Nastran Advanced Dynamics

MSC.
Nastran Version 70
Advanced Dynamic Analysis
Users Guide
Corporate
MSC.Software Corporation
2 MacArthur Place
Santa Ana, CA 92707 USA
Telephone: (800) 345-2078
Fax: (714) 784-4056
Europe
MSC.Software GmbH
Am Moosfeld 13
81829 Munich, Germany
Telephone: (49) (89) 43 19 87 0
Fax: (49) (89) 43 61 71 6
Asia Pacific
MSC.Software Japan Ltd.
Shinjuku First West 8F
23-7 Nishi Shinjuku
1-Chome, Shinjyku-Ku, Tokyo 160-0023, Japan
Telephone: (81) (03) 6911 1200
Fax: (81) (03) 6911 1201
Worldwide Web
www.mscsoftware.com
Disclaimer
This documentation, as well as the software described in it, is furnished under license and may be used only in accordance with the terms of
such license.
MSC.Software Corporation reserves the right to make changes in specifications and other information contained in this document without
prior notice.
The concepts, methods, and examples presented in this text are for illustrative and educational purposes only, and are not intended to be
exhaustive or to apply to any particular engineering problem or design. MSC.Software Corporation assumes no liability or responsibility to
any person or company for direct or indirect damages resulting from the use of any information contained herein.
User Documentation: Copyright ! 2004 MSC.Software Corporation. Printed in U.S.A. All Rights Reserved.
This notice shall be marked on any reproduction of this documentation, in whole or in part. Any reproduction or distribution of this
document, in whole or in part, without the prior written consent of MSC.Software Corporation is prohibited.
The software described herein may contain certain third-party software that is protected by copyright and licensed from MSC.Software
suppliers.
MSC, MSC/, MSC., MSC.Dytran, MSC.Fatigue, MSC.Marc, MSC.Patran, MSC.Patran Analysis Manager, MSC.Patran CATXPRES,
MSC.Patran FEA, MSC.Patran Laminate Modeler, MSC.Patran Materials, MSC.Patran Thermal, MSC.Patran Queue Manager and PATRAN
are trademarks or registered trademarks of MSC.Software Corporation in the United States and/or other countries.
NASTRAN is a registered trademark of NASA. PAM-CRASH is a trademark or registered trademark of ESI Group. SAMCEF is a trademark
or registered trademark of Samtech SA. LS-DYNA is a trademark or registered trademark of Livermore Software Technology Corporation.
ANSYS is a registered trademark of SAS IP, Inc., a wholly owned subsidiary of ANSYS Inc. ABAQUS is a registered trademark of ABAQUS
Inc. ACIS is a registered trademark of Spatial Technology, Inc. CATIA is a registered trademark of Dassault Systemes, SA. EUCLID is a
registered trademark of Matra Datavision Corporation. FLEXlm is a registered trademark of GLOBEtrotter Software, Inc. HPGL is a
trademark of Hewlett Packard. PostScript is a registered trademark of Adobe Systems, Inc. PTC, CADDS and Pro/ENGINEER are
trademarks or registered trademarks of Parametric Technology Corporation or its subsidiaries in the United States and/or other
countries.Unigraphics, Parasolid and I-DEAS are registered trademarks of Electronic Data Systems Corporation or its subsidiaries in the
United States and/or other countries. All other brand names, product names or trademarks belong to their respective owners.
C O N T E N T S
MSC.Nastran Advanced Dynamic Analysis Users Guide

Preface I About this Book, x
I List of MSC.Nastran Books, xii
I Technical Support, xiii
I www.mscsoftware.com, xv
I Documentation Conventions, xvi
I Permission to Copy and Distribute MSC Documentation, xvii
1
Introduction I Overview, 2
I Purpose of this Guide, 3
I Range of Readers, 4
I Summary, 5
2
Dynamic Modeling
Options
I Overview, 8
I Mass Modeling, 10
Mass Data Input Options, 10
Coupled Mass Matrix Terms, 11
Mass Effects in Moving Coordinate Systems, 13
I Modeling Damping Effects, 17
Viscous Damping, 17
Structural Damping, 17
Modal Damping, 19
Nonlinear Damping, 21
I DMIGs, Extra Points, and Transfer Functions, 23
Modal Transformations, 24
Direct Matrix Inputs, 24
Transfer Functions, 25
Pickups and Transducers, 26
Higher-Order Transfer Functions, 26
I Nonlinear Load Functions (NOLINi), 31
Theory for the Standard Transient Solutions, 32
Example of a Nonlinear Spring, 33
Velocity-Dependent Nonlinear Forces, 33
Nonlinear Transient Solution Sequences, 34
MSC.Nastran Advanced Dynamic
Analysis Users Guide
Recommendations for Using NOLIN1 Input, 35
Application Example Coupling of Rotating Structures, 36
I Debugging Dynamic Models, 39
Modeling Hints, 39
Testing Dynamic Models, 40
References, 60
3
System Matrix
Formulations
I Overview, 62
I Theoretical Basis for Reduction Methods, 63
Definition, 63
Statics, 63
Guyan Reduction, 66
General Dynamic Reduction, 66
I Dynamics of Rotating Structures, 70
Technical Approach, 70
DMAP Procedure to Add Rotation-Dependent Terms to MSC.Nastran
Dynamic Analyses, 79
I Overview of Aeroelastic Analysis, 81
Static Aeroelastic Response, 81
Aerodynamic Flutter, 82
Dynamic Aeroelastic Response, 82
Aeroelastic Optimization, 82
Aerodynamic Methods, 83
Special Features, 83
4
Boundary
Conditions and
Coupling to Other
Disciplines
I Overview, 86
I Enforced Motion with Loads, 87
Large Mass/Spring Method, 87
The Inertial Loads Approach, 89
Input Data, 90
Recommendations, 91
I Lagrange Multiplier Technique, 92
A Comparison of Condensation and Augmentation Solution Methods, 92
Matrix Theory for Dynamic Response, 94
User Interfaces, 99
Example Problem Descriptions, 103
Concluding Remarks, 104
I Axisymmetric Fluids in Tanks, 105
Input Data, 107
Solution Sequences, 110
Hydroelastic Data Processing, 111
Sample Hydroelastic Model, 113
I Virtual Fluid Mass, 117
User Interface, 117
Theoretical Summary, 119
Singularities for Enclosed Volumes, 120
Using Phantom Structural Boundaries, 122
Gravity Effects, 122
Examples, 123
I Coupled Acoustic Analysis, 124
Building the Models, 124
Setting Up MSC.Nastran, 127
Running the Jobs, 129
Advanced Methods, 131
5
Dynamic Solution
Techniques
I Overview, 140
I Review of Dynamic Excitations, 141
Subcases in Dynamic Analysis, 141
Loading Methods, 142
Recommendations, 144
I Complex Eigensolutions, 146
The Two Forms of Eigensolution, 146
The Complex Eigensolution Methods, 147
User Interface, 150
Modeling Techniques, 152
Example Problem: Complex Roots of a Friction Mechanism, 153
I Nonlinear Transient Response Analysis, 161
User Interface, 163
Case Control, 163
Implicit Integration Control: TSTEPNL Data, 164
Iteration Related Output Data, 168
Three-Point Method (NLTRD Module), 169
Two-Point Method (NLTRD2 Module), 170
Adaptive Time Stepping, 170
Quasi-Newton and Line Searches, 171
Restarts, 172
Expedient Iteration Strategies, 174
I Fourier Transform, 176
Theory, 176
Transformation of Loads to the Frequency Domain, 177
Calculation of Frequency Response, 178
Inverse Transformation of the Response to the Time Domain, 178
I Viscoelastic Material Properties, 181
Input Description, 182
I Free Body Techniques, 190
The SUPORT Option, 190
Geometric Shape Methods, 192
I Aeroelastic Solutions, 194
Introduction to Aeroelastic Analysis and Design, 196
Aerodynamic Theories, 196
Generation of Aerodynamic Matrices, 198
Static Aeroelastic Equations of Motion, 198
Flutter Solution Techniques, 199
Generalized Aerodynamic Matrices, 200
The K-Method of Flutter Solution, 201
Dynamic Aeroelastic Analysis, 203
Aeroelastic Transient Response Analysis, 203
Random Response Analysis, 204
I Dynamic Optimization and Design Sensitivity, 205
User Interface, 205
Case Control Commands, 205
Bulk Data Inputs, 205
The DRESP2, DEQATN, DCONSTR and DESOBJ Entries, 207
Example: A Clamped-Free Plate, 207
Example: A Portal Frame with Transient Loading, 209
6
Data Recovery I Overview, 214
I Data Recovery for Modal Methods, 215
Mode Displacement, Vector Method, 215
Matrix Method for Modal Response, 216
I Improving the Results, 217
Mode Acceleration Method, 217
Using Superelements for Data Recovery, 219
I Shock and Response Spectrum Analysis, 221
Theoretical Background, 221
Generating Response Spectra Curves, 223
Peak Response Calculation, 227
Response Spectrum Examples, 229
I Random Analysis with Coupled Excitations, 238
Theory, 239
Example: Random Analysis of a Simple Structure, 241
A
Theory of
Eigenvalue
Extraction
Methods
I Sturm Sequence Theory, 252
I Eigensolution Methods, 253
B
SSSALTER Library
For MSC.Nastran
I Overview, 270
I SSSALTER Library for MSC.Nastran Version 69.1, 271
I Alter to Perform Rigid Body Checks, Kinetic Energy Calculation, and Modal
Effective Weight Calculation, 277
I References, 279
C
Set Notations I Degree-of-Freedom Sets, 282
Degree-of-Freedom Set Definitions, 282
Degree of Freedom Set Bulk Data Entries, 285
I Multipoint Constraints, 286
I Single Point Constraints, 288
I Rigid Body Supports, 289
I Sets for Dynamic Reduction, 291
I Sets for Aerodynamics, 292
I Rules of Sets for Undefined Degrees-of-Freedom, 293
I Output Selection via Set Specification, 295
D
Selected
Bibliography
INDEX MSC.Nastran Advanced Dynamic Analysis Users Guide 301
Preface
I About this Book
I List of MSC.Nastran Books
I Technical Support
I Internet Resources
I Permission to Copy and Distribute MSC Documentation

About this Book
The MSC.Nastran Advanced Dynamic Analysis Users Guide extends many of the topics covered in
the MSC.Nastran Basic Dynamics Users Guide and serves as a starting point for other
dynamics-related topics. Basically, it is a compendium of new material as well as material from
many sources, including internal reports and other existing MSC documentation. The purpose
of this guide is to aid the experienced dynamics user in modeling and solving general problems
in dynamics using MSC.Nastran. Of particular interest are the topics dealing with non-structural
interactions such as control systems or fluids. Other useful topics cover model checkout
procedures and dynamics data recovery options.
Notes to the Reader
This book is not intended to be a stand-alone document. Because this guide has very little
overlap with the MSC.Nastran Basic Dynamics Users Guide or the MSC.Nastran Quick Reference
Guide, it does not cover basic MSC.Nastran dynamics input data and usage. You are encouraged
to keep these other texts handy when using this document.
Given the wide range of subject matter, it was difficult to organize the flow of this guide into a
meaningful sequence. As a compromise, the chosen sequence of topics roughly adheres to the
typical sequence of tasks performed by an engineer who builds a model and performs an
analysis. These five tasks are as follows:
1. Building the model.
2. Reducing the matrices.
3. Applying boundary conditions.
4. Solving for results.
5. Data recovery and postprocessing.
The contents in each chapter corresponding to these steps can be summarized as follows:
Modeling. The modeling chapter covers special modeling inputs used only by the dynamics
solutions. Discussions of the various mass and damping options are followed by instructions on
the use of direct input matrices, transfer functions, and nonlinear loads. A section on model
debugging describes the special checks often employed by dynamics specialists.
Matrices. The matrix formulation chapter focuses on matrix reduction methods and modeling
special effects with matrices. Here the type of analysis determines the matrix reduction options.
The formulations of cyclic symmetry and aeroelastic matrices are also included. Unfortunately,
the topics of superelements and modal synthesis were not available for this edition. For
information on Version 68 and earlier, the reader is referred to the MSC.Nastran Superelement
Users Guide. Version 70 contains many changes and will require a new user guide.
Boundary Conditions. The chapter on boundary conditions describes methods of enforcing
known boundary motions as well as coupled boundaries between structural and fluid models.
Unlike static analysis, enforced motions in dynamics are not processed directly, but must be
input indirectly through the dynamic loads. The other sections in this chapter describe the three
major methods of modeling fluids and their interfaces.
x Preface
Dynamic Solutions. The chapter on dynamic solutions focuses on the advanced dynamic
options for the numerical computations. Options such as complex eigenvalues, nonlinear
transient response, and Fourier transformations are additions to the solutions described in the
MSC.Nastran Basic Dynamics Users Guide. Aeroelastic solutions and dynamic design
optimization are related topics that require special solutions.
Data Recovery. The final chapter deals with dynamic data recovery and postprocessing
options. Several methods are available for recovering modal responses. Spectrum and random
analysis are tools which provide additional information on the expected structural response.
Acknowledgements
I wish to express my appreciation to all of the MSC staff and others who contributed to this book.
New material was contributed by Michael Gockel, William Rodden, and David Bella. Adapted
material from other sources was provided by Erwin Johnson, Ted Rose, Dean Bellinger, and Ken
Blakely. Wendy Webb worked closely with me to refine the content and prepare the book for
publication. My hearty thanks to all.
David N. Herting

List of MSC.Nastran Books
Below is a list of some of the MSC.Nastran documents. You may order any of these documents
from the MSC.Software BooksMart site at www.engineering-e.com.
Installation and Release Guides
Installation and Operations Guide
Release Guide
Reference Books
Quick Reference Guide
DMAP Programmers Guide
Reference Manual
Users Guides
Getting Started
Linear Static Analysis
Basic Dynamic Analysis
Advanced Dynamic Analysis
Design Sensitivity and Optimization
Thermal Analysis
Numerical Methods
Aeroelastic Analysis
Superelement
User Modifiable
Toolkit
x Preface

Technical Support
For help with installing or using an MSC.Software product, contact your local technical support
services. Our technical support provides the following services:
Resolution of installation problems
Advice on specific analysis capabilities
Advice on modeling techniques
Resolution of specific analysis problems (e.g., fatal messages)
Verification of code error.
If you have concerns about an analysis, we suggest that you contact us at an early stage.
You can reach technical support services on the web, by telephone, or e-mail:
Web Go to the MSC.Software website at www.mscsoftware.com, and click on Support. Here, you can
find a wide variety of support resources including application examples, technical application
notes, available training courses, and documentation updates at the MSC.Software Training,
Technical Support, and Documentation web page.
Phone
and
Fax
United States
Telephone: (800) 732-7284
Fax: (714) 784-4343
Frimley, Camberley
Surrey, United Kingdom
Telephone: (44) (1276) 67 10 00
Fax: (44) (1276) 69 11 11
Munich, Germany
Telephone: (49) (89) 43 19 87 0
Fax: (49) (89) 43 61 71 6
Tokyo, Japan
Telephone: (81) (3) 3505 02 66
Fax: (81) (3) 3505 09 14
Rome, Italy
Telephone: (390) (6) 5 91 64 50
Fax: (390) (6) 5 91 25 05
Paris, France
Telephone: (33) (1) 69 36 69 36
Fax: (33) (1) 69 36 45 17
Moscow, Russia
Telephone: (7) (095) 236 6177
Fax: (7) (095) 236 9762
Gouda, The Netherlands
Telephone: (31) (18) 2543700
Fax: (31) (18) 2543707
Madrid, Spain
Telephone: (34) (91) 5560919
Fax: (34) (91) 5567280
x Preface
Email Send a detailed description of the problem to the email address below that corresponds to the
product you are using. You should receive an acknowledgement that your message was
received, followed by an email from one of our Technical Support Engineers.
Training
The MSC Institute of Technology is the world's largest global supplier of
CAD/CAM/CAE/PDM training products and services for the product design, analysis and
manufacturing market. We offer over 100 courses through a global network of education
centers. The Institute is uniquely positioned to optimize your investment in design and
simulation software tools.
Our industry experienced expert staff is available to customize our course offerings to meet your
unique training requirements. For the most effective training, The Institute also offers many of
our courses at our customer's facilities.
The MSC Institute of Technology is located at:
2 MacArthur Place
Santa Ana, CA 92707
Phone: (800) 732-7211
Fax: (714) 784-4028
The Institute maintains state-of-the-art classroom facilities and individual computer graphics
laboratories at training centers throughout the world. All of our courses emphasize hands-on
computer laboratory work to facility skills development.
We specialize in customized training based on our evaluation of your design and simulation
processes, which yields courses that are geared to your business.
In addition to traditional instructor-led classes, we also offer video and DVD courses, interactive
multimedia training, web-based training, and a specialized instructor's program.
Course Information and Registration. For detailed course descriptions, schedule
information, and registration call the Training Specialist at (800) 732-7211 or visit
www.mscsoftware.com.
MSC.Patran Support
MSC.Nastran Support
MSC.Nastran for Windows Support
MSC.visualNastran Desktop 2D Support
MSC.visualNastran Desktop 4D Support
MSC.Abaqus Support
MSC.Dytran Support
MSC.Fatigue Support
MSC.Interactive Physics Support
MSC.Marc Support
MSC.Mvision Support
MSC.SuperForge Support
MSC Institute Course Information
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]

x Preface
Internet Resources
MSC.Software (www.mscsoftware.com)
MSC.Software corporate site with information on the latest events, products and services for the
CAD/CAE/CAM marketplace.
Simulation Center (simulate.engineering-e.com)
Simulate Online. The Simulation Center provides all your simulation, FEA, and other
engineering tools over the Internet.
Engineering-e.com (www.engineering-e.com)
Engineering-e.com is the first virtual marketplace where clients can find engineering expertise,
and engineers can find the goods and services they need to do their job
CATIASOURCE (plm.mscsoftware.com)
Your SOURCE for Total Product Lifecycle Management Solutions.

Close Options Options Close Options Options
Preface xix
Permission to Copy and Distribute MSC Documentation
If you wish to make copies of this documentation for distribution to co-workers, complete this
form and send it to MSC.Software Corporation. MSC will grant written permission if the
following conditions are met:
All copyright notices must be included on all copies.
Copies may be made only for fellow employees.
No copies of this manual, or excerpts thereof, will be given to anyone who is not an
employee of the requesting company.
Please complete and mail to MSC for approval:
Attention: Legal Department
2 MacArthur Place
Santa Ana, CA 92707
Name:_____________________________________________________________
Title: ______________________________________________________________
Company: _________________________________________________________
Address:___________________________________________________________
__________________________________________________________________
Telephone:_________________Email: __________________________________
Signature:______________________________ Date: ______________________
Please do not write below this line.
APPROVED: MSC.Software Corporation
Name:_____________________________________________________________
Title: ______________________________________________________________
Signature:______________________________ Date: ______________________
Close Options Options Close Options Options
Preface xx
_______________________
_______________________
_______________________
Place
Stamp
Here
Attention: Legal Department
2 MacArthur Place
Santa Ana, CA 92707
Fold here
Fold here
CHAPTER
1
Introduction
I Overview
I Purpose of this Guide
I Range of Readers
I Summary

1.1 Overview
MSC support personnel are frequently asked whether MSC.Nastran will solve a particular
unusual problem. The standard answer is of course we can solve this problem.
However, they add that the correct question is, How much time and effort can you afford
to spend on the analysis? They are usually guided by their conception of MSC.Nastran as
a toolkit, with the ability to be molded by the user into a custom system to solve a special
class of structural problems. The limitations of dynamic analysis in MSC.Nastran are
governed more by the conveniences of the code than by its capabilities.
From its original designs, MSC.Nastran has provided a comprehensive set of capabilities
for structural dynamics analysis. Although the designers were skilled in the methods used
by the aerospace industry, the capabilities have been applicable to many other problems,
such as earthquakes, automobiles, nuclear power plants, and many consumer products.
Because of these broad capabilities in the product, it has been difficult to document an
instruction document that explains how to use the system.
3 CHAPTER 1
Introduction
1.2 Purpose of this Guide
The purpose of this document is to guide the user into the recommended procedures for
solving the difficult problems with structural dynamics. Its predecessor is the
MSC.Nastran Basic Dynamics Users Guide, which is recommended for instructions on the
typical dynamics problems and use of the basic input data. This follow-on document
covers many aspects that would not accomodate the scope of the basic guide as well as new
topics related to other MSC.Nastran capabilities (such as aeroelasticity, optimization, or
random analysis). The basic principles also apply to related applications such as cyclic
symmetry.

1.3 Range of Readers
The target audience for this document covers a range of engineers bounded by two skill
levels. On one side we have the knowledgeable MSC.Nastran user who is familiar with
MSC.Nastran and the use of basic dynamics, but needs to learn more about the other
available dynamic techniques, including some of the advanced theory. On the other side
we have the dynamics specialist who is familiar with the advanced methods, or has
experience with other FE codes, but needs some instructions in the input formats and the
procedures. Many readers may only be interested in a few specialized topics and will only
be interested in a few chapters.
Most of the intended readers fit somewhere between these extremes. Because this guide
covers many disjoint topics, it is not intended to be casual reading material. This manual
does not overlap the MSC.Nastran Basic Dynamics Users Guide and is not intended for
beginners in both MSC.Nastran and dynamics solutions.
This guide covers all of the dynamics capabilities not covered completely in the
MSC.Nastran Basic Dynamics Users Guide. This includes advanced structural dynamics
topics such as random analysis as well as related nonstructural dynamics subjects such as
transfer functions and acoustics. The treatment of each topic depends on the availability
of other available documentation.
Topics such as aeroelasticity or nonlinear analysis are well-documented in separate users
guides, and their dynamics interfaces will only be introduced here. On the other hand, this
guide will be the sole source for topics such as hydroelasticity, which will be covered in
detail. Equations are included only when required to clarify an explanation; this guide will
not be a substitute for a theoretical manual. The goal is to provide sufficient detail for a
basic understanding of the use of each topic.
5 CHAPTER 1
Introduction
1.4 Summary
The chapters of this guide are divided into three groups: modeling practices, solution
options, and data recovery processes. The appendices contain reference material such as
eigenvalue theories, alter libraries, and set descriptions.
Dynamic Modeling Options on page 7, System Matrix Formulations on page 61, and
Boundary Conditions and Coupling to Other Disciplines on page 85 cover the
processes of building the matrices that represent the physical problems. The topics include
the finite elements, the special dynamic modeling options, and the available matrix
operations. Of special interest are the fluid boundary options such as acoustics.
Dynamic Solution Techniques on page 139 covers the methods of solving a set of
dynamics matrices. These include complex eigenvalues, transient analysis, frequency
response, and other variations such as aeroelastic flutter. Also included is a brief
discussion of design optimization in dynamics.
Data Recovery on page 213 covers the special aspects of data recovery in dynamics
analysis. Special topics include random vibrations and response spectrum analyses.

CHAPTER
2
Dynamic Modeling Options
I Overview
I Mass Modeling
I Modeling Damping Effects
I DMIGs, Extra Points, and Transfer Functions
I Nonlinear Load Functions (NOLINi)
I Debugging Dynamic Models

2.1 Overview
In general, the practice of building MSC.Nastran models for dynamic analysis is similar to
building statics or normal modes models. However, in dynamics the higher costs of
running detailed large order models usually results in a different approach. For instance,
a frequency response analysis using SOL 108 may require solutions at more than 100
frequencies. Each frequency will require an unsymmetric matrix solution using complex
variables, which will cost four to ten times more than a single static solution. If memory is
sufficient, the total result is a running time larger by a factor of 400 to 1000. If the required
memory (four times more is needed) is insufficient, the solution will spill, and even longer
run times will result. In addition, the database and scratch file space requirements will be
much larger and the size of the output data becomes unwieldy. The result is that many
users who simply take fine mesh static models directly into a dynamic analysis find this
process is very costly and may give up completely.
Strategies for Dynamic Analysis. The two major approaches to avoid the problems
described above are to use the matrix reduction methods available in MSC.Nastran or to
simply build another dynamic model with fewer points and elements. The matrix reduction
methods use sophisticated approximations to reduce the cost of the dynamic solutions.
However, they add complexity to the system and may also cost processing time overhead.
Instead of making decisions about the element mesh size, the user must consider the
frequency band and loading characteristics of the actual structure.
Although building a separate dynamic model will require extra work, it has several
advantages. The requirements for precise stress results for statics are usually not as
important for dynamics, thereby allowing the use of larger elements. The requirements for
accurate dynamic results are not as dependent on small elements since loads tend to be
distributed over wider areas by the inertial and damping effects. The actual accuracy
criteria for dynamic models is related to wavelength size relative to the mesh size. Also, in
a smaller dynamic model the job of interpreting outputs with plotters and other methods
becomes much easier.
It should also be noted that the process of changing finite element meshes is becoming
easier with modern geometry-based modeling systems and automated mesh generation.
Plan the Analysis. The following steps are recommended before starting a dynamic
analysis:
1. Estimate the frequency range of the structure. The first mode frequency can
usually be estimated by a simple equivalent beam calculation. A normal modes
analysis is highly recommended for dynamics regardless of the approach.
2. Consider the type and frequency content of the loads. Localized high frequency
sources such as brake squeal require different approaches from low frequency
distributed loads such as wind forces on a bridge.
9 CHAPTER 2
3. Use expected wavelengths to estimate required mesh size. A wavelength can be
estimated from the frequency and sound speed. Six elements per wave for
approximately 10% accuracy is a general rule-of-thumb. Also, this method may
be used to evaluate the quality of the calculated eigenvectors.
4. Always use SETs for output requests as a general practice. The use of Case
Control requests such as STRESS = ALL for a large transient analysis could
possibly exceed the size limits of your postprocessor.
5. Estimate the computer resources (CPU time, database size, output requirements)
for large problems before production runs are made. A good method is to run the
model with only a few time steps or frequencies, and examine the major time and
size messages in the F04 output file. The actual costs can then be extrapolated
with reasonable accuracy.
The basic modeling, constraints, and loading functions for dynamics are described in the
MSC.Nastran Basic Dynamics Users Guide. The following sections discuss some of the more
difficult and obscure topics.

2.2 Mass Modeling
Basic Definitions of Mass. The mass matrix in MSC.Nastran may contain much more
information than simple structural mass and inertia terms. In fact, it may contain control
system terms, fluid compressibility, or electromagnetic capacitance. One basic definition
is that any term which contributes to the generalized kinetic energy must create a
coefficient in the mass matrix. Another definition is that any generalized force, ,
proportional to an acceleration term produces a mass term , i.e.,
Eq. 2-1
where each component of the acceleration vector represents a generalized
degree-of-freedom.
The mass matrix is required for nearly all dynamic solution sequences. It is also used for
generating gravity and centrifugal loads for static solutions. Inertia relief solutions require
the mass matrix to balance the unbalanced forces on a free body. All solutions may
calculate the total mass and center of gravity (CG) information for printout.
Note that mass is not required for heat transfer dynamics.
Mass Data Input Options
MSC.Nastran provides the following means to specify the mass properties of the finite
element model:
1. The density (mass per unit volume) of the structural materials, which comprises
the finite element (RHO on the MATi Bulk Data entries).
2. Nonstructural mass per unit length of line elements or mass per unit area of
surface elements (NSM on the property Bulk Data entry). Examples of this
feature are coatings and thermal insulating materials.
3. Concentrated mass and inertia terms at a grid point via a CONM2 Bulk Data
entry. The provisions of the CONM2 entry are the mass, the offset of the center
of mass from the grid point, and the moments and products of inertia about the
center of mass. As an option, the center of mass may be measured from the origin
of the basic coordinate system rather than as an offset from the grid point.
4. A full symmetric matrix of mass coefficients at a grid point via the CONM1
Bulk Data entry.
5. Mass coupling between any two degrees-of-freedom via the CMASSi (i = 1,2,3,4)
Bulk Data entries. The form of the relationship is
Eq. 2-2
F
u
M
F { } M [ ] u
{ } =
u
{ }
6 6
f
1
f
2
" #
$ %
& '
M M
M M
u
1
u
2
" #
$ %
& '
=
1 CHAPTER 2
where and are the inertia forces acting at degrees-of-freedom 1 and 2,
respectively, and is the mass coefficient, specified on the CMASSi entry (or on
the PMASS entry if i = 2 or 4). In most applications, the second degree-of-freedom
is not specified. In this case, the entry generates the inertia force , and
is added to the mass matrix in the diagonal position corresponding to . An
important application of the CMASSi entry occurs in the recommended method
for specifying enforced motion at grid points (see Enforced Motion with Loads
on page 87).
6. Transfer functions defined on the TF Bulk Data entry may contribute terms to the
mass matrix.
7. Direct Matrix inputs may be added to the mass matrix via DMIG Bulk Data and
M2GG = or M2PP = Case Control commands.
Three Related Parameters. The bulk data input also includes three parameters which
relate to the specification of mass properties. They are:
1. PARAM,WTMASS,V1 Specifies a factor by which to multiply the mass of the
structure to obtain dimensionally correct mass. For example, if the ft-lb-sec
system is used, and the mass input unit is pounds, then V1 = 1/32.174 = .031081.
This factor operates on all mass terms (except those defined on DMIG entries;
PARAM,CM2,V1 may be used for these instead).
2. PARAM,COUPMASS,1 Requests that the coupled mass option be used rather
than the lumped mass option for all elements which support the coupled mass
option. In the lumped mass option, the distributed mass of an element is simply
divided up and the parts are located at the connected grid points. In the coupled
mass option, the distributed mass is replaced by a nondiagonal mass matrix
which couples the connected grid points. The latter method is sometimes more
accurate, but it uses more computer time. See the MSC.Nastran Reference Manual
for descriptions of the elements.
3. PARAM,GRDPNT,V1 Causes the Grid Point Weight Generator to be executed.
The value of the parameter, V1, is an integer which identifies a grid point at which
the rigid body mass properties of the structure will be computed. See the Getting
Started with MSC.Nastran Users Guide for an example of the output format which
is provided. This capability can be used as a check on mass and geometric input
data.
Coupled Mass Matrix Terms
A common difficulty for finite element beginners is to comprehend the generation of
off-diagonal or coupled terms in the mass matrix. These are caused by the following
actions.
f
1
f
2
M
f
1
Mu
1
=
M u
1

1. With PARAM,COUPMASS, the finite elements themselves generate coupled
terms to represent the distributed element mass more accurately than simple
lumped masses on the grid points. The total kinetic energy of the element is
represented by shape functions, which in turn, are coupled functions of
displacement and rotation.
2. The internal matrix operations in the solution sequences produce coupled mass
terms. MPCs and RBE-type elements cause the mass to be redistributed from the
dependent degrees-of-freedom to the remaining independent points. The ASET,
OMIT, and Superelement reduction operations may automatically fill the entire
mass matrix. These operations will maintain exact total mass and CG properties
and provide a good approximation to the inertia effects of the eliminated points,
but they can be costly.
3. Direct user inputs such as CMASSi scalar elements and CONMi mass elements
may produce off-diagonal mass terms. Other inputs that cause coupling are
direct matrix inputs (DMIG), and transfer functions (TF).
An example of a coupled scalar mass is illustrated in the example below:
A spring, , and two masses, and , are attached to a structural model. However,
when modeling fluids or other special connections, the user may wish to use the difference
in displacements, , instead of as the primary degree-of-freedom.
We may find the equivalent coupled mass matrix by the use of energy methods. The
potential energy, , and the kinetic energy, , for this part of the system are:
Eq. 2-3
From basic energy principles, the resulting equilibrium equations are
Eq. 2-4
The mass matrix for the two degrees-of-freedom becomes
Eq. 2-5
k
Structure
P
1
m
1
u
1
u
m
2
P
2
u
2
k m
1
m
2
u u
1
u
2
= u
2
V T
V ku
2
2 =
T m
1
u
1
2
m
2
u
1
u
( )
2
+ =
m
1
u
1
m
2
u
1
u
( ) + + P
1
P
2
+ =
ku m
2
u
1
u
( ) P
2
=
and
m [ ]
m
1
m
2
+ m
2
m
2
m
2
=
1 CHAPTER 2
where are the active degrees-of-freedom. Two CMASSi elements may be
used; one will be coupled to two points. Note that the spring, , is now connected to
only.
An alternate method would be to use the MPC constraints to produce the same effect.
Simply include all three DOFs in the model and use conventional lumped masses and a
connected spring. Choose as the MPC equation, and the reduced mass and
stiffness matrices will be identical to the system above.
Mass Effects in Moving Coordinate Systems
In most cases the motions of the grid points in MSC.Nastran are measured in a set of fixed
coordinate systems. Even large displacement nonlinear analysis uses displacements
measured in fixed directions. However, there are several cases in which it is more
convenient to use a moving coordinate system. Examples follow:
Spinning Bodies
If the entire structure is spinning at a constant angular velocity, both static centrifugal
forces and coupled dynamic inertial terms may be needed. An example is a high speed
turbine wheel that exhibits gyroscopic stability problems.
For dynamic analysis of rotating bodies, a special solution will be provided in the DMAP
alter library for the generation of extra matrix terms caused by the rotating coordinates.
For a body rotation defined by the angular velocity vector, , and for a location vector of
a point, , the absolute velocity vector of the point, , in fixed coordinates is
Eq. 2-6
The acceleration vector is
Eq. 2-7
The first term on the right-hand side of Eq. 2-7 is the static centrifugal force; the second
term is the centripetal stiffness; the third term is the Coriolis force; the last term, , is the
relative acceleration vector.
Another term that is calculated for spinning bodies is the so-called differential stiffness
matrix. It is proportional to the steady centrifugal preloads in the elements. These are the
terms that would stiffen a string if a weight on the string were swung in a circular motion.
These terms are important because they are approximately the same magnitude as the
Coriolis and centripetal stiffness terms defined in Eq. 2-7. (Centrifugal stiffness and
differential stiffness terms are of the same magnitude.)
The basic matrix equation for the forces in the moving system is
Eq. 2-8
The special MSC.Nastran process assembles the terms in Eq. 2-7 into matrices, that when
substituted into Eq. 2-8, result in
u u
1
u , =
k u
u
2
u
1
u + 0 =
r V
V r u + ( ) u
+ =

A r ( ) u ( ) 2 u
( ) u
+ + + =

u
M [ ] A { } B [ ] V { } K [ ] u { } + + P { } =

Eq. 2-9
where:
The system defined by Eq. 2-9 may be solved with a transient analysis, a frequency
response, or a complex eigenvalue calculation. The complex eigenvalues may be obtained
for a series of spin rates to determine the critical angular speeds. At each angular velocity
the complex roots are obtained in the form . Unstable conditions usually occur
when , where is a positive integer. By definition, the system is unstable if is a
positive number.
Inertia Relief
If a free body is accelerating due to constant unbalanced loads, the inertia relief solution
provides the ability to obtain static deflections relative to a set of reference points attached
to the moving coordinate system. An example is an airplane in a steady turn or
accelerating dive. Although this capability is a static solution, it is obtained from the
dynamics theory.
The basic matrix equation for the inertia relief method is
Eq. 2-10
where are displacements relative to the moving system and are the steady
accelerations to be determined from the mass and loads. If is a matrix whose columns
define the rigid body motions of the structure, then for a free body,
Eq. 2-11
where is called the rigid body tranformation matrix. However, since the full-sized
vector, ,

is a rigid body motion, it may be defined in terms of accelerations at a set of
reference coordinates, , by the equation
Eq. 2-12
Combining Eq. 2-12 into Eq. 2-11 and eliminating , we obtain
Eq. 2-13
where the total mass matrix for the reference coordinates is
Eq. 2-14
= generates the velocity-dependent Coriolis forces. Note that is not symmetric!
= the centripetal stiffness matrix
= the differential stiffness matrix
M [ ] u
{ } B B
c
+ [ ] u
{ } K K
c
K
d
+ + [ ] u { } + + P t ( ) { } =
B
c
[ ] B
c
K
c
[ ]
K
d
[ ]
p
n

n
i
n
=
n
P P
n
K [ ] u { } P { } M [ ] a
o
{ } =
u { } a
o
{ }
D [ ]
D [ ]
T
K [ ] u { } 0 { } D [ ]
T
P { } D [ ]
T
M [ ] a
o
{ } = =
D [ ]
a
o
{ }
a
r
{ }
a
o
{ } D [ ] a
r
{ } =
a
r
{ }
a
o
{ } D [ ] m [ ]
1
D [ ]
T
P { } =
m [ ] D [ ]
T
M [ ] D { } =
1 CHAPTER 2
The resulting set of equations defined in Eq. 2-10 may now be arbitrarily constrained since
the total load is balanced by the inertia forces.
Two different methods are used in MSC.Nastran to calculate the rigid body matrix . In
SOL 24 the SUPORT Bulk Data entry define the reference degrees-of-freedom. Any
number of degrees-of-freedom that provide a nonredundant set of supports may be
used. This option allows partially free bodies and extra mechanisms. In SOLs 61 and 101
the grid point geometry is used (PARAM,GRDPNT) to define six rigid body displacement
vectors. In these cases, only six free motions are allowed.
Base Excitations
One of several methods to solve problems with enforced motion is to constrain the point of
motion and solve the problem in the accelerating system. This method is related to inertia
relief but uses entirely different inputs. It is easy to use for earthquake analysis of
buildings, in which the base is accelerating uniformly.
If a structure is attached to a semi-rigid base that causes a known stress-free motion,
,

the total structural motion,
,
is
Eq. 2-15
where are displacements relative to the base motion. If the structure is not constrained
elsewhere, we may assume that the base displacements produce no force and
Eq. 2-16
If we move the known base motion to the right hand side it looks almost identical to
a gravity load:
Eq. 2-17
Acceleration Method. If the acceleration is uniform over the structure, a time-dependent
gravity load has exactly the same form as the last term of Eq. 2-17, and the GRAV input
load data may be used.
An example data input for a dynamic acceleration load is shown below. For more details
on these input formats, see the MSC.Nastran Basic Dynamics Users Guide on page v.
Case Control
LOADSET = 20 $ Requests LSEQ Id. 20 Process
DLOAD = 200 $ Requests Dynamic Load #200
Bulk Data
GRAV, 386, , 386.4, -
1.0
$ Defines Gravity Load in -x direction
D [ ]
u
r
u
0
{ }
u
A
{ }
u
A
{ } u
0
{ } u
g
{ } + =
u
g
{ }
u
0
{ }
M [ ] u
g
u
0
+ { } K [ ] u
g
{ } + P
g
{ } =
u
0
{ }
M [ ] u
g
{ } K [ ] u
g
{ } + P
g
{ } M [ ] u
0
t ( ) { } =

The net result is that the time-dependent inertia loads are applied to all points on the
structure in the x direction in proportion to the time-dependent function specified on the
TLOADi or RLOADi Bulk Data entry. The base points should be constrained and the
displacements will be calculated relative to the moving base. However, note that the
accelerations output from the solution will also be relative to the base motion and should
be corrected before being compared with accelerometer data.
A simpler alternative to base motion is the large mass approach, described in Enforced
Motion with Loads on page 87.
LSEQ, 20, 201, 386 $ Assembles GRAV load vector
$ Added to DAREA Id. 201
TLOAD1, 200, 201, etc. $ Dynamic Load using DAREA Id. 201
Case Control
1 CHAPTER 2
2.3 Modeling Damping Effects
The physical causes of damping in dynamic analysis are any processes which dissipate
energy or reduce the structural response through internal friction. Furthermore, the
internal velocities or displacements cause reactive damping forces which are irreversible
and nonconservative. Examples are mechanical devices such as shock absorbers, the
internal hysteresis that occurs in materials such as rubber, friction in joints, and other
nonlinear effects such as plastic strains in metals.
The four types of damping in MSC.Nastran are viscous, structural, modal, and nonlinear.
The basic input formats and applications for the damping coefficients in MSC.Nastran are
explained in Mass Input on page 17 of the MSC.Nastran Basic Dynamics Users Guide. The
following discussions are directed to advanced applications and the special problems of
damping.
Viscous Damping
The primary method for modeling viscous damping is through the CVISC and CDAMPi
Bulk Data entries. These produce forces which are linearly proportional to the velocities of
the connected grid points. Another source of viscous damping is structural damping that
must be converted to viscous damping in the transient response solutions. The
MSC.Nastran Basic Dynamics Users Guide covers this topic. Some additional details are
included below.
Structural Damping
Structural damping is intended to simulate the effects of linear material energy loss
proportional to the strains. In other words, this method approximates effects similar to
hysteresis. It is specified in the material definition input, MATi, and on the parameter, .
Note that in the frequency response and complex eigenvalue solutions, the structural
damping produces imaginary numbers in the complex stiffness matrix. In the transient
solutions, the matrix terms are converted to equivalent viscous damping; i.e., the forces
will be proportional to the derivative of strain with respect to time.
The physical basis for the phenomenon of imaginary stiffness is explained by Figure 2-1.
Shown is the force due to a stiffness matrix term, , plotted against the displacement of the
point. As the displacements cycle sinusoidally, the imaginary stiffness, , causes a phase
lag in the force response resulting in an elliptical path. The area enclosed by the curve is
equal to the dissipated energy.
G
K
G

Figure 2-1 Equivalent Hysteresis Path for Imaginary Stiffness Damping
The basic definition for the steady-state actual displacements, in terms of the complex
displacements is
Eq. 2-18
where is the actual real displacement, is normally a complex variable that we will set
to a real number, is the steady state frequency, and . If a structural
damping coefficient, , is added to the stiffness matrix, the resulting complex forces are in
the following form:
Eq. 2-19
The real part of the forces are
Eq. 2-20
The incremental work done over a period of time is
Eq. 2-21
where, from Eq. 2-18
Eq. 2-22
Combining Eq. 2-20 through Eq. 2-22 and integrating over a full cycle, results in the work
Real
Force
Real
Displacement
u
0
0 =
Ku
0
GKu
0
90 =
180 =
u
r
t ( ) Re u
0
( )e
i t
( ) =
u
r
u
0
e
i x
x ( ) i x ( ) sin + cos =
G
F ( ) 1 i G + ( )Ku
0
e
i t
=
F
r
Re F ( ) Ku
0
t G t sin cos ( ) = =
dW F
r
du
r
=
du
r
u
0
t sin ( )dt =
1 CHAPTER 2
Eq. 2-23
Evaluating the integral, we obtain work loss per cycle
Eq. 2-24
Note that the elastic energy terms average zero over the interval, but the energy dissipated
by the structural damping exists. The area inside the curve in Figure 2-1 is equivalent to
the damping energy loss, .
This damping method is a reasonable linear approximation to the classical hysteresis effect.
It may also be used to approximate other similar cyclic energy losses such as the effects of
loose joints and fasteners. (Hint: construct a similar trajectory curve and estimate the area.)
The main disadvantage of structural damping is that complex numbers must be converted
to real numbers in transient analysis; i.e., structural damping is converted to linear viscous
damping. This is good only when the response is dominated by a single known frequency.
Modal Damping
For the modal solution formulations, a special damping input is provided in addition to
the other damping terms described above. It is used primarily for efficiency and when test
results or contract specifications provide damping factors. These terms are applied only to
the uncoupled modal equations. When coupling effects are absent, the method avoids
expensive matrix calculations.
A second reason for using modal damping is that modal testing may provide accurate
damping inputs on a mode-by-mode basis. These damping factors may be converted to a
frequency-dependent table, TABDMP. If the MSC.Nastran normal mode frequencies are
close and the damping factors are small. The third reasonwhen damping factors are
specified by a third partyis frequently meant to keep the analyst more conservative. By
restricting the solutions to artificially low damping, this will force the staff to solve their
design problems for the worst case.
Modal damping is available only for the modal solutions method. The matrix terms
generated for modal analysis, and added to any direct matrix inputs, are
Eq. 2-25
In this equation, is the generalized coordinate of the i-th mode.
In accordance with modal definitions, the modal viscous damping coefficient may be
expressed as
Eq. 2-26
Ku
0
2
t t G t sin cos ( ) sin t d
0
2
(
=
W GKu
0
2
=
W
m
i
i
b
i
i
k
i
i
+ + P
i
=
i
b
i
b
i
g
i
i
m
i
=

where is the undamped vibration frequency (equal to ), expressed in radians per
unit time. The dimensionless input coefficient, , is equal to twice the critical damping
ratio of the mode. It is evaluated by linear interpolation of a user-specified function of
frequency, .
The physical meaning of modal damping is somewhat clouded. Its effect is viscous
damping, but it is proportional to the stiffness matrix, and varies with modal frequency.
The damping effects are distributed to the structure depending on the energy distribution
in each mode shape.
Another aspect of modal damping to remember is that the damping coefficient, , is
constant for each mode. If the mode is forcibly excited at a different frequency the initial
damping factor still applies. In other words, the damping at any particular frequency is a
function of several modal damping factors. This may cause unexpected results for
frequencies where a mode is not dominant.
The user specifies pairs of values on a TABDMP1 Bulk Data entry, which is selected by
the SDAMPING Case Control command. There are also provisions for expressing
damping as a fraction of critical damping ( ) or amplification quality factor ( ). If
accurate test results are available, the user can specify different damping coefficients,
obtained from modal tests, for the different frequency ranges. The user can ensure that the
desired damping has been obtained for a set of modes whose frequencies are known from
a previous run by providing the desired pairs on the TABDMP1 entry.
The complete damping matrix for modal transient analysis, , is
Eq. 2-27
where is a diagonal matrix whose elements are given by Eq. 2-26, is the matrix of
eigenvectors, and is the matrix of nonmodal damping terms. Note that the damping
effects which enter may well be duplicated by the effects included in so that, in
general, the user should be careful when using both forms simultaneously.
Also, the matrix in Eq. 2-27, or the equivalent complex stiffness matrix, is
generally coupled so that the efficient uncoupled methods of analysis cannot be used when
is present.
It is the accepted practice in many industries to express viscous damping as a fraction of
critical damping. Critical damping is defined as the value at which the homogenous
solution of Eq. 2-25 transitions from a damped sinusoid to a nonoscillating, decaying
exponential. A solution for the modal equation shows the value of critical damping, , to
be
Eq. 2-28
The fraction of critical damping, , is calculated from the equation
Note: In frequency response and complex eigenvalue analysis, the complete damping
matrix is similar, except that structural damping is treated as a complex stiffness
matrix.
i
k
i
m
i
g
i
g f ( )
b
f g ,
C C
c
Q
f,g
B [ ]
B [ ] b
i
[ ] [ ]
T
B
v
[ ] [ ] + =
b
i
[ ] [ ]
B
v
[ ]
B
v
[ ] b
i
[ ]
[ ]
T
B
v
[ ] [ ]
B
v
[ ]
C
c
C
c
2 k
i
m
i
=
i
2 CHAPTER 2
Eq. 2-29
Another form of damping specification is the amplification quality factor, , with the
definition
Eq. 2-30
All three forms of damping specification are available for modal damping.
An illustration of the comparison between modal damping and an equivalent function of
structural damping is shown in Figure 2-2, showing the different options for modal
damping tables. With unit modal masses, a constant modal damping table actually
corresponds to a viscous damping that increases with modal frequency. This method falls
between the two extremes (constant viscous damping and equivalent structural damping)
and serves as a general purpose compromise.
Figure 2-2 Damping Coefficients Versus Modal Frequency
Nonlinear Damping
This discussion explains the limitations in damping caused by the nonlinear transient
solution algorithms (SOL 129). Frequency response and complex eigenvalue solutions are
not available in nonlinear analysis. For specialized nonlinear dynamics topics, see
Nonlinear Transient Response Analysis on page 161. For basic information on the
MSC.Nastran nonlinear solutions, see the MSC.Nastran Handbook for Nonlinear Analysis or
the MSC.Nastran Reference Manual.
Finite elements may be either linear or nonlinear in the nonlinear formulations, and the
damping effects will be treated differently for each type. The basic rules for nonlinear
transient analysis are as follows:
i
b
i
C
c
------
b
i
2 k
i
m
i
------------------ - =
Q
i
Q
i
1
2
i
--------
1
g
i
---- = =
Modal Frequency
Modal Damping
Constant Viscous Damping
Structural Damping
b
i
b G
i
2
( )
b m
i
g
i
i
= ( )
i
b c = ( )

1. Plastic yield in the nonlinear materials automatically absorbs energy when the
structure follows a loading and unloading cycle. This is an actual hysteresis effect
that produces an accurate form of damping. However, note that strain rate effects
are not calculated directly. Strain rate effects must be modeled with structural
damping parameters, which are converted internally to viscous damping.
2. Viscous damping elements are always linear and will participate as constant
matrix terms.
3. Structural damping, defined with parameters and material bulk data inputs (the
GE field on the MATi entries), is recognized for both linear and nonlinear
elements. The damping matrix terms are calculated for the current material
stiffness moduli and geometry. Note, however, that the tangent matrices are only
updated periodically. The actual damping on nonlinear elements is
unpredictable and can change answers for different runs on the same problem
depending on the convergence rate and iteration strategy. It is recommended that
the matrix update strategy forces an update on the tangent matrix at every time
step.
4. A modal formulation (and therefore modal damping) is not available in a
nonlinear solution.
5. Superelements may be used to reduce the size of a nonlinear problem by
separating the linear elements into an upstream component. Component modal
synthesis may then be used to maintain accuracy. Structural damping is allowed
on superelements.
2 CHAPTER 2
2.4 DMIGs, Extra Points, and Transfer Functions
In dynamics modeling, we frequently need to include special nonfinite element effects
such as mechanical devices, servomechanisms, smart structures, and matrices from
external structures. These effects can usually be included as extra terms in the system
matrices along with extra degrees-of-freedom in the generalized displacement vectors.
The major difference between extra points (EPOINT data) and normal scalar points
(SPOINT data) is that the extra points are added to the system after the finite element
matrix assembly and real eigenvalue solution. Therefore, structural elements, constraints,
and static loads may not be connected to EPOINTs. Also, as with scalar points, they are
processed only in the residual superelement in SE formulations.
The EPOINT Set
Using standard set notation (see Set Notations on page 281), the extra point set merges
with the structural degrees-of-freedom according to the following diagram:
For Direct Formulations:
In direct solutions, the system matrices are formed by merging structural points and
e-points into the p-set. The actual solution matrices are assembled and added together at
the reduced, d-set level.
The set logic for modal formulations is as follows:
In modal solutions the system matrices are transformed to the modal coordinates and
merged with the extra point degrees-of-freedom. The modal coordinates include free body
modes, , and flexible modes, . Note that the coupling between the structural motions
and the extra points is only approximated by the truncated modal solution.
m
.
.
etc.
q
r
c
b
e
a
d
g
p
e
h
0

f

Modal Transformations
When extra points are included in the modal formulation, a transformation is needed
between the physical displacements and the modal coordinates. The displacements, ,
are defined by the transformation
Eq. 2-31
where:
and are the real eigenvectors. The loads and direct input matrices are transformed
consistently with the displacements. For the applied loads, ,
Eq. 2-32
Direct input matrices, K2PP, are transformed similarly in the form
Eq. 2-33
Mass and damping matrices are also transformed similarly.
Note that the extra points remain in the modal formulation. These are useful for modeling
transfer functions, initial conditions, and simple nonlinear functions.
Because these matrix terms couple the modal coordinates, this option is more expensive
than solving the simple uncoupled modal equations of motion. The coupled solution
requires the same type of matrix operations as the direct method.
Direct Matrix Inputs
Since direct matrix input (DMIG) data is covered in the MSC.Nastran Linear Static Analysis
Users Guide and the MSC.Nastran Basic Dynamics Users Guide, this section will be brief.
For dynamics modeling, the direct input matrices are defined for the p-set of
degrees-of-freedom for mass, damping, and stiffness matrices. The actual values are
defined by the user and may represent any type of linear solution, including fluids,
electrical circuits, and external structures. However, some practical recommendations are
as follows:
=
Note: DMIG matrices are selected by the K2PP =, B2PP =, and the M2PP = Case Control
commands.
u
p
u
p
{ }
u
g
u
e
-----
" #
$ %
& '
ph
[ ]

u
e " #
$ %
& '
= =
ph
[ ]

g
0
0 I
g
P
P
h
{ }
ph
[ ]
T
P
p
{ } =
K
hh
2
[ ]
ph
[ ]
T
K
pp
2
[ ]
ph
[ ] =
2 CHAPTER 2
1. Use the double field bulk data format (DMIG*) if more than three significant
digits of the input values are desired. Note that these solutions are sensitive to
the precision of the matrix terms. Use DMAP modules, INPUTi, for full double-
precision input.
2. Use the automatic selection of single- or double-precision data for the system.
The program sets the size depending on the word size of the computer.
3. Use the symmetric matrix option if possible. The unsymmetric flag will force all
operations into the unsymmetric mode, costing time and storage space.
An example problem that uses DMIG data for generating friction forces is given in
Complex Eigensolutions on page 146.
Transfer Functions
The MSC.Nastran transfer functions (TF inputs) are actually a convenient method for
generating special unsymmetric matrix input. When used with extra points they can
define second order operators with one output and multiple inputs. The basic equation
defining individual TF input is
Eq. 2-34
Here is the dependent degree-of-freedom, are the selected input degrees-of-freedom,
and the coefficients and are user-specified.
Internally, these coefficients are simply added to a single row in the matrix equation:
Eq. 2-35
The terms are added to the matrices in the following positions:
1. All terms in a single function are added to the row of the matrices corresponding
to .
2. is added to the diagonal term of . is added to the matrix, and is
added to the matrix on the diagonals, respectively.
3. For each independent point , , , , and are added to the column
corresponding to , and the row corresponding to , of the stiffness, damping,
and mass matrices, respectively.
Limitations
If the TF terms are the only occupants of the row of the matrix equation, the dynamic
solution will include the basic equation. However, no internal checks are made for
additional terms in the TF row in the matrices or loads. To avoid conflicts, the following
rules apply:
B
0
B
1
p B
2
p
2
+ + ( )u
d
i
A
oi
A
1i
p A
2i
p
2
+ + ( )u
i )
+ 0 =
u
d
u
i
A B
Mp
2
Bp K + + [ ] u { } P { } =
u
d
B
0
K [ ] B
1
B [ ] B
2
M [ ]
u
i
A
0i
A
1i
A
2i
u
i
u
d
u
d

1. should be the dependent variable for only one TF.
2. should not be a component of a structural grid point unless the transfer
function defines a load (see the discussion of transfer functions below).
3. External excitation signals, dynamic loads (DLOAD), and nonlinear functions
(NOLINi) may be applied to the point.
4. Unlike the multipoint constraints, the TF equations do not conserve energy. The
matrix terms are not symmetric, and no reciprocity forces are generated.
Pickups and Transducers
Connections to the structure may be needed for both inputs and outputs from the transfer
functions. To specify inputs, the structural displacements, velocities, or accelerations may
be referenced on the and input fields. The may be an extra point defining the
pickup voltage. This signal may then be processed by a series of transfer functions
(including feedback loops) until a servo generates a load on the structure.
It is recommended that the transfer function results applied to the structure be represented
by a force or moment. An enforced displacement, velocity, or acceleration is difficult to
model within the limitations described above.
To generate a force, on a structural point, , with a transfer function,
define a TF function of the form
Eq. 2-36
The coefficients must be zero since the structural point is already connected to the
finite elements. This transfer function is equivalent to adding a positive force on the
right-hand side of the matrix equation.
If the force or moment is acting on an actuator with a feedback loop, include the feedback
as another transfer function.
Higher-Order Transfer Functions
In many cases the control system includes a polynomial transfer function of order greater
than two and cannot be modeled directly. Since the TF inputs are limited to second-order
polynomials, the larger polynomials must be subdivided into several TF inputs with
intermediate extra points. For instance, if a point, , is defined by the polynomial
Eq. 2-37
this may be subdivided into two equations suitable for the TF format:
Eq. 2-38
u
d
u
d
u
d
A
i
u
i
u
d
F
j
u
j
F
j
A
0
A
1
p A
2
p
2
+ + ( )u
e
=
0 ( )u
j
A
0
A
1
p A
2
p
2
+ + ( )u
e
0 =
B
j
u
j
u
a
A Bp Cp
2
Dp
3
Ep
4
+ + + + ( )u
b
G p ( )u
a
=
A Bp Cp
2
+ + ( )u
b
p
2
u
e1
G p ( )u
a
+ 0 =
2 CHAPTER 2
and
Eq. 2-39
Here, is an intermediate extra point which is coupled to the degree-of-freedom in
both equations.
Note that this method may produce a singular mass matrix that causes problems for certain
complex eigenvalue options. This would occur in the example above if . A solution
is discussed in Complex Eigensolutions on page 146.
Alternate Method. If the polynomial is available in factored form, i.e.,
, a series of TF transfer functions can be used as illustrated in the
following sketch.
If is the dependent degree-of-freedom, the TF equations will be in the following form:
For polynomials in the denominator of a transfer function, use a feedback loop or coupled
equations as shown in the previous example. See the example below for a realistic
application.
Example
As an example of the general approach, consider the control system shown in Figure 2-3,
which might represent a simplified model of an optical device. It contains many of the
components found in control systems including attitude and rate sensors, signal
conditioners, and a nonlinear mechanical actuator with local feedback. The structure is
represented in Figure 2-3 by the displacement components, which are sensed by the
control system, or to which loads are applied.
The transfer functions (TFs) of the control system are listed in Table 2-1. The first five
transfer functions give the outputs of the devices labeled 1 to 5 in Figure 2-3 in terms of
their inputs. TF 6 defines the force on the structure, at point , which is a function of the
input to the mechanical actuator.
In the analysis, an extra point is assigned to each of the new variables, , ..., . The
coefficients of the transfer functions, expressed by TFs 1 to 6, are listed in Table 2-2. Note
the coefficient of 0.0 for the structural degree-of-freedom, , on TF 6. The existing
structure provides the stiffness and mass for this displacement.
u
e1
Dp Ep
2
+ ( )u
b
0 =
u
e1
u
b
E 0 =
G
1
p ( ) G
2
p ( ) G
3
p ( )
u
a
G
1
G
2
G
3
u
e1
u
e2
u
b
u
b
u
e1
G
1
u
a
0 =
u
e2
G
2
u
e1
0 =
u
b
G
3
u
e2
0 =
u
11
u
4
u
10
u
11

Figure 2-3 Example Control System
Table 2-1 Equations of Example Control System
Equation TF Number
1
2
3
Mechanical
Force
Structure
Input
Signal
Filter
Attitude
Sensor
Rate
Gyros
Actuator
Displacement
Feedback
Signal
Conditioner
4
3
2
1
6
5
u
6
u
3
u
8
u
11
u
2
u
1
u
4
u
5
t ( )
u
7
u
4
1
1
1
p +
------------------- Apu
2
Bpu
3
+ ( ) =
u
5
1
1
2
p +
------------------- Ru
1
=
u
6
1
1
3
p +
------------------- u
10
=
2 CHAPTER 2
Input Data File
The actual structure is a beam-like optical pointing device pivoting on the center. The data
file for a complex eigenvalue analysis is shown in Listing 2-1. The attitude sensor ( ),
the first rate gyro ( ), and a lumped mass are located on the right end (GRID 1). The
second rate gyro is attached to the next point (GRID 3). The actuator is connected to the
left end (GRID 11) in the direction. The units for the mechanical model are millimeters,
kilograms, seconds, milliNewtons, and radians. The units for the control system variables
(EPOINTs 4,5,6,7,8,10) are volts.
Listing 2-1 Data Listing for Optical Device Control System
4
5
6
Table 2-2 Transfer Function Table
Transfer
Function
No.
1 1 A
B
2 1 R
3 1 1
4
5 1 G
6 0.0 k
+k
Table 2-1 Equations of Example Control System (continued)
Equation TF Number
u
7
A
0
A
1
p +
B
0
B
1
p B
2
p
2
+ +
--------------------------------------------- Cu
4
Du
5
Eu
6
+ + ( ) =
u
8
Gu
11
1
4
p +
------------------- =
0 ( )u
11
k u
7
u
8
( ) =
u
d
b
0
b
1
b
2
u
i
a
0
a
1
a
2
u
4

1
u
2
u
3
u
5

2
u
1
u
6

3
u
10
u
7
B
0
B
1
B
2
u
4
u
5
u
6
CA
0
DA
0
EA
0
CA
1
DA
1
EA
1
u
8

4
u
11
u
11
u
7
u
8
u
5

z
=
u
2
u
y
=
y

ID ADUG,TFANDE
DIAG 8
SOL 107
TIME 5
CEND
TITLE = CONTROL SYSTEM EXAMPLE FOR ADUG
SUBTITLE = TRANSFER FUNCTIONS AND EXTRA POINTS
LABEL = COMPLEX MODES
SPC = 10
TFL = 6
CMETHOD = 200
SDISP=ALL
BEGIN BULK
$
EIGC,200,INV,MAX,,,,,,+EIG1
+EIG1,0.0,-1.0,0.0,1200.,20.0,10
$
$ STRUCTURE IS A BOX BEAM PIVOTING AT THE CENTER
GRDSET,,,,,,,1345
GRID,1,,500.0
GRID,3,,250.0
GRID,14,,0.0
GRID,15,,-250.0
GRID,11,,-500.0
CBEAM,1,1,1,3,,,1.0
CBEAM,2,1,3,14,,,1.0
CBEAM,3,1,14,15,,,1.0
CBEAM,4,1,15,11,,,1.0
PBEAM,1,1,1000.,125+6,125+6,,250.+6
MAT1,1,69.0+6,,0.3,5.0-6
$ OPTICAL DEVICE ON THE END
CONM2,6,1,,10.0,,,,,+CNM2
+CNM2,8.0+5,,4.0+5,,,4.0+5
$ PIVOT ON THE CENTER
SPC,10,14,12
$ EXTRA POINTS ARE VOLTAGES
EPOINT,4,5,6,7,8,10
$ TRANSFER FUNCTIONS IN ORDER
$ RATE GYROS PICK UP VELOCITIES
TF,6,4,,1.0,0.015,,,,+TF101
+TF101,1,2,,-0.25,,,,,+TF102
+TF102,3,2,,-0.5
$ ATTITUDE SENSOR MEASURES ROTATION RZ
TF,6,5,,1.0,0.02,,,,+TF201
+TF201,1,6,-1.2
$ INPUT SIGNAL, E10, IS FILTERED
TF,6,6,,1.0,0.001,,,,+TF301
+TF301,10,,-1.0
$ SIGNAL CONDITIONER COMBINES THE VOLTAGES
TF,6,7,,39.48,8.885-2,1.0-4,,,+TF401
+TF401,4,,-20.0,-0.4,,,,,+TF402
+TF402,5,,-200.0,-4.0,,,,,+TF403
+TF403,6,,100.,-20.0
$ DISPLACEMENT FEEDBACK TO THE ACTUATOR
TF,6,8,,1.0,0.005,,,,+TF501
+TF501,11,2,-1.0
$ ACTUATOR FORCES ARE ADDED TO ROW OF STRUCTURAL MATRICES
TF,6,11,2,0.0,,,,,+TF601
+TF601,7,,-1.0+5,,,,,,+TF602
+TF602,8,,1.0+5
$ TEMPORARILY GROUND POINT 10 WITH A DIAGONAL TERM
TF,6,10,,1.0
ENDDATA
3 CHAPTER 2
2.5 Nonlinear Load Functions (NOLINi)
The nonlinear load functions (NOLINi) are available in all of the transient solutions for the
purpose of generating direct forces from simple displacement and velocity functions. This
is a mature capability in MSC.Nastran that preceded the development of the material and
geometric nonlinear capabilities. It was intended to provide a direct method for modeling
certain mechanisms and special effects at the local level where only a few
degrees-of-freedom are coupled. It is not intended for general nonlinear analysis since the
logic cannot detect geometric changes and has no means of using element stresses or forces.
The main applications for the NOLINi functions are for local devices such as contact
problems, joints, and nonlinear dampers. For each nonlinear force component, the
dependencies with displacement and velocity degrees-of-freedom are explicitly defined on
one or more bulk data entries.
The limitations on the use of these functions are caused primarily by their simplicity. These
limitations are:
1. They have no memory or other path dependencies. Specifically, they provide no
direct means for including data from previous states.
2. Each input function only applies loads to a single degree-of-freedom, which
requires a great deal of effort to describe complex models.
3. The system treats these functions as simple forces rather than finite elements.
Therefore, the lack of a tangent matrix results in potential stability problems.
4. In the linear solutions, all degrees-of-freedom in the NOLIN functions must be in
the dynamic solution set.
The input data is simply a case control request: NONLINEAR, with one or more NOLINi
Bulk Data entries to define the set. The four basic options to define a scalar nonlinear force,
, in terms of the degrees-of-freedom, , are summarized in the table below.
The variables, , may be displacement or velocity components of grid, scalar, or extra
points in the solution set. In the DIRECT TRANSIENT solutions, the connected
degrees-of-freedom, and , etc., must remain in the solution set, . In the modal
transient solutions, only extra points are available for use by the NOLINi entries. In
nonlinear transient solutions all degrees-of-freedom are available, but unfortunately, extra
points are not supported.
Option Function Comments
NOLIN1 is a TABLEDi input
NOLIN2 Product of two variables
NOLIN3
is an input exponent
NOLIN4
Same except for negative
N
i
u
j
N
i
SF u
j
( ) = F u
j
( )
N
i
Su
j
u
k
=
N
i
S u
j
( )
A
, u
j
0 > = A
N
i
S u
j
( )
A
, u
j
0 < = u
u
u
i
u
j
u
d

