The

MacNeal-Schwendler
Corporation
815 Colorado Boulevard Tel: (213) 258-9111
Los Angeles, California 90041-1777 FAX: (213) 259-3838

Design Sensitivity and Optimization in MSC/NASTRAN

SEMINAR NOTES

NA∗V70.5∗Z∗Z∗Z∗SM-NAS107-NT1

February 1999
 DISCLAIMER

The concepts, methods, and examples presented in this text are for educational
purposes only and are not intended to be exhaustive or to apply to any particular
engineering problem or design. The MacNeal-Schwendler 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.

Printed in U.S.A.
1998 by The MacNeal-Schwendler Corporation
All rights reserved.
 The MacNeal-Schwendler Corporation
Seminar Questionnaire

Seminar ____________________________ Date ____________________________

Location ____________________________ Instructor ____________________________

Your Name ____________________________ Title ____________________________

Company _____________________________________________________________________

Your feedback is very important in helping MSC meet the training needs of its clients. Your specific
written comments are especially helpful. Please use the following scale:
1 = poor 2 = marginal 3 = average 4 = good 5 = very good

Rating Category Comments

Effectiveness of instructor in
explaining technical concepts.

Ability of instructor to relate

seminar content to real-world
problem solving.

Organization of material in a
clear and understandable
manner.

Balance of material (i.e., theory

vs. practical, applied
engineering).

(Continued on back)

Rev 5/93
 The MacNeal-Schwendler Corporation
Seminar Questionnaire (Cont.)

1 = poor 2 = marginal 3 = average 4 = good 5 = very good

Rating Category Comments

Quality of course handouts.

Usefulness of maps. Were you

able to locate the seminar room?
Any suggestions?

Impression of hotel
accommodations. At which
hotel did you stay? Any
problems?

How were the seminar lunches?

Overall impression of seminar.

Please offer suggestions for
improvement.

Would you recommend this instructor for future classes? Why or why not? ______________________

__________________________________________________________________________________

Rev 5/93
 TABLE OF CONTENTS

TITLE SECTION

1.0 INTRODUCTION

WHAT IS “DESIGN OPTIMIZATION”? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-1

BASIC FEATURES IMPLEMENTED IN MSC/NASTRAN . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2
MSC/NASTRAN IMPLEMENTATION OF STRUCTURAL OPTIMIZATION. . . . . . . . . . . . . 1-3
CONCEPTS PRESENTED IN THIS COURSE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-4
STRENGTHS OF MSC/NASTRAN STRUCTURAL OPTIMIZATION. . . . . . . . . . . . . . . . . . 1-5
GENERAL FUNCTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-6
SUMMARY OF NEW CAPABILITIES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-8
BASIC OPTIMIZATION PROBLEM STATEMENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-10
EXAMPLE FORMULATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-11
ONE BAR SUBJECT TO COMPRESSIVE LOAD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-16

2.0 NUMERICAL OPTIMIZATION

MINIMIZATION OF FUNCTIONS OF A SINGLE VARIABLE. . . . . . . . . . . . . . . . . . . . . . . . 2-1

GRADIENTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4
UNCONSTRAINED MINIMIZATION METHODS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-6
KUHN-TUCKER CONDITIONS FOR OPTIMALITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-10
SIMPLE CANTILEVER EXAMPLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-12
SERIES APPROXIMATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-15
SEQUENTIAL LINEAR PROGRAMMING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-19
METHOD OF FEASIBLE DIRECTIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-22
MODIFIED METHOD OF FEASIBLE DIRECTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-26
NUMERICAL IDENTIFICATION OF ACTIVE AND VIOLATED CONSTRAINTS. . . . . . . . . 2-32

3.0 DESIGN MODELING I

WHAT IS A DESIGN MODEL?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1

HOW ARE THE DESIGN AND ANALYSIS MODELS RELATED?. . . . . . . . . . . . . . . . . . . . 3-2
DESIGN AND ANALYSIS MODEL COMPARISON . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3
DESIGN MODEL DEFINITION PROCESS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-5
DEFINING THE ANALYSIS DISCIPLINES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-6
DEFINING THE DESIGN VARIABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-7
DESVAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-8
RELATING DESIGN VARIABLES TO PROPERTIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-10
DVPREL1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-11
IDENTIFYING THE DESIGN RESPONSES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-15
DRESP1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-16
DEFINING THE OBJECTIVE FUNCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-24
DESOBJ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-25
DEFINING THE DESIGN CONSTRAINTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-27
DCONSTR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-28
DCONADD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-31
DESSUB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-33
DESGLB. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-34
BASIC DESIGN PROCESS CONTROL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-38
DOPTPRM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-39

iii
 TABLE OF CONTENTS

TITLE SECTION

4.0 EXAMPLES I

EXAMPLE – 3-BAR TRUSS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1

10-BAR TRUSS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-6
10-BAR TRUSS RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-10
BEAM VIBRATION FREQUENCY AS A DESIGN CONSTRAINT (M. J. Turner 1970) . . . . 4-15
BEAM VIBRATION FREQUENCY AS A DESIGN CONSTRAINT RESULTS . . . . . . . . . . . 4-22

5.0 SPECIAL MODELING TOPICS I

DESIGN VARIABLE LINKING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-1

DLINK. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-2
REDUCED BASIS FORMULATIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-5
DESIGN RESPONSES AND CASE CONTROL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-9
MODE TRACKING FEATURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-11
MODE TRACKING REQUEST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-12
MODE TRACKING PARAMETERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-13
EQUALITY CONSTRAINTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-14
FREQUENCY MATCHING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-15

6.0 DESIGN MODELING II

DESIGN MODEL DEFINITION PROCESS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-1

USER-DEFINED DESIGN VARIABLE-TO-PROPERTY RELATIONS. . . . . . . . . . . . . . . . . 6-2
DVPREL2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-3
DEQATN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-5
DTABLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-9
EXAMPLE – DEFINING A RECTANGULAR SECTION BAR. . . . . . . . . . . . . . . . . . . . . . . . 6-10
USER-DEFINED RESPONSES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-11
DRESP2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-13
EXAMPLE – DEFORMATION CONTROL WITH TYPE-2 RESPONSES . . . . . . . . . . . . . . 6-16
EXAMPLE – BAR AXIAL PLUS BENDING STRESSES . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-19
RESTRICTIONS IN FORMING SYNTHETIC RESPONSES . . . . . . . . . . . . . . . . . . . . . . . . 6-20

7.0 EXAMPLES II

STIFFENED PANEL WEIGHT MINIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-1

25-BAR TRUSS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-12
CANTILEVER PLATE (REDUCED BASIS FORMULATION). . . . . . . . . . . . . . . . . . . . . . . . 7-26

8.0 SPECIAL MODELING TOPICS II

BEAM CROSS-SECTION LIBRARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-1

BEAM CROSS-SECTION LIBRAY OPTIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2
COMPARISON BETWEEN DESIGNING PBAR AND PBARL . . . . . . . . . . . . . . . . . . . . . . . 8-3
DVPREL1 INCLUDED ON DRESP2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-5
DESIGN VARIABLE LINKING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-6
CONSTRAINT EQUATION SCALING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-9
BEAM DESIGN – GENERAL ISSUES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-13
BEAM DESIGN – SIMPLE EXAMPLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-16

iv
 TABLE OF CONTENTS

TITLE SECTION

9.0 FUNDAMENTALS OF STRUCTURAL OPTIMIZATION

STRUCTURAL OPTIMIZATION – THE DIFFICULTIES. . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-1

STRUCTURAL OPTIMIZATION – THE SOLUTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-3
DESIGN VARIABLE LINKING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-7
DESIGN VARIABLE LINKING IN SHAPE REDESIGN . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-11
GENERATION OF CONSTRAINTS IN MSC/NASTRAN . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-15
CONSTRAINT SCREENING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-16
DSCREEN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-18
FORMAL APPROXIMATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-21
APRCOD – SELECTION OF APPROXIMATE FORM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-28
SENSITIVITY ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-30
SEMIANALYTIC SENSITIVITY ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-31
DELB – FINITE DIFFERENCE STEP SIZE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-33
SENSITIVITY ANALYSIS – IMPLEMENTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-35
ADJOINT SENSITIVITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-37
ADJOINT SENSITIVITY THEORY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-38
DESIGN SENSITIVITY COEFFICIENT MATRICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-47
DSAPRT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-49
FORMATTED SENSITIVITY PRINT EXAMPLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-51
MOVE LIMITS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-52
MOVE LIMITS IMPOSED ON ANALYSIS MODEL PROPERTIES . . . . . . . . . . . . . . . . . . . 9-53
MOVE LIMITS IMPOSED ON DESIGN VARIABLES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-55
AUTOMATIC UPDATES OF MOVE LIMITS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-56
MOVE LIMITS UPDATE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-57
PARAM OPTEXIT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-58
FLOW CHART FOR SOLUTION 200 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-59
CONVERGENCE AT THE DESIGN CYCLE LEVEL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-63
CONVERGENCE CRITERIA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-65
SUMMARY OF PARAMETERS FOR DESIGN OPTIMIZATION . . . . . . . . . . . . . . . . . . . . . 9-67

10.0 SHAPE OPTIMIZATION

BASIC EQUATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-1

BASIS VECTORS – EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-2
SHAPE BASIS VECTORS IN THE DESIGN MODEL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-6
DESIGN MODELING INPUT FOR
SHAPE OPTIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-7
MANUAL GRID VARIATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-8
DVGRID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-10
AUXILIARY MODELS IN SHAPE OPTIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-12
EXAMPLE – DIRECT INPUT OF SHAPES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-14
DVSHAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-19
AUXILIARY BOUNDARY MODELS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-24
EXAMPLE – ANALYTIC BOUNDARY SHAPES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-26
AUXCASE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-30
AUXMODEL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-31
BNDGRID. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-32
DVBSHAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-33

v
 TABLE OF CONTENTS

TITLE SECTION

SUMMARY OF SHAPE BASIS VECTOR GENERATION METHODS. . . . . . . . . . . . . . . . . 10-41

GUIDELINES, RECOMMENDATIONS AND LIMITATIONS . . . . . . . . . . . . . . . . . . . . . . . . 10-43

11.0 DYNAMIC RESPONSE OPTIMIZATION

DYNAMIC RESPONSE OPTIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-1

GENERAL CONSIDERATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-2
BASIC EQUATIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-3
EXAMPLE – MODAL FREQUENCY RESPONSE OPTIMIZATION. . . . . . . . . . . . . . . . . . . 11-6
EXAMPLE – ACOUSTIC OPTIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-18

12.0 SUPERELEMENT OPTIMIZATION

GENERAL CONSIDERATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-1

CASE CONTROL SPECIFICATION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-3
EXAMPLE – TWENTY FIVE BAR TRUSS, SUPERELEMENT OPTIMIZATION . . . . . . . . . 12-5

13.0 AEROELASTIC OPTIMIZATION

AEROELASTIC OPTIMIZATION MSC BACKGROUND . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-1

STATIC AEROELASTICITY DESIGN CONDITIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-2
STATIC AEROELASTICITY SENSITIVITY ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-3
FLUTTER DESIGN. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-4
FLUTTER RESPONSE PROPERTIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-5
FLUTTER RESPONSE SENSITIVITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-6
AEROELASTIC DESIGN EXAMPLE (HA200A and B) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-7
ANALYSIS CONDITIONS FOR EXAMPLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-9
DESIGN CONDITIONS FOR EXAMPLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-10
HA200B OBJECTIVE RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-12
HA200B DESIGN VARIABLE RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-13
HA200B FLUTTER RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-14
HA200A INPUT FILE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-15
DESIGN MODEL AND BULK DATA INPUT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-17
HA200A SELECTED RESULTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-23
HA200B SELECTED RESULTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-31

A CASE CONTROL COMMANDS AND BULK DATA ENTRIES

DESIGN MODELING INPUT DATA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-1

CASE CONTROL COMMANDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-3
AUXCASE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-4
AUXMODEL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-5
DESGLB. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-6
DESOBJ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-7
DESSUB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-8
MODTRAK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-9
BULK DATA ENTRIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-10
BNDGRID. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-11
DCONADD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-12
DCONSTR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-13
DEQATN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-15

vi
 TABLE OF CONTENTS

TITLE SECTION

DESVAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-19
DLINK. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-20
DOPTPRM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-21
DRESP1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-25
DRESP2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-31
DSAPRT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-34
DSCREEN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-37
DTABLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-39
DVBSHAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-40
DVGRID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-41
DVPREL1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-43
DVPREL2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-46
DVSHAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-48
MODTRAK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-49

vii
 SECTION 1

INTRODUCTION

WHAT IS “DESIGN OPTIMIZATION”? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-1

BASIC FEATURES IMPLEMENTED IN MSC/NASTRAN . . . . . . . . . . . . . . . . . . . . . 1-2

MSC/NASTRAN IMPLEMENTATION OF STRUCTURAL OPTIMIZATION. . . . . . . . 1-3

CONCEPTS PRESENTED IN THIS COURSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-4

STRENGTHS OF MSC/NASTRAN STRUCTURAL OPTIMIZATION. . . . . . . . . . . . . 1-5

GENERAL FUNCTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-6

SUMMARY OF NEW CAPABILITIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-8

BASIC OPTIMIZATION PROBLEM STATEMENT . . . . . . . . . . . . . . . . . . . . . . . . . . 1-10

EXAMPLE FORMULATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-11

ONE BAR SUBJECT TO COMPRESSIVE LOAD . . . . . . . . . . . . . . . . . . . . . . . . . . 1-16

 WHAT IS “DESIGN OPTIMIZATION”? 1

Automated modifications of the analysis model parameters to achieve

a desired objective while satisfying specified design requirements.

WHAT ARE THE POSSIBLE APPLICATIONS?

• Structural design improvements (optimization)

• Generation of feasible designs from infeasible designs

• Model matching to produce similar structural responses

• System parameter identification

• Configuration evaluations

• Sensitivity analysis

• Others - (depends on designer’s creativity)

1-1
BASIC FEATURES IMPLEMENTED IN MSC/NASTRAN

• Easy access to design synthesis capabilities

• Concept of design model

• Flexibility for design model representation

• User-supplied equation interpretation capability

• Efficient solution for problems of “any” size

• Number of finite element analyses as the measure of efficiency

1-2
 MSC/NASTRAN Implementation of Structural
Optimization

Initial
Design
Improved
Design

The required number of

iterations of the external loop
must be small.
Structural
Response
Analysis

Constraint Approximate
Optimizer
Screening Model

Sensitivity Many Times

Analysis

Finite Element
Analysis

One time around the loop is referred to as a design cycle or design

iteration.

1-3
 Concepts Presented in this Course

• Optimization

• Sensitivity Analysis

• Objective Function

• Constrants

• Design Model

• Approximate Problem

• Constraint Screening

• Move Limits

• Convergence

1-4
 STRENGTHS OF MSC/NASTRAN STRUCTURAL
OPTIMIZATION

• Efficient performance for small- to large-scale problems

• Reliable convergence characteristics

• Flexible user interface and user-defined equations

• Full implementation of approximation concepts

• Continuous enhancements

• Results dependent on the proven reliability of MSC/NASTRAN

analysis

• Commercial level support as a part of MSC/NASTRAN

• Access to the familiar analysis tools in MSC/NASTRAN

1-5
 GENERAL FUNCTIONS

• Solution Sequence

200 - supports design sensitivity and optimization

• Analysis Types

Statics
Normal modes
Buckling
Direct frequency*
Modal frequency*
Modal transient*
Static aeroelastic
Aeroelastic flutter

(*Includes acoustic responses)

• Design Variables

Sizing properties(including superelements)

Shape(superelements with manual grid variation only)

1-6
 SUMMARY OF NEW CAPABILITIES

Version 69

• Beam Cross Section Library

• Element and Property Input Checks

• Formatted Sensitivity Prints

• Mode Tracking

• Multiple Boundary Conditions

• Buckling

• Normal Modes

• Flutter

• New Natural Frequency Response Type

1-7
 SUMMARY OF NEW CAPABILITIES (Cont.)

Version 70

• Adjoint Sensitivity Method

• NVH Applications

• Over 10x Performance and Disk Space Improvements

• Response Type Dependent Approximation

1-8
 SUMMARY OF NEW CAPABILITIES (Cont.)

Version 70.5

• Machine Precision Optimization

• Presentation of Weight as a Function of Material ID

• Rigid Element Shape Sensitivities

• Simultaneous Static and Static Aeroelasticity Analyses

1-9
 BASIC OPTIMIZATION PROBLEM STATEMENT

• Design variables

Find {X} = (X1, X2, . . ., XN)

• Objective function:

Minimize F(X)

• Subject to:

• Inequality constraints:

Gj (X) ≤ 0 j = 1, 2, . . ., L

• Equality constraints:

Hk (X) = 0 k = 1, 2, . . ., M

• Side constraints:

x iL ≤ x i ≤ x iu i = 1 , 2 , . . ., N

1-10
 EXAMPLE FORMULATIONS

Example 1:

• Determination of equilibrium position in a two-spring system

(geometrically nonlinear deformation)

P2
l1 = 10 cm K1 = 8N/cm
X2
P2 = 5N P1
P1 = 5N
X1

l2 = 10 cm
K2 = 1N/cm

Undeformed Deformed

1-11
 EXAMPLE FORMULATIONS (Cont.)

• Potential energy of the system:

1 1
PE = --- K 1 ( ∆l 1 ) 2 + --- K 2 ( ∆l 2 ) 2 – P 1 X 1 – P 2 X 2
2 2

1 2
PE ( X 1, X 2 ) = --- K 1 [ X 12 + ( l 1 – X 2 ) 2 – l 1 ]
2

1 2
+ --- K 2 [ X 12 + ( l 2 – X 2 ) 2 – l 2 ] – P 1 X 1 – P 2 X 2
2

• Equilibrium position may be found by minimizing the total potential

energy of the system:

Find X = (X1, X2)

min F(X) = PE(X1, X2)

1-12
 EXAMPLE FORMULATIONS (Cont.)

Example 2: Gussetted Tube Connection

t3
t2

t1
h
P

l
---
4
l b

• Minimize structural mass with constraints on:

• Stresses in the gussets:

σ min plate ≤ σ plate ≤ σ max plate

• Buckling stresses in the tube:

σ tube ≤ σ buck

• Sizing and gauge requirements:

t 1 ≥ t 1min, 2t 1 ≤ b
t 2 ≥ t 2min, 2t 2 ≤ h
t 3 ≥ t 3min, t3 ≤ b

1-13
 EXAMPLE FORMULATIONS (Cont.)

• Optimization task statement

• Minimize:

2
ρv = ρt 3  --- + ρl [ bh – ( h – 2t 2 ) ( b – 2t 1 ) ]
l
 4

T Design
where: { x } = [ b h t 1 t 2 t 3 ] Variables
σ min plate – σ plate ≤ 0

σ plate – σ max plate ≤ 0 Response

Constraints
σ tube – σ buck ≤ 0

2t 1 – b ≤ 0
Design
2t 2 – h ≤ 0
Constraints
t3 – b ≤ 0

t1 – t1 ≤ 0
min

t2 – t2 ≤ 0 Side
min Constraints
t3 – t3 ≤ 0
min

1-14
 EXAMPLE FORMULATIONS (Cont.)

Example 3: Best choice design among candidate designs

• Assume three trial designs exist:

{ Y }1 , { Y }2 , { Y }3

where { Y } i = { A 1 , A 2 , r 1 , r 2 , t, l 1 , l 2 }
i

Not all designs may satisfy design requirements.

• Choose the appropriate combination of { Y } 1 , { Y } 2 , and { Y } 3

which will minimize the structural weight, yet still satisfy all the
design requirements.

• Proposed design:

{ Y′ } = α 1 { Y } 1 + α 2 { Y } 2 + α 3 { Y } 3

• Minimize:

W = F ( α1 , α2 , α3 )

• Subject to:

g j = g j ( α1 , α2 , α3 ) ≤ 0

hk = hk ( α1 , α2 , α3 ) = 0

1-15
 ONE BAR SUBJECT TO COMPRESSIVE LOAD

• Design problem:

Minimize the weight of the structure by changing the bar

cross-sectional area, without violating limits on compressive
stress

P = 1.0E4 N

L = 500 mm
2
Ainit = 75 mm
ρ, A, E
2
E = 7.0E4 N/mm
L
3
ρ = 2.7E–6 kg/mm
2
σmax = 200 N/mm

Optimal Design:

P
σ max = -------------
A opt
2
A opt = 1.0E4 N ⁄ 200 N ⁄ mm

2
= 50 mm

1-16
ONE BAR SUBJECT TO COMPRESSIVE LOAD (Cont.)

ID MSC, D200X9
TIME 5
SOL 200
CEND
TITLE = STRUCTURAL OPTIMIZATION TEST PROBLEM (STATIC LOADING)
SUBTITLE = SINGLE BAR, 1 DOF, SUBJECT TO COMPRESSIVE LOAD
STRESS = ALL
DISP = ALL
SPC = 1
LOAD = 1
ANALYSIS = STATICS
DESOBJ(MIN) = 20 $ OBJECTIVE FUNCTION DEFINITION
DESSUB = 100 $ CONSTRAINT SET SELECTION
BEGIN BULK
$
$ ANALYSIS MODEL
$
$ units are N, kg, mm:
$
GRID, 1, , 0.0, 0.0, 0.0
GRID, 2, , 0.0, 0.0, 500.0
SPC, 1, 1, 123456
SPC, 1, 2, 12456
CROD, 1, 1, 1, 2
PROD, 1, 1, 75.
$ mm**2
MAT1, 1, 7.0E4, , 0.3, 2.7E-6
$ N/mm**2 kg/mm**3
FORCE, 1, 2, , 1.0, 0.0, 0.0, -1.0E4
$ N
$
$ DESIGN MODEL DESCRIPTION
$
DESVAR, 1, A, 75., 10., 150.
DVPREL1,1, PROD, 1, 4, , , , , +DVP1
+DVP1, 1, 1.0
DRESP1, 10, STRESS, STRESS, PROD, , 2, , 1
DRESP1, 20, WEIGHT, WEIGHT
DCONSTR,100, 10, -200., 200.
$ N/mm**2
DSCREEN,STRESS, -0.5, 1
DOPTPRM,IPRINT, 2, DESMAX, 10, DELP, 0.5, P1, 1 +
+, P2, 15
ENDDATA

1-17
 ONE BAR SUBJECT TO COMPRESSIVE LOAD (Cont.)

***** NORMAL CONVERGENCE CRITERIA SATISFIED ***** (HARD CONVERGENCE DECISION LOGIC)

**************************************************************************************
CONVERGENCE ACHIEVED BASED ON THE FOLLOWING CRITERIA
(HARD CONVERGENCE DECISION LOGIC)

RELATIVE CHANGE IN OBJECTIVE 0.0000E+00 MUST BE LESS THAN 1.0000E-03

OR ABSOLUTE CHANGE IN OBJECTIVE 0.0000E+00 MUST BE LESS THAN 1.0000E-02
--- AND ---
MAXIMUM CONSTRAINT VALUE 3.1044E-04 MUST BE LESS THAN 5.0000E-03
(CONVERGENCE TO A FEASIBLE DESIGN)
--- OR ---
MAXIMUM OF RELATIVE PROP. CHANGES 0.0000E+00 MUST BE LESS THAN 1.0000E-03
AND MAXIMUM OF RELATIVE D.V. CHANGES 0.0000E+00 MUST BE LESS THAN 1.0000E-03
1-18

(CONVERGENCE TO A BEST COMPROMISE INFEASIBLE DESIGN)

**************************************************************************************
 ONE BAR SUBJECT TO COMPRESSIVE LOAD (Cont.)

***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************
(HARD CONVERGENCE ACHIEVED)
(SOFT CONVERGENCE ACHIEVED)
NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 3
NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 2
OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY
---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
CYCLE APPROXIMATE EXACT OF OF
NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------
INITIAL 1.012500E-01 -3.333334E-01
1 6.746810E-02 6.747905E-02 -1.623075E-04 3.104401E-04
2 6.747905E-02 6.747905E-02 0.000000E+00 3.104401E-04
---------------------------------------------------------------------------------------------------------------
1-19

1 STRUCTURAL OPTIMIZATION TEST PROBLEM (STATIC LOADING) APRIL 20, 1995 MSC/NASTRAN 8/17/94 PAGE 38
SINGLE BAR, 1 DOF, SUBJECT TO COMPRESSIVE LOAD

DESIGN VARIABLE HISTORY

-----------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
-----------------------------------------------------------------------------------------------------------------------------
1 | 1 | A | 7.5000E+01 : 4.9984E+01 : 4.9984E+01 :
-----------------------------------------------------------------------------------------------------------------------------
0*** USER INFORMATION MESSAGE 6464 (DOM12E)
RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 2.
1-20
 SECTION 2

NUMERICAL OPTIMIZATION

MINIMIZATION OF FUNCTIONS OF A SINGLE VARIABLE. . . . . . . . . . . . . . . . . . . 2-1

GRADIENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4

UNCONSTRAINED MINIMIZATION METHODS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-6

KUHN-TUCKER CONDITIONS FOR OPTIMALITY . . . . . . . . . . . . . . . . . . . . . . . . . 2-10

SIMPLE CANTILEVER EXAMPLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-12

SERIES APPROXIMATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-15

SEQUENTIAL LINEAR PROGRAMMING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-19

METHOD OF FEASIBLE DIRECTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-22

MODIFIED METHOD OF FEASIBLE DIRECTIONS. . . . . . . . . . . . . . . . . . . . . . . . . 2-26

NUMERICAL IDENTIFICATION OF ACTIVE AND VIOLATED CONSTRAINTS . . 2-32

 MINIMIZATION OF FUNCTIONS OF A SINGLE
VARIABLE 2

Consider the following function of a single variable X

F = F(X)

∗ dF
extremum is defined for X at which -------- = 0
dX

The numerical optimizer “samples” this continuous function and uses

this data to estimate an extremum. For example, if the following
information is available:

′
( X 0 , F 0 ), ( X 0 , F 0 ), ( X 1 , F 1 )

we could use a quadratic polynomial approximation of the form:

2
F˜ ( X ) = a 0 + a 1 X + a 2 X

2-1
 MINIMIZATION OF FUNCTIONS OF A SINGLE
VARIABLE (Cont.)

to construct 3 equations in 3 unknowns:

′
F 0 = a 1 + 2a 2 X 0
2
F0 = a0 + a1 X0 + a2 X0
2
F1 = a0 + a1 X1 + a2 X1

From which:

′
( F1 – F0 ) ⁄ ( ( X1 – X0 ) – F0 )
a 2 = -----------------------------------------------------------------------
( X1 – X0 )
′
a 1 = F 0 – 2a 2 X 0
2
a0 = F0 – a1 X0 – a2 X0

Therefore

˜ ( X ) = a + 2a X = 0
F′ 1 2

∗ a1
X ≅ – ----------
2a 2

F˜ ′′ ( X ) = 2a 2

∗
Minimum at X if a2 > 0
∗
Maximum at X if a2 < 0

2-2
 MINIMIZATION OF FUNCTIONS OF A SINGLE
VARIABLE (Cont.)

• Unconstrained minimum:

F(X)

Fopt

∗ X
X0 X X1

• Constrained minimum:

F(X)
u
X≤X

Fopt

u X
X0 X X1

u
Side constraint X≤X

u
X∗ = X

2-3
 GRADIENTS

Multivariate function minimization in MSC/NASTRAN is gradient-

based.

Let:

f = f ( X 1 , X 2 , …, X n )

i.e., a scalar function of n independent variables.

Gradient of a scalar function is defined as:

∂f ∂f ∂f
∇f = ---------- ⋅ e 1 + ---------- ⋅ e 2 + … + ---------- ⋅ e n
∂X 1 ∂X 2 ∂X n

th
where e 1 , e 2 , …, e n are unit vectors in the n direction.

Gradient operator is defined as:

∂ ∂ ∂
∇ = ---------- e 1 + ---------- e 2 + … + ---------- e n
∂X 1 ∂X 2 ∂X n

or in tensor notation as:

∂
∇ = --------e i
∂X i

2-4
 GRADIENTS (Cont.)

• Gradient of a scalar → vector field

2 2
Example: f ( X 1 , X 2 ) = X 1 + X 2

(Note that this is the equation for a family of circles in the X1, X2
plane with centers at the origin and radii of f .)

∂f ∂f
∇f ( X 1 , X 2 ) = ---------- e 1 + ---------- e 2
∂X 1 ∂X 2
= 2X 1 e 1 + 2X 2 e 2

 2X 1 
or ∇f ( X 1 , X 2 ) =  
 2X 2 

f and ∇f are continuous functions

Graphically: X2
∇f ( X 1, X 2 )

X1

Note that ∇f points in the direction of increasing function values.

2-5
 UNCONSTRAINED MINIMIZATION METHODS

• Example: Equilibrium of two-spring system

• Theoretical solution:

2
1 2 2
PE = --- K 1 X1 + ( l 1 – X2 ) – l 1
2
2
1 2 2
+ --- K 2 X1 + ( l 2 – X2 ) – l 2 – P1 X1 – P2 X2
2

minimum occurs at

 ∂ ( PE ) 
 ---------------
∂X
-
 1 
∇ ( PE ) =   = 0
∂ ( )
 ---------------- 
PE
 ∂X 
 2 

PEmin = −41.81 N − cm

X1∗ = 8.631 cm

X2∗ = 4.533 cm

The corresponding two-variable design space is illustrated on the

next page.

2-6
 UNCONSTRAINED MINIMIZATION METHODS (Cont.)

• Steepest descent:

Reductions in the objective function are sought by searching in a

direction given by the negative of the gradient.

2
X

8
100

80 6
60 X*
0 +
40 X
1 4
S
−40
30 −20 −30
1 20 −10
8
X 2 3 X
5
10 X 6
2 X
S 2
X 4 1
X X
−4 −2 0 2 4 6 8 10 12

−2

Design updates given by:

q q–1 q
X = X + α∗ S

2-7
 UNCONSTRAINED MINIMIZATION METHODS (Cont.)

where

q–1
 
 ∂F 
 ---------- 
∂X 1
 
 ∂F 
q q–1  ---------- 
S = – ∇F ( X ) = – ∂X 2 
 .
. 
 . 
 ∂F 
 ---------- 
 ∂X 1 
 

• This method is popular, but inefficient. In MSC/NASTRAN we only

use it as the first iteration of other, more efficient methods.

2-8
 UNCONSTRAINED MINIMIZATION METHODS (Cont.)

• Conjugate direction method

First iteration (q = 1)

q q–1
S = – ∇F ( X )

th
q iteration (q > 1)

q q–1 q–1
S = – ∇F ( X ) + βS

q–1 2 (Bias the new search direction

∇F ( X )
where β = ---------------------------------- with previous information)
q–2 2
∇F ( X )

2
X

100
8
3
X
80 6
60
0
40 X +
1 4 X*
S
−40
30 −20 −30
1 20 8 −10
X
S
2
10 2 5
3 3 X
X
2S
X 1
−2 X
−4 0 2 4 6 8 10 12

−2

• This method is easy to code and is a dramatic improvement over

the steepest descent method.

2-9
 KUHN-TUCKER CONDITIONS FOR OPTIMALITY

Kuhn-Tucker conditions for optimality follow directly from a

generalization of Lagrange multipliers.

An optimum design is at hand if:

∗
1. X is feasible

∗ j = 1, …, m
gj ( X ) ≤ 0

∗ k = 1, …, l
hk ( X ) = 0

∗ j = 1, …, m
2. λj g ( X ) = 0
j

λj ≥ 0

m l
∗ ∗ ∗
3. ∇F ( X ) + ∑ λj ∇gj ( X )+ ∑ λk + m ∇hk ( X ) = 0
j=1 k=1

λ k + m unrestricted in sign, but not used in MSC/NASTRAN

2-10
 KUHN-TUCKER CONDITIONS FOR
OPTIMALITY (Cont.)

X2
F ( X ) = constant

∗
∇F ( X )

g1 ( X ) = 0

X1
∗
∇g 2 ( X )
∗
∇g 1 ( X ) g2 ( X ) = 0

∗
∇F ( X )

∗
X

∗ ∗
λ 2 ∇g 2 ( X ) λ 1 ∇g 1 ( X )

∗
–∇ F ( X )

2-11
 SIMPLE CANTILEVER EXAMPLE

• Problem description

P = 2250 N

A
L = 500 cm

Section A-A

6 N
E = 1 × 10 ------------
2
cm

2-12
 SIMPLE CANTILEVER EXAMPLE (Cont.)

Minimize V = B • H • L

Subject to:

Mc
σ = -------- ≤ 700 Bending Stress
l

3
PL Tip Deflection
δ = ----------- ≤ 2.54
3El

H
---- ≤ 12 Aspect Ratio
B

1 ≤ B ≤ 20

Gauge Requirements
20 ≤ H ≤ 50

2-13
 SIMPLE CANTILEVER EXAMPLE (Cont.)

The Design Space

H/B = 12
65

60
V = 200000

175000
55 100000 150000
Height H (cm)

125000
H = 50
50
Optimum

45

δ = 2.54
40 75000
50000
σb = 700

35
2.5 3 4 5 6 7 7.5
Width B (cm)

2-14
 SERIES APPROXIMATIONS

Function gradient information can be used to construct first-order

Taylor Series approximations

2 2
o o df ∆x
f ( x + ∆x ) = f ( x ) + ------- d
⋅ ∆x + ---------f- ⋅ ---------- + …
dx x o 2 2!
dx x o

˜f ( x o + ∆x ) = f ( x o ) + df 2
------- o ⋅ ∆x + 0 ( ∆x )
dx x

where

2 2
0 ( ∆x ) ≡ error on the order of ∆x

2-15
 SERIES APPROXIMATIONS (Cont.)

Using the Simple Cantilever to illustrate:

• Minimize V = B • H • L

• Design variables B and H

• Subject to:

MC 6PL N
σ = --------- = ------------ ≤ 700 ------------
I 2 2
BH cm
3 3
PL 4PL
σ = ----------- = ---------------- ≤ 2.54 cm
3EI 3
BH E

2-16
 SERIES APPROXIMATIONS (Cont.)

• First-order approximations:

o o o o
V˜ ( B + ∆B, H + ∆H, L ) = V ( B , H , L )

∂V ∂V
- o o ⋅ ∆B + ------- o o ⋅ ∆H
+ ------
∂B B , H ∂H B , H

o o o o
σ̃ ( B + ∆B, H + ∆H, L ) = σ̃ ( B , H , L )

∂σ ∂σ
- o o ⋅ ∆B + ------- o o ⋅ ∆H
+ ------
∂B B , H ∂H B , H

o o o o
δ̃ ( B + ∆B, H + ∆H, L ) = δ̃ ( B , H , L )

∂δ ∂δ
- o o ⋅ ∆B + ------- o o ⋅ ∆H
+ ------
∂B B , H ∂H B , H

o o
• At ( B , H ) = ( 6, 45 )

o o 5 4 3
V ( B + ∆B, H + ∆H, L ) = 1.35 × 10 + 2.25 × 10 ∆B + 3.0 × 10 ∆H

o o
σ ( B + ∆B, H + ∆H, L ) = 555.56 – 92.593∆B – 24.691∆H

o o
δ ( B + ∆B, H + ∆H, L ) = 2.0576 – 0.34294∆B – 0.13717∆H

2-17
 SERIES APPROXIMATIONS (Cont.)

• The resultant linearized design space

65 H/B = 12

60

175000

55 100000 150000
125000
Height H (cm)

H = 50
50
Optimum
Approximate
Optimum
0
45 X

75000
40 δ̃

σ̃

35
2.5 3 4 5 6 7 7.5

Width B (cm)

In this case, the approximate optimum actually violates some of the

true constraints. It does, however, form a good starting point for
the next iteration. This is the basis of a “sequential linearization”
method of solution.

2-18
 SEQUENTIAL LINEAR PROGRAMMING

• First linearize:

o o o
F˜ ( x + ∆x ) = F ( x ) + ∇F ( x ) ⋅ ∆x
o o o
g̃ j ( x + ∆x ) = g j ( x ) + ∇g j ( x ) ⋅ ∆x j ∈ J, (J ≡ set of
active constraints)
o
∆x = x – x

• Solve this linear approximate problem using Simplex or other

optimizer.

• Relinearize and repeat to convergence.

This option is available in MSC/NASTRAN. (Set METHOD to 2 on

the DOPTPRM Bulk Data entry.)

2-19
 SEQUENTIAL LINEAR PROGRAMMING (Cont.)

Typical design cycle history:

x2

o
∇F ( x )

o
x

1
x

∗
x

2
x
x1

Boxes around each design represent move limits.

2-20
 SEQUENTIAL LINEAR PROGRAMMING (Cont.)

• Why move limits are necessary:

X2 move limits

o
F˜ ( X )

unbounded linear approximation

true optimum

o
g̃ ( X ) g(X) = 0

0 X1

• Move limits are required because higher order terms have been
ignored

2-21
 METHOD OF FEASIBLE DIRECTIONS

• In contrast with SLP, the method of feasible directions treats

active, nonlinear constraints directly.

X2 o
∇F ( X )

o
∇g j ( X ) F = Constant

gj ( X )
X1

• Usable search direction:

∇F ⋅ S ≤ 0

• Feasible search direction:

∇g j ⋅ S ≤ 0

X2 Feasible

Usable/Feasible
S
Usable gj ( X )
X1

2-22
 METHOD OF FEASIBLE DIRECTIONS (Cont.)

Determination of S (direction finding subproblem)

• To maintain feasibility:

∇g j ⋅ S + θ j ≤ 0 (1)

where θ j is the push-off factor

• To maintain usability:

∇F ⋅ S + β ≤ 0 (2)

A nonzero θ j and β ensure a search direction that reduces the

th
objective function without violating the j constraint for some finite
move in that direction.

It is apparent that we would like to maximize β in order to make the

greatest possible reductions in the objective.

2-23
 METHOD OF FEASIBLE DIRECTIONS (Cont.)

• Sequence of iterations using cantilever design space as an

example:

H/B = 12

60

55
Height H (cm)

H = 50
50
Optimum

45 o
X

δ = 2.54
40

σ = 700

35
3 4 5 6 7
Width B (cm)

2-24
 METHOD OF FEASIBLE DIRECTIONS (Cont.)

• The problem with the method is that it may tend to zigzag between
active and inactive constraints as the optimum is approached.

X2
g1 ( X )

o
X g2 ( X )
1
X

3 F ( X ) = Constant
X 4 2
X X
X1

• In structural applications, unnecessary function evaluations must

be kept to a minimum.

• Faster convergence can usually be obtained if the constraint

surfaces are “followed” to the optimum.

2-25
 MODIFIED METHOD OF FEASIBLE DIRECTIONS

• Combines best features of the method of feasible directions and

generalized reduced gradients. This method is the MSC/NASTRAN
default

• In the direction-finding problem, all push-off factors θ j are set to

zero

Maximize – ∇F ⋅ s

Subject to: ∇g j ( X ) ⋅ s ≤ 0 j∈J

s⋅s≤1

2-26
 MODIFIED METHOD OF FEASIBLE
DIRECTIONS (Cont.)

• Search direction s (inequality constrained problem)

• If all g j < 0 j = 1, ..., m

s = – ∇F Direction of steepest descent

Use conjugate direction if unconstrained

for more than one iteration.

• If some gj = 0 j ∈ J , within a specified tolerance

(J = set of active constraints)

Direction finding subproblem:

maximize – ∇F ⋅ s

subject to:

∇g j ( X ) ⋅ s ≤ 0
for j ∈ J
s⋅s≤1

• If design is not feasible, use a large push-off factor for violated

constraints and find the search direction as in the method of
feasible directions.

2-27
 MODIFIED METHOD OF FEASIBLE
DIRECTIONS (Cont.)

One-dimensional Search Modifications

A move in direction S will go into the infeasible region for nonlinear

convex constraints. Using the gradients of the active constraints, a
step back to the constraint boundary is taken, so the design “follows”
the constraints.

• First take a step α

q q–1 q
X = X + αS

• Evaluate the constraints and perturb the proposed design back to

the constraint boundary by:

T T –1
δX = A [ AA ] G

T
where A contains the gradients of the active constraints and G
contains the constraint values.

This is effectively a “minimum norm” move back to the constraint

boundary and may need several applications.

2-28
 MODIFIED METHOD OF FEASIBLE
DIRECTIONS (Cont.)

• Features

• Rapidly obtains an optimum design

• Deals with both inequality and equality constraints

• Usually satisfies the constraints with high precision at the optimum

2-29
 MODIFIED METHOD OF FEASIBLE
DIRECTIONS (Cont.)

• Sequence of iterations using cantilever design space as an

example

H/B = 12

60

55
Height H (cm)

H = 50
50
Optimum

o
45 X

δ = 2.54
40

σ = 700

35
3 4 5 6 7
Width B (cm)

2-30
 MODIFIED METHOD OF FEASIBLE
DIRECTIONS (Cont.)

• Recovery from infeasible designs

x2
F = Constant

∇F
Minimize Constraint
Violation

o
∇G x
x1

Minimize Objective
Function Compromise Direction
x2

∗
x

x1
o
x

No Feasible Design

2-31
 NUMERICAL IDENTIFICATION OF ACTIVE AND
VIOLATED CONSTRAINTS

• Parameters of interest

• Gj(X) is inactive if Gj(X) < CT.

• Gj(X) is active if CT ≤ Gj(X)≤ CTMIN.

• Gj(X) is violated if Gj(X) > CTMIN.

• CT and CTMIN may both be changed using the DOPTPRM Bulk

Data entry.

x2

Feasible Region
gj ( x ) < 0

g j ( x ) = CT

Infeasible Region gj ( x ) = 0
gj ( x ) > 0 g j ( x ) = CTMIN
x1

2-32
 NUMERICAL IDENTIFICATION OF ACTIVE AND
VIOLATED CONSTRAINTS (Cont.)

Gj(X)
Feasible Violated

Active

CTMIN

X
0
CT Numerical Constraint
Boundary

Exact Constraint
Boundary

CT = −0.03

CTMIN = 0.003

2-33
 SECTION 3

DESIGN MODELING I

WHAT IS A DESIGN MODEL? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1

HOW ARE THE DESIGN AND ANALYSIS MODELS RELATED? . . . . . . . . . . . . . . 3-2

DESIGN AND ANALYSIS MODEL COMPARISON . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3

DESIGN MODEL DEFINITION PROCESS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-5

DEFINING THE ANALYSIS DISCIPLINES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-6

DEFINING THE DESIGN VARIABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-7

DESVAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-8

RELATING DESIGN VARIABLES TO PROPERTIES . . . . . . . . . . . . . . . . . . . . . . . 3-10

DVPREL1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-11

IDENTIFYING THE DESIGN RESPONSES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-15

DRESP1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-16

DEFINING THE OBJECTIVE FUNCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-24

DESOBJ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-25

DEFINING THE DESIGN CONSTRAINTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-27

DCONSTR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-28

DCONADD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-31

DESSUB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-33

DESGLB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-34

BASIC DESIGN PROCESS CONTROL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-38

DOPTPRM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-39
 WHAT IS A DESIGN MODEL? 3

A design model is a statement of allowable changes that can be made

in the search for an optimal design. It also defines what is meant by
improvement.

It must:

• Define the design variables

• Relate design variables to changes in property or shape

• Define the objective function (scalar measure of quality)

• Place bounds on responses

3-1
 HOW ARE THE DESIGN AND ANALYSIS MODELS
RELATED?

DESIGN MODEL

DESIGN DESIGN OBJECTIVE,

IMPROVEMENT VARIABLES CONSTRAINTS

CAD MODEL, GEOMETRY, LOADS,

PROTOTYPE PROPERTIES BOUNDARY RESPONSES
CONDITIONS

ANALYSIS MODEL

3-2
 DESIGN AND ANALYSIS MODEL COMPARISON

Consider a structure consisting of I-section beams:

h tw tf

Analysis model states properties and connectivity:

CBAR 101 21 ... ...

A, I1, I2, I12, J, ...
...

...

...

...

...
PBAR 21 6 2.36 ...

The design model is design variable based:

h, b, tw, tf

3-3
 DESIGN AND ANALYSIS MODEL
COMPARISON (Cont.)

Analysis model properties are functions of the design variables:

A = 2∗b∗t f + ( h – 2t f )∗t w

3 (b – t )
b∗h w 3
I 1 = -------------- – --------------------∗ ( h – 2∗t f )
12 12

Analysis Cross-sectional properties

Design Cross-section dimensions

3-4
DESIGN MODEL DEFINITION PROCESS

Flowchart for Design Modeling

Define Analysis Disciplines

(Executive Control: SOL 200;
Case Control: Analysis =)

Define Design Variables

(Bulk Data: DESVAR)

Relate Design Variables to

Analysis Model:
(Bulk Data: DVPREL1)

Define Design Responses

(Bulk Data: DRESP1)

Define Objective (Case Control: DESOBJ)

and Constraints (Bulk Data: DCONSTR, DCONADD;
Case Control: DESGLB, DESSUB)

Provide Optional Parameters

(Bulk Data: DOPTPRM)

3-5
 DEFINING THE ANALYSIS DISCIPLINES

• Executive Section

SOL 200

• Case Control Section

STATICS – statics
MODES – normal modes
BUCK – buckling
DFREQ – direct frequency*
Analysis =
MFREQ – modal frequency*
MTRAN – modal transient*
SAERO – static aeroelasticity
FLUTTER – flutter

*Includes acoustics

3-6
 DEFINING THE DESIGN VARIABLES

Design variables are the quantities the optimizer modifies in its

search for an improved design.

The DESVAR entry defines a design variable, its initial value, and
bounds. It also can provide design cycle move limits.

3-7
DESVAR Design Variable

DESVAR a design variable for design optimization.

Defines Design Variable

Format:

1 2 3 4 5 6 7 8 9 10
DESVAR ID LABEL XINIT XLB XUB DELXV

Example:

DESVAR 2 BARA1 35.0 10. 100. 0.2

Field Contents
ID Unique design variable identification number. (Integer > 0)

LABEL User-supplied name for printing purposes. (Character)

XINIT Initial value. (Real, XLB ≤ XINIT ≤ XUB)

XLB Lower bound. (Real, Default = –1.0E+20)

XUB Upper bound. (Real, Default= +1.0E+20)

DELXV Fractional change allowed for the design variable during approximate
optimization. (Real > 0.0, for Default see Remark 2.)

Remarks:
1. DELXV can be used to control the change in the design variable during one optimization
cycle.

2. If DELXV is blank, the default is taken from the specification of the DELX parameter on
the DOPTPRM entry. If DELX is not specified, then the default is 1.0.

3-8
 DEFINING THE DESIGN VARIABLES (Cont.)

Example:

$ESVAR, ID, LABEL, XINIT, XLB, XUB, DELXV

$
DESVAR, 10, AREA1, 0.05, 0.01, 0.1
DESVAR, 11, THICK1, 0.03, 0.01, 0.08
$

0.01 ≤ x 10 ≤ 0.1

0.01 ≤ x 11 ≤ 0.08

The entire set of DESVAR entries defines a vector of design variables:

T
X = [ x 1, x 2, …, x n ]

with corresponding bounds,

I u
xi ≤ xi ≤ xi

3-9
 RELATING DESIGN VARIABLES TO PROPERTIES

As the optimizer changes design variables, the analysis model

properties must also change.

The simplest relation is a linear one:

pj = Co + ∑ Ci xi
i

The above can be written using the DVPREL1 entry.

3-10
Design Variable to Property Relation DVPREL1
DVPREL1 the relation between an analysis model property and design variables.
Defines Design Variable to Property Relation

Format:

1 2 3 4 5 6 7 8 9 10
DVPREL1 ID TYPE PID FID PMIN PMAX C0

DVID1 COEF1 DVID2 COEF2 DVID3 -etc.-

Example:

DVPREL1 12 PBAR 612 6 0.2 3.0

4 0.25 20 20.0 5 0.3

Field Contents
ID Unique identification number. (Integer > 0)

TYPE Name of a property entry, such as “PBAR”, “PBEAM”, etc. (Character)

PID Property entry identification number. (Integer > 0)

FlD Field position of the property entry, or word position in the element property table
of the analysis model. (Integer ≠ 0)

PMIN Minimum value allowed for this property. If FID references a stress recovery
location, then the default value for PMIN is –1.0+35. PMIN must be explicitly set
to a negative number for properties that may be less than zero (for example, field
ZO on the PCOMP entry). (Real; Default = 1.0E–20)

PMAX Maximum value allowed for this property. (Real; Default = 1.0E–20)

C0 Constant term of relation. (Real; Default = 0.0)

DVIDi DESVAR entry identification number. (Integer > 0)

COEFi Coefficient of linear relation. (Real)

Remarks:
1. The relationship between the analysis model property and design variables is given by:

P i = C0 + ∑ COEFi ∗ DVIDi
i

2. The continuation entry is required.

(Continued)

3-11
DVPREL1 Design Variable to Property Relation

3. PTYPE = “PBEND” is not supported, either directly through FIDs or indirectly via word po-
sitions in the element property table.

4. FID may be either a positive or a negative number. If FID > 0, it identifies the field position
on a property entry. If FID < 0, it identifies the word position of an entry in the element
property table. For example, to specify the area of a PBAR, either FID = +4 or FID = –3
can be used. However, if PTYPE = “PBEAM”, FID must be negative. See the following
element property table for the word positions for PBEAM.

Table 1. EPT Section for PTYPE = “PBEAM”.

