, FZE Force in the local.x, y, or z directions, respectively
MXE, MYE, MZE Moment about the local x, y, or z axes, respectively.
10.6.2 Load and SCALE Fields

PLOAD1 can be used to define a concentrated force or moment at a point along the length
of the element; or, a distributed force or moment. The load is described by:
1. Specifying the region over which the load is to be distributed using the X1 and X2
fields where X1 defines the start of the loaded region and X2 defines its end. If a
concentrated load at X1 is desired then X2 must either be blank or equal to X1.
2. Specifying whether the length is defined as the actual element length or is scaled
by either actual length or projected length, using the following character strings in
the SCALE field.
3.Specifying the load intensity using the P1, P2 fields where
• A concentrated load is defined by a real value for P1 with P2 blank
• A uniform load between X1 and X2 is defined by setting P2 = P1
Nastran Primer 269

External Loads
10
Distributed Load on Beam - PLOAD1
• A linearly varying load is defined between X1 an X2 if P2 ≠ P1.
SCALE Field Description

Distance is actual length along the neutral axis. If X2 > X1 the
LE
load is a force or moment per unit length.
Distances are normalized with respect to total element length. If
FR
X2 > X1 the load is a force or moment per unit length.
Same as LE except loads are defined in terms of projected lengths
LEPR
of the neutral axis on the Basic coordinate axes.
Same as FR except loads are defined in terms of normalized
FRPR
projected length of the neutral axis on the Basic coordinate axes.
10.6.2.1 Example - Load Distribution on Beam Elements

Consider the specification of the loads applied to the structural system as shown by Figure
1-3 where the loads are as follows:
1. A uniform load, pa, having a magnitude 20 lb./in. applied to element 2 in the
negative x-direction in the Basic coordinate system.
2. A concentrated load Pb applied to element 2 at a distance of 10 inches from grid
point 2 (taken to be the A-end of the element) having a magnitude of 30 lb.
3. A concentrated force Pc of 20 lb. and a concentrated moment Md of 60 in.-lb. at the
midpoint of element 1.
2 10
l1 in
P c = 20 lb
M d = 60 in-lb
P a = 20 lb
2
1
2l 1
1/ P b = 30 lb
x
1 3
Figure 1-3 Distributed and Concentrated Forces Along Beam-Type Elements
Assuming that the element z-axes coincide with the Basic Z-axis, these loads could be
described by using the following PLOAD1 data statements.
1 2 3 4 5 6 7 8 9 10
PLOAD1 101 2 FX FRPR 0. -20. 1. -20. 0001
PLOAD1 101 2 FY LE 10. -30. 0002
PLOAD1 101 1 FYE FR 0.5 20. 0003
PLOAD1 101 1 MZE FR 0.5 60. 0004
where statements 0001 thru 0004 define Pa, Pb, Pc, and Md, respectively.
270 Nastran Primer

External Loads
Uniform Normal Pressure - PLOAD2
10.7 Uniform Normal Pressure - PLOAD2

The PLOAD2 data statement shown by Bulk Data Image 1-6 specifies a uniform normal
pressure on the surface of the two dimensional shell elements including the QUAD4, QUAD8,
TRIA3, and TRIA6
1 2 3 4 5 6 7 8 9 10
PLOAD2 SID PRESS EID1 EID2 EID3 etc.
or
PLOAD2 SID PRESS EID1 THRU EIDn
where
Field Description
SID Load set identification number, (integer, > 0)
PRESS Magnitude of the pressure. (real)
Element identification numbers. All elements referenced must
ElDi exist and must be two-dimensional. When using the alternate
form of EID1 THRU EIDn, all elements must exist. (Integer)
Bulk Data Image 1-6 Pload2 - Normal Pressure for Bending Elements
10.7.1 Sign Convention for Positive Pressure

PLOAD2 generates grid point forces in the element coordinate system. The positive sense
of the pressure is determined by the grid number sequence specified on the associated element
connection data statement. The sense of positive pressure is found by applying the right hand
rule to the grid point numbering sequence for the element vertices. A counterclockwise
sequence defines positive pressure in the negative element z-direction.
Since the element grid point numbering sequence defines the sense of positive pressure it
is important that a consistent numbering sequence be used.
Nastran Primer 271

External Loads
10
Nonuniform Surface Tractions - PLOAD4
10.8 Nonuniform Surface Tractions - PLOAD4

A nonuniform surface traction can be applied to the surface of the two dimensional shell
elements or any face of three dimensional TETRA, HEXA or PENTA elements by the
PLOAD4 data statement shown by Bulk Data Image 1-7.
1 2 3 4 5 6 7 8 9 10
PLOAD4 SID EID P1 P2 P3 P4 G1 G3 or G4
CID N1 N2 N3
or, alternatively for two dimensional elements
PLOAD4 SID EID1 P1 P2 P3 P4 THRU EID2
CID N1 N2 N3
where:
Field Description
The identification number for the load. Must be selected sing a
SID
LOAD = <SID> Case Control statement.
EID Identification number of element to be loaded.
P1, P2, P3, and P4 Value of surface traction at vertices of loaded face. (Real)
Grid point IDs used to specify loaded face of three dimensional
G1, G3
elements as described below.
THRU, EIDn Defines set of two dimensional elements to be loaded.
ID of coordinate system used to define vector direction of
CID
traction. (Integer)
N1, N2, N3 Components of traction direction. (Real)
Bulk Data Image 1-7 PLOAD4 - Nonuniform Surface Traction
The alternate form can be used to define a surface traction over a set of TRIA3, QUAD4,
TRIA6, and QUAD8 elements.
10.8.1 Direction of Surface Traction

The direction of the surface traction can be defined by specifying the three real
components of a vector (N1, N2, N3) with respect to the CID coordinate system on the optional
continuation statement. If the continuation is not present then the traction is taken to be normal
to the surface using the following sign convention.
1. For two dimensional elements positive pressure is the direction determined by
applying the right hand rule to the node number sequence as described in the
previous section on PLOAD2.
2. For three-dimensional elements positive pressure is in the negative direction of the
normal to the face.
10.8.2 Specifying Load Intensity

The fields P1, P2, P3, P4 are interpreted as
272 Nastran Primer

External Loads
Nonuniform Surface Tractions - PLOAD4
1. The values of the surface traction at the ordered set of connected grid points G1,
G2, G3, and G4 as defined on the connection statement for a two dimensional
element where P4 is blank if the element is triangular.
2. The values of surface traction at the ordered set of connected vertex grid points on
the face of a three dimensional element. The face is identified by specifying two
grid points, G1 and G3, which define a diagonal on the face as described below. The
field P1 is then taken to be the traction at G1 and P2, P3, (and P4) are tractions at the
other vertices in a sequence determined by applying the right hand rule to the
outward normal to the face.
10.8.3 Specifying the Loaded Face of Solid Elements

The loaded face of a solid element is specified using the G1 and G3 fields as described
below:
1.For a HEXA element, G1 and G3 are diagonally opposite grid point numbers on a
face.
2.For quadrilateral faces of a PENTA, G1 and G3 are diagonally opposite grid point
numbers on a face.
3.For triangular faces of a PENTA, G1 is a grid point number on the face and G3 is
blank.
4.For TETRA elements, G1 is a grid point number on the face and G3 is the grid point
number not on the face.
10.8.3.1 Example -Surface Traction on Three Dimensional Element

Consider the specification of the surface traction applied to the eight-node HEXA element
shown by Figure 1-4.
3 psi
5 psi 2 4 psi
z
1 3
4
EID = 1
6
5
x y 7
Figure 1-4 Surface Traction on Three Dimensional Element
The magnitude on the surface traction is taken to be

5 psi at node 1
3 psi at node 2
4 psi at node 3
Nastran Primer 273

External Loads
10
Gravity Loads - GRAV
0 psi at node 4
and varies linearly along the edge of the element. A surface traction of this magnitude can be
specified in the direction of the Basic X-axis by means of the following PLOAD4 data
statement.
1 2 3 4 5 6 7 8 9 10
PLOAD4 101 1 5. 0. 4. 3. 1 3 +P1
+P1 0 1. 0. 0.
where the face is identified by grid points 1 and three in fields 8 and 9 and P1, P2, P3, and P4
are associated with grid points 1, 4, 3, and 2, respectively.
10.9 Gravity Loads - GRAV

A uniform gravity load can be applied to the structure by specifying an acceleration vector
using the GRAV data statement, which is shown in Bulk Data Image 1-8.
1 2 3 4 5 6 7 8 9 10
GRAV SID CID A N1 N2 N3 MB
where
Field Description
Load set identification number. The SID cannot be the same as
SID any other static load sets. Load sets in Bulk Data can be combined
using the LOAD data statement (see section 11.11). (integer > 0)
Set identification of the coordinate system in which the vector
CID
components N1, N2, and N3 are defined. (integer > 0)
Scale factor such that the desired acceleration is given by
A {a} = A{N}
where {N} need not be a unit vector. (real ≠ 0.)
Components of a vector, {N}, taken in the sense of the
N1, N2, N3 displacement degrees of freedom shown by Fig. 7-4 for the
coordinate type specified by CID. (real)
Used for PART superements to specify which coordinate system
to used in an assembled model.
MB = 0, The coordinate systems in the Main Bulk Data section
MB are considered to be stationary with respect to the assembly Basic
coordinate system (default)
MB = -1, the CID coordinate system is defined in the main bulk
data section
Bulk Data Image 1-8 GRAV - Specification of Gravity Vector
The acceleration vector is transformed to the basic coordinate system and expanded to a
set of vectors acting on the g-set. Letting { a g } be the set of node point accelerations, the
gravity loads are calculated using the structural mass matrix as:
274 Nastran Primer

External Loads
Rotation About an Axis- RFORCE
{ P g } = [ M gg ] { a g } (Eq. 10-1)
where [Mgg] is the system mass matrix.

The inclusion of the GRAV loads in the input implies that the mass matrix must be
generated. The user must therefore insure that the mass for all elements is defined as described
in Specification of Element Inertia Properties (p. 320) of the text.
Sets of gravity loads may be specified by giving each set an identification number that is
unique among all load set identifications. The gravity load to be used as execution time is then
specified by the Case Control Directive
LOAD = SID
It is important to note that NASTRAN does not allow GRAV data statements to have the
same set number as the applied external loads described in previous sections. If both gravity
and applied external loads are desired they must be combined using LOAD data statement,
which is described in Combined Loading - LOAD (p. 277).
10.10 Rotation About an Axis- RFORCE

The force field due to angular velocity and acceleration about a grid point, is shown by
Figure 1-5.
z
Go
ri
Ro
i
Ri
x y
Figure 1-5 Angular Rotation Vector for Centrifugal Forces
The RFORCE data statement is shown in Bulk Data Image 1-9.
1 2 3 4 5 6 7 8 9 10
RFORCE SID G0 CID A R1 R2 R3 METHOD
RACC MB
Nastran Primer 275

External Loads
10
Rotation About an Axis- RFORCE
where
Field Description
Load set identification number. The SID cannot be the same as
SID any other static load sets. Load sets in Bulk Data can be combined
using the LOAD data statement (see section 11.11). (integer, > 0)
G0 ID of a Grid point lying on the axis of rotation. ( integer ≥ 0
Set identification of the coordinate system in which the vector
CID
components N1, N2, and N3 are defined. (integer, > 0)
The magnitude of the angular velocity about the axis of rotation
A
in revolutions per unit time, real.
Components of the rotation vector, R ,taken in the sense of the

R1, R2, R3 displacement degrees of freedom shown by Figure 6-4 for the
coordinate type specified by CID. (real)
Method used to compute centrifugal force vector. (integer = 1 or
METHOD
2, or blank, default = 1)
The magnitude of the angular acceleration about the axis of
RACC
rotation in revolutions per unit time squared, real.
Used for PART superements to specify which coordinate system
to used in an assembled model.
MB = 0, The coordinate systems in the Main Bulk Data section
MB are considered to be stationary with respect to the assembly Basic
coordinate system (default)
MB = -1, the CID coordinate system is defined in the main bulk
data section
Bulk Data Image 1-9 RFORCE - Specification of Angular Velocity Vector
The structure is assumed to rotate such that point Go lies on the axis of rotation. If the Go
field is left blank, then the spin axis passes through the origin of the Basic coordinate system.
The angular rotation vector is defined as
R = R1 i1 + R2 i2 + R3 i3 (Eq. 10-2)
The angular velocity vector in radians per unit time is
ω = 2πAR (Eq. 10-3)
and, the angular acceleration vector in radians per unit time squared is
α = 2π ( RACC )R (Eq. 10-4)
The acceleration at Grid point Gi in the structure is then given by
ai = α × ri + ω × ( ω × ri ) (Eq. 10-5)
where we see from Figure 1-5 that
276 Nastran Primer

External Loads
Combined Loading - LOAD
Ri = ri + R0 (Eq. 10-6)
Applying Newton’s second law of motion, the force at grid point Gi is then:
Fi = mi ai (Eq. 10-7)
Denoting the set of accelerations associated with all geometric grid point degrees of
freedom as {ag}, the centrifugal forces are then calculated to be
{ P g } = – [ M gg ] { a g } (10-8)
10.10.1 Remarks
1.The axis of rotation passes thorough the origin of the Basic Coordinate system if
GID = 0
2.If CID = 0 then the axis of rotation vector is defined relative to the Basic Coordinate
system
3.The inclusion of centrifugal forces in the analysis implies that the system mass
matrix must be calculated. The user must therefore insure that the mass for all
structural elements as described in Section 15.4 of the text is defined. The user is
also cautioned that the use of consistent mass may lead to erroneous centrifugal
forces if the default for METHOD is taken. If METHOD = 2 then the off-diagonal
mass terms which are associated with the consistent mass formulation are correctly
incorporated unless the model contains offset CONM2 data statements.
4.If angular acceleration is included METHOD = 2 ia always used.
5.The axis of rotation must coincide with the axis of symmetry for Cyclic Symmetry.
6.For superelement analysis the reference grid point must be in the residual structure.
10.11 Combined Loading - LOAD

The loads defined in the previous sections are identified by means of a set number. The
load to be used is specified by means of the following Case Control Directive
LOAD = SID
Combinations of loads may be required and can be specified by means of the LOAD data
statement shown by Bulk Data Image 1-10. The LOAD data statement is used to combine sets
of static loads as well as to combine RFORCE and GRAV loads with other static loads.
1 2 3 4 5 6 7 8 9 10
LOAD SID S S1 L1 S2 L2 S3 L3
S4 L4
Nastran Primer 277

External Loads
10
Enforced Displacements - SPCD
where
Set identification of the combined load set. The magnitude of

entire load is given by
SID N
L = S 0 ∑ S i ( LSID i )
i=1
Si (i = 0, 1, 2, N) are scale factors. (real)

(i = 1, 2, N) are LOAD set identification numbers of any load set
LSIDi which can be specified by a LOAD =< SID> Case Control
Directive.
Bulk Data Image 1-10 LOAD - Combined Load Sets.
The static load vector specified by the set of loads used at execution time can be displayed
by the Case Control Directive.
OLOAD = <set>
where <set> is a set of grid and scalar points defined using a SET Case Control Directive.
10.12 Enforced Displacements - SPCD

The SPCD data statement shown by Bulk Data Image 1-11 provides the user with the
capability of defining a constraint that will be treated as load associated with a degree of
freedom in the s-set. The use of the SPCD statement avoids the decomposition of the stiffness
matrix when only the magnitudes of enforced displacements are changed.
1 2 3 4 5 6 7 8 9 10
SPCD SID G1 C1 D1 G2 C2 D2
where:
Field Description
SID Load set identification number. The SID cannot be the same as
any other static load sets. Load sets in Bulk Data can be combined
using the LOAD data statement (see section 11.12, (integer, > 0)
G, C Degree(s) of freedom associated with the enforced displacement
set where G is the grid point number and C is the degree of
freedom code(s). (Integers > 0)
D Value of enforced displacement. (Real)
Bulk Data Image 1-11 SPCD - Displacement Boundary Conditions in Load Set
The SPCD is similar to the SPC data statement described in Chapter 6 except that the set
identification associated with the SPCD refers to a load set and is specified using a LOAD =
<SID> Case Control statement. The following rules must be followed in using the load-set
enforced displacement.
278 Nastran Primer

External Loads
Enforced Deformation - DEFORM
1. The degrees of freedom specified on the SPCD data statement must also be defined
to be members of the s-set by SPC or SPC1 data statements.
2. The value D replaces the value of the constraint defined by the SPC or SPC1
statements only if the appropriate load set is selected in Case Control.
3. The LOAD combination capability can not be used to select SPCD sets.
4. At least one external load specification (FORCE, MOMENT, etc.) must be present
in the input, however, the associated load may be zero.
The SPCD assists the analyst in performing rigid body checks for large structural models.
Using SPCD, displacement boundary conditions can be specified as loading conditions so that
rigid body checks can be performed in a single solution with only one matrix decomposition.
10.13 Enforced Deformation - DEFORM

An enforced axial deformation can be specified for the ROD-type and BAR-type
elements by means of the DEFORM data statement shown by Bulk Data Image 1-12
1 2 3 4 5 6 7 8 9 10
DEFORM SID EID D EID D EID D
where
SID Load set identification. (integer, > 0)

Element number of a ROD, TUBE, CONROD, BAR, or BEAM
EID
element. (integer, > 0)
D Enforced axial deformation, plus for elongation. (real)
Bulk Data Image 1-12 DEFORM - Enforced Axial Displacement
The set of enforced displacements that is to be used is specified by means of the Case
Control Directive
DEFORM = SID
where SID is the identification number of a DEFORM set in Bulk Data.
10.14 Thermal Loading

Element and grid point temperatures are used for the determination of:
1. Thermal loading
2. Temperature-dependent material properties
3. Stress recovery
The analyst selects a specific set of temperature data and specifies whether the
temperatures are to be used to determine temperature-dependent material properties or to
define equivalent thermal loads by the TEMPERATURE Case Control Directive, which has
the format
Nastran Primer 279

External Loads
10
Thermal Loading
 MATERIAL
TEMPERATURE  LOAD  = SID
 BOTH 
where BOTH is the default option. For example,

TEMP(LOAD) = 30
specifies that temperature set 30 is to be used to calculate thermal loads and

TEMP(MATERIAL) = 20
specifies that temperature set 20 is to be used to calculate temperature-dependent material

properties. These specifications are mutually exclusive; i.e. temperature-dependent material
properties will not be calculated if LOAD is specified, and thermal loads will not be calculated
if MATERIAL is specified. Both thermal loads and material-dependent properties are
specified by either of the equivalent forms
TEMP(BOTH) = SID
or
TEMP = SID
If the thermal effects are specified, then all elements must have a temperature field
defined either directly by an element temperature field or indirectly by grid point temperatures.
If both are given then precedence is always given to element temperatures.
The equivalent concentrated thermal loads are calculated as follows
∫ [ be ]
T
{ P ei } = [ E e ] { e ei } dΩ (Eq. 10-9)
Ω
where Ω is the region of the element and the initial strains, {eei},due to thermal effects and are
{ e ei } = { α }∆T (Eq. 10-10)
where {α} is the set of thermal coefficients defined on an appropriate MATi data statement, as
described in Chapter 9.
If grid point temperatures are used to calculate the thermal strains, then the element
temperature is assumed to be the average of the temperature of the connected node points for
the lineal elements. In this case the thermal strains for an element are constant over the element
and are given by:
{ e ei } = { α }T ave (Eq. 10-11)
The effective thermal loads are then given by
∫ [ E ] [ bc ] dΩ
T T
{ P ei } = T ave { α } (Eq. 10-12)
Ω
The temperature at interior points of the isoparametric elements is found by interpolation

of the temperatures at the corner nodes.
280 Nastran Primer

External Loads
Grid Point Temperatures - TEMP and TEMPD
10.15 Grid Point Temperatures - TEMP and TEMPD

The grid point temperature can be specified by means of the TEMP and TEMPD data
statements, which are shown in Bulk Data Image 1-13.
1 2 3 4 5 6 7 8 9 10
TEMP SID G1 T1 G2 T2 G3 T3
TEMPD SID1 T1 SID2 T2 SID3 T3 SID4 T4
where:
Field Description
SID Identification number of the temperature set.
G,T Pairs that define the temperature T (real) at grid point G (integer).
Pairs on the TEMPD statement defining default values for grid
SID, T point temperatures which have not been specified on a TEMP
statement for temperature set SID.
Bulk Data Image 1-13 Specification of Grid Point Temperatures
The TEMP statement specifies a scalar value of temperature at grid points, and the
TEMPD data statement specifies a default temperature for all grid points whose temperatures
are not defined by a TEMP statement.
The grid point temperatures may be calculated in a separate execution of NASTRAN by
specifying the HEAT approach in Executive Control. This feature is especially attractive since
the grid point temperatures are calculated using basically the same finite element model as that
used for the structural analysis.
10.16 . Thermal Field for Axial Elements - TEMPRB

The temperature is assumed to vary linearly over the length of an axial element. If only
axial thermal strains are of interest then it may be reasonable to define grid point temperatures
using the TEMP data statement described in the previous section.
For elements that exhibit bending behavior the situation is not quite so simple since the
temperature gradients through the thickness produce thermal moments. The effective thermal
gradients in the bending elements are calculated using the following relations:
1
T y = ---- ∫ zT ( y, z ) dA
Iy
A
(Eq. 10-13)
1
T z = ---- ∫ yT ( y, z ) dA
Iz
A
Nastran Primer 281