Note that these functions mimic the basic nonlinear function generators used in passive
analog computers (MSC.Softwares original business). They may be added together and
combined with other functions such as MPCs and scalar elements to handle a variety of
problems.
Theory for the Standard Transient Solutions
The basic linear transient solution integrates the matrix equation
Eq. 2-40
where are the nonlinear forces which are dependent on variable displacements, ,
and velocities, , of the unknowns. Note that if these functions were used to replace a
stiffness term, the proper definition of would be .
There are several options in MSC.Nastran for transient integration. For the basic purposes
of this introduction, we will discuss the simplest form, the three-point method which
performs a step-by-step calculation. At time step , the solution is and the step size
is . Dropping the brackets, the averaged values of displacement, , velocity, ,
acceleration, , and load, , in terms of the three discrete steps are as follows:
Eq. 2-41
Eq. 2-42
Eq. 2-43
Eq. 2-44
In the solution, the vectors at step are obtained by substituting Eq. 2-2 through Eq. 2-5
in Eq. 2-40. This method is actually a variation of the Newmark-Beta method and is
guaranteed to be stable if . For consistency and stability, it would be desirable to
have the vector also in this form. Ideally, the nonlinear loads would be consistent with
the the linear displacements and loads, i.e.,
Eq. 2-45
However, is not available and can only be approximated by
extrapolating the equation:
Eq. 2-46
Substituting this approximation into Eq. 2-45 we obtain the term used in MSC.Nastran:
Eq. 2-47
In other words, calculating the nonlinear function at the center step is a valid
approximation to the value averaged over three steps. Unfortunately, when is rapidly
changing, will lag behind and create unstable oscillations. The results of an unstable
M [ ] a { } B [ ] v { } K [ ] u { } + + P { } N { } + =
N { } u { }
v { }
N N Ku =
t
n
u
n
{ }
h u v
a P
u u
n 1 +
1 2 ( )u
n
u
n 1
+ + =
v u
n 1 +
u
n 1
{ } 2h =
a u
n 1 +
2u
n
u
n 1 +
+ ( ) h
2
=
P P
n 1 +
1 2 ( )P
n
P
n
+ + =
n 1 +
0.25 >
N
N N
n 1 +
1 2 ( )N
n
N
n 1
+ + =
N
n 1 +
N u
n 1 +
, v
n 1 +
( ) =
N
n 1 +
2N
n
N
n 1

N N
n
u
N
3 CHAPTER 2
integration usually grow with a period of two or three time steps per cycle. An alternate
option for averaging nonlinear loads is available with DIAG 10, although it is not
recommended.
Example of a Nonlinear Spring
An example of an unstable system can be illustrated with a single variable example. Attach
a nonlinear spring function to a mass as shown in the sketch below.
For small motions, if we define the function as an equivalent spring with , the
integration equation for is
Eq. 2-48
For stability analysis, we will assume a uniform growth rate, , where and
. Substituting into Eq. 2-48 we obtain a quadratic equation for :
Eq. 2-49
If the magnitude of is greater than 1, the solution will potentially diverge ( may be
complex). It is easy to show that if the system will have a real root , which is
an unstable oscillation that changes sign every step.
In other words, the stability limit for the nonlinear stiffness for this case is
Eq. 2-50
The cure for divergence is to either reduce the time step, , add linear stiffness to replace
some of the nonlinear portion, or add enough mass, , in parallel to the nonlinear function
to satisfy the criteria for every possible value of . Scalar mass elements CMASSi may be
added to couple two scalar points, in parallel with the nonlinear spring, without affecting
the total mass of the structure.
Velocity-Dependent Nonlinear Forces
When a NOLINi function references a velocity of a grid, scalar, or extra point component
as an input, the nonlinear force must be calculated from the existing displacements.
However, for consistency with the linear damping terms in Eq. 2-40, the nonlinear force
should be:
Eq. 2-51
M
u
N u ( )
N ku =
u
n 1 +
M u
n 1 +
2u
n
u
n 1
+ ( ) h
2
N ku
n
= =
r u
n
ru
n 1
=
u
n 1 +
ru
n
= r
r
2
2
kh
2
2M
--------- 1
* +
, -
r 1 + + 0 =
r r
kh
2
4M > r 1 <
N
u
------ k
4M
h
2
-------- =
h
M
k
N
v
B u, v ( ) u
n 1 +
u
n 1
( ) 2h =

However, the current displacements have not been calculated at this stage, so we
again use the assumption of Eq. 2-46 and obtain the approximation:
Eq. 2-52
The stability of velocity-dependent NOLINi systems is very marginal for most
applications. Replacing the nonlinear spring ( ) in the example problem above with a
nonlinear damper, , we obtain the following system equation:
Eq. 2-53
For stability analysis we may substitute and as in the stiffness example above, to
obtain a quadratic equation:
Eq. 2-54
Unstable roots ( ) will occur when
Eq. 2-55
Again, the problem will show up as a spurious diverging oscillation. Divergence may be
cured by decreasing the time step size, , by changing part of the nonlinear function, , into
a linear damper, or by adding mass in parallel with the nonlinear function. SOL 129 is
better suited than SOL 109 for unstable or badly conditioned problems.
Nonlinear Transient Solution Sequences
The general nonlinear transient solutions described in Nonlinear Transient Response
Analysis on page 161 will also process the NOLINi functions. They provide an advantage
in that they will give the same stability as a linear solution, and will control the diverging
solutions. These methods make an attempt to calculate, for time step , the
displacements and velocities by iterating on Eq. 2-40 through Eq. 2-45 in a loop. This will
provide more consistency with the linear terms. When the advanced methods such as line
search and quasi-Newton options are activated, the adverse effects of the nonlinear forces
are corrected effectively, but at the cost of additional solution steps.
However, the nonlinear methods may also attempt to change the tangent stiffness matrix
when errors become too large in the solution search. Unfortunately, the NOLINi functions
only provide forces on the right side and do not contribute their changes to the matrices.
Not only will results not be improved, cost will be increased as well. Furthermore, since
new self-adaptive method modifies the time step size based on the current values of the
critical nonlinear matrix terms, the NOLINi terms will be ignored and the process may still
diverge.
In summary, the nonlinear transient solutions may be adapted for severe cases of NOLINi-
caused instabilities but will need careful changes to the default control parameters on the
TSTEPNL inputs.
u
n 1 +
N
v
B u, v ( ) u
n
u
n 1
( ) h N u
n
u
n 1
( ) h [ ]
k
b
M u
n 1 +
2u
n
u
n 1
+ ( ) h
2
N b u
n
u
n 1
( ) h = =
ru
n
r
2
u
n
r
2 bh
M
------ 2
* +
, -
r 1
bh
M
------
* +
, -
+ + 0 =
r 1 >
b 2M h >
h b
n 1 +
3 CHAPTER 2
Recommendations for Using NOLIN1 Input
The following suggestions may be useful if problems occur because of nonlinear forces:
1. Use the nonlinear solution (SOL 129) for potentially better error controls and
stability. This approach may use more time and disk storage space and require
additional effort to select the proper control parameters. We recommend starting
with the following approximate TSTEPNL Bulk Data parameters:
Method = TSTEP
KSTEP > NDT
MAXITER = 2
MAXLS = 6
MAXDIV = 10
LSTOL = 0.1
2. Avoid functions that will overwhelm the linear elements attached to the same
points. See the discussions on stability above. Most nonlinear functions can be
defined as the sum of linear elements and a nonlinear correction using NOLINi
inputs. If the slope of the nonlinear correction function is smaller in magnitude
than the linear coefficients, the problem will be stable. Also, added mass in
parallel with the function always helps.
3. Never set up nonlinear functions in series, i.e., using the output of one NOLINi as
an input to a second NOLINi function. Each stage in effect causes a time delay of
one or more time steps for the resulting forces. If the results are a corrective force,
they will act much like a control system with a bad feedback loop. At a critical
frequency the corrective forces will be applied with a delay that causes them to
have the wrong sign.
4. The forces, , must have the same sign and direction as an externally applied load
on the structure. Note that this will have the opposite sign as the forces applied
to the GRID points by the finite elements. (The force for a spring element will
have a negative value for a positive displacement.) This frequently causes some
difficulty in debugging the input data.
5. Use MPC or TF data to simplify the NOLINi inputs. A single nonlinear spring in
three dimensions connecting two points could require up to nonlinear
functions. Alternatively, one MPC equation could be used to define the strain as
an added scalar point, and one nonlinear function would define the generalized
force (force times length) on the scalar point. Because the MPC constraints are
consistent, the force will automatically be distributed to all six
degrees-of-freedom.
6. Plan ahead to retain the nonlinear points in the solution set. Remember that in
SOLs 109 and 112 the NOLINi functions may not reference any dependent or
eliminated degree-of-freedom. Note that in a modal analysis, this leaves only
extra points.
7. Use the full nonlinear solution (SOL129) for small to medium-sized problems.
N
N
6 6 36 =

Application Example Coupling of Rotating Structures
An advanced use for the NOLINi functions is the connection between a rotating structure
and a flexible supporting structure. The example below shows how to connect two grid
points, with one rotating at a constant rate relative to the other. This method could be used
to analyze sudden transients in rotating machinery. However, cost considerations would
limit the analysis to a finite number of cycles.
As shown in the sketch, GRID 101 ( ) is attached to the rotating structure. The rate of
rotation is known, but the actual angle is a function of time. GRID 201 ( ) represents the
fixed bearing point coincident with point 101. It would typically be defined with an RBE3
as the average motion of the points on the bearing race.
The definition of the connection, as shown in the sketch is:
A simple MPC equation may be used to connect the axial coordinates . The rotations
may have similar definitions but are not shown. A similar equation defines the forces on
the fixed point

in terms of the rotating forces .
For the actual connection, we will use the Lagrange Multiplier technique (see Lagrange
Multiplier Technique on page 92 for details) by defining another GRID which defines
the constraint forces on point 101. The resulting matrix partition equation for the three
points is
Eq. 2-56
where:
u
r
u
f
fixed
101
201
rotating
u
f 1
t
u
r2
u
r1
u
r1
u
f 1
t ( ) cos u
f 2
t ( ) sin + =
u
r2
u
f 1
t ( ) sin u
f 2
t ( ) cos + =
u
3
N
f
Q
r
Q
r
M
f f
0 0
0 m
rr
0
0 0 0
u
r " #
. .
$ %
. .
& '
K
f f
0 0
0 K
rr
I
0 I 0
U
f
u
r
Q
r
" #
. .
$ %
. .
& '
+
P
f
N
f
+
P
r
N
q
" #
. .
$ %
. .
& '
=
3 CHAPTER 2
In order to generate the sine and cosine terms, we will add two scalar points and connect
unit masses and springs . Using initial conditions, these are the sine and cosine
functions defined as scalar displacements. The following input data will solve the
connection:
In Case Control
IC = 100 $ For sine and cosine
K2PP = RCOUP $ For matrix terms
NONLINEAR=10 $ For NOLIN2
In Bulk Data
$ 50 radian/sec oscillators ( Note: Dont use PARAM,G)
SPOINT,2001,2002 $ sine and cosine
CMASS4,2001,1.0,2001
CMASS4,2002,1.0,2002
CELAS4,2003,2500.0,2001
CELAS4,2004,2500.0,2002
TIC,100,2001,,,50.0 $ sine
TIC,100,2002,,1.0 $ cosine
$ Lagrange Multiplier Grid and unit Matrix Terms.
GRID,3001, , , , , ,3456
DMIG,RCOUP,0,6,1,2
DMIG,RCOUP,101,1,,3001,1,1.0
DMIG,RCOUP,101,2,,3001,2,1.0
$ Note that other half is generated when using the SYM option
$ Nonlinear Functions = Products of Two Dof., Q
r
and sine/cosine
NOLIN2,10,201,1,1.0,3001,1,2002
NOLIN2,10,201,1,-1.0,3001,2,2001
NOLIN2,10,201,2,1.0,3001,1,2001
NOLIN2,10,201,2,1.0,3001,2,2002
$
$ Products of U
1
and sine/cosine
NOLIN2,10,3001,1,1.0,201,1,2002
NOLIN2,10,3001,1,1.0,201,2,2001
NOLIN2,10,3001,2,-1.0,201,1,2001
NOLIN2,10,3001,2,1.0,201,2,2002
$ end for one pair of points
Any type of structure may be connected to the axis points. However, for best stability place
all of the mass at the axis on the active side, GRID 201, where the nonlinear force is applied.
For inertia effects such as centripetal and Coriolis forces, the rotating body may require
matrix inputs described in the section on moving coordinates. In order to keep these terms
from affecting the fixed structure, superelement partitioning is recommended.
=
=
N
f
{ }
P
r1
t ( ) cos P
r2
t ( ) sin
P
r1
t ( ) sin P
r2
t ( ) cos +
" #
$ %
& '
N
Q
{ }
u
f 1
t ( ) cos u
r2
t ( ) sin
u
f 1
t ( ) sin u
r2
t ( ) cos +
" #
$ %
& '
k
2
=

2.6 Debugging Dynamic Models
This section outlines a procedure for systematically checking and documenting a finite
element model. Currently many models are generated by one organization and
transmitted to other groups for subsequent analysis. Since these models have had various
levels of analytical validation and documentation, checkout procedures are needed to
ensure that the models will be consistent and mathematically well-conditioned. These
procedures, however, are not substitutes for the independent verification phase of
analysis. They are intended to remove modeling errors in the design process, rather than
during the test updating phase, which occurs well after the hardware is built.
Modeling Hints
There is no single checklist that will ensure a complete check of a comprehensive finite
element model. Also, there is no substitute for actual test correlation with the model, nor
is there a substitute for the analysts engineering interpretation of the output and ones
intuition. A results prediction, which is determining gross results before the analysis, can
be used to good advantage. Simple load paths or natural frequencies of equivalent simple
beam/mass systems, etc., can be used to remove redundancies and predict the results. In
fact, this must be accomplished to some degree to size the model initially. This will also
provide baseline data, and the effects of finer modeling will then be known.
Relative to preparation of the pilot model, the following are a few suggestions for
eliminating or reducing modeling problems.
Start construction of a simple pilot model in which the user should:
Use beams and plates instead of solids.
Use RBE2s and RBE3s where they will simplify.
Simplify modeling offsets and local modeling details.
Ignore minor discontinuities such as holes and fillets.
Further refinement after this initial modeling should yield acceptable changes in the
results.
Do not rely on bending capability of thin plates and long, thin axial members to
render the model kinematically stable.
Make an initial run with membrane-only properties and pinned ended bars, and
check for irregularities.
Avoid use of AUTOSPC in the final model.
For shells, use the parameters K6ROT or SNORM.
The post-analysis assessment should include a check of the physical significance of the
loads and of the load path. Offsets whose moments are not properly accounted for may
overly weaken a very stiff load path. Also, large moments in relatively weak bending
members or plates may indicate modeling problems.
3 CHAPTER 2
Stress analysis should be performed at the detailed part level with the loads from the
model. The use of element stresses directly from the output of the model requires detailed
review in most cases. In fact, model properties may be intentionally different from the
actual hardware to obtain correct load distributions, and to match test data or dynamic
characteristics. Effective thicknesses or reduced bending properties may have been used
to reflect panel cutouts or partial beam and fixity. In this event, the finite element model
loads should be used with the actual drawing or as-built dimensions for detail stress
analysis. This piece-part assessment ensures a check and balance of the finite element
model and the stress distributions visualized and treated by the element selection. Also,
the source of the components of stress are known, that is, whether the predominant stress
component is due to bending or axial loads. Load transformation matrices are useful for
isolating critical design conditions but are not necessarily a sufficient basis for computing
the margin of safety.
One area in which an underestimation of load could occur is the local response of small
masses during a dynamic analysis. These should be addressed in the detailed stress
analysis with both the model predictions and an alternate loading such as a specified
loading condition. For the model to give correct loads for the local response of a mass, one
needs all of the following:
Mass must be represented by enough points to characterize the energy of the
critical local mode (a single-point mass may not be sufficient).
Mass must be supported by proper elastic elements to represent the local mode
(RBE2 or RBE3 may not be sufficient).
Mass must be in the ASET.
Model and all analysis (input spectra, etc.) must be carried beyond this local
critical mode (as far as frequency is concerned).
Testing Dynamic Models
Once the finite element model is completed and all documentation (such as model
schematics (road maps) as well as material and geometric property calculations) are
updated to the final model version, the following series of tests should be performed in
order to validate the model. It is recommended that these tests be run on the model and
subsystem models during the development stages as well.
Geometry Plots
The MSC.Nastran plotting package MSC.Patran, or another preprocessor graphics package
should be used to obtain visual images of the finite element model from many views in
such a way as to provide a clear representation of each element in at least one view and to
verify overall geometry and placement of elements. Figure 2-4 shows the Galileo
spacecraft finite element model and Figure 2-5 shows the Wide Field/Planetary Camera
finite element model. A shrink option should be used if possible to make sure all elements
are present (see Figure 2-6). This is particularly helpful when bars or beams are used to
model stringers along the edges of plate elements. Discontinuities show up only when the
shrink option is invoked.

Figure 2-4 Galileo Finite Element Model
4 CHAPTER 2
Figure 2-5 Galileo Wide Field Planetary Camera Finite Element Model

Figure 2-6 Shrink Option
Checkout Run RF24D32
A checkout run using the rigid format alter RF24D32 should be used to check connectivity
and duplicate element numbering. Any duplicate numbering should be corrected so that
errors with plot sets will not occur. The connectivity table does not include MPCs or rigid
elements (RBE2, RBE3, or RBAR). The checkout run also supplies all grid locations in basic
coordinates and lists the lengths, areas, and volumes of all elements. Listing 2-2 gives a
listing of the parameters used and Listing 2-3 through Listing 2-10 give examples of the
output. These items can be useful for finding anomalies not apparent in plots.
a) Shears and Bars, No Shrink
b) Bars Regular Size, Shears Shrunk
c) Bars and Shears Both Shrunk
4 CHAPTER 2
Listing 2-2 Checkout Run
$ BEGINNING OF RF ALTER 24$32
$
$ CHECK INPUT DATA
$ R I G I D F O R M A T 24
$
$ CHECK CONTROL INPUT
$ CASE CONTROL DECK THAT WILL BE USED IN PRODUCTION RUN.
$
$ UNDEFORMED PLOTS MAY BE REQUESTED INCLUDING PROPERTY IDS
$ AND SPC POINTS
$
$ BULK DATA INPUT
$ COMPLETE BULK DATA DECK TO BE SUBMITTED FOR PRODUCTION RUN
$ THE FOLLOWING PARAMETERS MUST BE SET TO THE INDICATED VALUES TO OBTAIN
$ OUTPUT FOR-
$ 1. GPL . SET PARAMETER GPL TO GPL
$ 2. EQEXIN - SET PARAMETER EQEXIN TO EQEXIN
$ 3. GPDT - SET PARAMETER GPDT TO GPDT
$ 4. CSTM . SET PARAMETER CSTM TO CSTM
$ 5. BGPDT - SET PARAMETER BGPDT TO BGPDT
$ 6. MGG - SET PARAMETER PRTMGG TO +1
$ 7. ETT - SET PARAMETER GPTT TO GPTT
$ 8. GPECT - SET PARAMETER GPECT TO +1 TO OBTAIN LIST OF ELEMENTS
$ CONNECTED TO GRIDS.
$ 9. PROUT -- SET TO POSITIVE VALUE TO OBTAIN SORTED LIST OF
$ ELEMENT IDS.
$ 10. EST - SET TO POSITIVE VALUE TO OBTAIN LENGTHS, AREAS,
$ OR VOLUMES OF ELEMENTS.
$ 11. PG -- SET PRTPG=+1 TO PRINT THE LOAD VECTOR MATRIX
$ 12. A SUMMATION OF LOAD IN BASIC WILL AUTOMATICALLY BE PRINTED
$
$ THE ITEMS CHECKED INCLUDE - GRID CONNECTION AND PROPERTY CARDS, MASS
$ PROPERTIES OF STRUCTURE, LOADS, TEMPERATURES AND CONSTRAINTS
$ AS WELL AS SOME GEOMETRY AND MATERIAL PROPERTIES
$ FOR EACH SUBCASE IN ALL BOUNDARY CONDITIONS.
$
$ THE ITEMS NOT CHECKED INCLUDE - GEOMETRY, MATERIAL PROPERTIES AND
$ GRID POINT SINGULARITIES
$
Output from the Grid Point Weight Generator is recommended. Set PARAM,GRDPNT
point to a positive integer. The value of the integer indicates the grid point to be used as
the reference point. If the integer is zero or an undefined grid, the origin of the basic
coordinate system is the reference point.

Listing 2-3 Formatted List of Table Data Block EQEXIN (Record 1)
Grid IDs as found in the Bulk Data are listed in numerical increasing order from
left to right (four per row) and are listed under the column heading External Grid
or Scalar ID.
The corresponding internal ID number is given to the immediate right of the grid
number under the column heading Internal Number.
In the example above, grid point number 25 has internal sort number 14.
The first column marked External Sort ID may be ignored (it simply increments
by four because there are four External Grid/Internal Numbers per line).
RF24D32 provides similar information in the following tables: GPL Record 1,
GPL Record 2, and EQEXIN Record 2. The table shown above was chosen
because its format is easier to explain.
This output is obtained by using RF Alter RF24D32 and
PARAM,EQEXIN,EQEXIN.
Listing 2-4 Formatted List of Table Data Block GPDT (Record 1)
This output is obtained by using RF Alter RF24D32 and PARAM,GPDT,GPDT.
4 CHAPTER 2
Listing 2-5 Formatted List of Table Data Block CSTM (Record 1)
This output is obtained by using RF Alter RF24D32 and PARAM,CSTM,CSTM.
Listing 2-6 Formatted List of Table Data Block BGPDT (Record 1)
This output is obtained by using RF Alter RF24D32 and PARAM,BGPDT,BGPDT.

Listing 2-7 Element Type Count
This output is obtained by using RF Alter RF24D32 and PARAM,PROUT,+1.
Elements not accounted for include MPCs, RBARs, RBE1s, RBE2s, RBE3s.
Listing 2-8 Elements Listed in Numerical Order
This output is obtained by using RF Alter RF24D32 and PARAM,PROUT,+1.
Listing 2-9 Grid Point-Element Connection List
This output is obtained by using RF Alter RF24D32 and PARAM,GPECT,+1.
Elements not accounted for include MPCs, RBARs, RBE1s, RBE2s, RBE3s.
4 CHAPTER 2
Listing 2-10 Element Length, Area, or Volume
This is special output and is obtained by using RF Alter RF24D32 and
PARAM,EST,+1.
Mass Distribution
PARAM,GRDPNT uses the Grid Point Weight Generator (GPWG), which gives the mass,
the CG, the moments of inertia, and the principal moments of inertia and their direction
cosines (see Table 2-3). Full use should be made of this diagnostic tool to correlate the
model with existing hardware or mass properties calculations. The GPWG uses only the
weight properties and geometry to calculate mass properties. The resultant mass
properties are also called the rigid body mass properties. Note that PARAM,WTMASS
does not affect the GPWG outputit is factored out of the printed results.

Table 2-3 PARAM,GRDPNT Weight Inertia Matrix
Static Loading
Static loading, such as a simple gravity load, can be helpful in checking out various
properties of finite element models. Displacements, element forces, and support reactions
(SPC forces) derived from 1G loading conditions provide a first check on mass, stiffness,
and determinacy of supports. Weight and CG can be calculated from SPC forces, which
should also be compared to any applied loads or weight. Load paths can also be assessed
using the element forces. Epsilon, Max Ratio, and SPC forces (at grids other than legitimate
boundary conditions) describe the overall health of the stiffness matrix. See Table 2-4 and
Table 2-5 for example printout data.
Table 2-4 Epsilon, Strain Energy and Max Ratio From Statics Run
4 CHAPTER 2
Table 2-5 SPC Forces From Statics Run
Allowable values for these quantities are as follows:
The 1G cases also provide a rough approximation of the frequency of the first mode. This
approximation can be accomplished by using the displacement ( ) at the CG in the
equation .
A 1G static load case can easily be obtained through the GRAV entry. This is preferred over
the inertia relief type method.
Epsilon*
* Epsilon is machine dependent. The above data is for CDC 64-bitword. Other machines
should give smaller numbers ( and ). Mechanisms or symmetry condi-
tions may require reevaluation of SPC force limits.
(large model)
(small model)
Max Ratio
SPC Forces (at internal points)
(model weight)
SPC Moments (at internal points) (model weight) (unit length)
1.0 10
8
1.0 10
11
1.0 10
6

1.0 10
9

1.0 10
+5

1.0 10
5

1.0 10
3

D
FN
1
2
------ G D

Static forces and moments can be applied to generate displacements, element, and SPC
forces. The magnitude and point of application of the forces should be representative of
typical structural loading, thereby allowing the analyst a good feel for the size of
displacements and forces as in the 1G cases. It is also helpful to use element strain energy
and grid point force balance with these runs. For information on the details of requesting
these capabilities, see Case Control Commands, ESE on page 261 and GPFORCE on
page 282 of the MSC.Nastran Quick Reference Guide.
Free Body Equilibrium Check
Using Solution 101 and the set of CHECKA alters in the SSSALTER Library, an equilibrium
check can be run to further validate the stiffness matrix. This check calculates the strain
energy resulting from unit translations and rotations.
All SUPORTs and SPCs should be removed. Include the following PARAMs in the Bulk
Data: AUTOSPC,YES; GPL; SEQOUT,1. The output from the AUTOSPC processor should
be reviewed carefully (see Table 2-6 for an example of AUTOSPC). This table will include
all degrees-of-freedom that have no stiffness. Each degree-of-freedom should be checked
and, if singular, it should be SPCd; if it is not intended to be singular, then the model must
be corrected.
Table 2-6 Grid Point Singularity Table
The KRBF matrix (see Table 2-7) is printed out and is a measure of the force required for
the rigid body displacements. All elements should be small, e.g.,
Diagonal translations <
Diagonal rotation <
Off-diagonal terms <
Table 2-7 KRBF Matrix
1.0 10
2
2.0 10
2
2.0 10
2
5 CHAPTER 2
Forces due to translations (circled) should be less than 1.E2.
Moments due to rotations (rectangle) should be less than 2.0E+2.
Off diagonal terms should be less than 2.0E+2.
Also printed in the matrix KRBFN, the forces at the grid points normalized to a maximum
of 1.0 (see Table 2-8). As seen from Table 2-8, this check easily identifies which grid points
and DOFs are causing problems.
Table 2-8 KRBFN Matrix
Rigid Body Check
If the model is primarily used for dynamic analysis, the CHECKA DMAP alter performs
an equilibrium check in Solution 103. In this case, the unit translations and rotations are
considered to be the rigid body displacements of the structure in output coordinates about
the SUPORT point. Strain energy, which should be very small, is calculated and printed
out.
The strain energy, the Epsilons, and the Max Ratio should also be checked at this point, as
they were for the static case (see Table 2-9).

Table 2-9 Epsilon, Strain Energy, and Max Ratio from Rigid Body Modes Run
The SUPORT entry is required for epsilon and strain energy.
This option requires a SUPORT entry. The DOF supported must form a determinate
interface or high strain energy will result. The model displacements printed represent the
displacements caused by moving a support DOF one unit while holding the other support
DOFs fixed. The model displacements should be checked for unit value (see Table 2-10).
Table 2-10 Rigid Body Mode Shapes
Eigenvalues and cycles are set equal to zero by the DMAP alter package.
Please note that these alters assume max normalization on the EIGR entry.
Thermal Test Case
As a further check on connectivity and the stiffness matrix, an isothermal expansion test
case can be run with a statically determinate interface. This is done on SOL 101 with a
TEMPD for the constant temperature load. All of the coefficients of expansion should be
5 CHAPTER 2
set to the same value. This check will not be rigid if rigid elements or bar offsets are
present. Rigid elements will not expand and may generate distortion forces and stresses
unless the appropriate degrees-of-freedom are released.
Modal Analysis
There are several diagnostic tools that can be used to further assess the integrity of the
model. These tools (effective mass, strain energy, kinetic energy, deformed plots) are
outlined in the following paragraphs.
The two major means of reducing the number of dynamic degrees-of-freedom in the modal
analysis are outlined below.
A major concern in dynamic analysis is the choice of an appropriate ASET when using the
Guyan reduction. The quality of the solution depends upon the reduced mass matrix
formed by this ASET. Will it retain sufficient mass in correct distribution to adequately
predict mode shapes and frequencies?
The results of a GDR run can be used to select an ASET for use in future processing if
desired (see the flow chart in Figure 2-7). See the description for CHECKA in SSSALTER
Library For MSC.Nastran on page 269. A proven rule-of-thumb for ASET selection is to
include all DOFs that have more than 2% of the system KE for all major modes (determined
by comparing EFMASS to the system weight) and more than 5% for the other modes. One
must be careful in the case of assemblies with a fine mesh; although the whole assembly
may be moving in a mode, there may not be any individual DOFs with greater than 2% of
the system kinetic energy. This is usually evident when a mode has a sizeable EFMASS,
and either no terms with a large kinetic energy or the terms with noticeable kinetic energy
do not account for the EFMASS of the mode. If this is the case, then there is no substitute
for common sense determination of DOFs that describe the subsystem motion.

Figure 2-7 ASET Selection Flow Chart
Note also that this removes the rigid body mode check from the run, since the SUPORT
entry must be removed (unless the run is a free-free run).
Run GRD
Loop on Grids
Mechanical
No
No
No
Yes
Yes
Yes
Intuitive
In ASET
Does the
ASET Describe the
Motion of the Major
Subsystems
Run Dynamics
With ASET
Do Modes
Compare to
LANCZOS
Select DOF That
Describe the Desired
Motion
Done
2% in Major Modes
5% in Minor Modes
K.E > 2% or 5%
5 CHAPTER 2
MER MREMER MR EFMASS
An important diagnostic tool is the elastic-rigid coupling matrix (MER), a triple product of
the elastic mode shape matrix, the ASET mass matrix, and the rigid body displacement
matrix with respect to the interface. This matrix gives the square root of the effective
mass in each retained mode so that the product of each element of this matrix will itself
give the effective mass contained in that mode in the associated direction. These values are
in EFMASS. This matrix can be used to determine which modes are energetic in terms of
interface loading. Table 2-11 shows a sample MER matrix taken from Reference 3. The
elements of this matrix have been squared and units converted to give recognizable weight
units.
N 6

Table 2-11 Galileo Spacecraft Baseline Model Modal Weight and Inertia
(Lb, Lb-in
2
)
5 CHAPTER 2
The product of MER with itself transposed (MREMER) gives the total effective mass
retained using the ASET DOF number of modes. The diagonal of MREMER is compared
with that of the rigid mass matrix (MR) to determine if sufficient mass is contained in the
selected modes to consider the model valid. The generation of the MR matrix also makes
use of the stiffness matrix (see the sketch below). The MR matrix should also be compared
to the mass properties of the Grid Point Weight Generator. The inertias are calculated
about the SUPORT grid and, unless this is the point selected with PARAM,GRDPNT, the
inertias will not agree. A typical allowable is a 5% loss of mass. This comparison can then
be used to modify the choice of ASET DOF or to increase the number of modes. This
diagonal is shown as the TOTAL line of Table 2-11 and is compared against the full model
weight.
Table 2-11 Galileo Spacecraft Baseline Model Modal Weight and Inertia
(Lb, Lb-in
2
) (continued)

Table 2-12 Galileo Baseline S/C Modal Frequencies
To calculate the MER and MREMER and to print them along with the MR, one should use
the CHECKA DMAP alter from the SSALTER library when there is a SUPORT entry in the
Bulk Data Section. The model must be cantilevered from the SUPORT; that is, the SUPORT
DOFs are the only constraints preventing rigid-body motion. The factor
is in the ALTER. If another value is desired, it must be substituted.
Kinetic and Strain Energy
The CHECKA alters may also be used for printing the kinetic energy distribution as a
fraction of the total, in the form of eigenvectors, for each mode. This information is useful
for showing the energy distribution within a mode, whereas the MER matrix is useful for
G 385.0886 in/sec
2
5 CHAPTER 2
determining which mode is energetic. The various grid points within a model can be
grouped together and the energy then given by subsystem. Table 2-12 shows a sample
taken from Reference 3. Examination of Table 2-11 and Table 2-12 show that for modes 1
and 2 the PWS subsystem has 100% of the KE, but in terms of effective mass it is negligible
(the Plasma Wave Subsystem (PWS) is a very light antenna). In contrast, the probe in
mode 5 not only has the predominant percent KE but is also an energetic mode
(Table 2-12).
Element strain energy can also be used to determine which elastic elements are
participating within a given mode. These two diagnostic tools can help isolate a weak or
very flexible area of the model that may only be a modeling error.
The whole package is intended for use with SOL 3. The model must have a determinate
support point without any SPCs to fix the base.
Deformed Plots
Another diagnostic tool is plotting. Deformed geometry plots, the eigenvector output, and
MER help to identify mode shapes and classify them. Plots will also highlight excessive
deflections that point to stiffness matrix problems.
Information from these diagnostics will feed back to both the stiffness and mass matrices
and the ASET degrees-of-freedom. In only a few iterations, all mass and stiffness problems
can be rectified and the model will be a realistic representation of the true structure.
The quality of the above checks depends on the quality of ASET chosen. It is good practice
to make at least one run using Generalized Dynamic Reduction and to verify the model by
comparing the resulting modes to those obtained using the ASET. If the two runs are not
comparable, the LANCZOS run should be used to select an ASET. The inclusion of all
terms with significant KE usually gives a good ASET, but since large, stiff assemblies might
not have large energy at individual grid points, common sense should also be used.
References
1. MSC.Nastran Handbook for Linear Analysis, Version 64, The MacNeal-Schwendler
Corporation, August 1985.
2. Equivalent Spring-Mass System for Normal Modes, R. M. Bamford, B. K. Wada,
and W. H. Gayman, NASA Technical Memorandum 33-380, February 1971.
3. Galileo S/C NASTRAN Model, Descriptive Data, and Checkcase Results,
R. Calvet and G. Wang, May 1985.
4. Checks that Pay, J. C. Auker, Proceedings, the MSC.Nastran European Users
Conference, Munich, West Germany, May 1984.
5. JSC-20545, Simplified Design Options for STS Payloads, D. A. Hamilton,
May 1985.

CHAPTER
3
System Matrix Formulations
I Overview
I Theoretical Basis for Reduction Methods
I Dynamics of Rotating Structures
I Overview of Aeroelastic Analysis

3.1 Overview
The solutions used in dynamic analysis are more complicated than the solution of a simple
matrix equation. Compared to linear static analysis, the running time for a dynamics
problem is more costly, the required file space is more excessive, and the larger amounts
of output data can be difficult to digest. Therefore, attention should be paid to reducing
the number of solution equations to manageable levels.
Other special effects enter into many cases of dynamic analysis. Certain loads, such as
those used in rotating structures and in aerodynamics, require special matrix formulations
and solution strategies. These problems use conventional matrices to represent structural
properties, but they can also require extra degrees-of-freedom and may need to account for
fundamental changes such as nonlinear effects. For example, in rotating structures the
connections vary with phase angles. In aeroelasticity the effects of the air flow vary with
frequency.
This chapter begins with an explanation of the various options available in MSC.Nastran
for reducing the size of the solution matrices in dynamic analysis. Direct methods, modal
formulations, and combinations are discussed.
The next section presents methods for the analysis of finite element models in rotating
coordinate systems. Applications include turbo machinery, disk drives, and flexible
shafts.
The last section is an introduction to the use of aeroelasticity, which can be applied to any
problem with a steady fluid flow attached to a flexible structure.
6 CHAPTER 3
3.2 Theoretical Basis for Reduction Methods
In the following development we will start from the full-size structural matrix equations
and derive the equations for Static Condensation, Guyan Reduction, Generalized Dynamic
Reduction (GDR), and Component Mode Synthesis. These operations will apply to both
single structures and superelement models. We will also try to explain the physical
consequences of the assumptions involved in reducing the systems.
Definition
The basic dynamic equation before reduction is given in the set (after SPC and MPC
constraints have been applied, but before DMIGs and extra points). The standard matrix
equation to be reduced is:
Eq. 3-1
where:
Note that all free-body motions must be included in the partition. Otherwise, will
be singular. The bar quantities ( , etc.) indicate unreduced values.
Statics
For statics problems, we may ignore the mass and damping effects and solve the lower
partition of Eq. 3-1 for :
Eq. 3-2
The two parts of Eq. 3-2 become the Guyan matrix and the static corrective
displacement :
Eq. 3-3
Eq. 3-4
, , are the displacements, velocities, and accelerations of the analysis (a) set, to
be retained.
, , are the displacements, velocities, and accelerations of the omit (o) set, to be
eliminated.
M, B, K are the mass, damping, and stiffness matrices (assumed to be real and
symmetric).
, are the applied loads.
u
f
M
aa
M
ao
M
oa
M
oo
u
a
u
o
" #
$ %
& ' B
aa
B
ao
B
oa
B
oo
u
a
u
o
" #
$ %
& ' K
aa
K
ao
K
oa
K
oo
u
a
u
o
" #
$ %
& '
+ +
P
a
P
o
" #
$ %
& '
=
u
a
u
a
u
a
u
o
u
o
u
o
P
a
P
o
u
a
K
oo
P
u
o
u
o
{ } K
oo
1
[ ] K
oa
[ ] u
a
{ } P
o
{ } + ( ) =
G
o
u
o
o
G
o
[ ] K
oo
1
[ ] K
oa
{ } =
u
o
o
{ } K
oo
1
[ ] P
o
{ } =

The exact static solution system is obtained by substituting Eq. 3-2 through Eq. 3-4 into the
upper partition terms of Eq. 3-1, resulting in the reduced equations used in the static
solution
Eq. 3-5
and
Eq. 3-6
where:
Eq. 3-7
Eq. 3-8
In actual practice the size of the set is usually small compared to , but the reduced
matrices are dense, resulting in no savings in cost. The savings in solving Eq. 3-5 are
usually offset by the costs of calculating and .
However, for dynamics, we also may approximate the vectors and to reduce the
order of the system. A good place to start is to use the static properties. From Eq. 3-6,
define the transformation
Eq. 3-9
where:
Eq. 3-10
Eq. 3-11
Here are the incremental displacements relative to the static shape. The system
described in Eq. 3-1 may be transformed to the new coordinates with no loss of accuracy.
The stiffness matrix in the transformed system is
Eq. 3-12
Performing the multiplication and substituting Eq. 3-3 results in
Eq. 3-13
K
aa
[ ] u
a
{ } P
a
{ } =
u
o
{ } G
o
[ ] u
a
{ } u
o
o
{ } + =
K
aa
[ ] K
aa
[ ] K
ao
[ ] G
o
[ ] + =
P
a
{ } P
a
{ } G
o
T
[ ] P
o
{ } + =
u
a
u
o
G
o
u
o
o
u
o
u
o
u
f
{ }
u
a
u
o
" #
$ %
& '
H
f
[ ] u
f
{ } = =
u
f
{ }
u
a
u
o
o
" #
$ %
& '
=
H
f
[ ]
I 0
G
o
I
=
u
o
o
K
f f
[ ]
I G
o
T
0 I
K
aa
K
ao
K
oa
K
oo
I 0
G
o
I
=
K
f f
[ ]
K
oo
0
0 K
oo
=
6 CHAPTER 3
Although the stiffness matrix becomes decoupled, the mass and damping matrices tend to
have more coupling than the original system. Since the damping terms have the same form
as the mass, we will not include them here. The exact transformed system becomes
Eq. 3-14
where:
Eq. 3-15
Eq. 3-16
Eq. 3-17
The damping matrix terms of are similar in form to the mass matrix partitions. An
alternative derivation which does not rely on symmetric transformation is given below.
Starting from Eq. 3-1 through Eq. 3-8, we may estimate the acceleration effects of the
omitted points by the equation
Eq. 3-18
Substituting Eq. 3-18 into the lower partition of Eq. 3-1, and solving for , with damping
neglected, we obtain the approximation
Eq. 3-19
Substituting Eq. 3-3 for and Eq. 3-16 for the mass terms into Eq. 3-19, we obtain
Eq. 3-20
Substituting Eq. 3-18 and Eq. 3-20 into the upper half of Eq. 3-1 (ignoring damping), we
obtain
Eq. 3-21
Combining the terms, we obtain the same results as Eq. 3-14 through Eq. 3-17.
The significance of this exercise is to show that the Guyan transformation has very
interesting properties, namely:
1. The approximation occurs only on the acceleration terms (Eq. 3-18).
2. The stiffness portion of the reduced system is exact.
3. The interior displacements defined by Eq. 3-20 and Eq. 3-14 are nearly identical.
M
aa
M
ao
M
oa
M
oo
a
u
o
" #
$ %
& ' K
aa
0
0 K
oo
u
a
u
o
" #
$ %
& '
+
P
a
P
o
" #
$ %
& '
=
M
aa
[ ] M
aa
[ ] M
ao
[ ] G
o
[ ] G
o
[ ]
T
M
oa
M
oo
G
o
+ [ ] + + =
M
ao
[ ] M
oa
T
[ ] M
ao
[ ] G
o
T
M
oo
[ ] + = =
M
oo
[ ] M
oo
[ ] =
B
f f
o
{ } G
o
[ ] u
a
{ }
u
o
u
o
{ } K
oo
1
[ ] P
o
{ } K
oa
[ ] u
a
{ } M
oa
M
oo
G
o
+ [ ] u
a
{ } ( ) =
K
oa
u
o
{ } G
o
[ ] u
a
{ } K
oo
1
P
o
{ } M
oa
[ ] u
a
{ } [ ] +
M
aa
M
ao
G
o
+ [ ] u
a
{ } K
aa
K
ao
G
o
+ [ ] u
a
{ } K
ao
K
oo
1
[ ] M
oa
M
oo
G
o
+ [ ] u
a
{ } +
P
a
{ } K
ao
[ ] K
oo
1
[ ] P
o
{ } =

The significant aspects of the partially decoupled system described by Eq. 3-9 through
Eq. 3-21 are that most of the MSC.Nastran reduction methods are easily developed from
this form and the approximations are conveniently explained in these terms. The Guyan
reduction, the GDR process, and the modal synthesis methods are described below.
Guyan Reduction
In the MSC Guyan reduction process, the omitted relative accelerations, , in Eq. 3-14 are
approximated by Eq. 3-18 and the solution system is
Eq. 3-22
The system described by Eq. 3-18 has several desirable properties. The overall mass and
center of gravity properties are preserved in the mass matrix. Also note that the static
stiffness is exact. If accelerations occur, the errors may be estimated by solving for after
is obtained. The omitted points, , could be recovered by solving the lower part of
Eq. 3-14.
However, in most applications, the stiffness terms of Eq. 3-14 dominate and the normal
modes of the omitted degrees-of-freedom are of higher frequency than the solution set.
Therefore, we may ignore the left-hand mass and solve for directly from Eq. 3-2 by the
equation
Eq. 3-23
The errors in Eq. 3-23 are proportional to the vector , which becomes small when the
individual masses and applied loads in the omit set are small. Because of the same
assumption, the errors of the upper half of Eq. 3-14 are also small. (In MSC.Nastran, is
neglected in the dynamic data recovery process.)
General Dynamic Reduction
In most applications, GDR is used directly to find the modes of the set and no points
are retained. The method then becomes a straightforward Rayleigh-Ritz/Lanczos
procedure. However, it is a recommended practice in MSC.Nastran to extend the set to
include any large masses to avoid numerical roundoff problems and provide more
accurate results.
Approximate Mode Shapes. If grid point displacements are retained in the GDR
process, the set is approximated by a set of generalized coordinates, , where
Eq. 3-24
where are a set of approximate eigenvectors or natural shapes. In its simple form
1
,
with replacing , the GDR transformation corresponding to Eq. 3-9 is
1
In these solution sequences includes and the previous equations for the set are
replaced by the set.
u
o
o
M
aa
[ ] u
a
{ } K
aa
[ ] u
a
{ } + P
a
{ } =
u
o
o
u
a
u
o
o
u
o
u
o
[ ] G
o
[ ] u
a
{ } =
u
o
o
{ }
u
o
o
u
f
u
a
u
a
u
a
u
o
o
u
q
u
o
o
{ }
oq
[ ] u
q
{ }
oq
u
a
u
q
u
a
u
t
u
b
u
c
u
r
+ + =
u
t
u
a
6 CHAPTER 3
Eq. 3-25
The transformed matrix system is then
Eq. 3-26
where is defined by Eq. 3-15 with replacing , as follows:
Eq. 3-27
Eq. 3-28
Eq. 3-29
Eq. 3-30
Eq. 3-31
Eq. 3-32
Error Analysis. Note that, as in Guyan Reduction, the exact static stiffness is retained. If
accelerations exist, errors will occur when the modal displacements do not match the exact
solution for . Assume that the displacement error, , is:
Eq. 3-33
The modal force error is
Eq. 3-34
However, from Eq. 3-14, we obtain
Eq. 3-35
and from the lower half of Eq. 3-26 and Eq. 3-30 through Eq. 3-32, we obtain
Eq. 3-36
Substituting Eq. 3-35 and Eq. 3-36 into Eq. 3-34, we obtain
Eq. 3-37
u
t
u
o
" #
$ %
& '
I 0
G
o

oq
u
t
u
q
" #
$ %
& '
=
M
t t
M
t q
M
qt
M
qq
u
t
u
q
" #
$ %
& ' K
t t
0
0 K
qq
u
t
u
q
" #
$ %
& '
+
P
t
P
q
" #
$ %
& '
=
M
tt
u
t
u
a
M
t q
[ ] M
ot
T
[ ]
oq
[ ] G
o
T
M
oo
[ ]
oq
[ ] + =
M
qq
[ ]
oq
t
M
oo
oq
[ ] =
K
t t
[ ] K
t t
[ ] K
t o
[ ] G
o
[ ] + =
K
qq
[ ]
oq
T
K
oo
oq
[ ] =
P
t
{ } P
t
{ } G
o
T
[ ] P
o
{ } + =
P
q
{ }
oq
T
[ ] P
o
{ } =
u
o
o
o
{ }
q
[ ] u
q
{ } u
o
o
{ } =
q
{ }
oq
T
K
oo
[ ]
o
{ } K
qq
[ ] u
q
{ }
oq
T
K
oo
[ ] u
o
o
{ } = =
K
oo
[ ] u
o
o
{ } P
o
{ } M
oo
[ ] u
o
o
{ } M
ot
[ ] u
t
{ } =
K
qq
[ ] u
q
{ }
a
T
[ ] P
o
{ }
a
T
M
oo
o
[ ] u
q
{ }
o
T
M
ot
[ ] u
b
{ } =
q
{ }
oq
T
M
oo
[ ] u
o
o
oq
[ ] u
q
{ }
" #
$ %
& '
=

Therefore, we observe that the loading errors for GDR will be proportional to the size of
the omitted masses, , and with the quality of the approximation
Eq. 3-38
Relative to Eq. 3-37, the equivalent load error for Guyan reduction was
Eq. 3-39
Even when the mode shapes are only approximate, as in GDR, we conclude that the results
(for the same set) will be improved from the Guyan results. However, note that when
a set is not used, the modes of the system have the same lower frequencies as the
system modes and the approximation might not improve. In other words, the use of
physical points in the GDR process combines the best parts of both Guyan and
Rayleight-Ritz reductions.
Component Mode Synthesis
In MSC.Nastran, the Component Mode Synthesis procedure is identical to the GDR
process except that:
1. The approximate eigenvectors are replaced with results from the module which
calculates modes of the superelement.
2. Boundary set points ( , , or ) will be retained for connection to other
superelements.
3. Special operations are performed for free-body motions.
Note that in Guyan Reduction the relative displacements, , for the omitted point are only
approximated in dynamics. In the GDR and modal synthesis methods, interior results are
obtained by solving for its eigenvector displacements and added directly into the matrix
solution system.
A special case, known as the inertia relief effect, occurs when the structural component is
excited at low frequencies (below the first modal frequency). The effect is defined by
Eq. 3-20 when is a low frequency excitation, having little effect on the vibration modes,
resulting in a relative quasi-static displacement, , where
Eq. 3-40
The displacement vector is called the inertial relief shape when is a rigid body
acceleration vector. It may cause a significant redistribution of the internal forces in the
structure.
In the solution of a single structural component, the inertia relief effects may be obtained
after the solution, , is obtained. However, for multiple superelements connected at the
boundary points, the load distributions on the boundary will be affected by these
internal loads. Note that the errors defined in Eq. 3-23 and Eq. 3-34 will be affected since
the eigenvectors may only approximate the vectors defined in Eq. 3-37.
M
oo
u
o

q
[ ] u
q
{ }
o
{ } M
oo
[ ] u
o
o
{ } =
u
o
u
t
u
o
o
u
t
u
t
u
c
u
r
u
o
o
u
a
u
o
o
u
o
o
{ } K
oo
[ ]
1
M
oa
[ ] u
a
{ } =
u
a
u
a
u
t
6 CHAPTER 3
The solution used in MSC.Nastran provides for inertia relief modes by adding six vectors
to the approximation set . These vectors are obtained from Eq. 3-40 and use the
equation:
Eq. 3-41
where is a matrix of six rigid body motions defined by the grid point geometry. Each
column of the matrix is scaled to a reasonable value and added to the matrix, and a
corresponding displacement is also included.
o
i r
[ ] K
oo
[ ]
1
M
oa
[ ] D
a
[ ] =
D
a
o
ir
o
u
q

3.3 Dynamics of Rotating Structures
The following sections provide a description of the dynamic loads on a rotating structure
defined in the rotating coordinate system. While the stiffness and damping terms
measured in a rotating system are the same as those measured in a stationary system, the
terms due to inertial resistance are dependent upon the rotation of the structure. These
inertial dependent terms need to be determined and added to the total impedance of the
structure before static and dynamic analyses in a rotating system can be performed.
Technical Approach
A rotating structure is usually modeled under the assumption that the coordinate system
used in the description of the structure will rotate at a constant rate about a fixed axis.
Displacements of the structure and forces applied to the structure are measured in this
rotating system. Because the rotating system is accelerating relative to a stationary inertial
system, the inertial or mass dependent impedance cannot be directly calculated in the
rotating system. The inertial impedance in the rotating system must first be determined in
the stationary reference frame, and then transformed to the rotating system.
The impedance due to stiffness and damping is unaffected by the choice of reference
frames. The stiffness and damping impedances in a rotating reference system are identical
to those in a stationary reference system.
To determine the inertial impedance in a rotating system requires the development of
general transformation techniques between stationary and rotating systems. The
following section describe the general transformation of a vector between stationary and
rotating system coordinates. The second section discusses inertial forces defined in a
rotating coordinate system. The third section applies these results to develop the
impedance of a rotating structure. The fourth section describes the DMAP procedure to
add the gyroscopic terms to the structural matrices for analysis in the rotating system.
Vector Transformation from Stationary to Rotating Coordinates
It is important to keep in mind that the essential features of a vector are magnitude and
direction, but not location. A vector defined in one coordinate system can just as easily be
defined in a different coordinate system through the use of a coordinate transformation
matrix. Also keep in mind that vectors may be time dependent in both their magnitude
and direction and that coordinate systems may rotate with respect to each other resulting
in a time-dependent transformation matrix.
The remainder of this discussion concerns transforming a vector defined in stationary
system coordinates to rotating system coordinates. The choice of which system is rotating
and which system is stationary is completely arbitrary at this point. No physical
significance (such as position, velocity, etc.) is assigned to the vector; the only important
qualities are its magnitude and direction.
The general transformation of a time-dependent vector from a stationary coordinate
system to a rotating coordinate system at the same origin may be written as
7 CHAPTER 3
Eq. 3-42
where:
The above transformation is valid for any vector, real or complex.
The fundamentals of this transformation are demonstrated for the two systems shown
below.
Figure 3-1 Stationary and Rotating Coordinate Systems
Assume the rotating system spins about the z-axis at a constant rate, , relative to the
stationary system. , , and are the time-dependent components of the vector
in the stationary system. , , and are the time-dependent components of the
vector in the rotating system.
For this example the transformation of the coordinates from the stationary to rotating
system is
Eq. 3-43
where . Equation Eq. 3-43 is simply an expanded form of Eq. 3-42.
Using complex identities for and ,
Eq. 3-44
Eq. 3-45
= rotating coordinate system definition of time-dependent vector.
= time-dependent transformation matrix from rotating to stationary system.
= stationary coordinate system definition of time-dependent vector.
v t ( )
r
{ } A t ( ) [ ] v t ( )
s
{ } =
v t ( )
r
{ }
A t ( ) [ ]
v t ( )
s
{ }
X
Y
Stationary Rotating
X
Y
Y
X
x t ( )
s
y t ( )
s
z t ( )
s
x t ( )
r
y t ( )
r
z t ( )
r
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
cos sin 0
sin cos 0
0 0 1
x t ( )
s
y t ( )
s
z t ( )
s
" #
. .
$ %
. .
& '
=
t =
cos sin
cos
1
2
--- e
i
e
i
+ ( ) =
sin
i
2
--- e
i
e
i
( ) =

Equation Eq. 3-43 is written as
Eq. 3-46
or
Eq. 3-47
where:
Eq. 3-48
and the conjugate matrix is
Eq. 3-49
Eq. 3-50
Additionally, we can write the vectors in terms of complex components
Eq. 3-51
where:
, , and are time-dependent magnitudes and
, , and are time-dependent phase functions.
Substituting the above definition into Eq. 3-47 produces the complex transformation
x t ( )
r
y t ( )
r
z t ( )
r
" #
.
$ %
.
& '
e
i
2
-------
1 i 0
i 1 0
0 0 0
e
i
2
----------
1 i 0
i 1 0
0 0 0
0 0 0
0 0 0
0 0 1
+ +
* +
/ 0
/ 0
/ 0
, -
x t ( )
s
y t ( )
s
z t ( )
s
" #
. .
$ %
. .
& '
=
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
e
i
T
1
[ ] e
i
T
1
[ ]
T
0
[ ] + + ( )
x t ( )
s
y t ( )
s
z t ( )
s
" #
. .
$ %
. .
& '
=
T
1
[ ]
1
2
---
1 i 0
i 1 0
0 0 0
=
T
1
[ ]
1
2
---
1 i 0
i 1 0
0 0 0
=
T
0
[ ]
0 0 0
0 0 0
0 0 1
=
x t ( )
y t ( )
z t ( )
" #
. .
$ %
. .
& '
x t ( )e
i
x
t ( )
y t ( )e
i
y
t ( )
z t ( )e
i
z
t ( )
" #
. .
. .
$ %
. .
. .
& '
=
x t ( ) y t ( ) z t ( )
x
t ( )
y
t ( )
z
t ( )
7 CHAPTER 3
Eq. 3-52
Inspection of the above equation shows that a vector defined in stationary system
coordinates can be defined as three vectors in rotating system coordinates. One vector will
have the phasing of its motion advanced by with respect to the stationary system
definition, the second will have its phasing delayed by , and the third will have its phase
unaffected.
Progressive and Regressive Vectors.
Eq. 3-53
where
Eq. 3-54
Eq. 3-55
Eq. 3-56
It is convenient to specify the vector defined by Eq. 3-54 as the progressive vector in
rotating system coordinates, the vector defined by Eq. 3-55 as the regressive vector in
rotating system coordinates, and the vector defined by Eq. 3-56 as the collective vector in
rotating system coordinates. Care must be taken when discussing progressive and
regressive vectors to establish whether the definition is in terms of the rotating or
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
T
1
[ ]
x t ( )
s
e
i
x
t ( ) t +
y t ( )
s
e
i
y
t ( ) t +
z t ( )
s
e
i
z
t ( ) t +
" #
. .
. .
$ %
. .
. .
& '
T
1
[ ]
x t ( )
s
e
i
x
t ( ) t
y t ( )
s
e
i
y
t ( ) t
z t ( )
s
e
i
z
t ( ) t
" #
. .
. .
$ %
. .
. .
& '
T
0
[ ]
x t ( )
s
e
i
x
t ( )
y t ( )
s
e
i
y
t ( )
z t ( )
s
e
i
z
t ( )
" #
. .
. .
$ %
. .
. .
& '
+
+
=
t
t
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
+
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
0
+ + =
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
+
e
i t
T
1
[ ]
x t ( )
s
y t ( )
s
z t ( )
s
" #
. .
$ %
. .
& '
=
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
e
i t
T
1
[ ]
x t ( )
s
y t ( )
s
z t ( )
s
" #
. .
$ %
. .
& '
=
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
0
T
0
[ ]
x t ( )
s
y t ( )
s
z t ( )
s
" #
. .
$ %
. .
& '
=

nonrotating system. By reformulating Eq. 3-43 as a transformation from rotating to
nonrotating coordinates, one can define progressive and regressive vectors in terms of the
nonrotating system.
Eq. 3-57
In this discussion, progressive and regressive are defined relative to the rotating system.
Inertial Forces in a Rotating Coordinate System
The calculation of inertial forces in a rotating coordinate system requires the introduction
of an inertial reference system. An inertial reference system is defined as a coordinate
system in which the motion of a particle with respect to these coordinates adhere to
Newtons laws of motion. Implied in these laws is that acceleration of a particle is
measured in a nonaccelerating and nonrotating coordinate system. In the above
discussions the stationary coordinate system can be considered an inertial coordinate
system.
Inertial force is defined as the resistance of a particle to a change in its state of rest or
straight-line motion. It is expressed as Newtons well-known equation
Eq. 3-58
or
Eq. 3-59
where:

The term is referred to as the inertial force. Inertial force represents the resistance of a
particle to a change in its velocity vector. It is equal in magnitude to the applied force but
acts in the opposite direction.
To determine the inertial forces in a rotating coordinate system, it is necessary to transform
the vector representing the acceleration of a particle in the inertial coordinate system to the
rotating coordinate system. This requires: 1) determining the acceleration vector in the
inertial coordinate system, then 2) transforming the acceleration vector from the inertial
system coordinates to the rotating system coordinates. This results in the inertial
coordinate system acceleration vector being defined in terms of rotating system vectors.
F = applied force
m = mass of the particle
a = acceleration of the particle
x t ( )
s
y t ( )
s
z t ( )
s
" #
. .
$ %
. .
& '
e
i
T
1
[ ]
e
i
T
1
[ ] T
0
[ ] + + ( )
x t ( )
r
y t ( )
r
z t ( )
r
" #
. .
$ %
. .
& '
=
F ma =
F ma 0 =
ma
7 CHAPTER 3
Eq. 3-57 defines a vector transformation from rotating system coordinates to stationary
system coordinates. Let a vector be defined as the position vector for a particle in the
stationary system coordinates, and be defined as the position vector in the rotating
system coordinates.
Eq. 3-60
Taking the second derivative with respect to time produces
Eq. 3-61
The equation above defines the acceleration vector in the stationary or inertial coordinate
system.
Eq. 3-47 represents the vector transformation from stationary system coordinates to
rotating system coordinates. Applying this transformation to the stationary system
acceleration vector given by Eq. 3-61 results in the inertial system acceleration defined in
rotating system coordinates, as follows:
Eq. 3-62
Using the identities
Eq. 3-63
Eq. 3-64
Eq. 3-65
Eq. 3-66
Eq. 3-67
r
t ( )
s
{ } e
i
T
1
[ ]
e
i
T
1
[ ] T
0
[ ] + + ( ) t ( )
r
{ } =
d
2
t ( )
s
{ }
dt
2
-------------------------- e
i
T
1
[ ]
e
i
T
1
[ ] T
0
[ ] + + ( )
d
2
t ( )
r
{ }
dt
2
--------------------------
i 2 e
i
T
1
[ ]
e
i
T
1
[ ] ( )
d t ( )
r
{ }
dt
-----------------------
2
e
i
T
1
[ ]
e
i
T
1
[ ] + ( ) t ( )
r
{ }
+
=
d
2
t ( )
s
{ }
dt
2
--------------------------
* +
/ 0
, -
r
e
i
T
1
[ ] e
i
T
1
[ ]
T
0
[ ] + + ( ) e
i
T
1
[ ]
e
i
T
1
[ ] T
0
[ ] + + ( )
d
2
t ( )
r
{ }
dt
2
--------------------------
i 2 e
i
T
1
[ ] e
i
T
1
[ ]
T
0
[ ] + + ( ) e
i
T
1
[ ]
e
i
T
1
[ ] ( )
d t ( )
r
{ }
dt
-----------------------
2
e
i
T
1
[ ] e
i
T
1
[ ]
T
0
[ ] + + ( ) e
i
T
1
[ ]
e
i
T
1
[ ] + ( ) t ( )
r
{ }
+
=
T
1
[ ] T
1
[ ]
T
1
[ ]
T
1
[ ] 0 [ ] = =
T
1
[ ] T
0
[ ] T
1
[ ]
T
0
[ ] 0 = =
T
1
[ ] T
1
[ ] T
1
[ ] =
T
1
[ ]
T
1
[ ]
T
1
[ ]
=
T
0
[ ] T
0
[ ] T
0
[ ] =

Eq. 3-68
Eq. 3-69
Eq. 3-62 reduces to
Eq. 3-70
The term is commonly referred to as the Coriolis acceleration and the term
is referred to as the centripetal acceleration. This equation provides a means
of representing inertial or stationary system acceleration in terms of rotating system
coordinates.
Using Eq. 3-70, the inertial force on a particle whose position is measured relative to a
rotating coordinate system is written as
Eq. 3-71
This equation can be written as
Eq. 3-72
where (for rotation about the z-axis)
Eq. 3-73
Eq. 3-74
Eq. 3-75
T
1
[ ]
T
1
[ ] + 2Re T
1
[ ] ( ) =
T
1
[ ]
T
1
[ ] i 2lm T
1
[ ] ( ) i2Re i T
1
[ ] ( ) = =
d
2
t ( )
s
{ }
dt
2
--------------------------
* +
/ 0
, -
r
d
2
t ( )
r
{ }
dt
2
-------------------------- 4Re i T
1
[ ] ( )
d t ( )
r
{ }
dt
----------------------- 2
2
Re T
1
[ ] ( ) t ( )
r
{ } =
4Re i T
1
[ ] ( )
2
2
Re i T
1
[ ] ( )
F t ( )
i nert i al
{ }
r
M [ ]
d
2
t ( )
s
{ }
dt
2
--------------------------
* +
/ 0
, -
r
M [ ]
d
2
t ( )
r
{ }
dt
2
-------------------------- 4Re i T
1
[ ] ( )
d t ( )
r
{ }
dt
-----------------------
2
2Re T
1
[ ] ( ) t ( )
r
{ }
* +
/ 0
, -
=
=
F t ( )
i nert i al
{ }
r
M [ ]
d
2
t ( )
r
{ }
dt
2
--------------------------
* +
/ 0
, -
B
c
[ ]
d t ( )
r
{ }
dt
-----------------------
* +
, -

2
K
c
[ ] t ( )
r
{ }
* +
/ 0
, -
=
M [ ]
m 0 0
0 m 0
0 0 m
=
B
c
[ ]
0 2m 0
2m 0 0
0 0 0
=
K
c
[ ]
m 0 0
0 m 0
0 0 0
=
7 CHAPTER 3
Structural Impedance of a Rotating Structure
In Eq. 3-72, the inertial forces are dependent on the time-varying position of a mass
particle. This position vector can be written in terms of an initial component and a
time-varying component.
Eq. 3-76
Eq. 3-72 can be rewritten using the above equation as
Eq. 3-77
Inertial forces due to the mass are resisted by structural stiffness and damping.
Eq. 3-78
Substituting Eq. 3-77 in the above equation and collecting all time-dependent terms on the
left-hand side results in
Eq. 3-79
The above equation is the equation of motion for a particle in a rotating elastic structure
with no externally applied forces. The loading on the right-hand side of Eq. 3-79 is often
called the centripetal force. The centripetal force is always present on a rotating structure
and produces an additional stiffness-like term. This additional term can be determined
from consideration of the standard static equilibrium equation for a single element.
Eq. 3-80
where:
K is the element stiffness in global coordinates (coordinates used to define motion).
u is the displacement in global coordinates.
P is the load in global coordinates.
The deformation, , produces element reaction forces, , which balance the applied load,
. The stiffness, , can be viewed as the change in the element reaction forces with respect
to a change in displacement.
t ( ) { }
r

r
{ } u t ( )
r
{ } + =
F t ( )
i nert i al
{ }
r
M [ ]
d
2
u t ( )
r
{ }
dt
2
--------------------------
* +
/ 0
, -
B
c
[ ]
d u t ( )
r
{ }
dt
-----------------------
* +
, -

2
K
c
[ ]
r
{ } u t ( )
r
{ } + ( )
*
+
/
0
,
-
=
B [ ]
d u t ( )
r
{ }
dt
-----------------------
* +
, -
K [ ] u t ( )
r
{ } + F t ( )
i nert i al
{ }
r
=
M [ ]
d
2
u t ( )
r
{ }
dt
2
--------------------------
* +
/ 0
, -
B [ ] B
c
[ ] ( )
d u t ( )
r
{ }
dt
-----------------------
* +
, -
K [ ]
2
K
c
[ ] ( ) u t ( )
r
{ } + +
2
K
c
[ ]
r
{ } =
Ku P =
u F
P K

Eq. 3-81
Eq. 3-82
Eq. 3-83
where:
F is the internal element force in global coordinates.
T is a transformation from element to global coordinates.
is the internal element force in element coordinates.
is a displacement in element coordinates.
For a linear material and a loading which is not dependent on displacement, the term
is constant, the element forces are linearly related to the element displacements.
The terms and are dependent only on the orientation of the element coordinate
system with respect to the global coordinate system. For small displacement problems,
and are assumed constant and are calculated from the undeformed geometry. The
three terms ( , , and ) are used to calculated the linear element stiffness
matrix in global coordinates. The element stiffness matrices are assembled to create the
complete linear stiffness matrix for the structure.
The second part of the stiffness calculation, , is usually ignored in standard linear
analysis. It arises from small changes in the element orientation with respect to the applied
load. For example, consider a bar under axial and transverse loading.
Figure 3-2 Cantilevered Bar With Axial and Transverse Loading
In a standard linear analysis, the -directed load is reacted by the bar bending and shear
forces. The -directed load is reacted by the bar axial forces. The calculated
displacement is independent of the -directed load. Likewise, the calculated
displacement is independent of the -directed load. On closer inspection, one would
find that as the free end of the bar rotates with respect to the applied loads, a component
of the bar axial force reacts against the -directed load. Likewise, a component of the bar
bending and shear react to the -directed load.
K
dF
du
-------
d TF
e
( )
du
------------------ = =
K T
dF
e
du
---------
* +
, -
F
e
dT
du
------
* +
, -
+ =
K T
du
e
du
---------
* +
, -
dF
e
du
e
---------
* +
/ 0
, -
F
e
dT
du
------
* +
, -
+ =
F
e
u
e
dF
e
du
e
T du
e
du
T
du
e
du
T du
e
du dF
e
du
e
F
e
dT du ( )
X
Y
F
y
F
x
y
x
y x
x y
y
x
7 CHAPTER 3
Figure 3-3 Bar Forces Due to Axial Loading
This results in a stiffness coupling which is not taken into account by standard linear
analysis. This effect is due to the term in Eq. 3-83 and is referred to as the
differential stiffness term. Its calculation requires knowledge of the element force and
the transformation matrix . The total stiffness matrix, often referred to as the tangent
stiffness matrix, is the sum of the linear and differential stiffnesses.
Eq. 3-79 shows that a centripetal load is always present in a rotating structure. This
constant loading results in a differential stiffness term which must be included for accurate
analysis of rotating structures. The differential stiffness is added to the stiffness terms on
the left-hand side of Eq. 3-79. The complete equation of motion for a rotating structure,
neglecting the centripetal loading is
Eq. 3-84
The above equation can be used to determine the motion of a rotating structure about the
deformation due to the centripetal loading.
DMAP Procedure to Add Rotation-Dependent Terms to
MSC.Nastran Dynamic Analyses
These alters may be used for the following problems:
To determine critical speeds of rotating machinery, such as shafts and turbines.
To predict effects of unbalanced forces at selected rotational speeds.
To simulate transient behavior relative to the steady-state rotation.
Note: In rotor dynamics shafts are often modeled with BEAM elements and the
axisymmetric disks as rigid bodies (CONM2s). Due to the polar movement of
inertia of these disks, there is a gyroscopic effect on the critical speeds.
Unfortunately, the DMAP alters discussed below do not allow for this gyroscopic
effect.
X
Y
x Component of
Force in Bar
y Component of
Force in Bar
Force in Bar
F
x
F
e
dT du ( )
F
e
T
M [ ]
d
2
t ( )
r
{ }
dt
2
------------------------------
* +
/ 0
, -
B [ ] B
c
[ ] ( )
d t ( )
r
{ }
dt
---------------------------
* +
, -
K [ ]
2
K
d
[ ]
2
K
c
[ ] + ( ) t ( )
r
{ }
+
+ 0 =

The procedure to add rotation-dependent terms to an MSC.Nastran analysis is done in two
steps: 1) Calculate the rotation-dependent terms and add them to the structural damping
and stiffness matrices, 2) Restart analysis with these modified matrices.
The first step is performed in SOL 101 (linear static analysis) or SOL 114 (cyclic linear static
analysis). A centripetal load is specified (RFORCE entry in the bulk data) as the only
loading (one subcase only). A DMAP alter uses this information to calculate all
rotation-dependent terms, adds them to the structural damping and stiffness matrices, and
stores the modified matrices on the database.
The second step is performed as a restart in any structured linear solution sequence. The
modified matrices are reduced and the analysis is performed. The user may change the
SPCs, MPCs, and loading, but not the actual structure.
Three alters are available in the SSSALTER library; they are:
A readme file (readme.gyro) and example problems are also included in the SSSALTER
library.
segyroa.vxx Adds gyroscopic and differential stiffness terms to noncyclic
symmetry superelement models. The alter is added to SOL 101.
Restarts to any structured superelement solution sequence are
allowed.
cygroa.vxx Adds gyroscopic and differential stiffness terms to cyclic
symmetric superelement models. The alter is added to SOL 114.
Restarts to any structured cyclic symmetry solution sequence is
allowed.
nlgyroa.vxx Adds gyroscopic terms to the normal modes calculation in
SOL 106 (nonlinear statics). No restart is required, normal modes
are calculated in SOL 106 if PARAM,NMLOOP is included in the
input file.
8 CHAPTER 3
3.4 Overview of Aeroelastic Analysis
MSC.Nastran provides efficient solutions to the problems of aeroelasticity, which is a
branch of applied mechanics that deals with the interaction of aerodynamic, inertial, and
structural forces. It is important in the design of airplanes, helicopters, missiles,
suspension bridges, and even tall chimneys and power lines. Aeroservoelasticity is a
variation in which the interaction of automatic controls requires additional consideration.
The primary concerns of aeroelasticity include flying qualities (stability and control),
flutter, and structural loads arising from maneuvers and atmospheric turbulence.
Methods of aeroelastic analysis differ according to the time dependence of the inertial and
aerodynamic forces that are involved. For the analysis of flying qualities and maneuvering
loads wherein the aerodynamic loads vary relatively slowly, quasi-static methods are
applicable. The remaining problems are dynamic, and methods of analysis differ
according to whether the time dependence is arbitrary (transient or random) or simply
oscillatory in the steady state.
MSC.Nastran considers three classes of problems in aeroelasticity:
Static Aeroelastic Response
Aerodynamic Flutter
Dynamic Aeroelastic Response
Each is described in the text that follows. In addition, information is provided on the
ability to include aeroelastic responses within the MSC.Nastran optimization capability,
the aerodynamic methods available, and on special features that make the MSC.Nastran
aeroelastic capability unique.
Static Aeroelastic Response
For the analyses of flying qualities and maneuvering loads, the assumption of quasi-steady
motion is valid, i.e., the dynamics of the flexible structure are neglected and quasi-static
methods are applicable. By assuming linear behavior of the aerodynamic, inertial and
structural forces during the motion, the equations of equilibrium in quasi-steady flight are
solved in closed algebraic (matrix) form. Linear and/or surface splines may be used to
connect the aerodynamic and structural grid points.
The static aeroelastic analysis solves for the trim condition in a prescribed maneuver. The
formulation of the equilibrium equations provides the aerodynamic stability and control
derivatives as an integral part of the trim process. The external flight loads and the
corresponding internal loads and stresses on the finite elements, are available as
postprocessing operations on the trim solution.
Static aeroelastic divergence is a non-oscillatory instability condition that can occur when
the aerodynamic forces overpower the stiffness of the structure. For free-flying vehicles,
this phenomenon is typically not of concern, but it can be critical in the structural design of
restrained wind tunnel models. Static aeroelastic divergence analysis can be performed as
an option within the overall static aeroelastic capability.