Word Type Item

1 I Property ID
2 I Material ID
3 I Number of segments
4 I Constant cross section flag (1 = yes, 0 = no)
5 I Unused
6 I Stress output request flag, SO (1 = yes, 0 = no)
7 R X/XB ratio; at end A, X/XB = 0.0
8 through 13 R A, I1, I2, I12, J, NSM
14 through 21 R C1, C2, D1, D2, E1, E2, F1, F2
22 I
23 R Repeat of words 6 through 21 for the 1st intermediate
24 through 29 R station
30 through 37 R
38 I
39 R
40 through 45 R 2nd intermediate station
46 through 53 R
54 I
55 R
56 through 61 R 3rd intermediate station
62 through 69 R
70 I
71 R
72 through 77 R 4th intermediate station
78 through 85 R
86 I
87 R
88 through 93 R 5th intermediate station
94 through 101 R

(Continued)

3-12
Design Variable to Property Relation DVPREL1

Word Type Item

102 I
103 R
104 through 109 R 6th intermediate station
110 through 117 R
118 I
119 R
120 through 125 R 7th intermediate station
126 through 133 R
134 I
135 R
136 through 141 R 8th intermediate station
142 through 149 R
150 I
151 R
152 through 157 R 9th intermediate station
158 through 165 R
166 I
167 R
168 through 173 R End B
174 through 181 R
182 through 189 R K1, K2, S1, S2, NSI(A), NSI(B), CW(A), CW(B)
190 through 197 R M1(A), M2(A), M1(B), M2(B), N1(A), N2(A), N1(B),
N2(B)

3-13
 RELATING DESIGN VARIABLES TO
PROPERTIES (Cont.)

Example:

Web
t i = 1.0 x 2
Web Cap
A i = 1.0 x 3

Plate
t i = 1.0 x 1

for the base plate thickness,

PSHELL, 1, 1, 0.15, 1
$
$...Define the design variables:
$
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
DESVAR, 1, T-PLATE,0.15, 0.001, 10.0
$
$...Relate the design variables to analysis model properties
$ (linear relations, so use DVPREL1)
$
$DVPREL1,ID, TYPE, PID, FID, PMIN, PMAX, C0, , +
$+, DVID1, COEF1, DVID2, COEF2, ...
DVPREL1,1, PSHELL, 1, 4, 0.01, , , , +DP1
+DP1, 1, 1.0

3-14
 IDENTIFYING THE DESIGN RESPONSES

Prior to defining the objective and constraint functions, the responses

on which they are based need to be identified.

The DRESP1 entry identifies responses that are computed from an

MSC/NASTRAN analysis.

3-15
DRESP1 Design Sensitivity Response Quantities

Defines a set of structural responses that is used in the designDesign

DRESP1 either as constraints
Sensitivity Responseor as an
Quantities
objective.

Format:

1 2 3 4 5 6 7 8 9 10
DRESP1 ID LABEL RTYPE PTYPE REGION ATTA ATTB ATT1

ATT2 -etc.-

Example:

DRESP1 1 DX1 STRESS PROD 2 3 102

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. See Remark 10 for defaults.
(Integer > 0)

ATTA, ATTB, Response attributes. See Table 1. (Integer > 0 or Real or blank)
ATTi

(Continued)

3-16
Design Sensitivity Response Quantities DRESP1
Table 1. Design Sensitivity Response Attributes.

Response Attributes
Response
Type ATTB (Integer > 0 or
(RTYPE) ATTA (Integer > 0) Real > 0.0) ATTi (Integer > 0)

WEIGHT Blank Blank SEIDi or ALL

VOLUME Blank Blank SEIDi or ALL

EIGN Normal Modes Mode Approximation Code. Blank

Number See Remark 19.

FREQ Normal Modes Mode Approximation Code. Blank

Number See Remark 19.
See Remark 18.
LAMA Buckling Mode Number Approximation Code. Blank
See Remark 19.

DISP Displacement Component Blank Grid ID

STRAIN Strain Item Code Blank Property entry (PID)

STRESS Stress Item Code Blank Property entry (PID)

FORCE Force Item Code Blank Property entry (PID)

CSTRAIN Strain Item Code LAMINA Number Property entry (PID)

(Integer; Default = 1)

CSTRESS Stress Item Code LAMINA Number Property entry (PID)

(Integer; Default = 1)

CFAILURE Failure Criterion Item LAMINA Number Property entry (PID)

Code (Integer; Default = 1)

FRDISP Displacement Component Frequency Value. (Real > Grid ID

0.0) See Remark 15.

FRVELO Velocity Component Frequency Value. (Real > Grid ID

0.0) See Remark 15.

FRACCL Acceleration Component Frequency Value. (Real > Grid ID

0.0) See Remark 15.

FRSPCF SPC Force Component Frequency Value. (Real > Grid ID

0.0) See Remark 15.

FRSTRE Stress Item Code Frequency Value. (Real > Property entry (PID)
0.0) See Remark 15..

(Continued)

3-17
DRESP1 Design Sensitivity Response Quantities

Response Attributes
Response
Type ATTB (Integer > 0 or
(RTYPE) ATTA (Integer > 0) Real > 0.0) ATTi (Integer > 0)

FRFORC Force Item Code Frequency Value. (Real > Property entry (PID)
0.0) See Remark 15.

TDISP Displacement Component Time Value. (Real) See Grid ID

Remark 16.

TVELO Velocity Component Time Value. (Real) See Grid ID

Remark 16.

TACCL Acceleration Component Time Value. (Real) See Grid ID

Remark 16.

TSPCF SPC Force Component Time Value. (Real) See Grid ID

Remark 16.

TSTRE Stress Item Code Time Value. (Real) See Property entry (PID)
Remark 16.

TFORC Force Item Code Time Value. (Real) See Property entry (PID)
Remark 16.

TRIM AESTAT or AESURF Blank Blank

Entry ID

STABDER AESTAT or AESURF Restraint Flag. (Integer 0 Component

Entry ID or 1) See Remark 13.

FLUTTER Blank Method (PK or PKNL, See Remark 14.

Default = PK).

Remarks:
1. Stress, strain, and force item codes can be found in the MSC/NASTRAN Quick Reference
Guide, Appendix A. For stress or strain item codes that have dual meanings, such as von
Mises or maximum shear, the option specified in the Case Control Section will be used;
i.e., STRESS(VONM) or STRESS(MAXS).

2. RTYPE=“CSTRESS”, “CSTRAIN”, and “CFAILURE” are used only with the PCOMP entry.
“CSTRESS” and “CSTRAIN” item codes are described under Table 1. (Element
Stress/Strain Item Codes) in the MSC/NASTRAN Quick Reference Guide, Appendix A.
“CFAILURE” item codes are described under Table Table 2. (Element Force Item Codes)
in the MSC/NASTRAN Quick Reference Guide, Appendix A. Only force item codes that
refer to failure indices of direct stress and interlaminar shear stress are valid.

(Continued)

3-18
Design Sensitivity Response Quantities DRESP1
The CFAILURE response type requires the following specifications on the applicable
entries:

a. Failure theory in the FT field on PCOMP entry.

b. Allowable bonding shear stress in the SB field on PCOMP entry.

c. Stress limits in the ST, SC, and SS fields on all MATi entries.

3. ATTB is used only for responses of composite laminae, dynamics, and stability deriva-
tives. For other responses, this field must be blank.

4. All grids associated with a DRESP1 entry are considered to be in the same region for
screening purposes. Only up to NSTR displacement constraints (see DSCREEN entry)
per group per load case will be retained in the design optimization phase.

5. DRESP1 identification numbers must be unique with respect to DRESP2 identification

numbers.

6. If PTYPE = “ELEM”, the ATTi correspond to element identification numbers.

7. If RTYPE = “DISP”, “TDISP”, “TVELO”, “TACCL” or “TSPCF”, multiple component num-

bers (any unique combination of the digits 1 through 6 with no embedded blanks) may be
specified on a single entry. Multiple response components may not be used on any other
response types.

8. If RTYPE = “FRDISP”, “FRVELO”, “FRACCL”, or “FRSPCF” only one component number

may be specified in the ATTA field. Numbers 1 through 6 correspond to real (or magni-
tude) components and 7 through 12 imaginary (or phase) components. If more than one
component for the same grid is desired, then a separate entry is required.

9. Real/imaginary representation is the default for complex response types. Magni-

tude/phase representation must be requested by the corresponding Case Control
command; e.g., DlSP(PHASE) = ALL.

10. REGION is used for constraint screening. The NSTR field on DSCREEN entries gives the
maximum number of constraints retained for each region per load case.

IF RTYPE = “WEIGHT”, “VOLUME”, “LAMA”, “EIGN” or “FREQ”, no REGION identification

number should be specified. For all other responses, if the REGION field is left blank, the
default specified in Table 2 is used. Usually, the default value is appropriate.

If the REGION field is not blank, all the responses on this entry as well as all responses on
other DRESP1 entries that have the same RTYPE and REGION identification number will
be grouped into the same region.

(Continued)

3-19
DRESP1 Design Sensitivity Response Quantities

Table 2. Default Regions for Design Sensitivity Response Types.

Response Type Default Region

WEIGHT No region

VOLUME No region

LAMA No region

EIGN No region

FREQ No region

DISP One region per DRESP1 entry

FRDISP One region per DRESP1 entry

FRVELO One region per DRESP1 entry

FRACCL One region per DRESP1 entry

FRSPCF One region per DRESP1 entry

TDISP One region per DRESP1 entry

TVELO One region per DRESP1 entry

TACCL One region per DRESP1 entry

TSPCF One region per DRESP1 entry

FLUTTER One region per DRESP1 entry

OTHER One region per PROPERTY entry. If PTYPE = “ELEM”,

then one region per DRESP1 entry

11. REGION is valid only among the same type of responses. Responses of different types
will never be grouped into the same region, even if they are assigned the same REGION
identification number by the user.

12. If RTYPE = “WEIGHT” or “VOLUME”, field ATTi = “ALL” implies total weight/volume of all
superelements except external superelements.

13. RTYPE = “STABDER” identifies a stability derivative response. ATTB is the restraint flag
for the stability derivative. ATTB = 0 means unrestrained, and ATTB = 1 means re-
strained. For example, ATTA = 4000, ATTB = 0, and ATT1 = 3 reference the unrestrained
Cz derivative for the AESTAT (or AESURF) entry ID = 4000.

(Continued)

3-20
Design Sensitivity Response Quantities DRESP1
14. RTYPE = “FLUTTER” identifies a set of damping responses. The set is specified by ATTi:

ATT1 = Identification number of a SET1 entry that specifies a set of modes.

ATT2 = Identification number of an FLFACT entry that specifies a list of densities.

ATT3 = Identification number of an FLFACT entry that specifies a list of Mach numbers.

ATT4 = Identification number of an FLFACT entry that specifies a list of velocities.

15. For RTYP = “FRDISP”, “FRVELO”, “FRACCL”, “FRSPCF”, “FRFORC”, and “FRSTRE”,
ATTB specifies a frequency value in cycles per unit time. If ATTB is specified, then the
responses are evaluated at the closest frequency selected by the OFREQ command. The
default for ATTB is all frequencies selected by the OFREQ command.

16. For RTYPE = “TDISP”, “TVELO”, “TACCL”, “TSPCF”, “TFORC”, and “TSTRE”, ATTB
specifies a time value. If ATTB is specified, then the responses are evaluated at the clos-
est time selected by the OTIME command. The default for ATTB is all time steps selected
by the OTIME command.

17. Intermediate station responses on CBAR elements due to PLOAD1 and/or CBARAO en-
tries may not be defined on the DRESP1 entry.

18. RTYPE = “EIGN” refers to normal modes response in terms of eigenvalue (radi-
an/time)∗∗2 while RTYPE = “FREQ” refers to normal modes response in terms of natural
frequency or units of cycles per unit time.

19. For RTYPE = LAMA, EIGN or FREQ, the response approximation used for optimization
can be individually selected. (Approximation Code = 1 = direct linearization, = 2 = Inverse
Linearization).

3-21
 IDENTIFYING THE DESIGN RESPONSES (Cont.)

Example: Selecting Displacements

$
$DRESP1, ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
$
$ TRANSLATIONAL DISPLACEMENTS AT GRIDS 501-505:
$
DRESP1, 201, TXYZ1, DISP, , , 123, , 501, +
+, 502, 503, 504, 505

Example: Selecting Stresses

$
$DRESP1, ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
$
$ AXIAL STRESSES FOR ROD ELEMENTS IN PID GROUPS 150,160,170,180:
$
DRESP1, 250, SIG1, STRESS, PROD, , 2, , 150, +
+, 160, 170, 180

3-22
 IDENTIFYING THE DESIGN RESPONSES (Cont.)

A single DRESP1 entry can produce a large number of design

responses:

EID1
EID2
PID1 .
.
.
EIDn
PID2

DRESP1 .
.
.

PIDm

3-23
 DEFINING THE OBJECTIVE FUNCTION

The objective function must be a single scalar quantity. It is identified

by:

DESOBJ DRESP1 (or DRESP2)

(Case Control
command)

3-24
Design Objective DESOBJ
DESOBJ the DRESP1 or DRESP2 entry to be used as the design objective.
Selects Design Objective

Format:

MAX
DESOBJ ( ) = N
MIN

Examples:
DESOBJ = 10
DESO = 25

Describer Meaning
MIN Specifies that the objective is to be minimized.

MAX Specifies that the objective is to be maximized.

n Set identification of a DRESP1 or DRESP2 Bulk Data entry. (Integer > 0)

Remarks:
1. A DESOBJ command is required for a design optimization task and is optional for a sensi-
tivity task.

2. If the DESOBJ command is specified within a SUBCASE, the identified DRESPi Bulk Data
entry uses a response only from that subcase. If DESOBJ appears above all SUBCASE
commands and there are multiple subcases, it uses a global response.

3. The referenced DRESPi entry must define a scalar response.

3-25
 DEFINING THE OBJECTIVE FUNCTION (Cont.)

Example:

$ FOR WEIGHT MINIMIZATION,

$
$ IN CASE CONTROL:
$
DESOBJ(MIN) = 10
$
$ IN BULK DATA:
$
$DRESP1, ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
$
DRESP1, 10, WEIGHT, WEIGHT, , , , , ALL
$ ^^^
$ (DEFAULT IS 0: JUST THE RESIDUAL STRUCTURE)
$

A common error is to reference a DRESP1 that identifies multiple

responses.

3-26
 DEFINING THE DESIGN CONSTRAINTS

Constraints are formed by placing bounds on responses.

L U
rj ≤ rj ( x ) ≤ rj

The DCONSTR entry supplies these bounds.

3-27
DCONSTR Design Constraints

DCONSTR
Define design constraints. Design Constraints

Format:

1 2 3 4 5 6 7 8 9 10
DCONSTR DCID RID LALLOW UALLOW

Example:

DCONSTR 10 4 1.25

Field Contents
DCID Design constraint set identification number. (Integer > 0)

RID DRESPi entry identification number. (Integer > 0)

LALLOW Lower bound on the response quantity. (Real, Default = –1.0E20)

UALLOW Upper bound on the response quantity. (Real, Default = 1.0E20)

Remarks:
1. The DCONSTR entry may be selected in the Case Control Section by the DESSUB or
DESGLB command.

2. DCID may be referenced by the DCONADD Bulk Data entry.

3. For a given DCID, the associated RID can be referenced only once.

4. The units of LALLOW and UALLOW must be consistent with the referenced response de-
fined on the DRESPi entry. If RID refers to an “EIGN” response, then the imposed bounds
2
must be expressed in units of eigenvalue, (radian/time) . If RID refers to a “FREQ” re-
sponse, then the imposed bounds must be expressed in cycles/time.

5. LALLOW and UALLOW are unrelated to the stress limits specified on the MATi entry.

6. Constraints are computed as follows:

LALLOW – r
g = -------------------------------- for lower bound constraints
GNORM

r – UALLOW
g = --------------------------------- for upper bound constraints
GNORM

where r is the response defined on the DRESPi entry and if the RID refers to a “FREQ”
response, then the imposed bounds must be expressed in cycles/time.

(Continued)

3-28
Design Constraints DCONSTR

LALLOW for lower bounds if LALLOW > GSCAL

GNORM = UALLOW for upper bounds if UALLOW > GSCAL
GSCAL otherwise

GSCAL is specified on the DOPTPRM entry (Default = 0.001)

7. As Remark 6 indicates, small values of UALLOW and LALLOW require special processing
and should be avoided. Bounds of exactly zero are particularly troublesome. This can be
avoided by using a DRESP2 entry that offsets the constrained response from zero.

3-29
 DEFINING THE DESIGN CONSTRAINTS (Cont.)

Constraint sets can be merged using the DCONADD entry.

3-30
Design Constraint Set Combination DCONADD
DCONADDthe design constraints for a subcase as a union of DCONSTRDesign
Defines entries.
Constraint Set Combination

Format:

1 2 3 4 5 6 7 8 9 10
DCONADD DCID DC1 DC2 DC3 DC4 DC5 DC6 DC7

DC8 -etc.-

Example:

DCONADD 10 4 12

Field Contents
DCID Design constraint set identification number. (Integer > 0)

DCi DCONSTR entry identification number. (Integer > 0)

Remarks:
1. The DCONADD entry is selected by a DESSUB or DESGLB Case Control command.

2. All DCi must be unique from other DCi and DCID.

3-31
 DEFINING THE DESIGN CONSTRAINTS (Cont.)

Constraint sets are selected in Case Control using either:

DESSUB Subcase-dependent constraints

DESGLB Global constraints (subcase independent, e.g.,

WEIGHT, VOLUME, or DRESP2s that are not functions
of DRESP1s)

3-32
Design Constraints Request at the Subcase Level DESSUB
DESSUBthe design constraints to be used in a design optimization
Select task for the
Design Constraints current
Request subcase.
at the Subcase Level

Format:
DESSUB = n

Examples:
DESSUB = 10
DESS = 25

Describer Meaning
n Set identification of a DCONSTR or DCONADD Bulk Data entry identification
number. (Integer ≥ 0)

Remark:
1. A DESSUB command is required for every subcase for which constraints are to be
applied.

3-33
DESGLB Request Design Constraints at the Global Level

DESGLB the design constraints to be applied at the global level

Selects in a design
Request optimization
Design Constraints task.
at the Global Level

Format:
DESGLB = n

Examples:
DESGLB = 10
DESG = 25

Describer Meaning
n Set identification of a DCONSTR or DCONADD Bulk Data entry identification
number. (Integer > 0)

Remarks:
1. A DESGLB command is optional and invokes constraints that are to be applied indepen-
dent of a particular subcase. These constraints could be based on responses that are
independent of subcases (e.g., WEIGHT or VOLUME).

2. The DESGLB command can be used to invoke constraints that are not a function of
DRESP1 entries; e.g., DRESP2 responses that are not functions of DRESP1 responses
are subcase independent.

3-34
 DEFINING THE DESIGN CONSTRAINTS (Cont.)

Example: Placing Bounds on a Bar Stress (Axial+Bending):

$
$ IN CASE CONTROL:
$
SUBCASE 20
ANALYSIS = STATICS
DESSUB = 100
(ETC.)
$
$ IN BULK DATA:
$
$DRESP1, ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
$
$ MAXIMUM STRESS AT END A:
DRESP1, 110, SAMAX, STRESS, PBAR, , 7, , 2010, +
+ 2020, 2030
$
$ MAXIMUM STRESS AT END B:
DRESP1, 120, SBMAX, STRESS, PBAR, , 14, , 2010, +
+ 2020, 2030
$
$ BOUNDS APPLIED TO BOTH ENDS A & B:
$DCONSTR,DCID, RID, LALLOW, UALLOW
DCONSTR,100, 110, -85., +85. $ N/mm**2 untreated struct. steel,
DCONSTR,100, 120, -85., +85. $ FS=2 on fatigue limit
$

• 7 and 14 pick the maximum of four stresses at ends A and B of bar

elements

3-35
 DEFINING THE DESIGN CONSTRAINTS (Cont.)

A single DCONSTR can place bounds on a large number of

responses.

EID1
EID2
PID1 .
.
.
PID2
EIDn
DCONSTR DRESP1 .
.
.
PIDm

3-36
 DEFINING THE DESIGN CONSTRAINTS (Cont.)

Constraints are implemented in a normalized form in the code.

L
rj – rj ( x )
g 2j – 1 ( x ) = ------------------------- ≤ 0
L
rj

U
rj ( x ) – rj
g 2j ( x ) = -------------------------- ≤ 0
U
rj

Normalization provides:

• Expression of all constraints in standard form (basic optimization

problem statement)

• Ranking of constraints regardless of response magnitude

It is best to use reasonable bounds, avoiding 0.0 if possible.

3-37
 BASIC DESIGN PROCESS CONTROL

Basic design optimization process control is provided by the

DOPTPRM entry, and the parameters NASPRT and DESPCH.

• Override of design process control parameters can be done using

the DOPTPRM Bulk Data entry.

3-38
Design Optimization Parameters DOPTPRM
DOPTPRM default values of parameters used in design optimization.
Overrides Design Optimization Parameters

Format:

1 2 3 4 5 6 7 8 9 10
DOPTPRM PARAM1 VAL1 PARAM2 VAL2 PARAM3 VAL3 PARAM4 VAL4

PARAM5 VAL5 -etc.-

Example:

DOPTPRM IPRINT 5 DESMAX 10

Field Contents
PARAMi Name of the design optimization parameter. Allowable names are given in
Table 1. (Character)
VALi Value of the parameter. (Real or Integer, see Table 1.)

Remarks:
1. Only one DOPTPRM entry is allowed in the Bulk Data Section.

Table 1. PARAMi Names and Descriptions.

Name Description, Type, and Default Value

APRCOD Approximation method to be used. 1 = Direct Linearization; 2 = Mixed Method

based on response type; 3 = Convex Linearization. APRCOD = 1 is recommended
for shape optimization problems. (Integer 1, 2, or 3; Default = 2)

CONV1 Relative criterion to detect convergence. If the relative change in objective

between two optimization cycles is less than CONV1, then optimization is termi-
nated. (Real > 0.0; Default = 0.001)

CONV2 Absolute criterion to detect convergence. If the absolute change in objective

between two optimization cycles is less than CONV2, then optimization is termi-
nated. (Real > 0.0; Default = 1.0E-20)

CONVDV Relative convergence criterion on design variables. (Real > 0.0; Default = 0.001)

CONVPR Relative convergence criterion on properties. (Real > 0.0; Default = 0.001)

(Continued)

3-39
DOPTPRM Design Optimization Parameters

Name Description, Type, and Default Value

CT Constraint tolerance. Constraint is considered active if current value is greater

than CT. (Real < 0.0; Default = –0.03)

CTMIN Constraint is considered violated if current value is greater than CTMIN.

(Real > 0.0; Default = 0.003)

DABOBJ Maximum absolute change in objective between ITRMOP consecutive iterations

(see ITRMOP) to indicate convergence at optimizer level. F0 is the initial objective
function value. (Real > 0.0; Default = MAX[0.00 ∗ ABS(F0), 0.0001])

DELB Relative finite difference move parameter. (Real > 0.0; Default = 0.0001)

DELOBJ Maximum relative change in objective between ITRMOP consecutive iterations to

indicate convergence at optimizer level. (Real > 0.0; Default = 0.001)

DELP Fractional change allowed in each property during any optimization design cycle.
This provides constraints on property moves. (Real > 0.0; Default = 0.2)

DELX Fractional change allowed in each design variable during any optimization cycle.
(Real > 0.0; Default = 1.0)

DESMAX Maximum number of design cycles to be performed. (Integer > 0; Default = 5)

DOBJ1 Relative change in objective attempted on the first optimization iteration. Used to
estimate initial move in the one-dimensional search. Updated as the optimization
progresses. (Real > 0.0; Default = 0.1)

DOBJ2 Absolute change in objective attempted on the first optimization iteration. (Real
> 0.0; Default = 0.2 ∗ (F0))

DPMIN Minimum move limit imposed. (Real > 0.0; Default = 0.01)

DX1 Maximum relative change in a design variable attempted on the first optimization
iteration. Used to estimate the initial move in the one dimensional search. Updated
as the optimization progresses. (Real > 0.0; Default = 0.01)

DX2 Absolute change in a design variable attempted on the first optimization iteration.
(Real > 0.0; Default = 0.2 ∗ MAX[X(I)])

DXMIN Minimum design variable move limit (Real > 0.0; Default = 0.05).

GMAX Maximum constraint violation allowed at the converged optimum. (Real > 0.0;
Default = 0.005)

GSCAL Constraint normalization factor. See Remarks under the DSCREEN and
DCONSTR entries. (Real > 0.0; Default = 0.001)

(Continued)

3-40
Design Optimization Parameters DOPTPRM

Name Description, Type, and Default Value

IGMAX If IGMAX = 0, only gradients of active and violated constraints are calculated. If
IGMAX > 0, up to NCOLA gradients are calculated including active, violated, and
near active constraints. (Integer > 0; Default = 0)

IPRINT Print control during approximate optimization phase. Increasing values represent
increasing levels of optimizer information. (0 ≤ Integer ≤ 7; Default = 0)
0 no output (Default)
1 internal optimization parameters, initial information, and results
2 same, plus objective function and design variables at each iterations
3 same, plus constraint values and identification of critical constraints
4 same, plus gradients
5 same, plus search direction
6 same, plus scaling factors and miscellaneous search information
7 same, plus one dimensional search information

IPRNT1 If IPRNT1 = 1, print scaling factors for design variable vector. (Integer 0 or 1;
Default = 0)

IPRNT2 If IPRNT2 = 1, print miscellaneous search information. If IPRNT2 = 2, turn on print

during one-dimensional search process. (Warning: This may lead to excessive
output.) (Integer 0, 1, or 2; Default = 0)

ISCAL Design variables are rescaled every ISCAL iterations. Set ISCAL= –1 to turn off
scaling. (Integer; Default = NDV (number of design variables))

ITMAX Maximum number of iterations allowed at optimizer level during each design cycle.
(Integer; Default = 40)

ITRMOP Number of consecutive iterations for which convergence criteria must be satisfied
to indicate convergence at the optimizer level. (Integer; Default = 2)

ITRMST Number of consecutive iterations for which convergence criteria must be met at
the optimizer level to indicate convergence in the Sequential Linear Programming
Method. (Integer > 0; Default = 2)

JTMAX Maximum number of iterations allowed at the optimizer level for the Sequential
Linear Programming Method. This is the number of linearized subproblems
solved. (Integer ≥ 0; Default = 20)

JPRINT Sequential Linear Programming subproblem print. If JPRINT > 0, IPRINT is turned
on during the approximate linear subproblem. (Default = 0)

JWRITE If JWRITE > 0, file number on which iteration history will be written. (Integer > 0;
Default = 0)

(Continued)

3-41
DOPTPRM Design Optimization Parameters

Name Description, Type, and Default Value

METHOD Optimization Method: (Integer 1, 2, or 3; Default = 1)

1: Modified Method of Feasible Directions. (Default)
2: Sequential Linear Programming
3: Sequential Quadratic Programming

P1 Print control items specified for P2. (Integer ≥ 0; Default = 0) Initial results are
always printed prior to the first approximate optimization. If an optimization task
is performed, final results are always printed for the final analysis unless
PARAM,SOFTEXIT,YES is specified. These two sets of print are not controllable.
n: Print at every n-th design cycle.

P2 Items to be printed according to P1: (Integer; Default = 1)

0: No print.
1: Print objective and design variables. (Default)
2: Print properties.
4: Print constraints.
8: Print responses.
16: Print weight as a function of a material ID (note that there is not a design
quantity so that only inputs to the approximate design are available).
n: Sum of desired items. For example, P2 = 10 means print properties and
responses.

PTOL Maximum tolerance on differences allowed between the property values on

property entries and the property values calculated from the design variable values
on the DESVAR entry (through DVPRELi relations). PTOL is provided to trap ill-
posed design models. (The minimum tolerance may be specified on user
parameter DPEPS. See MSC/NASTRAN Quick Reference Guide, Section 6.)
(Real > 0.0; Default = 1.0E+35)

STPSCL Scaling factor for shape finite difference step sizes, to be applied to all shape
design variables. (Real > 0.0; Default = 1.0)

3-42
 BASIC DESIGN PROCESS CONTROL (Cont.)

Some basic, useful parameters on this entry include:

DESMAX Maximum allowable number of design cycles

P1 Frequency of optimization summary output

P2 Quantity of optimization summary output

IPRINT Quantity of optimizer output

METHOD Numerical optimization method

• The frequency of MSC/NASTRAN data recovery is determined by:

PARAM,NASPRT,n

Output consists of standard MSC/NASTRAN output (Case Control

output requests only).

n<0 no data recovery

n=0 first and last design cycle (default)

n>0 n-multiples of design cycles, plus the last

3-43
 BASIC DESIGN PROCESS CONTROL (Cont.)

• The frequency of design optimization punch output is determined

by:

PARAM,DESPCH,n

Output consists of updated DESVAR and GRID entries (for shape).

If mode tracking is performed, updated DRESP1 entries may be
written in the *.pch file.

n<0 never

n=0 final design cycle only (default)

n>0 n-multiples of design cycles, plus the last

3-44
 SECTION 4

EXAMPLES I

EXAMPLE – 3-BAR TRUSS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1

10-BAR TRUSS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-6

10-BAR TRUSS RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-10

BEAM VIBRATION FREQUENCY AS A DESIGN CONSTRAINT

(M. J. Turner 1970) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-15

BEAM VIBRATION FREQUENCY AS A DESIGN CONSTRAINT RESULTS . . . . . 4-22

 EXAMPLE – 3-BAR TRUSS 4

This is similar to the Three-Bar Truss example in Section 7.1 of the

MSC/NASTRAN Design Sensitivity and Optimization User’s Guide).

• Design model description

• Design variables – cross-sectional areas A1 and A2

• Objective – weight minimization

• Constraints:

• Stress allowable: 20 ksi tension

15 ksi compression

• Displacements at grid 4: x direction ± 0.2 in

y direction ± 0.2 in

4-1
 3-BAR TRUSS (Cont.)

10" 10"

2 3
1
x
A1 A2
3
1 10"
2
A1

Subcase 2 Subcase 1
20,000 lbs 20,000 lbs
x: –16,000 lbs x: 16,000 lbs
y: –12,000 lbs y: –12,000 lbs

4-2
 3-BAR TRUSS (Cont.)

ID MSC, D200X1
TIME 10 $
SOL 200 $ OPTIMIZATION
CEND
TITLE = SYMMETRIC THREE BAR TRUSS DESIGN OPTIMIZATION - D200X1
SUBTITLE = BASELINE - 2 CROSS SECTIONAL AREAS AS DESIGN VARIABLES
ECHO = SORT
SPC = 100
DISP = ALL
STRESS = ALL
DESOBJ(MIN) = 20 $ (DESIGN OBJECTIVE = DRESP ID)
DESSUB = 21 $ DEFINE CONSTRAINT SET FOR BOTH SUBCASES
ANALYSIS = STATICS
SUBCASE 1
LABEL = LOAD CONDITION 1
LOAD = 300
SUBCASE 2
LABEL = LOAD CONDITION 2
LOAD = 310
BEGIN BULK
$
$------------------------------------------------------------------------
$ ANALYSIS MODEL
$------------------------------------------------------------------------
$
$ GRID DATA
$ 2 3 4 5 6 7 8 910
GRID, 1, , -10.0 , 0.0, 0.0
GRID, 2, , 0.0 , 0.0, 0.0
GRID, 3, , 10.0 , 0.0, 0.0
GRID, 4, , 0.0 , -10.0, 0.0
$ SUPPORT DATA
SPC, 100, 1, 123456, , 2, 123456
SPC, 100, 3, 123456, , 4, 3456
$ ELEMENT DATA
CROD, 1, 11, 1, 4
CROD, 2, 12, 2, 4
CROD, 3, 11, 3, 4
$ PROPERTY DATA
PROD, 11, 1, 1.0
PROD, 12, 1, 2.0
MAT1, 1, 1.0E+7, , 0.33, 0.1
$ EXTERNAL LOADS DATA
FORCE, 300, 4, , 20000., 0.8, -0.6
FORCE, 310, 4, , 20000., -0.8, -0.6
$
$------------------------------------------------------------------------
$ DESIGN MODEL
$------------------------------------------------------------------------
$

4-3
 3-BAR TRUSS (Cont.)

$...DESIGN VARIABLE DEFINITION

$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV(OPTIONAL)
DESVAR, 1, A1, 1.0, 0.1, 100.0
DESVAR, 2, A2, 2.0, 0.1, 100.0
$
$...DEFINITION OF DESIGN VARIABLE TO ANALYSIS MODEL PARAMETER RELATIONS
$DVPREL1,ID, TYPE, PID, FID, PMIN, PMAX, C0, , +
$+, DVID1, COEF1, DVID2, COEF2, ...
DVPREL1,10, PROD, 11, 4, , , , , +DP1
+DP1, 1, 1.0
DVPREL1,20, PROD, 12, 4, , , , ,+DP2
+DP2, 2, 1.0
$
$...STRUCTURAL RESPONSE IDENTIFICATION
$DRESP1,ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
DRESP1, 20, W , WEIGHT
DRESP1, 21, U4, DISP , , , 1, , 4
DRESP1, 22, V4, DISP , , , 2, , 4
DRESP1, 23, S1, STRESS, PROD, , 2, , 11
DRESP1, 24, S2, STRESS, PROD, , 2, , 12
$
$...CONSTRAINTS
$DCONSTR,DCID, RID, LALLOW, UALLOW
DCONSTR,21, 21, -0.20 ,0.20
DCONSTR,21, 22, -0.20 ,0.20
DCONSTR,21, 23, -15000.,20000.
DCONSTR,21, 24, -15000.,20000.
$
$...OPTIMIZATION CONTROL:
$
DOPTPRM,IPRINT, 5, DESMAX, 10, DELP, 0.5, P1, 1, +
+, P2, 15
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......0
ENDDATA

4-4
 3-BAR TRUSS (Cont.)

***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************
(HARD CONVERGENCE ACHIEVED)
NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 6
NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 5
OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY

-------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
CYCLE APPROXIMATE EXACT OF OF
NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
--------------------------------------------------------------------------------------------------
INITIAL 4.828427E+00 -3.234952E-01
1 3.007897E+00 3.008492E+00 -1.977251E-04 -3.737402E-03
2 2.821953E+00 2.821638E+00 1.118734E-04 -1.967246E-02
3 2.734469E+00 2.734299E+00 6.217039E-05 -7.241016E-03
4-5

4 2.708915E+00 2.708921E+00 -2.024285E-06 -2.369141E-04

5 2.702065E+00 2.702063E+00 7.941219E-07 -2.666992E-04

--------------------------------------------------------------------------------------------------

DESIGN VARIABLE HISTORY

-------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
-------------------------------------------------------------------------------------------------------------------------
1 | 1 | A1 | 1.0000E+00 : 7.1020E-01 : 7.8436E-01 : 8.1374E-01 : 8.1739E-01 : 8.3569E-01 :
2 | 2 | A2 | 2.0000E+00 : 9.9976E-01 : 6.0313E-01 : 4.3271E-01 : 3.9699E-01 : 3.3838E-01 :
-------------------------------------------------------------------------------------------------------------------------
*** USER INFORMATION MESSAGE 6464 (DOM12E)
RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 5.
 10-BAR TRUSS

• Design model description

• Design variables – 10 cross-sectional areas

• Objective – structural weight

• Constraints:

• Tensile stress ≤ 25,000 psi

• Compressive stress ≥ –25,000 psi

360" 360"
5 3 1
1 2
7 10

5 6
360"
8 9

3 4
6 4 2

100,000 lbs 100,000 lbs

Note: Several variations of this problem appear in the structural op-

timization literature. Multiple load paths are available to sup-
port the loads at grids 2 and 4.

4-6
 10-BAR TRUSS (Cont.)

ID MSC, D200X2
TIME 10
SOL 200 $ OPTIMIZATION
CEND
$
TITLE = OPTIMIZATION TEST CASE 2
SUBTITLE = TEN BAR TRUSS D200X2
DESOBJ = 11 $ OBJECTIVE FUNCTION DEFINITION
DESSUB = 1 $ CONSTRAINT SET SELECTION
ANALYSIS = STATICS
DISP = ALL
STRESS = ALL
SPC = 100
LOAD = 300
BEGIN BULK
$
$------------------------------------------------------------------------
$ ANALYSIS MODEL:
$------------------------------------------------------------------------
$
GRDSET, , , , , , , 3456
GRID, 1, , 720., 360., 0.
GRID, 2, , 720., 0., 0.
GRID, 3, , 360., 360., 0.
GRID, 4, , 360., 0., 0.
GRID, 5, , 0., 360., 0.
GRID, 6, , 0., 0., 0.
CROD, 1, 1, 5, 3
CROD, 2, 2, 3, 1
CROD, 3, 3, 6, 4
CROD, 4, 4, 4, 2
CROD, 5, 5, 3, 4
CROD, 6, 6, 1, 2
CROD, 7, 7, 5, 4
CROD, 8, 8, 6, 3
CROD, 9, 9, 3, 2
CROD, 10, 10, 4, 1
PROD, 1, 1, 5.0
PROD, 2, 1, 5.0
PROD, 3, 1, 5.0
PROD, 4, 1, 5.0
PROD, 5, 1, 5.0
PROD, 6, 1, 5.0
PROD, 7, 1, 5.0
PROD, 8, 1, 5.0
PROD, 9, 1, 5.0
PROD, 10, 1, 5.0
MAT1, 1, 1.0E7, , , 0.1, , , , +
+, 25000., 25000.
FORCE, 300, 2, , 100000.,0.0, -1.0, 0.0
FORCE, 300, 4, , 100000.,0.0, -1.0, 0.0

4-7
 10-BAR TRUSS (Cont.)

SPC1, 100, 12, 5, 6

$
$------------------------------------------------------------------------
$ DESIGN MODEL:
$------------------------------------------------------------------------
$
DESVAR, 1, A1, 5.0, 0.001, 1000.0
DESVAR, 2, A2, 5.0, 0.001, 1000.0
DESVAR, 3, A3, 5.0, 0.001, 1000.0
DESVAR, 4, A4, 5.0, 0.001, 1000.0
DESVAR, 5, A5, 5.0, 0.001, 1000.0
DESVAR, 6, A6, 5.0, 0.001, 1000.0
DESVAR, 7, A7, 5.0, 0.001, 1000.0
DESVAR, 8, A8, 5.0, 0.001, 1000.0
DESVAR, 9, A9, 5.0, 0.001, 1000.0
DESVAR, 10, A10, 5.0, 0.001, 1000.0
DVPREL1,1, PROD, 1, 4, , , , , +
+, 1, 1.0
DVPREL1,2, PROD, 2, 4, , , , , +
+, 2, 1.0
DVPREL1,3, PROD, 3, 4, , , , , +
+, 3, 1.0
DVPREL1,4, PROD, 4, 4, , , , , +
+, 4, 1.0
DVPREL1,5, PROD, 5, 4, , , , , +
+, 5, 1.0
DVPREL1,6, PROD, 6, 4, , , , , +
+, 6, 1.0
DVPREL1,7, PROD, 7, 4, , , , , +
+, 7, 1.0
DVPREL1,8, PROD, 8, 4, , , , , +
+, 8, 1.0
DVPREL1,9, PROD, 9, 4, , , , , +
+, 9, 1.0
DVPREL1,10, PROD, 10, 4, , , , , +
+, 10, 1.0
DRESP1, 1, S1, STRESS, PROD, , 2, , 1
DRESP1, 2, S2, STRESS, PROD, , 2, , 2
DRESP1, 3, S3, STRESS, PROD, , 2, , 3
DRESP1, 4, S4, STRESS, PROD, , 2, , 4
DRESP1, 5, S5, STRESS, PROD, , 2, , 5
DRESP1, 6, S6, STRESS, PROD, , 2, , 6
DRESP1, 7, S7, STRESS, PROD, , 2, , 7
DRESP1, 8, S8, STRESS, PROD, , 2, , 8
DRESP1, 9, S9, STRESS, PROD, , 2, , 9
DRESP1, 10, S10, STRESS, PROD, , 2, , 10
DRESP1, 11, W, WEIGHT
DCONSTR,1, 1, -25000.,25000.
DCONSTR,1, 2, -25000.,25000.
DCONSTR,1, 3, -25000.,25000.
DCONSTR,1, 4, -25000.,25000.

4-8
 10-BAR TRUSS (Cont.)

DCONSTR,1, 5, -25000.,25000.
DCONSTR,1, 6, -25000.,25000.
DCONSTR,1, 7, -25000.,25000.
DCONSTR,1, 8, -25000.,25000.
DCONSTR,1, 9, -25000.,25000.
DCONSTR,1, 10, -25000.,25000.
DOPTPRM,IPRINT, 2, DESMAX, 20, DELP, 0.8
$DSCREEN STRESS -100. 1
$DSCREEN DISP 1
$.......2.......3.......4.......5.......6.......7.......8.......9.......0
ENDDATA

4-9
 10-BAR TRUSS RESULTS

***** NORMAL CONVERGENCE CRITERIA SATISFIED ***** (HARD CONVERGENCE DECISION LOGIC)

**************************************************************************************
CONVERGENCE ACHIEVED BASED ON THE FOLLOWING CRITERIA
(HARD CONVERGENCE DECISION LOGIC)

RELATIVE CHANGE IN OBJECTIVE 2.0186E-03 MUST BE LESS THAN 5.0000E-03

OR ABSOLUTE CHANGE IN OBJECTIVE 3.2241E+00 MUST BE LESS THAN 5.0000E+00
--- AND ---
MAXIMUM CONSTRAINT VALUE 4.4045E-03 MUST BE LESS THAN 5.0000E-03
(CONVERGENCE TO A FEASIBLE DESIGN)
--- OR ---
MAXIMUM OF RELATIVE PROP. CHANGES 4.0000E-01 MUST BE LESS THAN 5.0000E-01
AND MAXIMUM OF RELATIVE D.V. CHANGES 4.0000E-01 MUST BE LESS THAN 1.0000E-03
4-10

(CONVERGENCE TO A BEST COMPROMISE INFEASIBLE DESIGN)

**************************************************************************************
 10-BAR TRUSS RESULTS (Cont.)

***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************

(HARD CONVERGENCE ACHIEVED)

(SOFT CONVERGENCE ACHIEVED)

NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 11

NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 10

OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY

---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
CYCLE APPROXIMATE EXACT OF OF
NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------
4-11

INITIAL 2.098234E+03 6.370801E-01

1 1.891479E+03 1.891557E+03 -4.091475E-05 2.185570E-02

2 1.727591E+03 1.727633E+03 -2.430620E-05 2.338859E-02

3 1.686985E+03 1.686986E+03 -3.618000E-07 2.917891E-03

4 1.649626E+03 1.649633E+03 -4.291912E-06 4.533672E-03

5 1.633149E+03 1.633161E+03 -7.474479E-06 1.653125E-03

6 1.622287E+03 1.622258E+03 1.745734E-05 3.786953E-03

7 1.610578E+03 1.610584E+03 -3.259081E-06 9.830468E-04

8 1.605705E+03 1.605713E+03 -5.017484E-06 5.797110E-03

9 1.597221E+03 1.597220E+03 2.292802E-07 2.474219E-04

10 1.593996E+03 1.593996E+03 -2.297439E-07 4.404531E-03

---------------------------------------------------------------------------------------------------------------
 10-BAR TRUSS RESULTS (Cont.)

DESIGN VARIABLE HISTORY

----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | A1 | 5.0000E+00 : 7.1656E+00 : 6.7879E+00 : 7.5020E+00 : 7.4191E+00 : 7.5574E+00 :
2 | 2 | A2 | 5.0000E+00 : 1.8086E+00 : 1.2854E+00 : 9.8005E-01 : 8.0972E-01 : 6.3827E-01 :
3 | 3 | A3 | 5.0000E+00 : 9.0000E+00 : 9.2569E+00 : 8.5793E+00 : 8.5881E+00 : 8.4451E+00 :
4 | 4 | A4 | 5.0000E+00 : 3.0356E+00 : 3.3196E+00 : 3.3145E+00 : 3.4713E+00 : 3.5531E+00 :
5 | 5 | A5 | 5.0000E+00 : 2.6696E+00 : 1.1306E+00 : 6.7799E-01 : 4.0678E-01 : 2.4407E-01 :
6 | 6 | A6 | 5.0000E+00 : 1.8086E+00 : 1.2854E+00 : 9.8005E-01 : 8.0972E-01 : 6.3827E-01 :
7 | 7 | A7 | 5.0000E+00 : 6.7380E+00 : 7.5133E+00 : 6.5181E+00 : 6.4789E+00 : 6.2868E+00 :
8 | 8 | A8 | 5.0000E+00 : 5.3387E+00 : 4.0911E+00 : 5.0533E+00 : 4.8353E+00 : 5.0332E+00 :
9 | 9 | A9 | 5.0000E+00 : 4.8307E+00 : 4.6313E+00 : 4.6936E+00 : 4.9093E+00 : 5.0248E+00 :
10 | 10 | A10 | 5.0000E+00 : 2.2236E+00 : 1.3882E+00 : 1.2902E+00 : 9.7214E-01 : 8.3043E-01 :
----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | 6 : 7 : 8 : 9 : 10 : 11 :
4-12

----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | A1 | 7.6879E+00 : 7.7337E+00 : 7.8566E+00 : 7.8665E+00 : 7.9267E+00 :
2 | 2 | A2 | 4.6169E-01 : 3.6299E-01 : 2.4340E-01 : 1.8105E-01 : 1.1949E-01 :
3 | 3 | A3 | 8.3136E+00 : 8.2719E+00 : 8.1506E+00 : 8.1431E+00 : 8.0690E+00 :
4 | 4 | A4 | 3.6911E+00 : 3.7331E+00 : 3.8545E+00 : 3.8673E+00 : 3.9281E+00 :
5 | 5 | A5 | 1.4644E-01 : 8.7864E-02 : 5.2718E-02 : 3.1658E-02 : 1.8995E-02 :
6 | 6 | A6 | 4.6169E-01 : 3.6299E-01 : 2.4340E-01 : 1.8105E-01 : 1.1949E-01 :
7 | 7 | A7 | 6.1055E+00 : 6.0343E+00 : 5.8732E+00 : 5.8515E+00 : 5.7556E+00 :
8 | 8 | A8 | 5.2173E+00 : 5.2757E+00 : 5.4508E+00 : 5.4628E+00 : 5.5563E+00 :
9 | 9 | A9 | 5.2201E+00 : 5.2795E+00 : 5.4454E+00 : 5.4669E+00 : 5.5591E+00 :
10 | 10 | A10 | 6.4001E-01 : 5.1250E-01 : 3.4406E-01 : 2.5765E-01 : 1.6734E-01 :
*** USER INFORMATION MESSAGE 6464 (DOM12E)
RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 10.
10-BAR TRUSS RESULTS (Cont.)

Objective Function

4-13
10-BAR TRUSS RESULTS (Cont.)

4-14
 BEAM VIBRATION FREQUENCY AS A DESIGN
CONSTRAINT (M. J. Turner 1970)

Purpose: To illustrate design optimization subject to constraints on

eigenvalue.

A1 A2 A3

6" t1 t2 t3 X

20" 20" 20"

• Objective

• Structural weight minimization

• Constraint

• Fundamental transverse vibration frequency ≥ 20 Hz

• Design variables

• A1, A2, A3 (bar cross-sectional areas)

• t1, t2, t3 (web thickness)

4-15
 BEAM VIBRATION FREQUENCY AS A DESIGN
CONSTRAINT (Cont.)

nastran oldq4k
ID MSC, D200X6 $
TIME 10
SOL 200 $ OPTIMIZATION
CEND
TITLE = VIBRATION OF A BEAM. D200X6
ECHO = UNSORT
DISP = ALL
STRESS = ALL
METHOD = 1
ANALYSIS = MODES
DESOBJ(MIN) = 1 $ OBJECTIVE FUNCTION DEFINITION
DESSUB = 10 $ CONSTRAINT SET SELECTION
$
BEGIN BULK
$
$------------------------------------------------------------------------
$ ANALYSIS MODEL:
$------------------------------------------------------------------------
$
EIGRL, 1, , ,
GRID, 1, , 0.0, 0.0, -3.0, , 123456
GRID, 2, , 20.0, 0.0, -3.0, , 2456
GRID, 3, , 40.0, 0.0, -3.0, , 2456
GRID, 4, , 60.0, 0.0, -3.0, , 2456
GRID, 5, , 0.0, 0.0, 3.0, , 123456
GRID, 6, , 20.0, 0.0, 3.0, , 2456
GRID, 7, , 40.0, 0.0, 3.0, , 2456
GRID, 8, , 60.0, 0.0, 3.0, , 2456
$
CROD, 1, 201, 5, 6
CROD, 2, 202, 6, 7
CROD, 3, 203, 7, 8
CROD, 7, 201, 1, 2
CROD, 8, 202, 2, 3
CROD, 9, 203, 3, 4
PROD, 201, 1, 1.0, 0.0
PROD, 202, 1, 1.0, 0.0
PROD, 203, 1, 1.0, 0.0
$
CQUAD4, 4, 204, 1, 2, 6, 5
CQUAD4, 5, 205, 2, 3, 7, 6
CQUAD4, 6, 206, 3, 4, 8, 7
PSHELL, 204, 1, 0.2
PSHELL, 205, 1, 0.2
PSHELL, 206, 1, 0.2
$

4-16
 BEAM VIBRATION FREQUENCY AS A DESIGN
CONSTRAINT (Cont.)