External Loads
10
Thermal Field for Surface Elements
where T(y,z) is the temperature distribution over the plane normal to the neutral axis, and Iy
and Iz are the area moments of inertia of the cross-section about the element y- and z-axes,
respectively. In this case the temperature field is defined using the TEMPRB data statement,
which is shown in Bulk Data Image 1-14.
1 2 3 4 5 6 7 8 9 10
TEMPRB SID EID1 Ta Tb Tya Tyb Tza Tzb
TCa TDa TEa TFa TCb TDb TEb TFb
EID2 EID3 EID4 EID5 EID6 EID7 -etc.-
Alternate form of second continuation
EID2 THRU EIDi EIDj THRU EIDK
where
Field Description
SID Temperature set identification. (integer, > 0)
Unique element identification numbers of elements with specified
ElDi
temperature field. (integer, > 0)
Ta, Tb Average temperature at ends a and b, respectively. (real)
Effective thermal gradient for calculation of thermal moment in a
Tya, Tyb, Tza, Tzb
bending element. (real)
Temperatures at points C, D, E, and F at end i (i = a, b) at which
TCi, TDi, TEi, TFi
stresses are to be recovered in a bending element. (real)
Bulk Data Image 1-14 TEMPRB - Temperature Field for Axial Elements
The bending stresses in a BEAM-type element that result from the mechanical loads are
modified by the element temperatures if at least one of the temperatures TCa, TDa, etc. are
different from zero. The thermal stresses at end a are then given by:
∆σ c = – Eα ( TC a – T ya CY a – T za CZ a – T a )
∆σ d = – Eα ( TD a – T ya DY a – T za DZ a – T a )
(Eq. 10-14)
∆σ e = – Eα ( TE a – T ya EY a – T za EZ a – T a )
∆σ f = – Eα ( TF a – T ya FY a – T za FZ a – T a )
where the material properties are defined using MAT1 data statement and where the distances
to the four points C, D, E and F on the cross-section are specified on an appropriate property
statement. The thermal stress at end b of the element are given by a set of equations that are
similar in form to (Eq. 10-14).
10.17 Thermal Field for Surface Elements

For two-dimensional elements that have bending behavior, a thermal moment will result
if the temperature distribution through the thickness is an odd function of the thickness relative
to the plate neutral surface. The user may specify the effect of the temperature variation
through the thickness by specifying the effective gradient:
282 Nastran Primer

External Loads
1
T′ = --- ∫ zT ( z ) dz (Eq. 10-15)
I
t
on a TEMPP1 data statement as shown by Bulk Data Image 1-15
1 2 3 4 5 6 7 8 9 10
TEMPP1 SID EID1 TBAR TPRIME T1 T2
EID2 EID3 EID4 EID5 EID6 EID7 -etc.-
alternative form for second continuation
“THRU
EID2 EIDI EIDj THRU EIDk
”
1 THRU 10 30 THRU 61
where
SID Load set identification. (integer, > 0)

ElDi Element identification numbers. (integer, > 0)
T Average temperature of the element. (real)
Temperatures at the upper and lower surfaces used only for stress
T1, T2
recovery. (real)
T’ Effective thermal gradient. (real)
Bulk Data Image 1-15 TEMPP1 - Thermal Gradients for Plate
Nastran Primer 283

External Loads
10
284 Nastran Primer

11 Static Analysis
Solution 1 and 101 are used for linear static analysis based on the finite element method
in Legacy and MSC NASTRAN, respectively. These solutions supports all of the modern
NASTRAN features including:
• Modern element library
• Constraint and rigid elements
• Automatic purging of unconnected degrees of freedom
• Distributed surface tractions on two and three dimensional elements
• Inertia relief
Other solution sequences for static analysis include nonlinear strain displacement relations or
nonlinear material behavior and are therefore beyond the scope of the present text.
11.1 System Matrices

The system matrices can include the mass matrix, [ M gg ] , as well as the stiffness matrix,
[ K gg ] and the load vector [ P g ] . These matrices can be
• Generated from a finite element model as described in previous chapters
• Specified directly using, Direct Matrix Input (DMIG)
• Specified using mass elements described in this chapter
• Specified element stiffness using GENEL.
The system matrices associated with the finite element model can be defined and
calculated by NASTRAN by
• Defining degrees of freedom using GRID and SPOINT data statements as described in
Chapter 6.
• Defining structural element connectivity and properties using the recommended
elements described in Chapter 5.
• Defining material properties using MAT-type data statements described in Chapter 10.
• Defining inertia properties as described in this chapter.
11.1.1 Direct Specification of System Matrices

The stiffness and mass matrices associated with g-set of displacements are of the
following form:
285 Nastran Primer

Static Analysis
11
System Matrices
1 2
[ K gg ] = [ K gg ] + [ K gg ]
1 2
[ M gg ] = [ M gg ] + [ M gg ] (Eq. 11-1)
1 2
{ Pg } = { Pg } + { Pg }
where the subscript, g, refers to the all inclusive set of displacements defined on Grid and
1 1
SPOINT data statements; and, where [ K gg ] and [ M gg ] are the stiffness and mass matrices
2 2
generated using the finite element model and where [ K gg ] and [ M gg ] are stiffness and mass
coefficients which can be defined directly by using the DMIG data statement shown by Bulk
Data Image 11-1.
1 2 3 4 5 6 7 8 9 10
*Header statement
DMIG NAME 0 FORM TIN TOUT 0001
*Column definition statement
DMIG NAME Gcol Ccol Gr1 Cr1 REr1,col IMr1,col 0002
+001 Gr2 Cr2 REr2,col IMr2,col Gr3 Cr3 REr3,clo IMr3,col 0003
+002 Gr4 Cr4 REr4,col IMr4,col etc. 0004
where:
Field Description
NAME The user-defined character tag for the matrix data block. (Unique
Character string, 1 to 8 characters)
FORM Form code for matrix data block as defined below. (integer > 0)
TIN Type of matrix element defined in DMIG as described below.
(integer > 0)
TOUT Type of matrix to be generated as described below. (integer > 0)
Gcol, Ccol Column number defined by grid number and a single degree of
freedom code. (Integers)
Gri, Cri Row number defined by grid number and a single degree of
freedom code, (Integers)
Real and imaginary components of matrix coefficient associated with row
REr1,col, IMr3,col
(Gri, Cri) in column (Gcol, Ccol). (Real or double precision)
Bulk Data Image 11-1 DMIG - Direct Matrix Input
The matrix data block is associated with an internal name by the ‘NAME’ field which is
a character string that must begin with an alphabetic character. The matrix is then specified by
• A header data statement having a zero, (0), in field 3 and where
FORM Matrix form defined by Table 11-2. (integer >0)

TIN Type of input defined by Table 11-3. (integer >0)
TOUT Type of output defined by Table 11-3 (integer >0)
• A DMIG statement having the same NAME as the header for each non-null column
286 Nastran Primer

Static Analysis
System Matrices
Table 11-2 Matrix FORM Codes

FORM CODE MATRIX FORM
1 Square but not Symmetric
2 General Rectangular
3 Diagonal (N = 1, M = no. of rows)
4 Upper Triangular Factor
5 Lower Triangular Factor
6 Symmetric
7 Row (M = no. of Columns, N = 1)
8 Unit (M = no. of rows, N = 1)
Table 11-3 Matrix Type Codess

TYPE CODE NUMERICAL REPRESENTATION OF MATRIX
0 Pick best for machine configuration
1 Real, single precision
2 Real, double precision
3 Complex, single precision
4 Complex, double precision
Each nonzero coefficient in the column must be specified by identifying the row
associated with the element followed by the value of the matrix element. DMIG has the general
capability of defining complex matrix data blocks so that the row designation is followed by
two fields, the first, REij, is the real part and the second, IMij, is the imaginary part. Since only
real matrices are used in static and normal modes analysis the second field must be left blank.
The more general capability for defining complex matrices is described in Bulk Data Image
11-1.
2 2
The matrices defined on DMIG Bulk Data can be selected as K gg or M gg type input by the
following Case Control Directives
2
• M2GG=<name> for [ M gg ] input
2
• K2GG=<name> for [ K gg ] input
where <name> is the NAME assigned to the matrix data block on the DMIG statement.
Consider the specification of the stiffness matrix for the cantilever beam shown by
Figure 11-1, for example
z
2EI = 1
1 2 3
x
l=1 l=1
Figure 11-1 Cantilever Beam Defined by DMIG
Nastran Primer 287

Static Analysis
11
System Matrices
The stiffness matrix for the beam was determined for a bending in the x-y plane in Chapter
5, (Eq. 4-26). It is not difficult to show that the stiffness matrix associated with bending in the
x-z plane is:
6 – 3l – 6 – 3l
2EI – 3l 2l 2 3l l 2
[ k ee ] = --------
3
- (Eq. 11-2)
l – 6 3l 6 3l
2 2
– 3l l 3l 2l
Substituting the values for l and EI gives
1 –3 –6 –3
[ k ee ] = 3
– 2 3 1
–6 3 6 3
–3 1 3 2
Using the techniques described in Chapter 6 to assemble the system matrices and apply the
displacement constraints then leads to the following 4 x 4 matrix for the solution set
displacements
12 0 –6 –3
[ K ll ] = 0 4 3 1
–6 3 6 3
–3 1 3 2
where for the purpose of this example the g-set includes the following physical degrees of
freedom:
T
{ u g } = u z1 θ y1 u z2 θ y2 u z3 θ y3
and the l-set is:
T
{ u l } = u z2 θ y2 u z3 θ y3
The stiffness matrix could then be defined by the following data statements.
1 2 3 4 5 6 7 8 9 10
GRID 2 0 1. 0. 0.
GRID 3 0 2. 0. 0. 0001
DMIG BEAM 0 6 1 0 0002
DMIG BEAM 2 3 2 3 12. 0003
+003 3 3 -6. 3 5 -3. 0004
DMIG BEAM 2 5 2 5 4. 0005
+005 3 3 3. 3 5 1. 0006
DMIG BEAM 3 3 3 3 6. 0007
+007 3 5 3. 0008
288 Nastran Primer

Static Analysis
Constraint and Static Condensation
1 2 3 4 5 6 7 8 9 10
DMIG BEAM 3 5 3 5 2. 0009
where NAME = BEAM. The first DMIG, statement 0002 in field 10, is the header specifying
a symmetric matrix, (FORM = 6), whose elements are real single precision, (TIN = 1), whose
output type will be appropriate for the type of computer (TOUT = 0).
1. The specification of symmetric matrix on DMIG implies that a given off-diagonal
element can be input either above or below the diagonal, but not both. Thus only
the diagonal and either upper or lower off-diagonal elements are to be entered for a
symmetric matrix.
2. The external column and row codes represent the non-null degrees of freedom in
the displacement set.
3. The stiffness matrix would be selected by the following Case Control Directive
K2GG = BEAM
11.2 Constraint and Static Condensation

The stiffness matrix associated with the set of all grid and scalar degrees of freedom can
be modified by means of various constraint and partitioning operations. These include
1. Specification of multipoint constraints Multipoint Constraints - MPC (p. 175))
2. Specification of single point constraints Single Point Constraints - SPC (p. 184)
3. Static Condensation Static Condensation -OMIT and ASET (p. 189)
4. Support of free bodies Support for Free Bodies - SUPORT (p. 193)
5. Specification of matrix elements using Flexibility to Stiffness Transformation
(p. 196)
The set of displacements remaining after these operations is called the l-set which is the
solution set for static analysis.
11.3 Static Loads

The set of static loads that is applied to the grid and scalar points can result from a variety
of sources, including
1. Concentrated forces and moments applied directly to the grid points Concentrated
External Forces at Grid Points (p. 264)
2. Concentrated loads applied to scalar points Load Applied to Scalar Point - SLOAD
(p. 268)
3. Distributed pressure load on one-, two-, and three-dimensional elements Uniform
Normal Pressure - PLOAD2 (p. 271)
4.Gravity loads Gravity Loads - GRAV (p. 274)
5. Centripetal acceleration Rotation About an Axis- RFORCE (p. 275)
6. Enforced boundary displacement Enforced Displacements - SPCD (p. 278)
7. Enforced axial deformation Enforced Deformation - DEFORM (p. 279)
8. Temperature field Grid Point Temperatures - TEMP and TEMPD (p. 281)
Nastran Primer 289

Static Analysis
11
Inertia Relief
The set of node point forces from all these effects is reduced to the l-set of displacements
by using the transformation defined by the constraints and partitioning specifications. The
resulting set of equations is of the form
[ K ll ] { u l } = { P l } (Eq. 11-3)
This set of equations is solved by first decomposing the stiffness matrix into its lower and
upper triangular factors. A forward-backward substitution is then performed for all load
subcases which have the same set of constraints.
11.4 Inertia Relief

A free body may be subject to a set of external forces which are not in static equilibrium
and which therefore produce rigid body accelerations. If the rate of change of the external
forces is sufficiently small compared to the lowest natural frequency, we may consider the
external forces to be in equilibrium with inertia forces.
The analyst may choose to apply a set of forces that represents a non equilibrium set. The
analyst may be interested in the effect of the inertia forces. In this case, a condition must be
defined in which the effects of the inertia forces are not present, hence the term "inertia relief."
The general procedure for removing the effects of inertia is
1. Select a set of determinate support points, e.g., the r-set, which is defined using a
SUPORT data statement
2. Determine the rigid body acceleration of the r-set from the loads applied to the
a-set of degrees of freedom
3. Calculate the rigid body accelerations of the l-set of degrees of freedom
4. Subtract the inertia forces that result from accelerations of the f-set from the
external load, {Pl}, and solve for the set of displacements obtained by constraining
the r-set. The forces of reaction of the r-set of degrees of freedom will be zero.
The equations of motion that include inertia relief are obtained by noting that the a-set is
partitioned into the l- and r-sets where the degrees of freedom in the r-set are just sufficient to
remove rigid body motion. The l- and r-sets are then related by the rigid body transformation
matrix defined by (7-63). The reduced mass matrix which is obtained by applying the same set
of transformations that were used to reduce the stiffness is then obtained by partitioning the
a-set of equations into the l- and r-sets as:
[ M ll ] [ M lr ]
[ M aa ] = (Eq. 11-4)
[ M rl ] [ M rr ]
˜
The reduced mass matrix is obtained by means of the transformation:
T
[ M rr ] = [ G ar ] [ M aa ] [ G ar ] (Eq. 11-5)
Where the transformation matrix relates the a- and r-sets and is expressed in terms of the rigid
body transformation matrix (Eq. 6-21) as:
[ G lr ]
{ ua } = { u r } = [ G ar ] { u r } (Eq. 11-6)
[ I rr ]
290 Nastran Primer

Static Analysis
Data Recovery
The reduced mass matrix then becomes:
T T T
[ M rr ] = [ M rr ] + [ M lr ] [ G lr ] + [ G lr ] [ M lr ] + [ G lr ] [ M ll ] [ G lr ] (Eq. 11-7)
˜
The set of forces associated with the r-set is reduced in a similar way to give:
T
{ P l } = [ G lr ] { P a } (Eq. 11-8)
The equations of motion for the supported r-set can be written as:
[ M rr ] { u·· r } = { P r } (Eq. 11-9)
The accelerations of the r-set can now be determined as:
–1
{ u·· r } = – [ M rr ] { P r } (Eq. 11-10)
and by using (Eq. 6-21), the accelerations of the l-set become:
{ u·· l } = [ G lr ] { u·· r } (Eq. 11-11)
The inertia forces associated with the l-set can now be determined in terms of the
partitions of the matrix as:
{ q l } = – [ M ll ] { u·· l } – [ M lr ] { u·· r } (Eq. 11-12)
where { q l } are the inertia forces. These forces can be expressed by using (Eq. 11-10) and
(Eq. 11-11) as:
–1
{ q l } = – ( [ M ll ] [ G lr ] + [ M lr ] ) [ M rr ] { P r } (Eq. 11-13)
The inertia forces are then added to the set of applied loads when inertia relief effects are
desired.
11.5 Data Recovery

The displacement sets associated with the various constraint and static condensation
operations are recovered to obtain the g-set. The details concerning this recovery process are
described in Chapter 6. Once the displacements have been recovered, the element forces,
stresses, forces of constraint, and strain energy are selectively calculated based on
user-supplied Case Control Directives that are described in Chapter 1.
11.6 Input Specifications
11.6.1 Executive Control Section

The Executive Control section is described in Chapter 1. The user specifies static analysis
by means of the SOL statement. A minimum Executive Control deck includes
SOL 101
CEND
Nastran Primer 291

Static Analysis
11
Input Specifications
11.6.2 Case Control Section

The Case Control section for static analysis and static analysis with inertia relief are
similar. The Case Control directives that apply to both are:
1.Output MUST be requested in Case Control. The following are appropriate for
Static Analysis
Directive Action
DISP Output a set of displacement as specified by a preceding
SET directive or the set ALL to write all displacements
on the selected file (the .out file by default or the
PUNCH file.)
ESE Calculate and write element strain energy to the selected
file (the .out file by default or the PUNCH file.)
GPFORCE Calculate and write grid point force balance to the
selected file (the .out file by default or the PUNCH file.)
MPCFORCE Calculate and write MPC forces at the grid points to the
selected file (the .out file by default or the PUNCH file.)
OLOAD Write the input forces at the grid points to the selected
file (the .out file by default or the PUNCH file.)
SPCFORCE Calculate and write the output to the selected file (the
out file by default or the PUNCH file.
STRAIN Calculate the element strain for the set of elements
specified by a preceding SET directive. The element
stresses and dispalcements are written to the .out file
and also to an external file called <name>.in where
<name> is the name of the input file with no extension.
The strain values at the grid points are the volume-
weighted average of the stress components for all
elements which connect at a grid point.
STRESS Calculate the element stress for the set of elements
specified by a preceding SET directive. The element
stresses and dispalcements are written to the .out file
and also to an external file called <name>.in where
<name> is the name of the input file with no extension.
The stress values at the grid points are the volume-
weighted average of the stress components for all
elements which connect at a grid point.
2.A separate subcase must be defined for each unique combination of constraints and
static loads
3. A static load must be defined above the subcase level with a LOAD,
TEMPERATURE (LOAD), DEFORM, or SPC Case Control Directive. Loads
defined within a subcase supersede the load defined above the subcase level. If
loading results from SPC sets, then at least one specified displacement degree of
freedom must be nonzero.
4. An SPC set must be specified for each subcase unless the body is a properly
supported free body, or all constraints are specified on GRID data statements, scalar
connections (i.e., springs to ground) or with the general element.
292 Nastran Primer

Static Analysis
Solution Sequence Output
5. Loading conditions associated with same set of constraints should be contiguous

subcases.
6. The REPCASE directive may be used to allow multiple data recovery
specifications for the same displacement set.
7. A single set of direct input matrices may be selected with K2GG, M2GG and P2G
Case Control directives above the subcase level.
For inertia relief, a SUPORT Bulk Data entity must be selected to remove at least one
rigid body degree of freedom. Constraints can be used to remove some, but not all, of the rigid
body degrees of freedom. If the SUPORT data statement is present then the mass matrix is
required. An error condition exists if the mass matrix is null.
11.6.3 Parameters
The parameters which are useful for static analysis are described in Appendix B. They
include
• GRDPNT -Turns on the weight generator.
• WTMASS -Conversion factor applied to mass matrix.
• AUTOSPC - Automatically purges unconnected degrees of freedom.
• COUPMASS -Optionally calculates consistent rather than lumped mass
11.7 Solution Sequence Output

Although most of the solution sequence output which describes the structural behavior is
optional and is requested in the Case Control section, some of the output is automatic or under
control of DIAG requests in the Executive Control section. The printer output is designed for
132 characters per line, with the lines per page controlled by the LINE directive in Case
Control. The default is LINE = 50. Optional titles defined by TITLE, SUBTITLE, and LABEL
Case Control Directives are printed at the top of each page. Titling directives may be defined
at the subcase level. The pages are automatically dated and numbered, and the NASTRAN
generation date is specified.
All of the output from data recovery and plot modules is optional, and its selection is
controlled by appropriate Case Control directives. The details of making printer selections in
Case Control are described in Section 1.11 for printer and punch file output. Detailed
information on the force and stress output available for each element type is described in
Chapter 6.
A few printer output items are under the control of PARAM statements. The use of
PARAM statements is described in Appendix B. The DIAG Executive Control directive is
used to control the printing of the compiled DMAP (DIAG 14) and executive tables such as
the File Allocation Table (FIAT) and the Operation sequence array (OSCAR). The use of the
DIAG statement is described of the Chapter 1.
11.7.1 Automatic Output

The first part of the output for a NASTRAN run is prepared during the execution of the
Preface, prior to the execution of the DMAP sequence. The following output is either
automatically or optionally provided during the execution of the Preface:
1. NASTRAN title page - automatic.
2. Executive Control echo - automatic.
Nastran Primer 293

Static Analysis
11
Fatal Errors
3. Case Control echo - automatic.