Aerodynamic Flutter
Flutter is the oscillatory aeroelastic instability that occurs at some airspeed at which energy
extracted from the airstream during a period of oscillation is exactly dissipated by the
hysteretic damping of the structure. The motion is divergent in a range of speeds above
the flutter speed. Flutter analysis utilizes complex eigenvalue analysis to determine the
combination of airspeed and frequency for which the neutrally damped motion is
sustained.
Three methods of flutter analysis are provided: the American flutter method (called the
K-method in MSC.Nastran), an efficient K-method (called the KE-method) for rapid flutter
evaluations, and the British flutter method (called the PK-method) for more realistic
representation of the unsteady aerodynamic loads as frequency dependent stiffness and
damping terms. The complex eigenvalue analysis is specified by the user with the
K-method, and the QR-transformation method is used with the KE- and PK-methods.
Again, linear and/or surface splines may be used to connect the aerodynamic and
structural grid points.
Dynamic Aeroelastic Response
The dynamic aeroelastic response problem is one of determining the response of the
aircraft to time-varying excitations. Atmospheric turbulence is the primary example of this
type of excitation, but store ejection loads and landing gear impact can also have an
aeroelastic component. Methods of generalized harmonic (Fourier) analysis are applied to
the linear system to obtain the response to the excitation. The turbulence model may be
regarded either as a stationary random loading or as a discrete gust.
The gust analysis capability computes response to random atmospheric turbulence and
discrete one-dimensional gust fields. The random response parameters calculated are the
power spectral density, root mean square response, and mean frequency of zero-crossings.
The response to the discrete gust is calculated by direct and inverse Fourier transform
methods since the oscillatory aerodynamics are only known in the frequency domain.
Time histories of response quantities are the output in the discrete case.
Aeroelastic Optimization
The integration of the aeroelastic analysis capability contained in MSC.Nastran with a
design sensitivity and optimization capability provides a new (Version 68) design tool for
the aeroelastician. Sensitivity analysis entails the determination of the effects that changes
in structural properties have on response quantities, such as displacements or stresses.
Optimization utilizes information on the response values and their sensitivities to
automatically determine a design that meets a design objective, such as limits on stresses,
deformations, or flutter characteristics. The static aeroelastic and flutter analyses are
available within the MSC.Nastran optimization capability and can be used in
multidisciplinary fashion along with standard static analysis, normal modes analysis, and
dynamic response analysis. Dynamic aeroelastic response is not available for
8 CHAPTER 3
optimization. Aeroelastic responses available for the sensitivity and optimization include
stability derivatives and trim settings from the static aeroelastic analysis and flutter
damping level from the flutter analysis.
Aerodynamic Methods
Five oscillatory aerodynamic theories are available for flutter analysis. There is one
subsonic method, the Doublet-Lattice Method with body interference, and three
supersonic methods: the Mach Box Method, Piston Theory, and a new (Aero II option)
multiple interfering surface method called ZONA51. The fifth method is rudimentary
Strip Theory, which can be applied at any Mach number. The static and dynamic
aeroelastic response solutions use both the subsonic Doublet-Lattice and supersonic
ZONA51 aerodynamic methods.
Special Features
Aeroelastic analysis in MSC.Nastran provides several advances in the state of-the art. The
fundamental problem of interconnecting the aerodynamic and structural grids in the finite
element models is solved by a closed form solution to an infinite plate over multiple
supports. This two-dimensional interpolation was developed in addition to a
generalization of the one-dimensional spline for a bending, twisting beam (elastic axis) on
multiple collinear supports.
The implementation of the lined-up British flutter method, called the PK-method in
MSC.Nastran, was the first attempt to popularize the British approach to flutter analysis in
the United States. This, along with the transfer function capability for control systems,
makes analysis of aeroservoelastic problems a routine matter. The transfer function
representation is for second order systems: a single output from multiple inputs.
The analysis of response to a discrete gust requires Fourier transform methods, because the
aerodynamics assume harmonic motion. First, a direct transform of the discrete gust
profile is necessary to place the forcing function in the frequency domain. Second, an
inverse Fourier transform of the forced frequency response is necessary to obtain the
transient response of vehicles to the gust. Both the direct and inverse Fourier transform
calculations have been implemented in MSC.Nastran.
The quasi-steady equations of motion of a free-flying vehicle require consideration of the
inertial relief effects. For the unrestrained vehicle, the inertial effects are contained in the
basic stability and control derivatives. However, derivatives that are independent of
weight distribution are desirable for use in flight simulators, and are obtained by assuming
the aircraft to be restrained in some reference support configuration. The equations of
motion using restrained aeroelastic derivatives require not only additional inertial
derivatives, but also the rotations of the mean axes relative to the support for each
aerodynamic variable (e.g., angle of attack, elevator rotation, pitch rate). These additional
aeroelastic coefficients permit the support to be unloaded and angular momentum to be
conserved. MSC.Nastran provides both restrained and unrestrained aeroelastic
derivatives, and in the restrained case, the inertial derivatives and mean axis rotations.

Control systems can also be included in dynamic response analysis using MSC.Nastran
transfer functions. With this feature, the aeroservoelastic interactions of ride comfort and
load alleviation systems can be investigated.
The MSC.Nastran procedure has general capabilities that are beyond those listed here, and
their application to aeroelastic design is limited only by the analysts ingenuity. For
example, aerothermoelasticity considers the effects of thermal stresses on structural
stiffness and the subsequent aeroelastic interactions. MSC.Nastran provides a capability
for nonlinear static analysis that includes temperature loadings. Aerothermoelastic
problems of high speed flight can therefore be addressed by restarting any of the three
aeroelastic analyses from a database created by the nonlinear analysis that has generated
the stiffness of the heated structure.
MSC.Nastran Advanced Dynamic Analysis Guide
CHAPTER
4
Boundary Conditions and Coupling to
Other Disciplines
I Overview
I Enforced Motion with Loads
I Lagrange Multiplier Technique
I Axisymmetric Fluids in Tanks
I Virtual Fluid Mass
I Coupled Acoustic Analysis

4.1 Overview
This chapter extends the discussion of basic dynamic loads and boundary conditions given
in the MSC.Nastran Basic Dynamics Users Guide. This chapter describes some of the more
advanced methods available in MSC.Nastran for dealing with these problems.
Although applying dynamic loads to a structure is a relatively straightforward task when
using the finite element method, the process of specifying a moving boundary requires
more work. Examples of enforced motions of a structural boundary include earthquakes
on a building, automobile suspensions, and any vibration affecting a light, flexible
structure connected to a massive, moving body. This chapter describes approximate and
exact solution techniques for this type of problem.
Another type of boundary condition is a fluid or gas coupled with the surface of a
structure. If these fluids are assumed to have small motions, they generate linear matrices
and allow the use of conventional solution methods. This chapter describes three
approaches to the coupled fluid-boundary problem, each with a unique set of capabilities.
8 CHAPTER 4
Boundary Conditions and Coupling to Other Disciplines
4.2 Enforced Motion with Loads
Three distinct methods are used in MSC.Nastran for dynamic enforced motions: the Large
Mass/Spring approach, the Inertial Loads method, and the Lagrange Multiplier (LMT)
method (described in Lagrange Multiplier Technique on page 92). The first two are
covered in this section.
Large Mass/Spring Method
The Large Mass/Spring approach is actually a modeling technique in which the user places
an element with a large mass or stiffness at the points of known acceleration or
displacement. In effect, this large element acts as a constraint on the connected point. The
user then supplies a corresponding large force via RLOADi or TLOADi inputs to produce
the desired motion. If the added element is sufficiently stiff or heavy, the reaction forces
from the actual structure will not affect the input motions. The actual input data and
controls are described in Enforced Motion on page 195 of the MSC.Nastran Basic
Dynamics Users Guide. Refer to them for modeling details. More advanced applications of
both large mass and large spring techniques are discussed below.
Using a large mass to enforce boundary motions is a standard practice in MSC.Nastran. It
works well on a simply supported structure when a single grid point is excited by a well
defined acceleration input. A typical example is the earthquake analysis of a tall building
where a single base input is assumed. However the method may be abused when other
types of boundary conditions are attempted.
Some additional considerations for the advanced analyst are listed below:
1. If enforced motion is applied to a redundant set of boundary points, a danger
exists that the large masses (or springs) may create fictitious forces and stresses in
the structure. This occurs when the enforced displacements inputs are not
precisely synchronized. Also, in a modal formulation, the extra masses will cause
fictitious low frequency modes to occur. The recommended procedure is to
connect the redundant points with RBE (rigid) elements to prevent their
independent motions. The Lagrange Multiplier method is much better, but will
not correct for errors in the loading functions. (See next note.)
2. Small errors in the loading history may cause large errors in the structural
response. When using enforced accelerations in a transient solution, a small bias
in the inputs (from instrumentation or processing) may cause a large spurious
drift where the structure displaces a large amount as a rigid body. Solutions to
remove the drift are as follows:
Supply a corrective load function, obtained from an initial run, to cancel
the measured drift;
Add dampers and springs in parallel with mass and tuned to filter the
input signal; or
Use the modal method and drop the zero-frequency modes. (See item 4
below.)

An example of a mechanical filter is shown in the sketch below. A large mass, ,
is attached to the base of a structure to allow enforced accelerations. For control
of the spurious displacements and velocities, attach a scalar spring, , and
damper element, , between the mass and ground for each direction. If the first
modal frequency is then set for a filter frequency approximately
one tenth of the first vibration mode frequency, . Set the damping coefficient to
a value near critical to eliminate spurious oscillations.
3. In the nonlinear transient solutions, large springs can affect the error tests and
convergence logic. The internally calculated error ratios are dimensionless
numbers obtained by dividing the errors by an average force or total energy. The
forces and energy created by the large mass/spring approach will dominate these
values, resulting in underestimates of the errors and false convergence. The
solution is to decrease the error allowables on the TSTEPNL Bulk Data inputs.
4. Using these methods in a modal formulation requires some attention. For output
of total displacements, the user should retain the zero or low frequency modes
that the large masses produce (i.e., set the parameter LFREQ = .01). Note that if
the low frequency modes are dropped from the dynamic solution, the output will
be the correct relative motion.
Large springs for enforced displacements are not recommended for the modal
formulations. They should generate high frequency modes that are usually
missing from the system. The resulting dynamic solution is not valid since the
large springs are not included in the modal stiffness matrix.
5. Numerical conditioning of the matrix solution may be affected by the method
used to connect the large mass or spring. Numerical roundoff of the results may
occur. MPCs, RBEis, and ASET operations all use a matrix elimination procedure
that may couple many degrees-of-freedom. If a large mass or spring is not
retained in the solution set, its matrix coefficient will be distributed to other
solution points. Then, matrix conditioning for decomposition operations
M
0
K
B
f
1
K 0.4 Mf
1
2
=
f
1
Structure
Seismic Mass
B K Filter
M
0
8 CHAPTER 4
becomes worse, when the large terms dominate the significant finite element
coupling terms). On the other hand, if the degrees-of-freedom with the large
terms remain in the solution set, they remain on the diagonal of the matrix and
the matrix decomposition is unaffected.
In summary, the large mass method is recommended for cases with known accelerations
at a single point. It works well with the modal formulation, providing good stress and
forces near the mass, and is easy to understand and use. In many cases, the structure is
actually excited by the motions of a large, massive, base (for instance, the geological strata)
which can actually be used as a value for the mass.
The large spring method is recommended for cases in which displacements are known at
one point and a direct formulation is used. The primary advantage is its simplicity.
However, the inertial loads approach or Lagrange Multiplier technique, described below,
are more general and have better reliability.
The Inertial Loads Approach
By changing to a moving coordinate system, the reference accelerations may be converted
to inertial loads. If the dynamic problem involves a free body connected to a single rigid
base, the GRAV and dynamic load inputs may be used to supply forces proportional to the
base acceleration. The solution will be defined by the displacements measured relative to
the fixed base.
Approach
The basic approach is shown in the sketch below:
We assume that the motions of the rigid base are known and the structure is fixed only to
the base. Define the displacement vector, , as a sum of the known base motion, and
the relative displacements, by the equation
Eq. 4-1
where is the rigid body transformation matrix that includes the effects of coordinate
systems, offsets, and multiple directions. If the structure is a free body and aerodynamic
effects are neglected, the base motions should not cause any static or viscous forces and
Eq. 4-2
Structure
Base
u
u
u
0
u u
0
u
u { } u { } D [ ] u
0
{ } + =
D [ ]
K [ ] D [ ] 0 =
and B [ ] D [ ] 0 =

where is the stiffness matrix and is the damping matrix. These properties allow us
to remove the reference motions from the solution as shown below.
The equilibrium equation for the whole system is
Eq. 4-3
Substituting Eq. 4-1 and Eq. 4-2 into Eq. 4-3, we obtain
Eq. 4-4
The solution matrices have not changed, but the solution, , must be zero at the base
attachment points. The right-hand term in Eq. 4-4 may be calculated in MSC.Nastran using
one of several options. One option is the response spectra analysis described in Shock
and Response Spectrum Analysis on page 221. Another is a recent MSC.Nastran
implementation that now allows time-dependent gravity loads. By definition, the GRAV
input generates load vectors of the form
Eq. 4-5
where is the gravity vector. Then if were time dependent, it could be used to replace
in Eq. 4-4 if
Eq. 4-6
Input Data
The inertial loads method is available for all dynamic response solutions. Refer to the
MSC.Nastran Basic Dynamic Analysis Users Guide for the complete input procedure for
using static loads in a dynamic response analysis. The method is valid for both direct and
modal formulations and is automatic in nonlinear and superelement models. Briefly, the
necessary input data is as follows:
1. LOADSET = N a Case Control request for processing a LSEQ (load sequence)
set.
2. LSEQ Bulk Data that connects the forces resulting from the GRAV data to a
dynamic DAREA set. Several load vectors may be included in the same LSEQ set.
3. GRAV Bulk Data defining an acceleration vector. Several may be included to
define different functions of time for the different directions.
4. SPC remember to fix the base attachments.
5. TLOADi or RLOADi Bulk Data for the dynamic load definition. Refers to the
DAREA set defined in the LSEQ data.
Other necessary inputs are the conventional DLOAD and TABLEDi data to help define the
time functions. DAREA Bulk Data is not necessary.
K [ ] B [ ]
M [ ] u
{ } B [ ] u
{ } K [ ] u { } + + P { } =
M [ ] u
{ } B [ ] u
{ } K [ ] u { } + + P { } M [ ] D [ ] u
0
{ } =
u
P
g
{ } M [ ] D [ ] g { } =
g g
u
0
g t ( ) { } u
0
t ( ) [ ] =
9 CHAPTER 4
Recommendations
The following comments describe the benefits and drawbacks to the method:
1. The solution will avoid the problem of rigid body drift that occurs in the other
methods when small errors are present in the input accelerations.
2. By eliminating large constant terms, the stress and force calculations may be
numerically more precise.
3. The method is compatible with superelement and nonlinear solution sequences.
4. No special sets or SUPORT data is necessary.
5. The main drawback is that the output accelerations are also relative and will not
correlate to measured accelerations.
For a method that solves for total response directly, see Lagrange Multiplier Technique
on page 92.

4.3 Lagrange Multiplier Technique
The Lagrange Multiplier Technique (LMT) is a special option for processing constraints in
matrix solutions. It can be used for all constraints and reduction methods such as SPC,
MPC, and rigid elements. LMT can also be used for enforced motion analysis in dynamic
analysis. Unlike the approximate large mass approach, the LMT is an exact method. The
LMT requires adding extra degrees-of-freedom to the matrix solution that are used as force
variables for the constraint functions. Coefficients are added to the matrices for the
equations that couple the constrained displacement variables to the points at which
enforced motion are applied.
The LMT produces indefinite system matrices that require special resequencing of
variables for numerical stability by either the user or by a special solver. The sparse matrix
solver first introduced in Version 67 includes an automatic row and column interchange
capability that allows reliable and efficient solution of indefinite matrices without the need
for the user to sequence variables for numerical stability. This solver is installed in several
more modules in the current version of MSC.Nastran such as the transient and frequency
response solver modules. The alter package takes advantage of their capability.
The DMAP alter introduces the LMT variables automatically without the need for special
user input. The user interface is basically unchanged. The only new inputs are the
INCLUDE entry to bring in the alter package, some optional new PARAM entries for
special operations, and a new enforced motion specification that is less complex than that
of the large mass method for enforced motion while providing better accuracy and
numerical stability. A few capabilities are blocked in this implementation as described at
the end of the user interface description. Details of the user interface are given below.
In summary, the DMAP implementation of the LMT method may be considered for any
linear analysis. It brings the largest advantages when the number of boundary constraint
equations is large or when superelement models are involved.
A Comparison of Condensation and Augmentation Solution
Methods
A small model is used here to contrast the conventional method of constraint elimination
and the LMT method of constraint augmentation. Consider the statics model shown
below.
k
u
1
u
2
enforced motion Y
1
applied
force P
2
applied
m
1
m
2
9 CHAPTER 4
Given this model, static equations of motion are as follows:
Eq. 4-7
Eq. 4-8
contains the unknown constraint forces required to cause the enforced motion and
react to the applied force. is the known enforced displacement, and is the known
applied force.
In the LMT method, the constraint equations are included in the system equations. The
displacement constraint can be written in matrix form after is moved to the left side,
Eq. 4-9
This row and a corresponding column are added to the system matrix to produce
Eq. 4-10
In the equation , the known vector quantities are now all on the right side, and
the unknown vector quantities on the left side. This equation can be solved directly
without the need for partitioning or constraint elimination. In dynamic analysis, enforced
motion can be input on velocity and acceleration variables directly by similar techniques.
The system has a unique solution even though there are null terms on the diagonal of .
These null terms are an indication of an indefinite matrix.
For a conventional modal solution, the unconstrained equation for eigensolution is
Eq. 4-11
When the constraint reduction process is used, the first row and column are eliminated.
The resulting equation for the eigenvalue is then
Eq. 4-12
Alternatively, using the LMT process, the system is augmented with the constraint
variables, with the resulting determinant
K [ ] u { } P { } q { } + =
u
1
Y
1
=
q { }
Y
1
P
2
q
1
0 1 0
q
1
u
1
u
2
Y
1
=
K [ ]
0 1 0
1 k k
0 k k
, P { }
Y
1
0
P
2
" #
. .
$ %
. .
& '
, u { }
q
1
u
1
u
2
" #
. .
$ %
. .
& '
= = =
K [ ] u { } P { } =
K [ ]
k m
1
k
k k m
2
u
1
u
2
" #
$ %
& '
0 { } =
k m
2
=

Eq. 4-13
When this determinant is evaluated, the only non-null product results from the (12)(21)(33)
terms, resulting in the same expression, .
Three counter-intuitive consequences are as follows:
1. does not enter into the determinant at all, even though it is present in the
matrix. In a constraint reduction process, by contrast, any mass term that appears
in the matrix used for eigensolution affects the eigenvalues.
2. Although the matrix has three rows and columns and two mass terms on its
diagonal, only one value of is defined. One might expect that this system would
have two or three values for , but it does not. Other values of beyond this
single value are neither zero nor infinity nor any other value. They are simply
undefined.
3. The results from a condensation solution and an augmentation solution are
identical. This is a general rule.
The resulting eigenvector is
Eq. 4-14
Where is the constrained eigenvector for the displacements and is the reaction force.
Note the scale factor of the whole eigenvector may be affected by if the normalization
method is MAX. For a dynamic solution using a modal formulation, the matrices are
transformed using the modes above. In the stiffness, mass, and damping matrices, the
constraint coefficients have no effect. However, if is an enforced displacement, the
generalized modal forces are
Eq. 4-15
The result of applying an enforced displacement on the LMT coordinates, , will be a
modal force in the opposite of the direction of motion. The results of a modal solution will
be relative to . See below for a solution that provides total displacements.
Matrix Theory for Dynamic Response
Notation
The notation of the conventional solutions is described in Constraint and Set Notation
on page 321 of the MSC.Nastran Reference Guide. It is maintained with the following
changes. The Lagrange Multiplier Technique requires the definition of enforced motion
0 1 0
1 k m
1
k
0 k k m
2
0 =
k m
2
=
m
1

{ }
r
c1
c2
" #
. .
$ %
. .
& '
k
0
1.0
" #
. .
$ %
. .
& '
= =
c
r
r
Y
1
P
c
T
P
g
r
T
Y
1
P
2 k
Y
1
= =
Y
1
Y
1
9 CHAPTER 4
variables. Their DOFs are placed in the , , and sets. These sets are members of the
USET table but are not used in the conventional solution method. They are defined on
USETi Bulk Data entries. They are regarded as members of the b-set in the alter package,
and have the following meanings:
New Unique Sets for Enforced Motion
The Unique Sets
Almost all of the letters of the alphabet are already in use to describe sets in the
conventional solutions. Therefore, sets with integers for names are added to the present
set names which use letters of the alphabet instead.
Direct Solutions in the Frequency and Time Domains
The statement of the frequency response problem is
Eq. 4-16
where:
Eq. 4-17
, , and are the stiffness, damping, and mass matrices, respectively, and is the
excitation frequency. If static or dynamic reduction are not used, most of the constraints
are placed in the stiffness matrix. The advantage of this placement is that the constraints
are not multiplied by . This allows the solution of the equations at an excitation
frequency near to zero. The remaining constraints are associated with enforced motion
variables. These constraints are placed in the stiffness, damping, or mass matrix
depending on whether the enforced motion is the displacement, velocity, or acceleration
type, respectively. If static or dynamic reduction is performed, all m- and s-set constraints
are eliminated along with the o-set. The -set is then the only constraint set in the d-set.
The DOFs listed on USET entries are input points for enforced motion. Data
conventionally used to define dynamic loads applied to these points is interpreted to be
time- or frequency-varying enforced motions. The value of the SETNAME entry on the
USET entries determines the type of input. A value of indicates an enforced
displacement type. A value of indicates an enforced velocity type, and a value of
indicates an enforced acceleration type. These dynamic quasi-loads are transferred to the
enforced motion variables by the transformation matrix as follows:
Eq. 4-18
Enforced displacement set
Enforced velocity set
Enforced acceleration set
u
1
u
2
u
3
u
1
u
2
u
3
A
dd
( ) [ ] u
d
{ } P
d
( ) { } =
A
dd
( ) [ ] K i B
2
M + [ ]
dd
=
K [ ] B [ ] M [ ]
u
3
u
1
u
2
u
3
T
4p
[ ]
P
4
( ) [ ] T
4p
[ ] P
p
( ) [ ] =

is the load vector output by the FRLG module. In this context it may contain both
applied forces and enforced motions. has a unit term in each row at the column
corresponding to the -set point where enforced motion is applied. The enforced motions
are removed from the displacement variables by the transformation matrix ,
Eq. 4-19
where is an identity matrix except for the rows associated with enforced motion
terms, where it has a zero row. The -set and -set components are merged to form
.
The stiffness and constraint components are separated into and , respectively,
where . This is done in the GMA subDMAP to ensure that the
constraints are not multiplied by the damping factor iG.
The equations for transient response analysis are similar to those for frequency response
analysis. The subDMAPs for the two types of analysis closely parallel each other.
A new set of partitions not actually used in the DMAP implementation is introduced here.
This set aids in understanding the effects of the enforced motion variables. Let
Eq. 4-20
and
Eq. 4-21
The 3-set contains the constraint force and displacement variables associated with enforced
motion. is the vector of constraint forces and the vector of constraint
displacements. Both are partitions of the vector and is their complement.
is the vector of enforced displacements, additional forces placed on the enforced
motion points, and their complement. The enforced motions of any LMT variables
contained in are zero by definition. contains both displacement and constraint
force unknown variables, while contains both known input displacements and
applied forces.
Consider the case of enforced displacement. then has the partitions
Eq. 4-22
P
p
[ ]
T
4p
[ ]
u
i
T
pp
x
[ ]
P
p
( ) [ ] T
pp
x
[ ] P
p
( ) [ ] =
T
pp
x
[ ]
p u
4
P
p
( ) [ ]
K
aa
g
[ ] R
aa
[ ]
K
aad
[ ] K
aa
g
[ ] R
aa
[ ] + =
u
d
{ }
q
3
u
3
u
c
" #
. .
$ %
. .
& '
=
P
d
{ }
Y
3
P
3
P
c
" #
. .
$ %
. .
& '
=
q
3
{ } u
3
{ }
u
d
{ } u
c
{ } Y
3
{ }
P
3
{ }
P
c
{ }
P
c
{ } u
c
{ }
P
c
{ }
K [ ]
K [ ]
0 R
33
R
3c
K
33
K
3c
Sym. K
cc
=
9 CHAPTER 4
is a subset of the constraint matrix . By inspection, , an identity
matrix, and is null. The form of the equation of motion is then
Eq. 4-23
For this case and have no terms for the -set variables. Their partitions are not
shown. Evaluating the first row of this equation, we obtain
Eq. 4-24
Note that is the force of constraint with a sign change.
Enforced Accelerations. Now consider the case of enforced acceleration. The equation
of motion is then
Eq. 4-25
Again, from the first row,
Eq. 4-26
The input quantity is an acceleration vector because the first partition of is
equated to an acceleration vector. That is, the type of the matrix where the constraint
equations are placed determines the type of motion of . In this DMAP alter package
enforced motion variables may be placed in any or all of the matrices. Mixed types of
enforced motion are therefore supported.
Note that in this case, is the force of constraint.
Modal Equations in the Frequency and Time Domains
All constraint equations are initially placed in the stiffness matrix to allow the solution of
the eigenvalue problem when rigid body modes are present. The system eigensolution is
obtained from the equation
Eq. 4-27
where is a natural frequency, is the corresponding eigenvector, and is the number
of eigenvectors calculated. After the equations are reduced to modal variables, the
constraints for enforced motion are moved to the damping or mass matrix for enforced
velocity or enforced acceleration, respectively, using the same user interface as the direct
solutions.
R
33
[ ] R
4g
[ ] R
33
[ ] I
33
[ ] =
R
3c
[ ]
M
dd
[ ] u
d
{ } B
dd
[ ] u
d
{ }
0 I
33
0
I
33
K
33
K
3c
0 K
c3
K
cc
q
3
u
3
u
c
" #
. .
$ %
. .
& '
+ +
Y
3
P
3
P
c
" #
. .
$ %
. .
& '
=
M
dd
[ ] B
dd
[ ] u
3
u
3
{ } Y
3
{ } =
q
3
0 I
33
0
I
33
M
33
M
3c
0 M
c3
M
cc
q
3
u
3
u
c
" #
. .
$ %
. .
& '
B
11
[ ] u
d
{ } K
dd
{ } u
d
{ } + +
Y
3
P
3
P
c
" #
. .
$ %
. .
& '
=
u
3
{ } Y
3
{ } =
Y
3
{ } P
d
{ }
Y
3
{ }
q
3
K
aa

2
M
aa
2
[ ]
az
{ } 0 =

az
{ } z

Each static constraint mode is computed by fixing all constraint DOFs except one and
applying a unit motion to it. The motion of the physical variables is the superposition of
the motion of the flexible modes and the constraint modes.
Let , where is defined in Eq. 4-20. is the transformation between
the modal and physical dynamic variables. , the unknown variable vector for modal
analysis, has the components
Eq. 4-28
is the vector of generalized displacements for the flexible modes. and are
the unknown constraint force variables and displacement variables for the constraint
modes. The corresponding known variables are
Eq. 4-29
is the modal force vector and and the known enforced motion variables and
applied forces of the enforced motion points.
is obtained from the equation
Eq. 4-30
where is defined in Eq. 4-21.
The equation relating and is used to replace in Eq. 4-16. The resulting
equation is pre-multiplied by to form the reduced basis equations of motion as
follows:
Eq. 4-31
When the multiplication of Eq. 4-31 is carried out, the components of the modal force
vector expressed in the partitions of Eq. 4-29 become:
Eq. 4-32
Eq. 4-33
Eq. 4-34
u
d
{ } T
dh
[ ] u
h
{ } = u
d
{ } T
dh
[ ]
u
h
{ }
u
z
q
3
u
3
" #
. .
$ %
. .
& '
u
h
{ } =
u
z
{ } q
3
{ } u
3
{ }
P
h
{ }
P
z
Y
3
P
3
" #
. .
$ %
. .
& '
=
P
z
{ } Y
3
{ } P
3
{ }
P
h
{ }
P
h
{ } T
dh
[ ]
T
P
d
{ } =
P
d
{ }
u
d
{ } u
h
{ } u
d
{ }
T
dh
T
[ ]
T
dh
[ ]
T
Z
dd
( ) [ ] T
dh
[ ] u
h
{ } T
dh
[ ]
T
P
d
{ } P
h
{ } = =
P
z
{ } q
3z
[ ]
T
Y
3
{ }
3z
[ ]
T
P
3
{ }
cz
[ ]
T
P
c
{ } + + =
u
3
{ } u
3
{ } Y
3
{ } = =
P
3
{ } q
33
[ ]
T
Y
3
{ } P
3
{ } u
c3
[ ]
T
P
c
{ } + + =
9 CHAPTER 4
The modal forces have components from all three partitions of . is
identically . includes loads applied directly to 3-set points , as well as the
constraint forces needed to maintain the motion of the unknown input point variables
at the prescribed input motion . (In this implementation the forces are moved to
.)
If there are enforced motion terms, the rows and columns of the constraint terms, and
, are moved to the corresponding stiffness, damping or mass matrices. The load vector
has enforced motion inputs transferred to the LMT variables by the same technique
used in the direct solutions. Note that the presence of enforced motion variables results in
nondiagonal h-set matrices, requires the use of the coupled solution algorithms for
frequency response and transient analysis. The matrices are sparse, however, with
off-diagonal terms occurring only in the columns associated with enforced motion so that
the effect on solution cost is small.
User Interfaces
A description of the user interface for the current DMAP implementation of the LMT
method is given below, followed in turn by a description of the nomenclature and theory
used in the implementation. Please refer to the SSSALTER Library delivered with
MSC.Nastran for the most current filenames and descriptions. (These are also provided on
the MSC Web page at http://www.mscsoftware.com.)
User Interface for the DMAP Alter Implementation
The alter package names and contents are listed in SSSALTER Library For
MSC.Nastran on page 269. The files are:
Most input is conventional. The exceptions are described here.
Executive Input File
Add INCLUDE [filename].vxx before the CEND statement (where xx is the current
version number for MSC.Nastran).
File Name Applies to Solution Sequence
Imt01 101, Static Analysis
Imt03 103, Real Eigensolution Analysis
Imt07 107, Direct Complex Eigensolution Analysis
Imt08 108, Direct Frequency Response Analysis
Imt09 109, Direct Transient Analysis
Imt10 110, Modal Complex Eigensolution Analysis
Imt11 111, Modal Frequency Response Analysis
Imt12 112, Modal Transient Analysis
P
z
{ } P
d
{ } u
3
{ }
u
3
{ } P
3
{ } P
3
{ }
u
3
{ }
Y
3
{ } P
3
{ }
Y
3
{ }
u
3
{ }
q
3
{ }
P
d
{ }

Case Control Input File
The input here is conventional. In dynamic analysis the reaction forces required to
produce enforced motion can be printed or plotted with SPCFORCE requests.
Bulk Data Input File
The Bulk Data entries for static and dynamic analysis are conventional. A new option is
available for enforced motion analysis. There are new parameters for other new options
described below. Some input types are also not allowed, as described below.
Place all DOFs at which enforced motion is to be applied on USET Bulk Data entries. The
SETNAME field of the USET entry specifies the type of input as shown in the example
below. For example, if one grid point is to be given three different types of input, the Bulk
Data entries for it would be:
Input Details. All DOFs listed on USETi entries must be in the b-set of the model. If there
are no OMITi, ASETi, BSETi or CSETi entries present in the Bulk Data file all free DOFs are
in the b-set by default. No other input is necessary. If there are ASETi entries, the DOFs
listed on USETi entries should also be listed on ASETi entries. All of these DOFS are then
in both the b-set and the a-set. If there are BSETi or CSETi entries, all DOFS listed on USETi
entries should also be listed on BSETi entries. An error exit is taken if any ui DOFs are not
in the b-set. All rigid body modes of the model must be constrained by combinations of
SUPORT entries, SPC entries, and USETi entries. The r-set retains its traditional meaning.
It is fixed in static analysis and free in dynamic analysis. The , , and sets comprise
a superset which is fixed in all types of analysis.
The TLOAD1 and TLOAD2 Bulk Data entries offer a provision for a TYPE input for use
with the large mass approach for enforced motion. (See field 5 on the TLOAD1 on
page 2125 of the .) When TYPE is set to the values 1 or 2, the input function or table is
differentiated to convert a velocity or displacement signal into an acceleration signal. This
is the only type of signal that the large mass approach can handle directly. The LMT
approach, by contrast, allows any of these signals to be used directly. For this reason, it is
recommended that the TYPE field be set to zero (or blank, which is equivalent). Use the
SETNAME values on the USET entry as illustrated above instead. They offer the necessary
and sufficient input.
Put the enforced motion time history or frequency variation on the Bulk Data entries
normally used for dynamic load inputs. (This technique is illustrated in the example
problem descriptions below.) This data is treated as enforced motion values rather than as
applied forces when the DOFs referenced are also listed on USET entries. You should
ensure that these points are not inadvertently loaded by other applied loads such as
$USET SETNAME GID CID
USET U1 12 2 $ U1 SELECTS ENFORCED DISPLACEMENT
USET U2 12 3 $ U2 SELECTS ENFORCED VELOCITY
USET U3 12 4 $ U3 SELECTS ENFORCED ACCELERATION
u
1
u
2
u
3
1 CHAPTER 4
pressure loads on elements or inertial loads. This condition can be checked by inspecting
the motion at the input points to ensure that the only signal present is that of the enforced
motion desired. The large mass approach can also be affected by this user error, but it is
less serious in that context because the large scale factor used on the enforced motion tends
to be much larger in magnitude than these unintended enforced motions. The enforced
motion time history or frequency variation appears in the displacement, velocity, and
acceleration output and should be checked when loads and enforced motions are applied
simultaneously to ensure that only the motion intended is applied to these points.
Option for Component Mode Synthesis of the Residual Structure
The Lanczos method of eigensolution is preferred for the LMT formulation. When this
option is selected, eigenvectors are computed for the v-set of the residual structure. The
v-set requires that an o-set be present. It also includes c-set and r-set DOFs when they exist.
These vectors are used to reduce the v-set in a manner analogous to GDR, except that exact
eigenvectors are used in contrast to the approximate eigenvectors of the GDR method. The
following additional entries must be present:
The residual structure must contain QSETi entries and SPOINT or GRID entries
for them.
A METHOD(FLUID) entry must present in Case Control.
An EIGR or EIGRL Bulk Data entry must be present.
This option is not available for fluid-structure analysis, where the METHOD(FLUID)
command is used to select a solution method for the fluid elements. The large mass
method may be used with this alter for those with a desire to continue with this method.
In this event, the user interface is the same as without the alter package.
Optional Parameters and Blocked Capabilities
Some capabilities are blocked or not implemented because they require module changes or
because the DMAP alters for their implementation with the LMT approach are excessive in
scope. These conditions are detected by DMAP operations and cause a DMAP User Fatal
Message (UFM) which describes the symptoms of the error and a corrective action.
Generalized Dynamic Reduction. Use the Lanczos option called by the
METHOD(FLUID) command instead. GDR input causes a fatal error message
(UFM).
Complex eigensolutions when the model contains enforced motion variables are
not possible, and cause a UFM. The HESS method with static or modal
condensation is the most reliable method to use with this implementation.
Skipped output steps in transient analysis are not supported. The skip factor
(field 5 on the TSTEP Bulk Data entry) must be set to 1. A UFM results when the
skip factor is larger than 1.
Initial conditions in transient dynamic analysis (requested by an IC Case Control
command) cause a UFM.

Mode acceleration in modal analysis results in a UFM. This is the feature selected
by PARAM,MODACC.
Virtual mass causes a UFM.
Inertia relief shapes for dynamic analysis.
Computation Time
The ratio of computation time between a conventional and an LMT solution is problem
dependent. The LMT solution should be appreciably faster when the matrix is sparse and
of large dimension, and when there are many multipoint constraints. These constraints
tend to increase the connectivity of the reduced matrix used in conventional analysis. The
cost of the matrix multiplication used to remove the dependent degrees-of-freedom grows
rapidly with the percentage of dependent DOFs. A series of test problems were used to
test cost ratios. These problems are known as the BSTAT series and are often used by MSC
in benchmark studies. They are square plate models constrained along opposite edges by
single-point constraints.
Use of the Superelement Capability
Superelements may also be used to alleviate numerical conditioning problems. For
example, if there are a few very stiff elements in a model made primarily of softer elements,
the stiff elements may be placed in superelements. Placing the stiff elements in the o-set
has a similar effect. Unlike the previous alter package, all superelement reductions are
done with the LMT approach.
Scaled Response Spectra
The user interface is similar to that for the standard solution sequences with the large mass
approach except that the LMT interface for enforced motion is used. The input file changes
which need to be made when converting a large mass approach input file to the LMT
approach are given below:
OFREQ Case Control Command (Optional) This entry is used to select the mode
shapes to be used for the calculations. When converting a model set up for the
large mass method for which the OFREQ command is present, delete the
frequencies from the selected set that are related to the very low frequency modes.
If it is desired to delete the effects of any constraint mode from the calculations,
set its input spectrum to a zero value at zero frequency. A nonzero value for
PARAM,LFREQ will remove all constraint modes.
SUPORT Entry Replace any SUPORT Bulk Data entries used for enforced
motion with USET entries with a SETNAME value of U3.
Large Masses Delete large masses used at enforced motion DOFs from the input
file. They are not needed.
1 CHAPTER 4
The remainder of the input file needs no changes. The alter directory delivered with
MSC.Nastran contains two annotated input files that show the conversion process between
the two approaches. (See the files RFLAG4.DAT and RFLAG5.DAT, which are based on
problems D0312D and D0312R documented in the MSC.Nastran Demonstration Problem
Manual.)
Example Problem Descriptions
A series of verification and demonstration problems are available on the RFA library where
the alter packages are stored. These problems are related to the first model described here.
Problems RFLAG1 through RFLAG6 have been converted to Version 68 with annotations.
Verification Test Model RFLAG1, Direct Frequency Response
A spring of unit value has a unit mass placed at node 2 and is given a unit sinusoidal
displacement input at node 1. A uniform structural damping rate of 0.1 is applied to the
model. A sketch of this model and the input file for this model are given below:
$ FILE RFLAG1. DAT VERIFICATION PROBLEM. VERSION xx
INIT MASTER(S) $ DISCARD DATA BASE AT END OF RUN
TIME 5
SOL 108 $ DIRECT FREQUENCY RESPONSE
INCLUDE rflag. vxx $ Assumes a copy of the alter is in the current
$ directory
CEND $
TITLE = FREQ. RESP. OF SINGLE DOF OSCILLATOR RFLAG1
SUBTITLE = ENFORCED DISPLACEMENT, PARAM G DAMPING
DLOAD = 2 $ SELECT THE ENFORCED MOTION INPUTS
FREQ = 3 $ SELECT THE EXCITATION FREQUENCIES
DISP = ALL$ PRINT THE INPUT AND OUTPUT
SPCF = ALL$ PRINT THE FORCE REQUIRED AT THE INPUT
BEGIN BULK
SPOINT, 1 2 $ SCALAR POINTS DEFINING X1, X2
CELAS2, 12, 1.0 1, , 2 $ SPRING BETWEEN X1, X2
CMASS2, 2, 1.0, 2 $ UNIT MASS ON X2
USET, U1, 1, 0 $ SPECIFY ENFORCED DISPLACEMENT INPUT
DAREA, 100, 1, 0, 1.0 $ UNIT SCALE FACTOR ON INPUT
RLOAD2, 2, 100, , , 200 $ APPLY FUNCTION 200 ON DAREA 100
TABLED1,200
, 0.0, 1.0, 100., 1.0, ENDT $ APPLY UNIT
$ FUNCTION OVER ALL FREQUENCIES
FREQ, 3, 0.0, 0.159155, 15.9155 $ EXCITATION FREQUENCIES OF
$ 0.0; 1/(2*PI), THE NATURAL FREQUENCY WN; AND 100 TIMES WN,
$ CYCLES/UNIT TIME
PARAM, G, 0.1 $ DAMPING CONSTANT TO MULTIPLY TIMES K
ENDDATA
1
2
k = 1.0
m = 1.0 response measured at this point
m = 0.0 motion enforced at this point

The solution sequences supported for this capability may all use the same DMAP alters
that are stored in rflag.vxx. The Case Control input file is conventional for dynamic
analysis except that the SPCFORCE command requests the printing of the applied force
required to cause the prescribed enforced motion, as well as conventional SPC force
output. The method used to generate DOFs for constraint variables is described in the
Theory section. The scalar point variables have only one component, while the DAREA
entry lists the same GID and C value as the USET entry. The RLOAD2 entry correlates the
DAREA and TABLED1 entries. PARAM,G is used to apply structural damping over all
elements in the model.
SUBCASE 1 $ SUPERELEMENT 1 SUBCASE
SUPER = 1 $
METHOD = 3 $ FOR CMS MODES OF S.E. 1
The SUPER entry names the SEID of the superelement being described. The METHOD
entry selects an EIGRL entry which determines the number of eigenvectors of the
component to be computed. Bulk Data entries are added to define variables for
generalized coordinates (SPOINT). Other entries attach the variables to the superelement
and place them in the generalized coordinate set (SEQSET1).
Concluding Remarks
The current MSC.Nastran implementation of the rflag.vxx alter should be considered for
use in the following conditions:
When it is desired to reduce the amount of storage space and computation time.
This is especially true when many MPC entries and rigid elements are used.
When an alternate method to the large mass method is desired for enforced
motion in dynamic analysis.
1 CHAPTER 4
4.4 Axisymmetric Fluids in Tanks
MSC.Nastrans axisymmetric hydroelastic capability allows the user to solve a variety of
fluid problems with small motion, compressibility, and gravity effects. A complete
derivation of the MSC.Nastran model and an explanation of the assumptions are given in
The NASTRAN Theoretical Manual, Section 16.1. The input data and the solution logic have
many similarities to those for a structural model. The standard normal modes analysis,
transient analysis, complex eigenvalue analysis, and frequency response solutions are
available with minor restrictions. The differences between a fluid model and an ordinary
structural model are due to the physical properties of a fluid. The characteristics of an
axisymmetric fluid model are summarized as follows:
1. The independent degrees-of-freedom for a fluid are the Fourier coefficients of the
pressure function (i.e., harmonic pressure) at fluid points in a cylindrical or
spherical coordinate system. The locations of these points are on the r-z plane.
2. Much like the structural model, the fluid data will produce stiffness and
mass matrices. Because they now relate pressures and flow instead of
displacement and force, their physical meaning is quite different. The user may
not apply loads, constraints, sequencing, or omitted coordinates directly on the
fluid points involved. Instead, the user supplies information related to the
boundaries and MSC.Nastran internally generates the correct constraints,
sequencing, and matrix terms. Indirect methods, however, are available to the
user for utilizing the internally generated points as normal grid or scalar points.
See Hydroelastic Data Processing on page 111 for the identification code.
3. When a physical structure is to be connected to the fluid, the user supplies a list
of fluid points on a plane and a related list of special structural grid points on the
circles corresponding to the fluid parts. Unsymmetric matrix terms define the
actual physical relations. A special provision is included in MSC.Nastran in the
event that the structure has planes of symmetry. The user may, if so desired,
define only a section of the boundary and solve the problem with symmetric or
antisymmetric constraints. The fluid-structure interface will take the missing
sections of structural boundary into account.
4. Because of the special nature of fluid problems, some user convenience options
are absent. The fluid elements and harmonic pressures are not available in the
structural plots. Plotting the harmonic pressures versus frequency or time may
not be directly requested. Because mass matrix terms are automatically
generated if compressibility or free surface effects are present, the weight and
C.G. calculations with fluid elements present may not be correct and should be
avoided. Also, the inertia relief rigid format uses the mass matrix to produce
internal loads, and if fluids are included, these special fluid terms in the mass
matrix may produce erroneous results.

5. The general acoustic capability in MSC.Nastran, described in Coupled Acoustic
Analysis on page 124, shares the same small-motion pressure formulation and
also produces unsymmetric matrices. For more generality, the acoustic fluid
model is defined by three-dimensional HEXA and PENTA elements and the
structural boundary is defined with conventional GRID points. Although it lacks
an option for gravity waves on the free surface, it can replace the axisymmetric
hydroelastic modeling for most problems.
In spite of the numerous differences between a structural model and a fluid model, the
similarities allow the user to formulate a problem with a minimum of data preparation and
obtain efficient solutions to large-order problems. The similarities of the fluid model to the
structural model are as follows:
1. The fluid is described by points in space and finite element connections. The
locations of the axisymmetric fluid points are described by rings (RINGFL) about
a polar axis and defined by their r-z coordinates. The rings are connected by
elements (CFLUIDi) which have the properties of density and bulk modulus of
compressibility. Each fluid ring produces, internally, a series of scalar points
and (i.e., harmonic pressures), describing the pressure function, , in the
equation
where the set of harmonics , and are selected by the user. If the user desires
the output of pressure at specific points on the circular ring, he may specify them
as pressure points (PRESPT) by giving a point number and an angle on a specified
fluid ring. The output data will have the values of pressure at the angle given
in the above equation. The output of free surface displacements normal to the
surface (FREEPT) are also available at specified angles, . The Case Control
command AXISYM = FLUID is necessary when any harmonic fluid
degrees-of-freedom are included. This command may not be used when F =
NONE on the AXIF Bulk Data entry (no harmonics specified).
2. The input file may include all existing options except the axisymmetric structural
element data. All existing Case Control options may be included with some
additional fluid Case Control requests. All structural element and constraint data
may be used but not connected to RlNGFL, PRESPT, or FREEPT fluid points. The
structure-fluid boundary is defined with the aid of special grid points (GRIDB)
that may be used for any purpose that a structural grid point is currently used.
3. The output data options for the structural part of a hydroelastic model are
unchanged from the existing options. The output values of the fluid will be
produced in the same form as the displacement vectors but with format
modifications for the harmonic data.
p
n
p
n*
P ( )
P ( ) P
0
p
n
n cos
n 1 =
N
)
p
n*
n 0 N 100 < < sin
n 1 =
N
)
+ + =
0 n n*
1 CHAPTER 4
Pressures and free surface displacements, and their velocities and
accelerations, may be printed with the same request (the Case Control
command PRESSURE = SET is equivalent to DlSP = SET) as structural
displacements, velocities, and accelerations.
Structural plots are restricted to GRID and GRIDB points and any elements
connected to them.
X-Y plot and Random Analysis capabilities are available for FREEPT and
PRESPT points if they are treated as scalar points.
The RINGFL point identification numbers may not be used in any plot
request; instead, the special internally generated points used for
harmonics may be requested in the X-Y plots and random analysis. (See
Hydroelastic Data Processing on page 111 for the identification number
code.)
No element stress or force data is produced for the fluid elements.
As in the case of the axisymmetric conical shell problem, the Case Control
command HARMONlCS = N is used to select output data up to the Nth
harmonic.
Input Data
Several special Bulk Data entries are required for fluid analysis problems. These entries are
compatible with structural entries. A brief description of the uses for each Bulk Data entry
follows:
AXlF
AXIF controls the formulation of the axisymmetric fluid problem. It is a required entry if
any of the subsequent fluid-related entries are present. The data references a fluid-related
coordinate system to define the axis of symmetry. The gravity parameter is included on
this entry rather than on the GRAV entry because the direction of gravity must be parallel
to the axis of symmetry. The values of density and elastic bulk modulus are conveniences
in the event that these properties are constant throughout the fluid. A list of harmonics and
the request for the nonsymmetric (sine) coefficients are included on this entry to allow the
user to select any of the harmonics without producing extra matrix terms for the missing
harmonics. A change in this list, however, will require a RESTART at the beginning of the
problem.
RINGFL
The geometry of the fluid model about the axis of symmetry is defined with RINGFL
entries. The RINGFL data entries serve somewhat the same function for the fluid as the
GRID entries serve in the structural model. In fact, each RINGFL entry will produce,
internally, a special grid point for each of the various harmonics selected on the AXIF data
entry. They may not, however, be connected directly to structural elements (see the GRIDB
and BDYLlST entries). No constraints may be applied directly to RINGFL fluid points.

CFLUIDi
CFLUIDi defines a volume of fluid bounded by the referenced RlNGFL points. The
volume is called an element and logically serves the same purpose as a structural finite
element. The physical properties (density and bulk modulus) of the fluid element may be
defined on this entry if they are variables with respect to the geometry. If a property is not
defined, the default value on the AXlF entry is assumed. Two connected circles (RINGFL)
must be used to define fluid elements adjacent to the axis of symmetry. A choice of three
or four points is available in the remainder of the fluid.
GRIDB
GRIDB provides the same functions as the GRID entry for the definition of structural grid
points. It will be attached to a particular RINGFL fluid point. The particular purpose for
this entry is to force the user to place structural boundary points in exactly the same
locations as the fluid points on the boundary. The format of GRIDB is identical to the
format of GRID except that one additional field is used to identify the RINGFL point. The
GRDSET entry, however, is not used for GRIDB data, and no superelement partitioning is
allowed.
GRIDB entries may be used without a fluid model. This is convenient in case the user
wishes to solve the structural problem first and to add the fluid effects later without
converting GRID entries to GRIDB entries. The referenced RINGFL point must still be
included in a boundary list (BDYLIST; see below), and the AXIF entry must always be
present when GRIDB entries are used. (The fluid effects are eliminated by specifying no
harmonics.)
FREEPT, PRESPT
FREEPT and PRESPT are used to define points on a free surface for displacement output
and points in the fluid for pressure output. No constraints may be applied to these points.
Scalar elements and direct matrix input data may be connected to these points, but the
physical meaning of the elements will be different from in the structural sense.
FSLlST, BDYLIST
FSLIST and BDYLIST define the boundaries of the fluid with a complete freedom of choice.
The FSLIST entry defines a list of fluid points which lie on a free surface. The BDYLIST
data make up a list of fluid points to which structural GRIDB points are connected. Points
on the boundary of the fluid for which BDYLlST or FSLIST data are not defined are
assumed to be rigidly restrained from motion in a direction normal to the surface.
With both of these lists, the sequence of the listed points determines the nature of the
boundary. The following directions will aid the user in producing a list:
1. Draw the z-axis upward and the r-axis to the right. Plot the locations of the fluid
points on the right-hand side of z.
1 CHAPTER 4
2. If one imagines himself traveling along the free surface or boundary with the fluid
on his right side, the sequence of points encountered is used for the list. If the
surface or boundary touches the axis, the word AXIS is placed in the list. AXIS
may be used only for the first and/or last point in the list.
3. The free surface must be consistent with static equilibrium. With no gravity field,
any free surface consistent with axial symmetry is allowed. With gravity, the free
surface must be a plane perpendicular to the z-axis of the fluid coordinate system.
4. Multiple free surface lists and boundary lists are allowed. A fluid point may be
included in any number of lists.
Figure 4-1 illustrates a typical application of the free surface and structural boundary lists.
Figure 4-1 Examples of Boundary Lists
FLSYM
FLSYM allows the user an option to model a portion of the structure with planes of
symmetry containing the polar axis of the fluid. The first plane of symmetry is assumed at
and the second plane of symmetry is assumed at where is an integer
specified on the entry. Also specified are the types of symmetry for each plane, symmetric
(S) or antisymmetric (A). The user must also supply the relevant constraint data for the
FLUID
Structure
Boundary
List #2
Free Surface List
Axis of
Symmetry
Structure
Boundary
List #1
FSLIST: 36, 37, 38, 39, 40
BDYLIST #1: AXIS, 22, 23, 29, 36
BDYLIST #2: 40, 35, 28, 21, 14, 7, 6, 5, 4, 3, 2, 1, AXIS
1 2 3 4 5 6
22
29
36
23
37 38 39
40
35
28
21
14
7
0.0 = 360 M = M

structure. The solution is performed correctly only for those harmonic coefficients that are
compatible with the symmetry conditions, as illustrated in the following example for
quarter symmetry, M= 4.
DMlAX
DMIAX is used for special purposes such as the specification of surface friction effects.
DMIAX is equivalent to DMIG except harmonic numbers are specified for the
degrees-of-freedom. A matrix may be defined with either DMIG or DMlAX entries, but not
with both.
Solution Sequences
Restrictions. The characteristics of the fluid analysis problems which cause restrictions
on the type of solution are as follows:
1. The fluid-structure interface is mathematically described by a set of unsymmetric
matrices. Since most solution sequences are restricted to the use of symmetric
matrices, the fluid-structure boundary is ignored. Thus, for any of these solution
sequences, MSC.Nastran solves the problem for a fluid in a rigid container with
an optional free surface and an uncoupled elastic structure with no fluid present.
2. No means are provided for the direct input of applied loads on the fluid. The only
direct means of exciting the fluid is through the structure-fluid boundary. The
fluid problem may be formulated in any solution sequence. However, only some
will provide nontrivial solutions.
The suggested solution sequences for the axisymmetric fluid and the restrictions on each
are described as follows:
SOL 103 or 200 Normal Modes Analysis
The modes of a fluid in a rigid container may be extracted with a conventional solution
request. Free surface effects with or without gravity may be accounted for. Any structure
data in the section will be treated as a disjoint problem. (The structure may also produce
normal modes.) Normalization of the eigenvectors using the POINT option will cause a
fatal error.
Series Plane 1
Plane 2
S A
Cosine S
A
0,2,4,...
none
1,3,5,...
none
Sine
(*)
S
A
none
1,3,5,...
none
2,4,6,...
1 CHAPTER 4
SOL 107 (with RF28D78) Direct Complex Eigenvalue Analysis
If no damping or direct input matrices are added, the resulting complex roots will be
purely imaginary numbers whose values are the natural frequencies of the system. The
mode shape of the combination may be normalized to the maximum quantity (harmonic
pressure or structural displacement) or to a specified structural point displacement.
SOL 108 or 200 (with RF26D78) Direct Frequency and Random
Response
This solution may be used directly if the loads are applied only to the structural points. The
use of overall structural damping (parameter g) is not recommended since the fluid
matrices will be affected incorrectly.
SOL 109 (with RF27D78) Direct Transient Response
Transient analysis may be performed directly on the fluid-structure system if the following
rules apply:
1. Applied loads and initial conditions are given only to the structural points.
2. All quantities are measured relative to static equilibrium. The initial values of the
pressures are assumed to be in equilibrium.
3. Overall structural damping (parameters and ) must not be used.
SOLs 110, 111, 112, and 200 Modal Formulation
Although these solution sequences may be used in a fluid dynamics problem, their
practicality is limited. The modal coordinates used to formulate the dynamic matrices will
be the normal modes of both the fluid and the structure solved as uncoupled systems.
Even though the range of natural frequencies would be typically very different for the fluid
than for the structure, the program will select both sets of modes from a given fixed
frequency range. The safest method with the present system is the extraction of all modes
for both systems with the Tridiagonalization Method. This procedure, however, results in
a dynamic system with large full matrices. The Direct Formulation is more efficient in such
cases. At present, the capability for fluid-structure boundary coupling is not provided
with the modal formulation. However, the capability may be provided by means of an
alter using the same logic as in the direct formulations.
Hydroelastic Data Processing
The fluid-related Bulk Data entries are converted by the program into equivalent grid
point, scalar point, element connection, and constraint data entry images. Each specified
harmonic, , of the Fourier series solution produces a complete set of special grid point
and connection entry images. In order to retain unique identification numbers, the user
identification numbers are encoded by the algorithm below:
3
g
N

RlNGFL Points
where:
CFLUIDi Connection Entries
where is defined above for each harmonic .
For example, if the user requested all harmonics from zero to two, including the sine series,
each RINGFL entry will produce five special grid entries internally. If the users
identification number (in Field 2 of the RINGFL data entry) were 37, the internally
generated grid points would have the following identification numbers:
These equivalent grid points are resequenced automatically by MSC.Nastran to be adjacent
to the original RlNGFL identification number. A RINGFL point may not be resequenced
by the user.
The output from matrix printout, table printout, and error messages will have the fluid
point labeled in this form. If the user wishes, he may use these numbers as scalar points
for Random Analysis, X-Y plotting, or for any other purpose.
In addition to the multiple sets of points and connection entries, the program may also
generate constraint sets. For example, if a free surface (FSLIST) is specified in a
zero-gravity field, the pressures are constrained to zero. For this case, the internally
generated set of single-point constraints are internally combined with any user-defined
structural constraints and will always be automatically selected.
If pressures at points in the fluid (PRESPT) or gravity-dependent normal displacements on
the free surface (FREEPT) are requested, the program will convert them to scalar points
and create a set of multipoint constraints with the scalar points as dependent variables.
The constraint set will be internally combined with any user-defined sets and will be
selected automatically.
= cosine series
= sine series
Harmonic ID
0 1,000,037
1* 1,500,037
1 2,000,037
2* 2,500,037
2 3,000,037
Grid point ID User ring ID 1,000,000 I
N
+ =
I
N
N 1 +
I
N
N 1 2 +
Element ID User ring ID 1,000 I
N
+ =
I
N
N
1 CHAPTER 4
The PRESPT and FREEPT scalar points may be used as normal scalar points for purposes
such as plotting versus frequency or time. Although the FREEPT values are displacements,
scalar elements connected to them will have a different meaning than in the structural
sense.
Sample Hydroelastic Model
Table 4-1 contains a list of the input data for a sample hydroelastic problem. Figure 4-1
describes the problem and lists the parameters. The relatively small number of grid points
were chosen for purposes of simplicity and not accuracy. The symbols for the fields in the
hydroelastic data entries are placed above each group. Structural data entries are included
in their standard forms. The explanations for the data are given in the following notes:
1. The AXISYM = FLUID entry is necessary to control the constraint set selections
and the output formats for a fluid problem. It must appear above the subcase
level.
2. DISPLACEMENT and PRESSURE Case Control commands are pseudonyms.
DISP = ALL produces all structure displacements, all free surface displacements,
and all fluid pressure values in the output. The HARMONICS control is a limit
on the harmonic data and has the same function as in an axisymmetrical conical
shell problem.
3. The AXIF entry defines the existence of a hydroelastic problem. It is used to
define overall parameters and control the harmonic degrees-of-freedom.
4. The RINGFL entries included define the five points on the fluid cross section.
5. The CFLUIDi entries are used to define the volume of the fluid as finite elements
connected by the RINGFL points. Since parameters p and B are missing, the
default values on the AXIF entry are used.
6. The FSLIST entry is used to define the free surface at z = 10.0. The density factor
is placed on the entry in this case. If blank, the default value on the AXIF entry
is used.
7. The fluid-structure boundary is defined on the BDYLIST entry. The AXIF default
density is used.
8. The GRIDB entries define the structure points on the fluid boundary. Points 3
through 6 are connected to fluid number 2 ring. The rotation in the r direction
(4 in field 8) is constrained.
9. The fact that one-quarter symmetry was used for the structure requires the use of
the FLSYM entry. Symmetric-antisymmetric boundaries indicate that only the
cosine terms for the odd harmonics interact with the structure. If
symmetric-symmetric boundary conditions were chosen on FLSYM, only the
even harmonics of the cosine series would interact with the structure.
10. The PRESPT entries define locations of pressure points in the fluid for pressure
output.
11. The FREEPT entries define locations on the free surface for displacement output.