CONM2, 10, 2, , 15.0

CONM2, 11, 3, , 15.0
CONM2, 12, 4, , 15.0
CONM2, 14, 6, , 15.0
CONM2, 15, 7, , 15.0
CONM2, 16, 8, , 15.0
$
MAT1, 1, 1.03E7, , 0.3, 0.1
PARAM, WTMASS, 0.002588
PARAM, GRDPNT, 1
$
$------------------------------------------------------------------------
$ DESIGN MODEL:
$------------------------------------------------------------------------
$
$...Define the design variables
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
$
DESVAR, 1, A1, 1.0, 0.01, 100.0
DESVAR, 2, A2, 1.0, 0.01, 100.0
DESVAR, 3, A3, 1.0, 0.01, 100.0
DESVAR, 4, T1, 0.2, 0.001, 10.0
DESVAR, 5, T2, 0.2, 0.001, 10.0
DESVAR, 6, T3, 0.2, 0.001, 10.0
$
$...Relate the design variables to analysis model properties
$DVPREL1,ID, TYPE, PID, FID, PMIN, PMAX, C0, , +
$+, DVIDD1, COEF1, DVID2, COEF2, ...
$
DVPREL1,1, PROD, 201, 4, , , , , +DP1
+DP1, 1, 1.0
DVPREL1,2, PROD, 202, 4, , , , , +DP2
+DP2, 2, 1.0
DVPREL1,3, PROD, 203, 4, , , , , +DP3
+DP3, 3, 1.0
DVPREL1,4, PSHELL, 204, 4, , , , , +DP4
+DP4, 4, 1.0
DVPREL1,5, PSHELL, 205, 4, , , , , +DP5
+DP5, 5, 1.0
DVPREL1,6, PSHELL, 206, 4, , , , , +DP6
+DP6, 6, 1.0
$
$...Identify the analysis responses to be used in the design model
$DRESP1,ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
$
DRESP1, 1, W, WEIGHT
DRESP1, 2, F1, EIGN, , , 1
$

4-17
 BEAM VIBRATION FREQUENCY AS A DESIGN
CONSTRAINT (Cont.)

$...Use these responses to define the objective (in case control)

$ and the constraints:
$DCONSTR,DCID, RID, LALLOW, UALLOW
DCONSTR,10, 2, 15791., 200000.0 $ lower bound = 20 Hz
$
$...Optional override of design optimization parameters:
DOPTPRM, IPRINT,2, DESMAX, 10, DELP, 0.5, P1, 1, +
+, P2, 15
$
ENDDATA

4-18
 BEAM VIBRATION FREQUENCY AS A DESIGN CONSTRAINT (Cont.)

***** NORMAL CONVERGENCE CRITERIA SATISFIED ***** (HARD CONVERGENCE DECISION LOGIC)

**************************************************************************************
CONVERGENCE ACHIEVED BASED ON THE FOLLOWING CRITERIA
(HARD CONVERGENCE DECISION LOGIC)

RELATIVE CHANGE IN OBJECTIVE 2.1191E-04 MUST BE LESS THAN 1.0000E-03

OR ABSOLUTE CHANGE IN OBJECTIVE 1.4758E-03 MUST BE LESS THAN 1.0000E-20
--- AND ---
MAXIMUM CONSTRAINT VALUE 2.2716E-03 MUST BE LESS THAN 5.0000E-03
(CONVERGENCE TO A FEASIBLE DESIGN)
--- OR ---
4-19

MAXIMUM OF RELATIVE PROP. CHANGES 2.8511E-02 MUST BE LESS THAN 1.0000E-03

AND MAXIMUM OF RELATIVE D.V. CHANGES 2.8511E-02 MUST BE LESS THAN 1.0000E-03
(CONVERGENCE TO A BEST COMPROMISE INFEASIBLE DESIGN)
**************************************************************************************
 BEAM VIBRATION FREQUENCY AS A DESIGN CONSTRAINT (Cont.)

***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************

(HARD CONVERGENCE ACHIEVED)

NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 7

NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 6

OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY

---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
CYCLE APPROXIMATE EXACT OF OF
NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------
4-20

INITIAL 1.920000E+01 N/A

1 9.599569E+00 9.599999E+00 -4.480283E-05 9.583256E-02

2 7.769873E+00 7.771017E+00 -1.472050E-04 1.124312E-02

3 7.153115E+00 7.153544E+00 -5.999172E-05 -8.152450E-03

4 7.007634E+00 7.007762E+00 -1.830387E-05 6.831794E-04

5 6.964279E+00 6.964198E+00 1.157139E-05 2.233088E-03

6 6.962721E+00 6.962722E+00 -2.054529E-07 2.271554E-03

---------------------------------------------------------------------------------------------------------------
 BEAM VIBRATION FREQUENCY AS A DESIGN CONSTRAINT (Cont.)

DESIGN VARIABLE HISTORY

----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | A1 | 1.0000E+00 : 5.0000E-01 : 7.5000E-01 : 7.9516E-01 : 8.1118E-01 : 8.2894E-01 :
2 | 2 | A2 | 1.0000E+00 : 5.0000E-01 : 3.3388E-01 : 4.2236E-01 : 4.3278E-01 : 4.3948E-01 :
3 | 3 | A3 | 1.0000E+00 : 5.0000E-01 : 2.5000E-01 : 1.5584E-01 : 1.3189E-01 : 1.1434E-01 :
4 | 4 | T1 | 2.0000E-01 : 1.0000E-01 : 1.0296E-01 : 6.6512E-02 : 5.3757E-02 : 5.0283E-02 :
5 | 5 | T2 | 2.0000E-01 : 1.0000E-01 : 5.0000E-02 : 4.6829E-02 : 4.7047E-02 : 4.3168E-02 :
6 | 6 | T3 | 2.0000E-01 : 1.0000E-01 : 5.0000E-02 : 2.5000E-02 : 2.4560E-02 : 2.5978E-02 :

----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | 6 : 7 : 8 : 9 : 10 : 11 :
4-21

----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | A1 | 8.2998E-01 :
2 | 2 | A2 | 4.3973E-01 :
3 | 3 | A3 | 1.1371E-01 :
4 | 4 | T1 | 4.9550E-02 :
5 | 5 | T2 | 4.2818E-02 :
6 | 6 | T3 | 2.6718E-02 :
*** USER INFORMATION MESSAGE 6464 (DOM12E)
RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 6.
BEAM VIBRATION FREQUENCY AS A DESIGN
CONSTRAINT RESULTS

4-22
BEAM VIBRATION FREQUENCY AS A DESIGN
CONSTRAINT RESULTS (Cont.)

4-23
BEAM VIBRATION FREQUENCY AS A DESIGN
CONSTRAINT RESULTS (Cont.)
 SECTION 5

SPECIAL MODELING TOPICS I

DESIGN VARIABLE LINKING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-1

DLINK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-2

REDUCED BASIS FORMULATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-5

DESIGN RESPONSES AND CASE CONTROL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-9

MODE TRACKING FEATURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-11

MODE TRACKING REQUEST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-12

MODE TRACKING PARAMETERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-13

EQUALITY CONSTRAINTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-14

FREQUENCY MATCHING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-15

 DESIGN VARIABLE LINKING 5

Design variable linking enforces a design variable dependence on

other, independent variables.

The DLINK entry is used to write this dependence.

5-1
DLINK Multiple Design Variable Linking

Relates
DLINK one design variable to one or more other design variables. Multiple Design Variable Linking

Format:

1 2 3 4 5 6 7 8 9 10
DLINK ID DDVID C0 CMULT IDV1 C1 IDV2 C2

IDV3 C3 -etc.-

Example:

DLINK 10 2 0.1 0.33 2 2.0 6 –1.0

8 7.0

Field Contents
ID Unique identification number. (Integer > 0)

DDVID Dependent design variable identification number. (Integer > 0)

C0 Constant term. (Real; Default = 0.0)

CMULT Constant multiplier. (Real; Default = 1.0)

IDVi Independent design variable identification number. (Integer > 0)

Ci Coefficient i corresponding to IDVi. (Real)

Remarks:
1. DLINK defines the relationship

DDVID = C0 + CMULT ∑ Ci ∗ IDVi

i

2. This capability provides a means of linking physical design variables such as element
thicknesses to nonphysical design variables such as the coefficients of interpolating
functions.

3. CMULT provides a simple means of scaling the Ci. For example if Ci = 1/7, 2/7, 4/7, etc.
is desired, then CMULT = 1/7 and Ci = 1, 2, 4, etc., may be input.

4. An independent IDVi must not occur on the same DLINK entry more than once.

5. ID is for user reference only.

5-2
 DESIGN VARIABLE LINKING (Cont.)

For example, consider a simple plate, in which the redesigned

thickness distribution is to vary linearly:

t1 t2 t3 t4

x
0 10. 20. 30. 40.

t4 – t1
t i = ---------------- x i + t 1

 30 

2 1
so, t 2 = --- t 1 + --- t 4
3 3

1 2
t 3 = --- t 1 + --- t 4
3 3

5-3
 DESIGN VARIABLE LINKING (Cont.)

The DLINK entries can be written as:

DESVAR, 1, T1, 1., 0.01, 5.

DESVAR, 1, T2, 1., 0.01, 5.
DESVAR, 1, T3, 1., 0.01, 5.
DESVAR, 1, T4, 1., 0.01, 5.
$
DLINK, 11, 2, 0., 0.333, 1, 2., 4, 1.
DLINK, 12, 3, 0., 0.333, 1, 1., 4, 2.
$
DVPREL1,21, PSHELL, 101, 4, 0.01, 5., , , +
+, 1, 1.0
DVPREL1,22, PSHELL, 102, 4, 0.01, 5., , , +
+, 2, 1.0
DVPREL1,23, PSHELL, 103, 4, 0.01, 5., , , +
+, 3, 1.0
DVPREL1,24, PSHELL, 104, 4, 0.01, 5., , , +
+, 4, 1.0
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......0

The only independent variables are t1 and t4. t2 and t3 are the
dependent quantities. The design space is now two-dimensional,
rather than four.

Reducing the number of independent design variables offers greater

advantages for large design problems.

5-4
 REDUCED BASIS FORMULATIONS

Reduced basis formulations are conceptually similar to design

variable linking, but can be implemented in a number of ways.

Consider a plate to be designed with uniform thickness distribution:

Plate Thickness Distribution

x=0

t1
t2

t10

x=L

5-5
 REDUCED BASIS FORMULATIONS (Cont.)

Basis Functions

F(x/L)
F1
1.0
F2
F3

x/L
0.5 1.0

F 1  --- = C 1
x
 L

F 2  --- = C 1  1 – ---
x x
 L  L

x 2
F 3  --- = C 1  1 – ---
x
 L  L

C 1 = 1.0

5-6
 REDUCED BASIS FORMULATIONS (Cont.)

These variations can be written as linear combinations of basis

vectors:

 t1 
  1.0 1.0 1.0
 t2 
  1.0 0.9 0.81
t 
 3 1.0 0.8 0.64
t 
 4 1.0 0.7 0.49
t   x1 
 5 1.0 0.6 0.36  
  =  x2 
 t6  1.0 0.5 0.25  
   x3 
 t7  1.0 0.4 0.16
 
 t8  1.0 0.3 0.09
 
t  1.0 0.2 0.04
 9
t  1.0 0.1 0.01
 10 

In general, a reduced basis formulation is expressed as:

{ p } M = [ T ] MXN { x } N

where M » N

5-7
 REDUCED BASIS FORMULATIONS (Cont.)

Since these are linear design variable to property relations, we can

use DVPREL1 entries:

$PSHELL,PID, MID1, T, MID2, 12I/T3 MID3

PSHELL, 101, 100, 1.0, 100, , 100
PSHELL, 102, 100, 1.0, 100, , 100
PSHELL, 103, 100, 1.0, 100, , 100
PSHELL, 104, 100, 1.0, 100, , 100
PSHELL, 105, 100, 1.0, 100, , 100
PSHELL, 106, 100, 1.0, 100, , 100
PSHELL, 107, 100, 1.0, 100, , 100
PSHELL, 108, 100, 1.0, 100, , 100
PSHELL, 109, 100, 1.0, 100, , 100
PSHELL, 110, 100, 1.0, 100, , 100
$
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
DESVAR, 1, X1, 0.33, -1.0, +1.0
DESVAR, 2, X2, 0.33, -1.0, +1.0
DESVAR, 3, X3, 0.33, -1.0, +1.0
$
$DVPREL1,ID, TYPE, PID, FID, PMIN, PMAX, C0, , +
$+, DVID1, COEF1, DVID2, COEF2, ...
DVPREL1,201, PSHELL, 101, 4, 0.01, , , , +
+, 1, 1.0, 2, 1.0, 3, 1.0
DVPREL1,202, PSHELL, 102, 4, 0.01, , , , +
+, 1, 1.0, 2, 0.9, 3, 0.81
DVPREL1,203, PSHELL, 103, 4, 0.01, , , , +
+, 1, 1.0, 2, 0.8, 3, 0.64
DVPREL1,204, PSHELL, 104, 4, 0.01, , , , +
+, 1, 1.0, 2, 0.7, 3, 0.49
DVPREL1,205, PSHELL, 105, 4, 0.01, , , , +
+, 1, 1.0, 2, 0.6, 3, 0.36
DVPREL1,206, PSHELL, 106, 4, 0.01, , , , +
+, 1, 1.0, 2, 0.5, 3, 0.25
DVPREL1,207, PSHELL, 107, 4, 0.01, , , , +
+, 1, 1.0, 2, 0.4, 3, 0.16
DVPREL1,208, PSHELL, 108, 4, 0.01, , , , +
+, 1, 1.0, 2, 0.3, 3, 0.09
DVPREL1,209, PSHELL, 109, 4, 0.01, , , , +
+, 1, 1.0, 2, 0.2, 3, 0.04
DVPREL1,210, PSHELL, 110, 4, 0.01, , , , +
+, 1, 1.0, 2, 0.1, 3, 0.01

5-8
 DESIGN RESPONSES AND CASE CONTROL

Specifying DRESP1 entries ensures that all necessary data recovery

is done automatically. (Internally, design optimization Case Control
Sections are written to ensure all necessary responses are
recovered.)

In general, Case Control output requests are for output/

postprocessing purposes only. (PARAM,NASPRT,n can be used to
control frequency of this output.)

Exceptions:

• Dynamic Responses

Magnitude/phase or real/imaginary representations are selected

using Case Control

Example:

SUBCASE 10
ANALYSIS = DFREQ
SET 200 = 1000,1001,1003
DISPLACEMENT(PHASE) = 200

All displacement components will be in magnitude/phase form both

in the design model and in the analysis output. The SET request is
not required.

5-9
 DESIGN RESPONSES AND CASE CONTROL (Cont.)

• Stress/Strain

von Mises, or maximum shear, are selectable in Case Control, both

for data recovery and for the design model.

Example:

SUBCASE 20
ANALYSIS = STATICS
SET 200 = 1000,1001,1003
STRESS(MAXS) = 200

Stresses for all elements are represented as maximum shears,

both for data recovery of the analysis model and for response
retrieval of the design model.

5-10
 MODE TRACKING FEATURES

• Useful when eigenvalues (e.g., first roof bending, first torsional,

etc.) are being designed

• Modes are “tracked” based on a cross-orthogonality check:

t
Φi Mi Φi – 1 = ti

• Updated DRESP1 data available in the .pch file

$
$ Mode Tracking has been performed successfully. Updated
DRESP1 entries are:
$
DRESP1 301LFREQ FREQ 2
DRESP1 303HFREQ FREQ 4

• If mode tracking fails, correlation data for the untracked mode(s) is

printed and the run is terminated.

• Mode tracking is requested by Case Control entry: MODTRAK.

• Parameters for mode tracking are specified by Bulk Data Entry:

MODTRAK.

5-11
MODTRAK Mode Tracking Request

MODTRAKmode tracking options in design optimization (SOL 200).

Selects Mode Tracking Request

Format:
MODTRAK = n

Example:
MODTRAK = 100

Describer Meaning
n Set identification of a MODTRAK Bulk Data entry. (Integer > 0)

Remark:
1. Selection of a MODTRAK Bulk Data entry with the MODTRAK Case Control command
activates mode tracking for the current subcase. This request is limited to normal modes
subcases (ANALYSIS = MODES) in design optimization (SOL 200).

5-12
Mode Tracking Parameters MODTRAK
MODTRAK parameters for mode tracking in design optimization (SOL 200).
Specifies Mode Tracking Parameters

Format:

1 2 3 4 5 6 7 8 9 10
MODTRAK SID LOWRNG HIGHRNG MTFILTER

Example:

MODTRAK 100 1 26 0.80

Field Contents
SID Sets identification number that is selected in the Case Control Section with the
MODTRAK command. (Integer; No Default) See Remark 1.

LOWRNG Lowest mode number in range to search. See Remark 2. (Integer ≥ 0,

Default = 0. If nonzero, LOWRNG < HIGHRNG.)

HIGHRNG Highest mode number in range to search. See Remark 2. (Integer > 0,
Default = number of eigenvalues extracted. If nonzero, LOWRNG < HIGHRNG.)

MTFILTER Filtering parameter used in mode cross-orthogonality check. See Remark 3.

(Real, Default = 0.9)

Remarks:
1. Only the designed modes for the subcase will be tracked. A designed mode is one that is
used in the design model (in connection with either objective or constraints) and, there-
fore, identified on a DRESP1 entry.

2. The range of modes LOWRNG through HIGHRNG, inclusive, will be used to track the de-
signed modes. If LOWRNG and HIGHRNG are both blank, then all computed modes will
be used to search for the designed modes. Since large numbers of computed modes will
result in higher computational costs, limiting the search range with LOWRNG and HIGH-
RNG is recommended.

3. Modes are considered to correlate if their mass normalized cross orthogonalities are
greater than MTFILTER.

5-13
 EQUALITY CONSTRAINTS

Equality constraints are not provided for directly, although they can
be formulated in MSC/NASTRAN.

The simplest way is to provide lower and upper bounds that are
equivalent:

r ≤ r(x) ≤ r

This can be written using a single DCONSTR entry

$
$ EIGENVALUE FOR MODE 2:
DRESP1, 15, LAMA1, EIGN, , , 2
$
$ BOUNDS:
DCONSTR,100, 15, 3.5531E4,3.5531E4 $ 30 Hz
$

This approach is not recommended, since it usually doesn’t lead to

the best conditioned problem. The bounds will probably never be
solved with equality.

The best way is to provide some tolerance or to recast the problem;

these are shown in the next example.

5-14
 FREQUENCY MATCHING

• Design objective:

• Assume that the eigenvalues of a natural vibration problem

λ i, λ i + 1, λ i + 2, … are found from an analysis of the current model.
Since the results of the shake tests indicate that they should be
Λ i, Λ i + 1, Λ i + 2, … we wish to make adjustments to a set of
structural parameters to match both the analysis and test results

• Approach 1

• Use equality constraints or two narrowly defined inequality

constraints

Λi – ε ≤ λi ≤ Λi + ε

• Approach 2

• Form a new synthetic response as the sum of frequency errors

squared and minimize it

ERR = ( 1.0 – Λ i ⁄ λ i )∗∗2 + ( 1.0 – ( Λ i + 1 ⁄ λ i + 1 ) )∗∗2 + …

5-15
5-16
 SECTION 6

DESIGN MODELING II

DESIGN MODEL DEFINITION PROCESS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-1

USER-DEFINED DESIGN VARIABLE-TO-PROPERTY RELATIONS. . . . . . . . . . . . 6-2

DVPREL2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-3

DEQATN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-5

DTABLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-9

EXAMPLE – DEFINING A RECTANGULAR SECTION BAR . . . . . . . . . . . . . . . . . 6-10

USER-DEFINED RESPONSES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-11

DRESP2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-13

EXAMPLE – DEFORMATION CONTROL WITH TYPE-2 RESPONSES. . . . . . . . . 6-16

EXAMPLE – BAR AXIAL PLUS BENDING STRESSES . . . . . . . . . . . . . . . . . . . . . 6-19

RESTRICTIONS IN FORMING SYNTHETIC RESPONSES. . . . . . . . . . . . . . . . . . . 6-20

DESIGN MODEL DEFINITION PROCESS 6

Flowchart for Design Modeling

Define Analysis Disciplines

(Executive Control: SOL 200;
Case Control: Analysis +)

Define Design Variables

(Bulk Data: DESVAR)

Relate Design Variables to

Analysis Model:
(Bulk Data: DVPREL1, DVPREL2, DEQATN)

Define Design Responses

(Bulk Data: DRESP1, DRESP2, DEQATN)

Define Objective (Case Control: DESOBJ)

and Constraints (Bulk Data: DCONSTR, DCONADD;
Case Control: DESGLB, DESSUB)

Provide any necessary

parameter overrides
(Bulk Data: DOPTPRM)

6-1
 USER-DEFINED DESIGN VARIABLE-TO-PROPERTY
RELATIONS

Analysis model parameters can be classified depending on how

they’re referenced in the design model:

PL
Analysis = Po + C2 X DVPREL1
Model =
Parameters PNL
= f(X,C) DVPREL2

In order to define a property with an equation, DVPREL2, DEQATN,

and possibly, DTABLE entries are used.

6-2
Design Variable to Property Relation DVPREL2
Defines the relation between an analysis model property and design Design
DVPREL2 variables withtoaProperty
Variable user-supplied
Relation
equation.

Format:

1 2 3 4 5 6 7 8 9 10
DVPREL2 ID TYPE PID FID PMIN PMAX EQID

“DESVAR” DVID1 DVID2 DVID3 -etc.-

“DTABLE” LABL1 LABL2 LABL3 -etc.-

Example:

DVPREL2 13 PBAR 712 5 0.2 4

DESVAR 4 11 13 5

DTABLE PI YM

Field Contents
ID Unique identification number. (Integer > 0)

TYPE Name of a property entry, such as PBAR, PBEAM, etc. (Character)

PID Property entry identification number. (Integer > 0)

FID Field position of the property in the analysis model entry. (Integer ≠ 0)

PMIN Minimum value allowed for this property. If FID references a stress recovery
location field, then the default value for PMIN is –1.0+35. PMIN must be
explicitly set to a negative number for properties that may be less than zero (for
example, field ZO on the PCOMP entry). (Real; Default = 1.E–20)

PMAX Maximum value allowed for this property. (Real; Default = 1.0E20)

EQID DEQATN entry identification number. (Integer > 0)

“DESVAR” DESVAR flag. Indicates that the IDs of DESVAR entries follow. (Character)

DVIDi DESVAR entry identification number. (Integer > 0)

“DTABLE” DTABLE flag. Indicates that the IDs for the constants in a DTABLE entry follow.
This field may be omitted if there are no constants involved in this relation.
(Character)

LABLi Label for a constant on the DTABLE entry. (Integer > 0)

(Continued)

6-3
DVPREL2 Design Variable to Property Relation

Remarks:
1. The variables identified by DVIDi and LABLi correspond to variable names (x1, x2, etc.)
listed in the left-hand side of the first equation on the DEQATN entry identified by EQID.
The variable names x1 through xN (where N = m + n) are assigned in the order DVID1,
DVID2, ..., DVIDn, LABL1, LABL2, ..., LABLm.

2. If both “DESVAR” and “DTABLE” are specified in field 2, “DESVAR” must appear first.

3. FID may be either a positive or a negative number. If FID > 0, it identifies the field position
on a property entry. If FID < 0, it identifies the word position of an entry in EPT. For
example, to specify the area of a PBAR, either FID = +4 or FID = –3 may be used. How-
ever, if PTYPE = “PBEAM”, FID must be negative. See Remark 4 on the DVPREL1 entry
description for specification of the PBEAM element property tables.

4. PTYPE = “PBEND” is not supported, either directly through FIDs or indirectly via word po-
sitions in the element property table.

6-4
Design Equation Definition DEQATN
DEQATN one or more equations for use in design sensitivity or p-element analysis.
Defines Design Equation Definition

Format:

1 2 3 4 5 6 7 8 9 10
DEQATN EQID EQUATION

EQUATION (Cont.)

Example:

DEQATN 14 F1(A, B, C, D, R) = A + B ∗ C – (D ∗ ∗ 3 + 10.0) + sin(PI(1) ∗ R)

+ A ∗ ∗ 2/(B – C); F = A + B – F1 ∗ D

Field Contents
EQID Unique equation identification number. (Integer > 0)

EQUATION Equation(s). See Remarks. (Character)

Remarks:
1. EQUATION is a single equation or a set of nested equations and is specified in fields 3
through 9 on the first entry and may be continued on fields 2 through 9 on the continuation
entries. On the continuation entries, no commas can appear in columns 1 through 8. All
data in fields 2 through 9 must be specified in columns 9 through 72. The large-field for-
mat is not allowed.

A single equation has the following format:

variable-1 (x1, x2, ..., xn) = expression-1

A set of nested equations is separated by semicolons and has the format:

variable-1 (x1, x2, ..., xn) = expression-1;

variable-2 = expression-2;
variable-3 = expression-3;
etc...................
variable-m = expression-m

Expression-i is a collection of constants, real variables, and real functions, separated by

operators, and must produce a single real value. (x1, x2, ..., xn) is the list of all the vari-
able names (except variable-i) that appear in all expressions. Variable-i may be used in
subsequent expressions. The last equation, variable-m = expression-m, provides the
value that is returned to the Bulk Data entry that references EQID; e.g., DRESP2. The
example above represents the following mathematical equations:

(Continued)

6-5
DEQATN Design Equation Definition

2
3 A
1 = A + B ∗ C – ( D + 10 ) + sin ( PI ( 1 ) ∗ R ) + --------------
B–C
F = A + B – F1 ∗ D

where SIN and PI are intrinsic functions. See Remark 4.

2. EQUATION may contain embedded blanks. EQUATION must contain less than 12,500
nonblank characters. This is equivalent to approximately 195 continuation entries.

3. The syntax of the expressions follows FORTRAN language standards. The allowable
arithmetic operations are shown in Table 1 in the order of execution precedence. Paren-
thesis are used to change the order of precedence. Operations within parentheses are
performed first with the usual order of precedence being maintained within the
parentheses.

Table 1. DEQATN Entry Operators.

Operator Operation Sample Expressions Interpreted As

–, + Negative or Positive immediately X∗∗–Y X∗∗(–Y)
preceded by exponentiation

∗∗ Exponentiation –X∗∗Y (–X∗∗Y)

–, + Negative or Positive –X–Y (–X)–Y

∗, / Multiplication or Division X∗Y+Z (X∗Y)+Z

+, – Addition or Subtraction X+Y X+Y

4. The expressions may contain intrinsic functions. Table 2 contains the format and descrip-
tions of functions that may appear in the expressions. The use of functions that may be
discontinuous must be used with caution because they can cause discontinuous deriva-
tives. These are ABS, DIM, MAX, MIN, and MOD. For examples and further details see
the MSC/NASTRAN DMAP Module Dictionary.

(Continued)

6-6
Design Equation Definition DEQATN
Table 2. DEQATN Entry Functions.

Format Description Mathematical Expression

ABS(x) absolute value |x|

ACOS(x) arccosine -1
cos x

ACOSH(x) hyperbolic arccosine -1

cosh x

ASIN(x) arcsine -1
sin x

ASINH(x) hyperbolic arcsine -1

sinh x

ATAN(x) arctangent -1
tan x

ATAN2(x,y) arctangent of quotient -1

tan (x/y)

ATANH(x) hyperbolic arctangent -1

tanh x

ATANH2(x,y) hyperbolic arctangent of quotient -1

tanh (x/y)

COS(x) cosine cos x

COSH(x) hyperbolic cosine cosh x

DIM(x,y) positive difference x – MIN(x,y)

EXP(x) exponential x
e

LOG(x) natural logarithm loge x

LOG10(x) common logarithm log10 x

LOGX(x,y) base x logarithm logx y

MAX(x1, x2, ...) maximum maximum of x1, etc.

MIN(x1, x2, ...) minimum minimum of x1, etc.

MOD(x,y) remainder (modulo) x – y ∗ (INT(x/y))

PI(x) multiples of pi (π) x∗π

SIN(x) sine sin x

SINH(x) hyperbolic sine sinh x

SQRT(x) square root x

TAN(x) tangent tan x

TANH(x) hyperbolic tangent tanh x

6-7
DEQATN Design Equation Definition

5. If the DEQATN entry is referenced by the:

a. DVPREL2 entry, then xi represents the DVIDj and LABLk fields.

b. DRESP2 entry, then xi represents the DVIDj, LABLk, NRm, and Gp fields in that
order.

c. GMLOAD, GMBC, or TEMPF entries, then

x1 represents x in the basic coordinate system,
x2 represents y in the basic coordinate system, and
x3 represents z in the basic coordinate system.

d. GMCURV entry, then

x1 represents line parameter u.

e. GMSURF entry, then

x1 represents surface parameter u and
x2 represents surface parameter v.

6. If the DEQATN entry is referenced by the GMLOAD, GMBC, TEMPF, GMCURV, or

GMSURF entries and your computer has a short word length (e.g., 32 bits/word), then
EQUATION is processed with double precision and constants may be specified in double
precision; e.g., 1.2D0. If your machine has a long word length (e.g., 64 bits/word) then
EQUATION is processed in single precision and constants must be specified in single pre-
cision; e.g., 1.2.

7. The DMAP logical operators NOT, AND, OR < XOR, and XQV cannot be used as Xi
names.

8. Input errors on the DEQATN entry often result in poor messages. Substituting a “[” for a
parenthesis or violating the restriction against large field format are examples. Known
messages are UFM 215, SFM 233 and UFM 5199. If any of these messages are encoun-
tered then review the DEQATN entry input.

6-8
Table Constants DTABLE
DTABLE a table of real constants that are used in equations (see DEQATN entry).
Defines Table Constants

Format:

1 2 3 4 5 6 7 8 9 10
DTABLE LABL1 VALU1 LABL2 VALU2 LABL3 VALU3 LABL4 VALU4

LABL5 VALU5 LABL6 VALU6 -etc.-

Example:

DTABLE PI 3.142 H 10.1 E 1.0E6 T 0.1

G 5.5E5 B 100.

Field Contents
LABLi Label for the constant. (Character)

VALUi Value of the constant. (Real)

Remarks:
1. Only one DTABLE entry may be specified in the Bulk Data Section.

2. LABLi are referenced by the LABi on the DVPREL2 or DRESP2 entries.

6-9
 EXAMPLE – DEFINING A RECTANGULAR
SECTION BAR

b = 0.3 1
2 C 1 = – --- b
c h = 0.4 2
A = b⋅h 1
C 2 = --- h
2
h 1 3
bh
I 1 = ----------
12
3
hb
b I 2 = ----------
12

$...ANALYSIS MODEL PROPERTIES:

PBAR, 120, 220, 0.12, 9.0E-4, 1.6E-3, , , , +
+, -.15, .2
$
$...BAR CROSS-SECTIONAL DIMENSIONS:
$ESVAR, ID, LABEL, XINIT, XLB, XUB, DELXV
DESVAR, 10, B1, .3, .1, 1.0
DESVAR, 11, H1, .4, .1, 1.0
$
$...BAR PROPERTY RELATIONS, A, I1, I2:
$DVPREL2,ID, TYPE, PID, FID, PMIN, PMAX, EQID, , +
$+, DESVAR,DVID1, DVID2, ..., , , , , +
$+, DTABLE,CID1, CID2, ...
DVPREL2,250, PBAR, 120, 4, , , 501, , +
+, DESVAR, 10, 11
DVPREL2,251, PBAR, 120, 5, 1.0E-5, , 502, , +
+, DESVAR, 10, 11
DVPREL2,252, PBAR, 120, 6, 1.0E-5, , 503, , +
+, DESVAR, 10, 11
$
$...EQUATIONS:
DEQATN 501 AREA(B,H) = B*H
DEQATN 502 I1(B,H) = B*H**3/12.
DEQATN 503 I2(B,H) = H*B**3/12.
$
$...STRESS RECOVERY POINT LOCATIONS:
$DVPREL1,ID, TYPE, PID, FID, PMIN, PMAX, C0, , +
$+, DVID1, COEF1, DVID2, COEF2, ...
DVPREL1,260, PBAR, 120, 12, -.5, , , , +
+, 10, -.5
DVPREL1,261, PBAR, 120, 13, .05, , , , +
+, 11, .5

6-10
 USER-DEFINED RESPONSES

Design responses can be classified based on how they are defined:

DRESP1

Responses obtained directly from

Simple an MSC/NASTRAN analysis
Responses
(Type 1)
Design =
Responses
Synthetic
Responses DRESP2
(Type 2)
Responses written using equations

Type 2, or second-level responses, can be written as functions of:

• Design Variables

• Table Constants

• Type 1 Response(s)

• Grid Coordinates

• DVPREL1 Entries

6-11
 USER-DEFINED RESPONSES (Cont.)

• Examples:

• Generation of a new stress or strain failure criterion that is not

available in MSC/NASTRAN

• Imposing local buckling criteria based on element sizes as well as

stress components

• Programming proprietary design-sizing equations

• Generation of nonlinear displacement responses such as

displacement magnitudes:

2 2 2
U = ux + uy + uz

• More complex relations among displacements could be formed by

combining with MPC and RBE3 capabilities.

A second level response is written using a DRESP2 Bulk Data entry,

which in turn references a DEQATN entry.

6-12
Design Sensitivity Equation Response Quantities DRESP2
DRESP2 equation responses that are used in the design, either
Defines asSensitivity
Design constraints or asResponse
Equation an objective.
Quantities

Format:

1 2 3 4 5 6 7 8 9 10
DRESP2 ID LABEL EQID REGION

“DESVAR” DVID1 DVID2 DVID3 DVID4 DVID5 DVID6 DVID7

DVID8 -etc.-

“DTABLE” LABL1 LABL2 LABL3 LABL4 LABL5 LABL6 LABL7

LABL8 -etc.-

“DRESP1” NR1 NR2 NR3 NR4 NR5 NR6 NR7

NR8 -etc.-

“DNODE” G1 C1 G2 C2 G3 C3

G4 C4 -etc.-

“DVPREL1” DPIP1 DPIP2 DPIP3 DPIP4 DPIP5 DPIP6 DPIP7

DPIP8 DPIP9 -etc.-

Example:

DRESP2 1 LBUCK 5 3

DESVAR 101 3 4 5 1 205 209

201

DTABLE PI YM L

DRESP1 14 1 4 22 6 33 2

DNODE 14 1 4 1 22 3

2 1 43 1

DVPREL1 101 102

Field Contents
ID Unique identification number. (Integer > 0)

LABEL User-defined label. (Character)

(Continued)

6-13
DRESP2 Design Sensitivity Equation Response Quantities

EQID DEQATN entry identification number. (Integer > 0)

REGION Region identifier for constraint screening. See Remark 5. (Integer > 0)

“DESVAR” Flag indicating DESVAR entry identification numbers. (Character)

DVIDi DESVAR entry identification number. (Integer > 0)

“DTABLE” Flag indicating that the labels for the constants in a DTABLE entry follow.
(Character)

LABLj Label for a constant in the DTABLE entry. (Character)

“DRESP1” Flag indicating DRESP1 entry identification numbers. (Character)

NRk DRESP1 entry identification number. (Integer > 0)

“DNODE” Flag signifying that the following fields are “designed grid points.” See Remark 7.
(Character)

Gm Grid point identification number. (Integer > 0)

Cm Degree of freedom number of grid point Gm. (1 ≤ Integer ≤ 3)

DVPREL1 Flag indicating DVPREL1 entry identification number. (Character)

DPIPi DVPREL1 entry identification number. (Integer > 0)

Remarks:
1. DRESP2 entries may only reference DESVAR, DTABLE, DRESP1, and DVPREL1 en-
tries. They may not reference other DRESP2 entries.

2. Referenced DRESP1 entries cannot span analysis types or superelements.

3. DRESP2 entries must have unique identification numbers with respect to DRESP1
entries.

4. The “DESVAR”, “DTABLE”, “DRESP1”, “DNODE”, and “DVPREL1” flags in field 2 must
appear in the order given above. Any of these words, along with the identification num-
bers associated with them, may be omitted if they are not involved in this DRESP2
relationship. However, at least one of these four types of arguments must exist.

5. The REGION field follows the same rules as for the DRESP1 entries. DRESP1 and
DRESP2 responses will never be contained in the same region, even if they are assigned
the same REGION identification number. The default is to put all responses referenced
by one DRESP2 entry in the same region.

(Continued)

6-14
Design Sensitivity Equation Response Quantities DRESP2
6. The variables identified by DVIDi, LABLj, NRk, the Gm, Cm pairs and DPIPi are assigned
(in that order) to the variable names (x1, x2, x3, etc.) specified in the left-hand side of the
first equation on the DEQATN entry referenced by EQID. The variable names x1 through
xN (N = m + n + p + q) are assigned in the order DVID1, ..., DVIDm, LABL1, ..., LABLn,
NR1, ..., NRp, G1, ..., Gq, DPIP1, ..., DPIPr. In the example below,

DESVARs 101 and 3 are assigned to arguments A and B.

DTABLEs PI and YM are assigned to arguments C and D.
Grid 14, Component 1 is assigned to argument R.

DRESP2 1 LBUCK 5 3

DESVAR 101 3

DTABLE PI YM

DNODE 14 1

DEQATN 5 F1(A, B, C, D, R) = A + B ∗ C – (D ∗ ∗ 3 + 10.0) + sin(C ∗ R)

7. (Gm, Cm) refer to a designed grid component. Depending on the scheme used in gener-
ating basis vectors, a designed grid component can be one of the following:

a. A component defined by a DVGRID entry if manual input of DVGRID entries is used.

b. A grid component that is free to move in a full auxiliary model when the external dis-
placement fields are used.

c. When the auxiliary model boundary shape method is used, a designed grid compo-
nent can be either a loaded grid component in the auxiliary model or a grid
component with motion that is obtained from interpolation of boundary shapes.

d. When the interface using geometric boundary shapes is used, a designed grid com-
ponent is either a grid component defined on a DVGRID entry or one with motion that
is obtained from interpolation of boundary shapes.

6-15
 EXAMPLE – DEFORMATION CONTROL WITH
TYPE-2 RESPONSES

desired shape

deformed shape

u1 u2 u3 u4
1 2 3 4
original shape

Design objective:

• Bar displaces but does not bend

• Target displacements become:

2u 1 + u 4
ũ 2 = -----------------------
3

and

u 1 + 2u 4
ũ 3 = -----------------------
3

6-16
 EXAMPLE – DEFORMATION CONTROL WITH
TYPE-2 RESPONSES (Cont.)

• A new response that represents a measure of deformation may be

expressed:

2 2
DEF = ( u 2 – ũ 2 ) + ( u 3 – ũ 3 )

$ IN CASE CONTROL:
$
DESOBJ(MIN) = 21
$
$ IN BULK DATA:
$
$...IDENTIFY THE FIRST-LEVEL DISPLACEMENT RESPONSES:
$
$DRESP1,ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
DRESP1, 11, UY1, DISP, , , 2, , 1
DRESP1, 12, UY2, DISP, , , 2, , 2
DRESP1, 13, UY3, DISP, , , 2, , 3
DRESP1, 14, UY4, DISP, , , 2, , 4
$
$DRESP2,ID, LABEL, EQID, REGION, , , , , +
$+, DRESP1, NR1, NR2, ...
DRESP2, 21, DEF2, 100, , , , , , +
+, DRESP1, 11, 12, 13, 14
$
DEQATN 100 F1(A,B,C,D) = (2.*A + D)/3.0 ; +
+ F2 = (A + 2.*D)/3.0 ; +
+ F = SQRT( (B-F1)**2 + (C-F2)**2 )
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......0

6-17
 DRESP2 DEPENDENCY ON DRESP1 RESPONSES

Recall that a single DRESP1 entry may generate a number of responses.

There will likewise be as many DRESP2 responses generated when
referencing this type of DRESP1 response.

EID1
EID2
PID1 .
.
.
PID2
EIDn
DRESP2 DRESP1 .
.
.

PIDm

6-18
 EXAMPLE – BAR AXIAL PLUS BENDING STRESSES

Stress recovery locations for a BAR element yield only bending stress
components. DRESP2 entries can be used to include axial stress
component, yielding total stress.

$ ANALYSIS:
$
CBAR, 10, 100, 1, 2, 1., 1., 0.
CBAR, 11, 100, 2, 3, 1., 1., 0.
CBAR, 12, 100, 3, 4, 1., 1., 0.
CBAR, 13, 100, 4, 5, 1., 1., 0.
$
PBAR, 100, 101, 1.25E3, 2.6042E5,6.5104E4,, , , +
+, 12.5, 25., 12.5, -25., -12.5, -25., -12.5, 25.
$
$ DESIGN:
$
DRESP1, 110, SA1, STRESS, PBAR, , 2, , 100
DRESP1, 111, SA2, STRESS, PBAR, , 3, , 100
DRESP1, 112, SA3, STRESS, PBAR, , 4, , 100
DRESP1, 113, SA4, STRESS, PBAR, , 5, , 100
DRESP1, 114, AXIAL, STRESS, PBAR, , 6, , 100
$
DRESP2, 210, SR1, 303, , , , , , +
+, DRESP1, 110, 114
DRESP2, 211, SR2, 303, , , , , , +
+, DRESP1, 111, 114
DRESP2, 212, SR3, 303, , , , , , +
+, DRESP1, 112, 114
DRESP2, 213, SR4, 303, , , , , , +
+, DRESP1, 113, 114
$
DEQATN 303 SRI(POINT,AXIAL) = POINT+AXIAL

6-19
 RESTRICTIONS IN FORMING SYNTHETIC
RESPONSES

• Different types of responses (for example, weight and

displacement) can be mixed in one equation with caution. Each
associated DRESP1 entry must generate a single response only.

• Responses cannot be mixed across the subcases in one equation.

• Multiple equations must be separated by a semi-colon (;) and no

recursive reference is allowed.

6-20
 SECTION 7

EXAMPLES II

STIFFENED PANEL WEIGHT MINIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-1

25-BAR TRUSS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-12

CANTILEVER PLATE (REDUCED BASIS FORMULATION) . . . . . . . . . . . . . . . . . 7-26

 STIFFENED PANEL WEIGHT MINIMIZATION 7

Fz

10404 20204
Web Cap
Fx

10204
X
10203
Y
20200
10004
10400

10200

Web
Plate
10000

Load cases:

1. In-plane load Nx=1000 lbs/in and uniform pressure 50 psi in the

+Z direction.

2. Lumped vertical load of 10,000 lbs at grid 10203.

7-1
 STIFFENED PANEL WEIGHT MINIMIZATION (Cont.)

• Objective

• Weight minimization

• Constraints

• Stress:

• von Mises stress for plate elements ≤ 25 ksi for subcases

1, 2

• Axial stress for rod elements ≤ 25 ksi for subcases 1, 2

• Displacement:

• z-displacement at 10203 ≤ 0.1 in. for subcase 1

• z-displacement at 10203 ≤ 0.03 in. for subcase 2

7-2
 STIFFENED PANEL WEIGHT MINIMIZATION (Cont.)

• Design variables

• Thickness of web

• Thickness of plate

• Area of web cap

A 2 A 2
I 1 = I 1  ------- , I 2 = I 2  -------
0 A  0 0 A  0
1
---
2 2
I 12 = I 12  ------- , C j = C j  -------
A A
0  A 0 0  A 0

7-3
 STIFFENED PANEL WEIGHT MINIMIZATION (Cont.)

ID MSC, D200X7 $
TIME 10
SOL 200 $ OPTIMIZATION
CEND
$
TITLE = STATIC ANALYSIS OF A STIFFENED PLATE D200X7
ECHO = UNSORT
DISP = ALL
STRESS = ALL
SPC = 1
ANALYSIS = STATICS
DESOBJ(MIN) = 15 $ OBJECTIVE FUNCTION DEFINITION
$ (MIN IS THE DEFAULT)
SUBCASE 1
LABEL = LOAD CONDITION 1
LOAD = 1
DESSUB = 100 $ CONSTRAINT DEFININITION
SUBCASE 2
LABEL = LOAD CONDITION 2
LOAD = 2
DESSUB = 200 $ CONSTRAINT DEFININITION
BEGIN BULK
$
$------------------------------------------------------------------------
$ ANALYSIS MODEL:
$------------------------------------------------------------------------
$
GRID, 10000, , 0.0, 0.0, 0.0
GRID, 10001, , 2.5, 0.0, 0.0
GRID, 10002, , 5.0, 0.0, 0.0
GRID, 10003, , 7.5, 0.0, 0.0
GRID, 10004, , 10.0, 0.0, 0.0
GRID, 10100, , 0.0, 2.5, 0.0
GRID, 10101, , 2.5, 2.5, 0.0
GRID, 10102, , 5.0, 2.5, 0.0
GRID, 10103, , 7.5, 2.5, 0.0
GRID, 10104, , 10.0, 2.5, 0.0
GRID, 10200, , 0.0, 5.0, 0.0
GRID, 10201, , 2.5, 5.0, 0.0
GRID, 10202, , 5.0, 5.0, 0.0
GRID, 10203, , 7.5, 5.0, 0.0
GRID, 10204, , 10.0, 5.0, 0.0
GRID, 10300, , 0.0, 7.5, 0.0
GRID, 10301, , 2.5, 7.5, 0.0
GRID, 10302, , 5.0, 7.5, 0.0
GRID, 10303, , 7.5, 7.5, 0.0
GRID, 10304, , 10.0, 7.5, 0.0
GRID, 10400, , 0.0, 10.0, 0.0
GRID, 10401, , 2.5, 10.0, 0.0

7-4
 STIFFENED PANEL WEIGHT MINIMIZATION (Cont.)

GRID, 10402, , 5.0, 10.0, 0.0

GRID, 10403, , 7.5, 10.0, 0.0
GRID, 10404, , 10.0, 10.0, 0.0
GRID, 20100, , 0.0, 5.0, 1.0
GRID, 20101, , 2.5, 5.0, 1.0
GRID, 20102, , 5.0, 5.0, 1.0
GRID, 20103, , 7.5, 5.0, 1.0
GRID, 20104, , 10.0, 5.0, 1.0
GRID, 20200, , 0.0, 5.0, 2.0
GRID, 20201, , 2.5, 5.0, 2.0
GRID, 20202, , 5.0, 5.0, 2.0
GRID, 20203, , 7.5, 5.0, 2.0
GRID, 20204, , 10.0, 5.0, 2.0
$
CQUAD4, 1, 1, 10000, 10001, 10101, 10100
CQUAD4, 2, 1, 10001, 10002, 10102, 10101
CQUAD4, 3, 1, 10002, 10003, 10103, 10102
CQUAD4, 4, 1, 10003, 10004, 10104, 10103
CQUAD4, 5, 1, 10100, 10101, 10201, 10200
CQUAD4, 6, 1, 10101, 10102, 10202, 10201
CQUAD4, 7, 1, 10102, 10103, 10203, 10202
CQUAD4, 8, 1, 10103, 10104, 10204, 10203
CQUAD4, 9, 1, 10200, 10201, 10301, 10300
CQUAD4, 10, 1, 10201, 10202, 10302, 10301
CQUAD4, 11, 1, 10202, 10203, 10303, 10302
CQUAD4, 12, 1, 10203, 10204, 10304, 10303
CQUAD4, 13, 1, 10300, 10301, 10401, 10400
CQUAD4, 14, 1, 10301, 10302, 10402, 10401
CQUAD4, 15, 1, 10302, 10303, 10403, 10402
CQUAD4, 16, 1, 10303, 10304, 10404, 10403
$
CQUAD4, 21, 2, 10200, 10201, 20101, 20100
CQUAD4, 22, 2, 10201, 10202, 20102, 20101
CQUAD4, 23, 2, 10202, 10203, 20103, 20102
CQUAD4, 24, 2, 10203, 10204, 20104, 20103
CQUAD4, 25, 2, 20100, 20101, 20201, 20200
CQUAD4, 26, 2, 20101, 20102, 20202, 20201
CQUAD4, 27, 2, 20102, 20103, 20203, 20202
CQUAD4, 28, 2, 20103, 20104, 20204, 20203
$
CBAR, 31, 3, 20200, 20201, 0.0, 1.0, 0.0
CBAR, 32, 3, 20201, 20202, 0.0, 1.0, 0.0
CBAR, 33, 3, 20202, 20203, 0.0, 1.0, 0.0
CBAR, 34, 3, 20203, 20204, 0.0, 1.0, 0.0
$
PSHELL, 1, 1, 0.15, 1
PSHELL, 2, 1, 0.2, 1
PBAR, 3, 1, 0.144, 1.728-4,1.728-2,1.745-2,, , +PB3
+PB3, 0.06, 0.6, 0.06, -0.6, -0.06, -0.6, -0.06, 0.6

7-5
 STIFFENED PANEL WEIGHT MINIMIZATION (Cont.)

$
MAT1, 1, 1.0E+7, , 0.33, 0.283
$
FORCE, 1, 10004, , 2000.0, 1.0, 0.0, 0.0
FORCE, 1, 10104, , 2000.0, 1.0, 0.0, 0.0
FORCE, 1, 10204, , 2000.0, 1.0, 0.0, 0.0
FORCE, 1, 10304, , 2000.0, 1.0, 0.0, 0.0
FORCE, 1, 10404, , 2000.0, 1.0, 0.0, 0.0
FORCE, 2, 10203, , 10000.0,0.0, 0.0, 1.0
PLOAD2, 1, 50., 1, THRU, 16
$
SPC1, 1, 1236, 10000
SPC1, 1, 136, 10100, 10300, 10400
SPC1, 1, 36, 10001, 10002, 10003, 10004, 10104
SPC1, 1, 36, 10401, 10402, 10403, 10404, 10304
SPC1, 1, 3, 10204
SPC1, 1, 13, 10200
SPC1, 1, 5, 20100, 20101, 20102, 20103, 20104
SPC1, 1, 6, 10101, 10102, 10103, 10104
SPC1, 1, 6, 10301, 10302, 10303, 10304
$
PARAM, GRDPNT, 1
PARAM, WTMASS, 0.00259
PARAM, AUTOSPC, YES
$
$------------------------------------------------------------------------
$ DESIGN MODEL:
$------------------------------------------------------------------------
$
$...Define the design variables:
$
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
DESVAR, 1, T-PLATE,0.15, 0.001, 10.0
DESVAR, 2, T-WEB, 0.20, 0.001, 10.0
DESVAR, 3, A-BAR, 0.144, 0.001, 10.0
$
$...Relate the design variables to analysis model properties
$ (linear relations, so use DVPREL1)
$
$...Express shell thicknesses as functions of x1, x2:
$DVPREL1,ID, TYPE, PID, FID, PMIN, PMAX, C0, , +
$+, DVIDD1, COEF1, DVID2, COEF2, ...
DVPREL1,1, PSHELL, 1, 4, 0.01, , , , +DP1
+DP1, 1, 1.0
DVPREL1,2, PSHELL, 2, 4, 0.01, , , , +DP2
+DP2, 2, 1.0
$
$...Express bar cross sectional area as a function of x3:
DVPREL1,3, PBAR, 3, 4, 0.01, , , , +DP3

7-6
 STIFFENED PANEL WEIGHT MINIMIZATION (Cont.)