4. Unsorted Bulk Data echo - optional, selected in Case Control using ECHO =
UNSORT.
5. Sorted Bulk Data echo - automatic, unless suppressed using ECHO = NONE in
Case Control.
The Grid Point Singularity Table is automatically output if essential singularities are
identified in the stiffness matrix. This table lists singular degrees of freedom in the
displacement coordinate system. These singular degrees of freedom are automatically purged
unless inhibited by the user by an AUTOSPC as described in Grid Point Singularity Processing
(p. 295).
11.7.2 Output of System Response Variables

NASTRAN is designed to solve large problems which may include several tens of
thousand variables and can produce a very large amount of data. The data base for large
problems may be several gigabytes. Data recovery and printout is therefore performed under
user control. The output options provided as a part of the static solution include the following
Case Control directives
1. Displacements at selected GRID points - DISP
2. Nonzero components of applied loads - OLOAD
3. Nonzero components of single point constraint forces - SPCF
4. Grid point force balance of selected points - GPFO
5. Element forces in selected elements - ELFO
6. Stress in selected elements - STRESS
7. Strain energy in selected elements - ESE
11.8 Fatal Errors

There are two classes of fatal errors in NASTRAN:
1. Those that occur during the execution of a DMAP module
2. Those which are associated with a logical inconsistencies during execution of the
DMAP instructions
The first type of fatal error is identified with a message in the printout and is associated with a
message number. The message numbers are assigned in groups as follows
1-1000 Errors found in the form of data deck (termed Preface Errors)
1001-200 Errors found by the executive DMAP modules
0
2001-899 Errors found by the functional modules
9
For example, the message
*** USER FATAL ERROR 2025, UNDEFINED COORDINATE SYSTEM 102
means that the user has made an error. A reference has been made to coordinate system 102
but that coordinate system has not been defined on CORD-type data statement.
294 Nastran Primer

Static Analysis
Grid Point Singularity Processing
All user and system errors are tabulated by message number in Ref. [2] In most cases the
description of the errors in the reference is more informative than the terse message which is
printed in the output.
The other class of error is associated with logical errors in the DMAP Solution Sequence
which sets the values of certain variable parameters to cause a jump to an error condition in the
DMAP sequence. The contents of the variable parameter table, which includes the values of
all variable parameters, is then printed and NASTRAN terminates normally. In these cases the
fact that the variable parameter table has been printed is the clue which tells us that an error
condition has been found. The variable parameters which control this action, and the associated
logical error are described by Table 11-4.
Parameter Name Meaning

NOL There are no independent degrees of freedom in the f-set
NOMGG The mass matrix has not been generated and is required
NOKGG The stiffness matrix has not been generated
SING Matrix decomposition of a singular matrix has been attempted
NOGPDT No grid or scalar points have been defined
TEST An incompatibility exists between a SOL subset and the Case
Control looping directives
Table 11-4 Variable Parameters Used for Error Exit in Statics

The diagnostic messages may be either user or program diagnostics that contain
warnings, information, or indication of fatal errors. These messages are usually identified by
number and are documented in the NASTRAN User’s Manual. These diagnostic messages are
identified by three asterisks that precede the message. The messages are self-explanatory and
are thus not included in this text.
In addition to the numbered diagnostic messages there are a few self-explanatory
messages, such as the time required for matrix decomposition and the grid point singularity
table if potential singularities are found.
11.9 Grid Point Singularity Processing

The assembled [Kgg] stiffness matrix will be singular since it contains rigid body motion
and is therefore rank deficient. Hopefully proper use of SPC, MPC and SUPORT will remove
of rigid body motion. But even then it may happen that the run fails during the solve operation
due to rank deficiency. This could be caused by a mechanism in the model but is more than
likely caused by the presence of one or more zero terms on the diagonal. The objective of grid
point singularity processing is to detect and to allow the user to cause the singular degree of
freedom to be removed.
The details of the process used by Legacy NASTRAN and the MSC and NX versions to
detect singularities is different. In both cases the 3 by 3 submatrices of the translational and
rotational degrees of freedom are formed and each submatrix is tested for rank. Those degrees
of freedom that are identified as potentially singular are identified. Those that are not removed
by SPC, MPC or other means are flagged as being singular and unless removed will lead to
failure during the solve operation.
The user control of this process for Legacy NASTRAN and MSC Nastran are described
in following sections.
Nastran Primer 295

Static Analysis
11
11.9.1 Legacy NASTRAN

If singularities are detected, the resulting action is controlled by an integer parameter
called AUTOSPC. (A character parameter of the same name is used by MSC NASTRAN as
described below.) The action taken is then dependent of the value of AUTOSPS as described
by Table 11-5:
Table 11-5 AUTOSPC Actions
Value Action
0 (Default) Write a table of singular degrees of freedom and terminate
>0 Write a list of SPC1 Bulk Data that will be added to the current s-
set and proceed
<0 Write a list of SPC1 Bulk Data that will be added to the current s-
set, and, in addition, write the spc1 to the punch file before
proceeding
11.9.2 MSC and NX NASTRAN

The most common modeling situation that leads to singular degrees of freedom occurs
when the elements chosen to represent the structure do not connect all of the structural degrees
of freedom at a grid point. The criteria used to identify singular degrees of freedom are thus
biased toward that situation. The user should carefully review the list of singularities which are
identified and removed from the analysis set.
The local grid point singularities are identified by examining the stiffness matrix, [ K gg ] ,
by the following procedure.
1. Calculate the principal stiffness, K i , and principal direction of stiffness, { d i } , for
each 3 x 3 diagonal partitions of [ K gg ] .
2. Normalize the three principal stiffness as follows
Ki
R i = -----------
- if K max > 0 i = 1, 2, 3
K max (Eq. 11-14)
Ri = 0 if K max = 0
3. Compare the ratio, R i , to a small positive number, EPZERO, which is defined by

a parameter
If R i < EPZERO then a potential singularity exists and can be removed by including the degree
of freedom having a direction closest to the principal direction, { d i } , in the s-set. The list of
potentially singular degrees of freedom is then compared with the displacement set
membership as defined by MPCs and SPCs. A potentially singular degree of freedom which is
not included in the m- or s-sets is then classified as a true singularity. Singular degrees of
freedom are automatically purged by appending them to the user-defined s-set unless the
AUTOSPC parameter is explicitly set to NO by a PARAM statement.
11.9.2.1 User Control

The user can exercise control over the identification of potential singularities as well as
their removal by means of the following parameters.
296 Nastran Primer

Static Analysis
AUTOSPC Parameter
YES Automatic purge i.e. append identified singularities to the s-set.

(Default)
NO No purge
EPZERO Singularity test value
10.-8 (Default)
PRGPST Controls the printout of singularity messages
YES Print messages (Default)

NO No messages
SPCGEN Controls SPC generation

>0 Generate SPC Bulk Data on the system PUNCH file for all user
defined and internally generated SPCs
0 No data statements generated (Default)
11.9.2.2 Automatic Output

The grid point singularity table is automatically printed if the associated control
parameter PRGPST is YES. An example of the table which could result due to the definition
of a planar bending problem is as follows:
GRID POINT SINGULARITY TABLE
POINT Failed Stiffness Old New
ID TYPE Direction Ratio USET USET
1 G 1 0. L S
1 G 2 0. L S
1 G 4 0. L S
1 G 6 0. L S
2 G 1 0. L S
2 G 2 0. L S
2 G 4 0. L S
2 G 6 0. L S
3 G 1 0. L S
3 G 2 0. L S
3 G 4 0. L S
3 G 6 0. L S
This table contains one line for each degree of freedom which is not elastically connected
and which has not been explicitly purged by SPC-type input. Degrees of freedom which have
been purged by the action of PARAM, AUTOSPC, YES are identified by an asterisk following
the last "New Uset" entry. The list can be lengthy if, for example, AUTOSPC is used to purge
unconnected degree of freedom for a solid element model.
Nastran Primer 297

Static Analysis
11
11.9.3 Displacement Set Membership

It is often necessary for the analyst to know the displacement set membership of each
degree of freedom and the internal equation number associated with each degree of freedom in
a specific set. For example, it is inconsistent and improper to define a degree of freedom in
more than one of the removed displacement sets. That is, with reference to Table 6-6, a degree
of freedom cannot be defined in more than one of the m-, s-, o-, or r-sets. If the analyst attempts
to do so the following error message will be printed
***USER FATAL MESSAGE 2101A, GRID POINT 21 COMPONENT 3 ILLEGALLY
DEFINED IN SETS M AND O
The user control for printing the Set Membership Tables for Legacy and MCSC and NX
NASTRAN is decrined ine the following sections.
11.9.3.1 Legacy NASTRAN
298 Nastran Primer

Static Analysis
The print of the set memebership tables is controles by DIAG 21 and DIAG 22 as
descrined in the follwing table:
DISP Action
21 Print a table whoserows are degrees of freedom and
whose columns are:
Internal dof numbers
External GRID and dof code
SAUTO - S-set dof generated by GPSP
SB - SPC Bulk Data
SG - Permanent SPC
L - Set of solution set dof
A - Retained set associated with ASET/OMIT
F - Free set
N - MPC Independent set
G - Set of all dof
O - Omitted set
S - Union of all SPC sets
M - Dependent set on MPC
22 Contents of various displacement sets
The DIAG 21 format produces a very long table since there is one line per degree of
freedom. However, it is also relates the internal dof number to GRID dof.
11.9.3.2 MSC and NX NASTRAN

The set membership tables are printed where the user can optionally select the type and
number of tables to be printed by means of the parameters USETPRT and USETSEL as
follows:
Parameter Name Parameter Value Action

USETPRT Print a table showing membership of each
0 (Default)
degree of freedom
Print selected displacement sets specified
1
by USETSEL parameter
>1 Both of the above
USETSEL 1 M-set
2 S-Set
4 O-set
8 R-set
16 G-set
32 N-set
64 F-Set
128 A-Set
256 L-Set
Nastran Primer 299

Static Analysis
11
Grid Point Weight Generator
Parameter Name Parameter Value Action

512 SG-Set (PSPC on Grid statements)
1024 SB-Set (SPC statements)
647 Union of SB, M, O, and A sets
The USETSEL values define a bit position associated with a specific set. The above
values are an abbreviated set which are appropriate for static analysis. For the complete set of
values see the Quick Reference Guide.
11.10 Grid Point Weight Generator

Weight and balance information can be calculated in any of the solution sequences for
structural analysis. The calculation is requested by the inclusion of a GRDPNT parameter. The
value of the parameter defines the reference GRID point. A value of zero (0) is used to specify
the origin of basic coordinates as the reference point. A blank field for the value is not allowed.
The grid point weight generator calculates the mass, center of gravity and mass moments
of inertia for the finite element model of the structure. These data are calculated from the
system mass matrix by a rigid body transformation about a reference point that is specified on
the GRDPNT parameter.
11.10.1 Rigid Body Transformation Matrix

Let { u 0 } represent the six degrees of freedom associated with the reference point. The
rigid body transformation matrix, [ D g0 ] ,that relates the displacement of the set of all grid point
displacements to displacements of the reference point:
{ u g } = [ D g0 ] { u 0 } (Eq. 11-1)
where [ D g0 ] is a g-by-6 matrix.
11.10.2 Rigid Body Mass Matrix

The rigid body mass matrix is calculated with respect to the reference point by the
transformation:
T
[ M 0 ] = [ D g0 ] [ M gg ] [ D g0 ] (Eq. 11-2)
which results in a 6-by-6 rigid body mass matrix. The mass matrix may have directional
properties because of scalar mass effects.
11.10.3 Principal Mass Axes

The set of grid point displacements is partitioned into translational (subscript t) and
rotational degrees of freedom (subscript r) so that the mass matrix can be partitioned as:
[ M tt ] [ M tr ]
[ M0 ] = (Eq. 11-3)
T
[ M tr ] [ M rr ]
300 Nastran Primer

Static Analysis
Example Problems
If the [ M tt ] matrix is not diagonal (i.e. if the elements of the off-diagonal terms are larger
than some preassigned small value) then the principal masses and associated principal axes of
the translational mass matrix are calculated. The transformation, [ S ] , from basic coordinates
to principal directions is defined in terms of the normalized eigenvectors as:
[ S ] = { e1 } { e2 } { e3 } (Eq. 11-4)
Each of the partitions of [ M 0 ] is then transformed to the principal axes of [ M tt ] as follows:
T
[ M tr ] = [ S ] [ M tr ] [ S ]
T
[ M rr ] = [ S ] [ M rr ] [ S ] (Eq. 11-5)
T
[ M tt ] = [ S ] [ M tt [ S ] ]
11.10.4 Centroid
The position of the center of gravity may be different when taken with respect to different
mass directions since the diagonal elements of the translational mass elements may be different
if scalar masses are specified (see Specification of Element Inertia Properties (p. 320)). The
location of the center of mass is therefore calculated with respect to the reference point for each
mass direction.
11.10.5 Moments of Inertia

The moments of inertia with respect to the principal mass axes at the center of gravity are
determined using the parallel axis theorem for each of the mass systems. The resulting inertia
tensor is then labeled [ I ( [ S ] ) ] .
The set of principal directions, [ Q ] , is then found so that the principal moments of inertia
are given by:
T
[ I( [ Q]) ] = [ Q] [I[ (S )] ][ Q] (Eq. 11-6)
11.11 Example Problems

General purpose finite element programs such as NASTRAN are designed to solve large
problems represented by several thousand degrees of freedom. However, it is always a good
idea to build up a base of experience and confidence in our computational tool by exploring all
of its features and capabilities for the solution of relatively simple problems before moving to
the complex world of real-life structures.
We will, therefore, use NASTRAN to solve several problems for which solutions are
known including:
1. A uniformly loaded cantilever beam
2. A stringer stiffened panel
These problems will allow us to use- several of the elements and modeling features together
with data recovery features in NASTRAN including
• Use of BAR, BEAM, QUAD4 and HEXA elements
• Use of single point constraints (SPC)
Nastran Primer 301

Static Analysis
11
Cantilever Beam with Uniform Load
• Use of PLOAD1, PLOAD2 and PLOAD4

• Grid point weight generator
• Printing displacement set membership tables
• Data recovery including
• Node point displacements
• Element forces and stresses
• Element strains and curvatures
• Single point Constraint forces
While no one example problem will incorporate all of these features, each will be
included in at least one of the examples which are described in the following sections.
11.12 Cantilever Beam with Uniform Load

The structural element for this example is shown by Figure 11-1.
z y’
30° y
z‘
p = po
1
h
2
L 4 x
3
b
Figure 11-1 Beam Element
Where the beam has the following geometric and elastic properties
E = 30 x 106 psi
ν= 0.3
L = 10 in.
h = 0.5 in.
γ = 0.281 Ibf/in.3
b = 0.25 in.
302 Nastran Primer

Static Analysis
and transverse shear coefficients, Ky = Kz = 0.833. The uniform applied loading, po = 10 lb./in.,
acts in a plane rotated 60 degrees from the Basic X-Z plane as shown.
The element y-z coordinates of the cross section are chosen so that they coincide with the
Basic Y-Z coordinates. The load is then defined either in terms of its components in the Basic
Y and Z directions or in terms of components in the element y-z directions. For the purpose of
this example the loads will be described as components referred to the Basic Y-Z coordinates.
The loads are modeled using three load subcases using suitable SUBCASE directives
a. Uniform load of magnitude posinθ in Basic Y-Direction using PLOAD1 with
basic coordinate option (i.e., TYPE = FY).
b. Uniform load of magnitude pocosθ in basic Z-direction using PLOAD1 with
basic coordinate option (i.e., TYPE = FZ).
c. Combination of two previous subcases using the LOAD data statement that
combines the results of the two previous subcases to obtain response to actual
load vector.
11.12.1 Element Properties

With reference toFigure 11-1, (Eq. 5-3) through (Eq. 5-5) are used to calculate the
moments of inertia with respect to the principal axes (y, z). The principal moments of inertia
are:
Iyy = 0.00260 in4
Izz = 0.00651 in4
Iyz = 0
A = 0.125 in2
The orientation of element axes is described using the components of the orientation
vector on the CBAR data statement. Taking the components as:
{V} = {0., 1., 0.}
will align the element y-axis with the Basic Y-axis. The area properties associated with the
principal axes are then specified using the PBAR data statement. The CBAR and PBAR data
statements for this case are:
1 2 3 4 5 6 7 8 9 10
CBAR 1 269 1 2 0. 1. 0. 0001
PBAR 269 30 .125 6.51-4 2.604-4 0002
+001 -.125 -.25 .125 -.25 .125 .25 -.125 .25 0003
+002 .833 .833 0004
where the stress recovery points at the corners and the transverse shear factors are specified on
continuations.
11.12.2 Distributed Line Load

The distributed line load for this case is specified by two PLOAD1 as described above.
The PLOAD1 together with a CBARAO to recover element force and stress output at 10
intermediate points along the length of the Bar for this case are:
1 2 3 4 5 6 7 8 9 10
PLOAD1 100 1 FY FR 0. 5. 1. 5.
PLOAD1 200 1 FZ FR 0. 8.6603 1. 86603.
Nastran Primer 303

Static Analysis
11
1 2 3 4 5 6 7 8 9 10
CBARAO 1 FR 9 .1 .1
11.12.3 The Input File

We will now create the complete NASTRAN input file that
• Requests static analysis, solution 101
• Requests SPC set 23 which will define geometric boundary conditions at X = 0.
• Purges unconnected degrees of freedom using AUTOSPC
• Selects static load set 101 which defines a uniform load by means of PLOAD1 Bulk
Data
• Requests displacements, SPC forces, element forces, and stresses at corner points as
output
• Generates weight and balance information using GRDPNT parameter
The NASTRAN input is then
1 2 3 4 5 6 7 8 9
SOL 101 $ REQUESTS STATIC SOLUTION SEQUENCE 0001
TIME 100 $ SETS SOLUTION TIME TO 100 MINUTES 0002
CEND 0003
TITLE=UNIFORMLY LOADED CANTILEVER BEAM 0004
SUBTITLE= PRINCIPAL ELEMENT AXES 0005
SPC = 23 0006
DISP = ALL 0007
ELFO = ALL 0008
STRESS = ALL 0009
OLOAD = ALL 0010
SPCFORCE = ALL 0011
SUBCASE 1 0012
LOAD = 100 0013
SUBCASE 2 0014
LOAD = 200 0015
SUBCASE 3 0016
LOAD = 300 0017
BEGIN BULK 0018
PARAM GRDPNT 0 0019
PARAM AUTOSPC YES 0020
PARAM USETSEL 1746 0021
GRDSET 14 0022
GRID 1 0 10. 0. 0. 0023
GRID 2 0 0. 0. 0. 0024
CBAR 1 229 2 1 0. 1. 0. 0025
304 Nastran Primer

Static Analysis
1 2 3 4 5 6 7 8 9
PBAR 229 30 6.5104-4 2.604-3 0026
+028 -.125 -.25 .125 -.25 .125 .25 -.125 .25 0027
+029 0028
MAT1 30 3.+7 .3 0029
+025 1.+4 1.+4 .5+4 0030
SPC 23 2 2356 0. 0031
PLOAD1 100 1 FY FR 0. 5. 1. 5. 0032
PLOAD1 200 1 FZ FR 0. 8.6603 1. 8.6603 0033
CBARAO 1 FR 9 .1 .1 0034
LOAD 300 1. 1. 100 1. 200 0035
ENDDAT
0036
A
With reference to the statements numbers in field 10:
0001 SOL requests solution 101, Static analysis

0002 TIME specifies two minutes of cpu time for
0003 CEND terminates Executive Control
0004 - 0005 TITLE and SUBTITLE are optional titling directives
0006 SPC Specifies that SPC set 23 is to be used for all subcases
0007 thru
Output requests for all subcases
0011
0012 thru
Load subcases
0017
0018 BULK DATA terminates Case Control
0019 GRDPNT parameter requests calculation of weight and balance
data with reference to the origin of the Basic coordinate system
0020 AUTOSPC set equal to YES will result in the automatic
removal of singular degrees of freedom. This is the default
value for the parameter so its presence is optional
0021 USETSEL will display the SB, SG, S, G, F and A sets
0022 GRDSET specifies default value for PSPC = 14 if the PSPC
field of a Grid is left blank. This value will remove u1 and u4 at
the point.
0023 -0024 GRID points 1 and 2
0025 CBAR connects Grids 1 and 2 with a BAR bending element and
aligns elemental y axis with the Basic Y axis. Points to PBAR
with PID = 229
0026 PBAR specifies Iyy = 2.604 x 10-4 in4, Izz = 6.5104 x 10-4 in4,
relative to element principal axes and points to material set 30
0027 PBAR continuation specifies the four stress recovery points, as
shown by Fig. 12-2 in element coordinates
0028 Specifies transverse shear flexibility coefficients
Nastran Primer 305

Static Analysis
11
0029 MAT1 specifies an isotropic material with E = 30 x 106 psi and

ν= 0.3
0030 Stress limits in tension, compression and shear
0031 SPC Specifies constraints to physical degrees of freedom at
Grid point 1 (ux2 and θx2 are not connected since A = 0 and J =
0 on PBAR data statement). All unconnected degrees of
freedom will be purged using AUTOSPC.
0032 PLOAD1 specifies uniform pressure load of 5 lb./in. in the
Basic Y-direction along length of the BAR
0033 PLOAD1 specifies uniform pressure load of 8.6603 lb./in. in the
Basic Z-direction along length of the BAR
0034 CBARAO requests calculation of element forces and stresses at
9 intermediate points along the length of the Bar
0035 LOAD combines the loads for subcases 1 and 2 for subcase 3
0036 ENDDATA terminates BULK DATA
11.12.4 Running NASTRAN

Running NASTRAN depends on the installation and the it’s dialect. NASA NASTRAN
processes the external files differently than MSC NASTRAN. This means that a bash shell that
defines file allocation is required for NASA NASTRAN. These details are taken care of using
ASSIGN statement in MSC NASTRAN.
• If using NASA NASTRAN make sure the script file in the BIN directory is PATHed
and the enter
script.sh barPA
• Create a script file in the PATHed directory. The script file, which might be named
runit.bat contains the line of text:
c:\msc\bin\nastran %1
The program could then be executed by entering:
runit barPA.dat
Where it is assumed that the input file is named barPA.dat. In either case the results will be
written in the directory in which the data file exists.
11.12.5 The Output File

The contents of the output file include echoes of the Executive Control and Case Control
sections and the sorted bulk data which is not described here. The calculated output
components are described below.
11.12.5.1 Grid Point Weight Table

The weight and balance information requested by the inclusion of the GRDPNT
parameter. The units of the coefficients of the mass and mass moment of inertia matrices are
in weight units because the weight per unit volume was specified on MAT1. The weight of the
structure is the same in the three coordinate directions and is equal to 0.35375 lb.
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
306 Nastran Primer

Static Analysis
* 3.512500E-01 0.000000E+00 0.000000E+00 0.000000E+00 0.000000E+00 0.000000E+00 *