Table 4-1 Sample Hydroelastic Problem
$ FLUID MASS DEMO - D107D2 DEMO PROBLEM
SOL 107
TIME 2
CEND
1.
TITLE = SAMPLE HYDROELASTlC PROBLEM
SUBTITLE = EIGENVALUE ANALYSIS WITH FLEXIBLE BOUNDARY
AXISYM = FLUID
SPC = 3
CMETHOD = 1
OUTPUT
2.
PRESS = ALL
HARMONICS = ALL
ELFORCE = ALL
BEGIN BULK
BULK DATA FIELD
3.
$AXIF,CID,G,DRHO,DB,NOSYM
AXIF,2,32.2,0.03,,NO
$,N
1
,N
2
1,3
CORD2C,2,,0.,0.,0.,0.,0.,1.0
1.0,0.,0.
4.
$RINGFL,IDF,Xl,X2,X3,IDF,X1,X2,X3
RINGFL,1,4.0,,10.0,2,8.0,,10.0
RINGFL,7,4.0,,5.0,8,8.0,,5.0
RINGFL,13,4.0,,0.0
5.
$CFLUID,EID,IDF1,IDF2,IDF3,IDF4,RHO,B
CFLUID2,101,1,7
CLFUID2,102,7,13
CFLUID3,103,7,8,13
CFLUID4,104,1,2,7,8
6.
$FSLIST,RHO,IDF1,IDF2,IDF3
FSLIST,0.03,AXIS,1,2
7.
$BDYLIST,RHO,IDF1,IDF2,IDF3
BDYLIST,,2,8,13
1 CHAPTER 4
8.
$GRIDB,ID,,,PHI,,CD,PS,IDF
GRIDB,3,,,0.0,,2,4,2
GRIDB,4,,,30.0,,2,4,2
GRIDB,5,,,60.0,,2,4,2
GRIDB,6,,,90.0,,2,4,2
GRIDB,9,,,0.0,,2,,8
GRIDB,10,,,30.0,,2,,8
GRIDB,11,,,60.0,,2,,8
GRIDB,12,,,90.0,,2,,8
GRIDB,14,,,0.0,,2,,13
GRIDB,15,,,30.0,,2,,13
GRIDB,16,,,60.0,,2,,13
GRIDB,17,,,90.0,,2,,13
CQUAD4,10,11,3,9,10,4
CQUAD4,11,11,4,10,11,5
CQUAD4,12,11,5,11,12,6
CQUAD4,13,11,9,14,15,10
CQUAD4,14,11,10,15,16,11
CQUAD4,15,11,11,16,17,12
PSHELL,11,12,0.5
MAT1,12,10.6+6,,0.3,0.05
SPC1,3,246,3,9,14
SPC1,3,135,6,12,17
SPC1,3,135,14,15,16
9.
$FLSYM,M,S1,S2
FLSYM,4,S,A
10.
$PRESPT,IDF,,IDP,PHI,IDP,PHI,IDP,PHI
PRESPT,7,,27,30.,28,60.
PRESPT,8,,30,30.,31,60.
PRESPT,13,,33,30.,34,60.
11.
$FREEPT,IDF,,IDP,PHI,IDP,PHI,IDP,PHI
FREEPT,1,,40,15.,41,30.,42,45.
FREEPT,1,,43,60.,44,75.
FREEPT,2,,45,15.,46,30.,47,45.
FREEPT,2,,48,60.,49,75.
EIGC,1,INV,MAX
0.,0.,0.,5.,3.,2,2
ENDDATA
Table 4-1 Sample Hydroelastic Problem (continued)

Figure 4-2 Sample Hydroelastic Problem
7 8
9 10 11 12
6 5 4 3
2 1
17 16 15
14
13
17
16
15
14
9 3
10 4
11 5
12 6
Symmetric
Anti-
symmetric
Free Surface
Fluid
Points
Structure
Points
Fixed
r
4.0 4.0
5.0
5.0
Fluid: Density p = 0.03
Bulk Modulus B = 1
Gravity g = 32.2
Structure: Thickness t = 0.5
Density r = 0.05
1 CHAPTER 4
4.5 Virtual Fluid Mass
A virtual fluid volume produces a mass matrix which represents the fluid coupled to a
boundary consisting of structural elements and other effects, such as free surfaces, planes
of symmetry, and infinite fluids. The incompressible fluid produces a mass matrix defined
with full coupling between accelerations and pressures on the flexible structural interfaces.
For each requested volume, the boundaries may be combinations of the following:
1. Structural finite element faces, with one or two wetted sides.
2. Free surfaces with zero pressure, but no sloshing effects.
3. Planes of symmetry with symmetric or antisymmetric motion.
4. Infinite boundaries, by default, on nonclosed volumes.
The virtual mass method is well suited for the following problem types:
1. An infinite fluid surrounding part of a structure.
2. A fluid with a free surface contained within a flexible structure.
3. Multiple fluid volumes with combinations of the options above such as a half of
a ship floating in water with several internal tanks, each with free surfaces and
symmetric boundaries.
Note that compressibility and surface gravity effects are neglected. It is assumed that the
important frequency range for the structural modes is above the gravity sloshing
frequencies and below the compressible acoustic frequencies. It is further assumed that the
density within a volume is constant and no viscous (rotational flow) or aerodynamic (high
velocity) effects are present. In other words, a ship traveling at a high rate through a thick
oil patch would require some extra modeling effort.
User Interface
The user interface for a virtual fluid mass analysis is simple and straightforward. The
sketch below illustrates some of the features.

Figure 4-3 Virtual Mass Terminology
1. The fluid/structure interface is defined with ELIST Bulk Data, which specify a set
of wetted TRIA3 and QUAD4 elements that define the structural portion of the
fluid boundary.
2. Each fluid volume is defined on an MFLUID Bulk Data input, which defines the
fluid density, the ELISTs, and other boundaries.
3. A set of MFLUID volumes are requested in the case control request MFLUID =
SID.
4. If pressure outputs are desired they will be printed if the case control request
MPRES = is used to define a set of elements. The elements must also be active in
an ELIST/MFLUID volume.
An example set of input data for a small problem is shown below:
Case Control
MFLUID = 25
MPRES= ALL
etc.
Bulk Data
1 ,2 ,3 ,4 ,5 ,6 ,7 ,8 ,9 ,10
$MFLUID, SID , CID , ZFS , RHO , ELIST1, ELIST2, PLANE1, PLANE2
MFLUID , 25 , 0 , 115.4, .0246, 255 , , S
$ELIST , ELID ,E1 , E2 , E3 , etc.
ELIST , 255 , 1 , THRU , 25
ETC...
Fluid
Volume 2
Fluid
Volume 1
Free Surfaces
Element
List #2
Element
List #1
Structure
1 CHAPTER 4
In the input above we have defined a fluid volume with a free surface normal to the basic
z direction. The xz plane (PLANE1) is a plane of symmetry and the density is 0.0246.
ELIST set 255 defines the positive faces of elements in the range 1 through 25 (missing
numbers are allowed) as the structural/fluid boundary.
The connected elements may be wetted on one or both sides by the same volume. For
instance a baffle plate extends partially into a tank and have two sides in the same fluid
volume. If a plate completely separates the two parts of the tank, it is recommended that
a separate volume be used for each side. Each side of an element should be listed on a
separate ELIST.
Special terms are calculated if enclosed fluid volumes do not have a free surface or a plane
of anti-symmetry. Otherwise, the incompressible fluid would produce a spurious low
frequency mode with a mode shape corresponding to the volume change. This mode will
not affect the structural response, but it is eliminated. Because the free surface allows the
fluid volume to expand, it does not have these problems.
A free surface is defined as an x-y plane in any local rectangular coordinate system. The
user simply specifies a value of z as the upper limit of the fluid volume. Different fluid
volumes may have different levels and orientation. It is not required that the surface
coincide with the element properties since partially wetted elements are allowed.
The virtual mass fluid option may be used in all MSC.Nastran dynamics solutions,
including the following special approaches:
Superelements (Residual Only )
Nonlinear Analysis (SOLs 99 and 129)
Optimization (does not create sensitivity matrices.)
Theoretical Summary
The following is a brief overview of the virtual mass approach. For more information, refer
to the standard references on boundary elements.
The Helmholtz method used by MSC.Nastran solves Laplaces Equation by distributing a
set of sources over the outer boundary, each producing a simple solution to the differential
equation. By matching the assumed known boundary motions to the effective motion
caused by the sources, we can solve a linear matrix equation for the magnitude of the
sources. The values of the sources determine the effective pressures and, thereby, the
forces on the grid points. Combining all of these steps into a matrix equation results in a
virtual mass matrix as derived below.
If is the value of a point source of fluid (units are volume flow rate per area) located at
location , and is assumed acting over an area , the vector velocity at any other point
is
Eq. 4-35
j
r
j
A
j
u
i
r
i
u
j
e
i j
r
i
r
j
2
---------------------- A
j
d
A
j
(
j
)
=

where is the unit vector in the direction from point to point . Note that the gradient
of the vector is the potential function which satisfies Laplaces Equation on a term by
term basis.
The other set of necessary equations are the pressures, , at any point, , in terms of the
density, , sources and geometry, namely
Eq. 4-36
The results of integrating Eq. 4-35 and Eq. 4-36 over the finite element surfaces are
collected respectively in two matrices, and where
Eq. 4-37
and
Eq. 4-38
where are the forces at the grid points. The matrix is obtained by integrating
Eq. 4-36. An additional area integration is necessary to convert the pressures to forces. A
mass matrix may now be defined using Eq. 4-37 and Eq. 4-38 as
Eq. 4-39
where the virtual fluid mass matrix, is
Eq. 4-40
Singularities for Enclosed Volumes
Note that if is singular or nearly singular, the mass matrix will cause problems with the
coupled solution. The singularity is active when a set of nonzero sources can produce zero
velocities on the structural surfaces. This occurs when the fluid is completely enclosed by
the structure and with planes of only symmetric motion. Large pressures could occur
within the fluid volume from an infinitesimal volume change on the incompressible fluid.
An alternate calculation is performed when the matrix is potentially singular. See
Remark 8 on the MFUID Bulk Data description for the specific instances. An additional
source, , is added near the center of each MFLUID volume. The expanded versions of
Eq. 4-37 and Eq. 4-38 are
Eq. 4-41
Eq. 4-42
An additional equation provides that the sum of the squares of the ordinary sources in the
vector, is minimized and a set of Lagrange multipliers, are used to enforce Eq. 4-41
as constraints. The function, , to be minimized is defined as
e
i j
j i
u
i
p
i
i
p
i
j
e
i j
r
i
r
j
------------------- A
j
d
A
j
(
j
)
=
[ ] [ ]
u
{ } [ ] [ ] =
F { } [ ]
{ } =
F [ ]
F { } M
f
[ ] u
{ } =
M
f
[ ]
M
f
[ ] [ ] [ ]
1
=
[ ]
[ ]
s
2
u
{ } [ ] { }
2
{ }s
2
+ =
F { } [ ]
{ }
2
{ }s
2
+ =
1
{ }
U
1 CHAPTER 4
Eq. 4-43
Taking the derivatives of with respect to , , and , we obtain three sets of equations.
Combining them, we obtain the following matrix equation:
Eq. 4-44
The vector may be eliminated by solving the second row partition and substituting
into the first row partition of the matrix, with the result:
Eq. 4-45
Eq. 4-45 may be solved for and , which are then used to obtain pressure. To obtain
the pressure, we substitute for in Eq. 4-42 to obtain the matrix equation
Eq. 4-46
Eq. 4-45 and Eq. 4-46 may then be combined into a single matrix defining the fluid, namely
Eq. 4-47
where:
Eq. 4-48
Note that the size of the matrix equation is only one term larger than the nonconstrained
case. Although it will be well behaved for the enclosed volume case, the overall
incompressible constraint is lost.
Other methods to avoid the singular matrix are as follows:
1. Put one or more small holes in the boundary by removing an element ID in an
unimportant area from the ELIST. The fluid will then leak out to an infinite
domain of fluid.
2. Define a free surface near the top of the container and modify the ELIST to remove
elements above the surface.
3. Constrain the structure to eliminate any net change to the enclosed volume. In
other words generate an MPC equation such that
U
1
2
--- [ ]
T
{ } [ ]
T
u
{ } [ ] { }
2
{ }s
2
{ } + =
U
1
s
2
0
2
T
I 0
2
T
0 0
s
2 " #
. .
$ %
. .
& '
u
0
0
" #
. .
$ %
. .
& '
=
{ }
2
T
0
s
2 " #
$ %
& ' u
0
" #
$ %
& '
=
{ } s
2
{ }
p { }

T
2
" #
$ %
& '
=
F { } M
f
[ ] u
{ } =
M
f
[ ]

T
2
T
0
1
I
0
=

Eq. 4-49
where is a boundary grid point and and are the effective area and normal vector,
respectively. (This is not an easy task.)
Fortunately, there are no problems in which free surfaces or planes of symmetry with
antisymmetric motion are present since the pressure must be zero on these boundaries.
Using Phantom Structural Boundaries
In many cases an ELIST boundary is desired where no CQUAD4 or CTRIA3 elements exist.
Examples occur:
When grid points connected by other element types, such as CHEXA, CSHEAR,
CBEAMs, form the boundary.
When only a small hole connects two separate fluid volumes (since poor
numerical conditioning (less roundoff) is expected with a single MFLUID).
If the fluid contains two free surfaces due to entrapped air at a different pressure.
If the fluid volume is a complex labyrinth, such as a boiler made with many tubes.
The phantom boundary may be constructed from CTRIA3 and CQUAD4 elements which
have zero or nearly zero stiffness in the normal direction. The PSHELL bending thickness
is used for this purpose. The membrane stiffness is optional. If no other structural
elements are present, a small stiffness is desired for both directions to avoid automatic
constraints.
Note that this method will couple only the fluid displacements in the normal direction.
Edge/corner effects and tangential motions will be approximate.
Gravity Effects
The free surfaces produced by the Virtual Fluid Mass option are simple planes of
antisymmetric motion with a null pressure assumed at the location of the free surface
plane. Effects such as fluid sloshing due to gravity waves are assumed to be uncoupled
from the higher frequency structural modes.
An approximation to the gravity effects for fluids with finite boundaries may be modeled
using a phantom boundary instead of a free surface. Spread grid points and plate elements
over the surface and constrain the in-plane motions and rotations to zero. Give the plates
a small membrane thickness and no bending material property. Add scalar springs
(CELASi) in the direction normal to the plane with stiffnesses, , calculated from
Eq. 4-50
where is the area under the point, is the density, and is the gravitational constant.
Note that these springs will affect the rigid body motion of the whole system and this
method should be used with caution. Another drawback is that the extra gravity boundary
may result in a completely enclosed fluid volume.
A
i
n
i
i
)
u
i
0 =
i A
i
u
i
K
i
K
i
A
i
g =
A
i
g
1 CHAPTER 4
Examples
The legal and illegal types of fluid boundaries are illustrated in Figure 4-4. The categories
are GOOD = legal configuration, BAD = illegal boundary, and MAYBE = conditionally
legal, which could be permitted if a phantom boundary is used.
Figure 4-4 Fluid Combinations
Fs
GOOD BAD MAYBE
Air
Air
Fs
Air
Fs Fs
Fs
Fs
Fluid
Fluid
Fs
Air
Fluid
Hole
Infinite
Fluid
Free
Surfaces
Internal
Only
PB = Phantom
Boundary Between 2
Fluids
Air
Fluid
Fluid Fluid
PB
Fluid
Fs
Fluid
PB
Fluid
Fluid
PB
Fluid
Fluid Fluid

4.6 Coupled Acoustic Analysis
Acoustic analysis in MSC.Nastran is used to determine vibrations in structural cavities
filled with fluid or air. The boundaries of the cavities may be defined as rigid, open,
damped, or flexible as defined by constraints and finite elements. Typical applications are
automobile or aircraft interiors, musical instruments, loudspeaker boxes, and solid rocket
motors. Any type of problem that involves small, linear motions of the fluid and structure
may be a candidate. The analyst may obtain natural frequencies and damping factors
along with mode shapes for both the structure and the coupled fluid. Other solution
sequences may be used to obtain results for forced response in the frequency or time
domain.
The analyst starts with the geometry of the the acoustic cavity boundary. The interior fluid
volume is modeled in three dimensions with conventional MSC.Nastran solid finite
elements connected to a special set of GRID points. Boundary motions are included if the
structure lies on the boundary, or fixed if no structural elements are connected (by default).
The structural interface simply requires that finite elements lay on the exterior fluid
boundaries. Note that coupled interactions between the fluid and structural model are
generated internally using the geometry definitions of the elements of the two models and
require very little user intervention.
Other types of boundaries that are available are free surfaces, acoustic barriers, and
acoustic absorbers. The free boundary is modeled by simply constraining the fluid
pressure degrees-of-freedom to zero. The barrier and absorber elements provide a
spring/mass/damper system between the fluid and the structure.
Limitations. Very few limitations are placed on the analyst. Small motion theory
prohibits the use of the models in high velocity flows (more than Mach 0.5) such as
mufflers and jet engines. Exterior fluids may be modeled with extra effort by extending the
fluid model out to a large distance and by providing a damped and/or constrained
exterior. Surface sloshing effects caused by gravity such as those in fuel tanks can be
modeled with additional effort by simulating the gravity effects on the free surface with
scalar springs.
A formal theoretical discussion is provided in Additional Topics on page 529 of the
MSC.Nastran Reference Guide.
Building the Models
The following steps are recommended for avoiding difficulties later.
1. Define the geometry and select the mesh sizes.
The first step in the modeling process is to estimate the overall FE mesh size and
the level of detail needed in the geometric approximations. A set of
three-dimensional elements and special fluid GRID points (with CD = 1) are
used to define the fluid while normal structural elements and points define the
container. A rule of thumb for fluids is that six-elements-per-wave are acceptable
for the highest frequencies exciting the fluid. For acoustics this usually translates
1 CHAPTER 4
into a corresponding coarse mesh. For incompressible fluids, the only waves
occur due to gravity effects on the free surface. For heavy liquids, the mass effects
are usually more important than with air, and the fluid FE mesh should be
modeled to match the structural mesh.
It is generally easier to build the three-dimensional fluid mesh first and then use
the outer faces to model the structure (which is usually modeled with plates).
Unfortunately, most engineers start with the structure that has been used for
another analysis and encounter difficulties when they attempt to construct a
three-dimensional FE model within the enclosure. In most cases, the fluid model
will have a different number of points on opposite faces and the engineer cannot
construct a simple, matching mesh without having leftover points and lines. In
this case, he should proceed with a simple fluid model and let MSC.Nastran
control the mismatched connections.
Although MSC.Nastran is more efficient when the structural GRID points
coincide with the fluid boundaries, it will process misaligned meshes when the
fluid faces are larger than the adjacent structural elements. The basic model
interface definitions are defined on the ACMODL Bulk Data entry. The user may
define the sets of points on the interface and a tolerance to define the distance
allowed between a fluid surface element and a connected structural point. This
tolerance will be important when several points are close together and only one
is wetted by the fluid.
2. Building the Fluid Finite Element Model
Each GRID point that defines the fluid mesh will have one degree-of-freedom
defining the pressure at a point at a fixed location. The fluid itself is modeled with
existing three-dimensional connection data (CHEXA, CPENTA, and CTETRA
Bulk Data). The existence of a fluid is defined on the property data (Field 8 of the
PSOLID Bulk Data) which also references the ID of the fluid material properties
(on a MAT10 Bulk Data entry).
User-input fluid coefficients are density, bulk modulus, and/or compressible
wave speed. Incompressible fluids are approximated by using a large bulk
modulus or wave speed. (A typical value for incompressible fluid would be a
wave length of 10 times the size of the structure.) The integration options on the
PSOLID data should be full Gaussian Integration with no shear reduction
scheme.
The higher-order elements (with midside nodes) will generally give better results
than the basic linear elements. The CHEXA, CPENTA and CTETRA elements can
be connected together without the problems that occur in structural elements.
However, mixing high-order and low-order elements is not recommended.
At this stage in the modeling process, it is advisable to set up the loading and
desired output points on the fluid. Acoustic sources such as a small loudspeaker
may be specified directly by scalar loading functions (ACSRCE, SLOAD, DAREA,
and/or RLOAD Bulk Data) without the need to model the structural details. The
output displacement printout will actually be the scalar pressure values.

Guidelines on building the models are different for the fluid elements than for
normal MSC structural elements. The lack of shear locking allows larger aspect
ratios and more distorted shapes for individual elements. Unlike structures, the
fluid CHEXA elements have no particular accuracy advantages over the CTETRA
and CPENTA elements. However, in theory, fluid finite element accuracies will
be more sensitive to sharp corners and abrupt openings. In these cases, a model
should be refined using smaller elements around the discontinuity.
3. Automatic Boundaries
Fluid boundaries may be fixed, flexible, or free. With no applied loads, the outer
surfaces of the fluid elements define a fixed, rigid boundary condition,
approximating a normal pressure gradient of zero. However, if structural finite
elements are attached, they will generate flow into the fluid and affect the
pressures. The pressure degrees-of-freedom may be treated exactly like
displacement DOFs, i.e., constraints may be necessary to fix the pressure of an
open boundary to zero. Also, they may be constrained with SPC or MPC Bulk
Data for various other reasons, such as symmetry.
If no structure is connected and the problem involves only fluid, then the effects
of the boundary flexibility may be ignored and the problem is greatly simplified.
A stand-alone fluid model may be analyzed by any of the standard MSC.Nastran
solution sequences. For instance, the natural frequencies or resonances may be
obtained from SOL 103 using symmetric real methods at greatly reduced costs.
The user controls the structural interfaces with the ACMODL Bulk Data and the
locations of the boundary GRID points. The input defaults are recommended that
automatically prompt MSC.Nastran to search all elements for potential interfaces.
If only a set of fluid or structural points are desired to be connected, the searching
process can be restricted to user-selected points defined on a SET1 Bulk Data
entry. When identical meshes are used, an MSC.Nastran branch is also provided
to generate area factors more efficiently and accurately.
Problems may occur when plate elements are wetted on both sides by the fluid
and a single structural point must be connected to both sides. The recommended
method for this case is to use separate fluid points for each side of the panel with
no physical separation. (The thickness is ignored.) No provision is made for
other special cases such as flow through holes in the plates. No surface friction,
surface tension, gravity, or other indirect boundary effects are calculated for the
simple fluid-structure interface. Damping effects due to acoustic surface
materials are described in the next section.
4. Absorbers, Barriers, and Panels
The acoustic absorber and barrier are used for advanced analysis when special
boundary effects are needed to model soundproofing materials on the structural
surface or baffles within the fluid.
1 CHAPTER 4
The absorber elements (CHACAB and PACABS Bulk Data ) are used to attenuate
the reflections of the acoustic waves when they encounter the structural
boundary. They are connected between the structure and an additional set of
displacement GRID points which serve as the actual fluid interface. They provide
a simple tuned circuit with a mass on the point, and a spring and damper in
parallel connected to the structure.
The recommended procedure to implement the CHACAB elements is to move all
of the fluid boundary points a small distance away from the surface to avoid the
automatic connections. An additional set of GRID points corresponding 1:1 with
the structural points are placed on the new fluid boundary to replace the
structural points. These absorber points are constrained to allow only normal
displacements and connected to the structure with CHACAB elements. Note that
the damping factors are important in complex eigenvalue and frequency
response analysis but will be ignored in real eigenvalue analysis.
The barrier elements CHACBR are similar except that they provide masses on
both sets of points and have no damping matrix. They may be used without
structural elements to simulate heavy acoustic baffling panels that have little
stiffness in the frequency of interest. Either side may be connected to fluid
boundaries or one side may be constrained.
The PANEL Bulk Data entry is a convenient method for combining sets of
structural elements for output purposes.
Setting Up MSC.Nastran
A number of solution control parameters and commands may be used to guide the acoustic
solutions. These are as follows:
Executive and Case Control
The solution sequences that currently support the 3-D acoustic boundaries are the
Structured Dynamic Sequences, SOLs 107 through 112. These include both
modal and direct formulations for transient and frequency response, as well as
complex eigenvalue analysis. Superelement controls and restarts are automatic
in these solutions and no special data is needed for the acoustics.
The Case Control data should be set up as a normal dynamics solution. A
CMETHOD = request is necessary to obtain coupled eigenvalues. DLOAD and
FREQ requests are necessary to analyze forced response in the frequency domain.
DLOAD, IC, and TSTEP requests are necessary for transient analysis. The main
concern here is to avoid costly runs caused by an excessive number of time steps
or frequencies. Be aware that the unsymmetric dynamic solution matrices with
complex terms used in this system may run several times longer per solution
point than a corresponding structural model.

Estimating Job Costs
Although the fluid GRID points will only add one degree-of-freedom per point to
the model a three-dimensional mesh in the fluid can be several times the size of
the basic structural matrix. Furthermore the boundary coupling terms are
unsymmetric and probably damped causing more numerical processing. As a
rough estimate, each matrix solve step will be 4X to 10X the cost of a structural
static solution. In direct frequency analysis each frequency requires a matrix
decomposition. The complex eigenvalues will require approximately 2 to 5
solves. The transient solutions are dominated by vector operations which are
several times faster per time step. However, most transient problems require
many time steps for adequate accuracy.
Run Strategies
For large models, the superelement (SE) reduction methods are recommended to
reduce the size of the solution matrices. The interior points of the fluid and the
non-boundary structure grid points may each be defined as a separate Tip SEs,
which are reduced to a smaller solution size, while the boundary points remain
in the residual SE. The residual SE may also contain generalized modal
displacements in place of the interior of both fluid and structural points by using
the modal synthesis method.
For moderately large problems, the ASET/OMIT method of reduction is available
for both the fluid and structural degrees-of-freedom. However, the accuracy is
highly dependent on the analysts choice of retained degrees-of-freedom. For
instance, if all interior fluid points were omitted, the internal acoustic waves
would not be sinusoidal and the waves would instantly travel from end to end.
The most effective application would be a highly refined structural model
connected to a coarser 3-D fluid model. Removing most of the extraneous
structural DOF (i.e., rotations, tangential motion, and nonwetted points) could
reduce the costs significantly.
The recommended first analysis for the coupled problem is the Direct Complex Eigenvalue
Solution Sequence (SOL 107). This solution will indicate the overall dynamic behavior
dominated by the lowest frequency natural modes and resonant frequencies. The
following are recommendations:
1. The EIGC input to control complex eigensolutions is substantially different from
the real mode EIGR data. Read and understand the input definitions. Avoid the
default values.
2. The frequency range definitions are specified in units of Radians per Second.
Multiply the frequencies of cycle per second by .
3. Do not use damping on the initial runs. The results will be easier to debug and
the job will run faster.
2
1 CHAPTER 4
4. Beware that an additional mode with zero frequency occurs in closed containers.
This mode causes problems with both INVP and LANC methods. Avoid it by
specifying a lower frequency range nearly as large as the lowest nonzero mode.
An alternative is to add a CELAS spring connected between a fluid point and
ground.
The HESS method should be used only with some form of matrix reduction for
typical problems. See the comments above regarding the use of ASET/OMIT an
Superelement partitioning.
Running the Jobs
Several diagnostics will be generated automatically by MSC.Nastran. Examples are shown
in the following pages.
The following special output is produced by the program when acoustic boundaries are
encountered. The first shows the automatic connections found by matching GRID
locations:
The following message is a sample of a case when the boundaries do not match:
*******************************************************
* *
* MATCHING GRID TABLE *
* *
*******************************************************
* USER GRID IDS. *
* ACOUSTIC GRIDS * STRUCTURAL GRIDS *
* * *
*******************************************************
* * *
* 1 * 1001 *
* * *
* 2 * 1002 *
* * *
* 3 * 1003 *
* * *
* 4 * 1004 *
* * *
* 5 * 1005 *
* * *
* 11 * 1011 *
* * *
ETC
*******************************************************
*** USER WARNING MESSAGE 6151, ACOUSTIC COUPLING
MATRIX WILL NOT BE CREATED FOR FACE = 42 82 83 43
OF ELEMENT ID = 42 ,BECAUSE ITS FLUID GRID POINTS DO NOT HAVE CORRESPONDING
STRUCTURAL GRID POINTS.
USER ACTION: IF COUPLING IS DESIRED, THEN CHECK GRID POINT DATA.
*
*
*
*
ETC....ETC.

and:
The following message indicates the error factors in area for free body motions. Nonzero
numbers indicate a hole in the model as indicated by the T2 value.
Restarts
The only major reasons to restart a fluid/structure interaction job are to use
superelement processing and/or to change the output requests. In typical jobs
the cost in the solution phase is more than 50% of the run, and a restart for a model
or load change is not worth the cost of saving a large database.
If the uncoupled structural model is large compared to the size of the actual coupled
surface, it may be defined as one or more Tip superelements and run as a separate initial
job. The interior points in the fluid may also be treated in this manner. Many acoustic
analysis runs can then be restarted each from a copy of the initial database without
recalculating the large matrices.
Print and plot output requests may need many changes after the initial runs are made. An
output-only restart can be very cost efficient on a large job. Note that most graphical
displays do not display complex numbers on model views (i.e., contour plots or deformed
shapes). Printed output of selected grid point or element sets may be necessary and is
recommended.
Diagnosing Problems
Most of the problems in starting to use acoustics are expected to occur in the definition of
the area factors which define the fluid/structure boundary. Some restrictions are
necessary to control errors in the case of overlapping fluid and structural FE meshes, and
in most cases the user will be warned of discrepancies. Some recommended techniques are
as follows:
Carefully check the special diagnostic outputs shown above. Holes in the
boundaries will cause nonzero resultant area factors.
Run tests without the boundary coupling to estimate the frequencies.
Temporarily switch to an alternate eigenvalue method and/or a smaller range to
ensure modes are correct.
If a model is large, try subdividing it into smaller sections that can be checked
more conveniently.
^^^ DMAP INFORMATION MESSAGE 9055 (SEMG) - THE FL./STR. INTERFACE CHECK IS FORCES AND MOMENTS
RESULTING FROM A UNIT INCREASE IN PRESSURE, OR CHANGES IN THE FLUID PRESSURE RESULTING FROM RIGID BODY
MOTIONS OF THE STRUCTURE. THESE VALUES ARE DIRECTLY PROPORTIONAL TO THE OPEN SURFACE OF THE FLUID.
FL./STR. INTERFACE CHECK
T1 T2 T3 R1 R2 R3
1 2.3554080E-18 -8.4260993E-02 -1.4653091E-28 6.0271138E-10 2.6100844E-19 1.8405548E-10
1 CHAPTER 4
Try changing the fluid density to a large number (e.g., water) and run with free
structural boundaries. The fluid should not restrain the free-body motions and
these modal frequencies should remain at zero.
Advanced Methods
Many of the other MSC.Nastran features should be available for acoustic analysis. Some
potential advanced modeling and solution features that will be useful are as follows:
Modal formulation with superelements: The use of Component Mode Synthesis
is recommended for reducing the size of the matrices and the solution costs. The
degrees-of-freedom of the interior fluid points will be replaced by the modes of
the fluid with rigid container boundaries. The uncoupled structural points may
be replaced by modal coordinates defined by a structure in a vacuum.
Virtual Fluids: The MFLUID Bulk Data input may be used to represent
incompressible fluids attached to the structure. This may be used to approximate
the low frequency effects of an infinite exterior fluid.
Decoupled Response Solutions: In most acoustic problems with a load-carrying
structure, the structural modes are not affected by the small mass of the air. In
turn, the high stiffness of the structure is effectively rigid for the purposes of
computing acoustic modes in the air. Forced vibration problems may then be
analyzed as two decoupled symmetric solutions:
Constrain the entire structural model and perform a forced response
analysis on the air (use SLOAD inputs to define the volume inputs from a
known source) and save the resulting boundary pressures.
Restart the job with the fluid points constrained and a DMAP alter to add
the pressure forces to the frequency- or time-dependent load vectors. The
resulting structural model will respond to the pressures from the first run.
Random and Response Spectrum Methods: The MSC.Nastran random analysis
options will also be useful for acoustic analysis. This capability requires a set of
unit forced responses in the frequency domain (SOLs 108 and 111) and spectral
densities of the forcing functions (RANDPS inputs). It will calculate the
combination spectral densities and autocorrelations of selected output quantities.
In acoustic problems, it could be used for problems having several correlated
sources such as an automobile interior.

Figure 4-5 Example of Two-Dimensional Automotive Model With Absorbers
Figure 4-6 Acoustic Suspension Loudspeaker
1 CHAPTER 4
Test Problem Description
A simple test problem illustrates the fundamentals of acoustic analysis for enclosed
containers with flexible walls. The physical problem represents an initial attempt at design
and analysis of an enclosure for an acoustic suspension loudspeaker system. The objective
was to calculate the resonant frequencies and responses of the system without the mass
and impedance of the speaker coil and magnet components.
Physical Description
The structure consists of a simple rectangular box with wood walls as shown by the plate
model in Figure 4-6. A single cutout is provided for the speaker and a thin polyethylene
cone was modeled with triangular shells. The physical properties of the model are listed
below. Note the use of the MKS system of units, which illustrates the flexibility of
MSC.Nastran.
Executive and Case Control
The Executive and Case Control inputs are used for selecting the type of solution, loads,
boundaries, and outputs. Several different runs using the same model were made to check
results and investigate the effects of structural coupling. The configuration shown below
was a Direct Complex Modal solution using the Lanczos method to extract eigenvalues.
An SPC request was used to constrain the structure and an output set was defined to limit
the printed displacements. All other data was normal, recommended control and label
input.
Bulk Data
The key data of note in the Bulk Data section are as follows:
1. The ACMODL input uses the minimum form which connects fluid and structural
points at identical locations.
2. PARAM,NEWSEQ,-1. Turns the resequencer off.
3. Several eigenvalue methods were used. It is recommended that complex roots
always be checked by another method.
4. The speaker cone consisted of a light, thin plastic material. The virtual mass of the
exterior air was estimated and added as nonstructural mass on the PSHELL input.
Width: 0.5 M Depth: 0.4 M
Height: 0.6 M Hole Diam. 0.345 M
Box E Modulus: 11.61E9 Box Density: 562 Kg/M**3
Wall thickness: 0.015 M
Air Wave speed: 344 M/Sec Air Density: 1.11KG/M**3
Cone E Modulus: 3.4E9 Cone Density: 450.0
Cone thickness: 0.1E-3 M Cone Depth: 0.04 M

Results from MSC.Nastran
Three runs (out of many) are described below. They illustrate a recommended sequence
for the analysis process.
Run 1: Real Eigenvalue Analysis
Before the coupled structure is analyzed, it is important to understand the behavior of the
structural and fluid models separately. Fortunately they can be included in the same data
file and the Real Eigenvalue Solution Sequence (SOL 103) may by used. The results for the
uncoupled speaker box and the acoustic modes are shown below. Note that the natural
frequencies for both systems occur in the same range, which indicates that the enclosure
modes will interact with the acoustic cavity resonances. Results for the acoustics were
checked by one-dimensional wave solutions. Results for the box were checked against
calculated natural frequencies of simply-supported plates.
-etc.-
Run 2: Complex Eigenvalue Analysis, Coupled Modes.
The coupled natural frequencies must be obtained from an unsymmetrical matrix equation
which requires a Complex Eigenvalue method even in the undamped case. The CLAN
method is recommended for most large-order non-superelement jobs. The only changes
from the Real Modes job was the addition of an ACMODL input and the changes in
eigenvalue method.
Mode No.
Frequency,
Hz
Type
1 0.0 Air- Constant Pressure (fictitious)
2 292 Air- 1st z (up/down)
3 297 Box- u(y) Back panel
4 351 Air- 1st x (Left/Right)
5 359 Box- u(x) Sides in-Phase
6 379 Box- u(x) Sides out-of-phase
7 450 Box- u(z) Top/bottom
8 453 Air- 1st y (Fore/Aft)
Mode No. Frequency Type of Dominant Motion
1 0.0 Air- Constant Pressure
2 291 Air- u(z)
3 295 Coupled- u(y)- back panel
4 336 Coupled- u(x)- sides
5 371 Coupled- u(y)- cone
1 CHAPTER 4
Results above are interpreted by examining the frequency shifts from the uncoupled
system. The modes with u(y) fore-aft motion are most likely to be excited by the speaker.
Modes 1, 2, 3, 4, 6, and 7 are close to their uncoupled equivalent. Modes 5 and 8 are new
combinations of higher modes. The results were confirmed by several methods:
1. The job was rerun using the INV method on the EIGC Bulk Data and produced
nearly identical results.
2. A printout of the interface area matrix was obtained by using a DMAP alter and
verified by hand calculations.
3. The natural symmetry of the geometry produced symmetric and antisymmetric
results relative to the natural structural planes of symmetry (except for roots with
close frequencies).
Run 3: Complex Eigenvalue Analysis, Modal Formulation
A more efficient method for solving a coupled matrix problem is to reduce the size of the
matrices by using modal coordinates. In this case, both the structural displacements and
fluid pressures are replaced in the matrix solution by generalized coordinates representing
the uncoupled real modes. For the modal cases below the real modes (25 total) below
800 Hz. were used. In the coupled cases 13 modes were obtained below 500 Hz. All results
compared to within 1%.
On a Sun SPARCstation 1+ the CPU time comparisons are:
The conclusion is that the modal method (SOL 110) will reduce the costs with very little
effect on accuracy. The user is cautioned to use a liberal number of real modes (twice the
number of coupled modes) to represent the system.
6 373 Coupled- u(x)- sides
7 445 Coupled- u(z)- top/bottom
8 453 Coupled- u(y)- cone
Type Solution Solution Size CPU Time (Seconds)
Uncoupled Modes 538 110.2
Direct, No Modes 538 586.3
Struct. Modes Only 114 143.2
All Modes 25 106.9
Mode No. Frequency Type of Dominant Motion

ACOUSTIC TEST PROBLEM DATA FILES
ID SPEAK39F,DNH
SOL 107
DIAG 8,12 $ PRINTS MATRIX TRAILERS AND ROOT-TRACKING MESSAGES.
CEND
TITLE=SPEAKER BOX -WITH CONE, SIMPLE CORNER SUPPORTS
SUBTITLE = COUPLED BOUNDARY, NON-MATCHING ELEMENTS
ECHO= UNSORT
SEALL=ALL
SPC=20
$ USES MKS SYSTEM
CMETHOD = 7 $LANCZOS
SET 20= 3,13,23,43,82,83,84,91,93,95,103,113,123,
131,135,153,163,171,173,175,183,
1013,1023,1043,1082,1083,1084,1091,1095,1113,
1131,1135,1163,1173,1183,1193
DISP= 20 $ FOR MINIMUM PRINTOUT
$ DISP(PLOT)=ALL $ FOR MSC/XL CONTOUR PLOTS
$
BEGIN BULK
$
$ DEFINE FLUID/STRUCTURE INTERFACE: COINCIDENT POINTS
ACMODL,IDENT
$
PARAM,POST,0
PARAM,COUPMASS,1
$ SUPPRESS SEQUENCER
PARAM,NEWSEQ,-1
$
$ BOX PROPERTIES - WOOD
MAT1,11,11.61+9,,0.3,562.0
PSHELL,1000,11,.015,11,,11
$
$ SPEAKER CONE
MAT1,3,3.4+9,,0.3,450.0
PSHELL,10,3,0.1-3,3,,,0.223
$
$ PROPERTIES OF AIR
MAT10,100,131.94+3,1.115
PSOLID,100,100,,2,,1,PFLUID
$
$ EIGEN METHODS
EIGR,20,MGIV,1.0,600.0
EIGC,7, CLAN,,,,,,,+CLAN
+CLAN,0.0,10.0,0.0,1600.0,100.0,,20
$ USE INVERSE POWER TO CHECK LOW ROOTS
EIGC,107,INV,MAX,,,,,,+EC1
+EC1,0.0,100.0,0.0,1800.0,100.0,12,9
$
$ FIX BOX AT BOTTOM CORNERS
SPC1,20,123,1001,1005,1031,1035
$ COORDINATE SYSTEM AT CENTER OF HOLE
CORD1C,83,283,113,85
$
$
$FLUID GRID POINTS
$ NOTE VALUE OF -1 IN FIELD 7 INDICATES 1 DOF.
GRID,1,, -.25, -.2, -.3,-1
GRID,2,,-.125, -.2, -.3,-1
GRID,3,, 0.0, -.2, -.3,-1
GRID,4,, .125, -.2, -.3,-1
1 CHAPTER 4
GRID,5,, .25, -.2, -.3,-1
GRID,11,, -.25, -.0667, -.3,-1
GRID,12,,-.125, -.0667, -.3,-1
GRID,13,, 0.0, -.0667, -.3,-1
GRID,14,, .125, -.0667, -.3,-1
GRID,15,, .25, -.0667, -.3,-1
GRID,21,, -.25, 0.0667, -.3,-1
GRID,22,,-.125, 0.0667, -.3,-1
GRID,23,, 0.0, 0.0667, -.3,-1
GRID,24,, .125, 0.0667, -.3,-1
.25, -.2, 0.,-1
-
ETC
-
$ STRUCTURE GRIDS
GRID,1001,, -.25, -.2, -.3
GRID,1002,,-.125, -.2, -.3
GRID,1003,, 0.0, -.2, -.3
GRID,1004,, .125, -.2, -.3
GRID,1005,, .25, -.2, -.3
GRID,1011,, -.25, -.0667, -.3
GRID,1012,,-.125, -.0667, -.3,,6
GRID,1013,, 0.0, -.0667, -.3,,6
-
ETC.
-
$ OPTIONAL ASET DATA TO USE WITH HESS METHOD
$ASET1,1,1,THRU,195
$ASET1,123,1003,1012,1014,1022,1024,1033
$ASET1,123,1163,1172,1174,1182,1184,1193
$ASET1,123,1042,1043,1044,1081,1082,1083,1084
$ASET1,123,1065,1122,1123,1124
$ASET1,123,1072,1074,1112,1113,1114,1152,1154
$ASET1,123,1051,1061,1131,1141
$ASET1,123,1055,1065,1135,1145
$
$
$SOLID ELEMENTS FOR AIR
CHEXA 1 100 1 2 12 11 41 42 +001
+001 52 51
CHEXA 2 100 2 3 13 12 42 43 +011
+011 53 52
CHEXA 3 100 3 4 14 13 43 44 +021
+021 54 53
CHEXA 4 100 4 5 15 14 44 45 +031
+031 55 54
CHEXA 11 100 11 12 22 21 51 52 +041
+041 62 61
CHEXA 12 100 12 13 23 22 52 53 +051
+051 63 62
CHEXA 13 100 13 14 24 23 53 54 +061
+061 64 63
CHEXA 14 100 14 15 25 24 54 55 +071
+071 65 64
CHEXA 21 100 21 22 32 31 61 62
+081
+081 72 71
-
ETC
-

$STRUCTURAL ELEMENTS
CQUAD4 1001 1000 1001 1002 1012 1011
CQUAD4 1002 1000 1002 1003 1013 1012
CQUAD4 1003 1000 1003 1004 1014 1013
CQUAD4 1004 1000 1004 1005 1015 1014
CQUAD4 1011 1000 1011 1012 1022 1021
CQUAD4 1012 1000 1012 1013 1023 1022
CQUAD4 1013 1000 1013 1014 1024 1023
CQUAD4 1014 1000 1014 1015 1025 1024
-
ETC
-
$
$SIMPLE CONE
$
CTRIA3, 5042,10,1042,1043,1083
CTRIA3, 5043,10,1043,1044,1083
CTRIA3, 5044,10,1044,1084,1083
CTRIA3, 5084,10,1084,1124,1083
CTRIA3, 5124,10,1124,1123,1083
CTRIA3, 5123,10,1123,1122,1083
CTRIA3, 5122,10,1122,1082,1083
CTRIA3, 5082,10,1082,1042,1083
ENDDATA
CHAPTER
5
Dynamic Solution Techniques
I Overview
I Review of Dynamic Excitations
I Complex Eigensolutions
I Nonlinear Transient Response Analysis
I Fourier Transform
I Viscoelastic Material Properties
I Free Body Techniques
I Aeroelastic Solutions
I Dynamic Optimization and Design Sensitivity

5.1 Overview
While planning the modeling process and making provisions for the reduction steps and
boundary conditions, the dynamics engineer should also consider the various options in
the actual solution. In addition to the basic linear transient and frequency response
processes, which are described fully in the MSC.Nastran Basic Dynamics Users Guide, other
solution options and variations for dynamic analysis are available.
Complex eigenvalue analysis is used for determining stability and assessing the overall
dynamic response of a system. Complex eigenvalue analysis must be used for eigenvalue
analysis when unsymmetric matrix terms or large damping effects are present.
Nonlinear transient analysis accounts for nonlinear element effects caused by nonlinear
materials, variable surface contact, and large displacements. The procedures for the data
input and controls are significantly different than those for the basic linear transient
solutions.
An option for using Fourier transform methods is available for the solution of special
problems that involve repeating loads in time.
Viscoelastic parameters, which can be applied in the solution of problems involving
rubber-like materials, are available as a special process in the direct frequency response
solutions.
The solution methods for free bodies are both necessary analysis procedures for
unconstrained models and diagnostic tools for constrained bodies. For instance, an
airplane or missile in flight (without proper supports) will have a singular stiffness matrix
and (will) produce large displacements in the results.
Finally, aeroelastic solutions such as flutter analysis, gust response, and static aeroelasticity
are also discussed in this chapter because of their unique algorithms and solution
procedures.
1 CHAPTER 5
5.2 Review of Dynamic Excitations
The methods used for generating dynamic loads in MSC.Nastran are very different from
those used for static loads. Dynamic loads generally vary with time or frequency. They
may also be applied with different phases or time lags to different portions of the structure.
The most general case would be a structure having a different load history for each
direction on each point on the model, which could require thousands of tabular inputs.
Fortunately, most applications require only a simple pattern of load variations with time
and geometry. The engineer usually encounters a single time function applied to a portion
of the structure or a general time function applied to a few structural points. An example
of the first case would be wind gust loads on an aircraft. An example of the second case
would be an automobile traveling over a rough road. MSC.Nastran requires only a
minimum input for either case.
Because of the many possible types of dynamic loading conditions, the input options are
also numerous. Most of the basic options are described in Modal Versus Direct
Frequency Response on page 114 and Modal Versus Direct Transient Response on
page 162 of the MSC.Nastran Basic Dynamics Users Guide. Some of the advanced methods
are described below.
Subcases in Dynamic Analysis
The Case Control options in dynamic analysis are less flexible than statics and nonlinear
analysis. Some of the statics options that are not available in dynamic response analysis
and complex eigenvalue analysis include:
Changes in boundary conditions between subcases.
SUBSEQ and SUBCOM solution combinations.
Grid point forces and element strain energy outputs.
Otherwise, each type of solution uses the Case Control subcase for a different purpose.
Some uses for subcases in dynamics analysis are given below.
1. Subcases are essential for most dynamic optimization problems (SOL 200). Each
subcase may define a different type of analysis or a different load.
2. In a superelement analysis, the subcases may be used to specify different
upstream LOADSET data and output processing for different SE components.
3. Subcases are recommended in nonlinear transient analysis (SOL 129) for applying
loads in a sequence or changing the time step method. A single load may be
applied in steps or a second load may be added with new subcases. Changing
time step size or the solution method may be necessary to overcome a
troublesome divergence region. Subcases also define the allowable solutions that
may be used in restarts. Use them liberally.

4. Multiple subcases are available for frequency response analysis (SOLs 108, 111
and 118) for the purpose of solving multiple loading conditions more efficiently.
(Each frequency requires a matrix decomposition and each additional load vector
may be processed at this time with small cost.) Another use is in random analysis
where several loads need to be combined, each with a different spectral density
distribution.
5. Subcases are used in complex eigenvalue analysis (SOLs 107 and 110) for
processing multiple direct input matrices. Typical applications are multiple
control system parameters or externally generated aerodynamic matrices
representing a variety of flight conditions.
6. Cyclic symmetry analysis (SOLs 114 through 118) requires subcases to distribute
the loads over different petals (i.e., sections of the structure). The methods are
identical to cyclic static analysis.
7. Multiple subcases are not recommended for linear transient analysis.
Loading Methods
An overview of the dynamic load algorithm is shown in Figure 5-1. The input data follow
two major paths: the DAREA method versus the LSEQ method. The details of the data
inputs are given in the appendices and in the MSC.Nastran Basic Dynamics Users Guide.
1 CHAPTER 5
Figure 5-1 Input Data Relationships for Dynamic Loads
The LSEQ Bulk Data act much like a set of Case Control commands to generate
static loads and assign them to a specific static load vector. These vectors are
assembled, reduced, and combined for all superelements. A different load vector
may be required for each unique function of time or frequency.
Case Control
Bulk Data
FORCEi
PLOADi
GRAV
etc.
..
LSEQ
DAREA
DPHASE
DELAY
Load Vectors
P(x,y,z)
TLOADi
RLOADi
DLOAD
TABLEDi
F(t) or F(f)
Dynamic Loads
P(t) or P(f)
Available to Superelements
LOADSET =
DLOAD =
Dynamic
Coefficients
Static Loads
Time or Frequency
Functions

The DAREA, DPHASE, and DELAY Bulk Data provide a direct method of
distributing the dynamic loads over the grid points. The DAREA factors
correspond to a specific load factor for a specific degree-of-freedom, much like an
area under a pressure load. The DPHASE and DELAY data are used when the
same load versus time function is applied to different points at different times.
An example of a DPHASE application is the frequency response of an automobile
traveling over a highway with a sinusoidal undulation. As the speed increases
the input frequency changes, but the phase difference between the axles is
constant.
The TLOADi Bulk Data inputs each define a function of time with coefficients or TABLEDi
references. The RLOADi Bulk Data inputs each define a function of load versus frequency
with complex TABLEDi inputs. The DLOAD entry is used to combine the different load
functions, either time or frequency dependent. These functions may be associated with
LSEQ-type vectors or DAREA coefficients, or both.
Recommendations
The following guidelines should be observed when applying dynamic loads.
TABLEDi
1. Remember that the tables are extrapolated at each end from the first or last two
points. If the load actually goes to zero, add two points with values of y = 0.
2. Linear interpolation is used between tabular points. This may lead to
accumulated roundoff and drift errors over a long-period transient analysis. Fix
this problem by adding a correcting load function to the same points on
subsequent runs.
3. If a jump occurs (two points with equal values of X), the value of Y at the jump is
the average of the two points.
Static Preloads
1. An initial linear static preload that is released suddenly for a transient analysis
may be performed in two stages in one run. In stage 1, apply the static loads and
use large time steps to suppress the higher frequency motions. In stage 2, remove
the loads and use small time steps to capture the dynamic action. Alternatively,
a nonlinear analysis may be restarted from a nonlinear static analysis, making the
process nearly automatic.
2. Simple static loads may be included through the dynamic load methods. Define
a TLOAD2 or an RLOADi with a constant function and add it via the DLOAD
Bulk Data input.
3. Thermal loads may be applied via the LSEQ method, but note that the dynamic
stress recovery method will not account for the additional strain or stress.
1 CHAPTER 5
Matching Initial Conditions
1. In linear transient analysis, the load at time, is always ignored. If the load
is actually a preload, the initial conditions (TIC data) should correspond to this
condition. Otherwise, if the load is being suddenly applied, the loading functions
will start at .
2. If a free body is being accelerated by an applied load, the calculated response will
also be delayed by a fraction of a time step.
Setup for Random Analysis
An MSC.Nastran random analysis requires a preliminary frequency response analysis to
generate the proper transfer functions that define the output/input ratios. The squared
magnitudes of the results are then multiplied by the spectral density functions of the actual
loads. Normally, the inputs are unit loads (e.g., one g constant magnitude base excitation
or a unit pressure on a surface).
Fluids and Acoustic Loads
Normally, fluid and acoustic elements are not loaded directly, but are excited by the
connected structures. However, loads on these special grid points may be used to
represent sources of fluid motion such as a small loudspeaker. The actual units of these
loads are the second derivative of volume flow with respect to time. See Axisymmetric
Fluids in Tanks on page 105 for details.
t 0.0 =
t t =

5.3 Complex Eigensolutions
Complex eigenvalue analysis is necessary when the matrices contain unsymmetric terms,
damping effects, or complex numbers where real modes analysis cannot be used. It is used
for the analysis of aeroelastic flutter, acoustics, rotating bodies, and many other physical
effects.
The unforced motion of a system of equations can be expressed as the sum of the motion
of its eigenvectors, oscillating and decaying or expanding with terms of the form
Eq. 5-1
The value of gives a measure of the rate of decay or divergence of the i-th natural
dynamic mode. If the value of is nonzero, it gives a measure of the rate of oscillation of
the solution. For the most general case, , which represents the shape of the mode,
contains complex numbers. The relative size of these numbers indicates which parts of the
structure are most active in this mode of motion. The imaginary parts of signify phase
differences or lag times between the degrees-of-freedom.
The results of complex eigenvalue analysis are used for tasks such as measuring the effect
of damping materials on system performance and determining the stability of a system
when it contains sources of energy such as rotating components. The complex
eigensolution is an end product in MSC.Nastran. There are no provisions for using these
shapes for modal solutions of frequency response analysis or transient response analysis,
as is done for the real eigensolution capability.
The Two Forms of Eigensolution
There are many special classes of eigenvalue problems. MSC.Nastran breaks all of them
into two major categories. The real eigenvalue problem is expressed by the two matrix
equation
Eq. 5-2
where and are real and symmetric, and is positive semi-definite. This type of
eigenvalue problem is discussed in the MSC.Nastran Basic Dynamics Users Guide. All
damping effects are ignored. It can be shown that the eigenvalues and eigenvectors
of this problem must be real, not complex, and that the eigenvectors can be made
orthonormal with respect to the mass matrix and orthogonal to the stiffness matrix; that is,
Eq. 5-3
where is the identity matrix, and
Eq. 5-4
where has the eigenvalues stored along its diagonal. This property is used to separate
eigenvectors for repeated eigenvalues into linearly independent sets.
u t ( ) Real
i
{ }e

i
i
i
+ ( )t
)
( ) =
i
u { }
K
i
M [ ]
i
{ } 0 =
K M M
i

i
t
M I =
I
t
K =
1 CHAPTER 5
Any problem whose input matrices do not meet these restrictions is called a complex
eigenvalue problem in MSC.Nastran. In general, the solution may include real or complex
numbers, and the orthogonality properties of this more general case are less powerful.
The matrix equation used for the complex eigenvalue problem is
Eq. 5-5
and
where any of the matrices may contain real or complex numbers, and may also be
symmetric or unsymmetric. If all matrices are symmetric, there is a transformation similar
to Eq. 5-3 and Eq. 5-4 that can transform all three matrices to a diagonal form. If any of the
matrices is complex or unsymmetric, only a weaker orthogonality condition applies. If all
of the eigenvalues are distinct (that is, there are no repeated roots), each matrix can be
reduced to diagonal form by also computing the left hand eigenvectors , that is, the
eigenvectors of Eq. 5-5 when all of its matrices are transposed. One orthogonality
condition for any two roots of the system happens to be
Eq. 5-6
where are the right-hand eigenvectors of Eq. 5-5. Some of the eigensolution methods
use these orthogonality conditions to improve the linear independence of the eigenvectors,
while others do not. None of the MSC.Nastran complex eigensolution methods has a
special provision for repeated roots. They may return the same eigenvector for each of the
repeated roots.
A special form of Eq. 5-5, called the shifted form, is used by some methods to provide faster
convergence in the iterations. If we introduce a shift point, , the relative distance of the
root is
Eq. 5-7
and Eq. 5-5 becomes
Eq. 5-8
One may iterate Eq. 5-8 by solving for trial vectors on the left side using previous estimates
on the right side to produce vectors corresponding to the smallest values of . This
procedure is used in the Lanczos method to obtain sets of approximate vectors and in
Inverse Power to obtain actual vectors.
The Complex Eigensolution Methods
There are three basic methods available at present. They are all based on iteration in some
form or other, with the result that their computation costs are somewhat unpredictable.
They are discussed in the order of their general utility, the most widely used first. Details
of their implementation are given in the MSC.Nastran Numerical Methods Users Guide.
K pB p
2
M + + [ ] { } 0 =
p i + =
Li
T
p
i
p
j
+ ( )M B + [ ]
Rj
{ } 0 if p
i
p
j
=
R
p
0
p p
0
=
K p
0
B p
0
2
M + + [ ] { } p
0
B 2p
0
M M + + [ ] { } =

The Complex Hessenberg Method
The basic algorithm solves the single matrix eigenvalue problem
Eq. 5-9
is transformed to an upper Hessenberg form. In this form all terms below the
subdiagonal are null. All of the roots of this transformed problem are computed by the
double QR algorithm of Francis. The eigenvectors are then found by inverse iteration.
Convergence is usually within one or two iterations because the roots are known at this
point. There are similarities in this solution process with the tridiagonal methods of real
eigenvalue analysis.
The three-matrix form is transformed to the single matrix form by premultiplying by the
inverse of the mass matrix,
Eq. 5-10
If is null then and . If is not null the velocity vector is defined from
,
Eq. 5-11
Eq. 5-10 and Eq. 5-11 are combined to form
Eq. 5-12
If is null Eq. 5-10 is used to produce n eigenvalues (only roots with are calculated),
while if exists Eq. 5-12 will produce 2n eigenvalues. The user can control the number of
eigenvectors computed in the Inverse and Lanczos methods.
The fact that must have an inverse means that may not contain null rows or columns,
or other sources of singularity. The rows and columns for these singular DOFs may be
removed by omitting them with OMITi entries, or, what is often less labor for large models,
by retaining important DOFs with independent mass with ASETi entries. All other
unconstrained DOFs are then placed in the omitted set. In general, translation DOFs with
attached masses are safe candidates for the a-set. Care should be taken with rotational
DOFs. An example of a DOF with nonnull mass but which is not independent is a rotation
DOF to which a CONM2 element is attached, and the CONM2 element has a point mass
and an offset, but zero moments of inertia. The rotation DOF will be singular when the
matrix is inverted.
No special logic is used for repeated roots, which return the same eigenvector. There is a
system cell which requests that a newer method with spill capability be used. The default
value for this system cell uses an algorithm with no spill capability. System cells can be set
on the NASTRAN command. Refer to the EIGC on page 1336 of the for details.
This method does not take advantage of matrix sparsity or banding. Solution costs are
proportional to .
A
i
I [ ] u
i
{ } 0 =
A
M
1
K pM
1
B p
2
I + + [ ] u { } 0 =
B A M
1
K =
i
p
2
= B v
u
v pu =
M
1
B M
1
K
I 0
p
I 0
0 I
+
v
u
" #
$ %
& '
0 { } =
B 0
B
M M
n
4
1 CHAPTER 5
The Complex Lanczos Method
Ritz vectors (that is, trial eigenvectors) are computed from Eq. 5-8 by inverse iteration, after
inserting a trial eigenvalue called a shift. Shift points are selected by the user. These
vectors are used to reduce the problem to tridiagonal form using the Lanczos reduction
method. The roots are computed by the QL algorithm in this reduced basis, and the
eigenvectors are computed by inverse iteration. The eigenvectors are then
back-transformed to the original basis. Both left- and right-hand eigenvectors are
computed to improve the convergence of the method. The left-hand eigenvectors are an
optional output of the eigensolution module.
The roots are computed near the shifts specified by the user. The method attempts to
compute all roots near a shift before going on to the next shift, but it can not be guaranteed
that all roots near the shift will be found. If more roots are calculated than were requested,
all are output.
This method takes advantage of the bandedness and sparsity of the input matrices.
Solution costs are proportional to where is the semi-bandwidth and the
number of eigenvectors computed. Many roots can be computed from each shift.
The Complex Inverse Power Method
The basic equation is solved by inverse iteration from the shifted equation similar to the
initial stage used for the Lanczos method. Each eigenvector and its eigenvalue are
determined simultaneously until a solution has converged to a user-specified tolerance.
The user inputs subregions in the complex plane for root searches. The same qualifications
that all roots in a region may not be found applies here, too. Costs are proportional to
because several shifts are usually required for each root found. This method is useful for
finding a few roots in a small region where poorly conditioned matrices cause problems
with the other methods.
The Modal Hessenberg Method
For many problems this method combines the best features of the real eigensolution
methods and the Complex Hessenberg method. The real eigensolutions from the
structural stiffness and mass matrices are found first. The Lanczos and SINV real
eigenvalue methods use a Sturm sequence technique that guarantees that all roots in a
range have been found or identified as being missing. The tridiagonal real eigensolution
methods also find all roots of the problem, so that none can be missing. Unfortunately,
there is no similar technique for roots in the complex plane. Using a modal approach
provides the following benefits:
1. The user can be assured that all the undamped roots of the system in a region
specified by the user have been found with any of the real methods.
2. The complex eigenvalues are usually of the same order of magnitude as the real
eigenvalues.
3. The real eigenvectors are then used to reduce the damping matrix and any other
special dynamic effects that may be present to h-set size matrices.
r 10 ( )nb
2
b r
3rnb
2

4. The h-set mass matrix is guaranteed to be well conditioned for the inverse
operation unless very strange effects are input with the M2PP matrix or the
counterpoint terms of transfer functions, so there is no need to omit massless
physical DOF, a task which can be laborious on a complicated model. (It is
necessary, however, to ensure that any extra points that are used have nonnull
mass terms.)
5. All eigensolutions are then found in this reduced basis. The user is assured that
there are no solutions which have been skipped over.
6. The economic trends are also favorable because the real eigensolutions do take
advantage of matrix sparsity and banding (the Real Lanczos Method is the best in
this respect today), and they reduce the problem to a much smaller size before the
Hessenberg solution is attempted.
However, if strong nonstructural effects are present, such as those caused by direct input
matrices or transfer functions, a large number of real modes may be necessary to account
for these forces. An example would be a servo control that acts as a large mass on the
structure. This mass could reduce the frequency of some local high frequency modes and
move them into the important range.
The Complex Lanczos and Inverse Power methods may also be used in the modal method,
but the economic rewards are less, and they may also miss roots. They should be used as
backup methods in this context, used to check the solution produced by the Hessenberg
method when there is some question about its correctness.
User Interface
Various form of the eigenvalues are used as inputs and are provided as outputs. The real
eigensolution outputs the eigenvalues in three forms. The complex eigenvalue, , is
formed from the real quantities and where . If , the radian frequency,
of complex eigenvalue analysis is the same as that of real eigensolutions. The real part,
, is a measure of the decay rate of a damped structure, or if negative, the rate of
divergence of an unstable system. The imaginary part, , is the modified frequency in
radians/unit time. However, roots with negative values of should be treated as special
terms. The output labeled f is the circular frequency in cycles per unit time. It is equal to
.
The CMETHOD = [SID] Case Control command selects the EIGC Bulk Data entry, which
has the form
1 2 3 4 5 6 7 8 9 10
EIGC SID METHOD NORM G C E ND1
ALPHAAJ OMEGAAJ ALPHABJ OMEGABJ LJ NEJ NDJ
p
p i + = 0. =
2 ( )
1 CHAPTER 5
NORM, G, and C have to do with the specification of the method of normalization. The
eigenvectors may be normalized either to a unit value at grid point G for coordinate C, or
for the largest term to be of unit magnitude. E is used to specify the convergence criterion
of the solution. Each method has a different default value for this criterion, and each is
adequate for most problems. ND1 is the number of eigenvectors to be computed by the
Hessenberg method. Data in this field is only allowed when there are no continuation
entries, while the Hessenberg method ignores the data on the continuation entries.
The shift points for the Lanczos method are defined on the continuation entries. The
recommended practice is to specify one point and at the lower bound of the
expected range of eigenvalues, but not at 0.0 A second shift may be input at and
at the upper bound of the expected range. All ALPHABj and OMEGABj must be blank.
The number of Ritz vectors computed for the J-th shift is three times the number entered in
NDJ. While each Ritz vector may be capable of defining an eigenvector, it is more likely
that some of the Ritz vectors will be discarded because they are not orthogonal to
lower-numbered Ritz vectors during the Lanczos reduction process. After the poor vectors
are discarded, the remaining vectors are used to compute the eigensolutions.
The number of modes computed from all shifts up to the j-th shift may be more or less than
NDJ. If there are less, the processing continues with the next shift; otherwise, the process
ends and all eigensolutions are output. The ratio of discarded vectors to retained vectors is
problem dependent. However, the recommended practice is to ask for all of the desired
eigenvectors for ND1 on the first shift, but put in additional shifts. If it is possible to
compute all solutions with one shift, this is the most economical option. If more shifts are
needed, they will be used.
The subregions for searching for roots with the Inverse Power method are also specified on
the continuation entries. (ALPHAAJ, OMEGAAJ) define one point in the complex plane,
and (ALPHABJ, OMEGABJ) a second point. A line is drawn between these points, and a
box of width LJ is placed around this line. NEJ is larger than the number of roots expected
in the subregion, and NDJ the number desired. This defines the first subregion for
searching for eigenvalues. More continuation entries may be used to define more
subregions. Again, if more eigenvalues are computed than are requested, all are output.
The eigenvalue output for a sample problem is shown below:
C O M P L E X E I G E N V A L U E S U M M A R Y
ROOT EXTRACTION EIGENVALUE FREQUENCY DAMPING
NO. ORDER (REAL) (IMAG) (CYCLES) COEFFICIENT
1 9 -5.806441E+01 5.750383E+03 9.152019E+02 2.019497E-02
2 8 -6.294888E+01 6.293917E+03 1.001708E+03 2.000309E-02
3 7 -6.910709E+01 6.844852E+03 1.089392E+03 2.019243E-02
4 1 -3.300980E+02 1.667092E+04 2.653260E+03 3.960164E-02
5 5 -3.565692E+02 1.823559E+04 2.902285E+03 3.910694E-02
. . .
The column labeled (REAL) contains , and the column labeled (IMAG) contains . The
column labeled (FREQUENCY) contains the circular frequency. The last column is the
damping coefficient computed from the equation
1
0.0 =
1
2
0.0 =
1