+DP3, 3, 1.0
$
$...Proportionally relate bar’s I1, I2, I12 to changes in area (x3):
$ (nonlinear relations require use of DVPREL2+DEQATN)
$DVPREL2,ID, TYPE, PID, FID, PMIN, PMAX, EQID, , +
$+, DESVAR,DVID1, DVID2, ..., , , , , +
$+, DTABLE,CID1, CID2, ...
DVPREL2,11, PBAR, 3, 5, 1.0-6, , 101, , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, I1INIT
DEQATN 101 I1(X3,X3INIT,I1INIT) = I1INIT ;
DELTA = SQRT(X3/X3INIT) ;
I1NEW = I1*DELTA**4
DVPREL2,12, PBAR, 3, 6, 1.0-6, , 102, , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, I2INIT
DEQATN 102 I2(X3,X3INIT,I2INIT) = I2INIT ;
DELTA = SQRT(X3/X3INIT) ;
I2NEW = I2*DELTA**4
DVPREL2,13, PBAR, 3, 7, 1.0-6, , 103, , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, I12INIT
DEQATN 103 I12(X3,X3INIT,I12INIT) = I12INIT ;
DELTA = SQRT(X3/X3INIT) ;
I12NEW = I12*DELTA**4
$
$...Modify stress recovery points accordingly:
DVPREL2,14, PBAR, 3, 12, , , 104, , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, CYINIT
DVPREL2,15, PBAR, 3, 13, , , 104, , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, CZINIT
DVPREL2,16, PBAR, 3, 14, , , 104, , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, DYINIT
DVPREL2,17, PBAR, 3, 15, -1.0, , 104, , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, DZINIT
DVPREL2,18, PBAR, 3, 16, -1.0, , 104, , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, EYINIT
DVPREL2,19, PBAR, 3, 17, -1.0, , 104, , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, EZINIT
DVPREL2,20, PBAR, 3, 18, -1.0, , 104, , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, FYINIT
DVPREL2,21, PBAR, 3, 19, , , 104, , +

7-7
 STIFFENED PANEL WEIGHT MINIMIZATION (Cont.)

+, DESVAR, 3, , , , , , , +
+, DTABLE, X3INIT, FZINIT
$
$...Equation for stress recovery points:
DEQATN 104 NEWPOINT(X3,X3INIT,POINT) = POINT*SQRT(X3/X3INIT)
$
$...Table constants for all equations:
DTABLE, X3INIT, 0.144, I1INIT, 1.728-4,I2INIT, 1.728-2,I12INIT,1.745-2,+
+, CYINIT, 0.06, CZINIT, 0.6, DYINIT, 0.06, DZINIT, -0.6, +
+, EYINIT, -0.06, EZINIT, -0.6, FYINIT, -0.06, FZINIT, 0.6
$
$...Identify the design responses:
$
$DRESP1,ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
DRESP1, 1, SBARA, STRESS, PBAR, , 7, , 3
DRESP1, 2, SBARB, STRESS, PBAR, , 14, , 3
DRESP1, 3, S13, STRESS, PSHELL, , 9, , 1
DRESP1, 6, S16, STRESS, PSHELL, , 17, , 1
DRESP1, 9, S23, STRESS, PSHELL, , 9, , 2
DRESP1, 12, S26, STRESS, PSHELL, , 17, , 2
DRESP1, 13, D1, DISP, , , 3, , 10302
DRESP1, 14, D2, DISP, , , 3, , 10203
DRESP1, 15, W, WEIGHT
$
$...Place bounds on the responses:
$
$DCONSTR,DCID, RID, LALLOW, UALLOW
DCONSTR, 10, 1, -25000.,25000.
DCONSTR, 10, 2, -25000.,25000.
DCONSTR, 10, 3, -25000.,25000.
DCONSTR, 10, 6, -25000.,25000.
DCONSTR, 10, 9, -25000.,25000.
DCONSTR, 10, 12, -25000.,25000.
DCONSTR, 20, 13, -0.1, 0.1
DCONSTR, 30, 14, -0.03, 0.03
$
$DCONADD,DCID, DC1, DC2, ...
DCONADD, 100, 10, 20 $ summed constraint set for subcase 1
DCONADD, 200, 10, 30 $ summed constraint set for subcase 2
$
$...Optional override of optimization control parameters:
$
DOPTPRM,IPRINT, 1, DESMAX, 20, DELP, 0.5, P1, 1, +
+, P2, 15
$ (DELP=0.5 allows larger moves, thus overcoming constraint
$ violations quicker)
$
ENDDATA

7-8
 STIFFENED PANEL WEIGHT MINIMIZATION (Cont.)

***** NORMAL CONVERGENCE CRITERIA SATISFIED ***** (HARD CONVERGENCE DECISION LOGIC)
**************************************************************************************
CONVERGENCE ACHIEVED BASED ON THE FOLLOWING CRITERIA
(HARD CONVERGENCE DECISION LOGIC)
RELATIVE CHANGE IN OBJECTIVE 0.0000E+00 MUST BE LESS THAN 1.0000E-03
OR ABSOLUTE CHANGE IN OBJECTIVE 0.0000E+00 MUST BE LESS THAN 1.0000E-02
--- AND ---
MAXIMUM CONSTRAINT VALUE 4.1453E-04 MUST BE LESS THAN 5.0000E-03
(CONVERGENCE TO A FEASIBLE DESIGN)
--- OR ---
MAXIMUM OF RELATIVE PROP. CHANGES 0.0000E+00 MUST BE LESS THAN 1.0000E-03
AND MAXIMUM OF RELATIVE D.V. CHANGES 0.0000E+00 MUST BE LESS THAN 1.0000E-03
(CONVERGENCE TO A BEST COMPROMISE INFEASIBLE DESIGN)
**************************************************************************************
7-9
 STIFFENED PANEL WEIGHT MINIMIZATION (Cont.)

***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************
(HARD CONVERGENCE ACHIEVED)

(SOFT CONVERGENCE ACHIEVED)

NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 8

NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 7

OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY

---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
7-10

CYCLE APPROXIMATE EXACT OF OF

NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------

INITIAL 5.784520E+00 1.311878E+00

1 1.122267E+01 1.122355E+01 -7.868297E-05 4.089278E-01

2 1.636887E+01 1.636941E+01 -3.285838E-05 -1.610547E-03
3 1.294620E+01 1.294636E+01 -1.259646E-05 -8.604820E-02
4 8.748927E+00 8.751337E+00 -2.753791E-04 1.988992E-02
5 7.379158E+00 7.379156E+00 3.230973E-07 9.885500E-02
6 7.954716E+00 7.954452E+00 3.315011E-05 4.145312E-04
7 7.954473E+00 7.954452E+00 2.637622E-06 4.145312E-04
---------------------------------------------------------------------------------------------------------------
 STIFFENED PANEL WEIGHT MINIMIZATION (Cont.)

STATIC ANALYSIS OF A STIFFENED PLATE D200X7 JULY 26, 1994 MSC/NASTRAN 7/25/94 PAGE 92 DESIGN VARIABLE HISTORY
------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
-------------------------------------------------------------------------------------------------------------------------
1 | 1 | T-PLATE | 1.5000E-01 : 3.0000E-01 : 4.4321E-01 : 2.8106E-01 : 1.4038E-01 : 1.0879E-01 :
2 | 2 | T-WEB | 2.0000E-01 : 4.0000E-01 : 6.1265E-01 : 8.4962E-01 : 8.2809E-01 : 7.5169E-01 :
3 | 3 | A-BAR | 1.4400E-01 : 1.6592E-01 : 1.2682E-01 : 6.4814E-02 : 3.2407E-02 : 1.6165E-02 :
-------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | 6 : 7 : 8 : 9 : 10 : 11 :
-------------------------------------------------------------------------------------------------------------------------
1 | 1 | T-PLATE | 1.1169E-01 : 1.1169E-01 :
2 | 2 | T-WEB | 8.3887E-01 : 8.3887E-01 :
3 | 3 | A-BAR | 1.6165E-02 : 1.6165E-02 :
-------------------------------------------------------------------------------------------------------------------------
*** USER INFORMATION MESSAGE 6464 (DOM12E)
7-11

RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 7.

 25-BAR TRUSS

This design problem calls for a minimum weight structure subject to

member stress, simple Euler buckling, and joint displacement
constraints, under static loading conditions.

Each structural element is a hollow tube with a thickness-to-diameter

ratio of 1:10. Diameters will be the design variables.

6 3
4
5
10
7 Y

8
9 X

7-12
 25-BAR TRUSS (Cont.)

• Analysis model description

• Three-dimensional

• Symmetric with respect to the x-z plane and the y-z plane

• Material

• E = 1.0E7 psi

3
• Weight density = 0.1 lbs/in

• Two static loading conditions

• Design model description

• Minimization of structural weight

• Design variables

• Eight independent diameters, controlling all 25 elements.

Wall thickness-to-diameter ratio of 1:10

7-13
 25-BAR TRUSS (Cont.)

• Constraints

• Allowable stress – tensile 40,000 psi, compressive –40,000

psi

• Euler buckling constraint on all members

• Displacement constraints – ±0.35 inches for translational

degrees of freedom at grids 1 and 2

Recall that Euler buckling is caused by compressive stresses. We

require that the stresses in each member be numerically greater than
the stress that will induce buckling, or

σ cr ≤ F s ∗σ

where Fs is a factor of safety.

Rearranging,

F s ∗σ
-------------- ≤ 1
σ cr

For Euler buckling,

P cr 2
π EI
σ cr = --------- ; P cr = – ------------
A 2
L

7-14
 25-BAR TRUSS (Cont.)

or,

2
π EI
σ cr = – ------------
2
AL

With:

A = 2πR∗t
π 2
= ------D
10
3
I = πR ∗t
π 4
= ------D
80

(2)
Our synthetic responses r are:

2
(2) σ 8L F s σ
r = --------- = – --------------------- ≤ 1
σ cr ( πD ) E
2

7-15
 25-BAR TRUSS (Cont.)

ID MSC, D200X3 $
TIME 10
SOL 200 $ OPTIMIZATION
CEND
TITLE = OPTIMAL SIZING OF A 25-BAR TRUSS - D200X3
SUBTITLE = EIGHT INDEPENDENT CROSS SECTIONAL AREAS,
SUBTITLE = PLUS EULER BUCKLING
ECHO = UNSORT
OLOAD = ALL
DISP = ALL
SPCFORCE = ALL
ELFORCE = ALL
STRESS = ALL
SPC = 100
ANALYSIS = STATICS $
DESOBJ(MIN) = 11 $ OBJECTIVE FUNCTION DEFINITION
DESSUB = 12 $ CONSTRAINT DEFININITION
SUBCASE 1
LABEL = LOAD CONDITION 1
LOAD = 300
SUBCASE 2
LABEL = LOAD CONDITION 2
LOAD = 310
BEGIN BULK
$
$-----------------------------------------------------------------------------
$ ANALYSIS MODEL
$-----------------------------------------------------------------------------
$
GRDSET, , , , , , , 456
MAT1, 1, 10.0E6, , , 0.1
SPC1, 100, 123, 7, THRU, 10
GRID, 1 , , -37.5, 0.0, 200.0
GRID, 2 , , 37.5, 0.0, 200.0
GRID, 3 , , -37.5, 37.5, 100.0
GRID, 4 , , 37.5, 37.5, 100.0
GRID, 5 , , 37.5, -37.5, 100.0
GRID, 6 , , -37.5, -37.5, 100.0
GRID, 7 , , -100.0, 100.0, 0.0
GRID, 8 , , 100.0, 100.0, 0.0
GRID, 9 , , 100.0, -100.0, 0.0
GRID, 10, , -100.0, -100.0, 0.0
CROD, 1 , 1, 1, 2
CROD, 2 , 2, 1, 4
CROD, 3 , 2, 2, 3
CROD, 4 , 2, 1, 5
CROD, 5 , 2, 2, 6
CROD, 6 , 3, 2, 4

7-16
 25-BAR TRUSS (Cont.)

CROD, 7 , 3, 2, 5
CROD, 8 , 3, 1, 3
CROD, 9 , 3, 1, 6
CROD, 10, 4, 3, 6
CROD, 11, 4, 4, 5
CROD, 12, 5, 3, 4
CROD, 13, 5, 5, 6
CROD, 14, 6, 3, 10
CROD, 15, 6, 6, 7
CROD, 16, 6, 4, 9
CROD, 17, 6, 5, 8
CROD, 18, 7, 4, 7
CROD, 19, 7, 3, 8
CROD, 20, 7, 5, 10
CROD, 21, 7, 6, 9
CROD, 22, 8, 6, 10
CROD, 23, 8, 3, 7
CROD, 24, 8, 4, 8
CROD, 25, 8, 5, 9
$
PROD, 1, 1, 2.0, 0.0
PROD, 2, 1, 2.0, 0.0
PROD, 3, 1, 2.0, 0.0
PROD, 4, 1, 2.0, 0.0
PROD, 5, 1, 2.0, 0.0
PROD, 6, 1, 2.0, 0.0
PROD, 7, 1, 2.0, 0.0
PROD, 8, 1, 2.0, 0.0
$
FORCE, 300, 1, , 1.0, 1000., 10000.,-5000.
FORCE, 300, 2, , 1.0, 0., 10000.,-5000.
FORCE, 300, 3, , 1.0, 500., 0., 0.
FORCE, 300, 6, , 1.0, 500., 0., 0.
FORCE, 310, 1, , 1.0, 0., 20000.,-5000.
FORCE, 310, 2, , 1.0, 0., -20000.,-5000.
$
$-----------------------------------------------------------------------------
$ DESIGN MODEL
$-----------------------------------------------------------------------------
$
$...Define the design variables
$ average (mean) diameters of tubes:
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
$
DESVAR, 1, D1, 2.5, 0.01, 100.0
DESVAR, 2, D2, 2.5, 0.01, 100.0
DESVAR, 3, D3, 2.5, 0.01, 100.0
DESVAR, 4, D4, 2.5, 0.01, 100.0

7-17
 25-BAR TRUSS (Cont.)

DESVAR, 5, D5, 2.5, 0.01, 100.0

DESVAR, 6, D6, 2.5, 0.01, 100.0
DESVAR, 7, D7, 2.5, 0.01, 100.0
DESVAR, 8, D8, 2.5, 0.01, 100.0
$
$...Relate the design variables to the analysis model properties
$ with (average diameter)/thickness = 10., fixed for all sections:
$
$DVPREL2,ID, TYPE, PID, FID, PMIN, PMAX, EQID, , +
$+, DESVAR, DVID1, DVID2, ..., , , , , +
$+, DTABLE, CID1, CID2, ...
$
DVPREL2,1, PROD, 1, 4, , , 100, , +
+ ,DESVAR, 1
DVPREL2,2, PROD, 2, 4, , , 100, , +
+ ,DESVAR, 2
DVPREL2,3, PROD, 3, 4, , , 100, , +
+ ,DESVAR, 3
DVPREL2,4, PROD, 4, 4, , , 100, , +
+ ,DESVAR, 4
DVPREL2,5, PROD, 5, 4, , , 100, , +
+ ,DESVAR, 5
DVPREL2,6, PROD, 6, 4, , , 100, , +
+ ,DESVAR, 6
DVPREL2,7, PROD, 7, 4, , , 100, , +
+ ,DESVAR, 7
DVPREL2,8, PROD, 8, 4, , , 100, , +
+ ,DESVAR, 8
$
$DEQATN EQUID F() = ...
$
DEQATN 100 AREA(DAVG) = (PI(1.)/10.)*DAVG**2
$
$...Identify the responses to be used in the design model
$DRESP1,ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
$
DRESP1, 1 , S1, STRESS, PROD, , 2, , 1
DRESP1, 2 , S2, STRESS, PROD, , 2, , 2
DRESP1, 3 , S3, STRESS, PROD, , 2, , 3
DRESP1, 4 , S4, STRESS, PROD, , 2, , 4
DRESP1, 5 , S5, STRESS, PROD, , 2, , 5
DRESP1, 6 , S6, STRESS, PROD, , 2, , 6
DRESP1, 7 , S7, STRESS, PROD, , 2, , 7
DRESP1, 8 , S8, STRESS, PROD, , 2, , 8
DRESP1, 9 , D1, DISP , , , 123, , 1
DRESP1, 10, D2, DISP , , , 123, , 2
DRESP1, 11, W , WEIGHT
$

7-18
 25-BAR TRUSS (Cont.)

$...Formulate the second level responses (here, simple Euler buckling)

$
$DRESP2,ID, LABEL, EQID, REGION, , , , , +
$+, DESVAR, DVID1, DVID2, ..., , , , , +
$+, DTABLE, LABEL1, LABEL2, ..., , , , , +
$+, DRESP1, NR1, NR2, ..., , , , , +
$+, DNODE, NID1, DIR1, NID2, DIR2, ..., , , +
$
$ buckling is the equation return value.
$
DEQATN 110 NUM (DAVG,FS,E,L,SIGMA) = 0.0;
NUM = 8.0*L**2*SIGMA;
DENOM = (PI(1.)*DAVG)**2*E;
BUCKLING = - FS*NUM/DENOM
$
DRESP2, 16, BUCK1, 110, , , , , , +
+, DESVAR, 1, , , , , , , +
+, DTABLE, FS, E, L1, , , , , +
+, DRESP1, 1
DRESP2, 17, BUCK2, 110, , , , , , +
+, DESVAR, 2, , , , , , , +
+, DTABLE, FS, E, L2, , , , , +
+, DRESP1, 2
DRESP2, 18, BUCK3, 110, , , , , , +
+, DESVAR, 3, , , , , , , +
+, DTABLE, FS, E, L3, , , , , +
+, DRESP1, 3
DRESP2, 19, BUCK4, 110, , , , , , +
+, DESVAR, 4, , , , , , , +
+, DTABLE, FS, E, L4, , , , , +
+, DRESP1, 4
DRESP2, 20, BUCK5, 110, , , , , , +
+, DESVAR, 5, , , , , , , +
+, DTABLE, FS, E, L5, , , , , +
+, DRESP1, 5
DRESP2, 21, BUCK6, 110, , , , , , +
+, DESVAR, 6, , , , , , , +
+, DTABLE, FS, E, L6, , , , , +
+, DRESP1, 6
DRESP2, 22, BUCK7, 110, , , , , , +
+, DESVAR, 7, , , , , , , +
+, DTABLE, FS, E, L7, , , , , +
+, DRESP1, 7
DRESP2, 23, BUCK8, 110, , , , , , +
+, DESVAR, 8, , , , , , , +
+, DTABLE, FS, E, L8, , , , , +
+, DRESP1, 8
$
$...Table constants

7-19
 25-BAR TRUSS (Cont.)

$DTABLE,LABEL1, VALUE1, LABEL2, VALUE2, LABEL3, VALUE3, LABEL4, VALUE4 +

$+, LABEL5, VALUE5, ...
$
DTABLE, L1, 75.00, L2, 130.50, L3, 106.80, L4, 75.00, +
+, L5, 75.00, L6, 181.14, L7, 181.14, L8, 133.46, +
+, E, 1.0E7, FS, 1.25
$
$...Define the design constraints
$CONSTR,DCID, RID, LALLOW, UALLOW
$
DCONSTR,10, 1 , -40000.,40000.
DCONSTR,10, 2 , -40000.,40000.
DCONSTR,10, 3 , -40000.,40000.
DCONSTR,10, 4 , -40000.,40000.
DCONSTR,10, 5 , -40000.,40000.
DCONSTR,10, 6 , -40000.,40000.
DCONSTR,10, 7 , -40000.,40000.
DCONSTR,10, 8 , -40000.,40000.
DCONSTR,10, 9 , -0.35 ,0.35
DCONSTR,10, 10, -0.35 ,0.35
$
DCONSTR,11, 16, , 1.0
DCONSTR,11, 17, , 1.0
DCONSTR,11, 18, , 1.0
DCONSTR,11, 19, , 1.0
DCONSTR,11, 20, , 1.0
DCONSTR,11, 21, , 1.0
DCONSTR,11, 22, , 1.0
DCONSTR,11, 23, , 1.0
$
$...Combine the two constraint sets
$ (equivalent to just putting all into the same set to begin with)
DCONADD,12, 10, 11
$
$...Override optimization parameter defaults:
$
DOPTPRM,IPRINT, 3, DESMAX, 15, p1, 1, p2, 15
$
ENDDATA

7-20
 25-BAR TRUSS (Cont.)

***** NORMAL CONVERGENCE CRITERIA SATISFIED ***** (HARD CONVERGENCE DECISION LOGIC)

**************************************************************************************
CONVERGENCE ACHIEVED BASED ON THE FOLLOWING CRITERIA
(HARD CONVERGENCE DECISION LOGIC)

RELATIVE CHANGE IN OBJECTIVE 8.9169E-04 MUST BE LESS THAN 1.0000E-03

OR ABSOLUTE CHANGE IN OBJECTIVE 7.2791E-01 MUST BE LESS THAN 1.0000E-02
--- AND ---
MAXIMUM CONSTRAINT VALUE 8.8620E-04 MUST BE LESS THAN 5.0000E-03
(CONVERGENCE TO A FEASIBLE DESIGN)
--- OR ---
MAXIMUM OF RELATIVE PROP. CHANGES 2.0000E-01 MUST BE LESS THAN 1.0000E-03
AND MAXIMUM OF RELATIVE D.V. CHANGES 1.0557E-01 MUST BE LESS THAN 1.0000E-03
7-21

(CONVERGENCE TO A BEST COMPROMISE INFEASIBLE DESIGN)

**************************************************************************************
 25-BAR TRUSS (Cont.)
***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************

(HARD CONVERGENCE ACHIEVED)

NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 13

NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 12

OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY

---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
CYCLE APPROXIMATE EXACT OF OF
NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------

INITIAL 6.493687E+02 2.031843E-00

1 8.805606E+02 8.805025E+02 6.599126E-05 1.628765E-01

7-22

2 8.741666E+02 8.741613E+02 6.074461E-06 -2.911091E-04

3 8.605269E+02 8.605145E+02 1.439852E-05 1.001358E-04

4 8.496406E+02 8.496353E+02 6.321647E-06 2.777576E-04

5 8.410117E+02 8.410120E+02 -2.902939E-07 3.701448E-04

6 8.340327E+02 8.340328E+02 -2.195423E-07 6.405115E-04

7 8.280418E+02 8.280425E+02 -8.108119E-07 7.916689E-04

8 8.236416E+02 8.236408E+02 9.633533E-07 6.154776E-04

9 8.201868E+02 8.201854E+02 1.711575E-06 7.835627E-04

10 8.175861E+02 8.175862E+02 -7.465287E-08 8.921623E-04

11 8.163202E+02 8.163205E+02 -3.738431E-07 7.225275E-04

12 8.155923E+02 8.155926E+02 -2.993414E-07 8.862019E-04

 25-BAR TRUSS (Cont.)

DESIGN VARIABLE HISTORY

----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | D1 | 2.5000E+00 : 2.7227E+00 : 2.4340E+00 : 2.1757E+00 : 1.9447E+00 : 1.7382E+00 :
2 | 2 | D2 | 2.5000E+00 : 2.9058E+00 : 3.0266E+00 : 3.0244E+00 : 3.0220E+00 : 3.0196E+00 :
3 | 3 | D3 | 2.5000E+00 : 2.8047E+00 : 2.8786E+00 : 2.8798E+00 : 2.8813E+00 : 2.8831E+00 :
4 | 4 | D4 | 2.5000E+00 : 2.0226E+00 : 1.8091E+00 : 1.6181E+00 : 1.4472E+00 : 1.2945E+00 :
5 | 5 | D5 | 2.5000E+00 : 3.2012E+00 : 2.8632E+00 : 2.5610E+00 : 2.2906E+00 : 2.0488E+00 :
6 | 6 | D6 | 2.5000E+00 : 2.6980E+00 : 2.5463E+00 : 2.5437E+00 : 2.5409E+00 : 2.5379E+00 :
7 | 7 | D7 | 2.5000E+00 : 3.3196E+00 : 3.3112E+00 : 3.3106E+00 : 3.3104E+00 : 3.3101E+00 :
8 | 8 | D8 | 2.5000E+00 : 2.8360E+00 : 2.9661E+00 : 2.9710E+00 : 2.9759E+00 : 2.9816E+00 :
----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | 6 : 7 : 8 : 9 : 10 : 11 :
----------------------------------------------------------------------------------------------------------------------------------
7-23

1 | 1 | D1 | 1.5535E+00 : 1.3885E+00 : 1.2409E+00 : 1.1091E+00 : 9.9205E-01 : 8.8731E-01 :

2 | 2 | D2 | 3.0167E+00 : 3.0107E+00 : 3.0074E+00 : 3.0039E+00 : 2.9999E+00 : 2.9960E+00 :
3 | 3 | D3 | 2.8849E+00 : 2.8880E+00 : 2.8900E+00 : 2.8923E+00 : 2.8946E+00 : 2.8969E+00 :
4 | 4 | D4 | 1.1578E+00 : 1.0356E+00 : 9.2624E-01 : 8.2845E-01 : 7.4100E-01 : 6.6297E-01 :
5 | 5 | D5 | 1.8325E+00 : 1.6390E+00 : 1.4660E+00 : 1.3112E+00 : 1.1851E+00 : 1.1847E+00 :
6 | 6 | D6 | 2.5348E+00 : 2.5300E+00 : 2.5271E+00 : 2.5241E+00 : 2.5214E+00 : 2.5192E+00 :
7 | 7 | D7 | 3.3098E+00 : 3.3092E+00 : 3.3094E+00 : 3.3100E+00 : 3.3113E+00 : 3.3130E+00 :
8 | 8 | D8 | 2.9871E+00 : 2.9928E+00 : 2.9980E+00 : 3.0034E+00 : 3.0077E+00 : 3.0068E+00 :
---------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | 12 : 13 : 14 : 15 : 16 : 17 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | D1 | 8.5227E-01 :
2 | 2 | D2 | 2.9947E+00 :
3 | 3 | D3 | 2.8977E+00 :
4 | 4 | D4 | 5.9298E-01 :
5 | 5 | D5 | 1.1853E+00 :
6 | 6 | D6 | 2.5172E+00 :
7 | 7 | D7 | 3.3137E+00 :
8 | 8 | D8 | 3.0071E+00 :
----------------------------------------------------------------------------------------------------------------------------------
*** USER INFORMATION MESSAGE 6464 (DOM12E)
RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 12.
25-BAR TRUSS (Cont.)

7-24
25-BAR TRUSS (Cont.)

7-25
 CANTILEVER PLATE (REDUCED BASIS
FORMULATION)

Purpose: To illustrate a reduced basis formulation for property

optimization using the DLINK entry.

Design Model Description

p1
Y
p2

t1 t2 t3 t4 t5 t6 t7 t8 10"
40"

• Design variables – α1, α2, α3

• These three parameters determine the longitudinal thickness

distribution of the plate

• Initial design α1 = α2 = α3 = 1.0

• Design objective

• Weight minimization

7-26
 CANTILEVER PLATE (REDUCED BASIS
FORMULATION) (Cont.)

• Design constraints

• Tip displacement ≤ 2 inches

• Major principal stresses at centroid ≤ 50,000 psi

• von Mises stress at centroid ≤ 29,000 psi

Load case 1: Two 50,000 lbf loads at the plate tip

Load case 2: 60.44 psi uniform pressure acting on the upper surface

Design variable to plate thickness relation:

t1 1.0 1.0 1.0

t2 1.0 0.875 0.7656
t 3 = α 1 1.0 + α 2 0.75 + α 3 0.5625
. . . .
. . . .
. . . .
t8 1.0 0.125 0.0156

7-27
 CANTILEVER PLATE (REDUCED BASIS
FORMULATION) (Cont.)

ID MSC, D200X5 $
TIME 10
SOL 200 $ OPTIMIZATION
CEND
TITLE = CANTILEVERED PLATE - D200X5 D200X5
SUBTITLE = REDUCED BASIS FORMULATION
SPC = 100
OLOAD = ALL
DISP = ALL
STRESS = ALL
DESOBJ(MIN) = 35 $ OBJECTIVE FUNCTION DEFINITION
SUBCASE 1
ANALYSIS = STATICS
LABEL = LOAD CONDITION 1
LOAD = 300
DESSUB = 10
SUBCASE 2
ANALYSIS = STATICS
LABEL = LOAD CONDITION 2
LOAD = 310
DESSUB = 10
BEGIN BULK
$
$------------------------------------------------------------------------
$ ANALYSIS MODEL:
$------------------------------------------------------------------------
$
MAT1, 51, 1.0E+7, , 0.33, 0.1, , , , +M2
+M2, 50000., 50000., 29000.
SPC1, 100, 123456, 1, 11, 21
GRID, 1 , , 0., -5., 0.
GRID, 2 , , 5., -5., 0.
GRID, 3 , , 10., -5., 0.
GRID, 4 , , 15., -5., 0.
GRID, 5 , , 20., -5., 0.
GRID, 6 , , 25., -5., 0.
GRID, 7 , , 30., -5., 0.
GRID, 8 , , 35., -5., 0.
GRID, 9 , , 40., -5., 0.
GRID, 11, , 0., 0., 0.
GRID, 12, , 5., 0., 0.
GRID, 13, , 10., 0., 0.
GRID, 14, , 15., 0., 0.
GRID, 15, , 20., 0., 0.
GRID, 16, , 25., 0., 0.
GRID, 17, , 30., 0., 0.
GRID, 18, , 35., 0., 0.
GRID, 19, , 40., 0., 0.

7-28
 CANTILEVER PLATE (REDUCED BASIS
FORMULATION) (Cont.)

GRID, 21, , 0., 5., 0.

GRID, 22, , 5., 5., 0.
GRID, 23, , 10., 5., 0.
GRID, 24, , 15., 5., 0.
GRID, 25, , 20., 5., 0.
GRID, 26, , 25., 5., 0.
GRID, 27, , 30., 5., 0.
GRID, 28, , 35., 5., 0.
GRID, 29, , 40., 5., 0.
$
CQUAD4, 1, 1, 1, 2, 12, 11
CQUAD4, 2, 2, 2, 3, 13, 12
CQUAD4, 3, 3, 3, 4, 14, 13
CQUAD4, 4, 4, 4, 5, 15, 14
CQUAD4, 5, 5, 5, 6, 16, 15
CQUAD4, 6, 6, 6, 7, 17, 16
CQUAD4, 7, 7, 7, 8, 18, 17
CQUAD4, 8, 8, 8, 9, 19, 18
CQUAD4, 11, 1, 11, 12, 22, 21
CQUAD4, 12, 2, 12, 13, 23, 22
CQUAD4, 13, 3, 13, 14, 24, 23
CQUAD4, 14, 4, 14, 15, 25, 24
CQUAD4, 15, 5, 15, 16, 26, 25
CQUAD4, 16, 6, 16, 17, 27, 26
CQUAD4, 17, 7, 17, 18, 28, 27
CQUAD4, 18, 8, 18, 19, 29, 28
PSHELL, 1, 51, 3.0, 51, , 51
PSHELL, 2, 51, 2.640625,51
PSHELL, 3, 51, 2.3125, 51
PSHELL, 4, 51, 2.015625,51
PSHELL, 5, 51, 1.75, 51
PSHELL, 6, 51, 1.515625,51
PSHELL, 7, 51, 1.3125, 51
PSHELL, 8, 51, 1.140625,51
$ 2 3 4 5 6 7 8 9 10
FORCE, 300, 9, , 50000., 0.0, 0.0, -1.0
FORCE, 300, 29, , 50000., 0.0, 0.0, -1.0
PLOAD2, 310, -60.44, 1, THRU, 8
PLOAD2, 310, -60.44, 11, THRU, 18
$
$------------------------------------------------------------------------
$ DESIGN MODEL:
$------------------------------------------------------------------------
$
$...Define the design variables
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
$ (This group will be the dependent design variables:)

7-29
 CANTILEVER PLATE (REDUCED BASIS
FORMULATION) (Cont.)

DESVAR, 1, T1, 3.0, 0.001, 100.0

DESVAR, 2, T2, 2.640625,0.001, 100.0
DESVAR, 3, T3, 2.3125, 0.001, 100.0
DESVAR, 4, T4, 2.015625,0.001, 100.0
DESVAR, 5, T5, 1.75, 0.001, 100.0
DESVAR, 6, T6, 1.515625,0.001, 100.0
DESVAR, 7, T7, 1.3125, 0.001, 100.0
DESVAR, 8, T8, 1.140625,0.001, 100.0
$ (This group will be the independent design variables:)
DESVAR, 9, ALPHA1, 1.0, -1.+10, 1.+10
DESVAR, 10, ALPHA2, 1.0, -1.+10, 1.+10
DESVAR, 11, ALPHA3, 1.0, -1.+10, 1.+10
$
$...Explicit design variable linking
$
$DLINK, ID, DDVID, CO, CMULT, IDV1, C1, IDV2, C2, +
$+, IDV3, C3, ...
DLINK, 1 , 1, , , 9, 1.0, 10, 1.0, +DL1
+DL1 , 11, 1.0
DLINK, 2 , 2, , , 9, 1.0, 10, 0.8750, +DL2
+DL2 , 11, 0.765625
DLINK, 3 , 3, , , 9, 1.0, 10, 0.7500, +DL3
+DL3 , 11, 0.5625
DLINK, 4 , 4, , , 9, 1.0, 10, 0.6250, +DL4
+DL4 , 11, 0.390625
DLINK, 5 , 5, , , 9, 1.0, 10, 0.5000, +DL5
+DL5 , 11, 0.2500
DLINK, 6 , 6, , , 9, 1.0, 10, 0.3750, +DL6
+DL6 , 11, 0.140625
DLINK, 7 , 7, , , 9, 1.0, 10, 0.2500, +DL7
+DL7 , 11, 0.0625
DLINK, 8 , 8, , , 9, 1.0, 10, 0.1250, +DL8
+DL8 , 11, 0.015625
$
$...Express analysis model properties linearly in terms of design variables
$DVPREL1,ID, TYPE, PID, FID, PMIN, PMAX, C0, , +
$+, DVID1, COEF1, DVID2, COEF2, ...
DVPREL1,1, PSHELL, 1, 4, , , , , +DP1
+DP1, 1, 1.0
DVPREL1,2, PSHELL, 2, 4, , , , , +DP2
+DP2, 2, 1.0
DVPREL1,3, PSHELL, 3, 4, , , , , +DP3
+DP3, 3, 1.0
DVPREL1,4, PSHELL, 4, 4, , , , , +DP4
+DP4, 4, 1.0
DVPREL1,5, PSHELL, 5, 4, , , , , +DP5
+DP5, 5, 1.0

7-30
 CANTILEVER PLATE (REDUCED BASIS
FORMULATION) (Cont.)

DVPREL1,6, PSHELL, 6, 4, , , , , +DP6

+DP6, 6, 1.0
DVPREL1,7, PSHELL, 7, 4, , , , , +DP7
+DP7, 7, 1.0
DVPREL1,8, PSHELL, 8, 4, , , , , +DP8
+DP8, 8, 1.0
$
$...Identify the design responses
$DRESP1,ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
DRESP1, 1, S11, STRESS, PSHELL, , 7, , 1
DRESP1, 2, S12, STRESS, PSHELL, , 9, , 1
DRESP1, 3, S13, STRESS, PSHELL, , 16, , 1
DRESP1, 4, S14, STRESS, PSHELL, , 17, , 1
DRESP1, 5, S21, STRESS, PSHELL, , 7, , 2
DRESP1, 6, S22, STRESS, PSHELL, , 9, , 2
DRESP1, 7, S23, STRESS, PSHELL, , 16, , 2
DRESP1, 8, S24, STRESS, PSHELL, , 17, , 2
DRESP1, 9, S31, STRESS, PSHELL, , 7, , 3
DRESP1, 10, S32, STRESS, PSHELL, , 9, , 3
DRESP1, 11, S33, STRESS, PSHELL, , 16, , 3
DRESP1, 12, S34, STRESS, PSHELL, , 17, , 3
DRESP1, 13, S41, STRESS, PSHELL, , 7, , 4
DRESP1, 14, S42, STRESS, PSHELL, , 9, , 4
DRESP1, 15, S43, STRESS, PSHELL, , 16, , 4
DRESP1, 16, S44, STRESS, PSHELL, , 17, , 4
DRESP1, 17, S51, STRESS, PSHELL, , 7, , 5
DRESP1, 18, S52, STRESS, PSHELL, , 9, , 5
DRESP1, 19, S53, STRESS, PSHELL, , 16, , 5
DRESP1, 20, S54, STRESS, PSHELL, , 17, , 5
DRESP1, 21, S61, STRESS, PSHELL, , 7, , 6
DRESP1, 22, S62, STRESS, PSHELL, , 9, , 6
DRESP1, 23, S63, STRESS, PSHELL, , 16, , 6
DRESP1, 24, S64, STRESS, PSHELL, , 17, , 6
DRESP1, 25, S71, STRESS, PSHELL, , 7, , 7
DRESP1, 26, S72, STRESS, PSHELL, , 9, , 7
DRESP1, 27, S73, STRESS, PSHELL, , 16, , 7
DRESP1, 28, S74, STRESS, PSHELL, , 17, , 7
DRESP1, 29, S81, STRESS, PSHELL, , 7, , 8
DRESP1, 30, S82, STRESS, PSHELL, , 9, , 8
DRESP1, 31, S83, STRESS, PSHELL, , 16, , 8
DRESP1, 32, S84, STRESS, PSHELL, , 17, , 8
DRESP1, 33, D1 , DISP , , , 3, , 9
DRESP1, 34, D2 , DISP , , , 3, , 29
$
$...Define the response to be used as the objective function:
DRESP1, 35, W , WEIGHT
$

7-31
 CANTILEVER PLATE (REDUCED BASIS
FORMULATION) (Cont.)

$...Define the design constraints

$DCONSTR,DCID, RID, LALLOW, UALLOW
DCONSTR,10, 1, -50000.,50000.
DCONSTR,10, 2, -29000.,29000.
DCONSTR,10, 3, -50000.,50000.
DCONSTR,10, 4, -29000.,29000.
DCONSTR,10, 5, -50000.,50000.
DCONSTR,10, 6, -29000.,29000.
DCONSTR,10, 7, -50000.,50000.
DCONSTR,10, 8, -29000.,29000.
DCONSTR,10, 9, -50000.,50000.
DCONSTR,10, 10, -29000.,29000.
DCONSTR,10, 11, -50000.,50000.
DCONSTR,10, 12, -29000.,29000.
DCONSTR,10, 13, -50000.,50000.
DCONSTR,10, 14, -29000.,29000.
DCONSTR,10, 15, -50000.,50000.
DCONSTR,10, 16, -29000.,29000.
DCONSTR,10, 17, -50000.,50000.
DCONSTR,10, 18, -29000.,29000.
DCONSTR,10, 19, -50000.,50000.
DCONSTR,10, 20, -29000.,29000.
DCONSTR,10, 21, -50000.,50000.
DCONSTR,10, 22, -29000.,29000.
DCONSTR,10, 23, -50000.,50000.
DCONSTR,10, 24, -29000.,29000.
DCONSTR,10, 25, -50000.,50000.
DCONSTR,10, 26, -29000.,29000.
DCONSTR,10, 27, -50000.,50000.
DCONSTR,10, 28, -29000.,29000.
DCONSTR,10, 29, -50000.,50000.
DCONSTR,10, 30, -29000.,29000.
DCONSTR,10, 31, -50000.,50000.
DCONSTR,10, 32, -29000.,29000.
DCONSTR,10, 33, -2., 2.
DCONSTR,10, 34, -2., 2.
$
$...Override optimization parameter defaults (optional)
DOPTPRM,DESMAX, 10, DELP, 0.5, DPMIN, .01, DELX, 2.0, +
+, DELB, 0.01, CONV2, 0.1
$
ENDDATA

7-32
 CANTILEVER PLATE (REDUCED BASIS FORMULATION) (Cont.)

***** NORMAL CONVERGENCE CRITERIA SATISFIED ***** (HARD CONVERGENCE DECISION LOGIC)

**************************************************************************************
CONVERGENCE ACHIEVED BASED ON THE FOLLOWING CRITERIA
(HARD CONVERGENCE DECISION LOGIC)

RELATIVE CHANGE IN OBJECTIVE 0.0000E+00 MUST BE LESS THAN 1.0000E-03

OR ABSOLUTE CHANGE IN OBJECTIVE 0.0000E+00 MUST BE LESS THAN 1.0000E-01
--- AND ---
MAXIMUM CONSTRAINT VALUE 7.6893E-04 MUST BE LESS THAN 5.0000E-03
(CONVERGENCE TO A FEASIBLE DESIGN)
--- OR ---
MAXIMUM OF RELATIVE PROP. CHANGES 0.0000E+00 MUST BE LESS THAN 1.0000E-03
AND MAXIMUM OF RELATIVE D.V. CHANGES 2.2390E-07 MUST BE LESS THAN 1.0000E-03
7-33

(CONVERGENCE TO A BEST COMPROMISE INFEASIBLE DESIGN)

**************************************************************************************
 CANTILEVER PLATE (REDUCED BASIS FORMULATION) (Cont.)

***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************

(HARD CONVERGENCE ACHIEVED)

(SOFT CONVERGENCE ACHIEVED)

NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 9

NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 8

OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY

---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
CYCLE APPROXIMATE EXACT OF OF
7-34

NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT

---------------------------------------------------------------------------------------------------------------
INITIAL 7.843750E+01 1.072155E+01

1 1.170497E+02 1.170496E+02 6.518085E-07 4.135007E+00

2 1.616799E+02 1.616796E+02 1.604404E-06 1.589539E+00

3 2.065549E+02 2.065548E+02 2.954913E-07 5.076001E-01

4 2.363098E+02 2.363113E+02 -6.392500E-06 1.115919E-01

5 2.423183E+02 2.423193E+02 -4.093034E-06 3.467807E-02

6 2.447022E+02 2.447021E+02 4.988528E-07 1.603751E-02

7 2.461020E+02 2.461019E+02 1.240038E-07 7.689251E-04

8 2.461019E+02 2.461019E+02 0.000000E+00 7.689251E-04

---------------------------------------------------------------------------------------------------------------
 CANTILEVER PLATE (REDUCED BASIS FORMULATION) (Cont.)
7-35
 SECTION 8

SPECIAL MODELING TOPICS II

BEAM CROSS-SECTION LIBRARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-1

BEAM CROSS-SECTION LIBRAY OPTIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . . 8-2

COMPARISON BETWEEN DESIGNING PBAR AND PBARL. . . . . . . . . . . . . . . . . . 8-3

DVPREL1 INCLUDED ON DRESP2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-5

DESIGN VARIABLE LINKING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-6

CONSTRAINT EQUATION SCALING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-9

BEAM DESIGN – GENERAL ISSUES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-13

BEAM DESIGN – SIMPLE EXAMPLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-16

 BEAM CROSS-SECTION LIBRARY 8

• May specify cross section dimensions (e.g., height and width for a
rectangular section) rather than properties (e.g., I, A, etc.)

• Cross sectional properties are automatically calculated.

• Two new Bulk Data entries: PBARL and PBEAML

8-1
 BEAM CROSS-SECTION LIBRAY OPTIMIZATION

• You may define these sections in MSC/PATRAN V6 by selecting

Beam type in the “Element Properties” form.

• You may optimize section dimension without writing DVPREL2 and

DEQATN Bulk Data entries for the BAR element.

• The PBARL entry can be referenced by a DVPREL1 entry

• For BEAM element, you still have to write the DVPREL2 and
DEQATN entries to optimize section dimension.

8-2
COMPARISON BETWEEN DESIGNING PBAR AND PBARL

1
--- B, --- H
1
2 2 
C D

1 H

F E
B

*************************************
* Input Data for Designing PBAR *
*************************************
$
PBAR 9 1 5.0 10.4167 0.4167
-0.5 2.5 0.5 2.5 0.5 -2.5 -0.5 -2.5
C
$ ID LABEL INIT MIN MAX
DESVAR 1 HEIGHT 5.0 0.01 10.
DESVAR 2 WIDTH 1.0 0.01 10.
$
$ Explict expression for cross-section area
$ ID TYPE PID FID PMIN PMAX EQID
DVPREL2 1 PBAR 9 4 0.01 100
DESVAR 1 2
DEQATN 100 A(H,B)=H*B
$
$ Explict expression for first moment intertia
DVPREL2 2 PBAR 9 5 0.01 200
DESVAR 1 2
DEQATN 200 A(H,B)=B*H**3/12.
$
$ Explict expression for stress data recovery point D (0.5B, 0.5H)
$ ID TYPE PID FID PMIN PMAX C0
DVPREL1 11 PBAR 9 14 0.01
1 0.5
DVPREL1 12 PBAR 9 15 0.01
2 0.5

8-3
 COMPARISON BETWEEN DESIGNING PBAR AND
PBARL (Cont.)

*************************************
* Input Data for Designing PBARL *
*************************************
$
PBARL, 9, 1, , BAR
1.0 5.0
$ ID LABEL INIT MIN MAX
DESVAR 1 HEIGHT5.0 0.01 10.
DESVAR 2 WIDTH 1.0 0.01 10.
$
$ Dimension Height references to design variable 1
$ ID TYPE PID FID PMIN PMAX C0
DVPREL1 11 BARL 9 13 0.01
1 1
$
$ Dimension Width references to design variable 2
DVPREL1 12 PBARL 9 12 0.01
1 2

8-4
 DVPREL1 INCLUDED ON DRESP2

Design Sensitivity Equation Response Quantities

• Defines equation responses that are used in the design, either as

constraints or as an objective.

Format:

1 2 3 4 5 6 7 8 9 10
DRESP2 ID LABEL EQID REGION
“DESVAR” DVID1 DVID2 DVID3 DVID4 DVID5 DVID6 DVID7

DVID8 -etc.-
“DTABLE” LABL1 LABL2 LABL3 LABL4 LABL5 LABL6 LABL7

LABL8 -etc.-
“DRESP1” NR1 NR2 NR3 NR4 NR5 NR6 NR7

NR8 -etc.-
“DNODE” G1 C1 G2 C2 G3 C3

G4 C4 -etc.-
“DVPREL1” DPIP1 DPIP2 DPIP3 DPIP4 DPIP5 DPIP6 DPIP7

DPIP8 DPIP9 -etc.-

NEW!

8-5
 DESIGN VARIABLE LINKING

The DLINK entry is the only way to implement design variable linking,
when DVPREL2’s are involved.

Consider a uniformly tapering beam with rectangular cross sections.

Uniform taper over ten sections will be determined from constant,
linear, and quadratic basis vector superposition. The section height
will be the design variable.

2
2 A = b ⋅ h = 5.0E-5 m
3
bh 4
1 h I 1 = ---------- = 1.042E-10 m
12
3
hb 4
b I 2 = ---------- = 4.167E-10 m
12

$PBAR, PID, MID, A, I1, I2

PBAR, 101, 100, 5.E-5, 1.042-10,4.167-10
PBAR, 102, 100, 5.E-5, 1.042-10,4.167-10
PBAR, 103, 100, 5.E-5, 1.042-10,4.167-10
PBAR, 104, 100, 5.E-5, 1.042-10,4.167-10
PBAR, 105, 100, 5.E-5, 1.042-10,4.167-10
PBAR, 106, 100, 5.E-5, 1.042-10,4.167-10
PBAR, 107, 100, 5.E-5, 1.042-10,4.167-10
PBAR, 108, 100, 5.E-5, 1.042-10,4.167-10
PBAR, 109, 100, 5.E-5, 1.042-10,4.167-10
PBAR, 110, 100, 5.E-5, 1.042-10,4.167-10
$
$...INDEPENDENT DESIGN VARIABLE SET:
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
DESVAR, 1, X1, 1.0, -1.0, +1.0
DESVAR, 2, X2, 0.0, -1.0, +1.0
DESVAR, 3, X3, 0.0, -1.0, +1.0
$
$...DEPENDENT DESIGN VARIABLE SET:
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
DESVAR, 11, H1, .01, 1.E-4, 1.0
DESVAR, 12, H2, .01, 1.E-4, 1.0
DESVAR, 13, H3, .01, 1.E-4, 1.0

8-6
 DESIGN VARIABLE LINKING (Cont.)