* 0.000000E+00 3.512500E-01 0.000000E+00 0.000000E+00 0.000000E+00 1.756250E+00 *
* 0.000000E+00 0.000000E+00 3.512500E-01 0.000000E+00 -1.756250E+00 0.000000E+00 *
* 0.000000E+00 0.000000E+00 0.000000E+00 0.000000E+00 0.000000E+00 0.000000E+00 *
* 0.000000E+00 0.000000E+00 -1.756250E+00 0.000000E+00 1.756250E+01 0.000000E+00 *
* 0.000000E+00 1.756250E+00 0.000000E+00 0.000000E+00 0.000000E+00 1.756250E+01 *
S
* 1.000000E+00 0.000000E+00 0.000000E+00 *
* 0.000000E+00 1.000000E+00 0.000000E+00 *
* 0.000000E+00 0.000000E+00 1.000000E+00 *
DIRECTION
MASS AXIS SYSTEM (S) MASS X-C.G. Y-C.G. Z-C.G.
X 3.512500E-01 0.000000E+00 0.000000E+00 0.000000E+00
Y 3.512500E-01 5.000000E+00 0.000000E+00 0.000000E+00
Z 3.512500E-01 5.000000E+00 0.000000E+00 0.000000E+00
I(S)
* 0.000000E+00 0.000000E+00 0.000000E+00 *
* 0.000000E+00 8.781250E+00 0.000000E+00 *
* 0.000000E+00 0.000000E+00 8.781250E+00 *
I(Q)
* 0.000000E+00 *
* 8.781250E+00 *
* 8.781250E+00 *
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 *
Figure 11-1 Grid Point Weight Table
11.12.5.2 Displacement Set Membership Tables

The displacement set membership for each degree of freedom can be printed using the
USETPRT and USETSEL parameters. USETPRT controls whether the USET table is to be
printed. Setting USETPRT to 12 requests two type of tables using external grid point
identification numbers. The first type of table contains a row for each degree of freedom in the
model and identifies the set membership for each degree of freedom. The headings for the
columns in this table are described by Table 11-1. If the USETPRT parameter requests row
tables this means that the degrees of freedom for specific sets are to be printed. Setting the
USETSEL value equal to 1746 requests the S, SB, SG, G, A, and F sets.
Table 11-1 Column Heading of Uset Table
Column Heading Description

EXT. GP. DOF The external degree of freedom code. The first number is the grid
or scalar point number, and the second is the degree of freedom
code at the grid or scalar point.
INT DOF The internal number for the degree of freedom
INT GP The internal point number and type:
G - Grid point
S - Scalar point
E - Extra point
EXT. GP. DOF The external degree of freedom code. The first number is the grid
or scalar point number, and the second is the degree of freedom
code at the grid or scalar point.
Nastran Primer 307

Static Analysis
11
Table 11-1 Column Heading of Uset Table
Column Heading Description

SB Single point constraints specified on SPC data statements
SG Single point constraints specified on GRID Bulk Data data
statements
L Degrees of freedom in the l-set
A Degrees of freedom in the a-set
F Degrees of freedom in the f-set
N Degrees of freedom in the n-set
G Degrees of freedom in the g-set
R Degrees of freedom in the r-set
O Degrees of freedom in the o-set
S Degrees of freedom specified in SG and SB sets
M Degrees of freedom in the m-set
E Degrees of freedom in the e-set
The information contained in these tables can be extremely valuable, especially when the
user is attempting to find the cause of a singular stiffness or mass matrix.
11.12.5.3 Selected Output

The displacements, forces of constraint, grid point loads, forces in the BAR element, and
the stress in the BAR element are shown below. The GRID point-oriented output such as
displacements, grid point loads and SPC forces are sorted by GRID point number. The
components are then printed under the headings T1, T2, T3 for displacement or force
components and Rl, R2, R3 for rotation or moment components. The components are
interpreted with respect to the displacement coordinate system which is associated with grid
point displacements.
Displacements - Subcase ‘
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 3.201971E-01 0.0 0.0 0.0 4.266677E-02
Displacements - Subcase 2
1 G 0.0 0.0 0.0 0.0 0.0 0.0
2 G 0.0 0.0 1.389341E-01 0.0 -1.847649E-02 0.0
Displacements - Subcase 3
1 G 0.0 0.0 0.0 0.0 0.0 0.0
2 G 0.0 3.201971E-01 1.389341E-01 0.0 -1.847649E-02 4.266677E-02
Load Vector - Subcase 1

1 G 0.0 2.500000E+01 0.0 0.0 0.0 4.166667E+01
2 G 0.0 2.500000E+01 0.0 0.0 0.0 -4.166667E+01

308 Nastran Primer

Static Analysis
1 G 0.0 0.0 4.330150E+01 0.0 -7.216917E+01 0.0

2 G 0.0 0.0 4.330150E+01 0.0 7.216917E+01 0.0

1 G 0.0 2.500000E+01 4.330150E+01 0.0 -7.216917E+01 4.166667E+01
2 G 0.0 2.500000E+01 4.330150E+01 0.0 7.216917E+01 -4.166667E+01
Forces of Single- Point Constraint - Subcase 1

1 G 0.0 -5.000000E+01 0.0 0.0 0.0 -2.500000E+02

1 G 0.0 0.0 -8.660300E+01 0.0 4.330150E+02 0.0

1 G 0.0 -5.000000E+01 -8.660300E+01 0.0 4.330150E+02 -2.500000E+02
Force Distribution in BAR Elements - Subcase 1

0 ELEMENT STATION BEND-MOMENT SHEAR FORCE AXIAL
ID. (PCT) PLANE 1 PLANE 2 PLANE 1 PLANE 2 FORCE TORQUE
1 0.000 2.500000E+02 0.0 5.000000E+01 0.0 0.0 0.0
1 0.100 2.025000E+02 0.0 4.500000E+01 0.0 0.0 0.0
1 0.200 1.600000E+02 0.0 4.000000E+01 0.0 0.0 0.0
1 0.300 1.225000E+02 0.0 3.500000E+01 0.0 0.0 0.0
1 0.400 8.999999E+01 0.0 3.000000E+01 0.0 0.0 0.0
1 0.500 6.250000E+01 0.0 2.500000E+01 0.0 0.0 0.0
1 0.600 3.999998E+01 0.0 2.000000E+01 0.0 0.0 0.0
1 0.700 2.250000E+01 0.0 1.500000E+01 0.0 0.0 0.0
1 0.800 9.999990E+00 0.0 1.000000E+01 0.0 0.0 0.0
1 0.900 2.499978E+00 0.0 5.000000E+00 0.0 0.0 0.0
1 1.000 -3.814697E-06 0.0 0.0 0.0 0.0 0.0

1 0.000 0.0 4.330150E+02 0.0 8.660300E+01 0.0 0.0
1 0.100 0.0 3.507422E+02 0.0 7.794270E+01 0.0 0.0
1 0.200 0.0 2.771296E+02 0.0 6.928240E+01 0.0 0.0
1 0.300 0.0 2.121774E+02 0.0 6.062210E+01 0.0 0.0
1 0.400 0.0 1.558854E+02 0.0 5.196180E+01 0.0 0.0
1 0.500 0.0 1.082538E+02 0.0 4.330150E+01 0.0 0.0
1 0.600 0.0 6.928239E+01 0.0 3.464120E+01 0.0 0.0
1 0.700 0.0 3.897137E+01 0.0 2.598090E+01 0.0 0.0
1 0.800 0.0 1.732059E+01 0.0 1.732060E+01 0.0 0.0
1 0.900 0.0 4.330115E+00 0.0 8.660301E+00 0.0 0.0
1 1.000 0.0 7.629395E-06 0.0 0.0 0.0 0.0

1 0.000 2.500000E+02 4.330150E+02 5.000000E+01 8.660300E+01 0.0 0.0
1 0.100 2.025000E+02 3.507422E+02 4.500000E+01 7.794270E+01 0.0 0.0
1 0.200 1.600000E+02 2.771296E+02 4.000000E+01 6.928240E+01 0.0 0.0
1 0.300 1.225000E+02 2.121774E+02 3.500000E+01 6.062210E+01 0.0 0.0
1 0.400 8.999999E+01 1.558854E+02 3.000000E+01 5.196180E+01 0.0 0.0
1 0.500 6.250000E+01 1.082538E+02 2.500000E+01 4.330150E+01 0.0 0.0
1 0.600 3.999998E+01 6.928239E+01 2.000000E+01 3.464120E+01 0.0 0.0
1 0.700 2.250000E+01 3.897137E+01 1.500000E+01 2.598090E+01 0.0 0.0
1 0.800 9.999990E+00 1.732059E+01 1.000000E+01 1.732060E+01 0.0 0.0
1 0.900 2.499978E+00 4.330115E+00 5.000000E+00 8.660301E+00 0.0 0.0
1 1.000 -3.814697E-06 7.629395E-06 0.0 0.0 0.0 0.0
Nastran Primer 309

Static Analysis
11
Stress Distribution in BAR Element - Subcase 1

S T R E S S D I S T R I B U T I O N I N B A R E L E M E N T S ( C B A R )
0 ELEMENT STATION SXC SXD SXE SXF AXIAL S-MAX S-MIN
ID. (PCT)
1 0.000 4.800012E+04 4.800012E+04 -4.800012E+04 -4.800012E+04 0.0 4.800012E+04 -4.800012E+04
1 0.100 3.888010E+04 3.888010E+04 -3.888010E+04 -3.888010E+04 0.0 3.888010E+04 -3.888010E+04
1 0.200 3.072008E+04 3.072008E+04 -3.072008E+04 -3.072008E+04 0.0 3.072008E+04 -3.072008E+04
1 0.300 2.352006E+04 2.352006E+04 -2.352006E+04 -2.352006E+04 0.0 2.352006E+04 -2.352006E+04
1 0.400 1.728004E+04 1.728004E+04 -1.728004E+04 -1.728004E+04 0.0 1.728004E+04 -1.728004E+04
1 0.500 1.200003E+04 1.200003E+04 -1.200003E+04 -1.200003E+04 0.0 1.200003E+04 -1.200003E+04
1 0.600 7.680017E+03 7.680017E+03 -7.680017E+03 -7.680017E+03 0.0 7.680017E+03 -7.680017E+03
1 0.700 4.320011E+03 4.320011E+03 -4.320011E+03 -4.320011E+03 0.0 4.320011E+03 -4.320011E+03
1 0.800 1.920003E+03 1.920003E+03 -1.920003E+03 -1.920003E+03 0.0 1.920003E+03 -1.920003E+03
1 0.900 4.799969E+02 4.799969E+02 -4.799969E+02 -4.799969E+02 0.0 4.799969E+02 -4.799969E+02
1 1.000 -7.324237E-04 -7.324237E-04 7.324237E-04 7.324237E-04 0.0 7.324237E-04 -7.324237E-04

ID. (PCT)
1 0.000 4.157210E+04 -4.157210E+04 -4.157210E+04 4.157210E+04 0.0 4.157210E+04 -4.157210E+04
1 0.100 3.367340E+04 -3.367340E+04 -3.367340E+04 3.367340E+04 0.0 3.367340E+04 -3.367340E+04
1 0.200 2.660614E+04 -2.660614E+04 -2.660614E+04 2.660614E+04 0.0 2.660614E+04 -2.660614E+04
1 0.300 2.037033E+04 -2.037033E+04 -2.037033E+04 2.037033E+04 0.0 2.037033E+04 -2.037033E+04
1 0.400 1.496596E+04 -1.496596E+04 -1.496596E+04 1.496596E+04 0.0 1.496596E+04 -1.496596E+04
1 0.500 1.039303E+04 -1.039303E+04 -1.039303E+04 1.039303E+04 0.0 1.039303E+04 -1.039303E+04
1 0.600 6.651535E+03 -6.651535E+03 -6.651535E+03 6.651535E+03 0.0 6.651535E+03 -6.651535E+03
1 0.700 3.741491E+03 -3.741491E+03 -3.741491E+03 3.741491E+03 0.0 3.741491E+03 -3.741491E+03
1 0.800 1.662883E+03 -1.662883E+03 -1.662883E+03 1.662883E+03 0.0 1.662883E+03 -1.662883E+03
1 0.900 4.157177E+02 -4.157177E+02 -4.157177E+02 4.157177E+02 0.0 4.157177E+02 -4.157177E+02
1 1.000 7.324687E-04 -7.324687E-04 -7.324687E-04 7.324687E-04 0.0 7.324687E-04 -7.324687E-04

ID. (PCT)
1 0.000 8.957222E+04 6.428020E+03 -8.957222E+04 -6.428020E+03 0.0 8.957222E+04 -8.957222E+04
1 0.100 7.255350E+04 5.206699E+03 -7.255350E+04 -5.206699E+03 0.0 7.255350E+04 -7.255350E+04
1 0.200 5.732622E+04 4.113934E+03 -5.732622E+04 -4.113934E+03 0.0 5.732622E+04 -5.732622E+04
1 0.300 4.389039E+04 3.149729E+03 -4.389039E+04 -3.149729E+03 0.0 4.389039E+04 -4.389039E+04
1 0.400 3.224600E+04 2.314086E+03 -3.224600E+04 -2.314086E+03 0.0 3.224600E+04 -3.224600E+04
1 0.500 2.239305E+04 1.607005E+03 -2.239305E+04 -1.607005E+03 0.0 2.239305E+04 -2.239305E+04
1 0.600 1.433155E+04 1.028482E+03 -1.433155E+04 -1.028482E+03 0.0 1.433155E+04 -1.433155E+04
1 0.700 8.061503E+03 5.785198E+02 -8.061503E+03 -5.785198E+02 0.0 8.061503E+03 -8.061503E+03
1 0.800 3.582886E+03 2.571199E+02 -3.582886E+03 -2.571199E+02 0.0 3.582886E+03 -3.582886E+03
1 0.900 8.957146E+02 6.427924E+01 -8.957146E+02 -6.427924E+01 0.0 8.957146E+02 -8.957146E+02
1 1.000 4.499452E-08 -1.464892E-03 -4.499452E-08 1.464892E-03 0.0 1.464892E-03 -1.464892E-03
11.12.6 Discussion of Results

The element forces are defined using the displacement sign convention rather than the
equilibrium convention because PLOAD1 and CBARAO are present in the input file. The
headings under Force Distribution in BAR Elements are:
Element ID The identification number of the element

Station
The normalized length along the Bar
PCT
Plane 1 is Mz, the moment about the element z-axis
Moment
Plane 2 is My, the moment about the element y-axis
Shear Plane 1 is Vy, the transverse shear in the element y direction

Force Plane 2 is Vz, the transverse shear in the element y direction
Force Fx, The force in the element x -direction
Axial
Mx, the moment about the element x-axis
Torque
310 Nastran Primer

Static Analysis
The headings under Stress Distribution are:
Element ID The identification number of the element

SXC Bending stress at the four stress recovery points identified as points
SXD C, D, E, and F
SXE
SXF
AXIAL Axial stress, P/A
S-MAX The maximum combined stress at the four stress recovery points
S-MIN The minimum combined stress at the four stress recovery points
M.S The margin of safety for stress based on stress limit defined by
MAT1 Bulk DATA for the element
The displacement at the end of the Bar element from beam theory with no correction for
transverse shear deflection is given by:
4
p0 L
δ = ----------- (Eq. 11-1)
8EI
where p0 is the magnitude of the uniform line load, L is the length of the Bar, E is the elastic
modulus and I is the area moment of inertia of the cross section. Substituting the values used
for this problem we find that δy = 0.032 for p0Y = 5.0 lb/in and δz = 0.01358 in. for p0Z = 8.6603
lb./in.
You should recall that the stiffness for the Bar element was obtained by using the shape
functions for a beam loaded with end loads. The PLOAD1 generates work equivalent loads
which lead to the exact solution for a uniformly distributed load using only one element.
Now looking at the transverse shear distribution we see that it varies linearly from a value
of zero at the free end to p0L at the fixed end which agrees exactly with equilibrium
considerations. Similarly the moment at any cross section is given by:
1 2
M ( x ) = --- p 0 ( L – x ) (Eq. 11-2)
2
Looking at the moment distribution, for loading in the Y-direction for example, it can be
verified that the moment distribution agrees exactly with that predicted by Eq. 12-23. It is left
for an exercise to verify that the results for other loadings supported by PLOAD1 result in exact
agreement with beam theory for a single element.
Nastran Primer 311

Static Analysis
11
Simply-Supported Rib-Stiffened Plate
11.13 Simply-Supported Rib-Stiffened Plate

The next example, which is a stiffened flat plate shown by Figure 11-1, will allow us to
explore the use of the NASTRAN shell elements for modeling.
1.0 in. 1.0 in.

0.5 in. 0.5 in.
z
0.4 in.
15 in. 15 in.
y
x 6
15 in. 15 in. E = 10 × 10 psi
ν = 0.3
3
γ = 0.1 lb/in
Figure 11-1 Eccentrically Stiffened Plate
With reference to Fig. 12-4, we will consider the following solutions

a) A simply supported flat plate subject to uniform pressure load without stiffeners
using appropriate symmetry conditions to model 1/4 of the plate
b)A simply supported plate with Bar stiffeners where the middle of the plate is used as
the reference surface and where the stiffener is modeled with an offset Bar element.
11.14 Problems
6
[1] Given a uniform cantilever beam having and elastic modulus, E = 30 ×10 psi .a cross
section moment of inertia, I = 1 ⁄ 12 in4,a length of 10 in. and a load at the free end of 300 lb.
Model the beam using DMIG and determine the displacement and slope at the free end
normalized using the exact solution.
[2] Repeat Problem 1 using the GENEL element
[3] Given the rectangular plate described in Sec. 8.3.5, for the case where a ⁄ b = 1 and a
meshing factor of 4, remove all displacements at the interior node points using static
condensation. Determine the displacement at the center normalized using the exact value.
[4] Model the simply supported rectangular plate having an aspect ratio, a ⁄ b = 5 , using
Hexa elements with 8 nodes for a mesh density of 4 by 4 elements. Load the top of the plate
with a unit uniform pressure using Pload4, Determine the normalized center displacement.
[5] Repeat problem 4 using Hexa 20 elements. Determine the normalized center
displacement and verify that the total load is correct.
[6] Repeat problem 4 but load the plate having a weight density of 1 lb per sq. in. using the
Grav load. Determine the normalized center displacement.
[7] Determine the converged center displacement for the stiffened plate described in Sect.
12.11.2
[8] Determine the normalized tip displacement of the swept plate described by Cook (Ref.
[16], Chapter 7) using a 5 by 5 mesh of Quad4 elements.
312 Nastran Primer

Static Analysis
References
11.15 References
[1] R.J. Roark, Formulas for Stress and Strain, Third Edition, McGraw-Hill, New York,
1954, p. 198.
[2] Anon, NASTRAN Users Manual, NASA SP-222, Vol. I, 1986. Sec. 6
Nastran Primer 313

Static Analysis
11
References
314 Nastran Primer

12 Normal Modes Analysis
The general problem of Structural Dynamic analysis is addressed by several authors

including Bathe [1], Zienkiewicz [2]and Hughes [3] and is beyond the scope of the present text.
We will restrict our attention in this chapter to the real eigenvalue problem. In succeeding
sections we will present the equations of motion for dynamic systems and define the
eigenvalue problem. We will then describe the important properties of normal modes. Finally,
the various eigenvalue extraction techniques which are available in NASTRAN will be
described.
12.1 Dynamic Motion

The motion of structural dynamic systems is described by a set of equations expresses the
balance between external applied loads, the internal forces and the inertial forces. For a
structural system these forces are: the inertial forces, { f i } , the damping or dissipative
forces, { f d } , the internal elastic forces, { f e } , and the external time-dependent forces, { P } .
Using Newton’s second law of motion then leads to
{ fi ( t ) } + { fd ( t ) } + { fe ( t ) } = { P ( t ) } (Eq. 12-1)
where the internal forces can be expressed in terms of displacements and derivatives of
displacements as follows:
··
{ fi ( t ) } = [ M ] { u }
·
{ fd ( t ) } = [ B ] { u } (Eq. 12-2)
{ fe ( t ) } = [ K ] { u }
using these definitions for internal forces we can then write (Eq. 12-1) as:
·· ·
[ M]{ u} + [ B]{u } + [ K]{ u} = { P( t)} (Eq. 12-3)
where the stiffness matrix, [ K ] , is the same as that for the static case. We also note that
(Eq. 12-3) reduces to the static equilibrium equation when the time derivatives of
··
displacements, { u· } and { u } , are zero. The additional system matrices associated with a
dynamic system are [ M ] , the system mass matrix and [ B ] , the system damping matrix.
The solution of (Eq. 12-3) is beyond the scope of this text. However it is worthwhile to at
least describe the solution methods which can be used in order to provide motivation for the
topic of normal modes and frequencies. In our discussion we will assume that the dynamic
model is based on the structural model used for a static analysis. This implies that the number
of degrees of freedom can be large so that cost, measured in terms of time and computer
resources, will be a factor in solving the set of dynamic equations.
315 Nastran Primer

Normal Modes Analysis
12
The Eigenvalue Problem
Again, just thinking the problem through for a moment, we recognize that if solution time
is a factor (and it is) then we should be able to reduce the solution time by reducing the number
of degrees of freedom. The question is: How can the system of equations be reduced from some
large number which was used for static analysis to a fairly small number which is appropriate
for dynamic analysis.
Let's be a little more specific. Without worrying at the present time about
how [ M ] and [ B ] are defined, let's assume that MPC and SPC and OMIT operations have been
applied so that the solution set is reduced to the a-set, which is the solution set for normal
modes analysis. The differential equation, (Eq. 12-3) is then written using standard
NASTRAN set notation as:
[ M aa ] { u·· a } + [ B aa ] { u· a } + [ K aa ] { u a } = { P a } (Eq. 12-4)
Now suppose we can find a transformation, [ G ah ] ,which relates { u a } to a greatly reduced

set of degrees of freedom { u h } . That is:
{ u a } = [ G ah ] { u h } (Eq. 12-5)
We can then use this transformation to reduce(Eq. 12-4)to the h-set resulting in:
[ M hh ] { u·· h } + [ B hh ] { u· h } + [ K hh ] { u h } = { P h } (Eq. 12-6)
where:
T
[ M hh ] = [ G ah ] [ M aa ] [ G ah ]
T
[ B hh ] = [ G ah ] [ B aa ] [ G ah ]
(Eq. 12-7)
T
[ K hh ] = [ G ah ] [ K aa ] [ G ah ]
T
{ P hh } = [ G ah ] { P a }
There isn't any question about being able to define the transformation(Eq. 12-5). The real
question is: what transformations would be useful and would retain the essence of the dynamic
DNA, as it were. A particularly attractive one would be one for which the reduced system
matrices [ M hh ] , [ B hh ] , and [ M hh ] , are diagonal. The resulting set of equations, (Eq. 12-6),
would be completely decoupled and their solution would be reduced to quadrature using the
convolution integral which is described by Meirovich [4] and others. We will see in the next
section that the eigenvectors for the real eigenvalue problem transform the mass and stiffness
matrices to diagonal forms.
The study of and solution for eigenvectors is thus of great importance in transforming
large sets of coupled dynamic problems to a small set of uncoupled dynamic problems because
of their orthonormal properties and the fact they form a basis in a-dimensional solution space.
The calculation of the eigenvalues and eigenvectors is also of importance in its own right, of
course, because they define the undamped free vibrational modes and frequencies of the
system.
12.2 The Eigenvalue Problem