1
g ( ) 2 =

which is approximately twice the value of the conventional modal damping ratio. This
form was more popular with the aeroelastic flutter specialists who were the primary users
of this capability.
Note that if the magnitude of this term is computed to be less than , it is reset to
zero.
For small values, the damping coefficient is twice the fraction of critical damping for the
mode. The eigenvalues are sorted on , with the negative values sorted first (there are
none in this example), sorted on increasing magnitude, followed by the eigenvalues with
positive , again sorted on magnitude. Roots with equal values are sorted next on .
Modeling Techniques
All Methods
Repeated roots are usually due to planes of symmetry of a structure. For example, a pencil
has a pair of repeated roots for each bending mode, with their eigenvectors being at 90
degrees to each other. In physical structures there is never absolute symmetry because of
the effects of manufacturing tolerances, material inconsistencies, etc. The user may adopt
this point of view when repeated roots cause problems by making small perturbations to
the structure to destroy symmetry. For example, when doing an analysis for the damped
modes of a pencil, a small mass term can be added in one direction. A small perturbation
is usually adequate to separate the modes enough to eliminate numerical problems.
Hessenberg Method
1. If extra points or other DOFs are connected by only transfer functions or x2PP
matrix terms, they may need to be differentiated to provide a mass term on these
DOFs. For example, if the following equations are written for extra points and
, which are coupled to the structural DOFs and ,
In these forms, would appear in the stiffness term of the transfer function, and
in the damping term.
The first equation can be differentiated twice, and the second once, to provide
The terms are placed in mass terms slots in the transfer function, or in the M2PP
matrix.
2. Avoid coupled mass. One apparent choice for obtaining mass terms on rotation
DOFs is to request coupled mass by use of the PARAM entry that requests it (but
this is usually a poor choice). The theories for coupled mass for some elements in
MSC.Nastran do not provide rotational mass while other elements have theories
5.0 10
4

e
1
e
2
u
11
u
12
c
1
e
1
u
11
=
c
2
e
1
u
12
=
c
1
c
2
c
1
e
1
u
11
=
c
2
e
1
u
12
=
c
i
1 CHAPTER 5
that produce rotational mass in a manner that does not provide an independent
mass term, leading to either failure in the inversion of , or poor numerical
conditioning. Null masses on rotational DOFs cause no problems in modal
analysis.
In direct analysis, if all unconstrained rotations are omitted, this potential
singularity problem is avoided, with little or no loss in accuracy, and the cost of
static reduction of the system is less than the savings of solving a smaller sized
eigenproblem.
Example Problem: Complex Roots of a Friction Mechanism
This problem illustrates the modeling and solution of a model with unsymmetric matrix
inputs. Complex eigenvalue analysis can predict the existence of unstable or undamped
roots, which will be the cause of noise.
The model consists of three bodies illustrated in Figure 5-2, namely a rigid housing, a pad
with an expendable friction surface, and a moving surface contacting the pad.
Figure 5-2 Sliding Friction Complex Eigenvalue Analysis
The objective of the analysis is to determine whether the friction effects can cause the onset
of a divergent squeal mode. The following assumptions are made in the modeling:
1. The speed of the sliding surface is assumed to be much less than the speed of the
traveling vibrational waves. Therefore, the elements representing the surface
may be limited to small motions and the traveling wave effects are ignored.
2. Pure sliding friction is assumed. The magnitude of the pad vibration may be very
small for the onset of the unstable mode. The analysis will be invalid when the
vibrational velocities exceed the surface velocity.
3. A static preload is assumed to be large enough to maintain full contact on the pad
surface. The frictional coefficient is assumed to be constant. (However, it could
be varied over the contact region.)
Frictional forces on the pad and sliding surface are proportional to the normal contact
forces, which in turn may vary with dynamic response. For small dynamic motions the
relationship is
M [ ]
z
k
k
Pad
Moving Surface
V
Y
Housing

Eq. 5-13
where:
The value of is only available indirectly from the element forces or by the use of Lagrange
multipliers. The solution variables in a conventional model consist only of displacements,
velocities, and accelerations.
An approximation to may be generated from a penalty function method using large
springs to represent the contact constraint. If a large spring, (CELAS2 data), is connected
to the two normal displacements at a point, the normal force is a function of the
displacements
Eq. 5-14
where , are the displacements in the normal direction.
Combining Eq. 5-13 and Eq. 5-14, we obtain the following matrix relationship:
Eq. 5-15
Special Input
The matrix terms in Eq. 5-15 may be input to the problem for each contact point directly as
DMIG data. The resulting matrix will be unsymmetric. The Case Control command
required for DMIG input is
K2PP = FRICTN
where FRICTN is the name of a DMIG Bulk Data entry, which starts with the header entry
DMIG, FRICTN, 0, 1, 1, 0 . . .
where the inputs and their functions are as follows:
A DMAP alter is used to allow multiplying this matrix by the user parameter FRIC.
is the tangential force on the pad.
is the tangential force on the sliding surface.
is the friction coefficient.
is the normal compression force variable.
FRICTN DMIG name
0 Required for all DMIG headers
1 Input format is single precision
0 Output format depends on the type of computer being used. The best
format is chosen for each computer type.
F
py
F
sy
N = =
F
py
F
sy
N
N
N
K
N K u
pz
u
sz
( ) =
u
pz
u
sz
F
py
F
sy
" #
$ %
& '
K
1 1
1 1
u
pz
u
sz
" #
$ %
& '
etc. =
1 CHAPTER 5
Listing 5-1 shows the resulting MSC.Nastran input file. RBEi and CONM2 elements are
used to simplify the model and eliminate spurious roots. The unique inputs are the DMIG
data and the associated CELAS2 springs, (Ids = 311-314).
The Case Control for a complex LANCZOS method is shown here, with the provisions for
the other methods preceded by a $ symbol. (Lines starting with $ are treated as comments,
but are otherwise ignored.)
CMETHOD = 10 $CLAN
$CMETHOD = 11 $INVP
$CMETHOD = 12 $HESS
A METHOD command is required when using the modal method of SOL 110. This input
is for the direct method of SOL 110. The CMETHOD command is required for complex
eigensolution method selection. Eigenvector output is requested by the DISP command.
The Bulk Data entries selected by these commands are as follows:
EIGRL,1, , 4000.0, 12 $ real lanczos
EIGC, 10, CLAN, MAX, , , , , ,+ELAN1
+ELAN1, 0.0, 6000.0, , , , , 12,, +ELAN2
EIGC, 11, INV , MAX, , , , , ,+EINV
+EINV, 0.0, 10000.0, 0.0, 50000., 20000., 20, 12
EIGC, 12, HESS , MAX, , , , 12

Listing 5-1 Input Data for Complex Eigenvalue Stability Analysis
$ FILE CEADL693.DAT
INIT MASTER(S) $ DISCARD DATA BASE AT END OF RUN
DIAG 12 $ DIAGNOSTICS FOR LANCZOS
COMPILE GMA SOUIN=MSCSOU NOLIST NOREF
ALTER MTRXIN $ USE STRING-BASED ALTERS FOR VERSION 68.2 AND SUBSEQUENT
ADD K2PP, /K2PPX/V,Y,FRIC=(1.0,0.0) $
EQUIVX K2PPX/K2PP/-1 $
ENDALTER
$SOL 103 $ USE FOR CHECK OF REAL MODES
SOL 107
$SOL 110 $ USE FOR MODAL HESSENBERG
DIAG 8 $ MATRIX TRAILERS
TIME 5
CEND
$
TITLE = SIMPLE COMPLEX EIGENVALUE ANALYSIS
SUBTITLE = UNSYMMETRIC, LANCZOS, FRICT=0.05
LABEL = 1% DAMPING
K2PP = FRICTN
$METHOD = 1 $ USE FOR MODAL HESSENBERG ANALYSIS
CMETHOD= 10 $CLAN
$CMETHOD= 11 $INVP
$CMETHOD= 12 $HESS
$SPC=10
MPC = 200
SVECT = ALL $ PRINT REAL EIGENVECTORS
SDISP = ALL $ PRINT REAL MODE COMPONENTS OF COMPLEX EIGENVECTORS
DISP = ALL $ PRINT COMPLEX EIGENVECTORS
SET 1 = 102, THRU, 202
BEGIN BULK
PARAM, GRDPNT, 0
PARAM, G, .02
PARAM, FRIC, 0.05
$
$ RIGID HOUSING MOUNTED ON SPRINGS, YZ MOTIONS ONLY
GRID,100,,0.0,0.0,35.0,,156
GRID,101,,0.0,0.0,15.0,,1456
GRID,102,,-50.0,30.0,7.5,,1456
GRID,103,,50.0,30.0,7.5,,1456
RBE2,101,100,23,101,102,103
$ LUMPED MASS AT CG
CONM2,100,100,,2.0 , , , , ,+CM1
+CM1,7200.0,,7200.0 , , ,7200.0
$ SPRINGS TO GROUND
CELAS2,102, 78.96+6, 100, 2
CELAS2,103, 78.96+6, 100, 3
CELAS2,104, 2.846+11,100,4
$
$ PAD IS A SINGLE HEXA WITH A POINT FOR THE HOUSING
CHEXA,200,200, 211, 212, 214, 213, 221, 222,+CHX1
+CHX1, 224, 223
RBE3,201,,201,123456,1.0,123,221,222,+RBE3
+RBE3,223,224
CONM2,202,201,,0.2
GRID, 201, ,0.0,0.0,15.0
GRID, 211, , -50., -30., 0.0,,1456
GRID, 212, , 50., -30., 0.0,,1456
GRID, 213, , -50., 30., 0.0,,1456
GRID, 214, , 50., 30., 0.0,,1456
GRID, 221, , -50., -30., 15.0,,1456
GRID, 222, , 50., -30., 15.0,,1456
GRID, 223, , -50., 30., 15.0,,1456
GRID, 224, , 50., 30., 15.0,,1456
1 CHAPTER 5
PSOLID, 200, 200, , THREE
MAT1,200,1.0+6,,0.3,1.111-6,,,0.02
$
$ RIGID LINKS FROM PAD FOR TANGENTIAL STOPS
MPC, 200, 213,2, 1.0, 223,2, 1.0, , +MPC21
+MPC21,, 102,2, -2.0
$MPC, 200, 213,3, 1.0, 223,3, 1.0, , +MPC22
$+MPC22,, 102,3, -2.0
MPC, 200, 214,2, 1.0, 224,2, 1.0, , +MPC31
+MPC31,, 103,2, -2.0
$MPC, 200, 214,3, 1.0, 224,3, 1.0, , +MPC32
$+MPC32,, 103,3, -2.0
$
$RBE3,202,,102, 234, 1.0, 123, 213 ,223,+RB31
$+RB31,UM ,213,23, 223,2
$RBE3,203,,103, 234, 1.0, 23 ,214, 224,+RB32
$+RB32,UM ,214,23, 224,2
$ SPRINGS CONNECTING HOUSING TO CENTER OF PAD
CELAS2,201, 15.791+6, 201,3 ,101,3
CELAS2,202, 15.791+6, 201,2 ,101,2
$
$ MOVING SURFACE SIMULATED WITH PLATES
$
CQUAD4,301,300,301,302,304,303
CQUAD4,302,300,303,304,312,311
CQUAD4,303,300,311,312,314,313
CQUAD4,304,300,313,314,322,321
CQUAD4,305,300,321,322,332,331
$
GRID, 301,,-50., -150.0, 0.0,, 12356
GRID, 302,, 50., -150.0, 0.0,, 12356
GRID, 303,,-50., -90.0, 0.0,, 1256
GRID, 304,, 50., -90.0, 0.0,, 1256
GRID, 311,,-50., -30.0, 0.0,, 1256
GRID, 312,, 50., -30.0, 0.0,, 1256
GRID, 313,,-50., 30.0, 0.0,, 1256
GRID, 314,, 50., 30.0, 0.0,, 1256
GRID, 321,,-50., 90.0, 0.0,, 1256
GRID, 322,, 50., 90.0, 0.0,, 1256
GRID, 331,,-50., 150.0, 0.0,, 12356
GRID, 332,, 50., 150.0, 0.0,, 12356
$
PSHELL, 300, 300, 15.0, 300,,300
MAT1, 300, 7.106+9, , 0.3, 4.444-6,,, 0.02
$
$ DUMMY SPRINGS TO MEASURE NORMAL FORCES
CELAS2, 311, 4.0+9, 311,3, 211,3
CELAS2, 312, 4.0+9, 312,3, 212,3
CELAS2, 313, 4.0+9, 313,3, 213,3
CELAS2, 314, 4.0+9, 314,3, 214,3
$
$ DMIGS REPRESENT SLIDING FRICTION.
$ TANGENTIAL FORCES EQUAL NORMAL FORCES TIMES FRICTION COEFFICIENT
$ MU =1.0
DMIG,FRICTN, 0, 1, 1, 0
$
DMIG,FRICTN, 211,3,, 211,2, 4.0+9,,+C211
+C211, 311,2, -4.0+9
DMIG,FRICTN, 311,3,, 211,2, -4.0+9,,+C311
+C311, 311,2, 4.0+9
$
Listing 5-1 Input Data for Complex Eigenvalue Stability Analysis (continued)

MSC.Nastran Results
This model was run with three different types of analysis. For the modal complex method,
20 real modes were requested to produce the reduced problem for complex eigensolution
analysis. It was found that the twelfth mode was quite divergent, while all of the other
modes were stable. The eigenvalues of this mode and the preceding and following modes
by this method are given below:
When the same model is analyzed by the Lanczos method, without modal reduction, the
roots in the same region are as follows:
Note that the results obtained by the two methods agree within a few percent, even though
these special effects are rather strong. This agreement occurs because almost all of the real
modes were found, including those at much higher frequencies than that of the twelfth
complex mode. When the friction coefficient parameter was reduced from a value of 0.5 to
0.05, all roots became stable. The roots in the region of the twelfth mode are as follows:
DMIG,FRICTN, 212,3,, 212,2, 4.0+9,,+C212
+C212, 312,2, -4.0+9
DMIG,FRICTN, 312,3,, 212,2, -4.0+9,,+C312
+C312, 312,2, 4.0+9
$
DMIG,FRICTN, 213,3,, 213,2, 4.0+9,,+C213
+C213, 313,2, -4.0+9
DMIG,FRICTN, 313,3,, 213,2, -4.0+9,,+C313
+C313, 313,2, 4.0+9
DMIG,FRICTN, 214,3,, 214,2, 4.0+9,,+C214
+C214, 314,2, -4.0+9
DMIG,FRICTN, 314,3,, 214,2, -4.0+9,,+C314
+C314, 314,2, 4.0+9
$
EIGRL,1, , , 20 $ USE FOR MODAL HESSENBERG
EIGC, 10, CLAN, MAX, , , , , ,+ELAN1
+ELAN1, 0.0, 6000.0, , , , , 12,, +ELAN2
EIGC, 11, INV , MAX, , , , , ,+EINV
+EINV, 0.0, 10000.0, 0.0, 50000., 20000., 20, 12
EIGC, 12, HESS , MAX, , , , 12
$ OMITS ARE NECESSARY WITH HESS TO REMOVE SINGULARITIES IN THE MASS MATRIX
$OMIT1, 4, 301, THRU, 332
ENDDATA
Listing 5-1 Input Data for Complex Eigenvalue Stability Analysis (continued)
C O M P L E X E I G E N V A L U E S U M M A R Y
0 ROOT EXTRACTION EIGENVALUE FREQUENCY DAMPING
NO. ORDER (REAL) (IMAG) (CYCLES) COEFFICIENT
. . .
11 13 -5.377758E+03 5.013208E+04 7.978768E+03 2.145436E-01
12 12 3.262632E+03 5.019004E+04 7.987992E+03 -1.300111E-01
13 5 -1.215334E+03 5.960500E+04 9.486430E+03 4.077961E-02
. . .
11 12 -5.339673E+03 5.010448E+04 7.974375E+03 2.131415E-01
12 11 3.294983E+03 5.022236E+04 7.993137E+03 -1.312158E-01
13 13 -1.200218E+03 5.960513E+04 9.486451E+03 4.027230E-02
1 CHAPTER 5
It would appear that there is a stability threshold somewhere between the friction
coefficient values of 0.05 and 0.5. The precise threshold can be determined by resetting the
value of this parameter and rerunning the model until the real value of this mode
approaches zero. Hardware elements that affect this unstable mode can be identified by
inspecting the eigenvector. The affect of hardware changes on the stability of this mode
can be evaluated by changing the most active elements and rerunning the analysis.
This study shows that the prudent method for checking this answer is the Lanczos method,
but the less expensive modal Hessenberg method can be used for the parameter variation
studies to find the stability boundary, with a final check made by the Lanczos method.
The eigenvector computed for this mode from the Direct Lanczos results is shown below.
This format contains two lines per grid point. The first line is the real value of
displacement. The second line indicates the imaginary part. Magnitude and phase angles
are an alternative choice.
Note that two points have a value of 1.0, except for a very small imaginary part. These are
the analysis set points with the largest value. (See points 223, 224.) Note that one point has
a value which exceeds unity. (See point 214.) This point is the dependent point on a
11 11 -9.727523E+02 5.035683E+04 8.014539E+03 3.863437E-02
12 12 -1.070261E+03 5.304468E+04 8.442323E+03 4.035319E-02
13 13 -1.200928E+03 6.031671E+04 9.599703E+03 3.982073E-02
COMPLEX EIGENVALUE = 3.294983E+03, 5.022236E+04
C O M P L E X E I G E N V E C T O R NO. 12
(REAL/IMAGINARY)

POINT ID. TYPE T1 T2 T3 R1 R2 R3
100 G .0 -2.587344E-03 -1.831422E-04 -1.900947E-05 .0 .0
.0 1.218755E-02 -2.493365E-04 9.222361E-05 .0 .0
101 G .0 -2.967534E-03 -1.831422E-04 .0 .0 .0
.0 1.403202E-02 -2.493365E-04 .0 .0 .0
102 G .0 -3.110105E-03 -7.534264E-04 .0 .0 .0
.0 1.472369E-02 2.517372E-03 .0 .0 .0
103 G .0 -3.110105E-03 -7.534264E-04 .0 .0 .0
.0 1.472369E-02 2.517372E-03 .0 .0 .0
201 G .0 8.990854E-02 6.901190E-02 -1.932251E-03 2.106063E-17 4.727121E-17
.0 -1.349169E-01 6.873457E-02 1.091643E-02 -4.813858E-17 -1.377165E-16
211 G .0 6.394567E-01 -1.057253E-02 .0 .0 .0
.0 6.715296E-01 -2.787052E-02 .0 .0 .0
212 G .0 6.394567E-01 -1.057253E-02 .0 .0 .0
.0 6.715296E-01 -2.787052E-02 .0 .0 .0
213 G .0 -1.006220E+00 -1.589278E-02 .0 .0 .0
.0 2.944739E-02 -2.169513E-03 .0 .0 .0
214 G .0 -1.006220E+00 -1.589278E-02 .0 .0 .0
.0 2.944739E-02 -2.169513E-03 .0 .0 .0
221 G .0 -8.201829E-01 1.269795E-01 .0 .0 .0
.0 -2.698337E-01 -2.587584E-01 .0 .0 .0
222 G .0 -8.201829E-01 1.269795E-01 .0 .0 .0
.0 -2.698337E-01 -2.587584E-01 .0 .0 .0
223 G .0 1.000000E+00 1.104436E-02 .0 .0 .0
.0 -2.834362E-17 3.962275E-01 .0 .0 .0
224 G .0 1.000000E+00 1.104436E-02 .0 .0 .0
.0 -1.627032E-15 3.962275E-01 .0 .0 .0
301 G .0 .0 .0 -1.812334E-04 .0 .0
.0 .0 .0 -4.347274E-04 .0 .0
302 G .0 .0 .0 -1.812334E-04 .0 .0
.0 .0 .0 -4.347274E-04 .0 .0
303 G .0 .0 -9.851459E-03 -1.239709E-04 .0 .0
.0 .0 -2.265664E-02 -2.428152E-04 .0 .0

selected MPC entry. Only analysis set points are considered when performing the
normalization. Dependent points for MPC equations and omitted points may be smaller
or larger than unity.
The meaning of the imaginary parts of the displacement at the other points is a measure of
the phase angle between motion at the normalizing point and the other points. A more
meaningful result for the imaginary part can be obtained from the modal analysis by
requesting SDISP output. This prints the modal contributions of the real modes which are
used to synthesize the complex modes. The output for the twelfth mode, from a Modal
Hessenberg analysis, is as follows:
Although the columns are labeled T1, T2, etc., they really represent the scalar modal
variables printed six per row. For example, the largest value occurs for the eleventh real
mode, printed in the row labeled 7 M under R2. (M symbolizes modal variable in this
context.) If the objective of the analysis is to find the parts of the structure that have the
largest effect on stability, when changed, it is often easier to understand the physics from
the real mode shapes rather than the displacement in a complex mode shape of physical
variables.
The a-set values of the real mode shape can be printed by the SVECT Case Control
command. The values of nonanalysis set DOFs, such as m-set or omitted points, appear as
blank spaces. If the complete real mode shapes are needed for plotting, or if element strain
energy is desired for the real modes, this input stream can be run in SOL 103, the real
modes analysis, by changing only the SOL entry. Inputs not related to real eigenvalue
analysis are ignored.
COMPLEX EIGENVALUE = 3.262632E+03, 5.019004E+04
C O M P L E X E I G E N V E C T O R N O . 12 (SOLUTION SET)
(REAL/IMAGINARY)

0 1 M 2.723618E-02 2.382100E-04 5.274842E-04 -2.809022E-14 -2.894890E-02 5.665277E-02
1.177777E-02 6.993240E-04 -8.778612E-04 -3.353533E-16 3.104551E-02 -6.190442E-02
0 7 M -9.730529E-15 2.098325E-14 -4.441694E-13 -4.738649E-02 1.000000E+00 1.575081E-01
-4.749197E-14 -2.213062E-14 3.112463E-13 -4.339022E-01 .0 -2.552028E-01
0 13 M 1.127128E-15 -1.005931E-01 -1.593949E-03 2.438817E-16 -6.407992E-05 -2.046400E-17
-3.111150E-14 3.855639E-02 3.900885E-04 1.721788E-15 4.127408E-05 6.070441E-16
0 19 M -8.610454E-17 4.902618E-06
-1.722265E-16 1.007778E-06
1 CHAPTER 5
5.4 Nonlinear Transient Response Analysis
Nonlinear effects in structures occur mainly due to nonlinear materials, preloads, and large
rotations. Contact problems exhibit nonlinear effects due to changes in boundary
conditions. All of these effects may be represented by nonlinear elements.
The nonlinear element library of MSC.Nastran consists of:
RODs, CONRODs, and TUBEs for unidirectional truss members.
BEAMs for axially and laterally deforming line members.
QUAD4s and TRIA3s for membrane, plate and shell modeling.
HEXAs, PENTAs, and TETRAs for solid modeling.
GAPs for contact and friction simulation.
BCONPs for slideline contact.
Nonlinear elements may be combined with linear elements for computational efficiency if
the nonlinear effects can be localized. Primary operations for nonlinear elements are the
updating of element coordinates and applied loads for large displacements and the
internal relaxation iteration for material nonlinearity. Refer to Table 5-1 for a summary of
the nonlinear element properties.
Table 5-1 Summary of Nonlinear Elements
Element Connectivity
BCONP Connection and type for slideline.
CBEAM Connection for beam element.
CGAP Connection for gap or frictional element.
CHEXA Connection for six-sided solid element.
CONROD Connection and properties for rod.
CPENTA Connection for five-sided solid element.
CQUAD4 Connection for quadrilateral element with bending and
membrane stiffness.
CQUADX Properties for axisymmetric hyperelastic solid.
CROD Connection for rod with axial and torsional stiffness.
CTETRA Connection for four-sided solid element.
CTRIA3 Connection for triangular element with bending and membrane
stiffness.
CTUBE Connection for a tube.

The geometric nonlinearity becomes discernible when the structure is subjected to large
displacement and rotation. Geometric nonlinear effects are prominent in two different
aspects: (1) geometric stiffening due to initial displacements and stresses, and (2) follower
forces due to a change in loads as a function of displacements. These effects are included
in the current release of MSC.Nastran, but the large deformation effect resulting in large
strains is yet to be implemented.
Material nonlinearity is an inherent property of any engineering material. Material
nonlinear effects may be classified into many categories. Included in the current release of
MSC.Nastran are plasticity, nonlinear elasticity, creep, and viscoelasticity. Creep and
viscoelasticity, implemented as a generalized nonlinear viscoelastic capability, may be
coupled with plasticity. Many sophisticated options are available for yield criteria and
hardening behavior in plasticity.
The primary solution operations are load and time steps, iterations with convergence tests
for acceptable equilibrium error, and stiffness matrix updates. The iterative process is
based on the modified-Newtons method combined with optional methods such as the
quasi-Newton (BFGS) update and the line search. The tangent matrix updates are
performed automatically to improve the computational efficiency, and may be overridden
at the users discretion.
For static analysis, a number of options such as arc-length methods is available for
snapthrough or post-buckling analysis. For the transient response analysis, a number of
options is available for implicit direct time integration, combined with adaptive and other
iteration strategies similar to those implemented for static analysis.
The adaptive method is implemented using the two-point recurrence (or one-step) formula
as its foundation. The optimum time step size, which is required for accuracy and
efficiency, changes continuously in the transient dynamic environment. The primary
concept of automatic time step adjustment is that the proper size of the time step can be
predicted based on the dominant frequency in the incremental deformation pattern at the
previous time step. This concept presents a deficiency of time lag involved in the
Element Properties
PBCOMP Properties for composite CBEAM.
PBEAM Properties for CBEAM.
PCOMP Properties for composite material laminate.
PGAP Properties for CGAP.
PLSOLID Hyperelastic properties for CHEXA, CPENTA, and CTETRA.
PROD Properties for CROD.
PSHELL Properties for CTRIA3 and CQUAD4.
PSOLID Properties for CHEXA, CPENTA, and CTETRA.
PTUBE Properties for CTUBE.
Table 5-1 Summary of Nonlinear Elements (continued)
1 CHAPTER 5
prediction process. Furthermore, changes in nonlinearity cannot be predicted from the
deformation pattern at the previous time step. This deficiency is overcome by the bisection
process, which is activated when any difficulties arise in terms of convergence during the
iteration.
User Interface
The input data required for SOL 129 is a combination of direct time integration control
data, similar to SOL 109 (for direct linear transient with superelements), and nonlinear
modeling data similar to SOL 106 (for nonlinear statics). The nonlinear properties and/or
effects are defined by nonlinear material data (MATS1 and TABLES1), gap elements (GAP)
for nonlinear interfaces, and PARAM,LGDISP for geometric nonlinearity. The transient
effects are produced by time-dependent loading functions (TLOADi, DAREA, LSEQ, etc.),
damping (parameters, elements and material data), and mass properties.
The unique data required for SOL 129 is supplied on the TSTEPNL Bulk Data entry. The
TSTEPNL entry in itself is a combination of the TSTEP entry for direct time integration and
the NLPARM entry for nonlinear iteration control. Restarts are controlled by parameters
(LOOPID, STIME, SLOOPID and SDATA) which can be specified either in the Case
Control Section or Bulk Data Section. Some optional parameters (TSTATIC, NDAMP) are
provided for additional control or capabilities.
Case Control
Each subcase defines a time interval starting from the last time step of the previous
subcase, subdivided into smaller time steps. The output time is labeled by the cumulative
time, including all previous subcases. There are advantages in dividing the total duration
of analysis into many subcases so that each subcase does not have excessive number of time
steps. The data blocks containing solutions are generated at the end of each subcase for
storage in the database for output processing and restarts. As such, converged solutions
are apt to be saved at many intermediate steps in case of divergence and more flexible
control becomes possible with multiple subcases.
The input loading functions may be changed for each subcase or continued by repeating
the same DLOAD request. However, it is recommended that one use the same TLOAD
Bulk Data for all the subcases in order to maintain the continuity between subcases,
because TLOADi data defines the loading history as a function of cumulative time. Static
loads (PLOADi, FORCEi, MOMENTi) may be associated with time-dependent functions
by the Bulk Data LSEQ, which can be selected by a Case Control command LOADSET.
However, no thermal loads or enforced displacements (SPCD) are allowed in the nonlinear
transient analysis. Nonlinear forces as functions of displacements or velocities (NOLINi)
may be selected and printed by the Case Control commands NONLINEAR and NLLOAD,
respectively. Each subcase may have a different time step size, time interval, and iteration
control selected by the TSTEPNL request. The Case Control requests that may not be
changed after the first subcase are SPC, MPC, DMIG, and TF.

Output requests for each subcase are processed independently. Requested output
quantities for all the subcases are appended after the computational process for actual
output operation. Available outputs are DISPLACEMENT, VELOCITY,
ACCELERATION, OLOAD, STRESS, FORCE, SDISPLACEMENT, SVELOCITY,
SACCELERATION, NONLINEAR (NLLOAD), and SPCFORCE. However, element force
output and GRID point stresses are not available for nonlinear elements.
Initial conditions (displacement or velocity) can be specified by the Bulk Data input, TIC,
selectable by the Case Control command IC. If initial conditions are given, all of the
nonlinear element forces and stresses must be computed to satisfy equilibrium with the
prescribed initial displacements. On the other hand, initial conditions can be generated by
applying static analysis for the preload using PARAM,TSTATIC in the first subcase. Then
the transient analysis can be performed in the ensuing subcases. Associated with the
adaptive time stepping method, the PARAM,NDAMP is used to control the stability in the
ADAPT method. The parameter NDAMP represents the numerical damping (a
recommended value for usual cases is 0.01), which is often required to improve the
stability and convergence in contact problems.
All the superelement model generation options and matrix reduction options are allowed
for the linear portion of the structure. The generalized dynamic reduction, component
mode synthesis, and Guyan reduction may be performed for upstream superelements.
The residual superelement may contain scalar degrees-of-freedom representing linear
modal formulations.
Implicit Integration Control: TSTEPNL Data
The input fields of the TSTEPNL Bulk Data entry specify the time step size, the number of
steps, and the output interval as well as the nonlinear iteration options. The TSTEPNL
Bulk Data is selectable by the Case Control command TSTEPNL. Although the same
TSTEPNL Bulk Data may be selected by more than one subcase, it is recommended to select
a different TSTEPNL entry for each subcase in preparation for changes in the restarts.
The choice of time step size is determined primarily by the frequency content of the input
load and the highest frequency mode of interest. A general guideline is that seven or more
steps per cycle be provided for reasonable accuracy. Modes with shorter periods (higher
frequency) will be attenuated by the numerical process. Highly nonlinear problems may
require smaller step size. However, the adaptive time stepping capability will
automatically adjust the time step size. Nevertheless, the initial time step size should be
estimated by the user according to the aforementioned principles.
A caution is necessary in using the automatic time step adjustment if the forcing function
is a short duration pulse. Since the automatic time step adjustment algorithm does not
consider the loading history, short pulses could be skipped if the time step is automatically
adjusted to a larger value than the pulse duration. It is advised that a drastic change in the
time step size between subcases be avoided. A drastic changee.g., ratio of the two time
steps exceeding 1000could cause a loss of accuracy upon subcase switch. In this case, an
intermediate subcase should be provided for a transition period of short interval to reduce
the ratio.
1 CHAPTER 5
The parameters for controlling the accuracy and stability of the incremental and iterative
process can be specified in the TSTEPNL Bulk Data entry. The controls are applicable to
the automatic time step adjustment and bisection process in addition to stiffness matrix
updates, BFGS updates, and line searches similar to those on the NLPARM Bulk Data
entry. Since default values have been selected from numerous test runs, the analysis
should be started with the default setting and changed if necessary. The TSTEPNL data
format is shown below with default values:
Format:
Example with Defaults (x is Required Input):
The TSTEPNL Bulk Data entry is selected using ID via the Case Control command
TSTEPNL. Each subcase (residual superelement solutions only) requires a TSTEPNL
entry. Multiple subcases are assumed to occur sequentially in time. Therefore, the initial
conditions of each subcase are defined by the end conditions of the previous subcase.
The NDT field specifies the number of time steps with DT as the size of each time step. The
total duration for the subcase can be assessed by multiplying NDT and DT (i.e., NDT*DT).
The time increment ( ) remains constant during the analysis in AUTO and TSTEP options,
and is equal to DT. However, the time increment ( ) changes during the analysis in the
ADAPT option and the actual number of time steps will not be equal to NDT. In the
ADAPT option, DT is used as an initial value for .
The NO field specifies the time step interval for output; i.e., every NO-th step solution is
saved for output. The data will be output at steps 0, NO, 2NO, ..., etc., and the last
converged step for printing and plotting purposes. The Case Control command OTIME
may also be used to control the output points.
The METHOD field selects an option for direct time integration and the stiffness matrix
update strategies among ADAPT, AUTO and TSTEP. If the AUTO option is selected,
MSC.Nastran automatically updates the stiffness matrix to improve convergence while the
KSTEP value is ignored. If the TSTEP option is selected, MSC.Nastran updates the stiffness
matrix every KSTEP-th increment of time. If the ADAPT option is selected, MSC.Nastran
automatically adjusts the incremental time and uses the bisection algorithm in case of
divergence. During the bisection process in the ADAPT option, stiffness is updated at
1 2 3 4 5 6 7 8 9 10
TSTEPNL ID NDT DT NO METHOD KSTEP MAXITER CONV +TNL1
+TNL1 EPSU EPSP EPSW MAXDIV MAXQN MAXLS FSTRESS +TNL2
+TNL2 MAXBIS ADJUST MSTEP RB MAXR UTOL RTOLB
1 2 3 4 5 6 7 8 9 10
TSTEPNL x x x 1 ADAPT 2 10 PW +TNL1
+TNL1 1.E-2 1.E-3 1.E-6 2 10 2 0.2 +TNL2
+TNL2 5 5 0 0.75 16.0 0.1 20.
t
t
t

every KSTEP-th successful bisection. The ADAPT method allows linear transient analysis,
but AUTO or TSTEP will abort the run if the model does not have any data representing
nonlinearity. The stiffness matrix is always updated for a new subcase or restart,
irrespective of the option selected.
The number of iterations for a time step is limited to MAXITER. If the solution does not
converge in MAXITER iterations, the process is treated as a divergent process; i.e., either a
bisection or stiffness matrix update takes place based on the value of MAXBIS. The sign of
MAXITER provides a control over reiteration in case of failure in convergence or bisection.
If MAXITER is negative, the analysis is terminated when the divergence condition is
encountered twice during the same time step or the solution diverges for five consecutive
time steps. If MAXITER is positive, MSC.Nastran computes the best attainable solution
and continues the analysis.
The convergence test is controlled by convergence test flags (U for displacement error test,
P for load equilibrium error test, W for work error test) and the error tolerances (EPSU,
EPSP and EPSW) which define the convergence criteria. All requested criteria (a
combination of U, P, and/or W) are satisfied upon convergence. It should be noted that at
least two iterations are necessary to check the displacement convergence criterion.
The MAXDIV field provides control over diverging solutions. Depending on the rate of
divergence, the number of diverging solutions (NDIV) is incremented by 1 or 2. The
solution is assumed to be divergent when NDIV reaches MAXDIV during the iteration. If
the bisection option is used with the ADAPT method, the time step is bisected upon
divergence. Otherwise, the solution for the time step is repeated with a new stiffness based
on the converged state at the beginning of the time step. If NDIV reaches MAXDIV twice
within the same time step, the analysis is terminated with a fatal message.
In transient analysis, the BFGS quasi-Newton updates and the line search process work in
the same way as in static analysis (except for the default settings). The MAXQN field
defines the maximum number of quasi-Newton vectors to be saved on the database and
the MAXLS defines the number of line searches allowed per iteration. Nonzero values of
MAXQN and MAXLS activate the quasi-Newton update and the line search process,
respectively.
The FSTRESS field defines a fraction of the effective stress ( ) which is used to limit the
subincrement size in the material routine. The number of subincrements in the material
routines is determined such that the subincrement size is approximately
(equivalent stress). FSTRESS is also used to establish a tolerance for error correction in the
elasto-plastic material, i.e.,
If the limit is exceeded at the converging state, MSC.Nastran will exit with a fatal error
message. Otherwise, the stress state is adjusted to the current yield surface.
In the ADAPT method, MAXBIS is the maximum number of bisections allowed for each
time step ( ). The bisection process is activated when divergence occurs
and . The number of bisections for a time increment is limited to MAXBIS. If
MAXBIS is positive and the solution does not converge after MAXBIS bisections, the best
FSTRESS
Error in yield function FSTRESS <
9 MAXBIS 9
MAXBIS 0
1 CHAPTER 5
solution is computed and the analysis is continued to the next time step. If MAXBIS is
negative and the solution does not converge in MAXBIS bisections, the analysis is
terminated.
The parameter ADJUST allows the user to control the automatic time stepping in the
ADAPT option. A value of zero for ADJUST turns off the automatic adjustment
completely. If ADJUST is positive, the time increment is continually adjusted for the first
few steps until a good value of is obtained. After this initial adjustment, the time
increment is adjusted every ADJUST time steps only. A value of ADJUST an order greater
than NDT will turn off adjustments after the initial adjustment. Since the automatic time
step adjustment is based on the mode of response and not on the loading pattern, it may be
necessary to limit the adjustable step size when the period of the forcing function is much
shorter than the period of dominant response frequency of the structure. It is the users
responsibility to ensure that the loading history is properly traced with the ADJUST
option. The ADJUST option should be suppressed for the duration of a short pulse
loading. If unsure, the user should start with a value of DT that is much smaller than the
pulse duration in order to properly represent the loading pattern.
MSTEP defines the desired number of time steps to obtain the dominant period response
accurately ( ). RB defines bounds for maintaining the same time step for
the stepping function in the automatic time step adjustment method ( ).
Parameters MSTEP and RB are used to adjust the time increment during the analysis in the
ADAPT option. The adjustment is based on the number of time steps desired to capture
the dominant frequency response accurately. The time increment is adjusted as
Eq. 5-16
where:
with:
The recommended value of MSTEP for nearly linear problems is 20. A larger value (e.g.,
40) is required for highly nonlinear problems. In the default options, MSC.Nastran
automatically computes the value of MSTEP based on the changes in the stiffness.
The MAXR field defines the maximum ratio for the adjusted incremental time relative to
DT allowed for time step adjustment ( ). MAXR is used to define the
upper and lower bounds for adjusted time step size, i.e.,
=
=
0.25 for
= 0.5 for
= 1.0 for
=
2.0 for
= 4.0 for
t
10 Integer 200
0.1 Real 1.0
t
n 1 +
f r ( )t
n
=
r
1
MSTEP
--------------------
2
n
------
* +
, -
1
t
n
--------
* +
, -
f r 0.5 RB <
f 0.5 RB r 2 <
f RB r 2 <
f 2. r 3. RB <
f r 3. RB
1.0 Real 32.0

Eq. 5-17
The UTOL field defines the tolerance on displacement increment below which there is no
time step adjustment ( ). UTOL is used to filter undesirable time step
adjustment; i.e., no time step adjustment is performed if
Eq. 5-18
The RTOLB field defines the maximum value of incremental rotation (in degrees) allowed
per iteration to activate bisection ( ). The bisection is activated if the incremental
rotation for any degree-of-freedom ( , , or ) exceeds the value specified for
RTOLB. This bisection strategy based on the incremental rotation is controlled by the
MAXBIS field.
Iteration Related Output Data
During the incremental and iterative computation, the process information consisting of
iteration data is printed at the end of each iteration or time step. The data is printed under
the following headings:
TIME Cumulative time for the duration of the analysis
ITER Iteration count for each time step
DISP Relative error in terms of displacements
LOAD Relative error in terms of load vectors
WORK Relative error in terms of work
LAMBDA(I) Rate of convergence in iteration
LAMBDA(T) Ratio of the load error for two consecutive time steps
LAMBDA-BAR Average of LAMBDA(T) over the last three steps, computed only
for AUTO or TSTEP method
DLMAG Absolute norm of the residual error vector, . The absolute
convergence is defined using DLMAG by
FACTOR Final value of the line search parameter
E-FIRST Divergence rate error before line search
E-FINAL Error at the end of line search
NQNV Number of quasi-Newton vectors appended
NLS Number of line searches performed during the iteration
ITR DIV Number of occurrences of divergence detected during the
adaptive iteration by the iteration module NLTRD2
MIN
DT
2
MAXBIS
----------------------- ,
DT
MAXR
-----------------
* +
, -
t MAXR DT
0.001 Real 1.0 <
U
n
U
max
-------------------- UTOL <
Real 2.0 >
x

y

z
R
R 10
12
<
1 CHAPTER 5
Three-Point Method (NLTRD Module)
The three-point method is chosen to be compatible with the MSC.Nastran linear transient
integration method. The Newmark Beta method based on the three-step integration is
combined with the modifications to Newtons method for nonlinear solutions. The
additional iteration steps provide equilibrium solutions at each time step, thereby
guaranteeing stability and accuracy for arbitrary time step size. This method can be used
by selecting the AUTO or TSTEP method in the TSTEPNL Bulk Data entry.
Basic Equations
We may calculate the load equilibrium error vector, at time step by the equation
Eq. 5-19
where:
The above equation is solved at the reduced ( ) displacement vector size. The
approximation errors due to dynamic reduction methods are not included in the error
vector . Applying Newmarks method over a finite time period, , the
average static forces are
Eq. 5-20
where is the Newmark Beta operator and is the nonlinear force due to a generalized
displacement vector . An identical definition occurs for from the applied loads at
each time step.
In summary, at each time step MSC.Nastran will iterate the displacements and forces until
passes the convergence tests or the number of passes reaches an iteration limit. With
a single step, i = i, calculating only , the results will be identical to the existing NOLINi
results in MSC.Nastran. For faster convergence the iterations may continue, the matrices
may be updated, and/or the time step size may be reduced. For more details on this
algorithm, refer to Chapter 9 of the MSC.Nastran Nonlinear Handbook.
MAT DIV Number of occurrences of bisection conditions in the material
routine (excessive stress increment) or in the rotation angle
(excessive rotation) during the iteration using the ADAPT method
NO. BIS Number of bisections executed for the current time interval
ADJUST Ratio of time step adjustment relative to DT within a subcase
= Average load over the time period ( ).
,
= Corresponding acceleration and velocity vectors.
= Average elasto-plastic element total force vector.
( for linear problems.)
R
n
{ } n
R
n
{ } P
n
Mu
n
Bu
n
F
n
{ } =
P
n
{ }
t
n 1
t
n
t
n 1 +
< <
u
n
{ } u
n
{ }
F
n
{ }
F Ku =
u
d
R
n
{ } t
n 1
t t
n 1 +
< <
F { } F u
n 1 +
( ) 1 2 ( )F u
n
( ) F u
n 1
( ) + + { } =
F u
n
( )
u
n
{ } P { }
R
n
i
{ }
R
0

Two-Point Method (NLTRD2 Module)
The multistep implicit integration method has difficulties when changing time step size.
Therefore, to allow self-adapting time step sizes, the two-point integration method is
introduced with module NLTRD2. This method can be selected by specifying the ADAPT
method in the TSTEPNL Bulk Data entry.
Newmark Integration
For the adaptive scheme, Newmarks method is employed with the two-point recurrence
formula for one-step integration, i.e.,
Eq. 5-21
and
Eq. 5-22
where , , and denote displacement, velocity, acceleration and the time step
increment, respectively. The subscript designates the time step and the parameters (
and ) are to be selected for the best solution. An equilibrium equation to be satisfied at
time step ( ) is
Eq. 5-23
where and denote mass and damping matrices, and and denote internal
and external forces, respectively.
An alternative expression for the load vector can be derived for by introducing
Eq. 5-24
By virtue of Eq. 5-24 the residual load error at each time step is effectively carried over to
the next step and the error propagation is reduced, while the computation is significantly
simplified.
Eq. 5-25
The iteration method calculates new values of until it is sufficiently small. The results
then become the starting point for the next time step.
Adaptive Time Stepping
It is desirable to have a fully automated method of time integration that renders effective
and accurate solutions of nonlinear problems. The dynamic response characteristics of the
structure may change due to the nonlinearity (geometric or material) or the mode of
U
n 1 +
{ } U
n
{ } t U
n
{ }
1
2
--- t
2
U
n
{ } t
2
U
n 1 +
U
n
{ } + + + =
U
n 1 +
{ } U
n
{ } t U
n
{ } t U
n 1 +
U
n
{ } + + =
U { } U
{ } U
{ } t
n
n 1 +
M U
n 1 +
{ } C U
n 1 +
{ } F U
n 1 +
( ) { } + + P
n 1 +
{ } =
M C F { } P
n 1 +
{ }
0.5 =
M U
n
{ } C U
n
{ } + P
n
F
n
{ } =
R
n 1 +
i
{ } P
n 1 +
F
n 1 +
i
{ }
4
t
----- M U
n
{ } P
n
F
n
{ }
4
t
2
-------- M
2
t
----- C + U
n 1 +
i
U
n
{ }
+ + =
R
n 1 +
i
1 CHAPTER 5
excitation. When the type of nonlinear behavior changes, the time step size should be
adjusted. The drawback is that in implicit methods, a change in the time step size requires
a decomposition of the tangent matrix. After the solution is achieved for the original time
step, the usual procedure will be resumed for the next time step.
In the present implementation of the automatic time stepping for nonlinear applications,
the adjustment is designed to be adaptive to the severity of the nonlinearity in the problem
by monitoring the changes in the stiffness. However, difficulties in the automatic time
stepping have been discovered when plasticity or GAP elements are involved. In such
cases an undesired time step adjustment can be caused by a drastic change in the stiffness,
which may only be a temporary difficulty. A filtering scheme has been devised to suppress
the effects of the spurious mass or stiffness in the automatic time step adjustments. In
addition, the bisection method is chosen to be activated only at the time when divergence
occurs. The bisection process is coordinated with the stiffness matrix update so that
changes in nonlinearities are properly reflected while the bisection is in progress.
After the time step is completely converged, the integration proceeds to the next time step
with the same increment ( ). Further bisection may be required after achieving
converged solution at the intermediate time steps.
If no further bisection is required and the solutions converge without any difficulty, some
recovery process such as doubling time step may be activated. However, any adjustment
to the time step would require additional matrix decomposition that offsets the advantages
of adjusting time step size. Therefore, it has been determined to continue stepping through
the specified time interval to completion. The normal process should be resumed when the
user-specified time step at which bisection is activated is fully processed.
When the automatic time stepping is combined with the bisection method and if the
bisection is activated, the automatic adjustment procedure will be deferred until the
divergence is trapped and the solution process is stabilized. The solution process is
considered stabilized when two successive solutions have converged without requiring
further bisection. This is the time when the automatic time step adjustment is resumed.
The maximum number of bisections is limited by a user-specified parameter MAXBIS
(defaulted to 5). The bisection process is activated on an as-needed basis. Users also have
an option to suppress bisection by specifying MAXBIS = 0. If the bisection is required more
than MAXBIS times, the solution process will continue without bisection by activating the
reiteration procedure, by which the same iteration process is repeated to find the best
attainable solution. The best attainable solution is a solution obtained from the iteration
that yields the least average error. The average error is defined by averaging relative errors
in displacements, loads and energy. If the i-th iteration yields the least average error, the
reiteration procedure will end at the i-th iteration and the normal time stepping procedure
will be resumed.
Quasi-Newton and Line Searches
The BFGS update and the line search processes are fully operational during the iteration.
Quasi-Newton (QN) vectors are continuously accumulated up to MAXQN pairs (MAXQN
is a user-specified parameter) until the new stiffness matrix is evaluated. Once the number
Atk

of QN vector pairs reaches MAXQN (defaulted to 10), the QN vectors will be updated
selectively based on the condition number of the QN update matrix. As a requirement for
the reiteration procedure, the QN vectors to be accumulated are stored on a temporary
basis until a converged solution is obtained. All the accumulated QN vectors are purged
if the stiffness matrix update or the decomposition (due to the change in ) is scheduled.
The iterative process for time step ( ) can be summarized as follows:
1. Given at the beginning of the time step are , , , , , , ,
, , and the triangularized matrices for .
2. Adjust the time step size, if applicable.
3. Initialize , , and .
4. Decompose, , if . Then, purge the QN vector file.
5. Compute .
6. Solve for by forward and backward substitution, using the BFGS update
if applicable.
7. Compute the line search error.
8. If divergence occurs, go to step 12.
9. Proceed to the next step if the line search error is less than the tolerance.
Otherwise, perform the line search process and go back to step 7.
10. Compute the global error function and check convergence.
11. If the solution has converged, go to step 16. Otherwise, save the QN vector on a
temporary basis (if applicable) and go to the next iteration in step 5.
12. If it is the first divergence or the divergence after the KSTEP-th converged
bisection step, update the stiffness matrix, and go back to step 4. Otherwise,
proceed to the next step.
13. Bisect and go back to step 3. However, if the maximum number of bisections
allowed for the time step is reached, proceed to the next step.
14. Go to step 5 to continue iteration. However, if the maximum number of iterations
allowed for an increment is reached, proceed to the next step.
15. Perform the reiteration procedure to find the best attainable solution.
16. Accept the solution ( and ) and append temporary QN vectors to the
permanent QN vector file.
17. Compute the velocity .
18. Advance to the next time step with .
Restarts
Since SOL 106 and SOL 129 share the same database storage formats for nonlinear tables
and matrices, the restart system for transient analysis can use either a previous static or
transient nonlinear analysis as its initial conditions.
t
n 1 +
t
n

ref
M [ ] C [ ] P
n
{ } F
n
{ } F
n
{ }
U
n
{ } U
n
{ } LDL
T
( ) A [ ]
4
t
2
--------M
2
t
-----C K
+ + =
U
n 1 +
0
{ } F
n 1 +
0
{ } P
n 1 +
0
{ }
A [ ] LKL
T
= t
n 1 +
t
n
R
n 1 +
i
{ }
U
i 1 +
{ }
t
U
n 1 +
F
n 1 +
U
n 1 +
n n 1 + =
1 CHAPTER 5
Restarting From SOL 106 Into SOL 129
The options for a restart from SOL 106 into SOL 129 are static to static and static to dynamic
analysis. For a restart from a previous static analysis, only the first subcase is affected.
Simply provide a database created in SOL 106 and specify the parameter
PARAM,SLOOPID,N
where N is the printed value of LOOPID for the desired static solution. The initial transient
load should be identical to static loads at the restart state. Constraint sets, direct input
matrices, mass, and damping may be changed.
Restarting Within SOL 129
Restarting within SOL 129 allows static to static, static to dynamic, and dynamic to
dynamic analysis. Restarts from a previous nonlinear transient execution are available for
a number of cases. If the same model is to be reexecuted, only the residual superelement
needs to be reassembled. If the final results from the previous transient run are to be used
as the initial conditions at t = 0, add dummy SUBCASE commands, corresponding to the
previous run, to start the residual Case Control execution and set the parameter,
STIME = 0.
The normal restart for a transient run is to be continued from the last step of a previous
subcase with different loads and/or TSTEPNL data. For the normal restart provide the
following parameters:
LOOPID = N: Start from the N-th subcase
STIME = t: Start from time t
Note that to avoid incompatible matrix sizes, constraint sets must not be changed. The
values of LOOPID and STIME, which are printed with the iteration information for each
subcase, can be directly read from the printout of the previous run. See the MSC.Nastran
Nonlinear Handbook for more details and some examples.
Eq. 5-26
where and are the number of time steps and the time increment of the i-th
subcase, respectively.
If a SOL 129 run is terminated abnormally in the middle of a subcase, it may or may not be
restartable depending upon the cause of the abnormal exit. If the job is stopped due to a
diverging solution, it can be restarted either from the end of a previous subcase or from the
last saved solution step. The restart procedure for the former is identical to that for the
normal restart as described in the preceding paragraph. The latter case also requires
parameters LOOPID and STIME; however, the input value for STIME differs depending on
the value of METHOD specified on the TSTEPNL entry.
t NDT
i
i 1 =
N
)
DT
i
=
NDT
i
DT
i

If METHOD = AUTO or TSTEP, STIME is the time corresponding to the last output step
which may be calculated based on the output skip factor (i.e., the NO on the TSTEPNL
entry). If METHOD = ADAPT, the last converged solution is always treated as an output
step and is always saved for the restart so that STIME can be the time of the last converged
step. The values of STIME and LOOPID can also be found in the printout, if the ADAPT
method is used.
Once STIME and LOOPID are known, determine the number of remaining time steps in
the subcase and create a new TSTEPNL entry for the remaining time. Insert a new subcase
that references the new TSTEPNL entry prior to the remaining subcases in the Case Control
Section.
A solution may be terminated in the middle of a subcase due to insufficient CPU time: (1)
the CPU time specified in the Executive Control Section is insufficient so the run is forced
to exit by MSC.Nastran, or (2) the CPU time exceeds the limit specified in the computer
system which leads to a sudden job abortion by the system. In the first case, MSC.Nastran
is able to detect the specified CPU time in the Executive Control Section and automatically
activate the exit procedure before time expiration. When completed, the solution can be
restarted from the termination point as in the solution diverging case. In the second case,
the solution can only be restarted from the end of a subcase.
Restarts may also be performed solely for data recovery by providing the following
parameters:
SDATA = -1 : Recover data without running the solution module
LOOPID = N : from the 1st through the N-th subcases
Note that solution sets DISP, VELO, ACCE, OLOAD, SPCF (printout and plotting) and
NLLOAD (plotting only) are recoverable while STRESS, SDISP, SVELO and SACCE sets
are not.
Expedient Iteration Strategies
The dynamic tangent matrix may change for the following reasons:
1. Stiffness (K) changes due to geometric, material, and/or kinematic nonlinearity.
2. Bisection or time-step adjustment causes changes in the effective stiffness.
While the second case only requires decomposition, the first case requires reformulation of
the new stiffness matrix as well. The adaptive method is based on the following matrix
update strategies:
1. Stiffness is updated at the previously converged position when the divergence is
detected for the first time at a given time step.
2. Stiffness matrix update is allowed only once within the same time step.
3. If the bisection is in progress, additional K-updates are allowed within a given
at every KSTEP-th converged bisection, where KSTEP is a user-specified
parameter in the TSTEPNL entry.
T
1 CHAPTER 5
4. Decomposition is performed within the module at every bisection or time step
adjustment.
5. Whenever the decomposition is performed, the iteration starts from the
previously converged position.
6. The stiffness matrix is updated at the beginning of each subcase.

5.5 Fourier Transform
The Fourier transform capability in MSC.Nastran allows a transient analysis to be
performed using a frequency response solution. Time-dependent applied loads are
transformed to the frequency domain and all frequency dependent matrix calculations are
completed. The frequency response solution variables are then transformed back into the
time domain.
Fourier transform methods have been implemented in MSC.Nastran to integrate the
equations of motion in order to obtain the aeroelastic response of fixed wing aircraft. This
capability is especially important for this type of analysis since the unsteady aerodynamic
matrices are known only in the frequency domain. The Fourier transform method may also
be used to solve for the transient response of conventional structural models (no
aerodynamic effects) subjected to periodic loads.
This capability is available in SOLs 108 and 111 for frequency response output data. For
transient type output, SOL 146 must be used. The transformation is performed when the
requested load is the TLOADi form.
Theory
Two forms of the Fourier transform are available: the Fourier series and the Fourier
integral. Both methods require necessary numerical compromises and hence produce
numerical approximations. The inverse transform includes an infinite sum, for which only
a finite number of terms are numerically evaluated. The inverse Fourier integral must be
numerically integrated, which may result in integration errors. The number of frequencies
at which the integrand is evaluated is limited by the cost of calculations.
In the Fourier series, the basic time interval is , with the function periodic. The
circular frequencies are given by
where is a large time equal to the period of the lowest forcing frequency.
The load transformation for a load at point is given by
Eq. 5-27
The response at point is given by
Eq. 5-28
where is the frequency response of any physical variable due to unit load. The
response in the time domain is given by
0 t T < <
n
2nf =
f
1
T
--- =
"
.
$
.
&
T
a
P
a

n
( ) P
a
t ( )e
i
n
t
t d
0
T
(
=
j
u
j

n
( ) H
j a

n
( )P
a

n
( ) =
H
ja

n
( )
1 CHAPTER 5
Eq. 5-29
In the Fourier Integral, the time interval is the limit as , ,and of the
Fourier series. Here, is a continuous variable. Eq. 5-27, Eq. 5-28, and Eq. 5-29 take the
form
Eq. 5-30
Eq. 5-31
Eq. 5-32
Transformation of Loads to the Frequency Domain
The transformation of the user-specified time dependent loads into the frequency domain
is given in Eq. 5-27. With this transformation, the user must define a function that vanishes
for .
For piecewise linear tabular functions (TLOAD1), a table of pairs ( , )( )
prescribes time intervals. If an X1 shift and an X2 scale factor are included, the
time-dependent load at point is given by
Eq. 5-33
where is an amplitude factor and is a delay factor that may depend upon the loading
point. Applying finite step-by-step integration to Eq. 5-27, the transformed load, , is
obtained for each requested frequency.
Likewise, the general function (TLOAD2) is defined by
Eq. 5-34
where
The value of must be an integer for transient analysis by the Fourier method. The
transformation to the frequency domain is also obtained by numerical integration.
These loads, which appear in the form required for frequency response, are transformed to
the modal coordinates exactly as in the modal frequency response method.
u
j
t ( )

--------
1
2
---
* +
, -
u
j
0 ( ) Re u
j

n
( )e
i
n
t
( )
n 1 =
)
+ =
T f 0 2nf
a
( ) P
a
t ( )e
i t
t d
0
(
=
u
j
( ) H
j a
( )P
a
( ) =
u
j
t ( )
1
---
* +
, -
Re u
j
( )e
i t
( ) d
0
(
=
t T >
x
i
Y
i
i 1, N =
N 1
a
P
a
t ( ) A
a
Y
T
t
a
X1
X2
-----------------------------
* +
, -
=
A
a

a
P
a
( )
P
a
T ( )
A
a
t
n
e
t
2f t
+ ( ) cos 0 t T
2
T
1
< <
0 Otherwise
=
t t T
1

a
=
n

One other source of loads for aeroelastic problems is a one-dimensional gust. The same
time dependencies are allowed as defined above; however, the amplitude ( ) and delays
( ) for the aerodynamic elements are computed from areas, dihedrals, and coordinates in
the flow direction.
Calculation of Frequency Response
Frequency responses are computed in the frequency domain by conventional
MSC.Nastran methods for coupled equations.
Inverse Transformation of the Response to the Time Domain
The response in the time domain is found either from the Fourier integral approximation
of Eq. 5-32 or from the Fourier series result of Eq. 5-29 (which can be thought of as a special
form of the integral approximation.) Three approximation methods are available to
evaluate the inverse transform which may be selected by the user via the parameter IFTM.
In all cases, the quantity is first calculated at a set of frequencies, , by the frequency
response module where the s do not need to be equally spaced. For all methods, is
set equal to zero outside the range of s computed. These methods are:
Consider Method 2. Solving the three-moment equations, the second derivatives , can
be found for each for which a frequency response has been computed. Then, in any
interval ,
Eq. 5-35
where:
Integrate Eq. 5-32 using Eq. 5-35 for , and sum over the integrals. Then collect the
terms for each with the result,
Eq. 5-36
Method 0
Approximate as a constant in each interval (the default
method). For equal frequency intervals, this method reduces to the
Fourier series approximation of Eq. 3.
Method 1
Fit with a piecewise linear function, and do not approximate .
Method 2
Fit with a cubic spline function, and do not approximate .
=
=
A
a
a
u
( )
i
i
u
( )
( )e
i t
u
( ) e
it
u
( ) e
it
u
( )
i

i 1 +
< <
u
( ) u

i
( ) s u

i 1 +
( ) r + [ ]

i 1 +

i
( )
2
6
------------------------------------
u

i
( )s s
3
u
i 1 +
( ) r r
3
( ) + [ ]
=
r

i
( )
i 1 +

i
( )
-----------------------------
s 1 r
u
( )
n
u t ( )
1
---
* +
, -
Re C
n
t ( )u

n
( ) D
n
t ( )u

n
( ) + [ ]e
i
n
t
" #
$ %
& '
n 1 =
N
)
=
1 CHAPTER 5
Eq. 5-37
Eq. 5-38
For the first terms in Eq. 5-36 ( ), use only the second terms on the right side of Eq. 5-37
and Eq. 5-38. For the last term in Eq. 5-36 ( ), use only the first terms on the right side
of Eq. 5-37 and Eq. 5-38.
Eq. 5-39
Eq. 5-40
The above form of the inverse transform has two advantages. First, numerical problems
for small values of are efficiently evaluated by choosing the series form of Eq. 5-40.
Also, the other two methods are easily derived as subcases. If terms are removed from
Eq. 5-35, a piecewise linear fit occurs. Thus, Method 1 results by deleting from Eq. 5-36,
i.e.,
Eq. 5-41
with defined by Eq. 5-38. Method 0 results if we replace by 1.0 in Eq. 5-38.
The above procedure for Method 0 always multiplies the first and last terms in the series
by one-half. In order to force agreement with the Fourier series, which is the limiting case
of equal frequency intervals, the first term in the series is multiplied by one-half only if the
value of the first frequency is zero.
Some special considerations are given in the equal frequency interval case. When all s
are equal and the first frequency is an integer multiple of , the time step is adjusted to
make , reducing the number of distinct values of and used
in Eq. 5-41. Also, and (Eq. 5-37 and Eq. 5-38) become independent of and do
not need to be computed at every frequency.
Other important practical considerations must be observed to use these methods
successfully. to illustrate one problem, consider the response of a simple damped oscillator
to a pulse (Figure 5-3). The upper three curves show the pulse and the response of the
system if it is very stable and slightly stable. Using the Fourier method, the pulse is
replaced by a series of pulses, with period .
C
n
t ( )

n

n 1
2
------------------------------ E
2
i t
n

n 1
( ) ( )

n 1 +

n
2
------------------------------ E
2
it
n 1 +

n
( ) ( ) + =
D
n
t ( )

n

n 1
( )
3
24
-------------------------------------- G i t
n

n 1
( ) ( )

n 1 +

n
( )
24
-----------------------------------
3
G i t
n 1 +

n
( ) ( ) =
n 1 =
n N =
G z ( ) 2E
2
z ( ) E
4
z ( ) =
E
K
z ( )
K!
z
k
------ e
z
z
k
k!
---- -
k 0 =
k K =
)
1
z
K 1 +
--------------
z
2
K 1 + ( ) K 2 + ( )
----------------------------------------
z
3
K 1 + ( ) K 2 + ( ) K 3 + ( )
------------------------------------------------------------- + + +
=
t
u
u
u t ( )
1
---
* +
, -
Re C
n
t ( )u
n
( )e
i
n
t
[ ]
n 1 =
N
)
=
C
n
E
2
f
f t
f t 1 integer =
n
t sin
n
t cos
C
n
t ( ) D
n
t ( ) n
1 f

Figure 5-3 Response of a Single Degree-of-Freedom for Three Types of Periodic
Loads
Guidelines for Use. As can be seen, this method gives good results if the system is
damped, but an incorrect impression if the system is lightly damped. Thus, in order for the
results of the Fourier method to be valid:
1. The system should be reasonably well damped.
2. The forcing function should be zero for some time interval to allow decay.
3. The frequency interval .
If the system has unstable modes, these will appear as a precursor before the pulse as a
stable mode in the reverse time. In general, Methods 1 and 2 are more accurate than
Method 0. However, these methods introduce positive artificial damping into the result
that may lead to erroneous conclusions in stability studies. To see this, consider the
function in Eq. 5-42. If equal frequency intervals are chosen, then
Eq. 5-42
Thus, Method 1 (and also Method 2) produces a decaying envelope that the user may
incorrectly interpret as additional damping.
The use of equal frequency intervals versus unequal intervals has been studied and results
are shown in the MSC.Nastran Aeroelastic Analysis Users Guide.
Input
(a) True Transient
(b) Periodic Representation
P(t) u(t)
u(t)
t
t
t
P(t)
t t
t
1
f
-----
Damped Damped
f 1 T
pul se
T
decay
+ ( )
C
n
t ( )
C
n
t ( )