DESVAR, 14, H4, .01, 1.E-4, 1.0

DESVAR, 15, H5, .01, 1.E-4, 1.0
DESVAR, 16, H6, .01, 1.E-4, 1.0
DESVAR, 17, H7, .01, 1.E-4, 1.0
DESVAR, 18, H8, .01, 1.E-4, 1.0
DESVAR, 19, H9, .01, 1.E-4, 1.0
DESVAR, 20, H10, .01, 1.E-4, 1.0
$
$...DESIGN VARIABLE LINKING, Hj = Hj(X1,X2,X3) ; j=11,20:
$DLINK, ID, DDVID, CO, CMULT, IDV1, C1, IDV2, C2, +
$+, IDV3, C3, ...
DLINK, 21, 11, , .01, 1, 1.0, 2, 1.0, +
+, 3, 1.0
DLINK, 22, 12, , .01, 1, 1.0, 2, 0.9, +
+, 3, 0.81
DLINK, 23, 13, , .01, 1, 1.0, 2, 0.8, +
+, 3, 0.64
DLINK, 24, 14, , .01, 1, 1.0, 2, 0.7, +
+, 3, 0.49
DLINK, 25, 15, , .01, 1, 1.0, 2, 0.6, +
+, 3, 0.36
DLINK, 26, 16, , .01, 1, 1.0, 2, 0.5, +
+, 3, 0.25
DLINK, 27, 17, , .01, 1, 1.0, 2, 0.4, +
+, 3, 0.16
DLINK, 28, 18, , .01, 1, 1.0, 2, 0.3, +
+, 3, 0.09
DLINK, 29, 19, , .01, 1, 1.0, 2, 0.2, +
+, 3, 0.04
DLINK, 30, 20, , .01, 1, 1.0, 2, 0.1, +
+, 3, 0.01
$
$...DESIGN EQUATION:
DEQATN 40 I2(H) = H*(0.005)**3/12.
$
$...TYPE-2 DESIGN VARIABLE TO PROPERTY RELATIONS:
$DVPREL2,ID, TYPE, PID, FID, PMIN, PMAX, EQID, , +
$+, DESVAR, DVID1, DVID2, ..., , , , , +
$+, DTABLE, CID1, CID2, ...
DVPREL2,41, PBAR, 101, 6, 1.0E-12,, 40, , +
+, DESVAR, 11
DVPREL2,42, PBAR, 102, 6, 1.0E-12,, 40, , +
+, DESVAR, 12
DVPREL2,43, PBAR, 103, 6, 1.0E-12,, 40, , +
+, DESVAR, 13
DVPREL2,44, PBAR, 104, 6, 1.0E-12,, 40, , +
+, DESVAR, 14
DVPREL2,45, PBAR, 105, 6, 1.0E-12,, 40, , +
+, DESVAR, 15

8-7
 DESIGN VARIABLE LINKING (Cont.)

DVPREL2,46, PBAR, 106, 6, 1.0E-12,, 40, , +

+, DESVAR, 16
DVPREL2,47, PBAR, 107, 6, 1.0E-12,, 40, , +
+, DESVAR, 17
DVPREL2,48, PBAR, 108, 6, 1.0E-12,, 40, , +
+, DESVAR, 18
DVPREL2,49, PBAR, 109, 6, 1.0E-12,, 40, , +
+, DESVAR, 19
DVPREL2,50, PBAR, 110, 6, 1.0E-12,, 40, , +
+, DESVAR, 20

8-8
 CONSTRAINT EQUATION SCALING

• Each constraint should be formulated so that it has sufficient

sensitivity with respect to the design variables

• Do not use 0.0 for the lower or upper bound, if possible

• Examples:

2 2
π E 2 2 8L
σ c + ---------- ( D + T ) ≥ 0 → ------------------------------------ σ c ≥ – 1.0
2 2 2 2
8L π E(D + T )

Synthetic Response

0.4ET D
σ c + ---------------- ≥ 0 → ---------------- σ c ≥ – 1.0
D 0.4ET

Synthetic Response

8-9
 CONSTRAINT EQUATION SCALING (Cont.)

Example: Synthetic Responses

σ1 + σ2
------------------- ≤ σ max
2

• Assume that the constraint was formulated as:

σ1 + σ2
------------------- – σ max ≤ 0
2

This zero upper bound is automatically replaced by a small number

ε to avoid division by zero. (ε = GSCAL, and can be defined on the
DOPTPRM entry. Its Version 68 default is 0.001.)

σ1 + σ2
 ------------------- – σ max – ε
 2 
g u = ------------------------------------------------------ ≤ 0
ε

∂g u 1
---------- = ------
∂σ 1 2ε
• Then

g
1
------
2ε
CTMIN
σ1
σ1
CT

Active

8-10
 CONSTRAINT EQUATION SCALING (Cont.)

• Instead, assume constraint is written as:

σ1 + σ2
------------------- ≤ σ max
2

• Then

σ1 + σ2
 -------------------  –σ
 2  max
g u = ---------------------------------------------- ≤ 0
σ max

∂g u 1
---------- = -----------------
∂σ 1 σ max

1
---------------------
2 σ max
CTMIN
σ1
σ1
CT

Active

8-11
 CONSTRAINT EQUATION SCALING (Cont.)

• Response could also be written as:

σ1 + σ2
------------------- ≤ 1
2σ max

producing the same results.

Note: Scaling of responses is essential in order to formulate a

well-conditioned optimization task. MSC/NASTRAN can do
some of the scaling for you, but not all of it. There is no
substitute for a well-posed problem!

8-12
 BEAM DESIGN – GENERAL ISSUES

BEAM elements differ from other element types in that properties

must be referenced according to their word positions in the element
property tables. This is necessary due to the generality of the BEAM
element and the fact that up to eight intermediate stations may be
defined for a single element.

In general, the cross section of a BEAM element may be either

constant, linear, or variable. The three possibilities are shown in the
following EPT configurations:

1. Constant section beam (only end A specified): Words 6-21 and

166-181 must be accounted for on DVPRELi entries.

End A End B
Words Words
6-21 166-181

8-13
 BEAM DESIGN – GENERAL ISSUES (Cont.)

2. Tapered beam (end A and end B specified): Words 6-21, 22-37,

and 166-181 must be accounted for on DVPRELi entries.

End A Intermediate End B

Words Words* Words
6-21 22-37 166-181

*This station is internally generated but must be

accounted for on DVPRELi entries. Words 22-37
are identical to Words 166-181 (End B).

8-14
 BEAM DESIGN – GENERAL ISSUES (Cont.)

3. Variable section beam (end A, intermediate station C and end B

specified). (This case is typical of any number of intermediate
stations up to and including 8.) Words 6-21, 22-37, 38-53, and
166-181 must be accounted for on DVPRELi entries. Thus, every
intermediate station plus one internally generated intermediate
station must be accounted for on the set of DVPRELi entries.

End A Station C Intermediate End B

Words Words Words* Words
6-21 22-37 38-53 166-181

*This station is internally generated but must be

accounted for on DVPRELi entries. Words 38-53
are identical to Words 166-181 (End B).

8-15
 BEAM DESIGN – SIMPLE EXAMPLE

Consider the uniformly tapered rectangular beam section:

b1

h1

h2

B
b2

With initial design:

b
 1  1.0 
h   
 1  2.0 
{x} =   =  
 b2   0.5 
   
 h2   1.0 

8-16
 BEAM DESIGN – SIMPLE EXAMPLE (Cont.)

The input can be expressed as:

$>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
$ BEAM MODELING TEST, VERSION 68
$
$ TAPERED BEAM WITH RECTANGULAR SECTION, DESIGN SENSITIVITY ANALYSIS. SINCE
$ BOTH ENDS A AND B ARE SPECIFIED ON THE PBEAM ENTRY, DESIGN CHANGES MUST
$ BE SPECIFIED NOT ONLY FOR THE ENDS, BUT ALSO FOR THE FIRST INTERMEDIATE
$ STATION, WHICH CONTAINS A COPY OF END B DATA.
$
$
$ PLANE 2
$ |
$ ----*----
$ | |
$ H | | --- PLANE 1
$ | |
$ ----*----
$ B * = STRESS RECOVERY LOCATIONS
$
$
$>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
$
TIME 10
SOL 200
CEND
$
$ CC FOR ANALYSIS:
DISP = ALL
STRESS = ALL
SPC = 100
LOAD = 300
$ CC FOR OPTIMIZATION:
ANALYSIS = STATICS
DESOBJ(MIN) = 8 $ OBJECTIVE FUNCTION DEFINITION
DESSUB = 10 $ CONSTRAINT SET SELECTION
$
BEGIN BULK
$PARAM, OPTIM, NO
PARAM, OPTEXIT, 4
$-------------------------------------------------------------------------------
$ ANALYSIS MODEL
$-------------------------------------------------------------------------------
MAT1 110 10.0E6 0.33 0.1 +M1
+M1 50000. 50000. 29000.
GRDSET 4
GRID 1 0.0 0.0 0.0
GRID 2 20. 0.0 0.0
CBEAM 1 100 1 2 0.0 1.0 0.0
$
$ PBEAM ENTRY INPUT WITH SLIGHT ‘ERROR’ IN TERMS. THIS HELPS VALIDATE THE DE-
$ SIGN MODEL BECAUSE USER WARNING MESSAGE WILL BE ISSUED, CONFIRMING OVERRIDE.

8-17
 BEAM DESIGN – SIMPLE EXAMPLE (Cont.)

$
PBEAM 100 110 2.01 .167 .667 +P11
+P11 0.0 1.01 0.0 -1.01 +P12
+P12 YES 1.0 0.51 .0104 .042 +P13
+P13 0.0 0.51 0.0 -0.51
$
SPC1 100 123456 1
FORCE 300 2 20000.0 0.0 0.0 -1.0
$-------------------------------------------------------------------------------
$ DESIGN MODEL
$-------------------------------------------------------------------------------
$
$...END A DATA: (A,I1,I2,C2,D2)
$
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
DESVAR, 1, B1, 1.0, 0.1, 10.0
DESVAR, 2, H1, 2.0, 0.2, 20.0
$
$DVPREL2,ID, TYPE, PID, FID, PMIN, PMAX, EQID, , +
$+, DESVAR, DVID1, DVID2, ..., , , , , +
$+, DTABLE, CID1, CID2, ...
DVPREL2,1, PBEAM, 100, -8, , , 101, , +
+, DESVAR, 1, 2
DVPREL2,2, PBEAM, 100, -9, , , 102, , +
+, DESVAR, 1, 2
DVPREL2,3, PBEAM, 100, -10, , , 103, , +
+, DESVAR, 1, 2
DVPREL2,4, PBEAM, 100, -15, , , 104, , +
+, DESVAR, 2
DVPREL2,5, PBEAM, 100, -17, , , 105, , +
+, DESVAR, 2
$
$...END B DATA: (A,I1,I2,C2,D2)
$
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
DESVAR, 3, B2, 0.5, 0.05, 10.0
DESVAR, 4, H2, 1.0, 0.1, 20.0
$
$DVPREL2,ID, TYPE, PID, FID, PMIN, PMAX, EQID, , +
$+, DESVAR, DVID1, DVID2, ..., , , , , +
$+, DTABLE, CID1, CID2, ...
DVPREL2,6, PBEAM, 100, -168, , , 101, , +
+, DESVAR, 3, 4
DVPREL2,7, PBEAM, 100, -169, , , 102, , +
+, DESVAR, 3, 4
DVPREL2,8, PBEAM, 100, -170, , , 103, , +
+, DESVAR, 3, 4
DVPREL2,9, PBEAM, 100, -175, , , 104, , +
+, DESVAR, 4
DVPREL2,10, PBEAM, 100, -177, , , 105, , +
+, DESVAR, 4
$
$...FIRST INTERMEDIATE STATION (COPY OF END B DATA): (A,I1,I2,C2,D2)
$

8-18
 BEAM DESIGN – SIMPLE EXAMPLE (Cont.)

$DVPREL2,ID, TYPE, PID, FID, PMIN, PMAX, EQID, , +

$+, DESVAR, DVID1, DVID2, ..., , , , , +
$+, DTABLE, CID1, CID2, ...
DVPREL2,11, PBEAM, 100, -24, , , 101, , +
+, DESVAR, 3, 4
DVPREL2,12, PBEAM, 100, -25, , , 102, , +
+, DESVAR, 3, 4
DVPREL2,13, PBEAM, 100, -26, , , 103, , +
+, DESVAR, 3, 4
DVPREL1,14, PBEAM, 100, -31, , , , +
+, 4, 0.5
DVPREL1,15, PBEAM, 100, -33, , , 105, , +
+, 4, -.05
$
DEQATN 101 AREA(B,H) = B*H
DEQATN 102 I1(B,H) = H*B**3/12.
DEQATN 103 I2(B,H) = B*H**3/12.
DEQATN 104 C2(H) = H/2.
DEQATN 105 D2(H) = -H/2.
$
$DRESP1,ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
DRESP1, 7, D2, DISP, , , 3, , 2
DRESP1, 8, W, WEIGHT
$
$DCONSTR,DCID, RID, LALLOW, UALLOW
DCONSTR, 10, 7, -3.0, 3.0
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......0
ENDDATA

8-19
 BEAM DESIGN – SIMPLE EXAMPLE (Cont.)

As proof that the override took place, one can refer to the comparison table between analysis and
design models:
----- COMPARISON BETWEEN INPUT PROPERTY VALUES FROM ANALYSIS AND DESIGN MODELS -----

---------------------------------------------------------------------------------------------------------------
PROPERTY PROPERTY FIELD ANALYSIS DESIGN LOWER UPPER DIFFERENCE
TYPE ID ID VALUE VALUE BOUND BOUND FLAG
---------------------------------------------------------------------------------------------------------------
PBEAM 100 -177 -5.100000E-01 -5.000000E-01 -1.000000E+35 1.000000E+20 WARNING
PBEAM 100 -175 5.100000E-01 5.000000E-01 -1.000000E+35 1.000000E+20 WARNING
PBEAM 100 -170 4.200000E-02 4.166667E-02 1.000000E-03 1.000000E+20 WARNING
PBEAM 100 -169 1.040000E-02 1.041667E-02 1.000000E-03 1.000000E+20 WARNING
PBEAM 100 -168 5.100000E-01 5.000000E-01 1.000000E-03 1.000000E+20 WARNING
8-20

PBEAM 100 -33 -5.100000E-01 -5.000000E-01 -1.000000E+35 1.000000E+20 WARNING

PBEAM 100 -31 5.100000E-01 5.000000E-01 -1.000000E+35 1.000000E+20 WARNING
PBEAM 100 -26 4.200000E-02 4.166667E-02 1.000000E-03 1.000000E+20 WARNING
PBEAM 100 -25 1.040000E-02 1.041667E-02 1.000000E-03 1.000000E+20 WARNING
PBEAM 100 -24 5.100000E-01 5.000000E-01 1.000000E-03 1.000000E+20 WARNING
PBEAM 100 -17 -1.010000E+00 -1.000000E+00 -1.000000E+35 1.000000E+20 WARNING
PBEAM 100 -15 1.010000E+00 1.000000E+00 -1.000000E+35 1.000000E+20 WARNING
PBEAM 100 -10 6.670000E-01 6.666667E-01 1.000000E-03 1.000000E+20 WARNING
PBEAM 100 -9 1.670000E-01 1.666667E-01 1.000000E-03 1.000000E+20 WARNING
PBEAM 100 -8 2.010000E+00 2.000000E+00 1.000000E-03 1.000000E+20 WARNING

1. IF FIELD ID IS LESS THAN ZERO, IT IDENTIFIES THE WORD POSITION OF AN ENTRY IN EPT.

2. IF FIELD ID IS GREATER THAN ZERO, IT IDENTIFIES THE FIELD POSITION ON A PROPERTY BULK DATA ENTRY.

3. THE DIFFERENCE FLAG IS USED TO CHARACTERIZE DIFFERENCES BETWEEN ANALYSIS AND DESIGN MODEL PROPERTIES:
IF THE FLAG IS NONE, THEN THERE IS NO SIGNIFICANT DIFFERENCE BETWEEN THE TWO VALUES.
IF THE FLAG IS WARNING, THEN THE USER IS ADVISED THAT DIFFERENCES EXIST.
IF THE FLAG IS FATAL, THEN THE DIFFERENCES ARE GREATER THAN 1.00000E+35 AND THE RUN WILL BE TERMINATED.
 SECTION 9

FUNDAMENTALS OF STRUCTURAL OPTIMIZATION

STRUCTURAL OPTIMIZATION – THE DIFFICULTIES . . . . . . . . . . . . . . . . . . . . . . . 9-1

STRUCTURAL OPTIMIZATION – THE SOLUTION. . . . . . . . . . . . . . . . . . . . . . . . . . 9-3

DESIGN VARIABLE LINKING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-7

DESIGN VARIABLE LINKING IN SHAPE REDESIGN . . . . . . . . . . . . . . . . . . . . . . 9-11

GENERATION OF CONSTRAINTS IN MSC/NASTRAN . . . . . . . . . . . . . . . . . . . . . 9-15

CONSTRAINT SCREENING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-16

DSCREEN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-18

FORMAL APPROXIMATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-21

APRCOD – SELECTION OF APPROXIMATE FORM . . . . . . . . . . . . . . . . . . . . . . . 9-28

SENSITIVITY ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-30

SEMIANALYTIC SENSITIVITY ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-31

DELB – FINITE DIFFERENCE STEP SIZE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-33

SENSITIVITY ANALYSIS – IMPLEMENTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-35

ADJOINT SENSITIVITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-37

ADJOINT SENSITIVITY THEORY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-38

DESIGN SENSITIVITY COEFFICIENT MATRICES . . . . . . . . . . . . . . . . . . . . . . . . . 9-47

DSAPRT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-49

FORMATTED SENSITIVITY PRINT EXAMPLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-51

MOVE LIMITS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-52

 SECTION 9

FUNDAMENTALS OF STRUCTURAL
OPTIMIZATION (Cont.)

MOVE LIMITS IMPOSED ON ANALYSIS MODEL PROPERTIES . . . . . . . . . . . . . 9-53

MOVE LIMITS IMPOSED ON DESIGN VARIABLES . . . . . . . . . . . . . . . . . . . . . . . . 9-55

AUTOMATIC UPDATES OF MOVE LIMITS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-56

MOVE LIMITS UPDATE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-57

PARAM OPTEXIT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-58

FLOW CHART FOR SOLUTION 200 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-59

CONVERGENCE AT THE DESIGN CYCLE LEVEL . . . . . . . . . . . . . . . . . . . . . . . . 9-63

CONVERGENCE CRITERIA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-65

SUMMARY OF PARAMETERS FOR DESIGN OPTIMIZATION . . . . . . . . . . . . . . . 9-67

 STRUCTURAL OPTIMIZATION – THE DIFFICULTIES 9

• Large numbers of design variables

• Large numbers of constraints, i.e., i-th component of stress for j-th

element of k-th load case σijk (with both upper and lower bounds)

• The structural responses which guide the design are implicit

functions of the design variables. Function evaluations are of
significant cost, and gradient information is necessary to
determine the search direction s.

• Problem formulation places a unique set of demands on design

engineers:

• Formulation of design objective

• Appropriate specification of constraints

• Analysis model integrity

9-1
 STRUCTURAL OPTIMIZATION –
THE DIFFICULTIES (Cont.)

• A “Brute Force” coupling of an analysis program and an

optimization program suffers from:

Too many design variables

Finite
Element Numerical
Too many design constraints Optimization
Analysis
Program Program
Too many detailed analyses

“Too many” implies redundant design information.

9-2
 STRUCTURAL OPTIMIZATION – THE SOLUTION

Approximation concepts are a formal implementation of methods

used by experienced design engineers.

Experienced Approximation
Engineers Concepts

Pick up the most Variable transformation

Design
important variables Apx1 = TpxN XNx1 p>>N 1
Variables
only 1 to 3 at one time.
Temporary deletion of
Pick up the most
nonactive constraints and
Constraints important constraints
dynamic updates of
only.
retained constraints.

Small number of Explicit approximations of

2
detailed structural a small number of the
Analysis analyses and tests retained constraints.
Results
Experience and Taylor Series
intuition Regression Analysis

Notes: 1. The challenge of the design engineer is to describe the

design problem as efficiently as possible, e.g., with the
smallest set of design variables.

2. Constraint screening and formal approximations are per-

formed automatically by the code.

9-3
 STRUCTURAL OPTIMIZATION –
THE SOLUTION (Cont.)

• Program flow using approximation techniques

• Perform detailed finite element analysis

• Calculate all constraints. Delete those constraints that are not

potentially critical (a common sense approach)

• Calculate the gradients of all “retained” constraints

• Create a “high-quality” approximation of the responses with

respect to the design variables

• Solve this “approximate” problem

• Update the analysis variables

• Perform a detailed analysis of the proposed design

• Calculate all constraints

• Check for convergence to an optimum

• Repeat if necessary

9-4
 STRUCTURAL OPTIMIZATION –
THE SOLUTION (Cont.)

• Approximate Design Model

Improved Designs

Finite
Numerical
Element
Optimization
Analysis
Program
Program

Approximate
Design
Model

• The Approximate Model now acts as the interface between the

analysis and the optimizer

• Properties of the Approximate Model

• It contains inexpensive and easily computed, yet high-quality,

explicit approximations of the objective function and all retained
constraints

• It includes the results of constraint deletion

• It is updated by a full finite element analysis every design cycle

9-5
 STRUCTURAL OPTIMIZATION –
THE SOLUTION (Cont.)

Program Structure with Approximate Model

Initial
Design
Improved
Design

The required number of

iterations of the external loop
must be small.
Structural
Response
Analysis

Constraint Approximate
Optimizer
Screening Model

Sensitivity Many Times

Analysis

Finite Element
Analysis

9-6
 DESIGN VARIABLE LINKING

Design variable linking is a choice that is up to the design engineer.

• Selecting design variables wisely such that the design task is

efficiently stated

• Coupling this with an expression of the practical aspects of design,

such as symmetry, manufacturability, etc.

Example - Structural Member Size Linking

• Design model description

• Design variables: Cross-sectional areas

• Objective: Weight minimization

• Constraints: Axial stress

360" 360"
5 3 1
1 2
7 10

5 6
360"
8 9

3 4
6 4 2

100,000 lbs 100,000 lbs

9-7
 DESIGN VARIABLE LINKING (Cont.)

• Areas are grouped as (1,2,3,4), (5,6), (7,8,9,10) and each group is

controlled by one independent variable

 A1 
  1.0 0.0 0.0
 A2 
  1.0 0.0 0.0
A 
 3 1.0 0.0 0.0
A 
 4 1.0 0.0 0.0
A   x1 
 5 0.0 1.0 0.0  
  =  x2 
 A6  0.0 1.0 0.0  
   x3 
 A7  0.0 1.0 1.0
 
 A8  0.0 0.0 1.0
 
A  0.0 0.0 1.0
 9
A  0.0 0.0 1.0
 10 

where Ai = cross-sectional area of the i-th member

xj = independent design variable

x1 = x2 = x3 = 10.0 initially

In this example, all designs considered by the optimizer will

consist of three characteristic cross-sectional areas, x1, x2, and x3
only.

9-8
 DESIGN VARIABLE LINKING (Cont.)

Example – Reduced Basis Approximations

360" 360"
5 3 1
1 2
7 10

5 6
360"
8 9

3 4
6 4 2

100,000 lbs 100,000 lbs

• Three designs are proposed:

1 1 1 1
• Proposal 1: A 1, A 2, A 3, …, A 10

2 2 2 2
• Proposal 2: A 1, A 2, A 3, …, A 10

3 3 3 3
• Proposal 3: A 1, A 2, A 3, …, A 10

9-9
 DESIGN VARIABLE LINKING (Cont.)

 
 A1  A1
1
A1
2
A1
3
 
A  1 2 3
 2 A2 A2 A2
 
 A3  1 2 3
  A3 A3 A3
 
 A4  A4
1
A4
2
A4
3
  x 
A  1 2 3  1
 5 A5 A5 A5  
  =  x2 
A  A
1
A6
2
A6
4  
 6  6  
  1 2 3  x3 
 A7  A7 A7 A7
 
  1 2 4
 A8  A8 A8 A8
 
A  A9
1
A9
2
A9
5
 9
  1 2 6
A  A 10 A 10 A 10
 10 

where Ai = cross-sectional area of the i-th member

The reduced basis method used here will yield a design consisting of
ten unique cross-sectional areas. Contrast this to the previous
example where linking of the properties (member size linking) yielded
3 characteristic cross-sectional areas. Though neither approach is
generally capable of yielding the global optimum, the techniques
often provide a convenient way of expressing manufacturability, or
other process constraints.

9-10
 DESIGN VARIABLE LINKING IN SHAPE REDESIGN

Example – Design Variable Linking to Enforce Symmetry

• Grid coordinates as design variables

x1 x2

1 2

3 4
h1
h2

5 6

• Linking to maintain symmetry

x1 = –x2
x5 = –x6

• Linking to keep the side as a straight line

1
x 4 = – ------ [ x 6 ( h 1 – h 2 ) + x 2 h 2 ]
h1
x3 = –x4

• Independent and dependent design variable sets are:

T
{ x }I = [ x2 x6 ]
T
{ x }D = [ x1 x3 x4 x5 ]

9-11
 DESIGN VARIABLE LINKING IN SHAPE
REDESIGN (Cont.)

Example - Design Variable Linking to Ensure Smooth Shapes

• 2-D and 3-D designs benefit from a parameterization of the

boundaries

In the figure below, independent motion at the grid points (non-

parameterized description) was allowed.

Optimizer will skew

elements to yield
“numerical stiffness.”

9-12
 DESIGN VARIABLE LINKING IN SHAPE
REDESIGN (Cont.)

Example – Parameterization of Boundaries

Circular Arc Straight Line

1 Grid Point 2 Grid Points
1 Parameter (R)

Cubic Polynomial
3 Grid Points
1 Slope

9-13
DESIGN VARIABLE LINKING IN SHAPE
REDESIGN (Cont.)

F ( ξ ) = x 1 F 1 ( ξ ) + x 2 F 2 ( ξ ) + x 3 F 3 ( ξ )…

F1 ( ξ )

F1 ( ξ )

F2 ( ξ )

F3 ( ξ )

.
.
.
Basis Functions

9-14
 GENERATION OF CONSTRAINTS IN MSC/NASTRAN

• For some given response:

rj ( x ) (DRESP1, 2)

specify upper and lower allowables:

rjL and rjU (DCONSTR)

• Designer’s intent is only to allow:

r jL ≤ r j ( x ) ≤ r jU

• MSC/NASTRAN constructs a pair of normalized constraints as

follows:

r jL – r j ( x )
g 2j – 1 ( x ) = ------------------------- ≤ 0
r jL
(automatic)
r j ( x ) – r jU
g 2j ( x ) = -------------------------- ≤ 0
r jU

• Constraint screening is applied to these normalized constraints

9-15
 CONSTRAINT SCREENING

• Purpose:

The number of constraints for structural design problems could be

extremely large, but most of them do not directly influence the
design process. Since the design sensitivity analysis becomes
very expensive for large numbers of constraints, and the amount of
data transferred to approximate optimization is proportional to the
number of constraints in that phase, it is necessary to delete all the
unnecessary constraints before moving into the sensitivity
analysis.

• Constraint deletion consists of two steps:

Step I - Deletion:

Based on constraint value. A constraint G(X) is deleted temporarily

if G(X) < TRS, where TRS is the truncation threshold value given by
a DSCREEN entry for each constraint type.

Gj(X)

x x x x x x x x x

TRS

–1.0 j
List of Constraints

9-16
 CONSTRAINT SCREENING (Cont.)

Step II - Regionalization:

Based on response regionalization. Both DRESP1 and DRESP2

entries have a field named REGION. All responses with the same
region ID and of the same type are recognized as belonging to the
same region. Even if there are many constraints above the
threshold values, only a small number (given by NSTR on the
DSCREEN entry) are retained per region and per subcase.

Assume NSTR = 2
G(X)

x x x x x x x x x

TRS

–1.0

Region 1 Region2 Region 3

1 0 2
Retained
Responses

9-17
DSCREEN Design Constraint Screening Data

DSCREENscreening data for constraint deletion.

Defines Design Constraint Screening Data

Format:

1 2 3 4 5 6 7 8 9 10
DSCREEN RTYPE TRS NSTR

Example:

DSCREEN STRESS –0.7 2

Field Contents
RTYPE Response type for which the screening criteria apply. See Table 1. (Character)

TRS Truncation threshold. (Real; Default = –0.5)

NSTR Maximum number of constraints to be retained per region per load case. See
Remark 3. (Integer > 0; Default = 20)

Remarks:
1. Displacement and displacement derivative constraints associated with one particular load
case are grouped by the specification of DRESP1 entries. From each group, a maximum
of NSTR constraints are retained per load case.

Table 1. Default Region Specifications for RTYPE.

Default Region Default Region

RTYPE Specification RTYPE Specification

DISP DRESP1 FRVELO DRESP1

STRESS Property ID FRACCL DRESP1
STRAIN Property ID FRSTRE Property ID
FORCE Property ID FRFORC Property ID
EQUA DRESP2 FRSPC DRESP1
CSTRESS Property ID TACCL DRESP1
CSTRAIN Property ID TDISP DRESP1
CFAILURE Property ID TVELO DRESP1
FLUTTER DRESP1 TSTRE Property ID
FRDISP DRESP1 TFORC Property ID
TSPC DRESP1

(Continued)

9-18
Design Constraint Screening Data DSCREEN
2. Stress-strain constraints are grouped by the property; i.e., all elements belonging to the
set of PIDs specified under ATTi on a DRESPi entry are regarded as belonging to the
same region. In superelement sensitivity analysis, if the property (PID) is defined in more
than one superelement, then separate regions are defined. A particular stress constraint
specification may be applied to many elements in a region generating many stress con-
straints, but only up to NSTR constraints per load case will be retained.

3. If RTYPE = “WEIGHT”, “VOLUME”, “EIGN”, “FREQ”, “LAMA”, “TRIM”, or “STABDER”

then NSTR is not applicable.

4. If a certain type of constraint exists but no corresponding DSCREEN entry is specified, all
the screening criteria used for this type of constraint will be furnished by the default
values.

5. Constraints can be retained only if they are greater than TRS. See the Remarks under
the DCONSTR entry for a definition of constraint value.

6. Constraint screening is applied to each superelement.

9-19
 CONSTRAINT SCREENING (Cont.)

• Recommendations

• Use the default values by not providing any DSCREEN entries. If

the constraints prove to be ill-conditioned, then modify screening
criteria according to constraint type

• Symptoms of ill-conditioned constraints

• Constraints that the optimizer has difficulty satisfying

• Constraints that switch from the active and inactive sets from one
design cycle to the next

• Constraints whose approximate values differ greatly (~20% or

more) from those obtained by a subsequent finite element analysis

9-20
 FORMAL APPROXIMATIONS

• These form a key ingredient in the approximate model in that it is

necessary to obtain high-quality and explicit approximations of the
implicit structural responses computed by the finite element
analyses

• High-quality – Maintain reasonable accuracy in predicting

responses for large variations of the independent design variables

• Reasonable – Direct the design toward a practical optimum design

• Large – At least 10 to 20% changes in the design variables

9-21
 FORMAL APPROXIMATIONS (Cont.)

• Taylor series expansion

2
∆x
f ( x + ∆x ) = f ( x ) + f′ ( x )∆x + f″ ( x ) ---------- + …
2!

• First-order, approximations, directly in terms of the design

variables:

n
∂f
∑  ------- (x – x )
o o
f(x) ≅ f(x ) +
∂x i x o i i
i=1

9-22
 FORMAL APPROXIMATIONS (Cont.)

• First order, approximations in terms of reciprocals of the design

variables:

Yi = Yi ( xi )

The Yi are termed intermediate variables

n
∂f
∑  -------
- ( Y ( x ) – Y i ( x i ) )
o o
f(x) ≅ f(x ) +
∂Y i x o i i
i=1

1
now, let Y i = ----
xi

∂Y i 1
then -------- = – ------
∂x i xi
2

n
∂f  o 2 1 1
∑  -------
o
and fR ( x ) = f ( x ) – ( x )  ---- – ------
 ∂x  x o i  x o
i i x
i=1 i

n o
∂f  x i o
∑  -------
o
= f(x ) – ------ ( x – x i )
 ∂x  x o x i
i i
i=1

9-23
 FORMAL APPROXIMATIONS (Cont.)

Question: Why are reciprocal approximations useful?

Example: Consider the structural responses of a ROD element.

A
P

Stress = P/A

DISPL = PL/AE

• Both are proportional to 1/A

σ̃

A˚ A

σ̃ is a first order, linear approximation to σ at A˚

9-24
 FORMAL APPROXIMATIONS (Cont.)

• If the design quantity is A, let Y = 1/A

1/A˚ 1/A

• For general cases this is only approximate (e.g., statically

indeterminate problems); however, this inverse relation indicates
the value of reciprocal approximations in structural optimization

9-25
 FORMAL APPROXIMATIONS (Cont.)

• Recall the general nonlinear basic optimization problem statement

Minimize F(x)

subject to:

gj ( x ) ≤ 0 j = 1,m

l u
xi ≤ xi ≤ xi i = 1,n

• Using Formal Approximations to recast the above leads to:

n
∂F
∑  -------
- ( Y i ( x i ) – Y i ( x i ) )
o o
Minimize: F˜ ( x ) = F ( x ) +
∂Y i x o
i=1

subject to:

n
 ∂g j
∑
o o
g˜ j ( x ) = g j ( x ) +  -------- Y i ( x i ) – Y i ( x i ) j = 1,m
 ∂Y i x o
i=1

Move limits:

l u
xi ≤ xi ≤ xi i = 1,n

where Yi = Xi for direct approximation

Yi = 1
----- for reciprocal approximation
Xi

9-26
 FORMAL APPROXIMATIONS (Cont.)

Question: How is the choice of direct versus reciprocal

approximations made?

Answer: For the mixed method (APRCOD = 2), the choice is

based entirely on response type.

• Direct Variable Approximations

• Weight

• Volume

• Internal force

• Buckling load factor

• Reciprocal Variable Approximations

• Displacement

• Stress

• Strain

• Eigenvalue/Natural Frequency

• Composite failure criteria

• Selecting the Formal Approximation Method:

• The selection is made with APRCOD on the DOPTPRM Bulk Data

entry

9-27
 APRCOD – SELECTION OF APPROXIMATE FORM

• The selection of variables with respect to which approximations

are formed is dependent upon the following

If analysis model properties are linearly related to the design vari-

ables, use independent design variables

If analysis model properties are expressed using the user-supplied

equations, use analysis model properties

• APRCOD = 1: Direct linearization of responses

N
∂g
g̃ ( X ) = g ( X o ) + ∑ -------
∂x i
( x i – x io )
i=1 x = xo

• APRCOD = 2: Mixed method based on response types (default)

Reciprocal variables – stress, strain, displacement, composite cri-

teria, eigenvalue

Direct variables – weight, volume, internal force, buckling load

9-28
 APRCOD – SELECTION OF APPROXIMATE
FORM (Cont.)

• APRCOD = 3: Convex linearization – selection of either direct or

reciprocal approximations depending on which is more
conservative (that is, provides the largest constraint value
estimate)

Since both approximations are available, look at the sign of the

difference between the two:

if g D ( X ) – g R ( X ) ≥ 0, use g D ( X )

if g D ( X ) – g R ( X ) < 0, use g R ( X )

Thus:

 N 
g(X ) + ∂g ∂g
 ∑
------- ( x i – x io ) if x i ⋅ ------- ≥ 0 
o ∂x i ∂x i
 i=1 x = xo 
 x = xo 
g̃ ( X ) =  
 N 
∂g ∂g
∑ x io  ---- – --------
2 1 1
 g(X ) – ------- if x i ⋅ ------- < 0
 o ∂x x x  ∂x i 
 i=1 i x=x i io
x = xo 
 o 

9-29
 SENSITIVITY ANALYSIS

• Formal Approximations are based on the assumption that gradient

information is available. Sensitivity Analysis yields this
information.

• Static Displacement Sensitivity Analysis

KU = P

• Differentiate with respect to a single design variable:

∂K ∂U ∂P
------- U + K ------- = -------
∂x i ∂x i ∂x i

∂U
• Solve for -------
∂x i

∂U ∂P ∂K
K ------- = ------- – ------- ⋅ U
∂x i ∂x i ∂x i

The equation above is exact. (The right-hand side is often referred

to as a pseudo-load vector.)

This first-order displacement sensitivity information allows the ap-

proximations of all other static response gradients to be made:

dr j ∆r j r j ( X + ∆x j , U + ∆u ) – r j ( X, U )
-------- ≅ -------- = ---------------------------------------------------------------------------
-
dx i ∆x i ∆x i

∂U
where ∆U = ------- ⋅ ∆x i
∂x i

9-30
 SEMIANALYTIC SENSITIVITY ANALYSIS

∂P ∂K
• The difficulty in practice arises from the fact that ------- and ------- in the
∂x i ∂x i
pseudo-load vector are not generally available analytically. Thus,
we must introduce approximations for these derivatives. The
result is a semianalytical approximation for the displacement
sensitivity components

Semianalytic Sensitivity Information

• First forward finite differences:

e o
e K ( x + ∆x ) – K ( x ) e o
∂K i
---------- ≅ ------------------------------------------------------------
∂x i ∆x i

the error is on the order of ∆x with this approximation

• Central differences:

e o
e K ( x + ∆x ) – K ( x – ∆x ) e o
∂K i i
---------- ≅ ---------------------------------------------------------------------------
∂x i 2∆x i

2
the error is reduced to the order of ∆x . Central Differences are
the default for shape optimization and forward differences are the
default for property optimization. PARAM,CDIF,YES or NO can be
used to override the defaults.

9-31
 SEMIANALYTIC SENSITIVITY ANALYSIS (Cont.)

• Eigenvalue sensitivity:

• Eigenvalue problem statement

[ K – λ j M ]φ j = 0

• Differentiating with respect to a design variable xi:

T ∂K ∂M
φ j ------- – λ j ------- φ j
∂λ ∂x i ∂x i
------ = ---------------------------------------------
∂x i T
φ j Mφ j

• Buckling sensitivity

Kφ k + λ k K d φ k = 0

Differentiating with respect to a design variable xi:

T  ∂K ∂K d
– φ k  ------- + λ k ---------- φ k
∂λ k  ∂x i ∂x i 
--------- = ------------------------------------------------------
∂x i T
( φk Kd φk )

9-32
 DELB – FINITE DIFFERENCE STEP SIZE

In computing gradients, finite differences with respect to properties

are used in the DSA modules. For example:

• Element stiffness matrix:

∂k ∂p j
e e
∂k
--------- = --------- ⋅ --------
∂x i ∂p j ∂x i

using first forward finite differences,

e o o e o
∂k
e k ( p j + DELB ⋅ P j ) – k ( p j )  ∂p j
--------- ≅ ------------------------------------------------------------------------------ ⋅  --------
∂x i DELB ⋅ p j
o  ∂x i x o

a similar expression can be written for central differences.

for DVPREL1,

pj = co + c1 x1 + … + ci xi + …

∂p j
-------- = c i
∂x i

9-33
 DELB – FINITE DIFFERENCE STEP SIZE (Cont.)

for DVPREL2,

p j = f ( x 1, x 2, …, x n )

∂p j p j ( x 1, x 2, …, x i + ∆x i, …, x n ) – p j ( x 1, x 2, x i – ∆x i, …, x n )
-------- = -------------------------------------------------------------------------------------------------------------------------------------------------------
∂x i 2∆x i

• Displacement perturbation:

∂U j o
∆U j = -------- ⋅ DELB ⋅ p i
∂p i

• Stress gradient:

∂σ ∂σ ∂p j
------- = - ⋅ --------
-------
∂x i ∂p j ∂x i

o o o
σ ( u + ∆u, p j + DELB ⋅ p j ) – σ ( u, p j )  ∂p j
≅ --------------------------------------------------------------------------------------------------- ⋅  --------
DELB ⋅ p j
o  ∂x i  x o

• If the value of pj is too small, then we may not have sufficient

perturbation to compute finite differences. Thus, it is then
necessary to use DELBM.

o o
• If DELB ⋅ p j ≤ DELBM , then DELB ⋅ p j in the equations given
above will be replaced by DELBM = 1.0E–5.

9-34
 SENSITIVITY ANALYSIS – IMPLEMENTATION

∂K ∂M
• In MSC/NASTRAN sensitivity analysis ------- and ------- are never formed
∂x i ∂x i
as G-size matrices

∂K ∂M
• Instead, vectors ------- U and ------- U are formed as the sums of the
∂x i ∂x i
products of the element level matrices and displacement vectors

M e
∂K  ∂k m e
------- U =
∂x i ∑  ----------- ⋅ U 
 ∂x i 
m=1

M e
∂M  ∂m m e
------- U =
∂x i ∑  ------------- ⋅ U 
 ∂x i 
m=1

• Multiplication involves only the DOFs associated with the m-th

element

9-35
 SENSITIVITY ANALYSIS – IMPLEMENTATION (Cont.)

• The most expensive calculations are:

• Back and forward substitutions

• It is important to reduce the number of right-hand-side

vectors. This is proportional to the number of load
cases ∗ the number of design variables.

• Data recovery (response sensitivity)

• Reduce the number of retained constraints

• Reduce the number of design variables

9-36
 ADJOINT SENSITIVITY

Motivation

• Consider a modal frequency analysis with

• Four subcases

• Two hundred frequencies

• 100,000 DOF

• Solution vectors require 1.28 gigabytes

• Now do a sensitivity analysis with 100 design variables

• Disk space requirement goes up by at least 100

• An alternative algorithm is available if:

• Only grid responses are active

• Design model is in the residual

• nresp < ndv × (nsub × nfreq)

9-37
 ADJOINT SENSITIVITY THEORY

Standard frequency response equation

2
[ – ω M + iωB + K ] { u } = { P }

Grid response (r) is extracted from the global displacement vector (u)

r = f(u)

Sensitivity of the response with respect to design variables is

T
dr ⁄ dx = ∂f ⁄ ∂ u ∂u ⁄ ∂x

In the direct method, ∂u ⁄ ∂x is obtained from:

2 2
[ – ω M + iωB + K ] { du ⁄ dx } = – [ – ω dM ⁄ dx + iωdB ⁄ dx + dK ⁄ dx ] { u }

The adjoint method solves for a solution vector of the form:

2 T
[ – ω M + iωB + K ] { λ } = { ∂ f ⁄ ∂ u }

Defining

2
[ FAC ] = [ – ω M + iωB + K ]

9-38
 ADJOINT SENSITIVITY THEORY (Cont.)

Then

T 2
dr ⁄ dx = – { λ } [ – ω dM ⁄ dx + iωdB ⁄ dx + dK ⁄ dx ] { u }

where the relationship

–1
[ FAC ] [ FAC ] = [ I ]

has been used.

9-39
 ADJOINT SENSITIVITY

WHAT IS SUPPORTED WHAT IS NOT SUPPORTED

Statics Static Aeroelasticity
Frequency Response Transient Response
Gravity Loads Statics Analysis with Inertia Relief or
Thermal Loads
Grid Responses p-elements
DISP FRDISP
FRVELO FRACCL
Design Model in Residual Element Responses
STRESS STRAIN FORCE
CSTRESS CSTRAIN CFAILURE
FRSTRE FRFORC FRSPCF
Multiple BC’s for Statics in V70.5 Design Model in Upstream
Superlements

9-40
 ADJOINT SENSITIVITY (Cont.)

Example - Visual Sensor

Model

Design Task

Minimize sensor jitter

nfreq

∑
2
φ = u si
i=1

Constraints

Weight ≤ 8.5

f 1 and f 2 ≤ 100 Hz

9-41
 ADJOINT SENSITIVITY (Cont.)

Example - Visual Sensor:

Model Statistics
Number of grids: 2335
Number of elements: 2129
Number of subcases: 1
Number of excitation frequencies: 101
Number of design variables: 13
Number of responses: 91

Performance Results

Parameter V70 V69.1

CPU Time 1913.4 secs 4709.9
Scratch Space 90.5 MB 608.4 MB
SCR 300 Space 70.4 MB 356.6 MB
No. of Design Cycles 9 8
Final Objective 1.3862 1.559

9-42
 ADJOINT SENSITIVITY (Cont.)

Example - Visual Sensor

9-43
ADJOINT SENSITIVITY (Cont.)

Example - Visual Sensor

9-44
 ADJOINT SENSITIVITY (Cont.)

Example - Static Car Body

Model Statistics
Number of grids: 2615
Number of elements: 2647
Number of subcases: 2
Number of design variables: 1075
Number of responses: 2

Performance Results

Parameter V70 V69.1

CPU Time 48.1 secs 410.2
Scratch Space 24.4 MB 554.0 MB
Scratch 300 Space 7.7 MB 259.1 MB

9-45
 ADJOINT SENSITIVITY (Cont.)

Example - Modal Frequency Car Body

Model Statistics
Number of grids: 14990
Number of elements: 16174
Number of subcases: 1
Number of frequencies: 1
Number of modes: 67
Number of design variables: 2514
Number of responses: 1

Performance Results

Parameter V70 V69.1

CPU Time 3504.2 secs 6787.7 secs
Scratch Space 541.0MB 5238. MB
SCR 300 Space 487.4MB 1953. MB
No. of Design Cycles 1 1
Final Objective 0.6379 0.6386
Final Maximum Constraint -0.4218 -.4151

9-46
 DESIGN SENSITIVITY COEFFICIENT MATRICES

• Used by the optimizer to construct approximations to structural

response quantities, or available as output for coupling with
external optimizers. (A design engineer might also act as an
“external optimizer”)

• Definition of design sensitivity coefficient

∂r j Response
λ ij = -------
∂x i Design Variable

Responses can be either first-level or direct (i.e., displacement,

stress, etc.), or second-level (synthetic), that is, defined via
DEQATN entries.

• To output; use PARAM,OPTEXIT,±4 in Bulk Data or Case Control.

As of Version 68, OPTEXIT = 7 yields sensitivity coefficients for the
optimal design.

• Recall that a number of analysis model properties may depend on

several design variables. To allow for this:

∆r j ∆r j ∆p k
--------- =
∆x i ∑ ---------- ⋅ ----------
∆p k ∆x i
k

Sensitivity DVPREL1,
Analysis DVPREL2

where pk = properties

9-47
DESIGN SENSITIVITY COEFFICIENT MATRICES (Cont.)

• Responses are listed in order in the DSCM2 Matrix. Design

variable order is printed in the design history summary at the end
of the .f06 file.

• Matrix DSCM2 contains the sensitivity coefficients

∆r j
---------
∆x i

where i, j = row, column

DSCM2 Matrix Structure

r1 rj r1 rj
.
.
.

.
.
.

.
.
.

.
.
.
Subcase 1 Subcase 1 Subcase 2 Subcase 2

x1  ∆r 1j  ∆r j 
 -----------  ----------
 ∆x 1  1  ∆x 1 1
x2
. . . .
. . . .
.
.
.

. . . .

xn  ∆r j 
 ----------
 ∆x n k

9-48
Design Sensitivity Output Parameters DSAPRT
Specifies
DSAPRT design sensitivity output parameters. Design Sensitivity Output Parameters

Format:

FORMATTED  ALL 
NOEXPORT  
DSAPRT ( UNFORMATTED , , [ START = i ], [ BY = j ], [ END = k ]) =  n 
EXPORT  
NOPRINT  NONE 

Examples:
DSAPRT(FORMATTED,EXPORT)
DSAPRT(FORMATTED,START = FIRST,BY = 3,END = LAST) = 101
DSAPRT(UNFORMATTED,START = FIRST)
DSAPRT(UNFORMATTED,EXPORT)
DSAPRT(FORMATTED,END = 4) = ALL
DSAPRT(UNFORMATTED,END = SENS) = ALL
DSAPRT(NOPRINT,EXPORT)

Describers Meaning
FORMATTED Output will be presented with headings and labels.

UNFORMATTED Output will be printed as a matrix print (see description of the MATPRN module
in the MSC/NASTRAN DMAP Module Dictionary).

NOPRINT No output will be printed.

EXPORT Output will be exported to an external binary file specified by PARAM,IUNIT.

NOEXPORT Output will not be exported to an external binary file.

START = i Specifies the first design cycle for output. (Integer > 0 or Character: “FIRST” or
“LAST”; Default = 1 or “FIRST”)

BY = j Specifies the design cycle interval for output. (Integer ≤ 1, or > 0; Default = 0)
See Remark 2.

END = k Specifies the last design cycle for output. (Integer > 0 or Character: “FIRST”,
“LAST”, or “SENS”; Default = “LAST”)

ALL All design responses (defined in DRESP1 and DRESP2 entries) will be output.

n Set identification of a previously appearing SET command. Only sensitivities of

responses with identification numbers that appear on this SET command will be
output. (Integer > 0)

(Continued)

9-49
DSAPRT Design Sensitivity Output Parameters

Remarks:
1. Only one DSAPRT may appear in the Case Control Section and should appear above all
SUBCASE commands.

2. Sensitivity data will be output at design cycles i, i + j, i + 2j, ..., k. Note that the BY = 0
default implies no sensitivity analysis at the intermediate design cycles.