The eigenvalue problem is associated with the solutions of the undamped (i.e., [ B ] = 0 )
unforced (i.e.,{ P } = 0 ) system of equations. We have then, from (Eq. 12-3):
316 Nastran Primer

··
[M]{u} + [K]{u} = 0 (Eq. 12-8)
Assuming separation of variables and harmonic motion gives the following:
iω t
n
{ u ( ( x, y, z ), t ) } = { φ ( x, y, z ) }e (Eq. 12-9)
where { φ ( x, y, z ) } is a vector of displacement amplitudes which defines the spacial

distribution of the displacement vector { u } . The substitution of { u } and { u·· } into (Eq. 12-8)
then gives the following equation which relates { φ } and ωn:
2 2
[ K ] { φ } –ωn [ M ] { φ } = ( { K } –ωn [ M ] ) { φ } = 0 (Eq. 12-10)
From (Eq. 12-10) we obtain the eigenvalue problem:
( [ K aa ] – λ [ M aa ] ) { φ a } = 0 (Eq. 12-11)
where the subscript a indicates the eigenvalue problem is associated with a-set degrees of
freedom and where:
2
λ = ωn (Eq. 12-12)
The eigenvalue problem, equation (Eq. 12-11) has non trivial solutions (i.e., a non null
values of { φ } ) only if the determinant of the coefficient matrix is zero; i.e.
det ( [ K ] – λ [ M ] ) = 0 (Eq. 12-13)
This determinant, which is called the characteristic equation, is a polynomial in terms of

powers of λ where the highest power is the same as the order of matrices [ M ] and [ K ] . The
values of λ which satisfy the characteristic equation are called the eigenvalues; and, since an
nth order polynomial has n roots there are exactly n eigenvalues.
Each one of the roots, λi (i = 1,2,...,n), then satisfies (Eq. 12-11) so that we have:
( [ K ] – λi [ M ] ) { φi } = 0 (Eq. 12-14)
where{ φ i }is the vector of displacement amplitudes, i.e., the eigenvector, associated with the
eigenvalue λi.
The procedure for determining eigenvalues and the associated eigenvector is conceptually
quite straightforward. The eigenvalues are obtained by solving for the roots of the
characteristic polynomial (Eq. 12-13). An eigenvalue is then substituted into (Eq. 12-14) and
the resulting set of algebraic equations is solved for the associated eigenvector.
As an example, consider a system with [ K ] and [ M ] given by:
[K] = 6 –2 [M] = 2 0 (Eq. 12-15)

–2 4 01
the characteristic equation (Eq. 12-13) then becomes:
det ( 6 – 2λ ) – 2 = 0 (Eq. 12-16)

–2 (4 – λ)
which has two distinct real roots:, λ1 = 2 and λ2 = 5.
Nastran Primer 317

12
The substitution of these two roots and the given [ K ] and [ M ] matrices into (Eq. 12-14)
then leads to the following eigenvectors:
 
for λ1 = 2 { φ1 } = α1  1 
 1 
(Eq. 12-17)
 
for λ2 = 5 { φ2 } = α2  2 
 –1 
where the coefficients, αi, indicate that we can only determine the relative values for the
amplitudes of the eigenvectors. We can generalize this last statement to say that if { φ i } is an
eigenvector, i.e., it satisfies (Eq. 12-13), then α i { φ i } is also an eigenvector.
Now that we know how to calculate eigenvalues and eigenvectors we can consider some
of their important properties and in the process see that the eigenvectors do have the special
property to which we alluded in the last section, i.e., they transform the mass and stiffness
matrices to diagonal form.
Let's suppose that we have two distinct eigenvalues and their associated eigenvectors
(λi,{ φ i }) and (λj, { φ j } ). Each of their pairs satisfy (Eq. 12-14) so that we have:
( [ K ] – λi [ M ] ) { φi } = 0
(Eq. 12-18)
( [ K ] – λj [ M ] ) { φj } = 0
T T
The premultiplication of the first of (Eq. 12-18) by { φ j } and the second by { φ i } followed
by the transposition of the second equation then leads to:
T
{ φj } ( [ K ] – λi [ M ] ) { φi } = 0
(Eq. 12-19)
T
{ φj } ( [ K ] – λj [ M ] ) = 0
where we have used the fact that [ K ] and [ M ] are symmetric matrices.
The subtraction of the second of (Eq. 12-19) from the first leads to:
T
( λj – λi ) { φj } [ M ] { φi } = 0 (Eq. 12-20)
Since λi and λj are distinct, it follows that:
T
{ φj } [ M ] { φj } = 0 i≠j (Eq. 12-21)
This relation can be generalized as:
T  M ii for i = j
{ φj } [ M ] { φi } =  (Eq. 12-22)
 0 for i≠j
Where Mii is the modal mass. This relation shows that the real eigenvectors are
orthonormal with respect to the mass matrix. It is a simple matter to show that the eigenvectors
are also orthonormal with respect to the stiffness matrix:
318 Nastran Primer

Standard Form of Eigenvalue Problem
T  K ii for i = j
{ aj } [ K ] { ai } =  (Eq. 12-23)
0 for i≠j
and that the modal mass, Mii and modal stiffness Kii are related to the eigenvalue as follows:
K ii
λ i = -------
- (Eq. 12-24)
M ii
Returning briefly to the transformation (Eq. 12-5), let us suppose that we have found h
eigenvectors each of which has a-rows. We can then define the transformation [ G ah ] to be a
rectangular array of the h eigenvectors, arranged in order of ascending eigenvalue as follows:
[ G ah ] = { a 1 } { a 2 } { a 3 } … { a h } (Eq. 12-25)
Since each of the eigenvectors is orthonormal with respect to the mass and stiffness
matrices, it follows that the use of the transformation (Eq. 12-5) will result in modal mass and
modal stiffness matrices which are diagonal. The use of real eigenvectors for the
transformation from physical set of degrees of freedom, { u a } , to modal h-set, { u h } ,
coordinates results in the modal formulation of dynamic algorithms in NASTRAN for transient
and frequency response using the modal formulation.
We noted earlier in this section that if { φ i } is an eigenvector which satisfies (Eq. 12-11)
then α i { φ i } , where αi is a scale factor for the ith mode is also a eigenvector since it satisfies
(Eq. 12-11) The value of αi depends on the method used to normalize the eigenvector. Two
popular methods incorporated in NASTRAN are MASS normalization where the eigenvectors
are scaled to make Mii = 1; and, MAX normalization where the scale factor is the largest
element of the eigenvector so that the largest element of the scaled eigenvector is equal to one.
12.3 Standard Form of Eigenvalue Problem

Several of the eigenvalue algorithms included in NASTRAN require that the eigenvalue
problem be transformed to the standard form of the eigenvalue problem which is:
([ J] – λ[I ] ){w } = 0 (Eq. 12-26)
where [ J ] is real and symmetric and [ I ] is the identity matrix. The transformation must be a
similarity transformation so that the symmetry of [ K ] and [ M ] is preserved and [K] and [J]
have the same eigenvalues with the same multiplicity, as shown by Strang [5]. The required
transformation is obtained by first finding the Choleski factors, [ C ] ,[6] of the mass matrix,
where:
T
[M] = [C][C] (Eq. 12-27)
The substitution of (Eq. 12-27) into (Eq. 12-11) then gives:
T
( – λ[ C][ C] + [K ] ){ a} = 0 (Eq. 12-28)
A set of transformed variables is then defined as:
T
{w} = [C] {a} (Eq. 12-29)
Nastran Primer 319

12
Specification of Element Inertia Properties
The substitution of (Eq. 12-29) into (Eq. 12-28) and subsequent premultiplication
–1
by [ C ] then leads to the form of equations (Eq. 12-26) where:
–1 –T
[ J] = [ C] [K ][ C] (Eq. 12-30)
The transformation, (Eq. 12-30), is called a similarity transformation. This type of

transformation preserves the character of the matrix [ K ] so that if it is real and
symmetric [ J ] will also have those characteristics.
It is important to note that the Choleski factors will be singular if [ M ] is singular. This
implies that an eigenvalue extraction algorithm which requires transformations to the standard
form using the algorithm described above will fail during the inversion of the Choleski factor.
The diagnostic message from NASTRAN will simply state that ‘MAA is singular’, where
‘MAA’ refers to the reduced mass matrix for the a-set. Special procedures are used in
NASTRAN to remove these singularities.
12.4 Specification of Element Inertia Properties

Element inertia properties can be generated using both structural mass and nonstructural
mass or by special the mass elements. The mass matrix will be calculated:
1. For all dynamic solution sequences, including normal modes and frequencies.
2. If the structural weight is requested by including the parameter GRDPNT in any of
the static solutions.
3. If external forces are to be generated using GRAV and RFORCE load
specifications.
12.4.1 Structural Mass

Structural mass is associated with the volume of the element and is defined by means of
the mass density field, RHO, on a MAT-type data statements. The value of RHO is used to
calculate the mass matrix for all elements for which a volume can be calculated.
12.4.2 Nonstructural Mass

Non structural mass allows the analyst to account for the inertia effects of components
that are non-structural in nature, such as hydraulic lines and wiring in an aircraft for example.
Non structural mass is associated with an element’s length or area rather than on its volume.
Non-structural mass is defined for one and two dimensional elements, by the NSM field
on the associated property data statement. The NSM will be interpreted as mass per unit length
for lineal elements and mass per unit volume for two dimensional elements. By its very nature,
NSM is meant to augment the structural mass. Thus, if both structural mass, RHO, and non
structural mass, NSM, are specified, the element mass matrix will be calculated using the sum
of structural and nonstructural mass values.
12.4.3 Consistent and Lumped Formulations

The masses associated with the structural finite elements can be calculated by two
methods using either a lumped or a consistent mass formulations. The lumped formulation is
generally the default unless the user requests the consistent formulation by means of the
COUPMASS parameter data statement which is described in a later section.
320 Nastran Primer

In the lumped mass method the total mass for the element is calculated using the RHO
and NSM values as appropriate for the element. The mass at each grid point is then calculated
by dividing the total element mass by the number of element vertices. This mass fraction is
then associated with each of the translational degrees of freedom for the element. The lumped
mass formulation does not generate inertia terms for rotational degrees of freedom. This means
that special care must be taken when using eigenvalue algorithms using the standard form of
the eigenvalue problem as described in the preceding sections since the associated mass matrix
in this case will be singular.
12.4.4 Mass Elements

Mass can be assigned to scalar and grid point degrees of freedom by means of the
CMASSi, CONM1 and CONM2 described in the following sections.
12.4.5 Scalar Mass - CMASS

A scalar mass, which is associated with a SPOINT or a single degree of freedom at a Grid
point is specified by one of the CMASS-type data statements shown by Bulk Data Image 12-1.
1 2 3 4 5 6 7 8 9 10
CMASS1 EID PID G1 C1 G2 C2
PMASS PID MASS
CMASS2 EID MASS G1 C1 G2 C2
CMASS3 EID PID S1 S2

PMASS PID MASS
CMASS4 EID MASS S1 S2
Where:
Field Description
EID Element identification number, integer >0. Must be unique among all
elements in the Bulk Data section.
PID Identification number of a PMASS data statement, integer >0.
G1, C1 The grid point number and degree of freedom code defining a single
degree of freedom to which mass is to assigned, integers; C1 is blank or
zero if G1 refers to a SCALAR point.
G2, C2 The grid point number and degree of freedom code defining a single
degree of freedom to which mass is to assigned, integers; C1 is blank or
zero if G1 refers to a SCALAR point.
S1 The identification number of a SCALAR point to which mass is to
assigned, integer > 0.
S2 The identification number of a SCALAR point to which mass is to
assigned, integer > 0.
MASS Mass coefficient to be assigned to the degree of freedom, real > 0.
Bulk Data Image 12-1 Scalar Mass Elements
Nastran Primer 321

12
The form of the scalar mass element allows the element to be connected to two degrees
of freedom which might be more appropriate for modeling passive electrical circuits than
structural systems. The MASS value is placed on the diagonal of the mass matrix associated
with the first of the two connected degrees of freedom and the off diagonal term associated
with the second. This allows the CMASS to be used to specify an inductance in an electrical
circuit element which is analogous to mass. When used to define the concentrated mass
associated with a single degree of freedom the second connected point must be left blank.
12.4.6 Concentrated Grid Point Mass - CONM1

The CONM1 data statement specifies the mass matrix coefficients relating all six degrees
of freedom at a grid point using the following CONM1 shown by Bulk Data Image 12-2.
1 2 3 4 5 6 7 8 9 10
CONM1 EID G CID M11 M21 M22 M31 M32 0001
+001 M33 M41 M42 M43 M44 M15 M25 M35 0002
+002 M45 M55 M16 M26 M36 M46 M56 M66
Where:
Field Description
EID Identification for the element, integer > 0. The element ID must be unique
among all elements in Bulk Data.
G The identification number of the Grid point for which the mass
coefficients are specified.
CID Identification number of the reference coordinate system in which the
mass coefficients are defined, integer =>0.
Mij The coefficients of the diagonal terms of the system mass matrix
associated with Grid point G, real.
Bulk Data Image 12-2 Grid Point Mass Matrix - CONM1
12.4.7 Offset Concentrated Grid Point Mass - CONM2

An offset concentrated mass at a grid point is specifies using the CONM2 as shown by
Bulk Data Image 12-3:
1 2 3 4 5 6 7 8 9 10
CONM2 EID G CID M X1 X2 X3 0001
+001 I11 I21 i22 i31 I32 I33
Where:
Field Description
EID Identification for the element, integer > 0. The element ID must be unique
among all elements in Bulk Data.
G The identification number of the Grid point for which the mass
coefficients are generated, integer > 0.
CID Identification number of the reference coordinate system in which the
offset, Xi, and moment of inertia tensor, Iij, are defined, integer =>0.
322 Nastran Primer

Real Eigenvalue Extraction Techniques
Field Description
M The offset concentrated mass, real.
Xi Components of the vector offset of the mass from the grid point
defined with reference to the displacement coordinate system at the grid
point if CID => 0. If CID = - 1 the offsets are taken to be the
difference between the grid point location and the position defined
by the offset coordinates.
Iij The coefficients of the mass moment of inertia tensor defined with
reference to the CID coordinate system.
Bulk Data Image 12-3 Concentrated Mass - CONM2
The form of the mass matrix generated at grid point G for the CONM2 specification is as
follows:
∫ ρ ( x2 + x3 ) dV
2 2
I 11 =
∫ ρ ( x3 + x1 ) dV
2 2
I 22 =
∫ ρ ( x1 + x2 ) dV
2 2
I 33 =
(Eq. 12-31)
I 12 = ∫ ρx1 x2 dV
I 23 = ∫ ρx2 x3 dV
I 31 = ∫ ρx3 x1 dV
M 0 0 0 0 0
0 M 0 0 0 0
0 0 M 0 0 0
[M] = (Eq. 12-32)
0 0 0 I 11 – I 12 – I 13
0 0 0 – I 21 I 22 – I 23
0 0 0 – I 31 – I 32 I 33
where the Grid point mass, M, is:
M = ∫ ρ dV (Eq. 12-33)
12.5 Real Eigenvalue Extraction Techniques

The eigenvalue problem is defined for the a-set degrees of freedom. The eigenvalue
problem in NASTRAN set notation is:
( – λ [ M aa ] + K aa ) { φ a } = 0 (Eq. 12-34)
Nastran Primer 323

12
Real Eigenvalue Extraction Techniques
In the “early days” of NASTRAN eigenvalue extraction was an expensive operation.

Since those days great strides in Linear Algebra have led to efficient methods that have made
most of the methods in the following table obsolete. Modern methods are based on the Lanczos
method. All algorithms are maintained in the program to support “dusty decks”, I suspect. That
said the correct algorithm is the “LANC” method for Legacy NASTRAN and the use of
EIGRL in NASTRAN derivatives for the Lanczos method,
Table 1-1 Eigen Solution Methods in NASTRAN
Method Name Description

INV Inverse power with shifts
GIV Givens method of triadiagonalization
LANC Lanczos method
FEER** Fast Eigenvalue Extraction
SINV* Inverse power with enhancements
MGIV* Modified Givens
HOU* Housholder’s method of triadiagonalization
MHOU* Modified Householder
AGIV* Automatic selection of Givens or modified Givens methods
AHOU* Automatic selection of Givens or modified Givens methods
Note: Those options identified by a single asterisk are only available in MSC and NX
NASTRAN. The FEER method, identified by double astericks, is only available in Legacy
NASTRAN
An algorithm together with its associated parameters is specified by an appropriate EIGR
or EIGRL data specification. EIGRL(p. 331) is used to specify the Lanczos method and the
EIGR is used to specify all other real eigenvalue extraction methods available in NASTRAN.
Inverse power is a technique that finds one eigenvalue and an associated eigenvector at a
time. The Givens and Householder algorithms use transformation techniques to find all the
eigenvalues by transforming the eigenvalue problem to a diagonal form. The associated
eigenvectors can then be selectively evaluated.
The Lanczos method is a subspace algorithm. The idea is to represent a subspace of the
eigenvectors using a linearly independent set of approximate transformation vectors which are
determined using a truncated inverse power solution and which span the subspace. The
matrices are then transformed to a subspace in which the eigenvalues and eigenvectors are
determined using Givens, House-holder or the Jacoby method. As a last step the subspace
eigenvectors are transformed to the solution set.
The relative efficiency of any of the methods is dependent on the matrix topology, the
order of the system matrices, and the number of eigenvalues and eigenvectors to be extracted.
Numerical studies have shown that the Lanczos technique is the most efficient of the
eigenvalue solvers for determining a large number of eigenvectors for large problem size.
The Givens method requires that the eigenvalue problem be cast in standard form as
defined by (Eq. 12-26). Since [ J ] in this form must be real and symmetric and since [ J ] is
obtained by the similarity transformation, (Eq. 12-30), this implies that the mass matrix must
be positive definite. Thus, the Givens method cannot be used for the buckling problem.
The following general observations can be made about the choice of eigenvalue
extraction technique
1.Lanczos is the recommended procedure for one-to-many eigenvalues
324 Nastran Primer

The Inverse Power Method
2.The inverse power method is obsolete. But, if used, is best suited for obtaining only
a few eigenvectors for large systems with sparse matrices.
3.The tridiagonal methods can for those cases where several eigenvectors are
required but the Lanczos method is to be recommended in their place,.
12.6 The Inverse Power Method

The following observations can be made about the inverse iteration with shifts procedure,
where one or more shift points λ0 are introduce in the frequency range that is specified on the
frequency range of interest.
1.The stiffness matrix can be singular, provided λ0 is nonzero, so rigid body modes
offer no special difficulty. Tests of this algorithm suggest that repeated rigid body
modes are evaluated with no difficulty.
2.The shift point, λ0, may be changed at any stage in the solution to improve
convergence or to improve accuracy and convergence rate after several roots have
been extracted. A change in starting point requires that an additional matrix
decomposition be performed.
3.The shift point can be specified so that eigenvalues can be obtained within a
specified frequency band rather than obtaining the smallest eigenvalue.
12.6.1 Specification of EIGR for Inverse Power with Shifts

The format for the EIGR data statement for the inverse power with shifts is:
1 2 3 4 5 6 7 8 9 10
EIGR Method SID F1 F2 Ne Nd 0001
+001 NORM G C
Where
Field Description
Method The value of this field is the character string, INV, for inverse power or
SINV for enhanced INV
SID The set identification number, integer >0
fmin, fmax Minimum and maximum frequencies defining the frequency range of
interest, real > 0. Cycles per unit time. ( 0. ≤ f min < f max )
Ne The number of estimated roots in the frequency range of interest, integer >
0.
Nd The number of eigenvalues desired. Integer >0 or blank. If blank the Nd =
3Ne.
NORM Method of normalizing the eigenvector. One of the character strings,
MAX, MASS or POINT
G, C The grid point number and degree of freedom code defining the degree of
freedom to be used for POINT normalization, integers.
Bulk Data Image 12-4 Inverse Power Method - INV and SINV
The data specified on EIGR are used to control the inverse iteration method in the
following manner:
Nastran Primer 325

12
1.The maximum and minimum frequencies of interest are converted to maximum and
minimum values of the eigenvalue λmin and λmax where
2
λ = ( 2πf )
2.The number of roots estimated in the range, Ne, is used to determine the number of
starting points, Ns, which is chosen to satisfy the inequality
Ne
N s – 1 < ------ ≤ N e
6
The starting points are then distributed, as shown in Figure 12-1 where the increment in the
eigenvalue is found from:
∆λ = λ max – λ min
and the position of the nth starting point is given by:
(n – 1 ⁄ 2)
λ ns = λ min + ----------------------- ∆λ
Ns
The algorithm assumes that approximately six good eigenvalues and associated
eigenvectors can be found from a single starting point. Thus, if the estimate number of roots is
six or less, the algorithm will use only one starting point, which is located in the middle of the
eigenvalue range of interest.
The eigenvalue analysis is terminated after Nd roots are found. If the user does not specify
Nd, the algorithm will terminate after finding three times the number of roots estimated.
∆λ- ∆λ ∆λ ∆λ-
-------- ------- ------- --------
2N s Ns Ns 2N s
λ s1 λ s2 λ s3
λ min λ max
Figure 12-1 Distribution of Starting Points for Inverse Power
12.6.2 Summary for the Inverse Iteration Method