1 t ( ) cos
1 2 t ( )
2
------------------------------------ 1
1
6
--- t ( )
2
+
* +
, -
=
Method 0
Method 1
"
.
$
.
&
=
1 CHAPTER 5
5.6 Viscoelastic Material Properties
The mechanical behavior of many glasses, rubbers, and high polymers under stress is
described in terms of a combination of elastic and viscous phenomena that may be
approximated by linear viscoelastic theory. The dynamic behavior of such material is
characterized by a strong dependence on frequency; and, under sinusoidal excitation, these
materials exhibit a steady-state response in which the stress lags the associated strain. The
general three-dimensional state of stress can then be described in terms of a complex
frequency-dependent shear modulus and a real constant value for Poissons ratio.
This method was developed for the analysis of rubber-like materials such as solid rocket
propellants. In that case, the dominant strain energy terms were the shear strains.
MSC.Nastran provides the ability to represent a single complex frequency-dependent
scalar material modulus of the form
Eq. 5-43
where:
The ratio
is denoted as the shear loss tangent.
The above formulation of viscoelastic (frequency-dependent) material properties may be
used in direct frequency analysis (SOL 108).
The stiffness and damping components of the dynamic matrices for direct frequency
response analysis are documented in the MSC.Nastran Basic Dynamics Users Guide in the
following form:
Eq. 5-44
Eq. 5-45
where:
= Shear storage modulus
= Shear loss modulus
= overall structural damping specified through the PARAM,G Bulk Data entry
= stiffness matrix for structural elements
= stiffness terms generated through direct matrix input, e.g., DMIG Bulk Data
entries
G f ( ) G f ( ) iG f ( ) + =
G
G
G f ( )
G f ( )
-------------- tan =
K
dd
[ ] 1 i g + ( ) K
dd
1
[ ] K
dd
2
[ ] K
dd
4
[ ] + + =
B
dd
[ ] B
dd
1
[ ] B
dd
2
[ ] + =
g
K
dd
1
[ ]
K
dd
2
[ ]

Eq. 5-44 is of particular interest in the current discussion of viscoelastic material properties
because the presence of these properties will be reflected in terms of this equation. For
discussion purposes, frequency-dependent material properties will be denoted as
viscoelastic materials and those material properties that are independent of frequency will
be denoted as elastic materials. Thus, if the stiffness properties for the viscoelastic elements
are initially computed on the basis of a representative reference modulus, , the
stiffness matrix for the viscoelastic elements (denoted by the subscript ) may be written
in the form
Eq. 5-46
Input Description
To use the viscoelastic capability, the following conditions are necessary:
1. Assume the matrix will be restricted only to the viscoelastic elements. This
restriction implies that elastic elements will have a blank or zero entry for on
their associated MATi Bulk Data entries. Conversely, all viscoelastic materials
must have representative reference values of , and entered on their
associated MATi Bulk Data entries. Then, by definition,
2. The TABLEDi tabular functions and are defined to represent the
complex moduli of all viscoelastic materials.
These two conditions may be combined in Eq. 5-44 to provide the following expression:
Eq. 5-47
A comparison of Eq. 5-46 and Eq. 5-47 yields the form of the tabular functions and
:
Eq. 5-48
= element damping matrix generated by the multiplication of individual
element stiffness matrices by an element damping, , entered on the MATi
Bulk Data entry associated with the element or elements in question
= damping matrix generated through CVISC and CDAMP1 Bulk Data entries
= damping terms generated through direct matrix input, e.g., DMIG Bulk Data
entries
K
dd
4
[ ]
g
e
B
dd
1
[ ]
B
dd
2
[ ]
G
REF
V
K
dd
f ( ) [ ]
V
G f ( ) i G f ( ) +
G
REF
-------------------------------------- K
dd
1
[ ]
V
=
K
dd
1
[ ]
g
e
g
e
G
REF
K
dd
4
[ ]
V
g
REF
K
dd
1
[ ]
V
=
TR f ( ) TI f ( )
K
dd
[ ]
V
1 ig + ( ) K
dd
1
[ ]
V
TR f ( ) i TI f ( ) + { } K
dd
4
[ ]
V
+ =
1 g
REF
TR f ( ) + ( ) i g g
REF
TI f ( ) + [ ] + { } K
dd
1
[ ]
V
=
TR f ( )
TI f ( )
TR f ( )
1
g
REF
-------------
G f ( )
G
REF
-------------- 1 =
1 CHAPTER 5
Eq. 5-49
Note that the direct input matrix, , from Eq. 5-44 is still available but not involved in
the definition of viscoelasticity.
Direct frequency response analyses that involve viscoelastic materials require some special
input data relative to analyses that involve only elastic materials. These special input
requirements are given below:
1. Executive Control Section:
None
2. Case Control Section:
SDAMPING = reference TABLEDi Bulk Data entry that defines the alternate
tabular form of
3. Bulk Data Section:
a. MATi Bulk Data entry
, the reference modulus
ratio for the viscoelastic material
, the reference element damping
All other entries on the MATi Bulk Data entry are utilized in the standard
manner.
b. TABLEDi Bulk Data entries:
A TABLEDi Bulk Data entry with an is used to define the function
of Eq. 5-48.
A TABLEDi Bulk Data entry with an is used to define the
function of Eq. 5-49.
All other input requirements to the MSC.Nastran Bulk Data entry are typical of direct
frequency response analysis. Note that the overall structural damping, , entered through
the PARAM Bulk Data entry (PARAM,G,XX) applies to all elastic materials.
Compare With Theory. The functional form of Eq. 5-48 and Eq. 5-49 requires the user to
perform some modest calculations that involve and . These two terms must also
be input to MSC.Nastran through the MATi Bulk Data entry for the viscoelastic materials.
In general, representative values of these parameters should be used. However, in those
cases where the MSC.Nastran OMIT feature is not used, the calculation of and
can be simplified. In this case, define
and select so that
TI f ( )
1
g
REF
-------------
G f ( )
G
REF
-------------- g =
K
dd
2
[ ]
n
TR f ( )
G G
REF
=
NU Poisson s =
GE g
REF
=
ID n =
TR f ( )
ID n 1 + =
TI f ( )
g
G
REF
g
REF
TR f ( ) TI f ( )
G
REF
G
G
REF
G
g
-------
g
REF

Then Eq. 5-48 and Eq. 5-49 reduce to the following form:
Eq. 5-50
Eq. 5-51
The above simplifications should not be used if OMIT calculations are involved in the
analysis in order to avoid possible matrix ill-conditioning.
Note that stress and force data recovery calculations are performed with the reference
moduli irrespective of frequency.
Example
To illustrate the representation of viscoelastic material properties in MSC.Nastran,
consider the following structure that may undergo both axial extension along the z-axis
and torsion about the z-axis:
where:
flywheel mass,
= 2.0
flywheel inertia,
= 10.0
axial stiffness,
=
torsional stiffness,
=
length,
= 2.0
area,
= .9
area moment,
= 2.0
G
REF
g
REF
1 =
TR f ( ) G f ( ) =
TI f ( ) G f ( ) =
z
l
M
z
I
z
K
z
E f ( )A
l
--------------
K
z
G f ( )J
l
--------------
l
A
J
1 CHAPTER 5
The symbols and imply that the extensional and torsional moduli are functions
of frequency, i.e., viscoelastic. For simplicity, it is assumed that and that these
quantities have the following frequency-dependent characteristics:
A model for the system is shown in the following schematic:
This model can be generated with the following MSC.Nastran Bulk Data entries:
f, hz G(f) G(f)
.8 1800. 180.
1.1 1850. 185.
1.4 1910. 191.
1.7 1970. 197.
2.0 2030. 203.
2.3 2070. 207.
2.6 2140. 214.
2.9 2210. 221.
1 2 3 4 5 6 7 8 9 10
GRID ID CP X1 X2 X3 CD PS
GRID 2 2. 1245
GRID 3 123456
E f ( ) G f ( )
E f ( ) G f ( ) =
3
21
2, 3
23
2, 6
24
23
i
i,j
k
= Grid Point ID
= Grid Point ID and Component No.
= Element ID
3
K
z
K
z
I
z
z

Note that the reference values of 2000. for both and are specified on the MAT1 Bulk
Data entry. The reference value for structural damping, , is set to .09 under the GE
field of the MAT1 Bulk Data entry. Once the reference values and have been
assigned, one can evaluate Eq. 5-48 and Eq. 5-49 for the values to be assigned to and
. The values for these functions are entered on TABLEDi Bulk Data entries. As elastic
portions of the structure may exist in addition to viscoelastic portions, assume that a value
of overall structural damping, g, is to be utilized for these elastic portions of the model. The
overall structural damping value of .06 is assigned through the following PARAM Bulk
Data entry.
This value of must be considered in Eq. 5-49. The evaluation of Eq. 5-48 and Eq. 5-49 will
result in the values shown in the following TABLED1 Bulk Data entries:
CMASS2 EID M G1 C1 G2 C2
CMASS2 21 2. 2 3 3 3
CMASS2 24 10. 2 6 3 6
CROD EID PID G1 G2
CROD 23 1 2 3
PROD PID MID A J
PROD 1 1 .9 2.
MAT1 MID E G NU RHO A TREF GE
MAT1 1 2000. 2000. .09
1 2 3 4 5 6 7 8 9 10
PARAM N V1 V2
PARAM G .06
1 2 3 4 5 6 7 8 9 10
TABLED1 ID
TABLED1 ID TABR1
X
1
Y
1
X
2
Y
2
X
3
Y
3
X
4
Y
4
+ABR1 .0 .0 .8 1.11111 1.1 .833333 1.4 .5 TABR2
X
5
Y
5
X
6
Y
6
X
7
Y
7
X
8
Y
8
+ABR2 1.7 .166667 2. .1666667 2.3 .3888889 2.6 .7777777 TABR3
E G
g
REF
G
REF
g
REF
TR f ( )
TI f ( )
g
1 CHAPTER 5
To demonstrate that elastic as well as viscoelastic elements may be included in the same
analysis, the following single degree-of-freedom is added to the Bulk Data Section:
The following Bulk Data entries are required to represent the foregoing single
degree-of-freedom oscillator:
X
9
Y
9
+ABR3 2.9 .5611111 ENDT
TABLED1 ID
TABLED1 11 TABI1
X
1
Y
1
X
2
Y
2
X
3
Y
3
X
4
Y
4
+ABI1 0. 0. .8 .3333333 1.1 .3611111 1.4 .3944444 TABI2
X
5
Y
5
X
6
Y
6
X
7
Y
7
X
8
Y
8
+ABI2 1.7 .4277778 2. .4611111 2.3 .4833333 2.6 .5222222 TABI3
X
9
Y
9
+ABI3 2.9 .5611111 ENDT
1 2 3 4 5 6 7 8 9 10
CELAS2 EID K G1 C1 G2 C2 GE S
CELAS2 1 1000. 1
3
1
i
j
= Scalar Point ID
= Element ID
2
B = 2.
g = .06
m = 10.
1 U
K = 10000.

The excitation for both disjoint models is a force (moment for the torsional system) with a
magnitude of . This function can be generated with the following Bulk Data entries:
To perform a frequency response analysis is necessary to provide a list of frequencies (Hz)
at which solutions are desired. The following FREQ1 Bulk Data entry is used for this
purpose.
CDAMP2 EID B G1 C1 G2 C2
CDAMP2 2 2. 1
CMASS2 EID M G1 C1 G2 C2
CMASS2 3 10. 1
1 2 3 4 5 6 7 8 9 10
RLOAD2 SID L M N TB TP
RLOAD2 1 1 0 0 1 0
TABLED1 ID
TABLED1 1 +ABC
X
1
Y
1
X
2
Y
2
X
3
Y
3
+ABC 0. 1. 1. 1. 100. 1. ENDT
DAREA SID P C A
DAREA 1 2 3 1.
DAREA 1 2 6 1.
DAREA 1 1 1.
FREQ1 SID F1 DF NDF
FREQ1 1 .5 .3 10
t cos
1 CHAPTER 5
The complete MSC.Nastran data entry for the two disjoint problems is listed below:
ID TEST,DAMPING
SOL 26
CEND
TITLE=FREQUENCY RESPONSE
SUBTITLE=VISCOELASTIC AND ELASTIC MATERIALS
DLOAD=1
SDAMP=10
FREQ=1
SPC=1
SET 1=1,2
DISP=1
VELO=1
ACCE=1
SET 7=23
STRESS=7
BEGIN BULK
$ INCLUDE ALL BULK DATA ENTRIES FOR BOTH
$ DISJOINT MODELS DISCUSSED IN THE PRECEDING
$ REMARKS
ENDDATA

5.7 Free Body Techniques
Free body motion in a structure occurs when a structure may move freely without applied
forces. Although the stiffness matrix will have one or more singularities, the combined
dynamic system, with mass and damping, may not be singular. Examples are flying
objects such as aircraft or rockets, and structures with kinematic mechanisms such as a
gyroscope or a pendulum.
In most cases, the dynamic response solutions in MSC.Nastran do not require any special
attention for free body motions. In general, MSC.Nastran will decompose matrix
combinations that are not singular. For example, an unbalanced load applied to a free body
has a unique dynamic solution, namely a constant acceleration. The only exception is that
of a frequency response analysis at a frequency of zero. In this case it is recommended that
a small frequency be used instead of zero for free bodies.
In some cases, it may be necessary to use the SUPORT Bulk Data to provide a method of
defining the free body modes of a structure. The primary use for this process is in inertia
relief option for static analysis. (See the MSC.Nastran Linear Static Analysis Users Guide.)
In other cases, the basic geometry is used for the definition of the rigid body displacements.
The parameter GRDPNT specifies a grid point or a location for the six rigid body
displacement vectorsthree translations and three rotations. This method is used in the
weight and CG calculations and is used for inertia relief with superelements.
The SUPORT Option
The SUPORT data defines a set of degrees-of-freedom that are used to restrain free motion
of the structure temporarily for the purposes of calculating normal modes, providing a
moving frame of reference, or measuring unbalanced loads. The basic theory and usage
are given in Overview on page 88 of the MSC.Nastran Basic Dynamics Users Guide, and
other MSC users guides. Other bulk data variations that perform this function are CYSUP
Bulk Data (cyclic symmetry) and SESUP Bulk Data (superelements).
In dynamic analysis, some specialized areas in which it is useful to apply free body
SUPORTs include:
1. Calculation of pure zero-frequency eigenvalues and, if necessary, their
eigenvectors. The result depends on the chosen eigenvalue extraction method. In
the Inverse Power method, the zero frequency eigenvectors are obtained from the
stiffness matrix. In the other real eigenvalue methods, the number of SUPORT
degrees-of-freedom indicate the number of zero-frequency modes to be
calculated.
2. Use of the large mass technique to enforce motion in a modal formulation. Here
the SUPORTs are essential to obtain frequencies that are exactly zero. (See
Enforced Motion with Loads on page 87.)
3. Response Spectra analysis with participation factors. Here the responses are
calculated relative to a moving base, defined by SUPORTs (see Shock and
Response Spectrum Analysis on page 221).
1 CHAPTER 5
4. Verification of new FE models with a free body test will find several types of
errors. It is useful to remove the constraints for a fixed structure to perform extra
check runs. For instance, the calculations for SUPORT inputs will indicate if
additional mechanism modes exist.
5. SUPORT data is required for mode acceleration data recovery of free bodies. In
this calculation it is necessary to solve a quasi static equation using the stiffness
matrix. The use of SUPORTs will eliminate singularities without
over-constraining the structure.
6. SESUP data are used for modal analysis of individual superelements. This is used
for Component Mode Synthesis when SECSET boundaries are used to obtain free
body modes.
Each of these topics is described elsewhere in this guide. However, they all have the same
limitations and requirements on the selection of the SUPORT set.
The SUPORT inputs define a set of degrees-of-freedom, , which belong to the partitions
Eq. 5-52
where are the dynamic QSET or component modal coordinates, and are the physical
degrees-of-freedom for the restrained structure. The free body motions of the structure
may be obtained by solving a problem with zero forces. The partitioned static matrix
solution is
Eq. 5-53
where is called the rigid body transformation matrix. It is obtained by solving the
equation:
Eq. 5-54
Depending on the type of solution, the matrix may be used for extracting
zero-frequency mode shapes, for finding relative deformation vectors and for measuring
resultants for load and force vectors. The number of columns of equals the number of
user-selected SUPORT points. This number may be less than six if other constraints are
applied, or may be greater than six if mechanisms are present. (Usually six
degrees-of-freedom are chosen at a hard point near the center of gravity.)
The requirements for the set are that they provide a nonredundant set of supports,
namely that:
u
r
u
a
{ }
q
u
l
u
r
" #
. .
$ %
. .
& '
=
q u
l
K
l l
K
l r
K
rl
K
rr
D
l
I
P { } 0 { } = =
D
l
[ ]
K
l l
[ ] K
l
[ ] K
l r
[ ] =
D
l
[ ]
D
l
[ ]
u
r

1. If the structure were to be constrained to zero displacement at the SUPORT DOF,
no singularities will occur in the resulting stiffness matrix partition; i.e.,
exists Eq. 5-55
2. If the unloaded structure were to be given enforced displacements at each of the
SUPORT DOF (with the others constrained), no stresses or forces should occur in
the structure; i.e.,
Eq. 5-56
3. The SUPORT DOF are connected to structural elements which provide real
supporting stiffness. (Points near the CG are preferred.)
4. In superelement analysis the final SUPORT set is restricted to the residual
structure. (However, SESUP data may be used to define separate sets for
modal analysis of upstream superelements.)
If any of these criteria are violated, the table output with User Information Message 3035
will contain large values for the epsilon and strain energy. The equations defining these
parameters are given in Overview on page 88 of the MSC.Nastran Basic Dynamics Users
Guide.
Geometric Shape Methods
In these cases, the rigid body shapes are defined geometrically by the locations of the grid
points and orientation of their displacement coordinate systems. The structure may or may
not be a free body. Also, mechanisms that produce free body differences in displacements
or rotations are not included.
The parameter GRDPNT is used normally to calculate the total weight, inertia, and CG
properties of a structure. It is also used to generate geometric rigid body shapes. The
matrix as defined in Eq. 5-56 may be calculated directly from the geometry. The
VECPLOT Module generates a matrix, , for every grid point, , from basic
kinematics. The definition is
Eq. 5-57
where is the vector of three displacements and three rotations in the local coordinate
system. represents three displacements and three rotations at GRDPNT (or the origin)
in basic coordinates. These partitions are merged into a full-size matrix, with six
rows for each grid point and six columns total. Note that the rows of are null for the
scalar points and is a unit matrix for the six rows corresponding to the GRDPNT parameter.
If SUPORT data is present, the matrix is transformed to define motion relative to the
supported degrees-of-freedom, . The internal calculations are given below
K
l l
[ ]
1
K
rl
[ ] D
l r
[ ] K
rr
[ ] + 0 [ ] =
u
r
D
l
[ ]
6 6 D
i
[ ] g
u
gi
{ } D
i
[ ] u
o
{ } =
u
gi
{ }
u
o
{ }
6 6 D
go
[ ]
D
go
[ ]
u
r
{ }
1 CHAPTER 5
Partition:
Eq. 5-58
Invert the support partition, and multiply:
Eq. 5-59
and merge:
Eq. 5-60
These matrices may be calculated at the superelement level as opposed to assembled
residual SE level required by the SUPORT partition method.
The applications include:
1. Static inertia relief solutions for SOL 101. The requirement to combine and reduce
mass matrices is eliminated with this method, saving nearly 50% of the execution
cost. For inertia relief effects, these solutions also require the input
PARAM,INREL,1 and a set of six SUPORT degrees-of-freedom.
2. Component Mode Synthesis for superelements provides an option for inertia
relief vectors. The advantage is improved accuracy in the final results with an
equal or lesser number of component modes. These are deformation shapes that
correspond to the six uniform acceleration loads on the constrained structure.
The deformation shape vectors are appended to the mode shapes and are
included as generalized degrees-of-freedom. The INRLM parameter controls this
option.
Note that in addition to the limitation of six free body motions, the geometric method will
not be correct for scalar points and fluid analysis methods.
D
go
[ ]
D
go
D
ro
1
D
gr
[ ] D
go
[ ] D
ro
[ ]
1
=
D
gr
[ ]
D
gr
I

5.8 Aeroelastic Solutions
The MSC.Nastran Aeroelastic Analysis Users Guide describes the theoretical aspects and the
numerical techniques used to perform aeroelastic analyses with MSC.Nastran. As
described in Overview of Aeroelastic Analysis on page 81, the system is used for flutter,
frequency response, gust response, and static analysis of aerodynamically loaded
structures. An outline of the capability is given here.
The aeroelastic analyses use the following features:
Structural Model. Any of the existing MSC.Nastran structural finite elements (except
axisymmetric and p-elements) can be used to build the structural model. The structural
stiffness, mass, and damping matrices required by the aeroelastic analyses are generated
by MSC.Nastran from the user input of geometric, structural, inertial, and damping data,
for subsequent use in the various aeroelastic analyses.
Fluid/Structure Connections. Matrices of aerodynamic influence coefficients are
computed only from the data describing the geometry of the aerodynamic finite elements.
The choice of aerodynamic grid points for the aerodynamic model is independent of the
location of the structural grid points. An automated interpolation procedure is provided
to relate the aerodynamic to the structural degrees-of-freedom. Splining techniques for
both lines and surfaces are used to generate the transformation matrix from structural grid
point deflections to aerodynamic grid point deflections where local streamwise slopes are
also computed. The transpose of this matrix transfers the aerodynamic forces and
moments at aerodynamic boxes to structural grid points.
Aerodynamic Theories. One subsonic and three supersonic lifting surface aerodynamic
theories are available in MSC.Nastran, as well as Strip Theory. The subsonic theory is the
Doublet-Lattice method, which can account for interference among multiple lifting
surfaces and bodies. The supersonic theories are the Mach Box method, Piston Theory, and
the ZONA51 method for multiple interfering lifting surfaces that was added to
MSC.Nastran in Version 67.
Static Aeroelastic Analysis. The structural load distribution on an elastic vehicle in
trimmed flight is determined by solving the equations for static equilibrium. The SOL 144
and SOL 200 processes will calculate aerodynamic stability derivatives (e.g., lift and
moment curve slopes and lift and moment coefficients due to control surface rotation) and
trim variables (e.g., angle of attack and control surface setting) as well as aerodynamic and
structural loads, structural deflections, and element stresses.
Modal Formulation. Dynamic aero solutions provide for modal reduction of the system
matrices. The number of degrees-of-freedom required for accurate solutions to dynamic
aeroelastic problems is generally far less than the number of physical degrees-of-freedom
used in the finite element structural model. The number of independent
degrees-of-freedom can be greatly reduced by using the (complex) amplitudes of a series
of vibration modes as generalized coordinates, e.g., by Galerkins method. MSC.Nastran
1 CHAPTER 5
can compute the vibration modes and frequencies and make the transformation to modal
coordinates. The matrices of aerodynamic influence coefficients are also transformed to
generalized aerodynamic forces by use of the vibration eigenvectors.
Flutter Analysis. The dynamic aeroelastic stability problem, flutter, is solved in SOL 145,
by any of three methods. The traditional American flutter method developed by the Air
Materiel Command (AMC) in 1942 is available in the first two methods. The first method
is called the K-method and is a variation of the AMC method. The second method, called
the KE-method, is more efficient from the point of view of tracking roots, but is limited in
input (no viscous damping) and output (no eigenvectors). The third method, called the
PK-method, is similar to the British flutter method, which was developed by the Royal
Aircraft Establishment.
Frequency Response. The coupling with aerodynamic loads has also been added to the
existing MSC.Nastran structural modal frequency response capability, SOL 146. Analyses
of frequency response to arbitrarily specified forcing functions can be carried out using the
oscillatory aerodynamic loads from any of the available aerodynamic theories. Frequency
response to a harmonic gust field can be calculated at subsonic speeds using the
Doublet-Lattice method for wing/body interference, and by the ZONA51 method for
interfering lifting surfaces at supersonic speeds.
Transient Response. Because unsteady aerodynamic loads are obtained only for
steady-state harmonic motion, they are known only in the frequency- and not the
time-domain. In SOL 146, Inverse Fourier Transform techniques provide the appropriate
methods by which transient response is obtained from the frequency response. Both
forward and inverse Fourier transforms are provided so that the time-varying forcing
function or the gust profile can be transformed into the frequency domain. Then, after
convolution with the system frequency response, the inverse transform leads to the
transient response of the system to the specified forcing function or gust profile.
Random Response. Stationary random response of the system, is available in SOL 146
from specified loadings and the power spectral densities of loads. Loads may be either
specified force distributions or harmonic gust fields. The statistical quantities of interest in
the response are , the ratio of standard deviations (rms values) of the response to that of
the input loading, and , the mean frequency of zero crossings (with a positive slope) of
the response. The capability to compute these quantities was added to MSC.Nastran by
modifying the existing random response module to include options to generate various
atmospheric turbulence power spectra and to perform the calculation of .
Design Sensitivities. The sensitivities of response parameters to changes in design
variables are calculated by the perturbation techniques developed for structural
optimization in MSC.Nastran and extended to include static aeroelasticity and flutter in
SOL 200. The basic aeroelastic sensitivities that can be obtained include stability
derivatives, trim variables, and flutter system dampings. The synthetic response technique
of MSC.Nastran optimization also permits the calculation of sensitivities of user-specified
functions of those standard response quantities.
A
N
o
N
o

Aeroelastic Optimization. Optimization of aeroelastic characteristics can be combined
with the other optimization features of MSC.Nastran in SOL 200, and vehicles can now be
designed optimally for aeroelastic loads, flying qualities, and flutter, as well as for strength,
vibration frequencies, and buckling characteristics.
Introduction to Aeroelastic Analysis and Design
Aeroelastic analysis and design solution sequences extend the range of capabilities in
MSC.Nastran beyond basic static and dynamic structural analysis.
Aeroelastic Modules. Options are available to:
Generate aerodynamic grid points.
Compute aerodynamic matrices.
Provide connection (interpolation) between the structural and aerodynamic grid
points.
Solve the equations for static aeroelasticity.
Solve the equations for flutter.
Solve the equations for dynamic aeroelastic response.
Calculate aeroelastic design sensitivities.
Optimize aeroelastic and related structural characteristics.
Aeroelastic DMAP Sequences. Four solution sequences are available:
1. SOL 144 for static aeroelastic analyses.
2. SOL 145 for modal flutter analyses by the K-, KE-, or PK-methods.
3. SOL 146 for modal dynamic aeroelastic response analyses due to gusts or control
surface deflections.
4. SOL 200 for design sensitivity and optimization including aeroelastic effects.
Since this fourth sequence has applications to many areas other than
aeroelasticity, the reader is referred to the MSC.Nastran Design Sensitivity and
Optimization Users Guide for a more comprehensive treatment. Aeroelastic
optimization is beyond the scope of this Guide and is not discussed further in this
section.
Aerodynamic Theories
MSC.Nastran has implemented six aerodynamic theories:
1. Doublet-Lattice subsonic lifting surface theory (DLM)
2. ZONA51 supersonic lifting surface theory
3. Subsonic wing-body interference theory (DLM with slender bodies)
4. Mach Box method
1 CHAPTER 5
5. Strip Theory
6. Piston Theory
Each of these methods is described in the MSC.Nastran Aeroelastic Analysis Users Guide.
They all share a common matrix structure.
Three matrix equations summarize the relationships required to define a set of
aerodynamic influence coefficients [see Rodden and Revell (1962)]. These are the basic
relationships between the lifting pressure and the dimensionless vertical or normal
velocity induced by the inclination of the surface to the airstream; i.e., the downwash (or
normalwash),
Eq. 5-61
the substantial differentiation matrix of the deflections to obtain downwash,
Eq. 5-62
and the integration of the pressure to obtain forces and moments,
Eq. 5-63
where:
The Aerodynamic Influence Coefficient Matrix. The three matrices of Eq. 5-62 and
Eq. 5-63 can be combined to give an aerodynamic influence coefficient matrix:
Eq. 5-64
= downwash (dimensionless)
= static aerodynamic downwash; it includes, primarily, the static incidence
distribution that may arise from an initial angle of attack, camber, or twist
= pressure on lifting element
= flight dynamic pressure
= reduced frequency, where is the angular frequency, is a
reference semichord, and is the free-stream velocity
= aerodynamic influence coefficient matrix, a function of Mach number ( ),
and reduced frequency ( )
,
= displacements and forces at aerodynamic grid points
,
= real and imaginary parts of substantial differentiation matrix, respectively
(dimensionless)
= integration matrix
w
j
{ } A
j j
[ ] f
j
q { } =
w
j
{ } D
j k
1
i k D
j k
2
+ [ ] u
k
{ } w
j
g
{ } + =
P
k
{ } S
kj
[ ] f
j
{ } =
w
j
w
j
g
f
j
j
q
k
k b V = b
V
A
j j
m, k ( )
m
k
u
k
P
k
D
j k
1
D
j k
2
S
kj
Q
kk
[ ] S
kj
[ ] A
j j
[ ]
1
D
j k
1
i k D
j k
2
+ [ ] =

All aerodynamic methods compute the , and matrices at user-supplied Mach
numbers and reduced frequencies. The Doublet-Lattice and ZONA51 theories compute
the matrix. Then, matrix decomposition and forward and backward substitution are
used in the computation of the matrix. The remaining methods compute directly
and use matrix multiplications to form . Details of the various methods are described in
the MSC.Nastran Aeroelastic Analysis Users Guide.
Generation of Aerodynamic Matrices
The aerodynamic equations described above form the basis of the aerodynamic
computations required for static aeroelastic analysis with some special purpose
modifications made for the MSC.Nastran implementation.
For static aeroelasticity, the downwash relation of Eq. 5-62 becomes
Eq. 5-65
where:
Static Aeroelastic Equations of Motion
The aerodynamic forces are transferred to the structure reduced to the a-set to form an
aerodynamic influence coefficient matrix, , which provides the forces at the structural
grid points due to structural deformations, i.e.,
and a second matrix, , which provides forces at the structural grid points due to unit
deflections of the aerodynamic extra points, :
= a vector of aerodynamic degrees-of-freedom (e.g., angles of attack)
= a vector of aerodynamic displacements (deformations)
= a vector of extra aerodynamic points used to describe, e.g., aerodynamic
control surface deflections and overall rigid body motions
= represents an initial static aerodynamic downwash. It includes, primarily, the
static incidence distribution that may arise from an initial angle of attack,
camber, or washout (twist)
= a substantial derivative matrix for the aerodynamic displacements. This is the
term of Eq. 5-62. The term is not used for this quasi-steady analysis.
= a substantial derivative matrix for the extra aerodynamic points
S D
1
D
2
A
Q A
1
Q
w
j
{ } D
j k
[ ] u
k
{ } D
j x
[ ] u
x
{ } w
j
g
{ } + + =
w
j
{ }
u
k
{ }
u
x
{ }
w
j
g
{ }
D
j k
[ ]
D
j k
1
D
j k
2
D
j x
[ ]
Q
aa
F
a
{ } Q
aa
[ ] u
a
{ } =
Q
ax
u
x
{ }
F
x
{ } Q
ax
[ ] u
x
{ } =
1 CHAPTER 5
The complete equations of motion in the a-set degrees-of-freedom require
The a-set equations are then:
Eq. 5-66
This is the basic set of equations used for static aeroelastic analysis. In the general case,
rigid body motions are included in the equations to represent the free-flying characteristic
of an air vehicle. This is addressed in MSC.Nastran by a requirement that the user identify
reference degrees-of-freedom equal in number to the number of rigid body motions using
the SUPORT Bulk Data entry. Eq. 5-66 is then partitioned into r-set (supported) and l-set
(left over) degrees-of-freedom, yielding
Eq. 5-67
where the notation
has been introduced.
At this point the MSC.Nastran implementation of aeroelastic analysis introduces a
mathematical technique that is based on the MSC.Nastran inertia relief analysis without
aeroelastic effects. The technique entails multiplying the first row of Eq. 5-67 by and
adding the result to the second row. Assuming that a steady-state condition exists, the
accelerations may be constrained and the system may be solved for steady-state loads or
divergence. The stability derivatives and static control system coefficients may also be
obtained.
Flutter Solution Techniques
Flutter is the dynamic aeroelastic stability problem. It can be solved in any speed regime
simply by selecting the appropriate aerodynamic theory. In the linear case assumed
throughout this guide, the solution involves a series of complex eigenvalue solutions; the
eigenvalue problem to be solved depends on the way in which the aerodynamic loads are
included in the equations of motion or whether certain damping terms are included.
Structural stiffness matrix
Structural mass matrix
Vector of applied loads (e.g., mechanical, thermal, and gravity loads
plus aerodynamic terms due to user input pressures and/or
downwash velocities)
K
aa
[ ]
M
aa
[ ]
P
a
{ }
K
aa
q Q
aa
[ ] u
a
{ } M
aa
[ ] U
a
{ } + q Q
ax
[ ] u
x
{ } P
a
{ } + =
K
l l
a
K
l r
a
K
rl
a
K
rr
a
u
l
u
r
" #
$ %
& ' M
l l
M
l r
M
rl
M
rr
u
l
u
r
" #
$ %
& '
+
K
l x
a
K
rx
a
u
x
{ }
P
l
P
r
" #
$ %
& '
+ =
K
aa
a
[ ] K
aa
q Q
aa
[ ] =
K
ax
a
[ ] q Q
ax
[ ] =
D
T

The manner in which the aerodynamic loads are included depends on how the
dimensionless oscillatory aerodynamic coefficients are defined. When Theodorsen (1935)
first developed the American method (K-method) of flutter analysis, he introduced the
aerodynamics into a vibration analysis as complex inertial terms and the flutter analysis
became a vibration analysis requiring complex arithmetic. At the same time, he introduced
an artificial complex structural damping, proportional to the stiffness, to sustain the
assumed harmonic motion. Flutter analysis is then a double eigenvalue problem in
frequency and velocity, and an iterative solution, using the reduced frequency of the
assumed harmonic motion as the iteration parameter, leads to the neutrally stable
conditions (flutter frequencies and velocities) at which no artificial damping is required.
The artificial damping is therefore seen not to be physically meaningful, other than,
perhaps, at speeds near flutter speeds.
Generalized Aerodynamic Matrices
Eq. 5-64 defines an aerodynamic influence coefficient matrix that is computed based
on the aerodynamic model. In order for this matrix to be useful in a flutter analysis, two
transformations must take place:
1. The matrices must be applied to the structural model using the spline techniques.
2. A modal reduction must be applied to obtain the matrices in generalized form.
Mathematically, those transformations can be expressed as
Eq. 5-68
where:
A level of complexity is added if the flutter analysis includes the use of extra points. Extra
points are used for the representation of control systems and are therefore required in
aeroservoelastic analyses. The flutter analysis then uses a merged matrix
Eq. 5-69
in which the h-set is a combination of the i-set normal modes and the e-set extra points. It
is seen that the lower e-set rows in the matrix are null. Physically, this indicates that the
normal mode deflections do not produce aerodynamic forces on the extra points ( )
and that the extra point deflections do not produce aerodynamic loads on the extra points
( ).
= the generalized aerodynamic matrix
= a matrix of i-set normal mode vectors in the physical a-set
= the spline matrix reduced to the a-set
WTFAC
T
= a weighting factor matrix defined by the user
Q
kk
Q
i i
[ ]
ai
[ ]
T
G
ka
[ ]
T
WTFACT [ ] Q
kk
[ ] G
ka
[ ]
ai
[ ] =
Q
i i
ai
G
ka
W
kk
Q
hh
[ ]
Q
i i
Q
i e
0 0
=
Q
ei
0 =
Q
ee
0 =
2 CHAPTER 5
The K-Method of Flutter Solution
The basic equation for modal flutter analysis by the K-method is
Eq. 5-70
where:
Note that , , and are not independent.
For the K-method of solution, the aerodynamic term is converted to an equivalent
aerodynamic mass
Eq. 5-71
The term involving in Eq. 5-71 has been multiplied by for mathematical
convenience, and is valid only at flutter, i.e., when . Eq. 5-71 is solved as an
eigenvalue problem for a series of values for parameters , , and . The complex
eigenvalue is , which can be interpreted as real values of and . The velocity,
, is recovered from . Flutter occurs for values of , , and for which .
The solutions are not valid except when , since the aerodynamic force terms are valid
only for sinusoidal motion and is not a physical damping.
The K-method of flutter analysis is a looping procedure. The values of , , and
are solved for various values of , , and . Plots of versus can be used to determine
the flutter speed(s) (where goes through zero to positive values). The KE-method and
= modal mass matrix, usually (but not necessarily) diagonal
= modal damping matrix
= modal stiffness matrix, usually (but not necessarily) diagonal; may be
complex (with actual structural damping); will be singular if there are rigid
body modes
m = Mach number
k = reduced frequency =
= reference length
= aerodynamic force matrix, which is a function of parameters and
= circular frequency =
g = artificial structural damping
= fluid density
V = overall forward velocity
= modal amplitude vector, sometimes called modal participation factors
M
hh
2
i B
hh
1 i g + ( )K
hh
1
2
--- V
2
* +
, -
Q
hh
m, k ( ) + + u
h
{ } 0 =
M
hh
B
hh
K
hh
.c 2V
c
Q
hh
m, k ( )
m k
2f
u
h
k V
M
hh
2
---
c
2k
------
* +
, -
2
Q
hh
m, k ( ) +

2
1 ig +
---------------- B
hh
i
1 i g +
-------------------- K
hh
+ + u
h
{ } 0 =
B
hh
1 ig +
g 0 =
m k
2
1 ig + ( ) g
V V c 2k = m k g 0 =
g 0 =
g
V g f 2 =
m k V g
g

the PK-method are the other major flutter options. These are discussed in the MSC.Nastran
Aeroelastic Analysis Users Guide. Typical flutter plots are shown in Figure 5-4 and
Figure 5-5 for the output from the PK-method.
Figure 5-4 V-f Curve for BAH Wing
Figure 5-5 V-g Curve for BAH Wing
Note that the bending branch goes to zero frequency before the mode goes unstable.
4.0
3.0
2.0
1.0
0
0 500 1000 1500 2000
F
r
e
q
u
e
n
c
y
,

f

(
H
z
)
Velocity (ft/s)
1st Mode (Bending)
2nd Mode (Torsion)
1st Mode (Bending)
2nd Mode (Torsion)
0.2
0.1
0.0
-0.1
-0.2
0 500 1000 1500 2000
D
a
m
p
i
n
g
,

g
Velocity (ft/s)
Flutter Divergence
2 CHAPTER 5
Dynamic Aeroelastic Analysis
Dynamic aeroelasticity differs from the flutter analysis described in the previous section in
that the right-hand side of Eq. 5-70 is no longer zero. Instead, loading, which can be in
either the frequency or the time domain, is applied. For both types of loading,
MSC.Nastran performs the primary analyses in the frequency domain. If the user has
supplied loadings in the time domain, Fourier Transform techniques are used to convert
the loadings into the frequency domain, a frequency response analysis is performed, and
the computed quantities are transformed back to the time domain using Inverse Fourier
Transform techniques. This section first describes the frequency response analysis that is
the basis of all MSC.Nastran dynamic aeroelastic analysis and then discusses the special
topics of transient response analysis and random response analysis.
Aeroelastic frequency response analysis in MSC.Nastran is performed in modal
coordinates and has a basic equation of the form
Eq. 5-72
where all terms on the left-hand side are identical to those of Eq. 5-70 and are defined with
that equation. The right-hand side provides the loading in modal coordinates, which can
be aerodynamic or nonaerodynamic in nature and is a function of the analysis frequency.
Nonaerodynamic generalized loads, designated , are obtained in the standard
fashion from the loadings applied to physical coordinates.
The solution of Eq. 5-72 entails solving for the generalized displacements by
decomposition/forward-backward substitution techniques applied to the coupled set of
complex equations. Because modal reduction techniques have been applied, the solution
costs are typically modest. Once the generalized displacements have been computed,
standard data recovery techniques can be used to determine physical displacements,
velocities, stress, etc.
Aeroelastic Transient Response Analysis
As discussed in the introduction to this section, Aeroelastic Transient Analysis relies on
Fourier transform techniques. Transient analysis by a Fourier transformation is separated
into three phases. First, the loads (defined as a function of time) are transformed into the
frequency domain. Second, the responses are computed in the frequency domain using the
algorithm of the preceding subsection. Third, these responses (in the frequency domain)
are transformed back to the time domain.
Transformation of Loads
The user specifies loads in the same manner as given in Modal Versus Direct Transient
Response on page 162 of the MSC.Nastran Basic Dynamics Users Guide. The two general
forms are the tabular, piecewise linear function and the general purpose function.
These loads, which are in the form required for frequency response, are transformed to the
modal coordinates exactly as in the modal frequency response method.
M
hh

2
iB
hh
1 ig + ( )K
hh
1
2
--- V
2
Q
hh
m, k ( ) + + u
h
{ } P ( ) { } =
PHF ( )

Inverse Transformation of the Response
The response is found from a numerical approximation to the inverse Fourier integral or
from the Fourier series, the Fourier series result, which can be regarded as a special form
of approximation to the integral. The quantity is first calculated at a set of frequencies,
, by the frequency response analysis. The do not need to be equally spaced and the
integral is evaluated only over the frequency range for which the frequency response has
been performed. This option is described in Fourier Transform on page 176.
Random Response Analysis
The major loads to which an aerospace vehicle is subjected can be predicted for the most
part from its design mission and maneuvering requirements. However, the total
environment cannot be predicted exactly and statistical methods based on the theory of
random processes must be employed to complete the description. Examples of random
processes in aeroelasticity include response to atmospheric gusts and to aerodynamic
buffeting. The random process theory considered in MSC.Nastran is based on generalized
harmonic analysis, i.e., frequency response techniques, and assumes that the system is
linear and that both the excitation and response are stationary with respect to time. See
Random Analysis with Coupled Excitations on page 238 for a description of the
methods.
u
( )
i

i
2 CHAPTER 5
5.9 Dynamic Optimization and Design Sensitivity
Design optimization is a major area of innovation in the development of MSC.Nastran.
The first version of the code, which was released in 1989, was the first to include a
comprehensive structural optimization capability. Later versions provided the ability to
automatically design a structure while considering limits on the structures static, modal
and buckling responses. This capability has been extended to superelement optimization,
aeroelastic optimization, shape optimization, and dynamic optimization, the subject of this
section. All of these capabilities can be made to couple with the same design task. A
mixture of design variables (i.e., shape and sizing) can be used while constraining or
minimizing responses from a number of analysis disciplines, such as statics, normal
modes, flutter and modal frequency response.
The specific dynamic analyses that are supported are direct frequency response (i.e., the
analysis takes place in physical coordinates), modal frequency response and modal
transient response (i.e., the analysis model is reduced using modal approximations). The
basis for performing the dynamic optimization is the dynamic sensitivity analysis
described in the MSC.Nastran Design Sensitivity and Optimization Users Guide. One
implementation consideration that is discussed in this reference, but that should be
reiterated here, is that eigenvector sensitivities are not computed as part of the sensitivity
analysis. Instead, the response sensitivities are computed directly from existing mode
shapes. As a rule of thumb, it is recommended that sensitivity and optimization tasks
retain at least twice as many modes as are felt to be adequate for the analysis.
User Interface
This section briefly discusses how the user invokes MSC.Nastrans design optimization
capability, primarily to document the types of responses that are available for the dynamic
process. Again, the reader is referred to the MSC.Nastran Design Sensitivity and
Optimization Users Guide for complete details on the use of the capability.
Case Control Commands
Subcases are used for multiple excitations and environments.
The objective function is a constraint set selected by the user.
The user selects the active design constraint set for a particular subcase.
Bulk Data Inputs
Example Response Definition. The DRESP1 Bulk Data entry defines a set of structural
responses that is used in the design either as constraints or as an objective. As an example,
the full Bulk Data description is included below.

Format:
Example:
The DRESP1 entry shown above is used to identify the responses that are to be used in the
design task. Responses are used for constraints on the design or as objective functions
either directly or through the use of user-supplied equations. The RTYPE field on this
entry selects the particular response type involved. There are ten optional RTYPEs
available with dynamic analysis:
1 2 3 4 5 6 7 8 9 10
DRESP1 ID LABEL RTYPE PTYPE REGION ATTA ATTB ATT1
ATT2 -etc.-
DRESP1 1 DX1 STRESS PROD 2 3 15 102
103
Field Contents
ID Unique entry identifier. (Integer > 0)
LABEL User-defined label. (Character)
RTYPE Response type. See table below. (Character)
PTYPE Element flag (PTYPE = ELEM) or property entry name. Used with
element type responses (stress, strain, force, etc.) to identify the
property type, since property entry IDs are not unique across property
types. (Character: ELEM, PBAR, PSHELL, etc.)
REGION Region identifier for constraint screening. (Integer > 0)
ATTA, ATTB,
ATTi
Response attributes. (Integer > 0 or Real or blank)
[FRDISP] Displacement in a frequency response analysis.
[FRVELO] Velocity in a frequency response analysis.
[FRACCL] Acceleration in a frequency response analysis.
[FRSTRE] Stress in a frequency response analysis.
[FRFORC] Force in a frequency response analysis.
[TDISP] Displacement in a transient response analysis.
[TVELO] Velocity in a transient response analysis.
[TACCL] Acceleration in a transient response analysis.
[TSTRE] Stress in a transient response analysis.
[TFORC] Force in a transient response analysis.
2 CHAPTER 5
A response may be associated with a specific time step, frequency, or subcaseor it may
be defined for all available solutions. In the latter case, an individual constraint is
generated for every available time step, etc. The specification of location of the responses
(e.g., grid ID or stress component) is performed using the ATTi fields on the entry. Note
that no distinction is made between modal and direct frequency analysis. Instead, the user
specifies the type of analysis by using Case Control requests.
The DRESP2, DEQATN, DCONSTR and DESOBJ Entries
Three additional Bulk Data entries are used for defining the design objective and
constraints:
1. The DRESP2 entry, coupled with the DEQATN entry, can be used to synthesize a
response based on DRESP1 responses and a user supplied equation. An example
of the use of this feature is to specify a response that is the difference of two
FRDISP responses. A synthesized response will be calculated for all time steps or
frequencies defined on the DRESP1 data.
2. The DCONSTR entry is used to place lower and upper bounds on the DRESP1
and/or DRESP2 responses.
3. The DESOBJ entry identifies the objective function that is to be minimized by
referring to a DRESP1 or DRESP2 entry. The objective must be a scalar quantity
such as mass, error, or cost.
On the other hand, a single DRESPi entry may generate many responses at each frequency
or time step the user specifies. Therefore, if the user invokes one of the dynamic response
quantities as the objective, it is necessary to identify which of the frequencies or time steps
is to be used for the objective.
Example: A Clamped-Free Plate
This example is a plate that is acted upon by a harmonically varying pressure loading
(Figure 5-6). The structural model is the same as the one used in Reference 3 to validate
the sensitivity analysis in the presence of dynamic loads. The plate is rectangular with a
length of 20.0 inches and a width of 10.0 inches. It is clamped at three edges and free on
the fourth. Since the pressure loading is uniform, only a one-half model with a symmetry
boundary condition is required. The material has an elastic modulus of and a
density of . The uniform pressure load has a magnitude of and is
applied at 500 Hz.
1.0 10
7
psi
0.1 lbm/in
3
1.0 lbf/in
2

Finite Element Model
Figure 5-6 The Clamped-Free Plate
Figure 5-7 Final Thickness Distributions for the Clamped-Free Plate as a Function of
the Imposed Frequency Constraints
The half model of the plate is modeled by a array of four-noded plate elements. The
elements are linked so that a single design variable controls a band of elements. That is, all
19 elements that are situated along the edge of the plate are controlled by the first design
variable, the 17 elements that are one element in from the edge are controlled by the second
Free Edge
Symmetric
Boundary Conditions
Grid 1110
y
x
t
1
x
1
=
10 10
2 CHAPTER 5
design variable and, finally, the single element at the center is controlled by the tenth
design variable. The magnitude of the displacement at plate center is constrained to be less
than 2.5 inches and the plate thickness is constrained to be greater than 0.01 inches.
Six different cases were run with the only difference being a constraint on the relation
between the natural frequencies and the excitation frequencies. Each test case was a
separate MSC.Nastran execution, which resulted in six different optimal designs. In the
first case, the first natural frequency was constrained to be greater than the 500.0 Hz
excitation frequency. For the remaining five cases:
Eq. 5-73
This frequency constraint was imposed to remove the disjoint design space phenomenon
that occurs when a local minima exists near the original modal frequencies. In other words,
we are forcing a major change in the modal frequency.
Results
Figure 5-7 shows the thickness distributions along either edge of the plate for the six cases.
In this example, only in the first test, where the first natural frequency is greater than the
excitation frequency, is the result intuitive. Table 5-2 shows the final weights of the six
designs. It can be seen that the weight is monotonically decreasing as the structure is made
more flexible (i.e., more modes are below the fixed excitation frequency).
Example: A Portal Frame with Transient Loading
The second example demonstrates the application of MSC.Nastrans optimization
capabilities to transient analyses. The example is a simple portal frame acted upon by an
enforced displacement at its base (Figure 5-8). This is a simple representation of a civil
engineering structure subjected to seismic loading. The problem was obtained from
Reference 9, which demonstrated the possibility of disjoint design spaces when working
in the time domain in an analogous fashion to the disjoint design space referred to earlier
in the frequency domain. The portal has the following parameters:
Table 5-2 Final Weights for the Clamped-Free Plate
Tes
t
Constraint Weight
1 5.163
2 1.963
3 1.831
4 1.310
5 0.740
6 0.613
i
500.0 Hz
i 1 +
i < < 1, 5 =
500.0 Hz.
1
<
1
500.0 Hz.
2
< <
2
500.0 Hz.
3
< <
3
500.0 Hz.
4
< <
4
500.0 Hz.
5
< <
5
500.0 Hz.
6
< <

Figure 5-8 The Portal Frame
The frame is subjected to horizontal ground motion in the form of a half sine pulse:
Eq. 5-74
The design task is to minimize the weight of the structure while imposing the limits on
structural displacement, stress and natural frequency listed in Table 5-3.
Finite Element Model
The frame was modeled using three finite element bars (the symmetry of the problem
could have been used to limit this to two elements). The stress is computed as the ratio of
the bending moment in the beam divided by the section modulus.
Table 5-3 Parameters for the Portal Frame Example
PARA
M
Meaning Value
E Young Modulus
psi
Material density
Minimum allowable inertia
NSM Nonstructural mass per unit length 10.0 lbm/in
Minimum allowable fundamental frequency 60 rad/sec
Maximum allowable displacement in
Maximum allowable stress ksi
L Height and width of the portal 180 in
30.0 10
6
0.28 lbm/in
3
I
mi n
290.0 in
4
mi n
max
3.0
max
3.0
15
15
1
1 2
2 3
4
u
g
t ( ) 1.0 30t ( ), t sin 30.0 =
2 CHAPTER 5
The weight of the beam is simply its volume times the structural density. One design
variable controls the bending moment of inertia of the vertical columns, while a second
variable controls the inertia of the horizontal beam. Equations for the section modulus, S,
and area, A, in terms of the moment of inertia of the beam, I, are taken from Reference 9 in
Selected Bibliography on page 299:
Eq. 5-75
Eq. 5-76
The transient response was performed over a period of 250 msecs at a time step of 1 msec.
Responses were retained at every 5 msecs and the stress and displacement constraints were
applied at each of the retained time steps. For the initial design, moment of inertia values
of were used for both design variables, making the initial objective value 1722.2 lbs
(this ignores the weight of the nonstructural mass).
Results
Figure 5-9 Transient Stress Response for the Initial and Final Designs
of the Portal Frame
S 60.6I 84000 + 290 =
A 0.465 I =
600 in
4

Figure 5-9 shows the transient stress response of the end of the horizontal beam for the
initial and final designs. It can be observed that the stresses for the initial design are well
below the 30 ksi limits, while the response is right at this limit for the final design. Since
there is no damping in the structural representation, the structure would continue to sway
if a longer record were obtained, but the imposed limits should not be exceeded.
Table 5-4 compares the results from this analysis with those given in Reference 9 in
Selected Bibliography on page 299. Note that the MSC.Nastran results are somewhat
higher than those given in the reference, which can be attributed to differences in the finite
elements used in the two analyses.
Although the examples shown here are elementary, it should be apparent that the
capabilities they demonstrate are quite powerful. MSC.Nastrans capabilities for dynamic
response optimization can be coupled with design tasks that have requirements for static
response, natural frequency, and buckling and aeroelastic responses. Note, too, that the
models can include superelements. Furthermore, although all the examples here have
properties for the design variables, the capability also interacts with MSC.Nastrans shape
optimization capability.
One additional area in which dynamic response optimization can be applied is the system
identification of structures to match modal test results. A modal test can be considered a
forced excitation. Therefore, the capability described here could be used modify structural
parameters in an MSC.Nastran analysis so that the responses obtained from an
MSC.Nastran frequency analysis could match test results.
Table 5-4 Comparison of Results for the Portal Frame
PARAM MSC.Nastran Reference
X1 (columns)
X2 (crossbeam)
Objective 1534. lbs 1463. lbs
578.6 in
4
498.0 in
4
301.7 in
4
330.0 in
4
CHAPTER
6
Data Recovery
I Overview
I Data Recovery for Modal Methods
I Improving the Results
I Shock and Response Spectrum Analysis
I Random Analysis with Coupled Excitations

6.1 Overview
In dynamic analysis, the size of the calculated output data could become several orders of
magnitude larger than the size of output for the equivalent statics analysis job. For
instance, a typical transient run could produce thousands of solution vectors. If caution is
not taken, the cost of processing the outputs for printing or plotting could exceed that of
the entire solution process. Given these concerns, the dynamics engineer should use the
SET logic in the Case Control Section of the MSC.Nastran input file to reduce the number
of outputs. Wherever possible, he should also use a graphical postprocessor to visualize
the results.
The choice of modal methods versus the direct method for dynamic solutions is discussed
in the MSC.Nastran Basic Dynamics Users Guide. Each method uses a different technique to
recover output answers. This chapter extends this discussion by describing several options
available in the modal method to improve accuracy while reducing computation costs.
This chapter describes two methods for calculating physical results from a solution using
modal degrees-of-freedom. The default method is the so-called matrix method, which is
the most effective technique for the majority of dynamics problems. The alternative
method uses the static approach to calculate results one vector at a time.
The mode acceleration method, also detailed in this chapter, is useful in some cases where
accurate stress results are required for only a few peak time steps or frequencies. However,
this option may be more expensive if many solution vectors exist.
In addition to the standard displacement, stress, and force outputs, dynamics engineers
frequently use other types of output that combine the effects of random or spectral inputs.
Shock and response spectral analysis provides a quick calculation of the probable peak
values in a modal solution for base accelerations. Random analysis is a more formal
method of determining power spectral densities, RMS averages, and other data for a
structure in a random load environment. This chapter devotes a section to each of these
topics.
2 CHAPTER 6
Data Recovery
6.2 Data Recovery for Modal Methods
Modal methods of analysis have the advantages of reducing the problem size, decoupling
the equations, and providing an intuitive feeling for the results. However, they require
additional calculation steps and may also result in a loss in accuracy. There are four
different methods for calculating the detailed results in the data recovery stage of analysis
using modal coordinates. Each of these options provide a unique tradeoff of efficiency vs.
accuracy for a given size of problem, number of solutions, and number of output requests.
Described below are the normal mode displacement method, the matrix method, the mode
acceleration method, and the use of superelements for improved modal results.
Mode Displacement, Vector Method
This method is almost identical to the statics method of data recovery in MSC.Nastran. It
is efficient when only a few time steps or frequencies are present in the solution and many
outputs are requested (such as for contour plotting). Starting with the modal
displacements, , and the eigenvectors, , we obtain a set of displacement vectors that are
expanded as with statics or normal mode analysis. The steps are summarized below.
Analysis Set Eq. 6-1
Free Set Eq. 6-2
SPC Forces Eq. 6-3
Independents Eq. 6-4
Grid Set Eq. 6-5
In the actual calculations, each vector at a particular frequency or time step becomes a
column of the solution matrix . Note that for all modal methods, no provisions are
made for thermal loads or enforced displacements (which are only calculated in statics
analysis.)
The element stresses and forces are calculated directly from the displacement vectors, one
element and one vector at a time. Note that if a large number of solutions exist (for instance
10000 time steps), the size of the matrices defined above may be quite large and the
calculation speed will be very slow.
[ ]
u
a
{ }
a
[ ] { } =
u
f
{ }
G
0
I
u
a
{ } =
Q
s
{ } K
sf
[ ] u
f
{ } P
s
{ } =
u
n
{ }
u
f
0 " #
$ %
& '
=
u
g
{ }
G
m
I
u
n
{ } =
u
g
[ ]

Matrix Method for Modal Response
This method is very efficient when the number of solutions is much larger than the number
of modes and when a limited number of output requests is made. The matrix method for
modal response is the default for the modern modal solutions. In this method, the same
operations as described above are performed on the eigenvectors and the results are saved.
A matrix is calculated for every output request using the eigenvector matrix instead of a
solution vector. The intermediate matrices are calculated from the output data recovery of
the modal solution. These are summarized below.
Displacements Eq. 6-6
SPC Forces Eq. 6-7
Stresses Eq. 6-8
Forces Eq. 6-9
Note that the number of rows of these matrices will correspond only to the requested
output sets (DISP =, STRESS =, etc.). The number of columns is equal to the number of
modes (which is usually smaller than the number of time steps of frequencies). For the
final output printing or xy-plotting, the equations for calculating the transient solution
outputs are as follows:
Displacements Eq. 6-10
Forces of SPC Eq. 6-11
Similar equations are used for SPC forces, stresses, and element forces, as well as for
frequency response, by simply combining the modal responses with the modal factors as
indicated above.
u
g
m
[ ]
g
[ ] =
Q
s
m
[ ] K
sf
[ ]
f
[ ] =
S
m
[ ] ( ) [ ] =
F
m
[ ] f ( ) [ ] =
u
g
o
t ( ) [ ] u
g
m
[ ] t ( ) [ ] =
Q
s
t ( ) [ ] Q
s
m
[ ] t ( ) [ ] =
2 CHAPTER 6
Data Recovery
6.3 Improving the Results
Mode Acceleration Method
The mode acceleration method of data recovery is used to reduce the modal truncation
errors that occur in the methods above. For good accuracy, a general recommendation is
to use many more modes than in the frequency range of the excitation loads. For most
applications this is adequate and the mode displacement method produces acceptable
results. However, some cases that may not produce accurate results are the following:
1. Errors may occur in the deformations and stresses in the area of a concentrated
load on a free point.
2. Concentrated damper elements or direct input matrices may change the
characteristic response displacements to be much different from simple modal
combinations.
3. Models with complex shapes may require an excessive number of modes to
represent the stress distributions. Accuracy of the modes becomes an issue when
more than a few hundred are necessary.
4. Massless points with external loads will have deformations and internal loads
that cannot be represented by finite frequency modes.
Using the direct method may eliminate these errors but may excessively increase the run
costs and file storage. The mode acceleration method is a useful compromise that
eliminates the errors described above yet is cheaper to run than the direct method.
However, it will cost more than the matrix method if a large number of solution vectors are
to be processed. The derivation is given below.
For convenience, we will show the equations for a frequency response solution. However,
the results are similar with a transient analysis. The ideal solution desired is that for the
direct matrix solution at frequency , which is
Eq. 6-12
where is the imaginary unit value.
We may obtain an exact solution if we use all of the modes in the system; namely, if N is
the order of the problem, then:
Eq. 6-13
Note that this set of modes also includes modes with infinite frequencies. These are present
with singular mass matrices. The modal identities are
Eq. 6-14
and for constant structural damping parameters, and
2
M j B K + + [ ] u ( ) { } P ( ) { } =
j
u { }
i
{ }
i
i 1 =
N
)
=
i
2
M [ ]
i
{ } K [ ]
i
{ } =
g
3

Eq. 6-15
Eigenvalue of Modal and Direct Solutions. Substituting Eq. 6-13, Eq. 6-14, and Eq. 6-15
into the terms defined in Eq. 6-12, we obtain the effect of each mode on the exact solution.
These terms are
Eq. 6-16
Eq. 6-17
and
Eq. 6-18
If all modes were included, we know that the results would be identical to a direct solution.
For the high frequency modes ( ), it is apparent that their contribution will be much
smaller for mass and damping effects than from their effect on the static solution of
Eq. 6-18. Therefore, if we truncate the modes such that

the primary error will be due to the
forces generated by the missing modes in Eq. 6-18. In the frequency domain the
approximations to Eq. 6-16 through Eq. 6-18 become
Eq. 6-19
Eq. 6-20
Eq. 6-21
The error can be expressed as a residual displacement vector .
Substituting Eq. 6-19 and Eq. 6-20 into Eq. 6-15, we obtain a corrected solution:
Eq. 6-22
where:
=
B [ ]
i
{ }
g
3
------
* +
, -
K [ ]
i
{ } =
2
M [ ] u { }

i
-----
* +
, -
2
K [ ]
i
{ }
i
i 1 =
N
)
=
j B [ ] u { }
j g
3
---------
* +
, -
K [ ]
i
{ }
i
i 1 =
N
)
=
K [ ] u { } K [ ]
i
{ }
i
i 1 =
N
)
=
i

m
>
2
M [ ] u { }
2
M [ ]
i
{ }
i
i 1 =
M
)

j B [ ] u { } j
i 1 =
M
)
B [ ]
i
{ }
i
K [ ] u { } K [ ]
i
{ }
i
K [ ] u { } +
i 1 =
M
)
=
u ( ) { }
K [ ] u ( ) { } P ( ) { } R ( ) { } =
R ( ) { }
j
i
{ }
i

2
i
{ }
i
i 1 =
M
)
i 1 =
M
)
2 CHAPTER 6
Data Recovery
An alternate method would be to solve the static part of Eq. 6-15 using Eq. 6-21. Since the
residual vector is orthogonal to the retained modes, then:
Eq. 6-23
We can assume that contains static response, if is nonsingular, plus modal
response in the form:
Eq. 6-24
Premultiplying Eq. 6-24 by and using Eq. 6-23, we obtain
Eq. 6-25
and, then from Eq. 6-24 and Eq. 6-25, we obtain a decoupled residual solution
Eq. 6-26
where:
The matrix, , is known as the residual flexibility matrix and has also been used for
modal synthesis modeling. Here it could be used as a data recovery step.
In the MSC.Nastran design, Eq. 6-22 has been chosen over Eq. 6-26 for the calculation of
the improved solution of large problems because the matrix could be very dense and
the singular free body case was easier to process.
Note that in the actual mode acceleration process other nonstructural effects, such as direct
input matrices or transfer functions, are not included in the matrix, , but are treated as
terms on the right-hand side and added to . For free bodies, the right hand loads are
converted to equilibrium loads to permit a decomposition of the singular stiffness matrix,
identically to the inertia relief solution. In fact, the entire process may be viewed as if all
the dynamic modal solutions were converted into equivalent static loads, and linear static
solutions were generated using the symmetric structural stiffness matrix.
Using Superelements for Data Recovery
If a structure has only a few points with dynamic loads or enforced boundary motion, a
simple superelement (SE) operation will have an improvement in accuracy similar to the
mode acceleration method. In transient analysis, the cost will be increased because the
uncoupled modal solution step is changed to coupled solution with a few extra
degrees-of-freedom. In frequency response analysis, the extra cost is very small. The basic
procedure is as follows:
=
[ ]
T
K [ ] u { } 0 { } =
u { } K [ ]
u { } K [ ]
1
P { } [ ] x { } + =
Q
T
K [ ]
x
i
m
i
i
2
( )
1
i
[ ]
T
P { } =
u ( ) { } Z [ ] P ( ) { } =
Z [ ] K [ ]
1
[ ]
.
m
i
i
2
( )
.
[ ]
T
Z [ ]
Z [ ]
K [ ]
P { }

1. Subdivide the structure into two superelements corresponding to the external
loads.
2. The residual (SE = 0) contains only the loaded structural points and scalar points
for the modal coordinates.
3. The upstream (SE = N) contains all of the remaining points and all of the
elements. It will also contain the residual SE points as boundary points.
4. Place a modal synthesis request for the upstream SE with a METHOD = request
in the corresponding Case Control subcase.
5. By default, the mode shapes will be fixed at the residual points. Free boundary
degrees-of-freedom may be specified on SECSET data.
6. The residual SE solution will contain both modal and grid point
degrees-of-freedom. If the resulting size is small, a direct solution for the forced
response is recommended.
7. The upstream solution will contain both modal response and static residuals
(described above) due to motion of the residual grid points. Stresses and forces
will be improved, especially on the interface elements.
In this method the static correction terms are assumed to be shapes corresponding to loads
on the boundary points. The upstream displacements, , are calculated from the
equation
Eq. 6-27
where:
The residual SE will obtain a dynamic solution for both and .
= the residual SE displacements
= the Guyan reduction matrix
= the partition of the eigenvectors
= the partition of the eigenvectors
= the modal degrees-of-freedom
u
o
{ }
u
o
{ } G
oa
[ ] u
a
{ }
o
[ ] G
oa
[ ]
a
[ ] [ ] q { } + =
u
a
{ }
G
oa
[ ]
o
[ ]
u
o
a
[ ]
u
a
q { }
u
a
q { }
2 CHAPTER 6
Data Recovery
6.4 Shock and Response Spectrum Analysis
Shock spectra analysis and response spectrum analysis are methods used by many
engineers to estimate the maximum dynamic response of a structure. Most applications
involve complicated time-dependent loads or accelerations that excite the base of a
structure, such as an earthquake ground motion on a building or an explosive shock on a
small component in a ship. (Note that the only difference between shock and response
spectra is whether output displacements are measured in a fixed frame of reference or
relative to the base motion.)
The advantage of these methods over a conventional transient analysis is economy and
simplicity. The only major calculation step is obtaining a sufficient number of normal
modes to represent the entire frequency range of the input excitation and resulting
response. The disadvantage of the method is that the accuracy may be questionable and
the requirement of special input data in the MSC.Nastran solution sequences. In many
cases, a direct transient analysis with the actual excitation load will be more accurate, easier
to use, and faster.
The procedure involves two stages. First the applied loads or base excitations are
converted in a direct transient solution (SOL 109) into a spectrum table consisting of peak
response magnitudes for a set of single degree-of-freedom oscillators. Each oscillator is a
scalar spring/mass/damper having a different natural frequency and damping ratio. This
stage is optional since the shock spectrum data is frequently given in the contractual design
specifications or, in the case of earthquakes, is available through governmental agencies.
The second stage of the analysis consists of a modal analysis of the structure, data recovery,
and the response calculation that combines the modal properties of the analysis model
with the spectrum data of the applied loads. This stage is performed in a modal analysis
solution sequence (SOL 103). If a database was saved from the first stage, a restart will
provide the spectrum data automatically. Otherwise, the spectrum data must be supplied
in a direct tabular input (response versus natural frequency for several damping ratios).
Theoretical Background
Starting with a modal transient analysis, the general approximation for a response
quantity, , is
Eq. 6-28
where and are the modal outputs and generalized displacements. The actual modal
equations are
Eq. 6-29
where is the vector of loading functions. For loading due to base accelerations, the
equivalent inertial loads are
Eq. 6-30
u
k
u
k
t ( )
i k
i
t ( )
i
)
=

i
g
i
i

i
2
i
+ +
i
T
P t ( ) { } =
P
P t ( ) { } M
aa
[ ] D
ar
[ ] u
r
t ( ) { } =

where the columns of represent vectors of rigid body motions of the whole structure
and the accelerations correspond to the base motions. Substituting Eq. 6-30 into Eq. 6-29
and combining terms we can separate the modal quantities from the transient solutions.
First we will develop the transient response functions. We begin by calculating the
responses
Eq. 6-31
where is a response function in direction , and is a function of the variables , , and
. The peak values of , obtained over a range of frequencies and damping factors is called
the response spectrum for the excitation, .
Next, from the normal mode analysis, we define the participation factors , for mode
and direction , as
Eq. 6-32
Then, from Eq. 6-28, the actual transient response at a physical point is
Eq. 6-33
The peak magnitudes of in Eq. 6-33 are usually dominated by the peak values of
occurring at the natural frequencies. In spectrum analysis the peak values of

are
approximated by combining functions of the peak values, , in the
approximation
Eq. 6-34
ABS Option. Eq. 6-33 and Eq. 6-34 define the ABS (Absolute Value) option. This method
assumes the worst case scenario in which all of the modal peak values for every point on
the structure are assumed to occur at the same time and in the same phase. Clearly in the
case of a sudden impact, this is not very probable because only a few cycles of each mode
will occur. However, in the case of a long term vibration, such as an earthquake when the
peaks occur many times and the phase differences are arbitrary, this method is acceptable.
A second way of viewing the problem is to assume that the modal magnitudes and phases
will combine in a probabalistic fashion. If the input loads are behaving randomly, the
probable (RMS) peak values are
Eq. 6-35
where the average peak modal magnitude, is
Eq. 6-36
D
ar
[ ]
x
r
gx
r