3. END = SENS requests design sensitivity analysis, and no optimization will be performed.

4. Prior to Version 69, PARAM,OPTEXIT,4, –4, or 7 requested design sensitivity output and
may still be used in Version 69. However, if both DSAPRT and PARAM,OPTEXIT,4, –4,
or 7 are specified, then DSAPRT overrides PARAM,OPTEXIT,4, –4, or 7.
PARAM,OPTEXIT values and the equivalent DSAPRT commands are as follows:

OPTEXIT Equivalent DSAPRT Command

4 DSAPRT(UNFORMATTED, END = SENS)

–4 DSAPRT(NOPRINT,EXPORT,END = SENS)

7 DSAPRT(UNFORMATTED,START = LAST)

9-50
 FORMATTED SENSITIVITY PRINT EXAMPLE

**********************************************************************
* *
* D E S I G N S E N S I T I V I T Y M A T R I X O U T P U T *
* *
* *
* R E S P O N S E S E N S I T I V I T Y C O E F F I C I E N T S *
* *
**********************************************************************

-----------------------------------------------------------------------------------------------------------------------
DRESP1 ID= 20 RESPONSE TYPE= WEIGHT SEID= 0
RESP VALUE DESIGN VARIABLE COEFFICIENT DESIGN VARIABLE COEFFICIENT
-----------------------------------------------------------------------------------------------------------------------
4.8284e+00 1 A1 2.8286e+00 2 A2 1.0006e+00
SYMMETRIC THREE BAR TRUSS DESIGN OPTIMIZATION - D200X1 JANUARY 18, 1996 MSC/NASTRAN 1/17/96 PAGE 24
9-51

BASELINE - 2 CROSS SECTIONAL AREAS AS DESIGN VARIABLES

-----------------------------------------------------------------------------------------------------------------------
DRESP1 ID= 23 RESPONSE TYPE= STRESS ELEM ID= 1 COMP NO= 2 SEID= 0
SUBCASE RESP VALUE DESIGN VARIABLE COEFFICIENT DESIGN VARIABLE COEFFICIENT
-----------------------------------------------------------------------------------------------------------------------
1 1.3530E+04 1 A1 -1.1895E+04 2 A2 -8.1543E+02
2 -9.0973E+03 1 A1 1.0742E+04 2 A2 -8.2031E+02

-----------------------------------------------------------------------------------------------------------------------
DRESP1 ID= 25 RESPONSE TYPE= STRESS ELEM ID= 3 COMP NO= 2 SEID= 0
SUBCASE RESP VALUE DESIGN VARIABLE COEFFICIENT DESIGN VARIABLE COEFFICIENT
-----------------------------------------------------------------------------------------------------------------------
1 -9.0973E+03 1 A1 1.0742E+04 2 A2 -8.2031E+02
2 1.3530E+04 1 A1 -1.1895E+04 2 A2 -8.1543E+02
 MOVE LIMITS

• Move limits are directly applied to the design variables as side

constraints

• Limits are also imposed as upper and lower bounds on each

analysis property related to the design model. These bounds are
applied to the optimization of the current approximate model and
are treated as constraints by the optimizer, not side constraints.

• Allowable move limits are problem dependent, and there is no

general a priori method of estimation

true function
F(X)
linear approximation

o
X = current design

o X
X

allowable
move limits

• For sizing problems, use relatively large move limits (such as 0.5–
1.0 initially), reducing them if convergence becomes a problem.

• Smaller move limits are recommended for dynamics problems

since such responses are often highly nonlinear.

9-52
 MOVE LIMITS IMPOSED ON ANALYSIS MODEL
PROPERTIES

• The property move limit is initialized by the value (or default)

supplied in the DOPTPRM entry in the DELP and DPMIN fields

• For property optimization, move limits are imposed on the analysis

model properties as well as the design variables

o o
• For the initial design x , analysis model properties p j , j = 1, 2, , ...,
J are determined. DELP specifies the bounds within which pj may
be modified as:

l u
pj ≤ pj ( x ) ≤ pj

where l = o o
pj p j – p j ⋅ DELP

u = o o
pj p j + p j ⋅ DELP

l o u
0 pj pj pj

9-53
 MOVE LIMITS IMPOSED ON ANALYSIS MODEL
PROPERTIES (Cont.)

• In case p j is small, both of the upper and lower bounds may be

too small imposing unnecessarily small move limits. To make sure
this does not happen, DPMIN is specified to ensure that at least
some move limits are available as follows:

l  l o 
p j = max  min [ p j ( DELP ), ( p j – DPMIN ) ], PMIN 
 

u  u o 
p j = min  max [ p j ( DELP ), ( p j + DPMIN ) ], PMAX 
 

l u
pj pj
l pjo u
pj pj

0 1

• Note that PMIN and PMAX from the DVPREL1 and 2 entries have
been accounted for as well.

• Because property limits are treated as constraints, it is possible

they may become violated during the design process if this helps
satisfy a more seriously violated constraint. In the extreme, a
physically meaningless design may result.

9-54
 MOVE LIMITS IMPOSED ON DESIGN VARIABLES

• Move limits are also imposed on the design variables

o o o o
xi – xi ⋅ DELTA ≤ x i ≤ x i + xi ⋅ DELTA

where DELTA = DELXV if specified on the DESVAR entry

or

DELX if specified on the DOPTPRM entry

or

1.0 by default

if DELXV is specified, it always has priority over DELX.

• As with the case of DPMIN for properties, DXMIN provides

minimum move limits on design variables if these are near zero

o
if x i ⋅ DELTA < DXMIN then

o o
xi – DXMIN ≤ x i ≤ xi + DXMIN

l u
However, x i and x i , provided on the DESVAR entry, are never
exceeded.

9-55
 AUTOMATIC UPDATES OF MOVE LIMITS

• Under the following conditions, move limits DELP, DPMIN and

DELX, DXMIN are reduced to one-half of the current values

• The current iteration cycle number is greater than or equal to three

• There is at least one violated constraint, and the level of constraint

violation is increasing

• Every time the move limits are changed, a message is printed and
the user is advised to include a revised DOPTPRM entry reflecting
the changes in the restart Bulk Data

9-56
 MOVE LIMITS UPDATE
^^^ USER INFORMATION MESSAGE 9052 (FEA)
^^^ STATIC ANALYSIS COMPLETED. DESIGN CYCLE NUMBER= 4

********************************************************
* *
* USER WARNING MESSAGE *
* *
* IF YOU WANT TO CONTINUE THE DESIGN AFTER THIS JOB *
* IS COMPLETED, YOU MUST INCLUDE A REVISED DOPTPRM *
* BULK DATA ENTRY IN THE BULK DATA SECTION WITH THE *
* FOLLOWING ITEMS MODIFIED AS SHOWN: *
* *
* DELP = 1.0000E-01 *
* DPMIN = 5.0000E-03 *
* DELX = 5.0000E-01 *
* DXMIN = 2.5000E-02 *
* *
********************************************************

********************************************************
9-57

* *
* IF A DELXV IS SPECIFIED ON A DESVAR BULK DATA ENTRY, *
* UPDATED DELXV VALUES ARE PRESENT ON THE DESVAR *
* ENTRIES CONTAINED IN THE PUNCH FILE. *
* *
* (NOTE: THERE MAY BE MORE THAN ONE MESSAGE LIKE *
* THIS. THE LAST ONE IN THIS RUN SHOULD *
* BE LOCATED AND USED.) *
* *
********************************************************

CONVERGENCE NOT ACHIEVED YET (HARD CONVERGENCE DECISION LOGIC)

----------------------------------------------------------------------------------------------

RELATIVE CHANGE IN OBJECTIVE : 2.8778E-02 MUST BE LESS THAN 1.0000E-03

OR ABSOLUTE CHANGE IN OBJECTIVE : 1.8917E-01 MUST BE LESS THAN 1.0000E-20
--- AND ---
MAXIMUM CONSTRAINT VALUE : 3.3610E-02 MUST BE LESS THAN 5.0000E-03
(CONVERGENCE TO A FEASIBLE DESIGN)
--- OR ---
MAXIMUM OF RELATIVE PROP. CHANGES : 0.0000E+00 MUST BE LESS THAN 1.0000E-03
AND MAXIMUM OF RELATIVE D.V. CHANGES : 4.0000E-01 MUST BE LESS THAN 1.0000E-03
(CONVERGENCE TO A BEST COMPROMISE INFEASIBLE DESIGN)
----------------------------------------------------------------------------------------------
 PARAM OPTEXIT

The OPTEXIT parameter provides a means of exiting a design task in

SOL 200.

OPTEXIT
Value Description
0 Do not exit. Proceed with optimization.
1 Exit after the initialization of the analysis and design model
but before finite element analysis begins.
2 Exit after finite element analysis and initial design response
and shape basis vector processing.
3 Exit after design constraint evaluation and screening.
4 Exit after design sensitivity analysis and print the matrix of
design sensitivity coefficients (DSCM2). This is equivalent to
the DSAPRT (UNFORM,END+SENS) Case Control
command.
–4 Exit after design sensitivity analysis and write the data blocks
related to sensitivity coefficients (DSCM2 and DSCMCOL) to
an external file using the OUTPUT2 and OUTPUT4 modules.
This is equivalent to the DSAPRT (NOPRINT,EXPORT
END+SENS) Case Control command. See related
parameters ITAPE, IUNIT, and OMAXR.
5 Exit after the first approximate optimization of the design
model.
6 Exit after the first update of the analysis model based on the
first approximate optimization of the design model.
7 Compute and output design sensitivity coefficients at the end
of normal program termination: hard convergence, soft
convergence, or maximum design cycles. This is equivalent
to the DSAPRT (UNFORM,START+LAST) Case Control
command.

9-58
 FLOW CHART FOR SOLUTION 200

MSC/NASTRAN Input File IFPL

Processing for Primary
and All Auxiliary Models
AXMDRV, AXMPR1

NO Does a Design
Model Exist?

YES

Generate Nonrepetitive Tables Used in

Design Sensitivity and Analysis.
Analysis Model Override by Design Model
DOPR1, DOPR2, DOPR4, DOM11
PREDOM

1 EXITOPT

Case Control Partitioning

Based on Analysis Type
MDCASE

PHASE0 Operations, Restart Checking for

Primary and Auxiliary Models (for Shape)
AXMDRV, AXMPR2 PHASE0

Design Optimization Data Initialization

DMPR, SDSA, SDSB DESINIT,
SETSOLAP

Begin Design Cycle Loop

Loop on Design Cycles

Database Cleanup From

Previous Design Cycle

9-59
 FLOW CHART FOR SOLUTION 200 (Cont.)

NO
If AMBS AMBS ≡ Auxiliary Model
Boundary Shapes
YES
Static Analysis for All Auxiliary Models,
Loop over Auxiliary Models, Boundary
Conditions (FEAOPT = Analysis)
FEA

NO If AMBS
GMBS ≡ Geometry Model
or GMBS
Boundary Shapes
YES
Collect Analysis Results, Generate
Enforced Boundary Displacements
for Auxiliary Structure Analysis
AXMDRV, DSAJ BNDSHP
Loop on Design Cycles

Finite Element Analysis and Data Recovery

for All Analysis Disciplines, Superelements,
and Subcases. Basis Vector Solution
(for Shape Optimization) FEA

NO If AMBS
or GMBS

YES
Partition Out Shape Basis Vector
Solutions, Generate Basis Vectors

If a Design Model Exists and NO

PARAM, OPTIM, YES EXIT

YES
Begin Design Sensitivity
and Optimization

Pre-Sensitivity Initialization Operations.

Shape Basis Vector Initialization
and Scaling Operations. PRESENS (First Design
DOM11, DOPR2, DOPR3, DOPR4, DOPR5 Cycle Only)

9-60
 FLOW CHART FOR SOLUTION 200 (Cont.)

Design Constraint Evaluation and Screening

DESCON
DSAD, DSAR, DSPRM, WEIGHT
SDR2STAT

Hard Convergence Check

DOM12

(and Design Cycle = 2 EXITOPT

6
e.g., One Pass Through Loop)

Maximum
YES
Design Cycles EXIT EXITOPT 7
Loop on Design Cycles

Completed?
NO

3 EXITOPT

If Hard YES
Convergence
Data Recovery and
NO Exit (FEAOPT =
DATARECOVERY”) FEA,
EXITOPT
*
Generate Pseudo-load Vectors for 7
Sensitivity Analysis Across Analysis
Disciplines, Subcases, and Superelements DESAERDR
DSABO, DSAE, DSAM, DSAN, ,PSLGDV,
DSAP, DSAR, DSVG1, DSVG2 SEDRDR

Generate Solutions for

Sensitivity Analysis
(FEAOPT = ’SENSITIVITY’)
FEA

*OPTEXIT=7 includes computation of sensitivities for the current design cycle (all operations up to DOM9).

9-61
 FLOW CHART FOR SOLUTION 200 (Cont.)

Compute Necessary Response

Sensitivities for All Retained Constraints RESPSEN,
DSAF, DSAH, DSAL, DSAW, DSDVRG, SEDRDR,
DSFLTF, DSFLUTE, DSPRM, DSVG1, DSVG4, SDR2STAT

Exit and Print Sensitivities

4
DSTAP2, DOM6, SDSC EXITOPT

Optimization with Respect

to Approximate Models
DOM9

Output Updated Grid Entries

DOM11
Loop on Design Cycles

Design Optimization Output

DOM10

Soft Convergence Check

DOM12

If Soft YES Data Recovery and

Exit (FEAOPT = FEA,
Convergence
’DATARECOVERY”) EXITOPT

NO 7

Analysis Model Updates

(Auxiliary + Primary)
AXMPR2 UPDATE

End Design
Cycle Loop

9-62
 CONVERGENCE AT THE DESIGN CYCLE LEVEL

IFP
Initialization
PREDOM

FEA Finite Element Analysis

DESCON Constraint Evaluation and Screening

DOM12 1 Hard Convergence

(Convergence Check Based on the
PSLGDV Data of Two Consecutive Designs)

FEA

RESPSEN Sensitivity Analysis and Optimization

with Approximate Models
DOM9

DOM10 Print Output

DOM12 Soft Convergence
2
(Check If the Optimizer Made Progress)
DOM11 Update Analysis Model

Note: Soft convergence 2 does not terminate the iteration

process, unless the SOFTEXIT parameter is YES
(default is NO).

9-63
 CONVERGENCE AT THE DESIGN CYCLE
LEVEL (Cont.)

Internal Default User

Variable Criterion Limit Parameter
(P) (P – 1)
OBJ – OBJ
CHGOBJ ------------------------------------------------------- 0.001 CONV1
(P – 1)
OBJ

ACHOBJ (P) (P – 1) 1.E-20 CONV2

OBJ – OBJ

 P (P) – P (P – 1) 
max  i i 
CHGPRP 
1 ≤ i ≤ NPROP 
---------------------------------------- 
(P – 1) 0.001 CONVPR
Pi 

 x (P) – x (P – 1) 
max  i i 
CHGDV 1 ≤ i ≤ NDV  --------------------------------------- 
(P – 1) 0.001 CONVDV
 xi 

CONMAX max { g ( x ) } 0.005 GMAX

k k

9-64
 CONVERGENCE CRITERIA

Soft Convergence Decision Logic

Compares the results of the approximate optimization with the

previous finite element analysis

CHGOBJ ≤ CONV1
No
or
ACHOBJ ≤ CONV2

YES
No
CHGPRP ≤ CONVPR

YES

CONMAX ≤ GMAX No
or
CHGDV ≤ CONVDV

YES

SOFTCV = TRUE SOFTCV = FALSE

9-65
 CONVERGENCE CRITERIA (Cont.)

Hard Convergence Decision Logic

Compares results of finite element analysis at the completion of a

design loop with the previous analysis results.

CHGOBJ < CONV1

or No
ACHOBJ < CONV2

Yes

Yes No
CONMAX < GMAX

CHGPRP<CONVPR
Yes and No
CHGDV<CONVDV

Convergence to a Best Compromise, Continue with

Unique Design Infeasible Design Optimization
Found
Nonunique If
CHGPRP > CONVPR or
CHGDV > CONVDV

9-66
 SUMMARY OF PARAMETERS FOR DESIGN
OPTIMIZATION

• The following parameters are unique to Design Sensitivity and

Optimization:

CDIF Controls the selection of finite difference scheme

used in sensitivity analysis.
YES Selects central differences (default
for shape optimization).
NO Selects forward differences (default
for property optimization only).
DESPCH Controls the frequency of updated DESVAR and
GRID Bulk Data entry output to the PUNCH file.
N<0 No output.
N=0 Final design cycle (default).
N>0 Every N-th, as well as final design
cycles.
DSNOKD Indicates if the differential stiffness effect is to be
included in buckling sensitivity analysis.
1.0 Include differential stiffness.
0.0 Do not include differential stiffness
(default).
NASPRT Controls the frequency of MSC/NASTRAN output.
-1 No output.
0 Output on initial and final design
cycles (default).
N Output every N-th iteration and also
prior to exit.

9-67
 SUMMARY OF PARAMETERS FOR DESIGN
OPTIMIZATION (Cont.)

OPTEXIT Instructs the program to exit at one of seven

predetermined exit points.
0 No user-defined exit (default).
N>0 Exit at one of the locations 1
through 7 (1 ≤ N ≤ 7).
SOFTEXIT Determines whether to terminate design cycles if
soft convergence is indicated.
NO Do not stop if soft convergence is
indicated (default).
YES Terminate design cycles if soft
convergence is achieved.
UPDTBSH Controls the update of the boundary shapes in the
analytic boundary shapes method for shape
optimization.
NO Do not update the boundary shapes
(default).
YES Update the boundary shapes.
Note: Regardless of the value of
UPDTBSH, shape basis vectors are
still updated for every design cycle
(interpolation to the interior grids).
UPDTBSH only controls updates on
the boundary shapes.

9-68
 SECTION 10

SHAPE OPTIMIZATION

BASIC EQUATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-1

BASIS VECTORS – EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-2

SHAPE BASIS VECTORS IN THE DESIGN MODEL. . . . . . . . . . . . . . . . . . . . . . . . 10-6

DESIGN MODELING INPUT FOR SHAPE OPTIMIZATION . . . . . . . . . . . . . . . . . . 10-7

MANUAL GRID VARIATION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-8

DVGRID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-10

AUXILIARY MODELS IN SHAPE OPTIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . 10-12

EXAMPLE – DIRECT INPUT OF SHAPES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-14

DVSHAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-19

AUXILIARY BOUNDARY MODELS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-24

EXAMPLE – ANALYTIC BOUNDARY SHAPES . . . . . . . . . . . . . . . . . . . . . . . . . . 10-26

AUXCASE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-30

AUXMODEL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-31

BNDGRID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-32

DVBSHAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-33

SUMMARY OF SHAPE BASIS VECTOR GENERATION METHODS . . . . . . . . . . 10-41

GUIDELINES, RECOMMENDATIONS AND LIMITATIONS. . . . . . . . . . . . . . . . . . 10-43

 BASIC EQUATION 10

MSC/NASTRAN uses shape basis vectors to describe shape changes:

Grid Point Basis Design Variable

Changes Vectors Changes
{∆G} = [T] ∗ {∆x}

where {∆G} = { G }
i + 1 – { G }i

= {x}
i + 1 – { x }i
{∆x}

i = current design

i+1 = updated design

A new shape is a linear combination of basis vectors.

The engineer determines HOW the structure can change ⇒ [T]

The optimizer determines HOW MUCH the structure can change ⇒

{∆x}
 BASIS VECTORS – EXAMPLES

Example 1 – Plate Height as Shape Design Variable

∆h = 1.0
1’ 2’
y
2
1
x 4’ h’
3’
1 ⁄ 2 ∆h
h
3 4

5 6

{ ∆G } = { G′ } – { G } = ∆x { T }

T
 
{ T } =  0. 1. 0. 0. 1. 0. 0. 0.5 0. 0. 0.5 0. 0. 0. 0. 0. 0. 0. 
 
GRID 1 GRID 2 GRID 3 GRID 4 GRID 5 GRID 6

10-2
 BASIS VECTORS – EXAMPLES (Cont.)

Example 2 – Radius of Circular Plate

6
4
θ
5 3
R 5’
3’
∆R = -1.0
1’ 1 2

{ ∆G } = { G′ } – { G } = ∆x { T }

T
 
{ T } =  – 1. 0. 0. 0. 0. 0. – 1. 0. 0. 0. 0. 0. – 1. 0. 0. 0. 0. 0. 
 
GRID 1 GRID 2 GRID 3

10-3
 BASIS VECTORS – EXAMPLES (Cont.)

Example 3

Consider a plate with a hole in the middle with unequal edge traction
forces. A quarter model is pictured below:

Ty = 1.0E4 N/m

0.15m Tx = 2.0E4 N/m

0.15m

Material:
aluminum, 7075–T6 sheet
2
E = 7.2E10 N/m
µ = 0.33
3
ρ = 2.8E3 kg/m

10-4
 BASIS VECTORS – AN EXAMPLE (Cont.)

We expect the optimal hole shape to be an ellipse, so a candidate set

of basis vectors might be:

Basis Vector {T}1 Basis Vector {T}2

MSC/NASTRAN will allow us to define design variables that represent

x and y axis intercepts as:

x1 = a
x2 = b

so that,

 ∆x 1  ∆a
{ ∆G } = [ { T 1 } { T 2 } ]  
 ∆x 2  ∆b

x-direction y-direction
variation variation

10-5
 SHAPE BASIS VECTORS IN THE DESIGN MODEL

There are two primary tasks in design modeling for shape:

Defining the shape basis vectors

Defining the design variables and correlating these to the basis

vectors

Defining shape basis vectors

Each basis vector component describes the direction and

magnitude of an individual grid component change, for a given
design variable change

This is a large amount of data.

MSC/NASTRAN provides four methods of generating shape basis

vectors

• Manual grid variation

• Direct input of shapes

• Geometric boundary shapes

• Analytic boundary shapes

10-6
DESIGN MODELING INPUT FOR SHAPE
OPTIMIZATION

Define Analysis Disciplines

(Executive Control: SOL 200;
Case Control: Analysis =)

Define Design Variables

(Bulk Data: DESVAR)

Relate Design Variables to

Analysis Model:
(Bulk Data: DVPREL1, DVPREL2)
for Shape (Bulk Data: DVGRID,
DVBSHAP, DVSHAP, BNDGRID)

Define Design Responses

(Bulk Data: DRESP1, DRESP2)

Define Objective (Case Control: DESOBJ)

and Constraints (Bulk Data: DCONSTR, DCONADD;
Case Control: DESGLB, DESSUB)

Provide any necessary

parameter overrides
(Bulk Data: DOPTPRM, DSCREEN)

10-7
 MANUAL GRID VARIATION

This method is recommended only for very simple cases, or in unique

situations.

A DVGRID entry is used to define motion for a single grid:

(Design Variable to GRID) Relationships

{∆G}i

z z
{G}i
y
y x

   
 ∆G ix   N1 
   
{ ∆G } i =  ∆G iy  = COEFF  N2  ⋅ ∆x DVID
   
 ∆G iz   N3 
   

1 2 3 4 5 6 7 8 9 10
DVGRID DVID GRID CID COEFF N1 N2 N3

10-8
 MANUAL GRID VARIATION (Cont.)

Each DVGRID entry defines a single basis vector component, as well

as identifying the related design variable.

Multiple references to the same grid-design variable result in vectorial

addition of the basis vector component.

• DVGRID entry format:

10-9
DVGRID Design Variable to Grid Point Relation

DVGRID the relationship between design variables and grid point locations.
Defines Design Variable to Grid Point Relation

Format:

1 2 3 4 5 6 7 8 9 10
DVGRID DVID GID CID COEFF N1 N2 N3

Example:

DVGRID 3 108 5 0.2 0.5 0.3 1.0

Field Contents
DVID DESVAR entry identification number. (Integer > 0)

GID Grid point (GRID) or geometric point (POINT) identification number. (Integer > 0)

CID Coordinate system identification number. (Integer ≥ 0; Default = 0)

COEFF Multiplier of the vector defined by Ni. (Real; Default = 0.0)

Ni Components of the vector measured in the coordinate system defined by CID.

(Real; at least one Ni ≠ 0.0.)

Remarks:
1. A CID of zero or blank (the default) references the basic coordinate system.

2. Multiple references to the same grid ID and design variable result in vectorial addition of
the participation vectors defined by CID, COEFF, and Ni. There is no restriction on the
number of DVGRID entries that may reference a given grid (GID) or design variable
(DVID).

3. The coordinate update equation is given as

∑ COEFF j ( X DVIDj – X DVIDj ) { N } j

0 0
{ g }i – { g }i =
i

T
where { g } i is the location of the i-th grid, [ g x g y g z ] .

T
The vector { N } = [ N x N y N z ] is determined from CID and Ni. Note that it is a change
0
in a design variable from its initial value X , and not the absolute value of the design
0
variable itself, that represents a change in a grid point location, { g } i – { g } i .

(Continued)

10-10
Design Variable to Grid Point Relation DVGRID
4. The DVGRID entry defines the participation coefficients (basis vectors) of each design
variable for each of the coordinates affected by the design process in the relationship

{ ∆g } i = ∑ { T }ij ∗ ∆X j
j

5. DVGRID entries that reference grid points on MPCs or RSSCON entries produce incorrect
sensitivities. Often the sensitivities are 0.0 which may result in a warning message indi-
cating zero gradients which may be followed by UFM 6499. Other rigid elements produce
correct results.

10-11
 AUXILIARY MODELS IN SHAPE OPTIMIZATION

The following methods use auxiliary models as an aid to shape basis

vector generation

• Direct input of shapes

• Geometric boundary shapes

• Analytic boundary shapes

• What is an auxiliary model?

It is a structure whose deformations may be used to generate basis

vectors.

Although the geometry is usually the same, boundary conditions,

loads, and possibly material types may differ.

10-12
AUXILIARY MODELS IN SHAPE OPTIMIZATION (Cont.)

Example: Elliptical hole variation in quarter plate

Loads applied to the hole boundary yield consistent deformations

that can be used as basis vectors:

{ U } Vectors from static analysis

{ ∆G } = [ T ] { ∆x }

10-13
 EXAMPLE – DIRECT INPUT OF SHAPES

(TPL problems D200CSX.DAT, D200CS.DAT)

The figure below is the initial design and von Mises stress resultants
for a two-dimensional, symmetric road support structure (culvert)

Pressure load on top surface

From the above stress distribution, we can observe that the culvert
interior profile is not optimal.

10-14
 EXAMPLE – DIRECT INPUT OF SHAPES (Cont.)

Shape Optimal Solution Approach:

• Build an auxiliary model

• Gather displacement solutions to use as basis vectors

• Define a design task

Auxiliary Model Generation

We require that the outer culvert profile remain unchanged. Interior

shape variations can be achieved by loading each interior grid
individually (7 separate load conditions).

10-15
 EXAMPLE – DIRECT INPUT OF SHAPES (Cont.)

ID,AUX1,VT100 $FEB 10,1990

TIME 10
SOL 101
CEND
TITLE=Culvert Example Using External Auxiliary Model
SUBTITLE=The External Auxiliary Model
SPC=25
$
$ seven load cases
$
SUBCASE 1
LOAD=100
DISP=ALL
SUBCASE 2
LOAD=101
DISP=ALL
SUBCASE 3
LOAD=102
DISP=ALL
SUBCASE 4
LOAD=103
DISP=ALL
SUBCASE 5
LOAD=104
DISP=ALL
SUBCASE 6
LOAD=105
DISP=ALL
SUBCASE 7
LOAD=106
DISP=ALL
BEGIN BULK
PARAM,POST,0
param,newseq,-1
$
$ The same GRID and CQUAD4 entries as the primary structure
$
GRID, 1,, 3.00000, 0.00000,.00
GRID, 2,, 4.00000, 0.00000,.00
GRID, 3,, 5.00000, 0.00000,.00
. . . . .
. . . . .
(see optimization input file)
. . . . .
. . . . .
GRID, 38,, 2.00000, 5.19600,.00
GRID, 39,, 2.50000, 5.19600,.00
GRID, 40,, 3.00000, 5.19600,.00

10-16
 EXAMPLE – DIRECT INPUT OF SHAPES (Cont.)

CQUAD4, 1,101, 1, 2, 6, 5
CQUAD4, 2,101, 2, 3, 7, 6
CQUAD4, 3,101, 3, 4, 8, 7
. . . . . . .
. . . . . . .
(see optimization input file)
. . . . . . .
. . . . . . .
CQUAD4, 25,101, 30, 31, 38, 37
CQUAD4, 26,101, 31, 32, 39, 38
CQUAD4, 27,101, 32, 33, 40, 39
PSHELL,101,102,.44
MAT1,102,2.+7,,.3
$
$ Additional CBAR elements maintain smoothness of the circular boundary
$
CBAR,31,1,13,14,,1.0
CBAR,32,1,14,15,,1.0
CBAR,33,1,15,16,,1.0
CBAR,34,1,16, 9,,1.0
CBAR,35,1, 9, 5,,1.0
CBAR,36,1,5 , 1,,1.0
PBAR 1 102 20.0 1.0 1.0
$
$ Seven load cases
$
FORCE,100,13,0,1.e5,0.,1.,0.
FORCE,101,14,0,1.e5,0.259,.9659
FORCE,102,15,0,1.e5,0.5,0.866,0.0
FORCE,103,16,0,1.e5,1.,1.,0.
FORCE,104,9,0,1.e5,0.866,0.5,0.0
FORCE,105,5,0,1.e5,0.9659,0.259
FORCE,106,1,0,1.e5,1.,0.,0.
$
$ Boundary conditions satisfy functional and manufacturing requirements
$
SPC1,25,345,1,THRU,40
SPC1,25,6,2,THRU,4
SPC1,25,6,6,THRU,8
SPC1,25,6,10,THRU,12
SPC1,25,6,17,THRU,19
SPC1,25,6,20,THRU,26
SPC1,25,6,27,THRU,33
SPC1,25,6,34,THRU,40
SPC1,25,12,33,THRU,40
SPC1,25,12,4,8,12,19,26
SPC1,25,1,13,20,27
SPC1,25,2,1,2,3
ENDDATA

10-17
 EXAMPLE – DIRECT INPUT OF SHAPES (Cont.)

Optimization input

• Displacement vectors are dblocate’d and defined as basis vectors

using DVSHAP Bulk Data entries.

10-18
Design Variable to Basis Vector(s) DVSHAP
Defines a shape basis vector by relating a design variable identificationDesign
DVSHAP number (DVID)
Variable to columns
to Basis Vector(s)
of a displacement matrix.

Format:

1 2 3 4 5 6 7 8 9 10
DVSHAP DVID COL1 SF1 COL2 SF2 COL3 SF3

Example:

DVSHAP 2 1 2.0 4 1.0

Field Contents
DVID Design variable identification number on the DESVAR entry. (Integer > 0)

COLi Column number of the displacement matrix. See Remark 2. (1 ≤ Integer ≤

maximum column number in the displacement matrix.)

SFi Scaling factor applied to the COLi-th column of the displacement matrix. (Real;
Default = 1.0)

Remarks:
1. DVID must be defined on a DESVAR entry.

2. COLi must be a valid column number in the displacement matrix.

3. Multiple references to the same DVID and/or COLi will result in a linear combination of
displacement vectors. In the example above, the shape basis vector is a linear combina-
tion of the fourth column and twice the second column.

4. The displacement matrix must have been created by MSC/NASTRAN and be available on
a database, which is attached via the DBLOCATE FMS statement shown below:

ASSIGN DISPMAT=’ physical filename of MASTER DBset ’

DBLOCATE DATABLK=(UG/UGD,GEOM1/GEOM1D,GEOM2/GEOM2D) ,
LOGICAL=DISPMAT

10-19
 EXAMPLE – DIRECT INPUT OF SHAPES (Cont.)

$
$ FMS section for retrieving the auxiliary displacement matrix
$
assign f1_aux=’culvert1.MASTER’
dblocate datablk=(ug/ugd,geom1/geom1d,geom2/geom2d) ,
logical=f1_aux
SOL 200 $
TIME 100
CEND
TITLE=CULVERT EXAMPLE USING EXTERNAL AUXILIARY STRUCTURE
SUBTITLE=THE PRIMARY STRUCTURE
ANALYSIS = STATICS
SPC=25
LOAD=1
DISP=ALL
STRESS=all
DESSUB = 10
desobj = 5
BEGIN BULK
PARAM,POST,0
$ PARAM,optexit,4
PARAM,NEWSEQ,-1
GRID, 1,, 3.00000, 0.00000,.00
GRID, 2,, 4.00000, 0.00000,.00
GRID, 3,, 5.00000, 0.00000,.00
GRID, 4,, 6.00000, 0.00000,.00
GRID, 5,, 2.89464, 0.78478,.00
GRID, 6,, 3.79369, 0.75885,.00
GRID, 7,, 4.69274, 0.73293,.00
GRID, 8,, 5.59178, 0.70700,.00
GRID, 9,, 2.60164, 1.49178,.00
GRID, 10,, 3.46229, 1.46585,.00
GRID, 11,, 4.32293, 1.43993,.00
GRID, 12,, 5.18357, 1.41400,.00
GRID, 13,, 0.00000, 3.00000,.00
GRID, 14,, 0.78478, 2.89464,.00
GRID, 15,, 1.49178, 2.60164,.00
GRID, 16,, 2.12100, 2.12100,.00
GRID, 17,, 3.00578, 2.12100,.00
GRID, 18,, 3.89057, 2.12100,.00
GRID, 19,, 4.77535, 2.12100,.00

10-20
 EXAMPLE – DIRECT INPUT OF SHAPES (Cont.)

GRID, 20,, 0.00000, 3.73200,.00

GRID, 21,, 0.68985, 3.66176,.00
GRID, 22,, 1.32785, 3.46643,.00
GRID, 23,, 1.91400, 3.14600,.00
GRID, 24,, 2.67052, 3.14600,.00
GRID, 25,, 3.42704, 3.14600,.00
GRID, 26,, 4.18357, 3.14600,.00
GRID, 27,, 0.00000, 4.46400,.00
GRID, 28,, 0.59493, 4.42888,.00
GRID, 29,, 1.16393, 4.33122,.00
GRID, 30,, 1.70700, 4.17100,.00
GRID, 31,, 2.33526, 4.17100,.00
GRID, 32,, 2.96352, 4.17100,.00
GRID, 33,, 3.59178, 4.17100,.00
GRID, 34,, 0.00000, 5.19600,.00
GRID, 35,, 0.50000, 5.19600,.00
GRID, 36,, 1.00000, 5.19600,.00
GRID, 37,, 1.50000, 5.19600,.00
GRID, 38,, 2.00000, 5.19600,.00
GRID, 39,, 2.50000, 5.19600,.00
GRID, 40,, 3.00000, 5.19600,.00
CQUAD4, 1,101, 1, 2, 6, 5
CQUAD4, 2,101, 2, 3, 7, 6
CQUAD4, 3,101, 3, 4, 8, 7
CQUAD4, 4,101, 5, 6, 10, 9
CQUAD4, 5,101, 6, 7, 11, 10
CQUAD4, 6,101, 7, 8, 12, 11
CQUAD4, 7,101, 9, 10, 17, 16
CQUAD4, 8,101, 10, 11, 18, 17
CQUAD4, 9,101, 11, 12, 19, 18
CQUAD4, 10,101, 13, 14, 21, 20
CQUAD4, 11,101, 14, 15, 22, 21
CQUAD4, 12,101, 15, 16, 23, 22
CQUAD4, 13,101, 20, 21, 28, 27
CQUAD4, 14,101, 21, 22, 29, 28
CQUAD4, 15,101, 22, 23, 30, 29
CQUAD4, 16,101, 27, 28, 35, 34
CQUAD4, 17,101, 28, 29, 36, 35
CQUAD4, 18,101, 29, 30, 37, 36
CQUAD4, 19,101, 16, 17, 24, 23
CQUAD4, 20,101, 17, 18, 25, 24
CQUAD4, 21,101, 18, 19, 26, 25
CQUAD4, 22,101, 23, 24, 31, 30
CQUAD4, 23,101, 24, 25, 32, 31
CQUAD4, 24,101, 25, 26, 33, 32
CQUAD4, 25,101, 30, 31, 38, 37
CQUAD4, 26,101, 31, 32, 39, 38
CQUAD4, 27,101, 32, 33, 40, 39
FORCE 1 34 0 1250. -1.
FORCE 1 35 0 2500. -1.

10-21
 EXAMPLE – DIRECT INPUT OF SHAPES (Cont.)

FORCE 1 36 0 2500. -1.

FORCE 1 37 0 2500. -1.
FORCE 1 38 0 2500. -1.
FORCE 1 39 0 2500.00 -1.
FORCE 1 40 0 1250. -1.
PSHELL,101,102,.44
MAT1,102,2.+7,,.3,0.731-3
SPC1,25,3456,1,THRU,40
SPC1,25,12,1,THRU,4
SPC1,25,1,13,20,27,34
$
$ design model
$
desvar 1 b1 3. -1.e6 1.e6 .2
desvar 2 b1 3. -1.e6 1.e6 .2
desvar 3 b3 3. -1.e6 1.e6 .2
desvar 4 b4 3. -1.e6 1.e6 .2
desvar 5 b5 3. -1.e6 1.e6 .2
desvar 6 b6 3. -1.e6 1.e6 .2
desvar 7 b7 3. -1.e6 1.e6 .2
$
$ A DVSHAP entry defines a shape basis vector by associating one design
$ variable to a dblocated displacement vector.
$
dvshap 1 1 66.773
dvshap 2 2 117.35 These scaling factors are chosen to
dvshap 3 3 216.33 normalize the basis vectors with the
dvshap 4 4 443.55
resulting maximum component of each
dvshap 5 5 220.89
dvshap 6 6 115.69 basis vector being unit.
dvshap 7 7 65.669
dresp1 5 volume volume
dresp1 2 von-mis stress pshell 9 101
DCONSTR 10 2 -3.100e43.100e4
doptprm DESMAX 25 APRCOD 1
param,nasprt,1
ENDDATA

10-22
 EXAMPLE – DIRECT INPUT OF SHAPES (Cont.)

Results:

Final volume is reduced by more than 20%. Maximum stess is

increased by 16%.

10-23
 AUXILIARY BOUNDARY MODELS

We have seen how auxiliary models can be used to generate shape

variations over the domain of the structure.

Auxiliary boundary models can likewise be used to generate shape

variations over the boundaries of structures. These can be used to
smooth out the applied shape-changing loads.

The geometric boundary shapes and analytic boundary shapes

methods both interpolate boundary variations to the structure’s
interior. The result is a set of shape basis vectors.

Auxiliary boundary model example:

Bar Elements

Primary Model Auxiliary Boundary Model

10-24
 AUXILIARY BOUNDARY MODELS

Auxiliary boundary model solutions yield the boundary shapes.

δ δ
Constant

Linear

Quadratic

Cubic

10-25
 EXAMPLE – ANALYTIC BOUNDARY SHAPES

Consider the following solid cantilever model, with tip loading as

shown:

165
154
143

132
121
110

Z
Y

10-26
 EXAMPLE – ANALYTIC BOUNDARY SHAPES (Cont.)

Initial von Mises stress distribution:

Z
Y

Suppose we would like to change the shape of the structure to

minimize mass subject to constraints on von Mises stresses.

To change the shape, we will modify the upper and lower planes of
the cantilever.

10-27
 EXAMPLE – ANALYTIC BOUNDARY SHAPES (Cont.)

Z
Y

The auxiliary boundary models will produce deformations over the

primary model boundaries. These deformations are then interpolated
to the interior. The resulting boundary + interior motion will yield a
shape basis vector.

From the auxiliary model analysis we have

 x1 
[ T ] = [ 1.0U SUBCASE200 1.0U SUBCASE300 ]  
 x2 

10-28
EXAMPLE – ANALYTIC BOUNDARY SHAPES (Cont.)

• Text input for analytic boundary shapes

AUXCASE

AUXMODEL

BNDGRID

DVBSHAP

10-29
AUXCASE Auxiliary Model Case Control Delimiter

AUXCASE Case Control Commands for an Auxiliary Model in SOL 200.

Delimits Auxiliary Model Case Control Delimiter

Format:
AUXCASE

Examples:
AUXCAS
AUXC

Remarks:
1. AUXCASE indicates the beginning of Case Control commands for an auxiliary model.
AUXCASE must follow the primary model Case Control commands.

2. All Case Control commands following this entry are applicable until the next AUXCASE or
BEGIN BULK command. Commands from preceding Case Control Sections are ignored.

3. Each auxiliary model Case Control must be delimited with the AUXCASE command.

4. The AUXMODEL command is used to associate the auxiliary model Case Control with a
particular auxiliary model.

10-30
Auxiliary Model Identification Number AUXMODEL
AUXMODEL an auxiliary model for generation of boundary shapes in Auxiliary
References shape optimization.
Model Identification Number

Format:
AUXMODEL = n

Examples:
AUXMODEL = 4
AUXM = 4

Describer Meaning
n Auxiliary model identification number. (Integer > 0)

Remarks:
1. AUXMODEL references a particular auxiliary model for analysis and may only be specified
in the auxiliary model Case Control Section.
2. See the BEGIN BULK command for the Bulk Data definition of an auxiliary model.

10-31
BNDGRID Boundary Grid Points

BNDGRIDa list of grid point identification numbers on design boundaries or surfaces

Specifies for shape
Boundary opti-
Grid Points
mization (SOL 200).

Format:

1 2 3 4 5 6 7 8 9 10
BNDGRID C GP1 GP2 GP3 GP4 GP5 GP6 GP7

GP8 -etc.-

Example:

BNDGRID 123 41 42 43 44 45 46 47

49

Alternate Format and Example:

BNDGRID C GP1 "THRU" GP2

BNDGRID 123 41 THRU 49

Field Contents
C Component number (any unique combination of integers 1 through 6 with no
embedded blanks). See Remark 1.

GPi Shape boundary grid point identification number. (0 < Integer < 1000000; For
THRU option, GP1< GP2)

Remarks:
1. C specifies the components for the listed grid points for which boundary motion is
prescribed.

2. Multiple BNDGRID entries may be used to specify the shape boundary grid point identifi-
cation numbers.

3. Both fixed and free shape boundary grid point identification numbers are listed on this
entry.

4. The degrees of freedom specified on BNDGRID entries must be sufficient to statically con-
strain the model.

5. Degrees of freedom specified on this entry form members of the mutually exclusive s-set.
They may not be specified on other entries that define mutually exclusive sets. See the
MSC/NASTRAN Quick Reference Guide, Appendix B for a list of these entries.

10-32
Design Variable to Boundary Shapes DVBSHAP
Associates a design variable identification number to a linear combination
DVBSHAP of to
Design Variable boundary shape
Boundary Shapes
vectors from a particular auxiliary model.

Format:

1 2 3 4 5 6 7 8 9 10
DVBSHAP DVID AUXMOD COL1 SF1 COL2 SF2 COL3 SF3

Example:

DVBSHAP 4 1 1 1.6

Field Contents
DVID Design variable identification number of a DESVAR entry. (Integer > 0)

AUXMOD Auxiliary model identification number. (Integer > 0)

COLi Load sequence identification number from AUXMODEL Case Control command.
(Integer > 0)

SFi Scaling factor for load sequence identification number. (Real; Default = 1.0)

Remarks:
1. Design variable DVID must be defined on a DESVAR entry.

2. Multiple references to the same DVID and/or COLi will result in the vector addition of the
referenced boundary shape vectors.

3. Multiple DVBSHAP entries may be specified.

10-33
 EXAMPLE – ANALYTIC BOUNDARY SHAPES (Cont.)

Modeling input:

TIME 600 $
SOL 200 $
CEND
TITLE = CANTILEVERED BEAM - HEXA **** D200AM3 **** D200AM3
DESOBJ = 15
DESSUB = 100
SUBCASE 100
ANALYSIS = STATICS
SPC = 1
LOAD = 1
DISPLACEMENT = ALL
OUTPUT(PLOT)
SET 1 = ALL
VIEW 90.0,0.0,90.0
PLOT SET 1
$SET 2 = ALL
$VIEW 34.0, 24.0, 0.0
$PLOT SET 2
$PLOT STATIC DEFORMATION SET 2
AUXCASE
TITLE = AUXILIARY MODEL 1
AUXMODEL = 1
SUBCASE 200
SPC = 200
LOAD = 220
LABEL = UPPER
SUBCASE 300
SPC = 300
LOAD = 330
LABEL = LOWER
BEGIN BULK
$
$------------------------------------------------------------------------
$ ANALYSIS MODEL:
$------------------------------------------------------------------------
$
PARAM AUTOSPC YES
PARAM POST 0
PARAM GRDPNT 0
PARAM MAXRATIO1.0E+8
$
CORD2S* 2 0 0.0 0.0+1A 2
*1A 2 0.0 0.0 0.0 1000.000+1B 2
*1B 2 1000.000 0.0 0.0 +1C 2
*1C 2
CORD2C* 1 0 0.0 0.0+1A 1
*1A 1 0.0 0.0 0.0 1000.000+1B 1
*1B 1 1000.000 0.0 0.0 +1C 1
*1C 1

10-34
 EXAMPLE – ANALYTIC BOUNDARY SHAPES (Cont.)

GRID 1 0 0.0 0.0 0.0 0

. . . . . . .
. . . . . . .
. . . . . . .
GRID 165 0 10.000 2.000 4.000 0
$GRDSET 456
CHEXA 1 1 1 2 13 12 34 35+EA 1
+EA 1 46 45
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
CHEXA 80 1 120 121 132 131 153 154+EA 80
+EA 80 165 164
MAT1* 1 2.0680E+05 0.28999999166+MA 1
*MA 1 1.00000000 1.169999996E-05 +MB 1
*MB 1 1500000.00 1500000.00 68000.00 +MC 1
*MC 1
PSOLID 1 1 0 0 0 0
SPC 1 1 123456 0.0
SPC 1 12 123456 0.0
SPC 1 23 123456 0.0
SPC 1 34 123456 0.0
SPC 1 45 123456 0.0
SPC 1 56 123456 0.0
SPC 1 67 123456 0.0
SPC 1 78 123456 0.0
SPC 1 89 123456 0.0
SPC 1 100 123456 0.0
SPC 1 111 123456 0.0
SPC 1 122 123456 0.0
SPC 1 133 123456 0.0
SPC 1 144 123456 0.0
SPC 1 155 123456 0.0
SPC1 1 456 1 THRU 165
FORCE 1 143 0 0.5 0.0 0.0 -50.0
FORCE 1 154 0 1.0 0.0 0.0 -50.0
FORCE 1 165 0 0.5 0.0 0.0 -50.0
$
$------------------------------------------------------------------------
$ DESIGN MODEL:
$------------------------------------------------------------------------
$
PARAM,DESPCH,1
PARAM,NASPRT,1
$
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
DESVAR 1 UPPER 1.0 .00 7.00 0.4
DESVAR 2 LOWER 1.0 .00 7.00 0.4
$
$DVBSHAP,DVID, AUXMID, COL1, SF1, COL2, SF2, ...
DVBSHAP 1 1 1 1.0
DVBSHAP 2 1 2 1.0

10-35
 EXAMPLE – ANALYTIC BOUNDARY SHAPES (Cont.)

$
$DLINK, ID, DDVID, CO, CMULT, IDV1, C1, IDV2, C2, +
$+, IDV3, C3, ...
DLINK 1 2 1.0 1 1.0
$
$ BOUNDARY CONDITIONS FOR SHAPE INTERPOLATIONS:
$
$ ---TOP SURFACE:
$
BNDGRID,C, GP1, GP2, GP3, GP4, GP5, GP6, GP7, +
$+, GP8, ...
BNDGRID 123 133 134 135 136 137 138 139
140 141 142 143 144 145 146 147
148 149 150 151 152 153 154 155
156 157 158 159 160 161 162 163
164 165
$
$ ---BOTTOM SURFACE:
BNDGRID 123 1 2 3 4 5 6 7
8 9 10 11 12 13 14 15
16 17 18 19 20 21 22 23
24 25 26 27 28 29 30 31
32 33
$
$ ---EXTERIOR SURFACES - INTERPOLATION IN X&Z DIRECTION ONLY:
BNDGRID 2 34 35 36 37 38 39 40
41 42 43 44
BNDGRID 2 56 57 58 59 60 61 62
63 64 65 66
BNDGRID 2 67 68 69 70 71 72 73
74 75 76 77
BNDGRID 2 89 90 91 92 93 94 95
96 97 98 99
BNDGRID 2 100 101 102 103 104 105 106
107 108 109 110
BNDGRID 2 122 123 124 125 126 127 128
129 130 131 132
$
$ ---TIP END:
BNDGRID 1 11 22 33 44 55 66
BNDGRID 1 77 88 99 110 121 132
BNDGRID 1 143 154 165
$
$ ---FIXED END:
BNDGRID 123 1 12 23 34 56 67 89
100 122 133 144 155
BNDGRID 1 45 78 111
$
$ FORMULATE WEIGHT-BASED SYNTHETIC RESPONSE: F = 1.E5*W
DRESP1 1 WEIGHT WEIGHT
DRESP2 15 WE1000 1
+ DRESP1 1
DEQATN 1 F(A)=100000.*A

10-36
 EXAMPLE – ANALYTIC BOUNDARY SHAPES (Cont.)