A summary of the eigenvalues found by the inverse power method is printed
automatically. To interpret this summary, the user must have some understanding of the
inverse iteration algorithm.
First of all, there is a difference between the starting point and the shift point. The
algorithm starts the eigenvalue search at a starting point and then may shift to another point
(indeed the same concept as a starting point) in order to speed convergence. If, for example,
the ratio of the first two shifted eigenvalues relative to the current starting point is
approximately equal to unity, the rate of convergence will be very slow. Since the value of the
closest eigenvalue and the time required to converge from both the current starting point and
the shifted point can be estimated, a decision to either shift or continue to iterate can be made.
326 Nastran Primer

The algorithms will continue to search for eigenvalues relative to the current starting point
until an eigenvalue is found in the range of the next starting point, where a certain overlap
between range of starting points is allowed. The algorithm finds successive eigenvalues, the
first of which is the closest to the starting point. The eigenvalues then increase in absolute
magnitude from the starting point. This means that one cannot be certain that the lowest root
has been found in a specified range unless all the roots in the range have been found. The
summary of the parameters associated with the algorithm and reasons for termination are as
follows.
1.Number of eigenvalues extracted
2.Number of starting points used
3.Number of starting points moves
4.Number of matrix decompositions
5.Number of iterations
6.Reason for termination
1.Two consecutive singularities encountered while performing matrix

decompositions.
If two successive singularities in ( [ K ] – λ 0 [ M ] ) are found, this
implies that there may be something abnormal about the problem.
Check the input to insure that the mass matrix has been specified.
2.Four shift points used while tracking a single root.
If four shifts are required to track a single root, this implies that the
first root is a long way from the first starting point and that the user
did not exercise proper care in specifying the frequency range of
interest. A preliminary estimate of the lowest root should be made
with changes to the frequency range, and perhaps, the number of
roots estimated in the range.
3.All eigenvalues found in the frequency range specified.
All the roots in the range have been found. An excellent reason for
terminating.
4.Three times the number of estimated eigenvalues have been found.
If the user has not specified the number of roots desired, the
algorithm will terminate after finding 3Ne roots. Generally
speaking, this is not a good reason for terminating.
5.All the eigenvalues that exist in the problem have been found.
A valid reason for termination, but one that the user should not
receive except for those cases where there are a very few degrees of
freedom. If the user desires all the roots, Givens method should be
specified since it is more economical than the inverse method if
several roots and associated vectors are desired.
6.The number of roots desired have been found.
same comment as reason 4
7.One or more eigenvalues have been found outside the specified
frequency range.
Nastran Primer 327

12
If one or more eigenvalues have been found outside the range, then
either all eigenvalues in the range have been found or there are no
roots in the range. A good reason for terminating, especially if only
one root is desired.
8.Insufficient time to find another root.
A bad reason for terminating. Re-execute the problem taking
advantage of current results, if possible, by restricting the range of
interest.
9.Unable to converge.
A bad reason for termination because the algorithm can be shown
to converge if [Maa] and [Kaa] are symmetric. Since the algorithm
uses system matrices generated by the finite element model, there
should be no convergence problem unless the user has modified the
[Maa] and/or [Kaa] matrices by some nonstandard technique.
7. Largest off-diagonal modal mass term and the numbers of the eigenvector pairs that fail
the criteria.
12.6.3 Finding Lowest Eigenvalue with Inverse Method

The inverse power method is frequently used to determine the lowest eigenvalue, i.e, the
lowest natural frequency for a normal mode solution. If one has no knowledge of the value for
the eigenvalue then one is on the horns of a dilemma because
1.The specification of a large frequency range for normal modes may lead to the
determination of "several eigenvalue-eigenvector pairs at great cost.
2.Even if all roots are found in the range there is always a possibility of having a root
below fmin if fmin > 0.
There isn't any magic prescription in Inverse Power that in essence says "find the lowest
root". All we can do is give some rather general guidelines which may help reduce the solution
costs and which will guarantee that the lowest root is found.
The suggested technique for finding the lowest eigenvalue is to specify a small frequency
range using fmin and fmax which is so small that the root will be outside the range. Then specify
that Ne = 1, which will generate one starting point in the middle of the range, and that Nd= 1
which will cause the algorithm to terminate after the extraction of one eigenvalue and
associated eigenvector. If we were successful in choosing the frequency range correctly then
the root is outside the range and is closest to the starting point, which is at the center of the
specified frequency range.
In order to guarantee that there are no roots below that calculated you can request that
parameters which are calculated within the eigenvalue extraction module be printed by using
DIAG 16 in Executive Control. While this data includes much which will either be
incomprehensible or useless it does include the number of roots below the current starting
point λ0. If this number is zero then we can be assured that the one root which was calculated
is indeed the lowest. A description of the data which is printed by DIAG 16 in Solution 1, ot
Solution 103 is included as part of the example problem in a later section.
328 Nastran Primer

The Lanczos Method
12.7 The Lanczos Method
12.7.1 Description of the Lanczos Procedure

The Lanczos procedure is described in detail by Nour-Omid[8] who notes that it is closely
related to inverse power which was described in the previous section. In this section we
summarize the description of the Lanczos procedure presented in [8], which should be
consulted for more detailed information.
In the inverse power with shifts procedure, given a pair of matrices,
[ A 0 ] = ( [ K ] – λ 0 [ M ] ) and [ M ] , and a starting vector, { r } , after n iterations inverse power
generates a sequence of vectors:
 –1 –1 2 –1 n  (Eq. 12-35)
 { r } ,[ A 0 ] [ M ] { r } ,( [ A 0 ] [ M ] ) { r } … ,( [ A 0 ] [ M ] ) { r } 
The sequence converges as n → ∞ to the eigenvector closest to the shift point.

In the Lanczos procedure each vector in the sequence is used to obtain the best
approximation to the eigenvectors. Instead of tossing the intermediate information away, all of
the vectors in the sequence are used to obtain a Rayleigh-Ritz approximation using the Gram-
Schmidt orthonormization procedure which is described by Bathe [7].
In describing the algorithm, assume the first n Lanczos vectors have been found,
 { q } ,{ q } ,{ q } … ,{ q }  (Eq. 12-36)
 1 2 3 n 
and that we want to construct the (n+1)st vector. Each of the preceding vectors has been
normalized with respect to the mass matrix and therefore satisfies the condition that
T
{ φ i } [ M ] { φ j } = δ ij (Eq. 12-37)
where δ ij is the Kroneker delta.

–1 n
To calculate the next Lanczos vector, { s n } = ( [ A 0 ] [ M ] ) { r } , must be normalized
with respect to all of the preceding { q n } . By definition
–1
{ sn } = ( [ A0 ] [ M ] ) { sn – 1 } . (Eq. 12-38)
Since s n – 1 }is already [M]-orthogonal to the previous (n - 2) Lanczos vectors it follows that
{ sn – 1 } = ∑ yi { qi } (Eq. 12-39)
i=1
where y i is the component of the Lanczos vector in the { s n – 1 } direction. The substitution 0f
(Eq. 12-39) into (Eq. 12-38) then gives
∑ yi [ A0 ]
–1
{ sn } = [ M ] { qi }
i=1
(Eq. 12-40)
n–1
∑ yi [ A0 ]
–1 –1
{ sn } = yn [ A0 ] [ M ] { qn } + [ M ] { qi }
i=1
Nastran Primer 329

12
The Lanczos Method
–1
Noting that each vector, [ A 0 ] [ M ] { q i } , can be written as the sum of the first n Lanczos
vectors so that
n–1
∑ yi { qi }
–1
{ sn } = yn [ A0 ] [ M ] { qn } + (Eq. 12-41)
i=1
where the last term of (Eq. 12-41) does not affect the determination of the next Lanczos vector
since it will disappear in the [M]-normalization of the {sn} against all preceding Lanczos
vectors. The next Lanczos vector is determines by first evaluating
–1
{ zn } = [ A0 ] [ M ] { qn } (Eq. 12-42)
Which is then [M]-normalized against all previous eigenvectors. Since { z n } contains

components of all previous Lanczos vectors it can be expressed as
{ zn } = { zn } + αn { qn } + βn { qn – 1 } + γn { qn – 2 } + … (Eq. 12-43)
where { z n } is the component of { z n } which is orthogonal to all previous Lanczos vectors.

Nour-Omid shows that all coefficients in this expansion except αn and βn vanish where
T
αn = { qn } [ M ] { zn }
(Eq. 12-44)
T
βn = { zn – 1 } [ M ] { qn }
Solving for { z n } by substituting (Eq. 12-43) into (Eq. 12-42) gives
–1
{ zn } = [ A0 ] [ M ] { qn } – αn { qn } – βn { qn – 1 } (Eq. 12-45)
The Lanczos vector the then obtained by normalizing { z n }
1
{ q n + 1 } = ---------------------------------------- { z n } (Eq. 12-46)
T
{ zn } [ M ] { zn }
After m Lanczos vectors have been generated they can be arranged in a rectangular array
of a-sized vectors, [Lam]. The eigenvector in the standard form of the eigenvalue problem,
(Eq. 12-26) is approximated by using the following transformation using the Lanczos vectors
{ w a } = [ L am ] { ξ m } (Eq. 12-47)
The substitution of (Eq. 12-47) into (Eq. 12-29) and the application of the [M]-orthonormal
properties of the Lanczos vectors then gives:
T –1
[ L am ] [ M ] ( [ A 0 ] [ M ] – [ Θ mm ] ) [ L am ] { ξ m } = { 0 } (Eq. 12-48)
where [ Θ mm ] is a diagonal matrix of eigenvalues. Because of the form of (Eq. 12-45),

(Eq. 12-48) reduces to the tridiagonal eigenvalue problem
( [ T mm ] – [ Θ mm ] ) { ξ m } = { 0 } (Eq. 12-49)
330 Nastran Primer

The Lanczos Method
where
T –1
[ T mm ] = [ L am ] [ M ] [ A 0 ] [ M ] [ L am ] (Eq. 12-50)
At each iteration of the Lanczos procedure the eigenvalue problem is transformed to a

tridiagonal form in the a subspace. All of the eigenvalues of the reduced problem can now be
economically determined using Givens method, which is described in the next section.
The implementation of the algorithm requires that sufficient Lanczos vectors be
generated to produce sufficient eigenvalues is a given frequency range. The use of the Strum
sequence check[9] which requires a decomposition of [A0] at the upper frequency of interest.
The number of eigenvalues is then equal to the number negative terms on the factor diagonal.
In practice the number of Lanczos vectors will be some multiple, greater than one, of the
number of roots at an below the upper frequency range. The algorithm also must determine
whether the roots in the subspace have converged by iteratively increasing the number of
Lanczos vectors and applying a convergence criterion.
The procedural steps define the Lanczos method. The actual implementation details,
which are not given here, are the key to rendering the method effective and reliable.
12.7.2 The Specification of the Lanczos Procedure - EIGRL

Not available in Legacy NASTRAN. Use the “FEER” option in EIGR ia an alternative.
The Lanczos method is specified by the EIGRL described by Bulk Data Image 12-5
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
where.
Field Description
SID Set identification number, integer > 0.
V1, V2 Frequency range of interest for vibration analysis; eigenvalue range for
buckling, real or blank.
Nd Number of roots desired, integer > 0 or blank.
MSGLVL Diagnostic message level, integer
0 - no messages (default)
1 - Prints eigenvalues accepted at each shift
>1 - increases level of diagnostic output
MAXSET Number of vectors in block or set, integer.
SHFSCH Estimate of the frequency of the first flexible mode, real or blank. A
reasonable estimate will improve performance when rigid body modes are
present.
NORM The method of normalizing the eigenvectors, Character string
MAX - Normalize largest component to one, default for vibration
MASS - Normalize so that modal mass is one, default for buckling
Bulk Data Image 12-5 Specification of Lanczos Method - EIGRL
Nastran Primer 331

12
Tridiagonal Methods
Remarks
1.The real eigenvalue method must be selected by the Case Control directive
METHOD = <SID>
2.The units of V1 and V2 are cycles per unit time for vibration analysis and are
multiples of the load for buckling.
3.MAX normalization is used for buckling analysis
4.The eigenvalues are found in increasing magnitude. The number and type of
eigenvalue is determine by the frequency range and the number of roots entries
according to Table 1-2
Table 1-2 Number and Type of Roots
V1 V2 Nd Action
V1 V2 Nd Lowest Nd or all in the range, whichever
is smaller
V1 V2 blank All in the range
V1 blank Nd Lowest Nd in the range ( V 1 ,∞ )
V1 blank blank Lowest root in the range ( V 1 ,∞ )
blank blank Nd Lowest Nd in the range ( – ∞ ,∞ )

blank blank blank Lowest root
blank V2 Nd Lowest Nd roots below V2
blank V2 blank All below V2
5.In vibration analysis the negative frequency range will be searched if V1 < 0.0. If
V1 is blank all roots less than 0.0 are calculated. Small negative roots generally
indicate rigid body modes. Significant negative values indicates a modeling
problem
6.The eigenvalues are sorted in increasing value for output. An eigenvector is found
for each eigenvector
7.Special control parameters are available when using parallel implementation on
selected computes. See the NASTRAN Quick Reference Guide for further details.
12.8 Tridiagonal Methods

NASTRAN incorporates two basic algorithms, Givens (GIV) and Householder as well as
modified Givens (MGIV) and modified Householder (MHOU) tridiagonal methods. Of these
only GIV is available in Legacy NASTRAN.
Wilkinson [12] notes that for a number of years it seemed that the Givens process was
likely to be the most efficient method of reducing a matrix to tridiagonal form. Householder
suggested that the reduction would be more efficient if Hermetian orthogonal matrices were
used instead of plane rotations used by Givens. Wilkinson notes that the two methods are
essentially the same but that Householders method is more efficient.
Both of these methods are based on the standard form of the eigenvalue problem given by
(Eq. 12-26) The difference between the two basic methods and their modifications is that the
traditional methods uses the Choleski factors of the mass matrix as shown by (Eq. 12-27)while
the modified methods reformulate the eigenvalue problem by defining the eigenvalue, λ, as:
332 Nastran Primer

Tridiagonal Methods
1 –λ
λ = ----- (Eq. 12-51)
Λ′ 0
where λ0 is a shift point. The substitution of ((Eq. 12-51)into ((Eq. 12-11) then gives:
[ – Λ′ ( [ K aa ] + λ 0 [ M aa ] ) + [ M aa ] ] { φ a } = 0 (Eq. 12-52)
The modified eigenvalue problem is thus reduced to standard form by finding the Choleski
factors of ( [ K aa ] + λ 0 [ M aa ] ) which will be non singular even if the a-set contains massless
degrees of freedom. A user-friendly feature is available for removing null columns which exist
in both [ K aa ] and [ M aa ] that is controlled by the ASING parameter which is described in a
later section.
The tridiagonal methods employ the following three steps to determine the eigenvalues
and eigenvectors:
1.Use the Givens or Householder method to reduce ( [ K aa ] + λ 0 [ M aa ] ) to
tridiagonal form
2.Use QR iteration to determine all of the eigenvalues
3.Use inverse iteration to determine eigenvectors in a specified frequency range.
12.8.1 EIGR for the Triadiagonalization Methods

The EIGR data statement for the Givens methods, GIV, MGIV, AGIV and the
Householder methods, HOU, MHOU and AHOU are shown by Bulk Data Image 12-6.
1 2 3 4 5 6 7 8 9 10
EIGR SID Method f1 f2 Nd +E1
+E‘ NORM G C
Where:
Field Description
SID The set identification number that is selected by a METHOD = SID Case
Control directive, integer > 0
Method The name of the requested algorithm, Character string:
GIV - Givens method
MGIV - Modified Givens method
AGIV - Automatic choice of GIV or MGIV
HOU - Householder method
MHOU - Modified Householder method
AHOU - Automatic choice of Householder method
f1, f2 The frequency range over which eigenvectors are to be calculated, real, f2
> f1.
Nd The number of eigenvectors to be calculated starting with the lowest,
integer.
NORM The method used for eigenvector normalization:
MASS - Mass normalization
MAX - Max normalization
POINT - Normalize with respect to displacement degree of freedom
G, C Grid point number and degree of freedom code used if NORM = POINT
Nastran Primer 333

12
Dynamic Degrees of Freedom
Bulk Data Image 12-6 Specification of Tridiagonal Eigenvalue Extraction
The data specified on EIGR is the same for all of the methods except for the second field
which specifies the specific method. Both of the Householder and Givens methods find all of
the eigenvalues in the a-set simultaneously. The recovery of eigenvectors is then controlled by
specifying:
• A frequency range (f1, f2) where f1, is the lowest and f2 is the highest frequency, in
cycles per unit time.
• The number of lowest frequencies for which eigenvectors are to be recovered - Nd
The Nd field is an integer number which defines the number of eigenvectors starting with
that associated with the lowest eigenvalue for either the transformations methods. If present
the Nd specification takes precedence over the frequency range for the transformation
methods.
The NORM field contains one of the literal strings MASS, MAX, or POINT which
defines how the eigenvectors are to be normalized. If POINT is specified then the eigenvectors
are normalized with respect to the displacement associated with the degree of freedom defined
by the pair (G, C) where G is a grid point identification number and C is a degree of freedom
code. The entry in field 3 contains one of the literal strings GIV, MGIV or AGIV to request
one of the Givens methods; or HOU, MHOU or AHOU to request one of the Householder
methods. In either case the SID field is an integer set identification number. The particular
eigenvalue algorithm which is to be incorporated in the analysis is then specified by the
following Case Control Directive.
METHOD = <SID>
12.9 Dynamic Degrees of Freedom

The workflow in perfomning a design simulation rarely starts and ends with a static
analysis. Linear and non-linear Dynamic analyses may be required to verify the design
integrity.
It is accepted practice, especially using Automated Multilevel Substructure analysis, [16],
that is included in Legaacy Nastran to validate each structure component by performing a
component mode anaysis. This has several advantages:
1.Validates the existance of six rigid body modes.
2.Identifies mechnisms if present. Mechanisms will lead to failure in statics unless
they are identified and removed.
3.Identifies normal modes. Are there suffient to span the space?
Performing a static simulation can be deceiving: only on beam element is required to find
the static solution. But the beam must be subdivied into many elements to represent, let’s say
the 4th mode. In that case perhaps 10 spacial points are required to adeqently represent a mode
so 40 element would be require to obtain good resoumtion og the 4th. This same ideal is also
true in buckling analyses.
334 Nastran Primer

Reducing Dynamic Degrees of Freedom
12.10 Reducing Dynamic Degrees of Freedom

On the other hand, NASTRAN can be used to model an almost unlimited variety of
modern structures. It is not uncommon to use several thousand grid points, with tens of
thousands of degrees of freedom in the static analysis of ships, aircraft, automobiles or nuclear
reactors to name a few. Realistic procedures for reducing the number of degrees of freedom
prior to performing dynamic analyses must therefore be formulated.
A number of techniques for reducing the dynamic degrees of freedom have been
presented in the literature including the Guyan Reduction[12], the use of static deflection
shapes and Component Mode Synthesis (CMS)[14]. CMS is, perhaps, more appropriate for
reducing the size of substructures and is not covered here. It is a technique that is supported in
the Automated Multilevel Substructure (AMSS) procedure that is included in Legacy
NASTRAN. See Reference [16].
12.10.1 Guyan Reduction

Static condensation is associated with the OMIT feature in NASTRAN which is described
in Chapter 6 where it is shown that the { u o } and { u a } displacements sets are related by
( { u o } = [ G oa ] { u a } )
–1
(Eq. 12-53)
[ G oa ] = – [ K oo ] [ K oa ]
The { u f } set is then related to the { u a } set by the transformation
( { u f } = [ G fa ] { u a } )
[ G oa ] (Eq. 12-54)
[ G fa ] =
[ I aa ]
If we use the static condensation for the dynamic equations we would run into certain
difficulties. As an illustration consider the following equations for normal modes analysis
partitioned into the a- and o-sets:
 
 – ω 2 [ M oo ] [ M oa ] + [ K oo ] [ K oa ]  { uo }  
=  0  (Eq. 12-55)
 n  { ua }
 [ M ao ] [ M̃ aa ] [ K ao ] [ K̃ aa ]   0 
If we attempt to solve for { u o } using the first matrix equation of (Eq. 12-55), as was done
in the static case, we find that the frequency appears in the transformation matrix which
relates { u o } and { u a } . An iterative procedure would therefore be required to find the
transformation associated with each natural frequency. An approximate technique which is
independent of frequency is therefore desirable. The approximate approach, called the Guyan
Reduction, uses the static condensation transformation (Eq. 12-53) for dynamics so that the
reduced mass matrix becomes:
T
[ M aa ] = [ G fa ] [ M ff ] [ G fa ]
(Eq. 12-56)
T T
[ M aa ] = [ M̃ aa ] + [ M ao ] [ G oa ] + [ G oa ] [ M oa ] + [ G oa ] [ M oo ] [ G oa ]
If the mass coefficients associated with { u o } are null then [ M aa ] = [ M̃ aa ] and the
transformation (Eq. 12-53) is exact.
Nastran Primer 335