2
x
r
+ + u
r
t ( ) =
x
r
r g
t x
r
u
ir
i
r
i r

i
T
M
aa
[ ] D
ar
{ } =
u
k
t ( )
i k
i r
x
r

i
, g
i
, t ( )
r
)
i
)
=
u
k
x t ( )
u
k
x
ri

i
,g
i
( ) max x
ri

i
,g
i
,t ( ) =
u
k
t ( )
i k

i r
x
ri

i
, g
i
( )
r
)
i
)
u
k

i k
i
( )
2
i
)
i

i r
x
r

i
, g
i
i
( ) ( )
2
r
)
=
2 CHAPTER 6
Data Recovery
SRSS Method. This approach is known as the SRSS (square root of sum-squared)
method. Note that the results in each direction are summed in vector fashion for each
mode first, followed by an SRSS calculation for all modes at each selected output quantity
. It is assumed that the modal responses are uncorrelated and the peak value for each
mode will occur at a different time. These results are optimistic and represent a lower
bound on the dynamic peak values.
The SRSS method may underestimate the actual peaks since the result is actually a
probable peak value for the period of time used in the spectrum analysis. The method is
normally augmented with a safety factor of 1.5 to 2.0 on the critical outputs.
NRL Method. As a compromise between the two methods above, the NRL (Naval
Research Laboratories) method was developed. Here, the peak response is calculated from
the equation
Eq. 6-37
where the j-th mode is the mode that produces the largest magnitude in the product .
The peak modal magnitudes, , are calculated with Eq. 6-36.
The rationale for the method is that the peak response will be dominated by one mode and
the SRSS average for the remaining modes could be added directly. The results will fall
somewhere between the ABS and SRSS methods.
Modes that are close in frequency may have their peak response occur at about the same
time (and with the same phase). For this reason, the SRSS and NRL methods contain a
provision to sum modal responses via the ABS method for modes that have closely spaced
natural frequencies. Close natural frequencies are defined by frequencies that meet the
following inequality:
The value for CLOSE is set by PARAM,CLOSE (the default is 1.0).
The modal summation option is set via PARAM,OPTION (the ABS method is the default).
Both PARAM,OPTION and PARAM,CLOSE may be set in any subcase, allowing
summation by several conventions in a single run.
Generating Response Spectra Curves
A response spectrum is generated as follows:
1. Assume that there is a series of small, single degree-of-freedom (SDOF) oscillators
each attached to the same location at the connection to the base structure. (In the
examples in the introduction to this section, the base structures are the building
and the spacecraft). These oscillators each have a different resonant frequency,
and all have the same modal damping ratio (2%, for example).
u
k
u
k

j k
j

i k
i
( )
2
i j
)
+
jk
j k
j
f
i 1 +
CLOSE f
i
<

2. Apply a transient excitation to the base structure. Use the base structures
damping when computing the base structures transient response at the location
of the SDOF oscillators.
3. Use the base structure transient response as input to each SDOF oscillator.
Compute the magnitude of peak response of each SDOF oscillator, and plot the
peak response versus oscillator resonant frequency.
4. Change the modal damping ratio for the oscillators (to 5%, for example) and
repeat Steps 2 and 3 for the range of expected damping.
The response spectrum, therefore, depicts the maximum response magnitude of an SDOF
system as a function of resonant frequency and damping. Figure 6-1 depicts the generation
of a spectrum.
Figure 6-1 Response Spectrum Generation
Note that the peak response for one oscillator does not necessarily occur at the same time
as the peak response for another oscillator. Note, too, that there is no phase information
since only the magnitude of peak response is computed.
Several values of peak response are computed for the oscillators. These include the
following response variable.
(a)
(b)
(c)
Peak
Response
Resonant Frequency
Transient excitation (a) is applied to a base structure (b), from which transient
response (c) is computed for each floor. This response is applied to a series of
oscillators (d), for which the peak response is plotted (e). Steps (d) and (e) are
repeated for different damping values to form response spectra as shown below.
Resonant Frequency (f)
(d)
= 5% critical damping
Peak
Response
(e)
Series of Oscillators
Transient
Response
Base
Structure
Transient
Excitation
f
1
f
2
f
3
. . . f
max
f
1
f
2
f
3
. . . f
max
f
n
2 CHAPTER 6
Data Recovery
Relative velocity and absolute acceleration are approximately related to the relative
displacement by
Eq. 6-38
Design spectra are most often defined in terms of , , and .
Note that for very low oscillator frequencies ( ),
Eq. 6-39
where is the motion of the base of the oscillator.
Similarly, for very high oscillator frequencies ( ),
Eq. 6-40
The approximate relationships between , , and , shown in Eq. 6-38, are not valid at
very low or very high oscillator frequencies or for large values of damping.
It is assumed in this process that the mass of each oscillator is very small relative to the base
structure, so the oscillators response does not influence the response of the base structure.
Dynamic Response Predictions. Once a spectrum is computed, it can be used for the
dynamic response analysis of an MSC.Nastran model of the component. For example, the
spectrum generated for a floor in a building that is subjected to an earthquake can then be
applied to a complex model of a piece of equipment attached to that floor. The peak
response of each mode of the equipment model is obtained from the spectrum, and these
peak responses are combined to create the overall response.
Because response spectrum generation involves transient response, similar modeling and
analysis considerations apply. In addition, the time step (field 4 on the TSTEP Bulk Data
entry, DT) should not be changed during the run, because MSC.Nastran uses only the
initial DT specification for the entire response spectrum generation run and therefore
wrong answers could occur.
The time step, DT, and time duration, (where is the number of time increments),
must take into account the loading, the base structure, and the frequency range of the
spectra generation. The time step must take into account the frequency content of the
applied excitation, the frequencies of the base structure, and the highest frequency for
which spectra are to be generated. There must be enough time steps per cycle of response
for both the base structure and the highest frequency oscillator in order to accurately
predict the peak response; 5 to 10 steps per cycle represents a typical value. In addition,
the time duration of the loading, the frequencies of the base structure, and the lowest
oscillator frequency must be considered when defining the time duration. There must be
X
r
X
r
2
X
r
X
r
X
r
X
0
X 0
X
r
U
b
U
b

X U
b
X
r
0
X
r
X
r
X
DT N N

a long enough time duration of response both for the base structure and the lowest
frequency oscillator in order to accurately predict the peak response. For short duration
loadings, the peak response often occurs well after the the load has peaked.
Initial conditions (specified via the TIC Bulk Data entry) are not used in response spectrum
generation. Initial conditions are used in the calculation of the transient response of the
base structure, but the calculation of the peak oscillator responses (i.e., the response
spectrum calculation) ignores any initial conditions.
User Interface for Response Spectra Generation
Response spectra are generated in the transient response solution sequences (SOL 109 for
direct and SOL 112 for modal). Transient response input is required to apply the transient
excitation to the base structure. Additional input required to generate response spectra are
described in Table 6-1 and Table 6-2.
The XYPLOT and XYPUNCH commands are included in the OUTPUT(XYPLOT) section.
Further details about the OUTPUT(XYPLOT) section are described in Structure Plotting
on page 243 of the MSC.Nastran Basic Dynamics Users Guide.
Examples of these commands include:
$ Plot absolute acceleration spectra for grid point 85, T3 component
XYPLOT ACCE SPECTRAL 1 /85(T3RM)
$ Punch relative displacement spectra for grid point 3, T1 component
XYPUNCH DISP SPECTRAL 1 /3(T1IP)
Relative and absolute spectra are denoted by IP and RM, respectively, in the parentheses
of the curve request.
Table 6-1 Case Control Input for Response Spectrum Generation
Case Control Command Description
XYPLOT SPECTRAL Compute spectra.
XYPUNCH SPECTRAL Punch spectra for subsequent use.
2 CHAPTER 6
Data Recovery
There are two FREQi entries: one to specify oscillator frequencies (i.e., frequencies for
which spectra will be computed) and the other to specify oscillator damping. (Note that
damping for the base structure is specified in another manner, such as with the TABDMP1
entry used for modal transient response analysis.)
Peak Response Calculation
Response spectra are applied to a structure for subsequent analysis. Note that the structure
in this case is not the base structure for which the spectra were computed but, rather, a
smaller structure such as a piece of equipment.
Response spectrum analysis approximates the peak structural response (typically stresses
and displacements, though not limited to those quantities). Approximations are made by
assuming:
1. Only the peak response is computed for the oscillators (see the preceding section).
2. There is no phase information or sign computed; only the magnitude of peak
response is computed.
3. The oscillator mass is small relative to the base structures mass.
4. The displacements, velocities, and accelerations are related by the approximate
relationships in Eq. 6-38.
5. The peak modal responses are combined to form the overall response via various
combination methods (see below).
The spectra themselves are often approximate. For example, design spectra have been
developed for seismic analysis, and these have conservatisms built into them by the fact
that safety factors are incorporated by either increasing the spectra values and/or by
decreasing the damping values.
These approximations make response spectrum analysis a tool that is useful for design and
efficiency. The effects of these approximations are further described in the remainder of
this section.
Table 6-2 Bulk Data Input for Response Spectrum Generation
Bulk Data Entry Description
PARAM,RSPECTRA,
0
Requests calculation of spectra.
DTI, SPSEL, 0
DTI, SPSEL, 1
Header for DTI.
Selects oscillator frequencies, oscillator
damping values, and grid points at
which spectra will be computed.
FREQi Specifies oscillator damping values.
FREQi Specifies oscillator frequencies.

Modeling and Analysis Considerations
Response spectrum application is a postprocessing function of normal modes analysis. It
is run in the normal modes solution sequences, so the modeling and analysis
considerations that apply for normal modes analysis also apply for response spectrum
application. The additional considerations also need to be followed:
1. The structure is run as an unrestrained model in the direction(s) of the load
spectrum application.
2. A large mass, on the order of to times the mass of the structure, must be
used at the structures base grid points where the input occurs. One way to do
this is to use RBEs or MPCs to connect the base points to a separate grid point, and
apply the large mass to that separate grid point. This separate grid point is where
the spectrum is applied.
3. A SUPORT Bulk Data entry is required at the spectrum input location. See Free
Body Techniques on page 190 for an explanation.
4. The modes must be mass normalized (which is the default).
The spectra that MSC.Nastran can apply are absolute acceleration, relative displacement,
and relative velocity spectra. You specify A, D, or V for acceleration displacement or
velocity, respectively, to specify the spectrum type.
Use all modes within the frequency range specified by the spectrum, but do not use modes
outside of the spectrum range. Usually, spectra to apply are considered to have zero values
outside of their range of definition; for example, an absolute acceleration spectrum defined
from 0 to 30 Hz is assumed to be zero beyond 30 Hz. However, MSC.Nastran extrapolates
spectral values for modes beyond the spectral range, which may lead to unexpected
answers. You can limit the number of modes used in the spectrum application by limiting
the number of computed modes (via the EIGRL or EIGR entry) or by using
PARAM,HFREQ,r (where r is the highest frequency mode to use) or PARAM,LMODES,n
(where n is the number of lowest modes to use).
Consider the entire response spectrum processgeneration and applicationas a
two-step process. Step 1 is generation of the response spectra and Step 2 is the application
of the response spectra. For a given input, transient applied to the base structure (Step 1),
the same stresses occur (Step 2) regardless of whether acceleration or displacement spectra
were computed in Step 1. However, displacements and accelerations are different, because
answers computed by using the absolute acceleration spectrum contain the rigid body
contribution, whereas answers computed by using the relative displacement spectrum do
not contain the rigid body contribution. Displacement and acceleration responses can be
made equal regardless of which spectra was used by using PARAM,LFREQ,0.01 (or some
other small number) to remove the rigid body mode contribution from the answers.
Stresses and other element quantities are unaffected by the contribution of any rigid body
modes. The same situation applies to relative velocity spectra as to relative displacement
spectra. However, because the relationships in Eq. 6-38 are approximate, all answers
(including stresses) will be slightly different depending on whether displacement, velocity,
or acceleration spectra were used.
10
3
10
6
2 CHAPTER 6
Data Recovery
User Interface for Spectrum Application
Response spectrum application is done in the normal modes solution sequences (SOL 103,
for example). In addition to the input for computing normal modes, input is required for
applying the spectra, as shown in Table 6-3.
All input listed in the table is required with the exception of PARAM,OPTION and
PARAM,CLOSE.
Response Spectrum Examples
This section provides an example of response spectra generation and an example of
response spectrum application. The base structure for which response spectra are
computed is a model of a five-story, shear-wall building, excited by an earthquake time
history. The resulting spectra at a point on the fourth floor are applied to a bracket model
in order to compute stresses, displacements, and accelerations.
Table 6-3 Input for Response Spectrum Application
Case Control Command Description
METHOD Selects eigenvalue extraction method.
SDAMP Selects the TABDMP1 Bulk Data entry.
DLOAD Selects the DLOAD Bulk Data entry.
Bulk Data Entry Description
PARAM,SCRSPEC,0 Requests response spectrum application.
EIGR or EIGRL Eigenvalue extraction method.
TABDMP1 Specifies damping for the structure.
DLOAD Defines spectrum multipliers.
DTI,SPECSEL,0 Header for DTI.
DTI,SPECSEL,1 Specifies type of spectrum (A, V, or D) and
selects damping.
A = absolute acceleration spectrum.
V = relative velocity spectrum.
D = relative displacement spectrum.
TABLED1 Specifies input spectrum values.
SUPORT Specifies input spectrum grid points.
CONM2,CMASS2, etc. Defines large mass used for the input spectrum.
PARAM,OPTION,a Specifies modal combination method (a = ABS
[default], SRSS, or NRL).
PARAM,CLOSE,r Specifies closeness parameter for modal
combinations (the default is 1.0).

Response Spectrum Generation
Figure 6-2 depicts the five-story building model. The fifteen grid points at the base of the
building are tied to a separate grid point (999) via RBE2 entries. Earthquake excitation is
applied to this grid point, in the x direction (T1). The absolute acceleration response
spectrum is computed for grid point 502. SOL 112, modal transient response, is used to
compute the spectrum. Damping of 5% critical is used for the building and damping of 2%
critical is used for the generated response spectrum. Listing 6-1 shows a portion of the
input file for this model.
Figure 6-2 Five-Story Building Model
Listing 6-1 Abridged Input File for Response Spectrum Generation
$ RESPONSE SPECTRUM GENERATION
$
$ Compute spectrum for absolute acceleration
$ for grid point 502 for 2% oscillator damping
$
ID MSC, DUG
TIME 15
SOL 112 $ MODAL TRANS. RESP.
CEND
$
TITLE = FIVE-STORY BUILDING, SHEAR MODEL
SUBTITLE = RESPONSE SPECTRUM GENERATION, X DIRECTION
$
SET 777 = 999, 101, 201, 301, 401, 502, 602
$
DISPLACEMENT(PLOT) = 777 $ AT LEAST DISP AND VELO MUST BE REQUESTED
VELOCITY(PLOT) = 777 $ FOR SPECTRUM GENERATION POINT
ACCELERATION(PLOT) = 777
2 CHAPTER 6
Data Recovery
$
METHOD = 999
DLOAD = 997
TSTEP = 995
SDAMPING = 991 $ DAMPING SPECIFICATION FOR THE BUILDING
$
OUTPUT(XYPLOT)
XGRID = YES
YGRID = YES
$
$ PLOT INPUT
XTITLE = TIME
YTITLE = ACCEL INPUT
XYPLOT ACCE /999(T1)
YTITLE = DISPL INPUT
XYPLOT DISP /999(T1)
$
$ RESPONSE SPECTRA GENERATION
XTITLE = FREQUENCY (HZ)
YTITLE = ABSOLUTE ACCELERATION
$
$ NOTE: IP -- IMAGINARY,PHASE -- IS RELATIVE
$ NOTE: RM -- REAL, MAGNITUDE -- IS ABSOLUTE
$
XYPUNCH ACCE SPECTRAL 1 /502(T1RM)
$
BEGIN BULK
$
$ REQUESTS CALCULATION OF SPECTRA
PARAM, RSPECTRA, 0
$
$ RESPONSE SPECTRUM GENERATION TABLE
$ DTI, SPSEL, RECNO, DAMPL, FREQL, Gi
$ RECNO = N, SELECTED BY XYPLOT
$ DAMPL = I, FREQi THAT SPECIFIES DAMPING VALUES
$ FREQL = FREQi THAT SPECIFIES FREQ. VALUES
$ Gi = K, L, ..., GRID POINTS FOR SPECTRA GENERATION
DTI, SPSEL, 0
DTI, SPSEL, 1, 885, 887, 101, 201, 301, 401,+DTI1
+DTI1, 502, 602
$
$ SPECIFY OSCILLATOR DAMPING VALUE (2%)
FREQ, 885, 0.02
$
$ SPECIFY FREQUENCIES FOR SPECTRA
FREQ1, 887, 0.01, 0.25, 121
$
$ -----------------------------------------------------
$
$ ... rest of model ...
$
ENDDATA
The XYPUNCH ACCE SPECTRAL command in the Case Control Section punches the
response spectrum (which will be used later, in the response spectrum application). In the
Bulk Data Section, PARAM,RSPECTRA,0 requests calculation of the spectrum. The
DTI,SPSEL entry references the FREQ entries; one of the FREQ entries defines the oscillator
damping and the other FREQ entry defines the frequency range over which to compute the
spectra. The DTI,SPSEL entry also defines grid points at which to compute spectra; spectra
are computed for grid points 101, 201, 301, 401, 502, and 602one per floorthough only
the one for grid point 502 is punched for subsequent use.

Note that a 2% damped spectrum was computed because that is the spectrum that will be
applied to a component model (see below). In practice, however, spectra are often
generated for multiple damping values (for example, 0%, 2%, and 5% damping).
The plotted absolute acceleration response spectra for grid points 999 and 502 are shown
in Figure 6-3. Note that the response spectrum peaks in the region of about 1.5-7 Hz for
grid point 999 and at about 2-3 Hz and again at about 7 Hz for grid point 502. The building
has acted as both a filter (note the difference in frequency content) and magnifier (note the
difference in spectral amplitudes). Floor response spectra computed closer to the base of
the building will show less filtering and magnification.
Figure 6-3 Absolute Acceleration Response Spectrum Plots for Grid Points 999 (top)
and 502 (bottom)
A portion of the printed output file is shown in Figure 6-4. A portion of the absolute
acceleration spectrum output for grid point 502 is shown. Spectra for the other grid points
(101, 201, etc.) are also contained in the output file, as are displacement and velocity
A
b
s
.

A
c
c
e
l
.

(
i
n
.
/
s

)
2
A
b
s
.

A
c
c
e
l
.

(
i
n
.
/
s

)
2
600
0
0
Frequency (Hz)
35
3000
0
0
Frequency (Hz)
35
2 CHAPTER 6
Data Recovery
spectra. The printed format for each spectrum is similar to that of frequency response
analysis, with the absolute spectrum output in the real location and the relative spectrum
output in the imaginary location. Because relative acceleration spectra are not calculated,
those components are zero. The punched spectrum for grid point 502 is shown in
Figure 6-5; this output is contained in the punch file.
Figure 6-4 Printed Output (Abridged)
-etc.-
Figure 6-5 Punched Response Spectrum
FRACTION OF CRITICAL DAMPING = .02 ABSOLUTE IN REAL LOCATION, RELATIVE IN IMAG. LOCATION
POINT-ID = 502
C O M P L E X A C C E L E R A T I O N V E C T O R
(REAL/IMAGINARY)
FREQUENCY TYPE T1 T2 T3 R1 R2 R3
1.000000E-02 G 8.300698E-02 5.819805E-06 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
2.600000E-01 G 3.203262E+01 2.812203E-04 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
5.100000E-01 G 7.052912E+01 1.082366E-03 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
2.951000E+01 G 2.725781E+02 3.764749E-02 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
2.976000E+01 G 2.703750E+02 3.731322E-02 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
3.001000E+01 G 2.681875E+02 3.697610E-02 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
3.026000E+01 G 2.659297E+02 3.663826E-02 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0
$ACCE 3 502 3 1
$ 2.000000E-02
TABLED1 2
.01 .083007 .26 32.0326 .51 70.5291 .76 113.748
1.01 365.142 1.26 371.666 1.51 648.599 1.76 1278.37
2.01 1927.95 2.26 2781.64 2.51 2914.5 2.76 2063.92
3.01 1700.21 3.26 1325.52 3.51 950.985 3.76 976.762
4.01 865.546 4.26 688.777 4.51 743.792 4.76 863.43
5.01 848.97 5.26 828.419 5.51 804.627 5.76 863.462
6.01 835.807 6.26 949.985 6.51 1254.88 6.76 862.748
7.01 707.527 7.26 715.255 7.51 680.692 7.76 679.488
8.01 631.878 8.26 567.551 8.51 530.414 8.76 496.182
9.01 499.051 9.26 501.652 9.51 506.18 9.76 514.3
10.01 533.526 10.26 532.416 10.51 503.248 10.76 482.305
11.01 459.648 11.26 440.015 11.51 423.567 11.76 414.748
12.01 414.575 12.26 412.973 12.51 411.023 12.76 408.342
13.01 403.66 13.26 396.885 13.51 389.822 13.76 386.508
14.01 384.621 14.26 382.541 14.51 380.549 14.76 378.697
15.01 376.879 15.26 375.092 15.51 373.363 15.76 371.656
16.01 369.979 16.26 368.299 16.51 366.666 16.76 365.027
17.01 363.346 17.26 361.717 17.51 360.109 17.76 358.52
18.01 356.969 18.26 355.399 18.51 353.801 18.76 352.254
19.01 350.629 19.26 348.945 19.51 347.336 19.76 345.734
20.01 344.09 20.26 342.414 20.51 340.668 20.76 338.938
21.01 337.25 21.26 335.535 21.51 333.848 21.76 332.16
22.01 330.43 22.26 328.715 22.51 326.973 22.76 325.293
23.01 323.609 23.26 321.914 23.51 320.195 23.76 318.484
24.01 316.684 24.26 314.863 24.51 313.035 24.76 311.191
25.01 309.398 25.26 307.578 25.51 305.594 25.76 303.672
26.01 301.641 26.26 299.688 26.51 297.633 26.76 295.617
27.01 293.648 27.26 291.57 27.51 289.516 27.76 287.5
28.01 285.367 28.26 283.266 28.51 281.219 28.76 279.023
29.01 276.953 29.26 274.742 29.51 272.578 29.76 270.375
30.01 268.188 30.26 265.93 ENDT

Response Spectrum Application
Figure 6-6 depicts a bracket model to which the spectrum will be applied. The twelve grid
points shown in the figure are connected via an RBE2 element to a separate grid point (888);
this grid point has a SUPORT applied in the y direction (T2) and a large mass attached to
it. The twelve grid points are constrained in all but the T2 direction. This is the grid point
to which the absolute acceleration spectrum generated in the previous run (see above) will
be applied, in the y direction. Displacements, accelerations, and stresses are computed. A
portion of the input file is shown in Listing 6-2.
Figure 6-6 Bracket Model
Listing 6-2 Abridged Input File for Response Spectrum Application
$ RESPONSE SPECTRUM APPLICATION
TIME 60
SOL 103
CEND
TITLE = BRACKET MODEL
SUBTITLE = APPLY ABS. ACCEL. SPECTRUM IN Y DIRECTION
SPC = 1
$
METHOD = 777
SDAMP = 851 $ SELECTS TABDMP1 FOR MODAL DAMPING
DLOAD = 852 $ SELECTS DLOAD ENTRY
$
DISPLACEMENT = ALL
STRESS = ALL
ACCELERATION = ALL
$
$
BEGIN BULK
2 CHAPTER 6
Data Recovery
PARAM,AUTOSPC,YES
$
$ SID V1 V2 ND
EIGRL,777, , ,3
$
$ PERFORM SHOCK SPECTRUM ANALYSIS
PARAM,SCRSPEC,0
$
$ 2% CRITICAL DAMPING
TABDMP1, 851, CRIT, , , , , , ,+TB1
+TB1, 0.0, 0.02, 50.0, 0.02, ENDT
$
$ SPECIFY SPECTRUM
DLOAD, 852, 1.0, 1.0, 1
$
$ RELATES SPECTRA LINES TO DAMPING
DTI,SPECSEL,0
DTI,SPECSEL,1, ,A, 2, 0.02
$
$ SUPORT POINT FOR SPECTRA
GRID, 888, , 3.0, 0.0, 0.0
CONM2, 888, 888, , 1.0E6
SPC, 1, 888, 13456
SUPORT, 888, 2
RBE2, 888, 888, 2, 14, 15, 22, 23, 133,+RB4
+RB4, 134, 141, 142, 145, 146, 158, 159
$
$ PUNCHED SPECTRUM FROM GENERATION RUN
$ACCE 3 502 3 1
$ 2.000000E-02
TABLED1 2
.01 .083007 .26 32.0326 .51 70.5291 .76 113.748
1.01 365.142 1.26 371.666 1.51 648.599 1.76 1278.37
2.01 1927.95 2.26 2781.64 2.51 2914.5 2.76 2063.92
3.01 1700.21 3.26 1325.52 3.51 950.985 3.76 976.762
4.01 865.546 4.26 688.777 4.51 743.792 4.76 863.43
5.01 848.97 5.26 828.419 5.51 804.627 5.76 863.462
6.01 835.807 6.26 949.985 6.51 1254.88 6.76 862.748
7.01 707.527 7.26 715.255 7.51 680.692 7.76 679.488
8.01 631.878 8.26 567.551 8.51 530.414 8.76 496.182
9.01 499.051 9.26 501.652 9.51 506.18 9.76 514.3
10.01 533.526 10.26 532.416 10.51 503.248 10.76 482.305
11.01 459.648 11.26 440.015 11.51 423.567 11.76 414.748
12.01 414.575 12.26 412.973 12.51 411.023 12.76 408.342
13.01 403.66 13.26 396.885 13.51 389.822 13.76 386.508
14.01 384.621 14.26 382.541 14.51 380.549 14.76 378.697
15.01 376.879 15.26 375.092 15.51 373.363 15.76 371.656
16.01 369.979 16.26 368.299 16.51 366.666 16.76 365.027
17.01 363.346 17.26 361.717 17.51 360.109 17.76 358.52
18.01 356.969 18.26 355.399 18.51 353.801 18.76 352.254
19.01 350.629 19.26 348.945 19.51 347.336 19.76 345.734
20.01 344.09 20.26 342.414 20.51 340.668 20.76 338.938
21.01 337.25 21.26 335.535 21.51 333.848 21.76 332.16
22.01 330.43 22.26 328.715 22.51 326.973 22.76 325.293
23.01 323.609 23.26 321.914 23.51 320.195 23.76 318.484
24.01 316.684 24.26 314.863 24.51 313.035 24.76 311.191
25.01 309.398 25.26 307.578 25.51 305.594 25.76 303.672
26.01 301.641 26.26 299.688 26.51 297.633 26.76 295.617
27.01 293.648 27.26 291.57 27.51 289.516 27.76 287.5
28.01 285.367 28.26 283.266 28.51 281.219 28.76 279.023
29.01 276.953 29.26 274.742 29.51 272.578 29.76 270.375
30.01 268.188 30.26 265.93 ENDT
$
$-------------------------------------------------------
$
$ ... rest of model ...
$
ENDDATA

Response spectrum application is invoked via PARAM,SCRSPEC,0. The DTI,SPECSEL
entry specifies that the input spectrum is acceleration (denoted by the A in field 5). The
TABDMP1 entry defines the modal damping (every mode has 2% critical damping). The
TABLED1 entry defines the input spectrum; this is the punch file that was generated from
the spectrum generation run.
Discussion of Results. Figure 6-7 shows a portion of the resulting printed output. The
eigenvalue summary shows the computed natural frequencies; note the rigid body mode,
which occurred because the bracket was not constrained in the y direction (there was a
SUPORT for that DOF). Matrix FN is the list of natural frequencies of the modes used for
analysis with response spectrum input (two in this casethe rigid body mode is not
included because our spectrum started at 0.01 Hz). Matrix PSIT lists the modal
participation factors in transposed form, with one column for each mode (three in this case)
and one row for each input point (one in this case). Note that the third mode (the second
elastic mode) cannot be readily excited by base motion, since its response to the load is 10
orders of magnitude less than the second (first elastic) mode. Matrix UHVR occurs once
for every analysis subcase with response spectrum input, and it lists the peak modal
response. The first column is displacement, the second is velocity, and the third is
acceleration. There is one row for each mode used in the response spectrum analysis (two
in this case). A portion of the resulting maximum displacements, accelerations, and
stressesquantities selected for output via Case Controlare also shown. These
quantities are computed using the ABS (default) method.
2 CHAPTER 6
Data Recovery
Figure 6-7 Printed Output (Abridged)
R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 0.0 0.0 0.0 1.000000E+00 0.0
2 2 3.930305E+03 6.269214E+01 9.977764E+00 1.000000E+00 3.930305E+03
3 3 2.878402E+04 1.696586E+02 2.700200E+01 1.000000E+00 2.878402E+04
^^^ DMAP INFORMATION MESSAGE 9047 (POSTREIG) - SCALED RESPONSE SPECTRA FOR RESIDUAL STRUCTURE ONLY
MATRIX FN (GINO NAME 101 ) IS A DB PREC 1 COLUMN X 3 ROW RECTANG MATRIX.
COLUMN 1 ROWS 2 THRU 3 --------------------------------------------------
ROW
2) 9.977767703979D+00 2.700200654232D+01
THE NUMBER OF NON-ZERO TERMS IN THE DENSEST COLUMN = 2
THE DENSITY OF THIS MATRIX IS 66.66 PERCENT.
PSIT
POINT VALUE POINT VALUE P0INT VALUE POINT VALUE POINT VALUE
COLUMN 1
888 T2 -1.00000E+03
COLUMN 2
888 T2 1.51209E-01
COLUMN 3
888 T2 8.54947E-11
U S E T D E F I N I T I O N T A B L E ( I N T E R N A L S E Q U E N C E , R O W S O R T )
R DISPLACEMENT SET
-1- -2- -3- -4- -5- -6- -7- -8- -9- -10-

1= 888-2
SCALED SPECTRAL RESPONSE, ABS OPTION, DLOAD = 852 CLOSE = 1.00
MATRIX UHVR (GINO NAME 101 ) IS A REAL 3 COLUMN X 3 ROW SQUARE MATRIX.
COLUMN 1 ROWS 2 THRU 3 --------------------------------------------------
ROW
2) -2.0431E-02 -8.7238E-13
COLUMN 2 ROWS 2 THRU 3 --------------------------------------------------
ROW
2) -1.2808E+00 -1.4801E-10
COLUMN 3 ROWS 2 THRU 3 --------------------------------------------------
ROW
2) -8.0299E+01 -2.5111E-08
THE NUMBER OF NON-ZERO TERMS IN THE DENSEST COLUMN = 2
THE DENSITY OF THIS MATRIX IS 66.66 PERCENT.
TIME = 0.000000E+00
D I S P L A C E M E N T V E C T O R

5 G 1.982064E-05 5.079640E-04 1.304063E-05 4.452030E-04 0.0 3.272993E-04
6 G 2.216715E-05 4.475303E-04 5.667240E-06 2.254874E-03 0.0 6.874480E-05
7 G 5.777859E-05 3.471966E-03 3.236318E-05 1.011459E-02 0.0 1.185161E-02
8 G 6.727507E-05 9.782481E-03 9.783542E-05 7.597393E-03 2.088506E-03 1.105588E-02
13 G 4.799800E-06 2.431473E-04 1.162851E-05 2.776468E-03 0.0 1.267932E-03
14 G 0.0 3.089312E-09 0.0 0.0 0.0 0.0
15 G 0.0 3.089312E-09 0.0 0.0 0.0 0.0
999 G 7.124796E-09 3.353467E-02 4.805730E-02 1.705154E-02 3.335225E-10 2.368395E-08
TIME = 0.000000E+00
A C C E L E R A T I O N V E C T O R

5 G 7.790122E-02 1.996455E+00 5.125370E-02 1.749785E+00 0.0 1.286387E+00
6 G 8.712371E-02 1.758932E+00 2.227400E-02 8.862347E+00 0.0 2.701882E-01
7 G 2.270876E-01 1.364589E+01 1.271973E-01 3.975345E+01 0.0 4.658046E+01
8 G 2.644117E-01 3.844816E+01 3.845233E-01 2.986009E+01 8.208470E+00 4.345302E+01
13 G 1.886469E-02 9.556437E-01 4.570363E-02 1.091238E+01 0.0 4.983363E+00
14 G 0.0 1.214195E-05 0.0 0.0 0.0 0.0
15 G 0.0 1.214195E-05 0.0 0.0 0.0 0.0
999 G 2.806829E-05 1.318016E+02 1.888800E+02 6.701779E+01 1.318517E-06 9.310856E-05
TIME = 0.000000E+00
S T R E S S E S I N Q U A D R I L A T E R A L E L E M E N T S ( Q U A D 4 )
ELEMENT FIBRE STRESSES IN ELEMENT COORD SYSTEM PRINCIPAL STRESSES (ZERO SHEAR)
ID. DISTANCE NORMAL-X NORMAL-Y SHEAR-XY ANGLE MAJOR MINOR VON MISES
4 -2.500000E-02 9.571945E+02 1.588326E+02 6.315743E+02 28.8527 1.305162E+03 -1.891354E+02 1.409281E+03
2.500000E-02 8.720101E+02 4.871640E+02 7.984738E+02 38.2254 1.500919E+03 -1.417454E+02 1.576578E+03
6 -2.500000E-02 1.215178E+04 9.352043E+03 5.037602E+03 37.2351 1.598040E+04 5.523426E+03 1.405755E+04
2.500000E-02 1.361088E+04 8.486970E+03 3.161832E+03 25.4915 1.511842E+04 6.979430E+03 1.310577E+04
235 -2.500000E-02 2.137487E+02 3.799783E+00 1.563414E+00 0.4266 2.137604E+02 3.788147E+00 2.118917E+02
2.500000E-02 1.987584E+02 4.320534E+00 9.967642E-01 0.2937 1.987635E+02 4.315422E+00 1.966413E+02

6.5 Random Analysis with Coupled Excitations
In MSC.Nastran, random response analysis is treated as a data reduction procedure that is
applied to the results of a frequency response analysis. First, the frequency response
analysis is performed for sinusoidal loading conditions, , each a separate subcase, at a
sequence of frequencies . Normal data reduction procedures are then applied to the
output of the frequency response analysis module, resulting in a set of output quantities
, corresponding to an output and subcase . The calculations of power spectral
densities and autocorrelation functions for the output quantities are performed in the
random analysis module.
Each loading condition subcase represents a unique random load source, which may be
applied to many grid points. Typically, these loads are chosen to be unit loads such as unit
g loads or unit pressures. The probabilistic magnitude of each load source is defined by
spectral density functions on RANDPS input data. If the load subcases are correlated, the
coupling spectral density is also defined on one or more RANDPS functions. An example
of coupled spectral density would be the forces on four wheels of a vehicle traveling over
a rough road.
Figure 6-8 is a simplified flow diagram for the random analysis module. The inputs to the
module are the frequency responses, , of quantities to loading conditions at
frequencies , and the auto- and cross-spectral densities of the loading conditions and
. The response quantities, , may be displacements, velocities, accelerations, internal
forces, or stresses. The power spectral densities of the response quantities are calculated
by different procedures depending on whether the loading conditions are correlated or
uncorrelated. The spectral densities due to all sources, considered independent, will be
combined into one set of outputs.
P
a
{ }
i
u
j a

i
( ) j a
H
ja

i
( ) u
j
P
a
{ }
i
S
a
S
ab
S
j
2 CHAPTER 6
Data Recovery
Figure 6-8 Flow Diagram for Random Analysis Module
Theory
The application of these frequency response techniques to the analysis of random
processes requires that the system be linear and that the excitation be stationary with
respect to time. The theory includes a few important theorems that will be reviewed.
An important quantity in random analysis theory is the autocorrelation function , of
a physical variable, , which is defined by
Enter
Combined
Loading
?
Compute
Autocorrelation
Functions
Cross
Spectral
Density
?
Exit
Compute rms
value
Yes No
Yes
No
Yes
No
Autocorrelation?
S
j
S
j a )
=
S
j
H
j a
H
j b
S
ab
b
)
a
)
= S
j a
H
j a
2
S
a
=
R
j
( )
u
j

Eq. 6-41
Note that is the time average value of , which is an important quantity in the
analysis of structural failure. The one-sided power spectral density of is defined
by
Eq. 6-42
It may be shown (using the theory of Fourier integrals) that the autocorrelation function
and the power spectral density are Fourier transforms of each other. Thus, we define the
autocorrelation function in terms of frequency response functions
Eq. 6-43
from the mean-square theorem, the rms (root mean squared) magnitude, , is
Eq. 6-44
The transfer function theorem states that, if is the frequency response of any
physical variable, , due to an excitation source, , which may be a point force, a loading
condition or some other form of excitation, i.e., if
Eq. 6-45
where and are the Fourier transforms of and , then the power spectral
density of the response , is related to the power spectral density of the source, ,
by
Eq. 6-46
Eq. 6-46 is an important result because it allows the statistical properties (e.g., the
autocorrelation function) of the response of a system to random excitation to be evaluated
via the techniques of frequency response. Another useful result is that, if sources , ,
, etc., are statistically independent, i.e., if the cross-correlation function between any pair
of sources
Eq. 6-47
is null, then the power spectral density of the total response is equal to the sum of the
power spectral densities of the responses due to individual sources. Thus
Eq. 6-48
R
j
( ) T
1
T
--- u
j
t ( )u
j
t ( ) t d
0
T
(
=
lim
R
j
0 ( ) u
j
2
S
j
( ) u
j
S
j
( ) T
2
T
--- e
i t
u
j
t ( ) t d
0
T
(
2
=
lim
R
j
( )
1
2
------ S
j
( ) ( ) cos d
0
(
=
u
j
u
j
2
R
j
0 ( )
1
2
------ S
j
( ) d
0
(
= =
H
ja
( )
u
j
Q
a
u
j
( ) H
j a
( ) Q
a
( ) =
u
j
( ) Q
a
( ) u
j
Q
a
S
j
( ) S
a
( )
S
j
( ) H
j a
( )
2
S
a
( ) =
Q
1
Q
2
Q
3
R
ab
( ) T
1
T
--- q
a
t ( )q
b
t ( ) d
0
T
(
=
lim
S
j
( ) S
j a
( )
a
)
H
j a
( )
2
S
a
( )
a
)
= =
2 CHAPTER 6
Data Recovery
If the sources are statistically correlated, the degree of correlation can be expressed by a
cross-spectral density, , and the spectral density of the response may be evaluated from
Eq. 6-49
where is the complex conjugate of .
In applying the theory, it is not necessary to consider the sources to be forces at individual
points. Rather, an ensemble of applied forces that are completely correlated (i.e., a loading
condition) should be treated as a single source. For example, a plane pressure wave from
a specified direction may be treated as a source. Furthermore, the response may be any
physical variable including internal forces and stresses as well as displacements, velocities,
and accelerations.
The power spectral densities, , are plotted versus frequency using the XYOUT
commands (XYPLOT, XYPRINT, XYPEAK, AND XYPUNCH). The autocorrelation
functions, , are plotted versus selected time delays, , using XYOUT requests.
Because of the potential computational costs, cross-correlations and cross-spectral
densities.
Example: Random Analysis of a Simple Structure
There are a number of physical phenomena that result in nondeterministic data where
specific responses cannot be predicted in a deterministic sense. Examples are earthquake
ground motion, the height and spacing of ocean waves, pressure gusts encountered by
aircraft in flight, by bridges during storms, and acoustic excitation such as jet engine noise.
The response of physical systems to such unpredictable values of excitation is an
application of random analysis methods.
In MSC.Nastran, ergodic random analysis is treated as a data reduction process that
expands the results of a frequency response analysis.
The inputs for these data recovery activities are output quantities from the frequency
response analysis, and user-supplied spectral data for the loading conditions. The outputs
are the power spectral density functions, the autocorrelation functions, the number of
positive crossings of the line by the response record in a unit time, and the rms
value of the response. These latter two items are of particular importance to the stress
analyst. The number of positive crossings, , is obtained from the joint probability
density function and is a measure of the apparent frequency of the response. The rms value
of the response is a measure of the dynamic magnitudes.
Simple output plots for the problem are shown in the following figures.
Model Description
The model illustrated here consists of three masses connected by springs and viscous
dampers. The base mass, , is considered to be the ground and is used to enforce a
excitation. The base is excited by a noise excitation displacement having a white noise
S
ab
S
j
H
j a
H
j b
S
ab
b
)
a
)
=
H
jb
H
jb
S
j
R
j
( )
y t ( ) 0 =
N
0
M
1
y t ( )

power spectral density . The middle mass is excited by a band-limited white
noise excitation force having a constant power spectral density . The two
loadings are correlated through a cross-spectral density .
Figure 6-9 Example Model
In the figure, the following notation has been used:
Mass of i-th grid point.
Stiffness between i-th and j-th grid points.
Viscous damping between i-th and j-th grid points.
Nondimensional damping ratio = .
Circular frequency = .
Mass Properties
EID G M
1 1
2 2
3 3
S
2
f ( )inch
2
Hz
F t ( ) S
1
f ( )lb
2
Hz
S
12
f ( )in-lb/Hz
Subcase 2
Subcase 1
M
3
M
1
M
2
M
3
M
2
=
k
23

3
2
M
3
=
k
12

2
2
M
2
=
C
23
2
3
3
M
3
=
C
12
2
2
2
M
2
=
y t ( )
F t ( )
M
i
k
i j
C
i j
2
j
C
i j
k
i j
M
j
j
k
ij
M
j
1.0E9 lb-sec
2
in
.1 lb-sec
2
in
1. lb-sec
2
in
2 CHAPTER 6
Data Recovery
For a check on the MSC.Nastran results, the responses at 200 Hz, 280 Hz, and 500 Hz are
tabulated. Results are given as magnitude/phase format.
Spring Properties
EID K
12 5.0E6 lb/in
23 10.0E6 lb/in
Damper Properties
EID B
120 14.14 lb-sec/in
230 63.25 lb-sec/in
Constraints
Node Constraint
1 13456
2 13456
3 13456
Frequencies (f)
200 Hz to 600 Hz by 10 Hz
Loads
Node Load Subcase
2 1
1 2
Spectral Densities
Excited Set (j) Applied Set (h)
S
jk
1 1
2 2
1 2
3 in-lb-hz
F
y
100. =
F
y
1.0E9 ( ) 2f ( )
2
lb =
1 10
5
lb
2
hz
1 10
4
in
2
hz

The auto-power spectral density of the j-th response is defined by the relationship
Eq. 6-50
The results for selected frequencies are tabulated below.
As discussed in the problem description, the number of positive crossings is a measure
of the apparent response frequency. This measure is compared below.
Frequency Response Theoretical MSC.Nastran
200 Hz
280 Hz
500 Hz
100 H
2f d
100 H
3f d
H
2d d
H
3d d
3.3705 10
5
/ 359.6029
4.0025 10
5
/ 359.5175
1.6853 / 179.8065
2.0013 / 179.7211
3.3705 10
5
/ 359.6029
4.0025 10
5
/ 359.5175
1.6853 / 179.8065
2.0013 / 179.7211
100 H
2f d
100 H
3f d
H
2d d
H
3d d
4.6806 10
4
/ 347.2289
6.7782 10
4
/ 346.9431
23.4031 / 167.5139
33.8912 / 167.2282
4.6806 10
4
/ 347.2288
6.7782 10
4
/ 346.9430
23.4031 / 167.5138
33.8911 / 167.2281
100 H
2f d
100 H
3f d
H
2d d
H
3d d
2.4206 10
7
/ 236.0504
1.0187 10
5
/ 180.4626
1.2103 10
2
/ 56.5594
.50935 / .9717
2.4025 10
7
/ 236.0506
1.0187 10
5
/ 180.4626
1.2103 10
2
/ 56.5600
0.50934 / .9717
Frequency
PSD M
2
Theoretical MSC.Nastran
200 Hz
280 Hz
500 Hz
S
j
f ( ) H
a
f ( )S
ab
H
b
f ( )
b
)
a
)
=
5.6805 10
5
5.6803 10
5
1.0955 10
2
1.0955 10
2
2.9307 10
9
2.9305 10
9
N
0
2 CHAPTER 6
Data Recovery
Listing 6-3 Data File for Example from Figure 6-9
ID FREQRSPN,RANDOM
TIME 5
DIAG 8
SOL 111
CEND
TITLE=FREQ RSPN - RANDOM ANALYSIS DEMONSTRATION
ECHO=UNSORT
METHOD=1 $ Reduce to Modal Coordinates
FREQUENCY=1
RANDOM=1 $ Basic request
SVEC=ALL $ Modal Coordinate output
SDISP(SORT2,PHASE)=ALL
DISP(SORT2,PHASE)=ALL
VELO(SORT2,PHASE,PLOT)=ALL
ACCEL(SORT2,PHASE,PLOT)=ALL
ELFORCE(SORT2,PHASE,PLOT)=ALL
ELSTRESS(SORT2,PHASE,PLOT)=ALL
SUBCASE 1
LABEL=FORCING FUNCTION ON MASS AT GRID POINT NUMBER 2
DLOAD=1
SUBCASE 2
LABEL=ENFORCED DISPLACEMENT ON GRID POINT NUMBER 1
DLOAD=2
$ ********************************************************
$ *** RANDOM ANALYSIS CASE CONTROL ***
OUTPUT (XYOUT)
XGRID=YES
YGRID=YES
XTITLE=FREQUENCY (HERTZ)
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
XTITLE=FREQUENCY (HERTZ)
YTITLE=VELOCITY RESPONSE LOAD CASE 1
YLOG=YES
XYPLOT VELO RESPONSE 1 /2(T2RM),3(T2RM)
YLOG=NO
YTITLE=PHASE
XYPLOT VELO RESPONSE 1 /2(T2IP),3(T2IP)
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
YLOG=YES
YTITLE=DISP P S D
XYPLOT DISP PSDF /2(T2)/3(T2)
YTITLE=ACCE P S
XYPLOT ACCE PSDF /2(T2)/3(T2)
YTITLE=FORCE P S D
XYPLOT ELFORCE PSDF /12(2)/23(2)
XTITLE=TIME (SEC)
YLOG=NO
YTITLE=VELOCITY AUTO CORROLATION
XYPLOT VELO AUTO /2(T2)/3(T2)
BEGIN BULK
$ *** STRUCTURE DEFINITION ***
Frequency
1st Mode
N
0
Theoretical
M
2
Displacement
M
2
284.2 Hz 282.4 Hz 284.3 Hz10

GRID 1 0. 0. 0. 13456
GRID 2 0. 1. 0. 13456
GRID 3 0. 2. 0. 13456
CONM2 1 1 1.+9
CONM2 2 2 .1
CONM2 3 3 1.
CELAS212 5.+6 1 2 2 2 .5
CELAS223 10.+6 2 2 3 2 .5
CDAMP2120 14.14 1 2 2 2
CDAMP2230 63.25 2 2 3 2
SUPORT1 2
$ *** EIGENVALUE SETUP ***
EIGR 1 MGIV 3 EIGR1
+EIGR1MASS
$ *** FREQUENCY SELECTION ***
FREQ1 1 200. 10. 40
$ *** APPLIED DYNAMIC LOAD SUBCASE 1 ***
DAREA 10 2 2 1.+2
RLOAD11 10 10
TABLED110 TABD1
+TABD11. 1. 10000.1. ENDT
$ *** ENFORCED DISPLACEMENT SUBCASE 2 ***
DAREA 20 1 2 1.+9
RLOAD12 20 110
TABLED4110 0. 1. 1. 10000. TABD4
+TABD40. 0. 39.478ENDT
$ *** RANDOM ANALYSIS DATA ***
RANDPS1 1 1 1.+5 1000
RANDPS1 2 2 1.-4 1000
RANDPS1 1 2 3. 1000
TABRND11000 TABR1
+TABR10. 0. 199.990. 200. 1. 600. 1. TABR2
+TABR2600.010. 10000.0. ENDT
RANDT11 100 0. .1
ENDDATA
The following is the key printed output:
FREQ RSPN - RANDOM ANALYSIS DEMONSTRATION AUGUST 31, 1994 MSC.Nastran 8/17/94 PAGE 6

R E A L E I G E N V A L U E S
MODE EXTRACTION EIGENVALUE RADIANS CYCLES GENERALIZED GENERALIZED
NO. ORDER MASS STIFFNESS
1 1 .0 .0 .0 1.000000E+00 .0
2 2 3.188542E+06 1.785649E+03 2.841949E+02 1.000000E+00 3.188542E+06
3 3 1.568106E+08 1.252240E+04 1.993002E+03 1.000000E+00 1.568106E+08


X Y - O U T P U T S U M M A R Y ( A U T O O R P S D F )
PLOT CURVE FRAME RMS NO. POSITIVE XMIN FOR XMAX FOR YMIN FOR X FOR YMAX FOR X FOR*
TYPE TYPE NO. CURVE ID. VALUE CROSSINGS ALL DATA ALL DATA ALL DATA YMIN ALL DATA YMAX
PSDF DISP 5 2( 4) 4.427970E-01 2.823553E+02 2.000E+02 6.000E+02 2.931E-09 5.000E+02 1.095E-02 2.800E+02
PSDF DISP 6 3( 4) 6.487856E-01 2.843350E+02 2.000E+02 6.000E+02 2.025E-06 6.000E+02 2.297E-02 2.800E+02
PSDF ACCE 7 2( 4) 1.400392E+06 2.856203E+02 2.000E+02 6.000E+02 2.855E+05 5.000E+02 1.049E+11 2.800E+02
PSDF ACCE 8 3( 4) 2.091627E+06 2.934378E+02 2.000E+02 6.000E+02 1.997E+08 2.000E+02 2.201E+11 2.800E+02
PSDF EL FOR 9 12( 2) 2.375014E+06 3.161015E+02 2.000E+02 6.000E+02 5.009E+08 2.200E+02 2.306E+11 2.800E+02
PSDF EL FOR 10 23( 2) 2.091485E+06 2.934356E+02 2.000E+02 6.000E+02 1.997E+08 2.000E+02 2.201E+11 2.800E+02
AUTO VELO 11 2( 4) 7.854708E+02 2.837833E+02 0.000E+00 1.000E-01 -5.550E+05 2.000E-03 6.170E+05 0.000E+00
AUTO VELO 12 3( 4) 1.158947E+03 2.872666E+02 0.000E+00 1.000E-01 -1.179E+06 2.000E-03 1.343E+06 0.000E+00
2 CHAPTER 6
Data Recovery
Figure 6-10 Velocity Response, Points 2 and 3 (Subcase 1)
Figure 6-11 Velocity Phase Angle, Points 2 and 3 (Subcase 1)

Figure 6-12 Power Spectral Density, Point 2 Displacement
Figure 6-13 Power Spectral Density, Point 3 Displacement
2 CHAPTER 6
Data Recovery
Figure 6-14 PSDF Forces in Element 12
Figure 6-15 Autocorrelation, Velocity of Point 2

APPENDIX
A
Theory of Eigenvalue Extraction
Methods
I Sturm Sequence Theory
I Eigensolution Methods

A.1 Sturm Sequence Theory
Methods of Computation
A numerical technique which is very helpful in a modal extraction is Sturm Sequence. In
this approach, a frequency is selected and computations are performed to determine the
number of roots which exist below that frequency. In some cases this frequency is chosen
manually by user input, other times it is selected automatically by MSC.Nastran. This
choice depends upon which modal extraction technique is being used.
Starting with the basic characteristic equation
Eq. A-1
a frequency, is substituted. This frequency is called the shift point. The
determinant is evaluated for the frequency . The resulting matrix is factored into
it components. The diagonal matrix, , is known as the Factor Diagonal
Matrix. The number of negative terms in the Factor Diagonal Matrix, known as the Sturm
Number, are the number of frequencies that exist below the shift point frequency, .
Sturm Sequence logic is a very important part of the many of the modal extraction and
dynamic reduction techniques. When it is utilized, the Sturm Sequence logic issues a
number of messages which indicate the various shifts within the frequency range of
interest. These messages help the various techniques to determine how many roots are in
specific ranges and also if all modes have been found in the range.
K
s
M + [ ] 0 =
s
=
K
s
M
s
L [ ] D [ ] L [ ]
T
D [ ]
s
2 APPENDIX A
Theory of Eigenvalue Extraction Methods
A.2 Eigensolution Methods
Methods of Eigensolution
Seven methods of real eigenvalue extraction are provided in MSC.Nastran. These methods
are numerical approaches to solving the equation of motion for its natural frequencies and
modes shapes. The reason for seven different numerical techniques is related to the fact
that no one method is the best for all problems. While most of the methods can be applied
to all problems, the difference between the methods many times reduces to the efficiency
of the solution process.
Most methods of algebraic eigenvalue extraction belong to one of two groups:
1. transformation methods, and
2. tracking methods.
In a transformation method, the eigenvalue equation is first transformed into a special
form (such as tridiagonal) from which eigenvalues may easily be extracted. In a tracking
method, the roots are extracted from the original dynamic matrix one at a time using an
iterative procedure.
Four of the real eigenvalue extraction methods available in MSC.Nastran are
transformation methods. These are:
1. Givens Methods
2. Householder Method
3. Modified Givens Method, and
4. Modified Householder Method.
These transformation methods use a tridiagonal solution method.
Two of the real eigenvalue extraction methods available in MSC.Nastran are classified as
tracking methods. These are:
1. Inverse Power Method, and
2. Sturm Modified Inverse Power Method.
The final real eigenvalue extraction method available in MSC.Nastran is the Lanczos
method. The Lanczos method combines the best characteristics of the tracking and
transformation methods.
The Standard Tridiagonal Methods (GIV and HOU)
The Givens(GIV) and Householder (HOU) modal extraction methods require a positive
definite mass matrix. There are no restrictions on the stiffness matrix other than that it
must be symmetric. These matrices will always result in real (positive) roots. The
tridiagonal methods are the most efficient methods for small problems and problems with
dense matrices when a large portion of the eigenvectors are needed. These methods find
all of the eigenvalues in the structure as a result of the transformation process. The user
must make a specific request for the required eigenvectors (mode shapes). The tridiagonal

methods do not take advantage of the sparse matrices, but they are efficient with the dense
matrices sometimes created using dynamic reduction. The steps used in the tridiagonal
methods are as follows:
1. Perform a Cholesky decomposition of the mass matrix:
Eq. A-2
where is a lower triangular matrix.
2. The symmetric eigenvalue equation can be written as:
Eq. A-3
where is the symmetric matrix and is the identity matrix. To do this,
premultiply Eq. A-1 by and substitute for from Eq. A-2 to set the
following:
Eq. A-4
Then make the transformation:
Eq. A-5
which reduces Eq. A-4 to Eq. A-3 using the following:
Eq. A-6
3. The matrix is converted to a tridiagonal matrix by a transformation method
according to the Givens or Householder method. (A tridiagonal matrix is a
matrix where the only nonzero terms in the i-th column are the i 1, i, i + 1 row.)
4. All the eigenvalues of the tridiagonal matrix are extracted using a modified QR
algorithm.
5. The eigenvectors are computed over a given frequency range or, at user option,
for a given number of eigenvalues using inverse iteration. Since the roots are
known to high accuracy at this point, the vectors converge rapidly.
6. Physical eigenvectors are recovered by performing the inverse transformation to
that made in Step 3 and then solving Eq. A-5 as follows:
Eq. A-7
7. If a SUPORT entry is used, the first eigenvectors computed are discarded. The
rigid body modes from the SUPORT computation are put in their place, and the
corresponding eigenvalues are set to zero. The SUPORT operation does cause
some redundant operations to be performed as the rigid body modes are
computed by two methods and then one set is discarded. This is generally a
minor issue.
M [ ] L [ ] L [ ]
T
=
L [ ]
J I [ ] w { } 0 =
J [ ] I [ ]
L [ ]
1
M [ ]
L [ ]
1
K [ ] u { } L [ ]
1
L [ ] L [ ]
T
u { } 0 =
w { } L [ ] u { } =
J [ ] L [ ]
1
K [ ] L [ ]
1, T
=
J [ ]
u { } L [ ]
T,1
w { } =
N
r
2 APPENDIX A
8. The back-transformed vectors are given a final stage of orthogonalization
with respect to the mass matrix. The modal mass matrix is calculated as follows:
Eq. A-8
then decomposed,
Eq. A-9
The refined eigenvectors, are found by a forward pass on the equation:
Eq. A-10
Note that the tridiagonal methods fail in Step 1 if the mass matrix is not positive definite.
In order to minimize this eventuality, degrees-of-freedom with null columns are removed
from ( ) by the automatic application of static condensation. This is called AUTO-OMIT.
(This procedure can be bypassed by using PARAM,ASING). The application of the
AUTO-OMIT process is a precaution and may not remove all possible causes of mass
matrix singularity such as, a point mass offset from a grid point, but it greatly improves the
reliability and convenience of the standard tridiagonal methods.
The Modified Tridiagonal Methods (MGIV, MHOU, AGIV, AHOU)
The modified tridiagonal methods, Modified Givens (MGIV) and Modified Householder
(MHOU), are similar to the standard tridiagonal methods (GIV, HOU) with the exception
that the mass matrix can be singular.
The steps used in the modified methods are as follows:
1. Perform a Cholesky decomposition of the positive definite matrix
Eq. A-11
where is a positive number automatically selected by the program to optimize
the reliability and accuracy of eigenvalue extraction. It is large enough to stabilize
all rigid body modes but also be as small as possible to reduce the numerical
contamination. The goal of the algorithm is to find a value near the first flexible
frequency of the system. It uses the diagonal terms of the mass and stiffness
matrices as follows:
Eq. A-12
PHI [ ]
M
modal
[ ] [ ]
T
M [ ] [ ] =
M
modal
[ ] L [ ]
T
L [ ] =
[ ]
T
L [ ] [ ]
T
[ ]
T
=
u
K
s
M + [ ] L [ ] L [ ]
T
=
s
1
N
1 2
M
i i
K
i i
--------
i 1 =
N
)
------------------------------------ =

where is the dimension of the matrices. If or (an under defined
degree-of-freedom) or (a likely very large mass term used to
approximate a rigid body mode), the corresponding term is omitted from the
summation. If the value calculated by this approximate method does not result
in a stable Cholesky decomposition, is increased by adding the arbitrary
amount of 0.01. If this value also results in failure, the original value is increased
by 0.02. If this value also fails, a fatal error is issued.
2. Write the eigen-equation standard form as:
Eq. A-13
where:
Eq. A-14
and:
Eq. A-15
by rearranging the terms to read as follows:
Eq. A-16
Then premultiply Eq. A-16 by and substitute for from Eq.
A-11 to obtain:
Eq. A-17
Premultiply by and then substitute:
Eq. A-18
resulting in Eq. A-13, Eq. A-14, and Eq. A-15.
3. The remaining steps are identical to those for the unmodified methods where:
Eq. A-19
Although the mass matrix is not required to be nonsingular in the modified
methods, a singular mass matrix can produce one or more zero eigenvalues, ,
which are equivalent to infinite eigenvalues, . Due to roundoff error, these will
appear in the output as very large positive or negative eigenvalues. To reduce the
incidence of such meaningless results, degrees-of-freedom with null masses are
eliminated by the AUTO-OMIT static condensation as in the case of the
unmodified methods. This step can be bypassed by using PARAM,ASING.
Many times the user is not conversant in the status of the mass matrix prior to the actual
matrix assembly. To assist the user in choosing the appropriate method, two options,
Automatic Givens (AGIV) and Automatic Householder (AHOU) are available. The
N M
i i
X
i i
0 =
M
i i
X
i i
10**8 >
s
J I [ ] w { } 0 =
1

s
+
---------------- =
J [ ] L [ ]
1
M [ ] L [ ]
1, T
=
K
s
M
s
+ ( )M + [ ] u { } 0 =
1
s
+ ( ) K
s
M + [ ]
M
LL
T

s
+
---------------- u { } 0 =
L [ ]
1
w { } L [ ]
T
u { } =
J [ ] J [ ] and w { } w { } = =
i
2 APPENDIX A
automatic methods, AHOU and AGIV, initially use the standard methods. In the first step
of the method, if the mass matrix is not well-conditioned for decomposition, the method
shifts to the corresponding modified method. The modified methods are more expensive
and introduce numerical noise due to the shift, but they resolve most of the numerical
problems of the ill-conditioned mass matrix.
Operations involved in the computation of the natural frequencies and mode shapes is
proportional to the number of degrees-of-freedom ( ). The time-consuming operations for
the tridiagonal methods are mass matrix decomposition and reduction to standard form
( for a full mass matrix, negligible for a diagonal matrix) tridiagonalization (
for HOU,

for GIV), QR root iteration ( ), inverse iteration for vectors ( per
vector), and inverse transformation of vectors ( per vector for HOU, for GIV). As
becomes large (greater than 100, perhaps), terms in dominate.
The Inverse Power Methods (INV, SINV)
The inverse power method with shifts is particularly effective when the mass and stiffness
matrices are sparse and when only a few of the eigenvalues are required. In MSC.Nastran,
the method is used as a stand-alone method to find all of the eigenvalues within a user-
specified domain.
If we define the eigenvalue problem as:
Eq. A-20
Let:
Eq. A-21
where is called the shift point.
The iteration algorithm is:
Eq. A-22
Eq. A-23
where is equal to the value of the element of with the largest absolute value.
is then placed on the right side of Eq. A-22 and iteration continues. It is easy to prove that
converges to , the shifted eigenvalue nearest to the shift point, and that
converges to the corresponding eigenvector of Eq. A-20. The following observations are
pertinent.
N
5 2 N
3
2 3 N
3
4 3 N
3
9 N
2
N
2
N
2
2 N
2
N N
3
K M [ ] u { } 0 =

o
+ =
o
K
o
M [ ] w
n
{ } M [ ] u
n 1
{ } =
u
n
{ }
1
C
n
------ w
n
{ } =
C
n
w
n
{ } u
n
{ }
1 C
n