$
$ CONSTRAINTS ON VON MISES STRESSES:
DRESP1 2 STRESS STRESS PSOLID 13 1
DSCREEN STRESS -1.0 10
DCONSTR 100 2 200.
$
$ OVERRIDE OF OPTIMIZATION PARAMETERS
DOPTPRM DESMAX 9 P1 1 P2 15
$
$------------------------------------------------------------------------
$ AUXILIARY BOUNDARY MODEL(S):
$------------------------------------------------------------------------
$
BEGIN BULK AUXMODEL=1
PARAM,PRGPST,NO
PARAM MAXRATIO1.0E+8
PARAM,AUTOSPC,YES
$
$ LOWER SURFACE:
CQUAD4 1000 2 1 2 13 12 0.0
CQUAD4 1001 2 2 3 14 13 0.0
CQUAD4 1002 2 3 4 15 14 0.0
CQUAD4 1003 2 4 5 16 15 0.0
CQUAD4 1004 2 5 6 17 16 0.0
CQUAD4 1005 2 6 7 18 17 0.0
CQUAD4 1006 2 7 8 19 18 0.0
CQUAD4 1007 2 8 9 20 19 0.0
CQUAD4 1008 2 9 10 21 20 0.0
CQUAD4 1009 2 10 11 22 21 0.0
CQUAD4 1010 2 12 13 24 23 0.0
CQUAD4 1011 2 13 14 25 24 0.0
CQUAD4 1012 2 14 15 26 25 0.0
CQUAD4 1013 2 15 16 27 26 0.0
CQUAD4 1014 2 16 17 28 27 0.0
CQUAD4 1015 2 17 18 29 28 0.0
CQUAD4 1016 2 18 19 30 29 0.0
CQUAD4 1017 2 19 20 31 30 0.0
CQUAD4 1018 2 20 21 32 31 0.0
CQUAD4 1019 2 21 22 33 32 0.0
$
$ UPPER SURFACE:
CQUAD4 950 2 133 144 145 134 0.0
CQUAD4 951 2 134 145 146 135 0.0
CQUAD4 952 2 135 146 147 136 0.0
CQUAD4 953 2 136 147 148 137 0.0
CQUAD4 954 2 137 148 149 138 0.0
CQUAD4 955 2 138 149 150 139 0.0
CQUAD4 956 2 139 150 151 140 0.0
CQUAD4 957 2 140 151 152 141 0.0
CQUAD4 958 2 141 152 153 142 0.0
CQUAD4 959 2 142 153 154 143 0.0
CQUAD4 960 2 144 155 156 145 0.0
CQUAD4 961 2 145 156 157 146 0.0
CQUAD4 962 2 146 157 158 147 0.0

10-37
 EXAMPLE – ANALYTIC BOUNDARY SHAPES (Cont.)

CQUAD4 963 2 147 158 159 148 0.0

CQUAD4 964 2 148 159 160 149 0.0
CQUAD4 965 2 149 160 161 150 0.0
CQUAD4 966 2 150 161 162 151 0.0
CQUAD4 967 2 151 162 163 152 0.0
CQUAD4 968 2 152 163 164 153 0.0
CQUAD4 969 2 153 164 165 154 0.0
$
MAT1 11 2.1E+5 0.8E+5 0.3 0.00
PSHELL 2 11 0.20 11 0.0
SPC1 200 123456 1 12 23
SPC1 200 12 11 22 33
SPC1 200 123456 34 THRU 165
SPCD 220 11 3 1.0 22 3 1.0
SPCD 220 33 3 1.0
SPC1 200 3 11 22 33
$
SPC1 300 123456 133 144 155
SPC1 300 12 143 154 165
SPC1 300 123456 1 THRU 132
SPCD 330 143 3 -1.0 154 3 -1.0
SPCD 330 165 3 -1.0
SPC1 300 3 143 154 165
ENDDATA

10-38
 EXAMPLE – ANALYTIC BOUNDARY SHAPES (Cont.)

Optimization results:

***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************
(HARD CONVERGENCE ACHIEVED)
(SOFT CONVERGENCE ACHIEVED)
NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 6
NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 5
OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY
---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
CYCLE APPROXIMATE EXACT OF OF
10-39

NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT

---------------------------------------------------------------------------------------------------------------
INITIAL 8.000000E+06 -3.827773E-01
1 7.401214E+06 7.401214E+06 0.000000E+00 -3.720914E-01
2 6.562918E+06 6.562912E+06 9.142283E-07 -3.570341E-01
3 5.389287E+06 5.389288E+06 -2.783299E-07 -4.672768E-02
4 5.350312E+06 5.350312E+06 0.000000E+00 8.081818E-04
5 5.350312E+06 5.350312E+06 0.000000E+00 8.081055E-04
----------------------------------------------------------------------------------------------------------------------
AUXILIARY MODEL 1 MARCH 16, 1994 MSC/NASTRAN 3/15/94 PAGE 226
DESIGN VARIABLE HISTORY
-------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
-------------------------------------------------------------------------------------------------------------------------
1 | 1 | UPPER | 1.0000E+00 : 1.4000E+00 : 1.9600E+00 : 2.7440E+00 : 2.7700E+00 : 2.7700E+00 :
2 | 2 | LOWER | 1.0000E+00 : 1.4000E+00 : 1.9600E+00 : 2.7440E+00 : 2.7700E+00 : 2.7700E+00 :
-------------------------------------------------------------------------------------------------------------------------
*** USER INFORMATION MESSAGE 6464 (DOM12E)
RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 5.
 EXAMPLE – ANALYTIC BOUNDARY SHAPES (Cont.)

Final Shape:

Z
Y

Final von Mises stress distribution:

Z
Y

10-40
 SUMMARY OF SHAPE BASIS VECTOR
GENERATION METHODS

• Manual grid variation

• Based entirely on DVGRID entries

• General, yet tedious and prone to input error

• Direct input of shapes

• Based on external auxiliary models

• DBLOCATE {U}’s, use to form [T]

• Entries: DBLOCATE, DVSHAP

• Geometric boundary shapes

• DVGRIDs specified over boundaries

• Code interpolates interior grid motion → [T]

• Efficient coupling with geometric modelers

• Entries: DVGRID, BNDGRID

10-41
SUMMARY OF SHAPE BASIS VECTOR GENERATION
METHODS (Cont.)

• Analytic boundary shapes

• Auxiliary boundary model solutions

• Code interpolates interior grid motion → [T]

• Interface completely within the MSC/NASTRAN text input

environment

• Entries: AUXCASE, AUXMODEL, BNDGRID, DVBSHAP

10-42
 GUIDELINES, RECOMMENDATIONS AND
LIMITATIONS

• Always preview shape basis vectors before starting an

optimization job to identify possible modeling errors and to check
whether the shape basis vectors satisfy design requirements.

• It is recommended to normalize a shape basis vector by its

maximum component. With a normalized basis vector, a user may
directly relate a physical parameter (e.g., radius, width) to a shape
design variable. Therefore, the actual lower and upper bounds can
be specified on DESVAR entries.

• Use smaller move limits for a shape design variable (delxv = 0.5 or
less)

• The new or updated finite element meshes can be shown using

Version 7.0 (or higher) of MSC/PATRAN. If other preprocessing
software is used, the updated GRID entries in the *.pch file may be
used.

• When an auxiliary model is used, rigid elements are useful to

generate certain basis vectors which change a feature’s location
but maintain the feature’s shape. In addition, temperature or
gravity loading may be useful to create special shape changes.

• Mode shapes generated from a normal modes analysis may be

used for basis vectors (Direct Input of Shapes only).

10-43
 GUIDELINES, RECOMMENDATIONS AND
LIMITATIONS (Cont.)

• The BNDGRID entries list grids whose displacements will be

specified, rather than interpolated. By default, these specified
displacements will be zero, unless a nonzero value is supplied via
either a DVGRID entry or an auxiliary boundary solution.

• As a limitation, a shape optimization task always starts with the

mesh given by GRID Bulk Data entries regardless of the initial
values on DESVAR and DVGRID entries.

10-44
 SECTION 11

DYNAMIC RESPONSE OPTIMIZATION

DYNAMIC RESPONSE OPTIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-1

GENERAL CONSIDERATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-2

BASIC EQUATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-3

EXAMPLE – MODAL FREQUENCY RESPONSE OPTIMIZATION . . . . . . . . . . . . . 11-6

EXAMPLE – ACOUSTIC OPTIMIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-18

 DYNAMIC RESPONSE OPTIMIZATION 11

• Available analysis disciplines in Solution 200

• Direct Frequency

• Modal Frequency

• Modal Transient

• Acoustic (Fluid-Structure Interaction)

• Available response types: (see also the DRESP1 entry)

• Displacement

• Velocity

• Acceleration

• SPC Force

• Stress

• Element force

• Equations (DRESP2+DEQATN)

• Weight

• Volume

11-1
 GENERAL CONSIDERATIONS

Design responses are computed just for the output frequencies and
output time steps:

• For frequency response:

⇒ OFREQ

• For transient response:

OTIME t after OTIME reduction

NOi t after NOi reduction

TSTEP t for all analysis steps

The design model can include:

• Property changes

• Shape changes

• Superelements

In short, all of the design modeling tools available for other analysis
disciplines are available for dynamic optimization as well.

11-2
 BASIC EQUATIONS

• Direct frequency response

• Equation of motion

2
[ – ω M + iωB + K ]u = F

• Displacement sensitivities, let

∂
∆ = -------
∂x i

2 2
[ – ω M + iωB + K ]∆u = ∆F – [ – ω ∆M + iω∆B + ∆K ]u

• Modal frequency response

• Modal transformation,

u = φξ

Differentiate

∂
-------
∂x i

and apply

[ ∆u = ∆φξ + φ∆ξ = ∆η ]

which avoids calculation of eigenvector sensitivities

11-3
 BASIC EQUATIONS (Cont.)

2 T
[ – ω φ Mφ + iωφ T Bφ + φ T Kφ ]∆η

T T 2
= φ ∆F – φ [ – ω ∆M + iω∆B + ∆K ]U

• Modal transient response

• Equation of motion

Mu̇˙ + Bu̇ + Ku = F

with modal transformation u = φξ differentiating, and using ∆η

T T T
[ φ Mφ ]∆η̇˙ + [ φ Bφ ]∆η̇ + [ φ Kφ ]∆η

T T
= φ ∆F – φ ( ∆Mu̇˙ + ∆Bu̇ + ∆Ku )

Limitations in Dynamic Response Sensitivity

• ∆F assumed zero in direct and modal frequency and modal

transient

This assumption is usually good, except for those situations in

which the following may be significant:

• Gravity (or other mass-related) loads

11-4
 BASIC EQUATIONS (Cont.)

• Follower forces (shape sensitivity)

• Thermal loading

• Complete stiffness contribution can be written as:

K = K 1 + igK 1 + i ∑ ge ke + K2
e

where K1 = structural stiffness

g = uniform structural damping coefficient PARAM,G

ge = structural damping coefficient on MAT entry

K2 = direct matrix input at grids (DMIG)

• Limitations:

g, ge, and K2 must be constant (no dependence on design vari-

ables).

Similarly, M2 and B2 are constant.

• Accurate sensitivities for the modal solutions typically require

retaining more modes (e.g., twice as many) as are required for
accurate modal analysis

11-5
 EXAMPLE – MODAL FREQUENCY RESPONSE
OPTIMIZATION

This example illustrates minimization of an rms displacement over a

given frequency range. (The example is also presented in the
MSC/NASTRAN Design Sensitivity and Optimization User’s Guide,
Section 7.7)

A simple pressure loaded flat plate is shown in the figure below. The
pressure loading is a constant 1.0 lbf/in**2 applied over a frequency
range of 20.0 to 200.0 Hz. We would like to minimize the transverse
mean square displacement of the free edge midside grid over the
same frequency range.

Grid 1110

X
(Uniform Pressure)

11-6
 EXAMPLE – MODAL FREQUENCY RESPONSE
OPTIMIZATION (Cont.)

Plate thicknesses are the design quantities, as shown in the following

figure:

Symmetric
y Boundary Conditions

Grid 1110

t1 = x1 Free
Edge

Ten independent design variables are used to describe the ten

“bands” of plate thicknesses.

The mean square displacement minimization will be subject to a

weight budget (minimize the displacement without appreciably
changing the weight):

7.99 ≤ vol ≤ 8.01

A mean square displacement can be written as:

100 200
2 2
∑ ∑
i 2i
minφ = ( u z, 1110 ) + 2 ( u z, 1110 )
i = 20 i = 51

11-7
 EXAMPLE – MODAL FREQUENCY RESPONSE
OPTIMIZATION (Cont.)

Input file:

$ Unit Test Data Deck

$ d200t4af: SOL200, V68
$
$ID MSC, D200t4af
ID MSC, D200RML3 $ V68.2
TIME 200
$TIME 100
SOL 200
CEND
TITLE = Synthesis of Responses across Different Frequencies: D200RML3
SET 10 = 1110
DISPL(PHASE,SORT1) = 10 $ MAGNITUDE/PHASE REPRESENTATION FOR RESPONSE
$ ANALYSIS AS WELL AS SENSITIVITY ANALYSIS
desglb = 10
subcase 1
SPC = 100
LOADSET = 720
DLOAD = 700
FREQ = 740
METHOD = 500
ANALYSIS = MFREQ
sdamping = 2000
DESOBJ = 1
output
disp(plot,phase) = 10
output(xyout)
cscale 2.0
ymax=4.0
plotter = nastran
ytitle = displacement at grid 1110
xyplot disp / 1110(t3)
$
BEGIN BULK
PARAM, WTMASS, .002588
$-------------------------------------------------------------------------------
$ ANALYSIS MODEL
$-------------------------------------------------------------------------------
$
$...GRID AND SPC DATA:
$
GRDSET, , , , , , , 6
GRID, 100, , 0., 0., 0.
=, *(1), , *(1.), =, =
=9
GRID, 200, , 0., 1.0, 0.
=, *(1), , *(1.), =, =
=9
GRID, 300, , 0., 2.0, 0.
=, *(1), , *(1.), =, =
=9

11-8
 EXAMPLE – MODAL FREQUENCY RESPONSE
OPTIMIZATION (Cont.)

GRID, 400, , 0., 3.0, 0.

=, *(1), , *(1.), =, =
=9
GRID, 500, , 0., 4.0, 0.
=, *(1), , *(1.), =, =
=9
GRID, 600, , 0., 5.0, 0.
=, *(1), , *(1.), =, =
=9
GRID, 700, , 0., 6.0, 0.
=, *(1), , *(1.), =, =
=9
GRID, 800, , 0., 7.0, 0.
=, *(1), , *(1.), =, =
=9
GRID, 900, , 0., 8.0, 0.
=, *(1), , *(1.), =, =
=9
GRID, 1000, , 0., 9.0, 0.
=, *(1), , *(1.), =, =
=9
GRID, 1100, , 0., 10.0, 0.
=, *(1), , *(1.), =, =
=9
$
SPC1, 100, 123456 100, 101, 102, 103, 104, 105, +
+, 106, 107, 108, 109, 110, 200, 300, 400, +
+, 500, 600, 700, 800, 900, 1000, 1100
SPC1, 100, 246, 1101, 1102, 1103, 1104, 1105, 1106, +
+, 1107, 1108, 1109
SPC1, 100, 246, 1110
$
$...ELEMENT DEFINITION AND PROPERTIES:
$ (ELEMENTS GROUPED BY PID SINCE THICKNESS OF ALL ELEMENTS IN A GROUP
$ ARE TO BE AFFECTED BY A SINGLE DESIGN VARIABLE)
$
MAT1, 150, 1.0E7, , 0.3, 0.1
$...ELEMENT GROUP 1:
CQUAD4, 101, 1, 100, 101, 201, 200
=, *(100), =, *(100), *(100), *(100), *(100)
=8
CQUAD4, 102, 1, 101, 102, 202, 201
=, *(1), =, *(1), *(1), *(1), *(1),
=7
PSHELL, 1, 150, .08, 150
$...ELEMENT GROUP 2:
CQUAD4, 202, 2, 201, 202, 302, 301
=, *(100), =, *(100), *(100), *(100), *(100)
=7
CQUAD4, 203, 2, 202, 203, 303, 302
=, *(1), =, *(1), *(1), *(1), *(1),
=6

11-9
 EXAMPLE – MODAL FREQUENCY RESPONSE
OPTIMIZATION (Cont.)

PSHELL, 2, 150, .08, 150

$...ELEMENT GROUP 3:
CQUAD4, 303, 3, 302, 303, 403, 402
=, *(100), =, *(100), *(100), *(100), *(100)
=6
CQUAD4, 304, 3, 303, 304, 404, 403
=, *(1), =, *(1), *(1), *(1), *(1),
=5
PSHELL, 3, 150, .08, 150
$...ELEMENT GROUP 4:
CQUAD4, 404, 4, 403, 404, 504, 503
=, *(100), =, *(100), *(100), *(100), *(100)
=5
CQUAD4, 405, 4, 404, 405, 505, 504
=, *(1), =, *(1), *(1), *(1), *(1),
=4
PSHELL, 4, 150, .08, 150
$...ELEMENT GROUP 5:
CQUAD4, 505, 5, 504, 505, 605, 604
=, *(100), =, *(100), *(100), *(100), *(100)
=4
CQUAD4, 506, 5, 505, 506, 606, 605
=, *(1), =, *(1), *(1), *(1), *(1),
=3
PSHELL, 5, 150, .08, 150
$...ELEMENT GROUP 6:
CQUAD4, 606, 6, 605, 606, 706, 705
=, *(100), =, *(100), *(100), *(100), *(100)
=3
CQUAD4, 607, 6, 606, 607, 707, 706
=, *(1), =, *(1), *(1), *(1), *(1),
=2
PSHELL, 6, 150, .08, 150
$...ELEMENT GROUP 7:
CQUAD4, 707, 7, 706, 707, 807, 806
=, *(100), =, *(100), *(100), *(100), *(100)
=2
CQUAD4, 708, 7, 707, 708, 808, 807
=, *(1), =, *(1), *(1), *(1), *(1),
=1
PSHELL, 7, 150, .08, 150
$...ELEMENT GROUP 8:
CQUAD4, 808, 8, 807, 808, 908, 907
=, *(100), =, *(100), *(100), *(100), *(100)
=1
CQUAD4, 809, 8, 808, 809, 909, 908
=, *(1), =, *(1), *(1), *(1), *(1),
PSHELL, 8, 150, .08, 150
$...ELEMENT GROUP 9:
CQUAD4, 909, 9, 908, 909, 1009, 1008
=, *(100), =, *(100), *(100), *(100), *(100)
CQUAD4, 910, 9, 909, 910, 1010, 1009

11-10
 EXAMPLE – MODAL FREQUENCY RESPONSE
OPTIMIZATION (Cont.)

PSHELL, 9, 150, .08, 150

$...ELEMENT GROUP 10:
CQUAD4, 1010, 10, 1009, 1010, 1110, 1109
PSHELL, 10, 150, .08, 150
$
$...EIGENVALUE EXTRACTION INFORMATION - 15 RETAINED MODES
$
EIGRL 500 15 0
$
$...FREQUENCY DEPENDENT LOADING DATA: (OSCILLATORY PRESSURE LOAD)
$
RLOAD1, 700, 710, , , 800
LSEQ, 720, 710, 730
PLOAD2, 730, 1.0, 101, THRU, 110
PLOAD2, 730, 1.0, 201, THRU, 210
PLOAD2, 730, 1.0, 301, THRU, 310
PLOAD2, 730, 1.0, 401, THRU, 410
PLOAD2, 730, 1.0, 501, THRU, 510
PLOAD2, 730, 1.0, 601, THRU, 610
PLOAD2, 730, 1.0, 701, THRU, 710
PLOAD2, 730, 1.0, 801, THRU, 810
PLOAD2, 730, 1.0, 901, THRU, 910
PLOAD2, 730, 1.0, 1001, THRU, 1010
TABLED1,800, , , , , , , , +
+, 0.0, 1.0, 1.0E3, 1.0, ENDT
$
FREQ1 740 20. 1. 180
tabdmp1 2000
0.0 0.20 1000.0 0.20 endt
$
$-------------------------------------------------------------------------------
$ DESIGN MODEL
$-------------------------------------------------------------------------------
$
$...SPECIFY DESIGN VARIABLES, RELATE LINEARLY TO PLATE THICKNESS
DESVAR, 1, T1, .08, .001, 1.0
DESVAR, 2, T2, .08, .001, 1.0
DESVAR, 3, T3, .08, .001, 1.0
DESVAR, 4, T4, .08, .001, 1.0
DESVAR, 5, T5, .08, .001, 1.0
DESVAR, 6, T6, .08, .001, 1.0
DESVAR, 7, T7, .08, .001, 1.0
DESVAR, 8, T8, .08, .001, 1.0
DESVAR, 9, T9, .08, .001, 1.0
DESVAR, 10, T10, .08, .001, 1.0
$
$...RELATE DESIGN VARIABLES TO PLATE THICKNESSES
DVPREL1,101, PSHELL, 1, 4, .01, , , , +00
=, *(1), =, *(1), =, =, =, =, =, *(1)
=8
+00, 1, 1.0
*(1), *(1), =
=8

11-11
 EXAMPLE– MODAL FREQUENCY RESPONSE
OPTIMIZATION (Cont.)

$
DRESP1 201 VOLUME VOLUME
DRESP1 20 g1110L FRDISP 3 20.0 1110
= *(1) = = = = = *(1.0) =
=79
DRESP1 102 G1110H FRDISP 3 102.0 1110
= *(2) = = = = = *(2.0) =
=48
$
DRESP2 1 UZ2 1
DRESP1 20 21 22 23 24 25 26
27 28 29 30 31 32 33
34 35 36 37 38 39 40
41 42 43 44 45 46 47
48 49 50 51 52 53 54
55 56 57 58 59 60 61
62 63 64 65 66 67 68
69 70 71 72 73 74 75
76 77 78 79 80 81 82
83 84 85 86 87 88 89
90 91 92 93 94 95 96
97 98 99 100 102 104 106
108 110 112 114 116 118 120
122 124 126 128 130 132 134
136 138 140 142 144 146 148
150 152 154 156 158 160 162
164 166 168 170 172 174 176
178 180 182 184 186 188 190
192 194 196 198 200
$
DEQATN 1 UZ2(U20,u21,u22,u23,u24,u25,U26,U27,U28,U29,U30,
U31,U32,U33,U34,U35,U36,U37,U38,U39,U40,
U41,U42,U43,U44,U45,U46,U47,U48,U49,U50,
U51,U52,U53,U54,U55,U56,U57,U58,U59,U60,
U61,U62,U63,U64,U65,U66,U67,U68,U69,U70,
U71,U72,U73,U74,U75,U76,U77,U78,U79,U80,
U81,U82,U83,U84,U85,U86,U87,U88,U89,U90,
U91,U92,U93,U94,U95,U96,U97,U98,U99,U100,
U102,U104,U106,U108,U110,U112,U114,U116,U118,U120,
U122,U124,U126,U128,U130,U132,U134,U136,U138,U140,
U142,U144,U146,U148,U150,U152,U154,U156,U158,U160,
U162,U164,U166,U168,U170,U172,U174,U176,U178,U180,
U182,U184,U186,U188,U190,U192,U194,U196,U198,U200)
= u20**2 + u21**2 + U22**2 + U23**2+ U24**2 +
U25**2 + U26**2 + U27**2 + U28**2 + U29**2 +U30**2 +
U31**2 + U32**2 + U33**2 + U34**2 + U35**2 +
U36**2 + U37**2 + U38**2 + U39**2 + U40**2 +
U41**2 + U42**2 + U43**2 + U44**2 + U45**2 +
U46**2 + U47**2 + U48**2 + U49**2 + U50**2 +
U51**2 + U52**2 + U53**2 + U54**2 + U55**2 +
u56**2 + u57**2 + u58**2 + u59**2 + u60**2 +
U61**2 + U62**2 + U63**2 + U64**2 + U65**2 +
U66**2 + U67**2 + U68**2 + U69**2 + U70**2 +

11-12
 EXAMPLE – MODAL FREQUENCY RESPONSE
OPTIMIZATION (Cont.)

U71**2 + U72**2 + U73**2 + U74**2 + U75**2 +

U76**2 + U77**2 + U78**2 + U79**2 + U80**2 +
U81**2 + U82**2 + U83**2 + U84**2 + U85**2 +
u86**2 + u87**2 + u88**2 + u89**2 + u90**2 +
U91**2 + U92**2 + U93**2 + U94**2 + U95**2 +
u96**2 + u97**2 + u98**2 + u99**2 + u100**2 +
2.0*(u102**2 + u104**2 + u106**2 + u108**2 + u110**2 +
u112**2 + u114**2 + u116**2 + u118**2 + u120**2 +
u122**2 + u124**2 + u126**2 + u128**2 + u130**2 +
u132**2 + u134**2 + u136**2 + u138**2 + u140**2 +
u142**2 + u144**2 + u146**2 + u148**2 + u150**2 +
u152**2 + u154**2 + u156**2 + u158**2 + u160**2 +
u162**2 + u164**2 + u166**2 + u168**2 + u170**2 +
u172**2 + u174**2 + u176**2 + u178**2 + u180**2 +
u182**2 + u184**2 + u186**2 + u188**2 + u190**2 +
u192**2 + u194**2 + u196**2 + u198**2 + u200**2)
$
DCONSTR 10 201 7.99 8.01
doptprm desmax 40 p1 1 p2 8 conv1 0.01
$.......2.......3.......4.......5.......6.......7.......8.......9.......0
$
$param, optexit, 2
ENDDATA

11-13
 EXAMPLE – MODAL FREQUENCY RESPONSE OPTIMIZATION (Cont.)

Results:
***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************

(HARD CONVERGENCE ACHIEVED)

NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 8

NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 7

OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY

---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
11-14

CYCLE APPROXIMATE EXACT OF OF

NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------
INITIAL 2.301266E+02 -1.249123E-03

1 2.077626E+02 1.858892E+02 1.176688E-01 -1.247635E-03

2 1.743351E+02 1.609400E+02 8.323036E-02 -1.242462E-03

3 1.552931E+02 1.549932E+02 1.935096E-03 -1.248490E-03

4 1.483332E+02 1.423191E+02 4.225774E-02 -1.247516E-03

5 1.352799E+02 1.396561E+02 -3.133548E-02 -1.240314E-03

6 1.376002E+02 1.359294E+02 1.229173E-02 -1.239836E-03

7 1.338581E+02 1.346727E+02 -6.048443E-03 -1.231839E-03

---------------------------------------------------------------------------------------------------------------
 EXAMPLE – MODAL FREQUENCY RESPONSE OPTIMIZATION (Cont.)

DESIGN VARIABLE HISTORY

----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | T1 | 8.0000E-02 : 9.6000E-02 : 9.2260E-02 : 1.0819E-01 : 8.6551E-02 : 1.0386E-01 :
2 | 2 | T2 | 8.0000E-02 : 7.9020E-02 : 9.0223E-02 : 8.1839E-02 : 9.2760E-02 : 8.0022E-02 :
3 | 3 | T3 | 8.0000E-02 : 6.4695E-02 : 7.0780E-02 : 7.0625E-02 : 6.8610E-02 : 6.8198E-02 :
4 | 4 | T4 | 8.0000E-02 : 6.6232E-02 : 5.2986E-02 : 4.2548E-02 : 4.6342E-02 : 3.8022E-02 :
5 | 5 | T5 | 8.0000E-02 : 7.4244E-02 : 6.0564E-02 : 5.0883E-02 : 4.8050E-02 : 4.8722E-02 :
6 | 6 | T6 | 8.0000E-02 : 8.0033E-02 : 7.2680E-02 : 7.1341E-02 : 7.4902E-02 : 7.4154E-02 :
7 | 7 | T7 | 8.0000E-02 : 8.5749E-02 : 8.7154E-02 : 8.9976E-02 : 9.4979E-02 : 9.4669E-02 :
8 | 8 | T8 | 8.0000E-02 : 9.6000E-02 : 1.1429E-01 : 1.2160E-01 : 1.3566E-01 : 1.3450E-01 :
9 | 9 | T9 | 8.0000E-02 : 9.6000E-02 : 1.1520E-01 : 1.2648E-01 : 1.5178E-01 : 1.5422E-01 :
10 | 10 | T10 | 8.0000E-02 : 9.6000E-02 : 1.1520E-01 : 1.2155E-01 : 1.4586E-01 : 1.4772E-01 :
11-15

----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | 6 : 7 : 8 : 9 : 10 : 11 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | T1 | 9.2234E-02 : 1.0245E-01 :
2 | 2 | T2 | 8.9215E-02 : 7.8416E-02 :
3 | 3 | T3 | 6.6226E-02 : 6.4560E-02 :
4 | 4 | T4 | 3.6059E-02 : 3.2658E-02 :
5 | 5 | T5 | 5.5572E-02 : 5.8001E-02 :
6 | 6 | T6 | 7.6552E-02 : 7.8469E-02 :
7 | 7 | T7 | 9.6333E-02 : 9.7656E-02 :
8 | 8 | T8 | 1.3458E-01 : 1.3406E-01 :
9 | 9 | T9 | 1.5716E-01 : 1.5932E-01 :
10 | 10 | T10 | 1.4968E-01 : 1.5128E-01 :
*** USER INFORMATION MESSAGE 6464 (DOM12E)
RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 7.
 EXAMPLE – MODAL FREQUENCY RESPONSE
OPTIMIZATION (Cont.)

Frequency-Dependent Displacement Results:

3.622 @ 58 Hz

2.929 @ 52 Hz

11-16
 EXAMPLE – MODAL FREQUENCY RESPONSE
OPTIMIZATION (Cont.)

Final Thickness Distribution:

Thickness

1 2 3 4 5 6 7 8 9 10

11-17
 EXAMPLE – ACOUSTIC OPTIMIZATION

This example illustrates minimization of acoustic response peaks

over a given frequency range. (The example is also presented in the
MSC/NASTRAN Design Sensitivity and Optimization User’s Guide,
Section 7.10.)

The figure below shows an acoustic box with an acoustic source and
a transducer located on opposite walls of the box (prop 3 ends). We
would like to modify the box plate thicknesses to minimize the
response peaks within the fluid (air). This is to be done without
changing the total weight of the box.

11-18
 EXAMPLE – ACOUSTIC OPTIMIZATION (Cont.)

One could use the following technique for minimizing response

peaks:

P(f)
c1 c2 c3
βx

f
f1 f2 f3

minimize βx

subject to P ( f1 ) – βx ≤ 0
P ( f2 ) – βx ≤ 0
P ( f3 ) – βx ≤ 0

One such constraint is written for every response of interest.

β can only be minimized if the peak pressures are lowered.

11-19
 EXAMPLE – ACOUSTIC OPTIMIZATION (Cont.)

Input File:

ID MSC-XL, MSC-NASTRAN
TIME 100
SOL 200 $ modal frequency response
CEND
SUBTITLE = acoustic and structural elements
LABEL = boxae1.dat
$
$ analysis case control:
set 20 = 11280
echo = sort(param,eigc,eigrl,freq,desvar,dconstr,dresp1,dresp2,deqatn,
dvprel1)
spc = 1
DISP(phase) = 20
method(structure) = 20
method(fluid) = 30
cmethod = 10
frequency = 200
dload = 100
partn = 20
$ optimization case control:
ANALYSIS = MFREQ
DESGLB = 5
DESSUB = 10
DESOBJ = 100
$
BEGIN BULK
$
EIGRL 20 200. max
eigrl ,30, 15., 155., 9, 0, , 105., max
eigc,10,clan,max,,,,
,15.,300.,,,,,4
$
$ sound pressure level
param,rms,yes
$ reference pressure for dB and dBA
param,prefdb,2.-5
$
PARAM AUTOSPC no
$
$ fluid/structure interface
acmodl,diff , , , ,0.01
$
$ structural damping:
param,g,0.02
$

11-20
 EXAMPLE – ACOUSTIC OPTIMIZATION (Cont.)

$-------2-------3-------4-------5-------6-------7-------8-------9-------10----
rload1,100,101,,,102
darea,101,1288,3,100.
tabled1,102
,0.,1.,1000.,1.,endt
$freq2, 200, 1.0, 200.0, 50
$-------2-------3-------4-------5-------6-------7-------8-------9-------10----
freq 200 40. 50. 60. 70. 80. 90. 95.
100. 105. 110. 120. 130. 140. 150. 160.
170. 180. 190. 200. 97.5 102.5
$-------2-------3-------4-------5-------6-------7-------8-------9-------10----
$
$
$ Define the Design Variables:
$
$ESVAR, ID, LABEL, XINIT, XLB, XUB, DELXV
DESVAR, 1, P1, 0.02493,0.0001, 1.
DESVAR, 2, P2, 0.01953,0.0001, 1.
DESVAR, 4, P4, 0.02047,0.0001, 1.
DESVAR, 5, P5, 0.02596,0.0001, 1.
DESVAR, 6, P6, 0.02175,0.0001, 1.
DESVAR, 7, P7, 0.02426,0.0001, 1.
DESVAR, 8, BETA 1.0, 0.001
$
$ Relate the Design Variables to changes in plate thicknesses:
$
$VPREL1,ID, TYPE, PID, FID, PMIN, PMAX, C0, , +
$+, DVID1, COEF1, DVID2, COEF2, ...
$
DVPREL1,1, PSHELL, 1, 4, 0.0001, , , , +
+, 1, 1.
DVPREL1,2, PSHELL, 2, 4, 0.0001, , , , +
+, 2, 1.
DVPREL1,4, PSHELL, 4, 4, 0.0001, , , , +
+, 4, 1.
DVPREL1,5, PSHELL, 5, 4, 0.0001, , , , +
+, 5, 1.
DVPREL1,6, PSHELL, 6, 4, 0.0001, , , , +
+, 6, 1.
DVPREL1,7, PSHELL, 7, 4, 0.0001, , , , +
+, 7, 1.
$
$ Define the synthetic Objective as a function of X8:
$
$RESP2, ID, LABEL, EQID, REGION, , , , , +
$+, DESVAR, DVID1, DVID2, ..., , , , , +
DRESP2 100 BETA 100
DESVAR 8
DEQATN 100 OBJ(BETA) = 10000.0 * BETA
$

11-21
 EXAMPLE – ACOUSTIC OPTIMIZATION (Cont.)

$ Define the constraint on weight (weight budget for optimization):

$
$RESP1, ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
$
DRESP1 2 WEIGHT WEIGHT
$CONSTR,DCID, RID, LALLOW, UALLOW
DCONSTR 5 2 2890. 2910.
$
$ Define the constraints on acoustic sound pressure levels
$ (one for each forcing frequency):
$
DRESP1 1 DRUCK FRDISP 1 11280
DRESP2 11 PRESBET 10
DESVAR 8
DRESP1 1
DEQATN 10 F(BETA,PRES) = 100.0 * BETA - PRES + 1000.
DCONSTR 10 11 1000.
$
$ Override miscellaneous Optimization Parameters:
$
DOPTPRM DESMAX 20 P1 1 P2 15 CONV1 1E-6
$
$
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
$ structural model
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
$
$
$ THIS SECTION CONTAINS THE PROPERTY AND MATERIAL BULK DATA ENTRIES
$
PSHELL 3 1 .100 1
$
PSHELL 1 1 .02493 1
PSHELL 2 1 .01953 1
pshell 4 1 .02047 1
pshell 5 1 .02596 1
pshell 6 1 .02175 1
pshell 7 1 .02426 1
$
$
$ THIS SECTION CONTAINS BULK DATA FOR SUPERELEMENT 0
$
$
GRID 1 0.0 0.0 0.0
GRID 2 2. 0.0 0.0
GRID 3 2. 0.0 1.
. . . . .
. . . . .
. . . . .

11-22
 EXAMPLE – ACOUSTIC OPTIMIZATION (Cont.)

GRID 1293 1.5 .5 1.

GRID 1294 1.6 .5 1.
GRID 1295 1.7 .5 1.
$
CQUAD4 1 1 1 9 29 28
CQUAD4 2 1 9 10 30 29
. . . . . . .
. . . . . . .
. . . . . . .
CQUAD4 999 3 1139 1140 275 296
CQUAD4 1000 3 1140 897 5 275
$
$ THIS SECTION CONTAINS THE LOADS,CONSTRAINTS, AND CONTROL BULK DATA ENTRIES
$
$
SPC 1 1 123 0.0
SPC 1 2 123 0.0
. . . . .
. . . . .
. . . . .
SPC 1 986 4
SPC 1 987 4
SPC 1 975 4
$
$
MAT1 1 2.+11 .3 7600.
$
$-------2-------3-------4-------5-------6-------7-------8-------9-------10----
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
$ acoustic model
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
$
$ THIS SECTION CONTAINS BULK DATA FOR SUPERELEMENT 0
$
$
$-------2-------3-------4-------5-------6-------7-------8-------9-------10----
GRID 10001 0.0 0.0 0.0 -1
GRID 10002 2. 0.0 0.0 -1
GRID 10003 2. 0.0 1. -1
. . . . . .
. . . . . .
. . . . . .
GRID 12540 2. 1. .2 -1
GRID 12541 2. 1. .1 -1
$
CHEXA 10001100 10004 10126 10127 10009 10018 10137 +
+ 10138 10019

11-23
 EXAMPLE – ACOUSTIC OPTIMIZATION (Cont.)

CHEXA 10002100 10009 10127 10128 10010 10019 10138 +

+ 10139 10020
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
CHEXA 11999100 12411 12530 12531 12412 12422 12540 +
+ 12541 12423
CHEXA 12000100 12412 12531 12532 12413 12423 12541 +
+ 10006 12424
$
psolid,100,100,,,,1,pfluid
$
mat10,100,,1.293,200.
$
ENDDATA

11-24
 EXAMPLE – ACOUSTIC OPTIMIZATION (Cont.)

Results:
***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************

(HARD CONVERGENCE ACHIEVED)

(SOFT CONVERGENCE ACHIEVED)

NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 16

NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 15

OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY

---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
11-25

CYCLE APPROXIMATE EXACT OF OF

NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------
INITIAL 1.000000E+04 1.106849E-01
1 5.334501E+03 5.334501E+03 0.000000E+00 4.063804E-01
2 1.246205E+03 1.246205E+03 0.000000E+00 2.724397E-01
3 1.809879E+03 1.809879E+03 0.000000E+00 1.475203E-01
4 2.076569E+03 2.076569E+03 0.000000E+00 1.426204E-01
5 2.075635E+03 2.075635E+03 0.000000E+00 9.813190E-02
6 2.076403E+03 2.076403E+03 0.000000E+00 4.730811E-02
7 2.103385E+03 2.103385E+03 0.000000E+00 6.921270E-02
8 2.104906E+03 2.104906E+03 0.000000E+00 6.877228E-02
9 2.133196E+03 2.133196E+03 0.000000E+00 9.842414E-02
10 2.478105E+03 2.478105E+03 0.000000E+00 2.656518E-02
11 1.858579E+03 1.858579E+03 0.000000E+00 8.776490E-03
12 1.393935E+03 1.393935E+03 0.000000E+00 5.736145E-03
13 1.424619E+03 1.424619E+03 0.000000E+00 2.339111E-03
14 1.068465E+03 1.068465E+03 0.000000E+00 1.823486E-03
15 1.068465E+03 1.068465E+03 0.000000E+00 1.823486E-03
---------------------------------------------------------------------------------------------------------------
 EXAMPLE – ACOUSTIC OPTIMIZATION (Cont.)

DESIGN VARIABLE HISTORY

----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | P1 | 2.4930E-02 : 3.4858E-02 : 3.2397E-02 : 2.8322E-02 : 3.4406E-02 : 3.5523E-02 :
2 | 2 | P2 | 1.9530E-02 : 2.6421E-02 : 2.6187E-02 : 3.3966E-02 : 3.3395E-02 : 3.2716E-02 :
3 | 4 | P4 | 2.0470E-02 : 1.0487E-02 : 1.2030E-02 : 1.3776E-02 : 7.8943E-03 : 6.4859E-03 :
4 | 5 | P5 | 2.5960E-02 : 2.0480E-02 : 2.1393E-02 : 1.3477E-02 : 2.3186E-02 : 2.2668E-02 :
5 | 6 | P6 | 2.1750E-02 : 1.8127E-02 : 2.1821E-02 : 2.3308E-02 : 2.1010E-02 : 1.9540E-02 :
6 | 7 | P7 | 2.4260E-02 : 1.9210E-02 : 1.5891E-02 : 1.4294E-02 : 9.1583E-03 : 1.0095E-02 :
7 | 8 | BETA | 1.0000E+00 : 5.3345E-01 : 1.2462E-01 : 1.8099E-01 : 2.0766E-01 : 2.0756E-01 :

----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | 6 : 7 : 8 : 9 : 10 : 11 :
11-26

----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | P1 | 3.4322E-02 : 2.9154E-02 : 2.4363E-02 : 2.7226E-02 : 2.4670E-02 : 2.4619E-02 :
2 | 2 | P2 | 3.1484E-02 : 2.6989E-02 : 3.1301E-02 : 2.5696E-02 : 2.3196E-02 : 2.3115E-02 :
3 | 4 | P4 | 5.6109E-03 : 6.5327E-03 : 5.8686E-03 : 5.4315E-03 : 4.3781E-03 : 3.8628E-03 :
4 | 5 | P5 | 2.4430E-02 : 3.1121E-02 : 3.3386E-02 : 3.8369E-02 : 4.0770E-02 : 4.0748E-02 :
5 | 6 | P6 | 1.8545E-02 : 1.7878E-02 : 1.7679E-02 : 1.4659E-02 : 1.6339E-02 : 1.6324E-02 :
6 | 7 | P7 | 1.3480E-02 : 1.9997E-02 : 2.1885E-02 : 2.5364E-02 : 2.7816E-02 : 2.7732E-02 :
7 | 8 | BETA | 2.0764E-01 : 2.1034E-01 : 2.1049E-01 : 2.1332E-01 : 2.4781E-01 : 1.8586E-01 :

----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | 12 : 13 : 14 : 15 : 16 : 17 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | P1 | 2.4258E-02 : 2.6005E-02 : 2.6021E-02 : 2.6021E-02 :
2 | 2 | P2 | 2.3058E-02 : 2.1841E-02 : 2.1834E-02 : 2.1834E-02 :
3 | 4 | P4 | 3.3257E-03 : 2.7535E-03 : 2.4141E-03 : 2.4141E-03 :
4 | 5 | P5 | 4.3248E-02 : 4.3566E-02 : 4.3587E-02 : 4.3587E-02 :
5 | 6 | P6 | 1.4020E-02 : 1.1520E-02 : 1.1632E-02 : 1.1632E-02 :
6 | 7 | P7 | 2.7829E-02 : 2.9923E-02 : 2.9964E-02 : 2.9964E-02 :
7 | 8 | BETA | 1.3939E-01 : 1.4246E-01 : 1.0685E-01 : 1.0685E-01 :

*** USER INFORMATION MESSAGE 6464 (DOM12E)

RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 15.
 EXAMPLE – ACOUSTIC OPTIMIZATION (Cont.)

Sound Pressure Levels: Initial and Final Distributions:

140.5 dB @ 100 Hz

115.9 dB @ 100 Hz

11-27
 EXAMPLE: ACOUSTIC OPTIMIZATION (Cont.)

Structural and Fluid Eigenfrequencies:

Structural
Eigenfrequencies
Mode Fluid
No. Initial Final Eigenfrequencies
1 75.99 7.861680E+01 50.01
2 95.29 8.777968E+01 100.41
3 104.16 9.964240E+01 100.41
4 130.78 1.187981E+02 100.41
5 133.35 1.262226E+02 112.19
6 143.74 1.309330E+02 112.19
7 153.16 1.448834E+02 142.00
8 173.69 1.457052E+02 142.00
9 – – 142.00

11-28
 SECTION 12

SUPERELEMENT OPTIMIZATION

GENERAL CONSIDERATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-1

CASE CONTROL SPECIFICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-3

EXAMPLE – TWENTY FIVE BAR TRUSS, SUPERELEMENT OPTIMIZATION . . . 12-5

 GENERAL CONSIDERATIONS 12

Superelements supported in SOL 200 are for old formulation only.

Superelement optimization allows the design model to span

superelement boundaries.

Advantages of using superelements in design optimization:

• Large problems can be broken into smaller, more manageable

pieces

• Incremental or concurrent processing using split database

technique (multi-master)

• Partial redesign requires partial solution

• Provides effective use of resources

Design variables:

• For sizing: can be related to property groups across superelements

• For shape: limited shape optimization support is available in

Version 68 – only the Manual Grid Variation method can be used in
a superelement model

12-1
 GENERAL CONSIDERATIONS (Cont.)

Design responses:

All solution 200 analysis disciplines are supported

• Weight and volume responses can be from any, or all,

superelements

• First level responses (DRESP1) can be from any superelement

(except eigenvalues, which are for the residual only)

• Second level responses (DRESP2) are supported, although the

argument list cannot span superelements or load cases

Supported Superelements:

• Primary and Image superelements are supported

• Image superelements reference the design variables and

responses of the Primary superelement

• Although External superelements can be used in the analysis, they

cannot be a part of the design model (they cannot be written as
functions of design variables, or contribute design responses)

12-2
 CASE CONTROL SPECIFICATION

Single Discipline:

• Condensed (SUPER=ALL)

• Expanded

Multidisciplinary

• Condensed (SUPER=ALL) case control is recommended

Example: Condensed Case Control

• Recommended Case Control structure

ID MSC, D200SE3 $
SOL 200 $ OPTIMIZATION
CEND
TITLE=STATIC ANALYSIS OF A 25-BAR TRUSS D200SE3
DISP=ALL
SPC=100
ANALYSIS=STATICS
SUPER=ALL
DESOBJ=15
SUBCASE 1
DESSUB=1
LABEL=LOAD CONDITION 1
LOAD=300
SUBCASE 2
DESSUB=2
LABEL=LOAD CONDITION 2
LOAD=310
BEGIN BULK

12-3
 CASE CONTROL SPECIFICATION (Cont.)

Example: Expanded Case Control

• Design constraints called out for each superelement and load case

ANALYSIS=STATICS
DESOBJ=15
SUBCASE 10
DESSUB=1
SUPER=1,1 $ SUPERELEMENT 1, LOAD SEQ 1
LOAD=300
SUBCASE 20
DESSUB=2
SUPER=1,2 $ SUPERELEMENT 1, LOAD SEQ 2
LOAD=310
SUBCASE 30
DESSUB=3
LOAD=300
SUBCASE 40
DESSUB=4
LOAD=310
BEGIN BULK

12-4
 EXAMPLE – TWENTY FIVE BAR TRUSS,
SUPERELEMENT OPTIMIZATION

This example illustrates the use of superelement optimization for a

simple case. (The example is also presented in the MSC/NASTRAN
Design Sensitivity and Optimization User’s Guide, Section 7.9)

The figure below shows a twenty-five bar truss structure that has
been partitioned into an upstream superelement 1, and a residual
structure, superelement 0. (Grids 1 and 2 have been defined as
interior to superelement 1.)

2
Superelement 1

6 3

5 4

10
7 Y
Superelement 0

X 8
9

Analysis Model Description

Three-dimensional truss
Symmetric with respect to x-y plane and y-z plane
Weight density = 0.1 lbs/in3
Materials: E = 1.0E7 psi
Two distinct loading conditions

12-5
 EXAMPLE: TWENTY FIVE BAR TRUSS,
SUPERELEMENT OPTIMIZATION (Cont.)

Design Model Description

Minimization of structural weight

Design variables: Cross-sectional areas linked to eight
independent design variables
Constraints: Allowable stress: Tensile = 40,000 psi
Compressive = –40,000 psi
Displacement constraints: ±0.35 inches at
grid 1 and 2 for all translational degrees of
freedom
Euler buckling constraints for compressive
members assuming tubular section diameter
to thickness ratio of approximately 100

The Euler Buckling stress for a pin-connected truss member can be

written as:

1  π Ei
Pb 2
σ b = ------- = ----  – ------------
A
L 
A 2

Assuming tubular section elements with a diameter to thickness ratio

of 100,

D
A ≅ πDt , where ---- ≅ 100
t

So the area moment of inertia can be approximated as

2 2
πDt ( D + t )
I ≅ ----------------------------------
8

12-6
 EXAMPLE – TWENTY FIVE BAR TRUSS,
SUPERELEMENT OPTIMIZATION (Cont.)

Buckling stresses in terms of the area design variables can be written

approximately as:

2
AE π ( 100 + 1 ) D
σ b = – --------∗ ------------------------------- where ---- = 100
2 8∗100 t
L

A feasible buckling condition is:

σ ≥ σb

Which can be normalized as:

σ
------ ≤ 1
σb

12-7
 EXAMPLE – TWENTY FIVE BAR TRUSS,
SUPERELEMENT OPTIMIZATION (Cont.)

Input data:

ID MSC, D200X3
TIME 10
SOL 200 $ OPTIMIZATION
CEND
super = all
ECHO = UNSORT
OLOAD = ALL
DISP = ALL
SPCFORCE = ALL
ELFORCE = ALL
STRESS = ALL
SPC = 100
ANALYSIS = STATICS $
DESOBJ(MIN) = 15 $ OBJECTIVE FUNCTION DEFINITION
DESSUB = 12 $ CONSTRAINT DEFININITION
SUBCASE 1
LABEL = LOAD CONDITION 1
LOAD = 300
SUBCASE 2
LABEL = LOAD CONDITION 2
LOAD = 310
BEGIN BULK
$
$-------------------------------------------------------------------------
$ ANALYSIS MODEL
$-------------------------------------------------------------------------
$
GRDSET, , , , , , , 456
SESET, 1, 1, 2
MAT1, 1, 10.0E6, , , 0.1, , , , +M1
+M1, 25000., 25000.
SPC1, 100, 123, 7, THRU, 10
GRID, 1 , , -37.5, 0.0, 200.0
GRID, 2 , , 37.5, 0.0, 200.0
GRID, 3 , , -37.5, 37.5, 100.0
GRID, 4 , , 37.5, 37.5, 100.0
GRID, 5 , , 37.5, -37.5, 100.0
GRID, 6 , , -37.5, -37.5, 100.0
GRID, 7 , , -100.0, 100.0, 0.0
GRID, 8 , , 100.0, 100.0, 0.0
GRID, 9 , , 100.0, -100.0, 0.0
GRID, 10, , -100.0, -100.0, 0.0
CROD, 1 , 1, 1, 2
CROD, 2 , 2, 1, 4

12-8
 EXAMPLE – TWENTY FIVE BAR TRUSS,
SUPERELEMENT OPTIMIZATION (Cont.)