12
Removing Matrix Singularities
In order to assess the approximation which is associated with the use of the static relation
(Eq. 12-53) in dynamics it is reasonable to consider the consequences of using (Eq. 12-55) to
define a relation between the o- and a-sets. From the upper partition we find that:
2 –1 2
{ u o } = – ( ω [ M oo ] + [ K oo ] ) ( – ω [ M oa ] + [ K oa ] ) { u a } (Eq. 12-57)
The substitution of (Eq. 12-57) into the lower partition of (Eq. 12-55) then gives the following
reduced equation:
2
[ – ω [ M aa ] + [ K aa ] (Eq. 12-58)
2 2 –1 2
+ ( – ω [ M ao ] + [ K au ] ) ( – ω [ M oo ] + [ K oo ] ) ( – ω [ M oa ] + [ K oa ] ) ] { u a } = 0
2 –1
The inverse term, ( – ω [ M oo ] + [ K oo ] ) , can be represented using a series of expansion
as follows:
2 –1 –1 2 –1 –1
( – ω [ M oo ] + [ K oo ] ) = [ K oo ] + ω [ K oo ] [ M oo ] [ K oo ] +… (Eq. 12-59)
The substitution of (Eq. 12-59) into (Eq. 12-57) then leads to the same reduced stiffness
matrix and reduced mass matrix as those obtained by using static condensation when only
terms through ω2 are retained.
The exact solution of the reduced problem is thus seen to involve an iterated procedure.
Such a procedure is not automated in the normal modes analysis solution algorithm so that
some means of selecting the OMIT degrees of freedom to minimize the frequency-dependent
error in the reduced mass matrix is required. Unfortunately, there is no automated procedure
for selecting those degrees of freedom to be included in OMIT. There are some guidelines,
based on experience, which suggest retaining
1.Degrees of freedom with large inertia.
2.Degrees of freedom with large accelerations (i.e., large relative displacements for
normal mode analysis).
3.Approximately 3.5 times as many well-chosen degrees of freedom as the number
of accurate modes desired in the solution.
4.Translational displacement degrees of freedom for bending problems.
These suggestions almost presuppose that the user knows the solution to the eigenvalue
problem in order to make an intelligent choice of retained physical degrees of freedom in the
a-set. While Guyan Reduction (i.e., the use of the OMIT feature) is inexpensive and is
therefore popular, it requires user prescience to obtain accurate solutions.
12.11 Removing Matrix Singularities

All of the eigenvalue extraction techniques perform a decomposition of the stiffness or
mass matrix or a linear combination of the two. Any degree of freedom which is unconnected
in both matrices is undefined, and its presence in the analysis set will lead to a singular matrix
that, by definition, has no inverse. A singularity check is made in the matrix decomposition
modules in NASTRAN so that a fatal error will be issued and program execution will be
terminated.
336 Nastran Primer

Removing Matrix Singularities
The unconnected degrees of freedom in the stiffness matrix can be purged either explicitly
by single point constraints or implicitly by means of the AUTOSPC parameter (See Chapter
11 and Appendix B). A singular mass matrix is then the analyst's real concern, especially in the
GIV or HOU methods.
The mass matrix associated with the a-set can be made non singular by using static
condensation to remove all massless degrees of freedom from the analysis set. This procedure
leads to a positive definite mass matrix which is required for the GIV and HOU methods and
while a positive definite mass matrix is not required by the MGIV or MHOU methods the
algorithm is generally faster if massless degrees of freedom are removed by static condensation
(OMIT). However, the use of static condensation leads to matrices with greater density and
increased active columns so that the INV generally runs slower if static condensation is
performed on massless degrees of freedom.
MSC and NX NASTRAN provides an user option that can be used to detect singularities
in the stiffness and mass matrices and to then perform matrix operations which are appropriate
for the eigensolution method to remove the singularity. This action is controlled by the integer
ASING parameter which provides two options for dealing with the singularity. The parameter
values and the associated actions are described below.
12.11.1 Detect Singularities and Exit - ASING = -1

ASING is set to - 1 all degrees of freedom with null columns in both the stiffness and mass
matrix are identified and a fatal error exit is taken. This is the proper option to use to assure
that no modeling errors exist prior to performing an expensive eigensolution.
12.11.2 Remove Massless Degrees of Freedom ASING = 0

If ASING is set to zero then all degrees of freedom with null columns are identified and
placed in the w-set where:
{ uw }
{ ua } = (Eq. 12-60)
{ ux }
The a-set mass and stiffness matrices are then partitioned so that
  {u }
 – ω 2 [ M xx ] [ M xw ] + [ K̃ xx ] [ K xw ]  x
= {0} (Eq. 12-61)
 [ M wx ] [ M ww ]  {u } {0}
 [ K wx ] [ K ww ]  w
where [ M xx ] , [ M wx ] and [ M xw ] , are null by definition. All zero diagonal terms in [ K ww ] are
replaced with unit diagonal terms so that we have from the second of (Eq. 12-61)
[ K wx ] { u x } + [ K ww ] { u w } = { 0 } (Eq. 12-62)
The inverse of [ K ww ] now exists because former null columns have been replaced with
unit diagonals so that we can solve for { u w } by using(Eq. 12-62) to obtain
{ u w } = [ G wx ] { u x } (Eq. 12-63)
where
–1
[ G wx ] = – [ K ww ] [ K wx ] (Eq. 12-64)
Nastran Primer 337

12
Input File for Normal Modes Analysis
The reduced set of equations for the x-set (i.e. the solution set for eigensolution) is then
found to be
2
( – ω [ M xx ] + [ K xx ] ) { u x } = { 0 } (Eq. 12-65)
where
˜ ] + [K ] [K ] T
[ K xx ] = [ K xx wx wx (Eq. 12-66)
This formulation, which is associated with ASING = 0, is appropriate for the GIV, HOU,
MGIV and MHOU methods. It will prevent fatal errors which are due to a singular mass matrix
in the GIV and HOU methods if the singularity is caused by null columns in [ M ww ] . However,
mechanisms cannot be detected. That type of singularity does not necessarily prevent
eigensolutions but can cause poor numerical stability.
The MGIV, AGIV MHOU and AHOU methods do not suffer from numerical instability
due to singular or nearly singular mass matrices so that they is more reliable than the GIV and
HOU methods, respectively. The removal of the massless degrees of freedom by static
reduction will reduce solution costs and will also eliminate the calculation of the eigenvalues
for high frequency "noise".
The static reduction of massless degrees of freedom does not introduce any of the
approximations which are associated with the Guyan Reduction. The expediency of replacing
the null columns of the stiffness matrix by unit diagonals is proper because the associated
columns of [ G wx ] are null. The same result could be achieved by first purging the unconnected
degrees of freedom o [ K gg ] followed by an OMIT of massless degrees of freedom, but at a
higher cost.
If the INV method is specified then there is no mass reduction. However, all degrees of
freedom having null columns in both mass and stiffness matrices are removed before
performing the eigensolution.
12.12 Input File for Normal Modes Analysis
12.12.1 Executive Control Section

The executive control section must specify that the normal modes and frequency analysis,
solution 103, is to be executed by means of the following Executive Control Directive.
SOL 103 in MSC/NASTRAN
or
SOL 3 in Legacy NASTRAN
12.12.2 Case Control Section

The following items relate to the selection of Bulk Data and subcase definitions for
normal modes.
1.A METHOD = SID Case Control Directive must be included to select an EIGR
Bulk Data statement defining an eigenvalue extraction method.
2.An SPC set must be selected unless the model is a free body or if the constraints are
specified as permanent constraints on a GRID card or by means of general
elements.
338 Nastran Primer

Parameters
3.Multiple subcases may be used only for output requests. A single subcase is
sufficient if the same output requests are appropriate for all modes. The use of
multiple subcases is discussed in Chapter One.
All of the output Case Control Directives that are associated with static analysis can be
specified for Normal Modes analysis.
12.13 Parameters
The following optional parameters may be used in normal modes analysis. (See Appendix
B of the Primer for the form of the PARAM Bulk Data card and additional discussion of use
of parameters)
1.AUTOSPC Purges unconnected degrees of freedom

YES - Remove singular DOF (default)
NO - Terminate if singular DOF are found
2.COUPMASS Causes the generation of consistent rather than lumped mass.
-1 Lumped formulation (default)
>0 Consistent formulation
3.WTMASS Factor which multiplies all terms in the mass matrix (real)
Other parameters of interest are described in Appendix B.
12.14 Example Problems

We will consider several example problems whose purpose is to show how to employ the
new capability which we have described in this chapter. We will thus use relatively simple
structural topologies to solve several eigenvalue problems including
1.A Cantilever beam
2.A simply supported rectangular plate
12.14.1 Cantilever Beam

The cantilever beam considered for this example is that presented by Fig. 12-2 where it
was noted that the weight density of the beam is 0.281 lb./in3. Since this value will be specified
on the Mat1 data statement we will need to include a WTMASS parameter to change from
weight to mass units.
In defining the finite element model we recognize that the mesh refinement is dictated by
the number of mode shapes that to be adequately represented. We will therefore suppose that
the model is to be capable of representing the fourth symmetric mode, and that approximately
10 equally spaced increments are required to represent a mode. The number of grid points is
thus 41 for the entire beam.
The boundary conditions for a fully fixed boundary are to be imposed at x=0 and the beam
is free at x = L.
The problem will be solved using the Inverse power method to solve for a lumped mass
formulation and Givens method to solve for the consistent mass formulation. The results for
the two formulations will then be compared with the theoretical solutions for natural
frequencies.
Common Model Components
Nastran Primer 339

12
Example Problems
The finite element model for the static problem described in Chapter 12 is not adequate
since it does not have sufficient Grid points to represent the modes of interest. A more refined
model that can serve as the basis for both the inverse power and Givens methods. In preparing
this model:
1.Use the area properties described in Chapter 12
2.Use the material properties specified in Chapter 12 including the weight density
3.Include the WTMASS parameter having a value of 1/g.
Lanczos
Lanczos is to be run for the lumped mass formulation. As we noted earlier, the
COUPMASS parameter controls the type of formulation. Its default value is -1 which results
in the lumped formulation.
Selected Output
The output data for this example problem is shown below where we note
1.The INPUT DATA DECK ECHO, which was requested by ECHO = BOTH in Case
Control prints the input Bulk Data deck exactly as it is read by NASTRAN. In this
case we have used free-field input together with generation and repeat features
2.REAL EIGENVALUES. The table of eigenvalues is automatically printed for
normal modes. The eigenvalues are sorted in increasing value in the 3rd column and
are converted to radians and cycles/second in columns 4 and 5. The generalized
mass is only determinable for calculated eigenvectors (three in this case) and is
equal to 1. because we requested MASS normalization. The generalized stiffness
for this case is seen to be equal to the eigenvalue for modes for which modes were
calculated.
3.REAL EIGENVECTOR - The displacements in the a-set for the four modes is
printed as a result of SVECTOR = ALL
Theoretical Results
The exact solution for the cantilever beam are presented by several authors including
Meirovich [13] who shows that the first three natural frequencies for a uniform cantilever beam
are:
2 EI -
ω 1 = ( 1.875 ) ------------
4
ρAL
ω 2 = ( 4.694 )
2 EI -
------------ (Eq. 12-67)
4
ρAL
ω 1 = ( 7.855 )
2 EI -
------------
4
ρAL
where I is either Iyy or Izz since the section is rectangular.
12.14.2 Simply Supported Plate

The structural model is a square plate with simply supported edges. The finite element
model, which is included in the inp directory of the NASA NASTRAN download from the
GitHub site uses a 10x20 mesh that represents one-half of the plate reducing the number of
GRID points by one-half. Symmetric boundary conditions are used on the midline (x = 0) as
described below.
340 Nastran Primer

Problems
Only bending modes are desired so in-plane displacements and rotations about the normal
are constrained. The FEER method is used for eigen extraction. Both structural mass and non
structural mass are used to define mass matrix.
Propertites
The properties of the square plate are shown by the following table:
Table 1-3 Properties of Square Plate
Property Value
l - Plate length 20 in.
w - Plate width 20 in.
t - Plate thickness 1 in.
E - Elastic Modulus 30 x 106 lbf/in2
ν − Poisson ratio 0.3
ρ - Mass density Structural mass = 200 lbf sec2/in4
Nonstructural mass = 6.039 lbf sec2/in4
Boundary Conditions
The boundary conditions are:
1.Symmetry conditions along x = 0: θx = 0, uy
2.Simple support along y = 0: uz = 0, θx = 0
3.Simple support along x = 10: uz = 0, θy= 0
4.Simple support along y = 20: uz = 0, θx = 0
Eigenvalue Data
1.Method = FEER
2.Center point for FEER: 0.87
3.Number of desired roots: 3
Elements
CQUAD1 elements are used.
Results
The following table lists the NASTRAN and theoretical natural frequencies (cps), where
the theoretical values are defined in Ref. [15]
Table 1-4 Comparison of Plate Natural Frequencies (cps)
Mode
1 .9069 .9056
2 2.2672 2.2634
3 4.5345 4.5329
12.15 Problems
[1] Determine the first three elastic bending modes and frequencies for a free-free beam
6
bending in the x-y plane where the elastic modulus is, E = 30 ×10 psi , the weight density, γ
3 4
= 0.1 lb/in , the cross section moment of inertia, I = 0.1 in , and the length, L = 10 in.
Nastran Primer 341

12
References
[2] Determine the first four bending modes and frequencies of the simply supported plate
described in Section 12.14.2. Plot the modes and normalize the frequencies using the exact
values.
[3] Using 30 elements along the length, use static condensation to remove the displacement
and rotation degrees of freedom at all even numbered node points. Compare the frequencies to
those found in problem 1.
[4] Using the model for problem 1, compare results for Givens, Lanczos or FEER, and
Inverse Power methods.
[5] Using the model for the square plate example, Section 12.14.2, change the element from
CQUAD1 to CQUAD4, making sure to also use the correct property type, and use the Lanczos
method for eigenvalue extraction.
12.16 References
[1] K-J. Bathe, Finite Element Procedures, Prentice Hall, Englewood Cliffs, N. J., 1996.
[2] O.C. Zienkiewicz and R.L Taylor, The Finite Element Method, Fourth Edition, Volume
2: Solid and Fluid Mechanics Dynamics and Non-linearity, McGraw-Hill, Berkshire England,
1991.
[3] T.J.R. Hughes, The Finite Element Method: Linear Static and Dynamic Analysis, Prentice
Hall, Englewood Cliffs, NJ, 1987.
[4] L. Meirovich, Elements of Vibration Analysis, McGraw-Hill, New York, 1972, pp. 65-70.
[5] G. Strang, Linear Algebra and its Applications, Academic Press, New York, 1976, p 221.
[6] ibid, p. 241.
[7] K-J. Bathe, Finite Element Procedures, Prentice Hall, Englewood Cliffs, N. J., 1996, pp.
906-908.
[8] B. Nour-Omid, “The lanczos algorithm for the solution of large generalized
eigenproblems”, in the book by T.J.R. Hughes, The Finite Element Method: Linear Static and
Dynamic Analysis, Prentice Hall, Englewood Cliffs, NJ, 1987., pp. 582-629
[9] K-J Bathe and E.L. Wilson, Numerical Methods in Finite Element Analysis, Prentice Hall,
Englewood Cliffs, N.J., 1976, pp 505-506
[10] ibid, pp. 465-471.
[11] NASTRAN Theoretical Manual, NASA SP-221, Section 5.5.3.
[12] R. Guyan, “Reduction of Stiffness and Mass Matrices,” AIAA J., 3,(2), 1965, pp. 380.
[13] L. Meirovich, Elements of Vibration Analysis, McGraw-Hill, New York, 1972, p. 212.
[14] R.R. Craig, “A Review of Time-Domain and Frequency Domain Component Mode
Synthesis Methods”, Int. J. Analytical and Experimental Modal Analysis, Vol. 2, No. 2, 1987
[15] W.F Stokley, “Vibration of Systems Having Distributed Mass and Elasticity”, Chap. 7,
Shock and Vibration Handbook, C.M. Harris and C.E. Crede, Editors, McGraw-Hill, 1961.
[16] Anon., “NASTRAN USER’S MANUAL”, NASA SP-222, VOL. I, Sec. 1.10.2.
342 Nastran Primer

13 Build Legacy NASTRAN
NASA has published the source code for NASA NASTRAN on their GitHub site.
(Googling for “NASA NASTRAN” should get you there). The complete download includes
the source code, the DMAP for all rigid formats, the NASA NASTRAN manuals and the
verifications problems. It’s quite a treasure trove for software archeologists to rummage
around in!
13.1 Before You Start

NASTRAN is a fairly complicated program. You really need some understanding of the
architecture to get started. I am going to assume that is the case. In what follows I will outline
the workflow for getting Legacy NASTRAN running. You can then judge what to do next.
13.1.1Go Windows
I have lots of experience with Compac FORTRAN Visual Studio 6.5 which is a wonderful
development environment. The compiler has all of the right stuff to compile and link Legacy
NASTRAN and the debugger is unequalled in my humble opinion. However, the current
FORTRAN compilers that support Windows are too dear for my budget so I continue to rely
on my old faithful. If you have access to either Portland Group or Intel compilers just make
sure they support Cray pointers.
With that as a background and having access to a suitable FORTRAN installation you will
have no problem with building an executable for Legacy NASTRAN. You will find some
compile errors, perhaps the same as those noted for gfortran in the next section, but they are
easy to overcome. The best part is that the compiler will support varaible formats a capability
in which gfortran failed.
So read through the next section, fire up your Visual Studio FORTRAN and copy all the
source files to whatever project you choose, perhaps nastran. Then build to find compiler
errors, fix them, build again and you should have it.
You will still need to create a script file; or, perhaps create a FORTRAN front end to serve
the same funtion. That is what I did since I have more confidence in my FORTRAN skills than
my Powershell skills. Either way you can use the script file in the NASA NASTRAN
installation directory as a guide.
13.1.2Go Linux
You need a FORTRAN compiler. I suggest using the GNU toolchain including the C,
C++ and gfortran compilers. You don’t need the C compilers right now but you will if you
want to add new code. If you have a Linux system installed all you need to do is install the
GNU compilers. If you don’t know how to do that you probably shouldn’t be thinking about
giving Legacy NASTRAN new life.
If your computer has windows installed you will need to install MinGW and MSYS. It’s
not hard to do, just do some Googling, it’s amazing what you can find on the web. I am going
to assume that you have done this and have installed the GNU tool chain as I described above.
343 Nastran Primer

Build Legacy NASTRAN
13
Compiling and linking.
You are almost there. Now download Legacy NASTRAN from the NASA GitHub site
into your Linux (MinGW) home directory. The path will be something like:
c:/mingw/msys/1.0/<your login name>. If you haven’t already done so open a command shell
in Windows, change directory to c:/mingw/msys and type “msys’ to open a bash shell and then
cd to the nastran directory (I am assuming you have downloaded to that directory).
13.2 Compiling and linking.

If you haven’t used a make file you need to do a little research on the subject. Make is a
utility that controls compiling and linking the source code. Once you have created the file
called makefile all you need to do is enter make in the bash shell. But I am getting ahead of
myself.
13.2.1Work directory
Create a work directory as a subdirectory in the nastran directory and, in the work
directory, create the following subdirectories:
Table 13-1 Sandbox Directories

Directory Name Description
sb sandbox for the executable
include include file
bin bash files
test test files
Right now all these directories are empty but we will start working on that by:
1. copying all *.f files from nastran/mds to work/sb

2. copying all *.f files from nastran/mis to work/sb
We still need to create a bash script for executing nastran and copy a model file from the
nastran/nid demonstration directory to the test directory but we have some serious work ahead
of us before we get that far. Once we have a successfully linked nastran program we will get
to those details.
13.2.2Build NASTRAN
Firstly. you will need a makefile. Copy the following and save it as “makefile” in work/sb:
# Declaration of variables
FC = gfortran
FFLAGS = -w -fno-range-check -fno-automatic -fcray-pointer -g
# file names
EXEC=nast
SOURCES = $(wildcard *.f)
OBJECTS = $(SOURCES:.f=.o)
344 Nastran Primer

# Main target
$(EXEC):$(OBJECTS)
$(FC) $(OBJECTS) -g -o $(EXEC)
# To obtain the object files

%.o:%.f
$(FC) $(FFLAGS) -g -c $< -o $@
# To remove generated files
clean:
rm -f $(EXEC) $(OBJECTS)
Then, create a new directory “nastprog” in the work directory and move nasthelp.f and
nastplot.f there. These are utility programs that are not a intrinsic part of nastran.
Then open a bash shell and cd to sb and enter make in the command window. The
following routines will fail to compile.
Table 13-2 sb Compile failures

Routine Problem
conmsg.f Error in time and date intrinsic
ffread.f Error in open statement at line 376
mapfns.f Error in calling LOC at line 138. Change %LOC to
LOC
nsinfo.f Error in open at line 68
pexit.f Error at line 45. LINK is external
rfopen.f Need to change open at line 81
sgino.f Need to change open statements.
sofut,f Error in calling rename. It is external
tdate.f Need to replace time and date intrinsic
After making these changes the routines will compile but fail to link. The link errors are:
1.Both errtrc.f and dummy.f produce the symbol “errtrc”. Remove the
symbol from dummy.f
2. remove endsys.f. It is no longer used.
3. move chkfil.f. to nastprog directory. It is a utility program.
4. remove shtrmd.f and shhmgd.f
Nastran should now link and create the executable called run.exe. Now the fun begins.
Nastran Primer 345

13
13.2.3Create the nastran bash file

In order to execute run.exe you need to create a bash script. You can use the script file in
the download as a guide or you can copy the following script file and save in it the work/test
directory.
#!/bin/bash
cwd=$(pwd)
echo The current work Directory is - $cwd
progname=”$1”
ft05=”$progname.nid”
ft06=”$progname.out”
nptp=”$progname.nptp”
echo Input file is: $ft05
sof1=”none”
sof2=”none”
rfdir=”/home/harry/nasanastran/rf”
nasthome=”/home/harry/nasanastran/work”
tmp=”$nasthome/test/tmp”
projectdir=”$nasthome/test”
nasexec=”$nasthome/sb/run”
mkdir -p $tmp
nasscr=$tmp
dbmem=”14000000”
ocmem=”14000000”
echo ==== NASTRAN is beginning execution of $progname ====
NPTPNM=$nptp;export NPTPNM
PLTNM=”$progname.plt”;export PLTNM
DICTNM=”none”;export DICTNM
PUNCHNM=”$progname.pnh” ;export PUNCHNM
OUT11=”$progname.out11” ;export OUT11
IN12=”$progname.in12” ;export IN12
FTN11=”none”;export FTN11
DIRCTY=$nasscr ;export DIRCTY
lognm=”$progname.log”
LOGNM=$lognm;export LOGNM
OPTPNM=”none”;export OPTPNM
RFDIR=$rfdir ;export RFDIR
SOF1=$sof1;export SOF1
SOF2=$sof2 ;export SOF2
FTN16=”none”;export FTB16
DBMEM=$dbmem;export DBMEM
OCMEM=$ocmem ;export OCMEM
FT05=$ft05;export FT05
346 Nastran Primer