1
u
n
{ }

1. A triangular decomposition of the matrix is required in order to
evaluate from Eq. A-22. The effort required to perform the triangular
decomposition is greatly reduced if is a narrow band matrix. Structural
analysis is characterized by narrowly banded stiffness and mass matrices. From
Sturms theorem, the number of negative terms on the factor diagonal is exactly
equal to the number of roots below . This number is printed out at every shift
and is identified as the Sturm sequence number.
2. It is unnecessary for to be nonsingular, so that rigid body modes cause no
special difficulty. Similarly, null rows in cause no problems. If they are
removed by static condensation, they will increase the cost of the solution
unnecessarily. This occurs primarily due to the increase in bandwidth inherent
in static condensation
3. The shift point may be changed (at the cost of an additional triangular
decomposition) in order to improve the rate of convergence toward a particular
eigenvalue or to improve the accuracy and convergence rate after several roots
have been extracted from a given shift point. The next shift point is computed
from the residual terms remaining after computing the present root by a root
estimation algorithm (See theMSC.Nastran Numerical Methods Users Guide for
details.)
4. can be placed to obtain the eigenvalues within a desired frequency band and
not just those that have the smallest absolute value.
5. The Sturm sequence data can be used to ensure that all eigenvalues within the
requested frequency band have been found. This can be done manually with
DIAG 16 for the INV method or automatically with the SINV method (see the
MSC.Nastran Numerical Methods Users Guide for details). In the SINV method, the
requested band is broken into subregions, and the Sturm number is determined
at each boundary of the subregion. If the root estimation algorithm predicts that
the next root is out of the subregion but the Sturm sequence data states that more
roots remain in the subregion, a bisection method is used to iterate to the missing
roots. All roots for a subregion must be found before the method is allowed to
enter the next subregion. The INV method tends to compute shifts that wander
more about the requested frequency band and computes more roots outside the
band than the SINV method does.
In the practical application of the tracking methods, most of the numerical efficiency issues
(trial eigenvectors, convergence criteria, etc.) are automatically handled by MSC.Nastran.
The shift points must be selected so that the numerical algorithms can operate in an
optimum manner. Generally, this is handled by MSC.Nastran, but the user can assist
MSC.Nastran through a careful choice of frequency range.
If the SUPORT entry is used, the rigid body modes calculated are used directly as the first
set of eigenvectors with their eigenvalues set to zero. Iteration then proceeds for the
nonzero roots. Use of the SUPORT feature reduces costs of the INV and SINV methods
considerably besides giving the aesthetic benefits of exact zero-frequency rigid body
modes. The SUPORT entry is recommended for use with these methods.
K
o
M [ ]
w
n
{ }
K
o
M [ ]
o
K [ ]
M [ ]
o
2 APPENDIX A
The Shifted Block Lanczos Method
The Lanczos method (LAN) available in MSC.Nastran is an extremely reliable and accurate
method for solving the large sparse eigenvalue problems typical of many structures.
A shifted block Lanczos algorithm is implemented in MSC.Nastran. A summary of that
method is given here with special emphasis on the effect of user inputs.
For the standard eigenvalue problem written in the form:
Eq. A-24
the Lanczos method computes a sequence of vectors called Lanczos vectors and scalars
. The basic recurrence relation for the j-th step of the Lanczos algorithm can be written
as follows:
Eq. A-25
Eq. A-26
Eq. A-27
Eq. A-28
The first j-steps of the method can be combined and written in matrix form:
Eq. A-29
where = unit vector with the j-th component equal to one and all other components
equal to zero
Eq. A-30
Eq. A-31
The eigenvalues of the tridiagonal matrix are called Ritz values. Some of these Ritz
values are good approximations to eigenvalues of . The corresponding Ritz vectors
are computed from the eigenvectors of and the Lanczos vectors. An eigenpair is
A [ ] x { } x { } =
q
Q
j
, A
r
j 1 +
{ } A [ ] q
j
{ }
j
q
j
{ }
j
q
j 1
{ } =
j
q
i
{ }
T
A [ ] q
j
{ } =
j 1 +
r
j 1 +
=
q
j 1 +
{ }
r
j 1 +
{ }
B
j 1 +
-------------------- =
A [ ] Q
j
[ ] Q [ ] T
j
[ ]
j 1 +
q
j 1 +
{ } e
j
{ }
T
=
e
j
{ }
Q
j
{ } q
1
{ }, q
2
{ }, ..., q
j
{ } ( ) with Q
j
T
[ ] Q
j
[ ] I [ ] = =
T
j
[ ]
1

2

2

2

2

3

3

4

j

j

j
=
T
j
[ ]
A [ ] x
i
( )
T
j
[ ]

defined as an eigenvalue and its corresponding eigenvector . Let be an
eigenpair of the tridiagonal matrix . Then an approximate eigenpair for the original
problem is where:
Eq. A-32
Eq. A-33
As implied by the formulation of the algorithm, the method computes a tridiagonal matrix
and a set of Lanczos vectors that become larger with each iteration. The algorithm in exact
arithmetic can be terminated when or when sufficiently many eigenvalues of
have been computed. Lanczos also handles the effects of roundoff.
The algorithm that has been incorporated in MSC.Nastran is a shifted block version of the
basic Lanczos recurrence relation. The blocking approach to the Lanczos algorithm
(working with blocks of Lanczos vectors simultaneously) enables the method to handle
eigenvalues of high multiplicity including shifting in the algorithm which helps to speed
the convergence of the method.
In this method, the matrix is replaced by the sum of the stiffness matrix plus shifted mass
matrix, similar to the approach in the modified tridiagonal methods. This allows solution
of problems with rigid body modes and generally faster convergence. The shifts (trial
roots) are influenced by user input. The user is required to input either a frequency band
of interest, a desired number of roots, or a combination of this data.
As an example, suppose that both the lower bound (V1) and upper bound (V2) of the
frequency band are defined by the user as well as the number of roots desired (ND). The
main objective for the shift strategy is to encompass the region containing the eigenvalues
of interest in which all of the eigenvalues have been computed and the number of
eigenvalues has been verified with a Sturm sequence check.
The first shift is at V1 to determine the Sturm sequence number. The factor matrix is stored
for possible later use. If V1 is not defined by the user, the first shift is made at . The
second shift point is chosen at V2.
After a Lanczos iteration based on the selected shift and the computation of the
corresponding Ritz values, some values will be found to be acceptable eigenvalues while
other eigenpairs will fail convergence tests. A decision is made as to whether it is more
efficient to re-shift or continue with further iteration. The re-shift point is chosen from the
existing Ritz values (smaller than V2) or from the stored factor from the shift at V2. At this
point, the number of accepted eigenvalues and the Sturm sequence numbers at each end of
the region are known.
If iteration with the factor from V2 does not produce all of the roots desired, two more
types of shift will be used. If it appears that the missing roots are near the lowest shift, an
approximate first mode eigenvalue is used. This is either a nonzero SHFSCL field from the
EIGRL entry or the a similar shift term used in the modified tridiagonal methods. If the
missing roots are elsewhere, bisection between prior shifts is used. If all requested
eigenpairs still cannot be found, a user information message is printed, and all accepted
s
i
( )
i k
{ }
ik
{ }, s
i
( )
T
j
[ ]
i
, x
i
{ } ( )
i

i
=
x
i
{ } Q
j
{ } s
i
{ } =
B
j 1 +
0 = A [ ]
p
A
s
2 APPENDIX A
eigenpairs are output. If some roots have been found, the module exits normally. A fatal
error is issued if no roots have been found. If accepted Ritz values are found outside the
requested region, they are discarded.
The strategy for selecting block size is to first find the minimum of N/2, 2E, p and
MAXSET. N is the problem size, E the number of roots desired, p the number of Lanczos
vectors that will fit into memory, and MAXSET the user selection for block size (which has
a default of 7).
Another efficiency feature unique to the Lanczos method is selective orthogonalization of
the trial eigenvectors. All of those contained in the same block are made orthogonal to one
another. Tests are made, at several steps, to determine if further orthogonalization
between vectors in different blocks is required. This avoids the cost term cubic on E (the
number of eigenvectors calculated) which becomes the dominant term for GDR and the
SINV methods as E becomes large.
If the SUPORT entry is used, the matrix is merged with an identity matrix to form
raw rigid body vectors . These raw vectors are used as initial vectors. If the r-set has
been well-chosen, the rigid body modes will converge rapidly. If they have been chosen
poorly, the Lanczos iterations will continue until a good set of eigenvectors have been
found. The corresponding roots are set to the aesthetically pleasing value of zero rather
than to the small numerical zeros likely to be produced by the actual computations.
The Lanczos method is more forgiving than the other methods since it repairs the effects of
a poor r-set selection. However, the repeatability of rigid body modes computed by
Lanczos with the SUPORT entry is not the same as with other extraction methods. Rigid
body modes computed using the SUPORT entry and a non-Lanczos method will produce
identical modes, irrespective of the method. If the rigid body modes produced by the
Lanczos method with SUPORT entries are compared with those produced by any of the
other methods, the Lanczos modes will be a proper linear combination of the modes from
the other methods and may be markedly different in appearance. Furthermore, if small
changes are made to the stiffness matrix, the Lanczos modes again may be markedly
different, although they are again a proper linear combination of the set produced by other
methods. This lack of repeatability makes the Lanczos method unusable with the scaled
response spectra capability of MSC.Nastran.
The cost for the LAN method is dominated by the decomposition, inverse iteration, and
orthogonalization operations at the size of the input matrices N. The number of roots per
decomposition varies by problem and is difficult to determine a priori.
User Interface for Real Eigenvalue Analysis
The EIGR and EIGRL Bulk Data entries are the heart of the user interface because they
define a method and select parameters that control the eigenvalue extraction procedure.
The EIGRL entry is used for the LANCZOS method, and the EIGR entries are used for all
of the other methods.
D
i r
[ ] I
rr
[ ]
D
ar
[ ]

The data used for the EIGR entry depend on the method selected. The basic data required
is related to the eigenvalue extraction method chosen and to the frequency range or
number of required roots. The basic format of the Bulk Data entry is as follows:
The SID (Field 2) is the Set ID associated with this EIGR entry. A particular EIGR entry is
activated in an analysis using the METHOD = SID entry in the Case Control Section. If the
Set ID on the METHOD and EIGR match, the eigenvalues will be extracted using the
method referenced on that entry.
The METHOD Entry (Field 3) selects the eigenvalue method from the following list:
1. INV Inverse Power
2. SINV Sturm Modified Inverse Power
3. GIV Givens Method
4. HOU Householder Method
5. MGIV Modified Givens Method
6. MHOU Modified Householder Method
7. AGIV Automatic Givens Method
8. AHOU Automatic Householder
The F1 Entry (Field 4) is used to specify the lowest frequency of interest in the eigenvalue
extraction.
The F2 Entry (Field 5) is used to specify the highest frequency of interest in the eigenvalue
extraction.
The NE entry (Field 6) is used by the INV method only. It defines the estimated number of
roots in the range. A good estimate will result in a more efficient solution.
The ND entry (Field 7) is used to specify the desired number of roots in the range of
interest.
The NORM (Field 2 continuation) entry on the continuation card is used to specify the
method of eigenvector normalization. The choices are:
1. MASS Mass normalization (Default if used continuation entry not required)
2. MAX Normalization to maximum A-set component
3. POINT Normalization to user-defined DOF
The POINT entry (Field 3 continuation) is used to specify the point of interest from the
analysis set for POINT normalization only.
The CID entry (Field 4 continuation) is used to specify the component of displacement for
POINT normalization only.
1 2 3 4 5 6 7 8 9 10
EIGR SID METHOD F1 F2 NE ND +BC
+BC NORM GID CID
2 APPENDIX A
There is an interrelationship between the F1, F2, and ND field on the EIGR entry.
The rules for METHOD = GIV, HOU, MGIV, MHOU, AGIV and AHOU are identical. If
any of these methods are selected, MSC.Nastran will find all of the eigenvalues but will
compute only the eigenvectors specified by F1 and F2 or those specified by ND (the
desired number). F1 and F2 specify the lower and upper bounds of the frequency range
in which eigenvectors are computed, and ND specifies the number of eigenvectors, starting
with the lowest (or the first rigid body mode, if present). If F1, F2, and ND entries are
present, ND takes precedence. Note that the default for F1 is 0.0.
The following examples demonstrate the use of the EIGR data entry.
In the first example (SID = 1), the automatic Householder method is selected with 10 roots
requested. Since the default MASS eigenvector normalization is requested, no
continuation entry is needed.
For the second example (SID = 2), the same method is requested, but all the roots below 100
cycles per unit time are requested with MAX vector normalization.
Table A-1 EIGR Input Parameters
When METHOD = INV
or SINV
When METHOD = GIV, MGIV,
HOU, or MHOU
F1,F2 Frequency range of interest
( ). F1 must be input.
If METHOD = SINV and ND is
blank, then F2 must be input.
Frequency range of interest ( ;
F1 < F2. (If ND is blank, F1 and F2 are
ignored. If ND is blank, eigenvectors are
found whosenatural frequencies lie in the
range between F1and F2.
NE Estimate of number of roots in
range (Required for
METHOD = INV (Integer > 0).
Not used by SINV method.
Not used.
ND Desired number of roots (Default
is 3*NE, INV only) (Integer > 0).
If this field is blank and
METHOD = SINV, then all roots
between F1 and F2 are searched
and the limit is 600 roots.
Desired number of eigenvectors
( ). If ND is zero, the number of
eigenvectors is determined from F1 and F2
(Default = 0). If all three are blank, then ND
is automatically set to one more than the
number of degrees-of-freedom listed on
SUPORT entries.
1 2 3 4 5 6 7 8 9 10
EIGR 1 AHOU 10
EIGR 2 AHOU 100.0 +2
+2 MAX
Real 0.0
Real 0.0
Integer 0

For the third example (SID = 3), the inverse iteration method with Sturm sequence checks
is requested for the first six roots in the range specified. The POINT normalization is
requested for grid point 32 in the R1 direction.
If METHOD = SINV, the values of F1, F2, and ND determine both the number of
eigenvalues and the eigenvectors that will be computed. These entries also provide hints
to help MSC.Nastran find the eigenvalues. F1 and F2 specify the frequency range of
interest within which MSC.Nastran will search for modes. MSC.Nastran will attempt to
find all of the modes in the range F1, F2 or the number specified by ND, whichever is less.
If searching stops because ND modes are found, there is no guarantee that they will be the
lowest eigenvalues. Iterative methods only guarantee the solution for modes nearest the
shift point. If ND modes are not found in the range of interest, SINV will usually find one
mode outside the range F1, F2 (or possibly more) before stopping the search.
The inverse power method is particularly efficient when only a small number of
eigenvalues and eigenvectors are wanted. Very often only the lowest mode is of interest.
The following example illustrates an EIGR entry which will extract only the lowest nonzero
eigenvalue.
It is assumed in the example that the frequency of the lowest mode is greater than 0.01
cycles per unit time. The program will find one eigenvalue outside the range F1, F2 and
stop the search. The eigenvalue found is guaranteed to be the lowest non zero eigenvalue
(or a member of the lowest closely spaced cluster of eigenvalues in cases with
pathologically close roots) provided that there are no negative eigenvalues and that the
SUPORT entry has been used to specify the correct number of zero eigenvalues. The NE
entry (field 6) is used only by the INV method.
The fields of the EIGRL entry are designed to select and set parameters for the Lanczos
method as well as request diagnostics.
The V1 field defines the lower frequency bound, the V2 field defines the upper frequency
bound, and the ND field defines the number of eigenvalues and eigenvectors desired in the
region. The V1 and V2 are expressed in units of cycles per unit time. Examples of the
results of using explicit or default values for the V1, V2, and ND fields are shown in Table
A-2. The defaults on the EIGRL entry are designed to provide the minimum number of
roots in cases where the input is ambiguous.
1 2 3 4 5 6 7 8 9 10
EIGR 3 SINV 0.0 100.0 6 +3
+3 POINT 32 4
EIGR SID METHOD F1 F2 NE ND
EIGR 13 INV 0.0 .011 1
EIGRL SID V1 V2 ND MSGL VL MAXSET SHFSCL
EIGRL 1 0.1 3.2 10
2 APPENDIX A
The MSGLVL field of the EIGRL entry is used to control the amount of diagnostic output
provided by Lanczos. The default value of 0 produces no diagnostic output. The values 1,
2, or 3 provide more output with the higher values providing increasingly more output.
Typically, the default value is sufficient. In some cases, higher diagnostics levels may be
in order to help resolve difficulties with special modeling problems.
The MAXSET field is used to control the block size, the number of appropriate eigenvectors
computed in the outermost iteration loop. The default value of 7 is recommended for most
applications. There may be special cases where a larger value may result in quicker
convergence of highly multiple roots or a lower value may result in more efficiency when
the structure is lightly coupled. However, the default value has been chosen after
reviewing results from a wide range of problem types on several different computer types
with the goal of minimizing the sum of CPU and I/O cost.
One common occurrence is for the block size to be reset by MSC.Nastran during the run
because there is insufficient memory for a block size of 7. Computational efficiency tends
to degrade as the block size decreases. It is, therefore, important to examine the
EIGENVALUE ANALYSIS SUMMARY output to determine that the program has
sufficient memory to use an efficient block size. A smaller block size may be more efficient
when only a few roots are requested. The minimum recommended block size is 2.
The SHFSCL field allows a user-designated shift to be used when all other shifting
strategies fail. This entry is only used under special circumstances. However, when used,
its value should be set to the expected first natural frequency.
The Lanczos method normalizes the computed eigenvectors using the MASS method. The
other normalization methods are not available.
Solution Control For Normal Modes
When used as an independent solution, normal modes analysis is available in SOL 103 of
the Structured Solution Sequences and SOL 3 and SOL 63 in the Unstructured Solution
Sequences. All eigenvalue extraction methods are available in all of these solutions, with
Table A-2 Model Outputs from EIGRL Input Options
Case V1 V2 ND Number and Type of Roots Found
1 V1 V2 ND Lowest ND or all in range; whichever is smaller
2 V1 V2 All in range
3 V1 ND Lowest ND in range [V1, +00]
4 V1 Lowest root in range [V1, +00]
5 ND Lowest ND roots in [-00, +00]
6 Lowest root
7 V2 ND Lowest ND roots below V2
8 V2 All below V2

one exception. If the Lanczos method is to be used in SOL 3, Rigid Format Alter RF3D83
must also be included. The Executive Control Section can also contain diagnostic DIAG 16
which will print the iteration information used in the INV or SINV methods.
One of the most important entries in the Case Control Section is the METHOD entry. This
entry is required. The Set ID referenced by the METHOD entry refers to the Set ID of an
EIGR entry or an EIGRL entry in the Bulk Data.
When a modal extraction is performed, the MSC.Nastran output file contains various
diagnostic messages and an Eigenvalue Table. Optional grid and element output are
available using standard Case Control output requests. These requests are summarized in
Table A-3.
Table A-3 Eigenvalue Extraction Output Requests
Grid Output
DISPLACEMENT
(or VECTOR)
Requests the eigenvector (mode shape) for a set of grid points
GPFORCE Requests the Grid Point Force Balance Table to be computed for
each mode
GPSTRESS Requests grid point stresses to be computed for a set of grid
points. This request must be accompanied by the ELSTRESS Case
Control request and the definition of Stress Surfaces and/or
Stress Volumes in the OUTPUT(POST) section of the Case
Control. This request also requires the use of Rigid Format Alter
RF3D81 when used in SOL 3.
SPCFORCE Requests Forces of Single Point Constraint to be computed for a
set of grid points for each mode.
Element Output
ELSTRESS
(or STRESS)
Requests the computation of modal stresses for a set of elements
for each mode
ESE Requests the computation of modal element strain energies for a
set of elements for each mode
ELFORCE
(or FORCE)
Requests the computation of modal element forces for a set of
elements for each mode
STRAIN Requests the computation of the modal element strains for a set of
elements
MODES A special Case Control request which permits selective output
requests to be processed on selective modes
2 APPENDIX A
In addition to Bulk Data entries required to define the structural model, the only required
Bulk Data entry is the eigenvalue selection entry, EIGR or EIGRL. The EIGR entry is used
to select the modal extraction parameters for the Givens, Householder, Modified Givens,
Modified Householder, Automatic Givens, and Automatic Householder methods. The
EIGRL entry is used to select the modal extraction parameters for the Lanczos method.

APPENDIX
B
SSSALTER Library For MSC.Nastran
I Overview
I SSSALTER Library for MSC.Nastran Version 69.1
I Alter to Perform Rigid Body Checks, Kinetic Energy Calculation,
and Modal Effective Weight Calculation
I References

B.1 Overview
The most current (MSC.Nastran V69.1)

SSSALTER filenames and their full descriptions are
maintained on the MSC.Software Corporate Web page at
http://www.mscsoftware.com
The SSSALTER files are delivered with MSC.Nastran on the MSC.Nastran delivery media.
MSC.Nastran clients may also request specific SSSALTERs from their authorized MSC
representative.
2 APPENDIX B
B.2 SSSALTER Library for MSC.Nastran Version 69.1
Attached is a brief description of each DMAP delivered with the SSSALTER library in
Version 69.1, some of which may be of interest to dynamics engineers.
Each of the DMAP alters was tested for a number of possible applications, but they have
not undergone the extensive testing required to be included in MSC.Nastran. It is possible
that your application may use a path through the DMAP alter that was not tested, so
always review your results carefully. Unlike the RFALTER directory, these DMAP alters
were developed independently and it is likely that a conflict will occur if more than one
alter is used in a run.
addstata.v691
Adds a static solution onto a transient solution. This is often necessary for structures that
have an existing (constant) static loading in addition to the transient load.
alter1ga.v691, alter2ga.v691, alter9ga.v691
These DMAP alters are used to create a reduced model of a component using the
Craig-Bampton method of component modal synthesis. A special feature of these files is
that they create output transformation matrices (OTM) for displacements, forces, and
stresses. They also allow for assembly plots that include a skeleton for the component.
This skeleton is based on grid points selected during the processing of the component. The
DMAP alters are:
alter1g.v691 (SOL 103) Creates reduced models (often referred to as Craig-Bampton
models) of components using superelements. Includes creation of transformation matrices
for output quantities.
alter2g.v691 (SOL 103) Reads the matrices created by alter1g into MSC.Nastran and stores
them as a superelement.
alter9ga.v691 (SOLs 111 and 112) Calculates the boundary forces for a superelement and
calculates the results using transformation matrices (includes sebloada.v691).
appenda.v691
Appends eigenvalue solutions.
checka.v691
Performs mass and stiffness checks in preparation for analysis. Calculates assembly weight
when using superelements. In modal solutions, calculates modal effective weight and
kinetic energy. If superelements are used, provides the total fractions of effective weight
and kinetic energy for each superelement and assembly. Will force Lanczos method to treat
a SUPORT entry like the other eigenvalue solution methods.

cova.v691
Computes the variance and standard deviation in response given the user-specified
variance in design variables. Calculations are performed by postprocessing design
sensitivity results. The following is computed:
where:
cygyroa.v691
Adds gyroscopic terms to cyclic symmetry analyses.
ddama.v691
Performs dynamic analysis of shipboard equipment using the dynamic design analysis
method (DDAM) for response spectra analysis.
delmodea.v691
Allows the user to delete individual modes from the set of system (residual) level modes.
dmigtrna.v691
Transforms the DMIG matrix selected by the K2GG Case Control command from the
coordinate system listed on PARAM,XCID to the basic coordinate system. It will then
punch this transformed DMIG matrix into the punch file.
dtranra.v691
Allows NOAP restarts in the Structured Solution Sequences.
evdsa.v691, evds103a.v691, evds200a.v691
Computes eigenvector derivatives in design sensitivity analysis.
gdrlana.v691
Incorporates Lanczos into GDR in order to replace the inverse iteration method (IIM).
genela.v691
Calculates the forces resulting from a GENEL element in SOL 112.
glforcea.v691
Calculates superelement interface loads.
COVR = Response variance
S = Sensitivity coefficient matrix
COVD = Design variable variance
COVR S
t ( )
COVD S =
2 APPENDIX B
iter101.v691, iter106.v691, iter108.v691, iter111.v691, iter24.v691, iter26.v691,
iter30.v691, iter66.v691
Use the iterative solver instead of the direct solver in SOL 101,
106, 108, 111, 24, 26, 30, and 66, respectively.
lmt01.v691, lmt03.v691, lmt07.v691, lmt08.v691, lmt09.v691, lmt10.v691, lmt11.v691,
lmt12.v691
Use the Lagrange Multiplier technique instead of the conventional method for SOL 101,
103, 107, 108, 109, 110, 111, and 112.
loop8a.v691
Allows for different FREQ requests in separate subcases in direct frequency response.
mfreqea.v691
Calculates the contribution of each mode to the strain energy. These values will help
identify which modes contribute to the solution.
mica.v691
Allows the use of initial conditions in a modal transient analysis. This DMAP alter imposes
user-specified displacement and velocity initial conditions in a modal transient solution.
In order to get a match on the initial conditions, the DMAP appends one or two vectors
onto the modes. These vectors represent the amount of the initial conditions which the
modes are not capable of representing. (They are automatically orthogonalized to the
modes.)
modevala.v691
Determines if the modes obtained are capable of representing the solution to a set of
applied loadings. A common practice in modal solutions is to assume that if the modes are
capable of representing the static solution to an applied loading, they contain sufficient
information to represent the high frequency response of the structure to a dynamic
application of the loading. This DMAP alter provides that information without having to
perform a dynamic solution.
mtacca.v691
Allows use of modal transient analysis to solve for a structure which has non-zero initial
accelerations. This DMAP alter assumes that the initial condition is a steady-state
acceleration based on the initial loading. The alter performs a static solution (in modal
coordinates) to the initial loading, subtracts the initial loading from all applied loadings,
solves the transient analysis, then adds the initial static solution to the results.
mtranea.v691
To calculate the contribution of each mode to the strain energy and kinetic energy of the
solution. These values will help identify which modes contribute to the solution.

mtranra.v691
Allows NOAP restarts in the Structured Solution Sequences (SOL 112).
nlgyroa.v691
Adds gyroscopic terms to nonlinear analysis.
norma.v691
Normalizes the output eigenvectors to a maximum of 1.0.
norma.v691
Normalizes the output eigenvectors to a maximum of 1.0.
If a reduction method is used, the eigenvectors are scaled based on the reduced solution
set. This DMAP alter scales the eigenvectors based on the largest value in the g-set
(excluding the q-set) being set to 1.0.
ortx7a.v691, ortx10a.v691
Scale comlex eigenvectors to produce unit generalized mass matrix.
pchdispa.v691
Creates DMIG entries containing the solution from a static run.
postmaca.v691
Computes the cross-orthogonality between test modes and modes of an MSC.Nastran
model.
Two cross-orthogonality checks are made: modal assurance criterion (MAC) and mass
orthogonality (ORTHOA).
premaca.v691
Computes the cross-orthogonality between modes of a full model and modes of a reduced
model. Three cross-orthogonality checks are made: modal assurance criterion (MAC) and
two types of mass orthogonality (ORTHOA and CHECKIT).
propa.v691, propf.for (a FORTRAN program)
Provides the propeller aerodynamic coefficients and wing/nacelle aerodynamic
interference loads for aircraft flutter analysis.
raleigh.v691
Includes raleigh damping in SOL 108.
resflexa.v691
Appends static residual vectors onto the eigenvectors in dynamic solutions for improved
accuracy. It works at the superelement and the system level. Similar to the
mode-acceleration and residual flexibility methods.
2 APPENDIX B
rflag.v691
Uses the Lagrange Multiplier technique instead of the conventional method. It
incorporates features in lmt01, lmt03, lmt07, lmt08, lmt09, lmt10, lmt11, and lmt12.
rflagb.v691
Uses the Lagrange Multiplier technique for enforced motion, but all other constraints are
processed by conventional constraint variable elimination techniques.
ridgyroa.v691
Adds gyroscopic terms to dynamic analysis using the structured solution sequences.
These solution sequences include direct and modal frequency response, direct and modal
transient response, and direct and modal complex eigenvalue analyses. The alter does not
support real eigenvalue analysis.
sebloada.v691
Calculates dynamic superelement interface loads.
sedampa.v691
Allows modal damping on superelement component modes.
segyroa.v691
Adds gyroscopic terms to superelement analysis.
solchk.v691, dcchk.v691, dfchk.v691, dtchk.v691, mfchk.v691, mcchk.v691,
mochk.v691, mtchk.v691, sechk.v691
This alter extends the same theory for load epsilon checks long used in static analysis to
dynamic analysis. The load residual
is computed. The load epsilon EPSM is computed for the first ten solutions.
spc101a.v691
Organizes the calculated SPCForces by the type of constraint placed upon a DOF; e.g., was
it a user-induced constraint such as the PS field on the GRID or an SPC card, or was it a
DOF constrained by the AUTOSPC function. User has further option of breaking the
user-induced constraints into those DOF constrained by the GRID PS field and those on an
SPC set. Answers may be independently filtered to remove smaller SPCForces from the
printouts. Resultant vectors for SPCForce contributions are also calculated.
spc24a.v691
Organizes the calculated SPCForces by the type of constraint placed upon a DOF; e.g.,
whether it was a user-induced constraint such as the PS field on the GRID or an SPC card,
or whether it was a DOF constrained by the AUTOSPC function. The user has the further
option of breaking the user-induced constraints into those DOF constrained by the GRID
R P K DISP B VELO M ACCEL =

PS field and those on an SPC set. Answers may be independently filtered to remove
smaller SPCForces from the printouts. Resultant vectors for SPCForce contributions are
also calculated.
zfreq.v691
Allows for modeling of frequency-dependent springs and dampers and multiple subcases
calling out different FREQi entries in modal frequency response analysis.
2 APPENDIX B
B.3 Alter to Perform Rigid Body Checks, Kinetic Energy
Calculation, and Modal Effective Weight Calculation
checka.v691
Performs mass and stiffness checks in preparation for analysis. Calculates assembly
weight when using superelements. In modal solutions, calculates modal effective weight
and kinetic energy. If superelements are used, provides the total fractions of effective
weight and kinetic energy for each superelement and assembly. Will force Lanczos
method to treat a SUPORT entry like the other eigenvalue solution methods.
SOL 101, 103 (possibly others).
Input
Executive Control
Include checka.v691.
The following parameters affect the operation of this DMAP alter:
GRDPNT (Default = -1). Used by grid point weight generator. This DMAP also uses this
grid point for assembly grid point weight generator and as the reference location for
rigid-body checks.
CHKSTIF (Default = 1). A positive value causes the DMAP to perform checks on the
stiffness as follows:
1 = check KGG, KNN, and KAA.
2 = check KGG and KNN.
3 = check KGG only.
The DMAP performs rigid-body checks on the stiffness matrices by generating geometric
rigid-body vectors about GRDPNT and multiplying the stiffness matrix by the vectors.
CHECKTOL (Default = 1.e-05) Acceptable tolerance on stiffness matrix checks. If the
strain energy for all checks of a stiffness matrix is less than this value, a message is printed
stating so. If not, then the strain energy and the resulting (normalized) reactions are
printed.
CHKMASS (Default = 1) Similar to CHKSTIF only for the mass matrices. Allows
verification that no mass was removed during processing.
KEPRT (Default = 1) Applicable only in modal solutions. A positive value of this
parameter results in the calculation of kinetic energy fractions for the modes (assuming
that modes are normalized to unit mass).
EFWGT (Default = 2) Applicable only in modal solutions. A positive value results in the
calculation of the modal effective mass for each mode (once again, assuming modes are
normalized to unit mass).

1 = calculates the effective weight and print for each mode.
2 = calculates the effective weight and fraction of the total for each mode and prints the
fraction (requires CHKMASSu0).
LANSUP (default = 1) forces Lanczos method to treat a SUPORT entry in the same
manner as the other eigenvalue solution methods.
1 = Lanczos will treat a SUPORT like the other methods
< >1 = > No change from the current way Lanczos handles a SUPORT
Various matrices representing the calculated quantities.
The alter does what is requested by the user. That is, if kinetic energy is requested in a
static solution, the DMAP is not smart enough to realize that the solution is static.
Therefore, it calculates and prints an incorrect quantity using the static vectors.
check1.dat (SOL 103) Run to perform stiffness and mass checks plus effective weight and
kinetic energy calculations.
2 APPENDIX B
B.4 References
JPL, MSC.Nastran Model Checkout, 1986 MSC World Users Conference Proceedings,
Paper No. 9, March 1986.
T. Rose, Using Superelements to Identify the Dynamic Properties of a Structure, 1988
MSC World Users Conference Proceedings, Vol. 1, Paper No. 41, March 1988.

APPENDIX
C
Set Notations
I Degree-of-Freedom Sets
I Multipoint Constraints
I Single Point Constraints
I Rigid Body Supports
I Sets for Dynamic Reduction
I Sets for Aerodynamics
I Rules of Sets for Undefined Degrees-of-Freedom
I Output Selection via Set Specification

C.1 Degree-of-Freedom Sets
Structural matrices are initially assembled in terms of all structural points, which excludes
only the extra points introduced for dynamic analysis. These matrices are generated with
six degrees-of-freedom for each geometric grid point and a single degree-of-freedom for
each scalar point. These degrees-of-freedom are partitioned into sets, based on user inputs.
The sets are used to successively eliminate variables during the solution process, as
described in General Operations on page 374 of the MSC.Nastran Reference Guide.
Degree-of-Freedom Set Definitions
Each degree-of-freedom is a member of one mutually exclusive "set". Set names have the
following definitions:
Each degree-of-freedom is also a member of one or more combined sets called "supersets."
Supersets have the following definitions:
Set Name Meaning
m degrees-of-freedom eliminated by multipoint constraints.
sb*
* Strictly speaking, sb and sg are not exclusive with respect to one another. Degrees-of-
freedom may exist in both sets simultaneously. Since these sets are not used explicitly in
the solution sequences, this need not concern the user. However, those who use these sets
in their own DMAPs should avoid redundant specifications when using these sets for
partitioning or merging operations. That is, a degree-of-freedom should not be specified
on both a PS field of a GRID entry (sg set) and on a selected SPC entry (sb set). Redundant
specifications will cause UFM 2120 in the VEC module and behavior listed in the
MSC.Nastran 2005 DMAP Programmers Guide for the UPARTN module. These sets are
exclusive, however, from other mutually exclusive sets.
deegrees-of-freedom eliminated by single-point constraints that are
included in boundary condition changes and by the AUTOSPC feature.
sg* degrees-of-freedom eliminated by single-point constraints that are
specified on the PS field on GRID Bulk Data entries.
o degrees-of-freedom omitted by structural matrix partitioning.
q generalized degrees-of-freedom for dynamic reduction or component
mode synthesis.
r reference degrees-of-freedom used to determine free body motion.
c degrees-of-freedom which are free during component mode synthesis or
dynamic reduction.
b degrees-of-freedom fixed for component mode analysis or dynamic
reduction.
e extra degrees-of-freedom introduced in dynamic analysis.
sa permanently constrained aerodynamic degrees-of-freedom.
k aerodynamic degrees-of-freedom.
2 APPENDIX C
Set Notations
The a-set and o-set are created in the following ways:
Set Name Meaning (+ Indicates Union of Two Sets)
All degrees-of-freedom eliminated by single-point constraints.
The structural coordinates remaining after the reference coordinates
are removed (degrees-of-freedom left over).
The total set of physical boundary degrees-of-freedom for
superelements.
The set assembled in superelement analysis.
The set used in dynamic analysis by the direct method.
Unconstrained (free) structural degrees-of-freedom.
Free structural degrees-of-freedom plus extra degrees-of-freedom.
All structural degrees-of-freedom not constrained by multipoint
constraints.
All structural degrees-of-freedom not constrained by multipoint
constraints plus extra degrees-of-freedom.
All structural (grid) degrees-of-freedom including scalar degrees-of-
freedom.
All physical degrees-of-freedom.
Physical and constrained aerodynamic degrees-of-freedom.
Physical set for aerodynamics.
Statically independent set minus the statically determinate supports
( ).
The set free to vibrate in dynamic reduction and component mode
synthesis.
m
sb
sg
o
q
r
c
b
e
sa
k
s
t
a
d
f
fe
n
ne
g
p
ps
pa
l
sets
supersets
s sb sg + =
l b c + =
t l r + =
a t q + =
d a e + =
f a o + =
fe f e + =
n f s + =
ne n e + =
g n m + =
p g e + =
ps p sa + =
pa ps k + =
f r o l + =
fr f q r =
v o c r + + =

1. If only OMITi entries are present, then the o-set consists of degrees-of-freedom
listed explicitly on OMITi entries. The remaining f-set degrees-of-freedom are
placed in the b-set which is a subset of the a-set.
2. If ASETi or QSETi entries are present, then the a-set consists of all degrees-of-
freedom listed on ASETi entries and any entries listing its subsets, such as QSETi,
SUPORTi, CSETi, and BSETi entries. Any OMITi entries are redundant. The
remaining f-set degrees-of-freedom are placed in the o-set.
3. If there are no ASETi, QSETi, or OMITi entries present but there are SUPORTi,
BSETi, or CSETi entries present then the entire f-set is placed in the a-set and the
o-set is not created.
4. There must be at least one explicity ASETi, QSETi, or OMITi entry for the o-set to
exist, even if the ASETi, QSETi, or OMITi entry is redundant.
In dynamic analysis, additional vector sets are obtained by a modal transformation derived
from real eigenvalue analysis of the a-set. These sets are as follows:
One vector set is defined that combines physical and modal degrees-of-freedom:
The membership of each degree of freedom can be printed by use of the Bulk Data entries
PARAM,USETPRT and PARAM,USETSEL.
= rigid body (zero frequency) modal degrees-of-freedom
= finite frequency modal degrees-of-freedom
=
, the set of all modal degrees-of-freedom
=
, the set of all modal degrees-of-freedom
i

o

f
+
u
h

i
u
e
+
2 APPENDIX C
Set Notations
Degree of Freedom Set Bulk Data Entries
Degrees-of-freedom are placed in sets as specified by the user on the following Bulk Data
entries:
In superelement analysis, the appropriate entry names are preceded by the letters SE, and
have a field reserved for the superelement identification number. This identification is
used because a boundary (exterior) grid point may be in one mutually exclusive set in one
superelement and in a different set in the adjoining superelement. The SE-type entries are
internally translated to the following types of entry for the referenced superelement:
Name Bulk Data Entry Name
m MPC, MPCADD, MPCAX, POlNTAX, RBAR, RBE1, RBE2, RBE3,
RROD, RSPLINE, RTRPLT, GMBC, GMSPC*
* Placed in set only if constraints are not specified in the basic coordinate system.
sb SPC, SPC1, SPCADD, SPCAX, FLSYM, GMSPC*, BNDGRID,
(PARAM,AUTOSPC,YES)
sg GRID, GRIDB, GRDSET (PS field)
o OMIT, OMIT1, OMlTAX, GRID (SEID field), SESET
q QSET,QSET1
r SUPORT, SUPORT1, SUPAX
c CSET,CSET1
b BSET, BSET1
e EPOINT
sa CAEROi
k CAEROi
a ASET, ASET1, Superelement exterior degrees-of-freedom, CSUPEXT
Entry Type Equivalent Type
SEQSETi QSETi
SESUP SUPORT
SECSETi CSETi
SEBSETi BSETi

C.2 Multipoint Constraints
Each multipoint constraint is described by a single equation that specifies a linear
relationship for two or more degrees-of-freedom. In static analysis, multiple sets of
multipoint constraints can be provided in the Bulk Data Section, with selections made at
execution time by using the subcase structure in Case Control as explained in Case
Control Section on page 11 of the MSC.Nastran Reference Guide. Multipoint constraints
are also discussed in Sections 3.5.1 of the MSC.Nastran Reference Manual and Section 5.4 of
The NASTRAN Theoretical Manual.
Multipoint constraints are defined on MPC, MPCADD, and MPCAX entries. MPC is the
basic entry for defining multipoint constraints. The first component specified on the entry
is the dependent degree-of-freedom, i.e., that degree-of-freedom that is removed from the
equations of motion. Dependent degrees-of-freedom may appear as independent terms in
other equations of the set; however, they may appear as dependent terms in only a single
equation. The MPCADD entry defines a union of multipoint constraints. The MPCAX
defines multipoint constraints in conical shell problems. Some uses of multipoint
constraints are:
To describe rigid elements and mechanisms such as levers, pulleys, and gear
trains. In this application, the degrees-of-freedom associated with the rigid
element that are in excess of those needed to describe rigid body motion are
eliminated with multipoint constraint equations. Treating very stiff members as
rigid elements eliminates the ill-conditioning associated with their treatment as
ordinary elastic elements.
To enforce zero motion in directions other than those corresponding with
components of the global coordinate system. In this case, the multipoint
constraint will involve only the degrees-of-freedom at a single grid point. The
constraint equation relates the displacement in the direction of zero motion to the
displacement components in the global system at the grid point.
To be used with scalar elements to generate nonstandard structural elements and
other special effects.
To describe parts of a structure by local vibration modes. This application is
treated in Section 14.1 of The NASTRAN Theoretical Manual. The general idea is
that the matrix of local eigenvectors represents a set of constraints relating
physical coordinates to modal coordinates.
In general, the user must provide the coefficients in the multipoint constraint equations.
However, several rigid elements have been introduced which will generate the MPC
equations for some applications:
RROD A pin-ended rod which is rigid in extension.
RBAR A rigid bending element with six degrees-of-freedom at each end.
RTRPLT A rigid triangular plate with six degrees-of-freedom at each
vertex.
2 APPENDIX C
Set Notations
The rigid elements will always meet equilibrium and continuity requirements, whereas
this is a user responsibility for MPC equations.
Multipoint forces of constraint may be output with MPCFORCE Case Control command in
SOLution Sequences 101 through 200.
RBE1 A rigid body connected to an arbitrary number of grid points. The
independent and dependent degrees-of-freedom can be arbitrarily
selected by the user.
RBE2 A rigid body connected to an arbitrary number of grid points. The
independent degrees-of-freedom are the six components of
motion at a single grid point. The dependent degrees-of-freedom
are specified at an arbitrary number of grid points.
RBE3 Defines the motion at a reference grid point as the weighted
average of the motions at a set of other grid points.
RSPLlNE Defines multipoint constraints for the interpolation of
displacements at grid points.

C.3 Single Point Constraints
A single point constraint (SPC) applies a fixed value to a translational or rotational
component at a geometric grid point or to a scalar point. Common uses of single point
constraints are to specify the boundary conditions of a structural model by fixing the
appropriate degrees-of-freedom and to eliminate unwanted degrees-of-freedom with zero
stiffness. Multiple sets of single point constraints can be provided in the Bulk Data Section,
with selections made at execution time by using the subcase structure in the Case Control
Section as explained in Case Control Section on page 11 of the MSC.Nastran Reference
Guide. This procedure is particularly useful in the solution of problems having one or more
planes of symmetry.
The elements connected to a grid point may not provide resistance to motion in certain
directions, causing the stiffness matrix to be singular. Single point constraints are used to
remove these degrees-of-freedom from the stiffness matrix. A typical example is a planar
structure composed of membrane and extensional elements. The translations normal to
the plane and all three rotational degrees-of-freedom must be constrained since the
corresponding stiffness matrix terms are all zero. If a grid point has a direction of zero
stiffness, the single point constraint needs not be exactly in that direction, but only needs
to have a component in that direction. This allows the use of single point constraints for
the removal of such singularities regardless of the orientation of the global coordinate
system. Although the displacements will depend on the direction of the constraint, the
internal forces will be unaffected.
One of the tasks performed by the Grid Point Singularity Processor (GPSP) is to examine
the stiffness matrix for singularities at the grid point level. Singularities remaining at this
level, following the application of the multipoint and single point constraints, are listed in
the Grid Point Singularity Table (GPST), which is automatically printed. The GPST lists all
singular degrees-of-freedom, in the global coordinate system, and the ratio of stiffness
between the softest and stiffest degree-of-freedom for the grid point. The user may request
that single point constraints be generated for all identified singularities by use of the Bulk
Data entry PARAM,AUTOSPC,YES.
Single point constraints are defined on SPC, SPC1, SPCADD and SPCAX entries. The SPC
entry is the most general way of specifying single point constraints. The SPC1 entry is a
less general entry that is more convenient when several grid points have the same
components constrained to a zero displacement. The SPCADD entry defines a union of
single point constraint sets specified with SPC or SPC1 entries. The SPCAX entry is used
only for specifying single point constraints in problems using conical shell elements.
Single point constraints can also be defined on the GRID entry. In this case, however, the
points are constrained for all subcases. The default value for enforced displacement on
points constrained on GRID entries is zero. The default value can be overridden at the
subcase level with SPC entries.
The printed output for single point forces of constraints can be requested in Case Control.
All nonzero forces are printed, whether they originate from SPCi entries, the PS field on
GRID entries, or by PARAM,AUTOSPC,YES.
2 APPENDIX C
Set Notations
C.4 Rigid Body Supports
In the following discussion, a free body is defined as a structure that is capable of motion
without internal stress; i.e., it has one or more rigid body degrees-of-freedom. The stiffness
matrix for a free body is singular with the defect equal to the number of stress-free, or rigid
body modes. A solid three-dimensional body has up to six rigid body modes. Linkages
and mechanisms can have a greater number. In order to permit the analysis of
mechanisms, no restriction is placed in the program on the number of stress-free modes.
Free-body supports are defined with a SUPORT or SUPORT1 entry. In the case of
problems using conical shell elements, the SUPAX entry is used. Free-body supports must
be defined in the global coordinate system. The SUPORT1 entry must be selected by the
SUPORT1 Case Control command.
In static analysis by the displacement method, the rigid body modes must be restrained in
order to remove the singularity of the stiffness matrix. The required constraints may be
supplied with single point constraints, multipoint constraints, or free body supports. If
free body supports are used, the rigid body characteristics will be calculated and a check
will be made on the sufficiency of the supports. Such a check is obtained by calculating the
rigid body error ratio and the strain energy as defined in the Rigid Body Matrix Generator
operation. This error ratio and the strain energy are automatically printed following the
execution of the Rigid Body Matrix Generator. The error ratio and the strain energy should
be zero, but may be nonzero for any of the following reasons:
Round-off error accumulation.
Insufficient free body supports have been provided.
Redundant free body supports have been provided.
The redundancy of the supports may be caused by improper use of the free body supports
themselves or by the presence of single point or multipoint constraints that constrain the
rigid body motions.
Static analysis with inertia relief is necessarily made on a model having at least one rigid
body motion. Such rigid body motion must be constrained by the use of free body
supports. These supported degrees-of-freedom define a reference system, and the elastic
displacements are calculated relative to the motion of the support points. The element
stresses and forces will be independent of any valid set of supports.
Rigid body vibration modes are calculated by a separate procedure provided that a set of
free body supports is supplied by the user. This is done to improve efficiency and, in some
cases, reliability. The determinant method, for example, has difficulty extracting zero
frequency roots of high multiplicity, whereas the alternate procedure of extracting rigid
body modes is both efficient and reliable. If the user does not specify free body supports
(or he specifies an insufficient number of them), the (remaining) rigid body modes will be
calculated by the method selected for the finite frequency modes, provided zero frequency
is included in the range of interest. If the user does not provide free body supports, and if
zero frequency is not included in the range of interest, the rigid body modes will not be
calculated.

Free body supports must be specified if the mode acceleration method of solution
improvement is used for dynamic problems having rigid body degrees-of-freedom (see
Rigid-body Modes on page 87 of the MSC.Nastran Basic Dynamics Users Guide). This
solution improvement technique involves a static solution, and although the dynamic
solution can be made on a free body, the static solution cannot be performed without
removing the singularities in the stiffness matrix associated with the rigid body motions.
2 APPENDIX C
Set Notations
C.5 Sets for Dynamic Reduction
There are several methods for reducing the size of models in dynamic analysis. The
method described here uses shpaes derived as generalized functions by static analysis. The
same entries may be used in static analysis with the same nomenclature being used
internally as in superelement analysis.
The statically independent degrees-of-freedom remaining after constraint elimination are
partitioned into two sets. The partitions are defined by listing the degrees-of-freedom for
one of the partitions on the ASETi entries. These degrees-of-freedom are referred to as the
analysis set. The remaining degrees-of-freedom in the omitted set with the remaining
degrees-of-freedom being placed in the analysis set. This is easier if the analysis set is large.
In the case of problems using conical shell elements, the OMITAX entry is used.
On of the more important applications of partitioning is Guyan reduction, described in
on page 311 of the MSC.Nastran Basic Dynamics Users Guide. This technique is a means for
reducing the number of degrees-of-freedom used in dynamic analysis with minimum loss
of accuracy. Its basis is that many fewer grid points are needed to describe the inertia of a
structure than are needed to describe it elasticity with comparable accuracy. The error in
the approximation is small provided that the set of displacements used for dynamic
analysis is judiciously chosen. Its members should be uniformly dispersed throughout the
structure, and all large mass items should be connected to grid points that are members of
the analysis set.
The use is cautioned to consider the fact that the matrix operations associated with this
partitioning procedure tend to create nonzero terms and to fill what were previously very
sparse matrices. The partitioning option is most effectively used if the members of the
omitted set are either a very large fraction or a very small fraction of the total set. In most
of the applications, the omitted set is a large fraction of the total and the matrices used for
analysis, while small, are usually full. If the analysis set is not a small fraction of the total,
a solution using the large, but sparser, matrices may be more efficient.
The a-set is further partitioned into the q-, r-, c-, and b-sets. The q-set is used to store the
coefficients for the generalized coordinates determined by dynamic reduction and/or
component mode calculations. The r-set is again used to determine rigid body modes. The
c-set contains coordinates considered free to move while computing the generalized
coordinate functions. The b-set contains coordinates fixed during this process.
The set used to describe the generalized functions for dynamic reduction and component
mode synthesis is the v-set, the union of the o-, r-, and c-sets.
The d-set (dynamic) for direct formulations is formed from the union of the structural
degrees-of-freedom in the a-set and the extra points of the e-set. The d-set can be used to
input unsymmetric terms into the equations of motion using DMIG and TF Bulk Data
entries. In the modal formulations, the h-set is composed of the union of the (modal set
and the e-set. Again, the DMIG and TF Bulk Data entries are a means to introduce
unsymmetric matrix coefficients.
i

C.6 Sets for Aerodynamics
Aerodynamic calculations are made in a Cartesian aerodynamic coordinate system. By
convention, the flow is in the positive x direction. The basic coordinate system was not
chosen, since it would place an undesirable restriction upon the description of the
structural model. Any MSC.Nastran Cartesian system may be specified, and flow will be
defined in the direction of its positive x-axis. All element and aerodynamic grid point data,
computed initially in the basic coordinate system, will be converted to the aerodynamic
coordinate system. The global (displacement) coordinate system of the aerodynamic grid
points will have its T1 direction in the flow direction. T3 is normal to the element for boxes,
and parallel to the aerodynamic T3 in the case of bodies. Coordinate system data are
generated for the aerodynamic grid points.
The grid points are physically located at the centers of the boxes and body elements.
Permanent constraints are generated for the unused degrees-of-freedom. A second set of
grid points, used only for undeformed plotting, is located at the element corners. All six
degrees-of-freedom associated with each grid point in this second set are permanently
constrained. Grid point numbers are generated based upon the element identification
number. For any panel, the external grid point numbers for the boxes start with the panel
identification number and increase consecutively.
Aerodynamic degrees-of-freedom, along with the extra points, are added after the
structural matrices and modes have been determined. This introduces the following
displacement sets:
k Aerodynamic box and body degrees-of-freedom
sa Permanently constrained degrees-of-freedom associated with
aerodynamic grid points
ps Union of the p-set (physical) and the sa-set
pa Union of k-set and ps-set (physical and aerodynamic)
2 APPENDIX C
Set Notations
C.7 Rules of Sets for Undefined Degrees-of-Freedom
Most of the Bulk Data entries used to list set membership specify mutually exclusive sets.
This means that degrees-of-freedom listed on more than one such entry will result in a fatal
error. For example, a degree-of-freedom may not be in the m-set on two RBAR entries, or
on a selected SPC entry and a SUPORT entry.
The a-set, by contrast, defines a combined set. Degrees-of-freedom listed on Bulk Data
entries that define its subsets, the q-, r-, c-, and b-set, may also be listed on ASETi Bulk Data
entries, although this is a redundant specification and is not necessary. However,
automatic restarts are more reliable if ASETi entries are present for all degrees-of-freedom
listed on SUPORT entries.
Bulk Data entries used to define mutually exclusive sets automatically place them in their
specified set and any combined set to which they belong. Since the ASETi entries define a
combined set, the subsets of the a-set, if not defined, must be placed in a mutually exclusive
set by an arbitrary convention. The conventions described below are used because they
allow the user to specify these sets with the minimum input for most cases. In
superelement analysis, exterior (boundary) degrees-of-freedom are automatically
determined and placed in the a-set of the superelement. Exterior degrees-of-freedom for
superelements are therefore equivalent to degrees-of-freedom listed on ASETi entries in
the discussion below.
The f-set is the union of the mutually exclusive o-, q-, r-, c-, and b-sets and includes the a-
set. The o-set exists under two conditions only:
1. If OMITi entries are present, the o-set consists of degrees-of-freedom listed
explicitly on OMlTi entries. The remaining f-set degrees-of-freedom are placed in
the b-set. They are then also members of the a-set.
2. If ASETi or QSETi entries are present, the a-set consists of all degrees-of-freedom
listed on ASETi entries and any entries listing its subsets, such as QSETi,
SUPORTi, CSETi, and BSETi entries. The remaining f-set degrees-of-freedom are
placed in the o-set. Note, however, there must be at least one explicit ASETi or
QSETi entry type present for the o-set to exist, even if it is a redundant
specification. If there are no ASETi or QSETi entries present, and SUPORT,
BSETi, or CSETi are present, the entire f-set is placed in the a-set; that is, the o-set
is not formed.

If both OMITi and ASETi entries are present, the OMITi entries are ignored, so that
condition 2, above, applies. Similarly, if ASETi entries are used in conjunction with entries
defining its subsets, undefined degrees-of-freedom in the a-set are placed in mutually
exclusive sets as shown in the following table (0 = no entry types present; 1 = entry type
present):
Set Defined by Entry Type
Case
1 2 3 4
c 0 0 1 1
b 0 1 0 1
Undefined a-set placed in b c b b
2 APPENDIX C
Set Notations
C.8 Output Selection via Set Specification
The membership of grid points and their components and scalar points in the various sets
is stored in the USET table as digits in a binary number. Some modules use input
parameters that are translations of these binary numbers as a shorthand notation. Three
formats are used:
Set Name
(DEFUSET, USET and USET1 Bulk Data entries, MATGPR, UPARTN, VEC, UMERGE,
UMERGE1, and PARAML (OP = USET) Modules)
The name of the set from the table below (which may be redefined by the DEFUSET entry).
Decimal Equivalent
(TABPRT, MATMOD (Option 17), and MATGEN (Option 11) Modules)
The decimal equivalent of the set is listed below. Several sets can be described by a single
number that is the sum of the listed equivalents. For example, to print the m and s sets, the
value of 3(1+2) is input to the TABPRT module. The value of -1 wiII cause the output of all
sets.
Bit Position
(VEC Module)
The location of the digit corresponding to the set in the binary number, starting with the
left-most digit, can be used to identify one set, as listed in Table C-1.
Table C-1 USET Table: Set Name, Decimal Equivalent, and Bit Position
Set
Name
Decimal
Equivalent
Bit
Positio
n
U1 -- 1
U2 -- 2
U3 -- 3
U4 -- 4
U5 -- 5
U6 -- 6
V 33554432 7
FR 16777216 8
T 8388608 9
O 4194304 10
B 2097152 11

The data block naming convention in the solution sequences for matrices related to the sets
is based on the set names. For example, the system stiffness matrix has the name
KGG. Some sets are used that do not appear in the USET table. The h-set, for example, is
the set used for modal formulations. Its size is not known until after the USET table is
formed, so it does not appear in the table. The j-set, a set synonymous with the g-set, is
used in superelement analysis. There are several other sets used that also do not appear in
the USET table.
C 1048576 12
PA 524288 13
K 262144 14
SA 131072 15
PS 65536 16
D 32768 17
FE 16384 18
NE 8192 19
P 4096 20
E 2048 21
SB 1024 22
SG 512 23
L 256 24
A 128 25
F 64 26
N 32 27
G 16 28
R 8 29
O 4 30
S 2 31
M 1 32
H,J,... -- --
Table C-1 USET Table: Set Name, Decimal Equivalent, and Bit Position (continued)
Set
Name
Decimal
Equivalent
Bit
Positio
n
K
gg
[ ]
2 APPENDIX C
Set Notations
Some names may be used for other purposes. The name H, for example, may be input to
the MATGPR module when one wants to label rows or columns of matrices sequentially,
rather than by their external sequence numbers. This feature may be used to number
vectors in load matrices or in other instances not related to modal analysis, even though H
is defined for modal analysis.

1. Miura, H., MSC/NASTRAN Handbook for Structural Optimization, The
MacNeal-Schwendler Corporation, 1988.
2. Moore, G.J. MSC/NASTRAN Design Sensitivity and Optimization Users Guide,
Version 67, The MacNeal-Schwendler Corporation, 1992.
3. Moore, G.J. and Nagendra, G.K., Dynamic Response Sensitivities in
MSC/NASTRAN Proceedings of the MSC 1991 World Users Conference, Paper
No. 4, March 1991.
4. Patel, H., Multidisciplinary Design with Superelements in MSC/NASTRAN,
AIAA CP-92-4732, Fourth AIAA/USAF/NASA/OAI Symposium on
Multidisciplinary Analysis and Optimization, Cleveland, Ohio, September1992.
5. Johnson, E.H., and Reymond, M.A., Multidisciplinary Aeroelastic Analysis and
Design using MSC/NASTRAN, AIAA CP-91-1097, 32nd SDM Conference,
Baltimore, MD, April 1991.
6. Kodiyalam, S., Vanderplaats, G., Miura, H., Nagendra, G. and Wallerstein, D.
Structural Shape Optimization with MSC/NASTRAN, AIAA CP 90-1007, 31st
SDM Conference, Long Beach, CA, April 1990.
7. Johnson, E.H., Disjoint Design Spaces in the Optimization of Harmonically
Excited Structures, AIAA Journal, Vol. 14, No. 2, February 1976, pp 259-261.
8. Introduction to Version 67 of MSC/NASTRAN, The MacNeal-Schwendler
Corporation, September 1991.
9. Cassis, J.H., Optimum Design of Structures Subjected to Dynamic Loads, PhD.
Thesis UCLA-ENG-7451, June 1974.
APPENDIX
D
Selected Bibliography

3
I N D E X

A
ABS option, 222
Acceleration loads
example, 15
Acceleration method, 15
Acoustic absorbers, 126
Acoustic analysis, 106, 124
advanced methods, 131
diagnostics, 130
example, 133
job cost estimation, 128
model building, 124
MSC.Nastran setup, 127
run strategies, 128
Acoustic barriers, 126
Acoustic loads, 145
Acoustic panels, 126
Adative time stepping, 170
Aerodynamic
degrees of freedom, 198
displacements, 198
extra points, 198
influence coefficient matrix, 197
matrices
generation of, 198
methods, 83
theories, 194
Aeroelastic
transient response analysis, 203
Aeroelastic analysis, 81
DMAP sequences, 196
dynamic response, 82
flutter, 83
modules, 196
optimization, 82, 196
overview, 196
special features, 83
static response, 81
theoretical implementation in
MSC.Nastran, 196
Aeroeleastic analysis
flutter, 82
Aeroeleastic solutions, 194
Applied loads, 199
ASET, 12, 40, 54, 55, 56, 60, 284
a-set, 284
Axisymmetric fluids
in tanks, 105
B
Base excitations, 15
Boundary conditions, 86
C
CHECKA, 52, 59
Complex eigensolutions, 146, 147
Complex Lanczos method, 149
example, 153
Hessenberg method, 148, 152
Inverse power method, 149
Modal Hessenberg method, 149
modeling techniques, 152
user interface, 150
Component mode synthesis, 63, 101, 131, 193
Computer resources, 9
conjugate matrix, 72
Control systems, 83, 84
example, 27
Coriolis forces, 13, 38
Coupled acoustic analysis
job output, 129
Coupled mass matrix terms, 11
Coupled scalar mass
example, 12
CVISC, 17
D
damper elements, 217
I N D E X
MSC.Nastran
Advanced
Dynamic
Analysis Users
Guide
INDEX
Damping, 17, 212
critical, 20
modal, 19
nonlinear, 21
structural, 17, 22
viscuous, 17, 20, 22
Data recovery, 214
and superelements, 219
for modal methods, 215
improving results, 217
Decoupled response solutions, 131
Deformed plots, 60
Design sensitivity, 195, 205
DIAG 10, 33
Direct matrix input (DMIGs), 23, 24
DoubletLattice method, 83, 194
Downwash, 197
Dynamic analysis
cost factors, 8
E
EFMASS, 54, 56
Elastic rigid coupling matrix (MER), 56
Enforced acceleration, 97
Enforced motion
with loads, 87
Excitations, 141
Extra points, 152, 200
F
Fluid
boundaries, 126
examples, 123
loads, 145
Fluidstructure analysis, 101
Flutter, 82, 195
Flutter analysis, 83, 195
PKmethod, 83
Fourier
coefficients, 105
transform methods, 176
transform, inverse, 204
transformations, 203
transforms, 240
Free bodies, 51, 190
Free surfaces, 119
Frequency range, 8
Frequency response, 178, 195
equations for, 95
G
General dynamic reduction, 66
Generalized aerodynamic matrices, 200
Generalized dynamic reduction, 60, 63, 101
Geometric shape methods, 192
Gravity effects, 122
Grid point weight generator, 44
Grid Point Weight Generator (GPWG), 48
Guyan reduction, 54, 63, 66
H
HARMONICS, 107
Hessenberg method, 148, 152
Hydroelastic modeling, 107
data processing, 111
sample, 113
solution sequences, 110
Hysteresis, 22
I
impedance, 70
Inertia relief, 10, 14, 15, 102, 193, 199, 289
Inertial forces
and rotating coordinate systems, 74
Inertial loads approach, 89
input data, 90
recommendations, 91
Initial conditions, 145
Integration matrix, 197
Interation strategies, 174
Inverse Fourier Transform, 204
Inverse power method, 149, 150
Inverse transformation, 178
K
Flutter analysis
Kmethod, 201
Kmethod, 201
Kinetic and strain energy, 59
3 INDEX
L
Lagrange multiplier method, 36, 87, 92
computation time, 102
DMAP alter implementation, 92, 99
example problems, 103
user interface, 99
Lanczos method, 55, 60, 101, 149, 150
Large mass method, 16, 89
Large mass/spring method, 87, 88
Large springs method, 88
Linear transient integration method, 169
Load tranformation
time to frequency, 177
Loading methods, 142
recommendations, 144
Loads, 8
M
Mach box method, 194
Mach number, 201
Mass distribution, 48
Mass matrix, 10, 199
Mass modeling, 10
and inertia relief, 10
effects in moving coordinate systems, 13
options, 10
parameters, 11
MATGEN module, 295
MATGPR module, 295
MATMOD module, 295
Matrix method
for modal response, 216
Matrix reduction, 8
Matrix reduction methods
theory, 63
Matrix theory
dynamic response, 94
Mechanical filter
example, 88
Mechanisms, 15
Modal analysis, 54, 215
Modal damping matrix, 201
Modal formulation
in aeroelastic solutions, 194
Modal Hessenberg method, 149
Modal mass matrix, 201
Modal stiffness matrix, 201
Modal transformations, 24
Mode acceleration, 102, 217
Mode displacement
vector method, 215
Model debugging, 39
deformed plots, 60
elasticrigid coupling matrix, 56
free body equilibrium check, 51
hints, 39
kinetic and strain energy, 59
modal analysis, 54
model checkout, 43
references, 60
rigid body check, 52
static loading, 49
testing, 40
thermal test case, 53
Modeling
debugging, 39
geometry plots, 40
recommendations, 8
testing, 40
N
Newmark integration, 170
NLTRD module, 169
NOLINi, 31, 36
NONLINEAR, 31
Nonlinear element library, 161
Nonlinear load functions (NOLINi), 31
example, 33, 36
input recommendations, 35
velocity dependent, 33
Nonlinear transient analysis (SOL 129), 21
solution sequences, 34
Nonlinear transient response, 161
user interface, 163
Numerical conditioning, 88
O
OMT, 12
Optimization
and dynamics, 205
coupled to other disciplines, 212
example, 207, 209
OPTION, 223
Options
automated meshing, 8
mass modeling, 10
o-set, 284
Output requests, 9
INDEX
P
PARAML module, 295
Peak response calculation, 227
Perturbations, 152
Phantom structural boundaries, 122
Pickups, 26
Piston theory, 194
Plastic yield, 22
Poisson's ratio, 181
PRESPT, 106
PRESSURE, 107
Q
QuasiNewton line searches, 171
R
Random response analysis, 131, 145, 195
aeroelastic, 204
example, 241
theory, 239
with coupled excitations, 238
Reduced frequency, 197, 201
Response spectra analysis, 190, 221, 228
examples, 229
generation, 223
user interface, 226
user interface, 229
Restarts, 172
Rigid body, 192, 289
checks, 52
modes, 97
Rigid Format alter
RF24032, 43
Rotating structures, 70
Rotationdependent terms
SSSALters for dynamic analysis, 79
r-set, 199
S
Scaled response spectra, 102
Set definitions, 282
SET logic, 214
Solution sequences
SOL 103, 229
SOL 129, 35
Spectral densities, 238
Spinning bodies, 13
SRSS method, 223
SSSAlter library, 80
CHECKA, 59
SSSAlters, 103
Static
aeroelasticity, 194
equations of motion, 198
condensation, 63
loading, 49
preloads, 144
Strip theory, 83, 194
Structural damping, 201
Sturm sequence technique, 149
Subsonic theory, 194
Superelements, 22, 102, 131, 193, 285
reduction operations, 12
T
TABPRT module, 295
Thermal test case, 53
Transducers, 26
Transfer functions, 11, 12, 23, 25, 152, 240
example, 27
higher order, 26
limitations, 25
Transient analysis, 221
solutions theory, 32
Transient response, 195
equations for, 96
TSTEPNL, 35
U
UMERGE module, 295
UMERGE1 module, 295
Unstable integration, 33
UPARTN module, 295
V
VEC module, 295
Vector transformations, 70
Vectors
progressive and regressive, 73
Virtual fluid mass, 117, 131
example, 118
theory, 119
user interface, 117
3 INDEX
Virtual mass, 102
phantom boundaries, 122
singularities for enclosed volumes, 120
Viscoelastic materials, 181
example, 184
Z
ZONA51, 83

Nastran Advanced Dynamics

Uploaded by

Copyright:

Available Formats

Nastran Advanced Dynamics

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

Nastran Advanced Dynamics

Uploaded by

Copyright:

Available Formats

MSC.

You might also like