FT06=$ft06;export FT06
CWD=$cwd;export CWD
PROJ=$projectdir;export PROJ
#printenv
$nasexec
#gdb $nasexec
echo ===== NASTRAN has completed problem $progname =====
You will need to edit the paths for rfdir and nasthome. You can also change the extensions
for ft05, the input filr and ft06, the output file. Many nastran users us “bdf” or “dat” for the
input file and “f06” for the output file extensions.
13.2.4Additional Modifications
Additional changes are required to eliminate run time errors associated with file I/O.
Firstly the size of the character variable associated with environment variables and an array
called dsnames must be changed. to do this make modifications to files in the sb directory as
follows:
1.In nastrn.f change the lengths of the character variables “value” and
“dsnames” from 44 to 80, and add an integer parameter
“len_dsnames” with a value of 80.
2. In nastrn.f, change “do 5 i=44,1,-1” to “do 5
i=len_dsnames,1,-1”.
3. In gino.f and ginio.f, change the size of the dsnames variable from
44 to 80.
4. In xsem00.f, add the following in the definition section:
! Set variables
integer error_id
integer nin, nout
character(80) proj,ft05,ft06,output,infile
5. In xsem00.f, add the following equivalence, “equivalence

(xx(3),nin).
6. In xsem00.f, add the following after “call btstrp”:
!----------------------------------------------------------------------
! hgs 12/06/2104 - The NASA delivery uses redirected input and out put
! so these files are not explicitly opened. I need
! to change the stdin and stdout to allow the use of
! GDB in the script file. Therefore the following mod-
! ifications explicitly open the stdin and stdout files
! using the FTN5 and FTN6 ENV set by the script.
!
c
c open inout and output files
c
call getenv(‘PROJ’,proj)
call getenv(‘FT05’,ft05)
call getenv(‘FT06’,ft06)
Nastran Primer 347

13
output = trim(proj)//’/’//trim(ft06)
error_id = 0
101 continue
ifile = nout
open(nout,file=output,form=’formatted’,
1status=’unknown’,iostat=ierr,err=102)
go to 103
102 continue
error_id = -2
go to 104
103 continue
ifile = nin
infile = trim(proj)//’/’//trim(ft05)
open(nin,file=infile,form=’formatted’
1 ,status=’unknown’, iostat =ierr,err=106)
go to 105
106 continue
error_id = -1
104 continue
c
c open error
c
write(nout,*) ‘Error in opening file =’,ifile,’ IOSTAT = ‘,ierr
select case(error_id)
case(-1)
write(nout,’(a)’) ‘File name: ‘,infile
case(-2)
write(nout,’(a)’) ‘File name: ‘,output
end select
call pexit ! close down and exit with ierror
return
105 continue
c
I notice that there are tab characters in the above code fragment so some editing might be
required.
7. In btstrp.f, at line 180 change the variable “mach” to 6.

8. run make in the sb directory. Hopefully there are no errors. If there
are rather than trying to sort them out you can got to
“www.mycosim.com” for help.
13.2.5Run a test file

Copy the file d01081a.nid to the test directory. Then change to the test directory and enter
“nast.sh d01081a”. Assuming no link errors the problem will apparently run to completion, and
it almost makes it to the end. However, there is a problem in the OFP module that leads to an
error. It turns out that OFP used a viable format write that gfortran won’t handle. So, bummer.
If you had used another compiler such as as visual studio fortran you won’t have this problem
and your run should go to s successful completion.
348 Nastran Primer

Don’t give up on gfortran, this is the only limitation I have found but we need a work-
around. That is to replace the current ofpprt.f and add a routine forwrt.f. I include those routines
here, but suggest that might make more sense to go to my github site and download a modified
nastran installation that I call “classicnastran”.
13.2.5.1 OFPPRT
SUBROUTINE OFPPNT (OUT,NWDS,FMT)
C
CWKBD LOGICAL DEBUG
CWKBR INTEGER OUT(NWDS), FMT(300)
INTEGER OUT(NWDS)
CHARACTER*1 FMT(1200)
CWKBI
COMMON /MACHIN/ MACHX
COMMON /SYSTEM/ SYSBUF, L
CWKBD DATA DEBUG / .FALSE. /
C
CWKBD IF (DEBUG) WRITE (L,10) (FMT(K),K=1,32)
10 FORMAT (‘ FMT=’,32A4)
CWKBR 5/95 IF ( MACHX.EQ.2 .OR. MACHX.EQ.5 )
IF ( MACHX.EQ.2 .OR. MACHX.EQ.5 .OR. MACHX .EQ. 21 )
* WRITE (L,FMT,IOSTAT=IOSXX) (OUT(K),K=1,NWDS)
CWKBR 5/95 IF ( MACHX.NE.2 .AND. MACHX.NE.5 )
IF ( MACHX.NE.2 .AND. MACHX.NE.5 .AND. MACHX .NE. 21 )
* CALL FORWRT (FMT, OUT, NWDS)
RETURN
END
13.2.5.2 FORWRT SUBROUTINE FORWRT ( FORM, INDATA,

NWDS )
C********************************************************************
C EXPECTED TYPES OF FORMAT CODES ARE AS FOLLOWS
C NH------ NENN.N NDNN.N NX
C NFNN.N NINN NGNN.N NAN
C NPENN.N NPFNN.N NPN(----) NP,ENN.N
C NP,FNN.N NP,N(----)
C SPECIAL CHARACTERS: /(),
C ICHAR = CURRENT CHARACTER NUMBER BEING PROCESSED IN “FORM”
C ICOL = CURRENT CHARACTER COLUMN POSITION WITHIN THE LINE
C NCNT = NUMBER OF VALUES OF IDATA AND DATA THAT HAVE BEEN
PROCESSE
C********************************************************************
CHARACTER*1 FORM(1000)
CHARACTER*1 SLASH , BLANK
CHARACTER*1 LPAREN, RPAREN, PERIOD, COMMA, NUMBER(10)
CHARACTER*1 H, E, D, X, F, I, G, A, P
CHARACTER*2 PFACT
CHARACTER*4 CDATA(200)
CHARACTER*132 LINE
CHARACTER*132 TFORM
INTEGER*4 INDATA(NWDS), IDATA(200)
REAL*4 DATA(200)
Nastran Primer 349

13
REAL*8 DDATA(100)
COMMON /SYSTEM/ ISYSBF, IWR
EQUIVALENCE (IDATA, DATA, DDATA, CDATA )
DATA H/’H’/, E/’E’/, D/’D’/, X/’X’/, F/’F’/
DATA I/’I’/, G/’G’/, A/’A’/, P/’P’/
DATA LPAREN /’(‘/, RPAREN/’)’/, PERIOD/’.’/
DATA COMMA /’,’/, SLASH /’/’/, BLANK /’ ‘/
DATA NUMBER /’0’,’1’,’2’,’3’,’4’,’5’,’6’,’7’,’8’,’9’/
IF ( NWDS .LE. 200 ) GO TO 2
PRINT *,’ LIMIT OF WORDS REACHED IN FORWRT, LIMIT=200’
CALL PEXIT
2 DO 3 KB = 1, NWDS
IDATA( KB ) = INDATA( KB )
3 CONTINUE
ILOOP = 0
ICHAR = 1
NCNT = 1
ICOL = 1
LINE = BLANK
PFACT = BLANK
ICYCLE= 0
5 IF ( FORM(ICHAR) .EQ. LPAREN ) GO TO 75
ICHAR = ICHAR + 1
IF ( ICHAR .LE. 1000 ) GO TO 5
GO TO 7702
70 IF ( ICHAR .GT. 1000 ) GO TO 7702
IF ( FORM(ICHAR) .EQ. BLANK ) GO TO 75
IF ( FORM(ICHAR) .EQ. SLASH ) GO TO 100
IF ( FORM(ICHAR) .GE. NUMBER(1) .AND.
& FORM(ICHAR) .LE. NUMBER(10) ) GO TO 200
IF ( FORM(ICHAR) .EQ. A ) GO TO 300
IF ( FORM(ICHAR) .EQ. I ) GO TO 400
IF ( FORM(ICHAR) .EQ. H ) GO TO 500
IF ( FORM(ICHAR) .EQ. X ) GO TO 600
IF ( FORM(ICHAR) .EQ. P ) GO TO 700
IF ( FORM(ICHAR) .EQ. F ) GO TO 800
IF ( FORM(ICHAR) .EQ. G ) GO TO 800
IF ( FORM(ICHAR) .EQ. D ) GO TO 800
IF ( FORM(ICHAR) .EQ. E ) GO TO 800
IF ( FORM(ICHAR) .EQ. LPAREN ) GO TO 1000
IF ( FORM(ICHAR) .EQ. RPAREN ) GO TO 1100
IF ( FORM(ICHAR) .NE. COMMA ) GO TO 7702
IF ( ICYCLE .EQ. 0 ) PFACT = BLANK
75 ICHAR = ICHAR + 1
GO TO 70
C PROCESS SLASH
100 CONTINUE
IF ( LINE .NE. BLANK ) WRITE ( IWR,900 ) LINE
900 FORMAT(A132)
IF ( LINE .EQ. BLANK ) WRITE ( IWR,901 )
901 FORMAT(/)
LINE = BLANK
IF ( ICYCLE .EQ. 0 ) PFACT = BLANK
ICOL = 1
350 Nastran Primer

GO TO 75
C GET MULTIPLIER FOR FIELD CONVERSION
200 CALL FORNUM ( FORM, ICHAR, IMULT )
GO TO 70
C PROCESS ALPHA FIELD--FORMAT(NNANNN) (NN=IMULT,NNN=IFIELD)
IF ( NCNT .GT. NWDS ) GO TO 1200
CALL FORNUM ( FORM, ICHAR, IFIELD )
ILEFT = NWDS - NCNT + 1
IF ( ILEFT .LT. IMULT ) IMULT = ILEFT
IF ( IMULT .EQ. 0 ) IMULT = 1
WRITE ( TFORM, 902 ) IMULT, IFIELD
902 FORMAT(‘(‘,I2,’A’,I2,’)’)
I1 = ICOL
LENGTH = IMULT*IFIELD
NEND = NCNT + IMULT - 1
LAST = ICOL + LENGTH - 1
WRITE( LINE(ICOL:LAST), TFORM ) (CDATA(KK),KK=NCNT,NEND)
ICOL = ICOL + LENGTH
NCNT = NCNT + IMULT
IMULT = 1
GO TO 70
C PROCESS INTEGER FIELD -- FORMAT(NNINNN) (NN=IMULT,NNN=IFIELD)
WRITE ( TFORM, 903 ) IMULT, IFIELD
903 FORMAT(‘(‘,I2,’I’,I2,’)’)
I1 = ICOL
WRITE( LINE(ICOL:LAST), TFORM ) (IDATA(KK),KK=NCNT,NEND)
NCNT = NCNT + IMULT
IMULT = 1
GO TO 70
C PROCESS HOLERITH FIELD -- FORMAT(NNH----) (NN=IMULT)
500 LAST = ICOL + IMULT - 1
ICHAR = ICHAR + 1
LCHAR = ICHAR + IMULT - 1
WRITE ( LINE(ICOL:LAST), 904 ) (FORM(KK),KK=ICHAR,LCHAR)
904 FORMAT(133A1)
ICOL = ICOL + IMULT
ICHAR = LCHAR
IMULT = 1
GO TO 75
C PROCESS X FIELD -- FORMAT(NNX) (NN=IMULT)
600 WRITE ( TFORM, 905 ) IMULT
905 FORMAT(‘(‘,I2,’X’,’)’)
LAST = ICOL + IMULT - 1
WRITE( LINE(ICOL:LAST), TFORM )
ICOL = ICOL + IMULT
Nastran Primer 351

13
IMULT = 1
GO TO 75
C PROCESS P FACTOR FOR FLOATING FORMAT
700 WRITE ( PFACT,904 ) FORM(ICHAR-1), FORM(ICHAR)
710 IF ( FORM( ICHAR+1 ) .NE. BLANK .AND. FORM( ICHAR+1 ) .NE.
& COMMA ) GO TO 75
ICHAR = ICHAR + 1
IF ( ICHAR .GT. 1000 ) GO TO 7702
GO TO 710
C PROCESS FLOATING FIELD -- FORMAT(NPNNXNNN.NNNN) WHERE
C (NP = PFACT, NN=IMULT, NNN=IFIELD, NNNN=IDEC)
800 ITYPE = ICHAR
ICHAR = ICHAR + 1
810 IF ( FORM( ICHAR ) .EQ. PERIOD ) GO TO 820
ICHAR = ICHAR + 1
GO TO 810
CALL FORNUM ( FORM, ICHAR, IDEC )
WRITE ( TFORM, 906 ) PFACT, IMULT, FORM(ITYPE),IFIELD, IDEC
906 FORMAT(‘(‘,A2,I2,A1,I2,’.’,I2,’)’)
I1 = ICOL
IF ( FORM(ITYPE) .EQ. D )
& WRITE( LINE(ICOL:LAST), TFORM ) (DDATA(KK),KK=NCNT,NEND)
IF ( FORM(ITYPE) .NE. D )
& WRITE( LINE(ICOL:LAST), TFORM ) (DATA(KK),KK=NCNT,NEND)
NCNT = NCNT + IMULT
IMULT = 1
GO TO 70
C PROCESS LEFT PAREN (NOT THE FIRST LEFT PAREN BUT ONE FOR A
GROUP)
C IMULT HAS THE MULTIPLIER TO BE APPLIED TO THE GROUP
1000 ICYCLE = IMULT-1
ICSAVE = ICHAR+1
ILOOP = 1
IMULT = 1
GO TO 75
C PROCESS RIGHT PAREN ( CHECK IF IT IS THE LAST OF THE FORMAT)
C IF IT IS PART OF A GROUP, THEN ICYCLE WILL BE NON-ZERO
1100 IF ( ICYCLE .GT. 0 ) GO TO 1110
IF ( ILOOP .NE. 0 ) GO TO 1120
C NO GROUP, THEREFORE MUST RE CYCLE THROUGH FORMAT
C UNTIL LIST IS SATISFIED
WRITE ( IWR,900 ) LINE
ICHAR = 2
LINE = BLANK
352 Nastran Primer

Taking Stock
PFACT = BLANK
ICOL = 1
GO TO 70
C GROUP BEING PROCESSED, DECREMENT COUNT AND RESET ICHAR TO
BEGINNING
C OF THE GROUP
1110 ICYCLE = ICYCLE - 1
ICHAR = ICSAVE
GO TO 70
C FINISHED WITH LOOP, CONTINUE WITH FORMAT
1120 ILOOP = 0
ICYCLE = 0
GO TO 75
1200 WRITE ( IWR,900 ) LINE
7000 CONTINUE
RETURN
7702 WRITE( IWR, 9901 ) ICHAR, FORM
9901 FORMAT(///’ SUBROUTINE FORWRT UNABLE TO DECIPHER THE
FOLLOWING’
& ,’ FORMAT AT CHARACTER ‘,I4,/,’ FORMAT GIVEN WAS THE
FOLLOWING:’
& ,/,(1X,131A1))
END
13.3 Taking Stock

If all has gone well you now have a working version of the NASTRAN installation that
you downloaded from the NASA NASTRAN GitHub site. It has been a struggle at times but
congratulations if you persevered. Now, where to we go from here?
There are lots of changes, the most pressing is an improved element library. The next is
to replace or augment the current solvers for statics and eigenvalue extraction with those using
sparse logic. There are many open source codes out there. Then there are utilities to be
upgraded such as replacing the current MPC module with one that finds the best m-set rather
than requiring the user to do so. And speaking of MPC equations, additional constraint
generators are required including contact and glue connections.
But the JEWEL that comes with Legacy NASTRAN is the substructure capability called
Automated Multilevel SubStructures (AMSS). It truly is a great start to a new paradigm for
changing the workflow in creating models of assemblies.
13.3.1Next Step
I have a lot of experience with NASTRAN development starting with an installation
delivered in 1995. In the two years after the 1993 release that is on the NASA GitHub, the
maintenance team at COSMIC made some important changes. Those changes are included in
the download for classicalnastran that is discussed on “www.mycosim.com”. I suggest starting
there. It’s a much improved baseline for future developments.
Nastran Primer 353

13
Taking Stock
354 Nastran Primer

A Use of Parameters
Parameters may be specified by the user by means of a PARAM-card in Bulk Data. The
PARAM-card is as shown by Card Image A-1.
1 2 3 4 5 6 7 8 9 10
PARAM N V1 V2
where:
N Parameter name
V1, V2 Parameter value depending on the parameter type as follows:
Type V1 V2
Integer Integer Blank
Real, single-precision Real Blank
Literal Literal Blank
Real, double-precision D.P. Blank
Complex, single-precision Real Real
Complex, double-precision D.P. D.P.
Bulk Data Image A-1 The PARAM Bulk Data Statement
The allowable PARAMeters that might be useful for rigid formats 1, 101, 3, and 103 are
summarized below.
ASING (MSC Only) Automatic elimination of matrix singularities for normal modes
analysis (Sec. Chapter 12)
AUTOSPC (MSC) Automatic purge of unconnected degrees of freedom Chapters 6
and 12). associated PARAMeters include
EPPRT
EPZERO
PRGPST
SPCGEN
USETPRT
USETSEL
AUTOSPC Automaticically detect and, if found, remove under user control.
(Legacy NASTRAN)) Integer
0 - Detect, create and print the Grid Point Singularity Table.
(GPST). (Will probaly result in failure in the solve operation.)
1 - Process GPST and create SPC1s to remove the singular degrees
of freedom
-1 - Same as = 1 but write the SPC1s to the PUNCH file
355 Nastran Primer

Use of Parameters
A
COUPMASS A positive value (default is -1) requests the coupled rather than the
lumped mass formulation.
GRDPNT A positive value (default is -1) requests the weight and balance
information described in Sec. 12.10 to be calculated and displayed
using the grid point number, defined as the value of the
PARAMeter as the reference point.
IRES A positive value (default is -1) requests that the residual for the a-
and o-sets be displayed.
MAXRATIO The presence of this parameter will cause the conditioning numbers
for each degree of freedom to be compared to the value of the
5
PARAMeter (default is 10 ). If they are greater than MAXRATIO
the matrix is considered to include a mechanism. The conditioning
numbers and the external degree of freedom codes are printed and
execution is terminated.
WTMASS The mass matrix is multiplied by the value of this parameter
(default is 1.) prior to dynamic analysis. The use of this
PARAMeter allows the analyst to define the density and
nonstructural mass in terms of weight units which are then
multiplied by a value of 1/g which is specified as the WTMASS
PARAMeter value.
356 Nastran Primer

Use of Parameters
Nastran Primer 357

(Harry G. Schaeffer) NASTRAN Primer (B-Ok - Xyz) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Harry G. Schaeffer) NASTRAN Primer (B-Ok - Xyz) PDF

Uploaded by

Copyright:

Available Formats

NASTRAN Primer

Static and Normal Modes Analysis

Copyright © 2015 by Harry G. Schaeffer

Warning and Disclaimer

Nastran Primer iii

Nastran Primer vii

viii Nastran Primer

xii Nastran Primer

1.1 Historical Introduction

1.1.1 The Finite Element Method

1.1.2 Solving NASA’s Simulation Needs

1.1.3 NASA NASTRAN

1.1.4 MSC NASTRAN

1.2 NASTRAN® Architecture

1.2.1 Independent Modules

1.2.3 Executive System

1.3 NASTRAN Technology

1.3.1 Engineering Mechanics

1.3.2 Numerical Analysis

1.3.3 Computer Science

1.3.4 Computer Graphics

1.4 Design of the Finite Element Model

1.4.1 Basic Modeling objects

1.4.2 The Design Procedure

1.5 Processing the Finite Element Model

1.5.1 Marshall the Model Data

1.5.2 Generation of Element Stiffness Matrices

1.5.3 Assemble the Unconstrained Stiffness Matrix

1.5.4 Detect and remove essential singularities

1.5.5 Apply Constraints

1.5.6 Generate Loads

1.5.7 Solve for Constrained Displacements

1.5.8 Recover Unconstrained Displacements

1.5.9 Recover Element Behavior

1.5.10 Stress Averaging at the Grid Points

1.5.11 Strain Energy Density

1.5.12 Results Files

1.5.13 Post Processing

1.6 User Program Interface

1.6.1 Input File

1.6.2 Types of Solutions.

Solution Number Description of Solution

Solution Number Description of Solution

Table 1-1 Solution Sequences in MSC.Nastran

1.7 The Input File

1.7.1 NASTRAN Statement

1.7.2 Problem Definition Section

Delimiter Name Purpose

Table 1-2 Delimiter Statements

1.7.3 Getting Started with MSC.Nastran

1.8 Specifying the Finite Element Model - Bulk Data

1.8.1 Free Field Bulk Data

1.8.2 Replication and Field Generation

1.8.3 Summary of Structurally Oriented Bulk Data

Data Statement Description Reference

1.8.3.1 Scalar Degrees of freedom

SPOINT Defines scalar point Defining Scalar

1.8.3.2 Coordinate Systems

1.8.3.3 Grid Points

1.8.3.4 Constraint and Partitioning

OMIT Specify degrees of freedom to be removed from the Static

1.8.3.5 General Element

1.8.3.6 Scalar Elements