CROD, 3 , 2, 2, 3
CROD, 4 , 2, 1, 5
CROD, 5 , 2, 2, 6
CROD, 6 , 3, 2, 4
CROD, 7 , 3, 2, 5
CROD, 8 , 3, 1, 3
CROD, 9 , 3, 1, 6
CROD, 10, 4, 3, 6
CROD, 11, 4, 4, 5
CROD, 12, 5, 3, 4
CROD, 13, 5, 5, 6
CROD, 14, 6, 3, 10
CROD, 15, 6, 6, 7
CROD, 16, 6, 4, 9
CROD, 17, 6, 5, 8
CROD, 18, 7, 4, 7
CROD, 19, 7, 3, 8
CROD, 20, 7, 5, 10
CROD, 21, 7, 6, 9
CROD, 22, 8, 6, 10
CROD, 23, 8, 3, 7
CROD, 24, 8, 5, 9
CROD, 25, 8, 4, 8
$
PROD, 1, 1, 2.0, 0.0
PROD, 2, 1, 2.0, 0.0
PROD, 3, 1, 2.0, 0.0
PROD, 4, 1, 2.0, 0.0
PROD, 5, 1, 2.0, 0.0
PROD, 6, 1, 2.0, 0.0
PROD, 7, 1, 2.0, 0.0
PROD, 8, 1, 2.0, 0.0
$
FORCE, 300, 1, , 1.0, 1000., 10000.,-5000.
FORCE, 300, 2, , 1.0, 0., 10000.,-5000.
FORCE, 300, 3, , 1.0, 500., 0., 0.
FORCE, 300, 6, , 1.0, 500., 0., 0.
FORCE, 310, 1, , 1.0, 0., 20000.,-5000.
FORCE, 310, 2, , 1.0, 0., -20000.,-5000.
$
$-----------------------------------------------------------------------------
$ DESIGN MODEL
$-----------------------------------------------------------------------------
$
$...Define the design variables
$DESVAR,ID, LABEL, XINIT, XLB, XUB, DELXV
$

12-9
 EXAMPLE – TWENTY FIVE BAR TRUSS,
SUPERELEMENT OPTIMIZATION (Cont.)

DESVAR, 1, X1, 2.0, 0.01, 100.0

DESVAR, 2, X2, 2.0, 0.01, 100.0
DESVAR, 3, X3, 2.0, 0.01, 100.0
DESVAR, 4, X4, 2.0, 0.01, 100.0
DESVAR, 5, X5, 2.0, 0.01, 100.0
DESVAR, 6, X6, 2.0, 0.01, 100.0
DESVAR, 7, X7, 2.0, 0.01, 100.0
DESVAR, 8, X8, 2.0, 0.01, 100.0
$
$...Relate the design variables to the analysis model properties
$DVPREL1,ID, TYPE, PID, FID, PMIN, PMAX, C0, , +
$+, DVIDD1, COEF1, DVID2, COEF2, ...
$
DVPREL1,1, PROD, 1, 4, , , , , +DP1
+DP1 ,1, 1.0
DVPREL1,2, PROD, 2, 4, , , , , +DP2
+DP2 ,2, 1.0
DVPREL1,3, PROD, 3, 4, , , , , +DP3
+DP3 ,3, 1.0
DVPREL1,4, PROD, 4, 4, , , , , +DP4
+DP4 ,4, 1.0
DVPREL1,5, PROD, 5, 4, , , , , +DP5
+DP5 ,5, 1.0
DVPREL1,6, PROD, 6, 4, , , , , +DP6
+DP6 ,6, 1.0
DVPREL1,7, PROD, 7, 4, , , , , +DP7
+DP7 ,7, 1.0
DVPREL1,8, PROD, 8, 4, , , , , +DP8
+DP8 ,8, 1.0
$
$...Identify the responses to be used in the design model
$DRESP1,ID, LABEL, RTYPE, PTYPE, REGION, ATTA, ATTB, ATT1, +
$+, ATT2, ...
$
DRESP1, 1 , S1, STRESS, PROD, , 2, , 1
DRESP1, 2 , S2, STRESS, PROD, , 2, , 2
DRESP1, 3 , S3, STRESS, PROD, , 2, , 3
DRESP1, 4 , S4, STRESS, PROD, , 2, , 4
DRESP1, 5 , S5, STRESS, PROD, , 2, , 5
DRESP1, 6 , S6, STRESS, PROD, , 2, , 6
DRESP1, 7 , S7, STRESS, PROD, , 2, , 7
DRESP1, 8 , S8, STRESS, PROD, , 2, , 8
DRESP1, 9 , D1, DISP , , , 1, , 1
DRESP1, 10, D2, DISP , , , 2, , 1
DRESP1, 11, D3, DISP , , , 3, , 1
DRESP1, 12, D4, DISP , , , 1, , 2
DRESP1, 13, D5, DISP , , , 2, , 2
DRESP1, 14, D6, DISP , , , 3, , 2

12-10
 EXAMPLE – TWENTY FIVE BAR TRUSS,
SUPERELEMENT OPTIMIZATION (Cont.)

DRESP1, 15, W , WEIGHT, , , , , ALL

$
$...Formulate the second level responses (here, simple Euler buckling)
$DRESP2,ID, LABEL, EQID, REGION, , , , , +
$+, DESVAR, DVID1, DVID2, ..., , , , , +
$+, DTABLE, LABEL1, LABEL2, ..., , , , , +
$+, DRESP1, NR1, NR2, ..., , , , , +
$+, DNODE, NID1, DIR1, NID2, DIR2, ..., , , +
$
DRESP2, 16, SC1, 1, , , , , , +DR11
+DR11 , DESVAR, 1, , , , , , , +DR12
+DR12 , DTABLE, L1, , , , , , , +DR13
+DR13 , DRESP1, 1
DRESP2, 17, SC2, 1, , , , , , +DR21
+DR21 , DESVAR, 2, , , , , , , +DR22
+DR22 , DTABLE, L2, , , , , , , +DR23
+DR23 , DRESP1, 2
DRESP2, 18, SC3, 1, , , , , , +DR31
+DR31 , DESVAR, 3, , , , , , , +DR32
+DR32 , DTABLE, L3, , , , , , , +DR33
+DR33 , DRESP1, 3
DRESP2, 19, SC4, 1, , , , , , +DR41
+DR41 , DESVAR, 4, , , , , , , +DR42
+DR42 , DTABLE, L4, , , , , , , +DR43
+DR43 , DRESP1, 4
DRESP2, 20, SC5, 1, , , , , , +DR51
+DR51 , DESVAR, 5, , , , , , , +DR52
+DR52 , DTABLE, L5, , , , , , , +DR53
+DR53 , DRESP1, 5
DRESP2, 21, SC6, 1, , , , , , +DR61
+DR61 , DESVAR, 6, , , , , , , +DR62
+DR62 , DTABLE, L6, , , , , , , +DR63
+DR63 , DRESP1, 6
DRESP2, 22, SC7, 1, , , , , , +DR71
+DR71 , DESVAR, 7, , , , , , , +DR72
+DR72 , DTABLE, L7, , , , , , , +DR73
+DR73 , DRESP1, 7
DRESP2, 23, SC8, 1, , , , , , +DR81
+DR81 , DESVAR, 8, , , , , , , +DR82
+DR82 , DTABLE, L8, , , , , , , +DR83
+DR83 , DRESP1, 8
$
$...Equations used to define second level responses (note: fixed-field form!!)
$DEQATN EQUID F() = ...
$
DEQATN 1 F(A,RL,S) = -S*RL**2/(A*1.E7*39.274)
$

12-11
 EXAMPLE – TWENTY FIVE BAR TRUSS,
SUPERELEMENT OPTIMIZATION (Cont.)

$...Table constants
$DTABLE,LABEL1, VALUE1, LABEL2, VALUE2, LABEL3, VALUE3, LABEL4, VALUE4 +
$+, LABEL5, VALUE5, ...
$
DTABLE, L1, 75.00, L2, 130.50, L3, 106.80,L4, 75.00, +
+, L5, 75.00, L6, 181.14,L7, 181.14, L8, 133.46
$
$...Define the design constraints
$CONSTR,DCID, RID, LALLOW, UALLOW
$
DCONSTR,10, 1 , -40000.,40000.
DCONSTR,10, 2 , -40000.,40000.
DCONSTR,10, 3 , -40000.,40000.
DCONSTR,10, 4 , -40000.,40000.
DCONSTR,10, 5 , -40000.,40000.
DCONSTR,10, 6 , -40000.,40000.
DCONSTR,10, 7 , -40000.,40000.
DCONSTR,10, 8 , -40000.,40000.
DCONSTR,10, 9 , -0.35 ,0.35
DCONSTR,10, 10, -0.35 ,0.35
DCONSTR,10, 11, -0.35 ,0.35
DCONSTR,10, 12, -0.35 ,0.35
DCONSTR,10, 13, -0.35 ,0.35
DCONSTR,10, 14, -0.35 ,0.35
$
DCONSTR,11, 16, -1.0E10, 1.0
DCONSTR,11, 17, -1.0E10, 1.0
DCONSTR,11, 18, -1.0E10, 1.0
DCONSTR,11, 19, -1.0E10, 1.0
DCONSTR,11, 20, -1.0E10, 1.0
DCONSTR,11, 21, -1.0E10, 1.0
DCONSTR,11, 22, -1.0E10, 1.0
DCONSTR,11, 23, -1.0E10, 1.0
$
$...Combine the two constraint sets
$ (equivalent to just putting all into the same set to begin with)
DCONADD,12, 10, 11
$
$...Override optimization parameter defaults:
$
DOPTPRM,IPRINT, 3, DESMAX, 15, DELP, 0.5, p1, 1, +
+, p2, 15
ENDDATA
$.......2.......3.......4.......5.......6.......7.......8.......9.......0

12-12
 EXAMPLE – TWENTY FIVE BAR TRUSS, SUPERELEMENT
OPTIMIZATION (Cont.)

Results:

***** NORMAL CONVERGENCE CRITERIA SATISFIED ***** (HARD CONVERGENCE DECISION LOGIC)

**************************************************************************************
CONVERGENCE ACHIEVED BASED ON THE FOLLOWING CRITERIA
(HARD CONVERGENCE DECISION LOGIC)

RELATIVE CHANGE IN OBJECTIVE 5.2768E-05 MUST BE LESS THAN 1.0000E-03

OR ABSOLUTE CHANGE IN OBJECTIVE 2.8809E-02 MUST BE LESS THAN 1.0000E-02
--- AND ---
12-13

MAXIMUM CONSTRAINT VALUE 3.8660E-04 MUST BE LESS THAN 5.0000E-03

(CONVERGENCE TO A FEASIBLE DESIGN)
--- OR ---
MAXIMUM OF RELATIVE PROP. CHANGES 3.8638E-02 MUST BE LESS THAN 1.0000E-03
AND MAXIMUM OF RELATIVE D.V. CHANGES 3.8638E-02 MUST BE LESS THAN 1.0000E-03
(CONVERGENCE TO A BEST COMPROMISE INFEASIBLE DESIGN)
**************************************************************************************
 EXAMPLE: TWENTY FIVE BAR TRUSS, SUPERELEMENT
OPTIMIZATION (Cont.)

***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************

(HARD CONVERGENCE ACHIEVED)

NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 9

NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 8

OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY

---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
12-14

CYCLE APPROXIMATE EXACT OF OF

NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------

INITIAL 6.614414E+02 1.102773E-01

1 5.985143E+02 5.985219E+02 -1.254311E-05 2.294609E-03

2 5.763787E+02 5.763831E+02 -7.730217E-06 8.119856E-04
3 5.605763E+02 5.605792E+02 -5.117301E-06 3.697191E-04
4 5.513699E+02 5.513676E+02 4.206515E-06 2.951452E-03
5 5.484974E+02 5.484994E+02 -3.560852E-06 3.337860E-05
6 5.465300E+02 5.465295E+02 7.817438E-07 1.285757E-04
7 5.459542E+02 5.459532E+02 1.788729E-06 3.235681E-06
8 5.459244E+02 5.459244E+02 0.000000E+00 3.865957E-04
---------------------------------------------------------------------------------------------------------------
 EXAMPLE – TWENTY FIVE BAR TRUSS, SUPERELEMENT
OPTIMIZATION (Cont.)

DESIGN VARIABLE HISTORY

---------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
---------------------------------------------------------------------------------------------------------------------------
1 | 1 | X1 | 2.0000E+00 : 1.0000E+00 : 9.3000E-01 : 4.6500E-01 : 2.3250E-01 : 1.1625E-01 :
2 | 2 | X2 | 2.0000E+00 : 2.0461E+00 : 2.0175E+00 : 2.0560E+00 : 2.0416E+00 : 2.0378E+00 :
3 | 3 | X3 | 2.0000E+00 : 3.0000E+00 : 3.1355E+00 : 3.0361E+00 : 3.0550E+00 : 3.0211E+00 :
4 | 4 | X4 | 2.0000E+00 : 1.0000E+00 : 8.6000E-01 : 4.3000E-01 : 2.1500E-01 : 1.0750E-01 :
5 | 5 | X5 | 2.0000E+00 : 1.0000E+00 : 8.6240E-01 : 4.3120E-01 : 2.1558E-01 : 1.0774E-01 :
6 | 6 | X6 | 2.0000E+00 : 1.0000E+00 : 6.1363E-01 : 7.1806E-01 : 6.8686E-01 : 6.6251E-01 :
7 | 7 | X7 | 2.0000E+00 : 1.5944E+00 : 1.5890E+00 : 1.5927E+00 : 1.5991E+00 : 1.6201E+00 :
8 | 8 | X8 | 2.0000E+00 : 2.5863E+00 : 2.7106E+00 : 2.6171E+00 : 2.6308E+00 : 2.6893E+00 :
12-15

---------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | 6 : 7 : 8 : 9 : 10 : 11 :
---------------------------------------------------------------------------------------------------------------------------
1 | 1 | X1 | 5.8125E-02 : 2.9063E-02 : 2.8529E-02 :
2 | 2 | X2 | 2.0443E+00 : 2.0427E+00 : 2.0427E+00 :
3 | 3 | X3 | 3.0237E+00 : 3.0059E+00 : 3.0061E+00 :
4 | 4 | X4 | 5.3750E-02 : 2.6875E-02 : 2.5837E-02 :
5 | 5 | X5 | 5.3789E-02 : 5.4376E-02 : 5.3374E-02 :
6 | 6 | X6 | 6.8365E-01 : 6.8306E-01 : 6.8303E-01 :
7 | 7 | X7 | 1.6118E+00 : 1.6216E+00 : 1.6215E+00 :
8 | 8 | X8 | 2.6649E+00 : 2.6690E+00 : 2.6691E+00 :
---------------------------------------------------------------------------------------------------------------------------
*** USER INFORMATION MESSAGE 6464 (DOM12E)
RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 8.
 SECTION 13

AEROELASTIC OPTIMIZATION

AEROELASTIC OPTIMIZATION MSC BACKGROUND . . . . . . . . . . . . . . . . . . . . . 13-1

STATIC AEROELASTICITY DESIGN CONDITIONS . . . . . . . . . . . . . . . . . . . . . . . . 13-2

STATIC AEROELASTICITY SENSITIVITY ANALYSIS . . . . . . . . . . . . . . . . . . . . . . 13-3

FLUTTER DESIGN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-4

FLUTTER RESPONSE PROPERTIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-5

FLUTTER RESPONSE SENSITIVITY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-6

AEROELASTIC DESIGN EXAMPLE (HA200A and B) . . . . . . . . . . . . . . . . . . . . . . 13-7

ANALYSIS CONDITIONS FOR EXAMPLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-9

DESIGN CONDITIONS FOR EXAMPLE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-10

HA200B OBJECTIVE RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-12

HA200B DESIGN VARIABLE RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-13

HA200B FLUTTER RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-14

HA200A INPUT FILE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-15

DESIGN MODEL AND BULK DATA INPUT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-17

HA200A SELECTED RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-23

HA200B SELECTED RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-31

AEROELASTIC OPTIMIZATION MSC BACKGROUND13

Aeroelasticity Design Optimization

Aeroelastic
Optimization

13-1
 STATIC AEROELASTICITY DESIGN CONDITIONS

• Standard response quantities, e.g.,

• Stress/strain

• Displacement

• Internal force

• Trim parameters, for example

• α ≤ 10.0°

• – 20.0° ≤ δ e ≤ 20.0°

• Stability derivative values – restrained or unrestrained

• Equations using stability derivatives can be used to design, e.g.,

roll effectiveness

13-2
 STATIC AEROELASTICITY SENSITIVITY ANALYSIS

• Similar to sensitivity of static analysis without aerodynamics

• Aerodynamic matrices are invariant with respect to design

variables

• Calculation varies depending on the design condition

• Stability derivative sensitivities require additional pseudo-

displacement vectors

13-3
 FLUTTER DESIGN

• The PK method solves for p at a set of specified velocities

• Given p:

ω = Imag ( p )

γ = Real ( p ) ⁄ ω

• The flutter response is the damping level (γ) at user specified:

• Velocities

• Densities

• Mach numbers

• Mode numbers

13-4
 FLUTTER RESPONSE PROPERTIES

• Computation of the flutter speed is not required

• Response evaluated only at velocities of interest

• “Hump” mode behavior is addressed

• Typically, only a few responses require gradients

13-5
 FLUTTER RESPONSE SENSITIVITY

• The gradient of the damping value is given by

∂γ 1 ∂p R ∂p l
------ = ----  ---------- – γ --------
∂x ω  ∂x ∂x 

• Gradient is computed analytically

• Sensitivity of the normal modes is considered negligible

• Adjoint flutter vector is utilized

• Similar to frequency constraint sensitivity

• Gradient calculation is conceptually straightforward but

algebraically complex

13-6
 AEROELASTIC DESIGN EXAMPLE (HA200A and B)

• Aerodynamic panels

(a) Plan View (b) Side View

2000 1000
Canard

2100

2007 1007

2103 1131

1100

2115 1119

Aileron
3100

2131
3115
Wing
3103

Rudder

This is a standard MSC/NASTRAN test case (Example HA144E).

13-7
 AEROELASTIC DESIGN EXAMPLE (Cont.)
(HA200A and B)

• Structural model

97 97
(a) Plan View (b) Side View

98 Designed Bars 98

90 90

221 99 121
99

220 120

222 122
211 111

210 110

212 112 100

100 311

310

312

This is a standard MSC/NASTRAN test case (Example HA144E).

13-8
 ANALYSIS CONDITIONS FOR EXAMPLE

• Five trim conditions:

• Level flight, M = 0.9, q = 1200 psf (sea level)

• Level flight, M = 1.2, q = 863 psf (20,000 ft.)

• Steady rolling pullout, M = 0.9, q = 1200 psf

• Abrupt rolling pullout, M = 0.9, q = 1200 psf

• Snap-roll entry, M = 0.45, q = 300 psf

• Two flutter conditions:

• Sea level, M = 0.9

• Sea level, M = 1.2

13-9
 DESIGN CONDITIONS FOR EXAMPLE

• Minimize the weight:

3
W = ∑ ρi Ai l i
i=1

With bar properties:

A   1.5000 
I   
 1  0.1736 
  =  X i
 2
I  2.0000 
   0.4630 
 J i  

where X1 = DV10 for the inboard wing bar

X2 = DV20 for the outboard wing bar

X3 = DV30 for the fin bar

and ρ1 = 3
38.490 lb/ft to give a weight of 333 lbs when X1 = 1.0

ρ2 = 3
38.490 lb/ft to give a weight of 667 lbs when X2 = 1.0

ρ3 = 3
5.7735 lb/ft to give a weight of 50 lbs when X3 = 1.0

13-10
 DESIGN CONDITIONS FOR EXAMPLE (Cont.)

Subject to:

6
• σt = 50,000 psi = 7.20 × 10 psf limit allowable tensile stress

6
• σc = 40,000 psi = 5.76 × 10 psf limit allowable compressive stress

( u5 ) – ( u5 )
120 100 
 1.0°, M = 0.9 , Level flight, q = 1200 psf
• ≤
( u5 ) – ( u5 )
100  
0.5°, M = 1.2 , Level flight, q = 863 psf
220

pb  0.60, M = 0.9, q = 1200 psf

• -------------- ≥ 
2Vδ a  0.43, M = 1.2, q = 863 psf

• g ≤ 0.00 at V = 1000, 1300 and 1500 ft/s for M = 0.9 and 1.2

13-11
 HA200B OBJECTIVE RESULTS

4000

3500
Wing Structural Weight (lbs)

3000

2500

2000

1500

1000

500

0
0 1 2 3 4 5 6 7
Iteration Number

13-12
 HA200B DESIGN VARIABLE RESULTS

2.0

= DV10
Design Variable

= DV20
= DV30
1.0

0.0
0 1 2 3 4 5 6 7
Iteration Number

13-13
 HA200B FLUTTER RESULTS

1.00
Optimized Design
0.75
Initial Design
0.50
Damping (g)

0.25

0.00

–0.25 VREQ
–0.50

–0.75

–1.00

6.0

5.0
Frequency (Hz)

4.0

3.0

2.0 Optimized Design

Initial Design

1.0
0.0 1000 1200 1400 1600 1800
Velocity (ft/sec)

13-14
 HA200A INPUT FILE

ID MSC, HA200A
$$$$$$$$ HANDBOOK FOR AEROELASTIC ANALYSIS EXAMPLE HA200A $$$$$$$$
$$
$ MODEL DESCRIPTION FULL SPAN 30 DEG FWD SWEPT WING $
$ WITH AILERON, CANARD AND AFT SWEPT $
$ VERTICAL FIN AND RUDDER. $
$ BAR MODEL WITH DUMBBELL MASSES.$
$$
$ SOLUTION QUASI-STEADY AEROELASTIC ANALYSIS$
$ AND UNSTEADY FLUTTER ANALYSIS USING$
$ DOUBLET-LATTICE METHOD$
$ AERODYNAMICS AT MACH NO. 0.9.$
$$
$ OUTPUT STANDARD AEROELASTIC OUTPUT PLUS$
$ A TABLE IDENTIFYING RESPONSES$
$ FOR WHICH SENSITIVITY RESULTS ARE$
$ AVAILABLE FOLLOWED BY A MATRIX OF$
$ SENSITIVITY VALUES.$
$ $
$$
$$$$$$$$ $$$$$$$$
TIME 30 $ CPU TIME IN MINUTES
SOL 200 $ OPTIMIZATION WITH AEROELASTICITY
CEND
TITLE = EXAMPLE HA200A: 30 DEG FWD SWEPT WING WITH CANARD AND FIN
SUBTI = DEMONSTRATION OF AEROELASTIC SENSITIVITY ANALYSIS
ECHO = BOTH
SPC = 1
DESOBJ = 10
DISP = ALL $
STRESS = ALL $
FORCE = ALL $
AEROF = ALL $
APRES = ALL $
SUBCASE 1
LABEL = SUBSONIC SYMMETRIC PULLOUT
ANALYSIS = SAERO
DESSUB = 1
TRIM = 1 $
SUBCASE 2
LABEL = SUPERSONIC SYMMETRIC PULLOUT
ANALYSIS = SAERO
DESSUB = 2
TRIM = 2 $
SUBCASE 3
LABEL = HIGH SPEED ROLLING PULLOUT
ANALYSIS = SAERO
DESSUB = 200
TRIM = 3 $

13-15
 HA200A INPUT FILE (Cont.)

SUBCASE 4
LABEL = HIGH SPEED PULLUP WITH ABRUPT ROLL
ANALYSIS = SAERO
DESSUB = 200
TRIM = 4 $
SUBCASE 5
LABEL = SUBSONIC ENTRY INTO SNAP ROLL
ANALYSIS = SAERO
DESSUB = 200
TRIM = 5 $
SUBCASE 6
LABEL = SUBSONIC FLUTTER ANALYSIS
ANALYSIS = FLUTTER
SET 10 = 1,THRU,100000
PARAM OPPHIPA,1
DISP = 10
STRESS = NONE $
FORCE = NONE $
AEROF = NONE $
APRES = NONE $
DESSUB = 6
METHOD = 20
FMETHOD = 30
SUBCASE 7
LABEL = SUPERSONIC FLUTTER ANALYSIS
ANALYSIS = FLUTTER
DISP = NONE $
STRESS = NONE $
FORCE = NONE $
AEROF = NONE $
APRES = NONE $
DESSUB = 7
METHOD = 20
FMETHOD = 40
BEGIN BULK

13-16
 DESIGN MODEL AND BULK DATA INPUT

$
$ * * * *
$
$ * * * OPTIMIZATION * * *
$
$
$ * * *
$
$ * THE DESIGN MODEL *
$
$ DEFINITION OF THE DESIGN VARIABLES
$
$ THE DESVAR ENTRY DEFINES A DESIGN VARIABLE FOR DESIGN
$ OPTIMIZATION. LISTED ARE A UNIQUE DESIGN VARIABLE ID
$ NUMBER, A USER SUPPLIED NAME FOR PRINTING PURPOSES, AN
$ INITIAL VALUE, A LOWER BOUND, AND AN UPPER BOUND.
$
$DESVAR ID LABEL XINIT XLB XUB
DESVAR 10 PBAR101 1.0 0.001 100.0
DESVAR 20 PBAR102 1.0 0.001 100.0
DESVAR 30 PBAR103 0.1 0.001 100.0
$
$ RELATIONSHIP OF DESIGN VARIABLES TO ANALYSIS MODEL PROPERTIES
$
$ THE DVPREL1 ENTRY EXPRESSES AN AN ANALYSIS MODEL PROPERTY AS
$ A LINEAR FUNCTION OF DESIGN VARIABLES. IT LISTS A UNIQUE ID,
$ AN ANALYSIS MODEL ENTRY TYPE ID STRING, A PROPERTY ENTRY ID,
$ THE FIELD POSITION OF THE PROPERTY ENTRY OR WORD POSITION IN
$ THE ELEMENT PROPERTY TABLE OF THE ANALYSIS MODEL, THE MINIMUM
$ AND MAXIMUM VALUES ALLOWED FOR THIS PROPERTY DURING OPTIMIZATION,
$ A CONSTANT TERM OF RELATION, A DESIGN VARIABLE ENTRY (DESVAR)
$ ID, AND A COEFFICIENT OF LINEAR RELATION. THE EQUATION IS PI
$ = CO + CJXJ AND THE CONTINUATION ENTRY CAN BE USED TO LIST
$ MORE THAN ONE CJ. IN THIS CASE, THERE IS ONLY ONE CJ FOR EACH
$ DVPREL1 ENTRY.
$
$DVPREL1ID TYPE PID FID PMIN PMAX C0 XXXX DVPREL1
DVPREL1 1014 PBAR 101 4
$ DVID1 COEF1 DVID2 COEF2 DVID3 COEF3 ....
10 1.5
DVPREL1 1015 PBAR 101 5
10 0.173611
DVPREL1 1016 PBAR 101 6
10 2.0
DVPREL1 1017 PBAR 101 7
10 0.462963
DVPREL1 1024 PBAR 102 4
20 1.5
DVPREL1 1025 PBAR 102 5
20 0.173611

13-17
 DESIGN MODEL AND BULK DATA INPUT (Cont.)

DVPREL1 1026 PBAR 102 6

20 2.0
DVPREL1 1027 PBAR 102 7
20 .462963
DVPREL1 1028 PBAR 103 4
30 1.5
DVPREL1 1029 PBAR 103 5
30 0.173611
DVPREL1 1030 PBAR 103 6
30 2.0
DVPREL1 1031 PBAR 103 7
30 .462963
$
$ * * *
$
$ * STRUCTURAL RESPONSES AND CONSTRAINTS *
$
$ THE DRESP1 ENTRY DEFINES A SET OF STRUCTURAL RESPONSES THAT
$ IS USED IN THE DESIGN EITHER AS CONSTRAINTS OR AS AN OBJECTIVE.
$ IT LISTS A UNIQUE ENTRY IDENTIFIER, A USER DEFINED LABEL, THE
$ THE RESPONSE TYPE, THE PROPERTY ENTRY TYPE OR ELEMENT ID FLAG
$ (ELEM), A REGION IDENTIFIER FOR CONSTRAINT SCREENING, AND A
$ NUMBER OF ATTRIBUTES DEFINED IN THE TABLE GIVEN IN THE DRESP1
$ BULK DATA ENTRY DESCRIPTION (SEE EITHER THE MSC/NASTRAN USERS
$ MANUAL, THE DESIGN OPTIMIZATION USERS GUIDE, OR THE QUICK
$ REFERENCE GUIDE).
$
$ DRESP1,10, IDENTIFIES THE
$ WEIGHT RESPONSE USED AS THE OBJECTIVE FUNCTION.
$
$DRESP1 ID LABEL RTYPE PTYPE REGION ATTA ATTB ATT1
DRESP1 10 WEIGHT WEIGHT
$
$ THE FOLLOWING DRESP1 ENTRIES IDENTIFY MAXIMUM AND MINIMUM
$ STRESSES AT END A OF ALL ELEMENTS IN PROPERTY GROUPS 101,
$ 102, AND 103. THE STRESS ITEM CODES LISTED IN THE ATTA
$ FIELDS CAN BE FOUND IN SECTION 4 OF THE MSC/NASTRAN USERS
$ MANUAL.
$
$
$ STRESS RESPONSES
$
$DRESP1 ID LABEL RTYPE PTYPE REGION ATTA ATTB ATT1
DRESP1 1001 TEN110 STRESS PBAR 7 101
102 103
DRESP1 1002 COM110 STRESS PBAR 8 101
102 103
$

13-18
 DESIGN MODEL AND BULK DATA INPUT (Cont.)

$ STRESS CONSTRAINTS
$
$ THE DCONSTR ENTRY DEFINES DESIGN CONSTRAINTS. LISTED ARE
$ A CONSTRAINT SET ID, THE DRESPI ENTRY ID AND THE LOWER AND
$ UPPER BOUND IMPOSED ON THIS RESPONSE QUANTITY. THE FOLLOWING
$ TWO ENTRIES IMPOSE LIMITS ON THE ALLOWABLE STRESSES IN THE
$ BARS IN UNITS OF POUNDS/FOOT**2.
$
$DCONSTR DCID RID LALLOW UALLOW
DCONSTR 200 1001 -5.76+6 7.20+6
DCONSTR 200 1002 -5.76+6 7.20+6
$
$
$ TIP TWIST RESPONSES
$
DRESP1 101 RTIPROT DISP 5 120
DRESP1 201 LTIPROT DISP 5 220
DRESP1 100 RTROT DISP 5 100
$
$ SECOND LEVEL RESPONSES FOR TIP TWIST
$
$ THE DRESP2 ENTRY DEFINES THE INPUT ARGUMENTS TO USER-SUPPLIED
$ EQUATIONS. THESE SECOND-LEVEL RESPONSES CAN THEN EITHER BE
$ USED AS CONSTRAINTS OR AS AN OBJECTIVE FUNCTION. INPUT MAY
$ CONSIST OF DESIGN VARIABLES (DESVAR), FIRST-LEVEL (DRESP1)
$ RESPONSES, TABLE CONSTANTS (DTABLE), AND GRID COORDINATES
$ (DVGRID). IT LISTS AN ID, A USER DEFINED LABEL, THE DEQATN
$ ENTRY ID, A REGION IDENTIFIER FOR CONSTRAINT SCREENING, A
$ STRING INDICATING DESVAR ID NUMBERS, A DESVAR ID, A STRING
$ INDICATING THAT THE LABELS FOR THE CONSTANTS IN A DTABLE ENTRY
$ FOLLOW, THE LABELS OF CONSTANTS IN THE DTABLE INPUT,
$ A STRING INDICATING DRESP1 ID NUMBERS, DRESP1 IDS, A STRING
$ SIGNIFYING THAT THE IDS AND DIRECTIONS FOLLOWING ARE NODE
$ NUMBERS AND CARTESIAN DIRECTION COMPONENTS, NODE NUMBERS,
$ AND CARTESIAN DIRECTIONS.
$
$DRESP2 ID LABEL EQID REGION XXXX XXXX XXXX XXXX
DRESP2 5 RHSTWIST5 +DR25
$ DRESP1 NR1 NR2 NR3 ETC
+DR25 DRESP1 101 100
$
DRESP2 6 LHSTWIST5 +DR26
+DR26 DRESP1 201 100
$
$ EQUATION DEFINING SECOND LEVEL RESPONSE FOR TIP TWIST
$
$ THE DEQATN ENTRY DEFINES THE EQUATION(S) USED IN THE DESIGN
$ PROCESS. IT LISTS A UNIQUE EQUATION ID, AND THE EQUATION IS
$ WRITTEN IN FORTRAN LIKE SYNTAX FOLLOWING THE RULES IN DEFINING
$ DMAP ASSIGNMENTS AND FUNCTIONS.

13-19
 DESIGN MODEL AND BULK DATA INPUT (Cont.)

$DEQATN EQID EQUATION

DEQATN 5 F(RTIP,RROOT) = RTIP - RROOT
$
$
$ TIP TWIST CONSTRAINTS
$
DCONSTR 50 5 -0.017450.01745
DCONSTR 50 6 -0.017450.01745
DCONSTR 60 5 -.008726 .008726
DCONSTR 60 6 -.008726 .008726
DCONADD 1 50 200
DCONADD 2 60 200
$
$ * * *
$
$ * AILERON ROLL EFFECTIVENESS *
$
$
$ RESPONSES REQUIRED FOR ROLL EFFECTIVENESS
$
DRESP1 1401 CLDELTA STABDER 517 0 4
DRESP1 1402 CLP STABDER 513 0 4
$
$
$DRESP2 ID LABEL EQID REGION XXX XXX XXX XXX +DRES2
DRESP2 2401 ROLLEFF 103 +DR2401
$ DRESP1 NR1 NR2 NR3 ETC
+DR2401 DRESP1 1402 1401
$DEQATN EQID EQUATION
DEQATN 103 F(A,B) = -B/A
$
$ SUBSONIC AILERON EFFECTIVENESS CONSTRAINT
$
DCONSTR 50 2401 0.60
$
$ SUBSONIC AILERON EFFECTIVENESS CONSTRAINT
$
DCONSTR 60 2401 0.43
$
$ * * *
$
$ * SUBSONIC AND SUPERSONIC FLUTTER *
$
$ RESPONSE FOR SUBSONIC FLUTTER
$
$DRESP1 ID LABEL FLUTTER XXXX REGION SID XXXX ID_MODE +DR
DRESP1 1 FLUTTER FLUTTER 88 +DR1
$ ID_DENS ID_MACH ID_VEL
+DR1 1 2 4

13-20
 DESIGN MODEL AND BULK DATA INPUT (Cont.)

$
$ SELECTION OF FLUTTER MODES FOR OPTIMIZATION
$
SET1 88 4 THRU 7
$
$ SELECTION OF VELOCITIES FOR IMPOSING FLUTTER CONSTRAINTS
$
FLFACT 4 1000.0 1300.0 1500.0
$
$ EQUATION FOR SECOND LEVEL FLUTTER RESPONSE TO CONSTRAIN SYSTEM DAMPING
$
DRESP2 4 GDAMP 4 +DR24
+DR24 DRESP1 1
DEQATN 4 F(A) = (A - 0.03)/0.1
$
$ CONSTRAINT ON AEROELASTIC SYSTEM DAMPING
$
DCONSTR 6 4 -1.0+20 -0.3
$
$
$ RESPONSE FOR SUPERSONIC FLUTTER
$
$DRESP1 ID LABEL FLUTTER XXXX REGION SID XXXX ID_MODE +DR
DRESP1 11 FLUTTER FLUTTER 89 +DR11
$ ID_DENS ID_MACH ID_VEL
+DR11 11 12 14
$
SET1 89 4 THRU 7
$
FLFACT 14 1000.0 1300.0 1500.0
$
DRESP2 21 GDAMP 4 +DR21
+DR21 DRESP1 11
$
DCONSTR 7 21 -1.0+20 -0.3
$
$ * * *
$
$ * OPTIMIZATION CONTROL PARAMETERS *
$
$ THE DOPTPRM ENTRY IS USED TO OVERRIDE DEFAULT OPTIMIZATION
$ PARAMETERS. THE DOPTPRM ENTRY IS USED
$ HERE TO SET THE MAXIMUM NUMBER OF DESIGN CYCLES TO 25 AND TO
$ REQUEST DETAILED RESULTS FOLLOWING EACH OPTIMIZATION CYCLE.
$
$ PARAM1 VAL1 PARAM2 VAL2 ETC
DOPTPRM DESMAX 25 P1 2 P2 15 DELB 0.01
$

13-21
 DESIGN MODEL AND BULK DATA INPUT (Cont.)

$ THE PARAMETER CDIF,YES FORCES THE SELECTION OF THE CENTRAL

$ DIFFERENCE SCHEME USED IN THE SEMI-ANALYTIC APPROACH
$ REGARDLESS OF THE TYPE OF OPTIMIZATION REQUESTED.
PARAM CDIF YES
$
$ THE BULK DATA PARAMETER, NASPRT, IS USED TO SPECIFY HOW OFTEN
$ MSC/NASTRAN OUTPUT IS TO BE PROVIDED. ITS VALUE INDICATES THAT
$ OUTPUT IS TO BE COMPUTED EVERY N-TH DESIGN CYCLE.
PARAM NASPRT 2
$
$ THE BULK DATA PARAMETER, OPTEXIT, IS USED TO CONTROL THE EXIT
$ POINT FROM A SOLUTION 200 OPTIMIZATION TASK. OPTEXIT=4 INDICATES
$ THAT A SENSITIVITY ANALYSIS IS TO BE PERFORMED WITHOUT OPTIMIZATION.
PARAM OPTEXIT 4
$
$ THE DSCREEN ENTRY IS USED HERE TO FORCE RETENTION OF ALL THE
$ CONSTRAINTS SO THAT THE DESIRED RESPONSE SENSITIVITIES WILL
$ BE PRINTED. CONSTRAINTS ARE RETAINED IF THEY ARE GREATER
$ THAN TRS.
$SCREEN RTYPE TRS NSTR
DSCREEN STRESS -2.0
DSCREEN EQUA -2.0
ENDDATA
INPUT BULK DATA CARD COUNT = 900

13-22
 HA200A SELECTED RESULTS

EXAMPLE HA200A: 30 DEG FWD SWEPT WING WITH CANARD AND FIN PAGE 31
DEMONSTRATION OF AEROELASTIC SENSITIVITY ANALYSIS

----- COMPARISON BETWEEN INPUT PROPERTY VALUES FROM ANALYSIS AND DESIGN MODELS -----
---------------------------------------------------------------------------------------------------------------
PROPERTY PROPERTY FIELD ANALYSIS DESIGN LOWER UPPER DIFFERENCE
TYPE ID ID VALUE VALUE BOUND BOUND FLAG
---------------------------------------------------------------------------------------------------------------
PBAR 101 4 1.500000E+00 1.500000E+00 1.000000E-03 1.000000E+20 NONE
PBAR 101 5 1.736110E-01 1.736110E-01 1.000000E-03 1.000000E+20 NONE
PBAR 101 6 2.000000E+00 2.000000E+00 1.000000E-03 1.000000E+20 NONE
PBAR 101 7 4.629630E-01 4.629630E-01 1.000000E-03 1.000000E+20 NONE
PBAR 102 4 1.500000E+00 1.500000E+00 1.000000E-03 1.000000E+20 NONE
PBAR 102 5 1.736110E-01 1.736110E-01 1.000000E-03 1.000000E+20 NONE
PBAR 102 6 2.000000E+00 2.000000E+00 1.000000E-03 1.000000E+20 NONE
13-23

PBAR 102 7 4.629630E-01 4.629630E-01 1.000000E-03 1.000000E+20 NONE

PBAR 103 4 1.500000E+00 1.500000E-01 1.000000E-03 1.000000E+20 WARNING
PBAR 103 5 1.736110E-01 1.736110E-02 1.000000E-03 1.000000E+20 WARNING
PBAR 103 6 2.000000E+00 2.000000E-01 1.000000E-03 1.000000E+20 WARNING
PBAR 103 7 4.629630E-01 4.629630E-02 1.000000E-03 1.000000E+20 WARNING
1. IF FIELD ID IS LESS THAN ZERO, IT IDENTIFIES THE WORD POSITION OF AN ENTRY IN EPT.
2. IF FIELD ID IS GREATER THAN ZERO, IT IDENTIFIES THE FIELD POSITION ON A PROPERTY BULK DATA ENTRY.
3. THE DIFFERENCE FLAG IS USED TO CHARACTERIZE DIFFERENCES BETWEEN ANALYSIS AND DESIGN MODEL PROPERTIES:
IF THE FLAG IS NONE, THEN THERE IS NO SIGNIFICANT DIFFERENCE BETWEEN THE TWO VALUES.
IF THE FLAG IS WARNING, THEN THE USER IS ADVISED THAT DIFFERENCES EXIST.
IF THE FLAG IS FATAL, THEN THE DIFFERENCES ARE GREATER THAN 1.00000E+35 AND THE RUN WILL BE TERMINATED.
 HA200A SELECTED RESULTS (Cont.)

----- WEIGHT/VOLUME RESPONSES -----

------------------------------------------
COLUMN DRESP1 RESPONSE
NO. ENTRY ID TYPE
------------------------------------------
1 10 WEIGHT
----- STATICS RESPONSES -----
----------------------------------------------------------------------------------------------------------
COLUMN DRESP1 RESPONSE GRID/ELM COMPONENT SUB PLY
NO. ENTRY ID TYPE ID NO. CASE NO.
----------------------------------------------------------------------------------------------------------
2 100 DISP 100 5 1
3 101 DISP 120 5 1
4 201 DISP 220 5 1
5 1001 STRESS 110 7 1
13-24

6 1001 STRESS 210 7 1

7 1001 STRESS 120 7 1
8 1001 STRESS 220 7 1
9 1001 STRESS 310 7 1
10 1002 STRESS 110 8 1
11 1002 STRESS 210 8 1
12 1002 STRESS 120 8 1
13 1002 STRESS 220 8 1
14 1002 STRESS 310 8 1
----- AEROELASTIC STABILITY DERIVATIVE RESPONSES -----
--------------------------------------------------------------------------------------------------------------
COLUMN DRESP1 RESPONSE SUBCASE RESTRAINED/ COMPONENT AESTAT/AESURF
NO. ENTRY ID TYPE ID UNRESTRAINED NO. ID
--------------------------------------------------------------------------------------------------------------
15 1401 STABDER 1 UNRESTRAINED 4 517
16 1402 STABDER 1 UNRESTRAINED 4 513
 HA200A SELECTED RESULTS (Cont.)

----- STATICS RESPONSES -----

----------------------------------------------------------------------------------------------------------
COLUMN DRESP1 RESPONSE GRID/ELM COMPONENT SUB PLY
NO. ENTRY ID TYPE ID NO. CASE NO.
----------------------------------------------------------------------------------------------------------
17 100 DISP 100 5 2
18 101 DISP 120 5 2
19 201 DISP 220 5 2
20 1001 STRESS 110 7 2
21 1001 STRESS 210 7 2
22 1001 STRESS 120 7 2
23 1001 STRESS 220 7 2
24 1001 STRESS 310 7 2
25 1002 STRESS 110 8 2
26 1002 STRESS 210 8 2
13-25

27 1002 STRESS 120 8 2

28 1002 STRESS 220 8 2
29 1002 STRESS 310 8 2
----- AEROELASTIC STABILITY DERIVATIVE RESPONSES -----
--------------------------------------------------------------------------------------------------------------
COLUMN DRESP1 RESPONSE SUBCASE RESTRAINED/ COMPONENT AESTAT/AESURF
NO. ENTRY ID TYPE ID UNRESTRAINED NO. ID
--------------------------------------------------------------------------------------------------------------
30 1401 STABDER 2 UNRESTRAINED 4 517
31 1402 STABDER 2 UNRESTRAINED 4 513
 HA200A SELECTED RESULTS (Cont.)

----- STATICS RESPONSES -----

----------------------------------------------------------------------------------------------------------
COLUMN DRESP1 RESPONSE GRID/ELM COMPONENT SUB PLY
NO. ENTRY ID TYPE ID NO. CASE NO.
----------------------------------------------------------------------------------------------------------
32 1001 STRESS 110 7 3
33 1001 STRESS 210 7 3
34 1001 STRESS 120 7 3
35 1001 STRESS 220 7 3
36 1001 STRESS 310 7 3
37 1002 STRESS 110 8 3
38 1002 STRESS 210 8 3
39 1002 STRESS 120 8 3
40 1002 STRESS 220 8 3
41 1002 STRESS 310 8 3
13-26

42 1001 STRESS 110 7 4

43 1001 STRESS 210 7 4
44 1001 STRESS 120 7 4
45 1001 STRESS 220 7 4
46 1001 STRESS 310 7 4
47 1002 STRESS 110 8 4
48 1002 STRESS 210 8 4
49 1002 STRESS 120 8 4
50 1002 STRESS 220 8 4
51 1002 STRESS 310 8 4
52 1001 STRESS 110 7 5
53 1001 STRESS 210 7 5
54 1001 STRESS 120 7 5
55 1001 STRESS 220 7 5
56 1001 STRESS 310 7 5
57 1002 STRESS 110 8 5
58 1002 STRESS 210 8 5
59 1002 STRESS 120 8 5
60 1002 STRESS 220 8 5
61 1002 STRESS 310 8 5
 HA200A SELECTED RESULTS (Cont.)

----- AEROELASTIC FLUTTER RESPONSES -----

--------------------------------------------------------------------------------------------------------------------------
COLUMN DRESP1 RESPONSE SUBCASE MODE DENSITY MACH VELOCITY
NO. ENTRY ID TYPE ID NO. NO.
--------------------------------------------------------------------------------------------------------------------------
62 1 FLUTTER 6 4 1.000 0.900 1000.000
63 1 FLUTTER 6 4 1.000 0.900 1300.000
64 1 FLUTTER 6 4 1.000 0.900 1500.000
65 1 FLUTTER 6 5 1.000 0.900 1000.000
66 1 FLUTTER 6 5 1.000 0.900 1300.000
67 1 FLUTTER 6 5 1.000 0.900 1500.000
68 1 FLUTTER 6 6 1.000 0.900 1000.000
69 1 FLUTTER 6 6 1.000 0.900 1300.000
70 1 FLUTTER 6 6 1.000 0.900 1500.000
71 1 FLUTTER 6 7 1.000 0.900 1000.000
13-27

72 1 FLUTTER 6 7 1.000 0.900 1300.000

73 1 FLUTTER 6 7 1.000 0.900 1500.000
74 11 FLUTTER 7 4 1.000 1.200 1000.000
75 11 FLUTTER 7 4 1.000 1.200 1300.000
76 11 FLUTTER 7 4 1.000 1.200 1500.000
77 11 FLUTTER 7 5 1.000 1.200 1000.000
78 11 FLUTTER 7 5 1.000 1.200 1300.000
79 11 FLUTTER 7 5 1.000 1.200 1500.000
80 11 FLUTTER 7 6 1.000 1.200 1000.000
81 11 FLUTTER 7 6 1.000 1.200 1300.000
82 11 FLUTTER 7 6 1.000 1.200 1500.000
83 11 FLUTTER 7 7 1.000 1.200 1000.000
84 11 FLUTTER 7 7 1.000 1.200 1300.000
85 11 FLUTTER 7 7 1.000 1.200 1500.000
----- IDENTIFICATION OF COLUMNS IN THE DESIGN SENSITIVITY -----
----- MATRIX THAT ARE ASSOCIATED WITH DRESP2 ENTRIES -----
 HA200A SELECTED RESULTS (Cont.)

----------------------------------------------------------
COLUMN DRESP2 SUB FREQ/
NO. ENTRY ID CASE TIME
----------------------------------------------------------
86 5 1
87 6 1
88 2401 1
89 5 2
90 6 2
91 2401 2
92 4 6
93 4 6
94 4 6
95 4 6
96 4 6
13-28

97 4 6
98 4 6
99 4 6
100 4 6
101 4 6
102 4 6
103 4 6
104 21 7
105 21 7
106 21 7
107 21 7
108 21 7
109 21 7
110 21 7
111 21 7
112 21 7
113 21 7
114 21 7
115 21 7
^^^ DMAP INFORMATION MESSAGE 9029 (DESOPT) - DESIGN SENSITIVITY COEFFICIENT MATRIX FOR DIRECT
AND SYNTHETIC RESPONSES - GRADIENTS OF RESPONSES WITH RESPECT TO INDEPENDENT DESIGN VARIABLES
 HA200A SELECTED RESULTS (Cont.)

MATRIX DSCM2 (GINO NAME 101 ) IS A REAL 115 COLUMN X 3 ROW RECTANG MATRIX.
COLUMN 1 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) 6.6667E+02 1.3333E+03 5.0000E+01
COLUMN 2 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) 2.3115E-04 5.5879E-06 5.3085E-05
COLUMN 3 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) -5.4311E-03 -4.8257E-03 8.5682E-05
COLUMN 4 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) -5.4313E-03 -4.8257E-03 8.5682E-05
COLUMN 5 ROWS 1 THRU 3 --------------------------------------------------
ROW
13-29

1) -1.2751E+06 1.3784E+05 5.2500E+03

COLUMN 6 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) -1.2751E+06 1.3784E+05 5.2500E+03
COLUMN 7 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) -1.2258E+05 -4.9729E+05 2.1250E+03
.
.
.
COLUMN 110 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) 5.2690E-03 3.3642E-02 8.6986E-04
COLUMN 111 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) 1.4705E+00 1.4853E+00 1.0002E-01
COLUMN 112 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) 1.1036E+01 5.7482E+00 3.5277E-01
 HA200A SELECTED RESULTS (Cont.)

COLUMN 113 ROWS 1 THRU 3 --------------------------------------------------

ROW
1) 5.7897E-01 7.2965E-01 2.5005E-02
COLUMN 114 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) 9.0316E-03 6.3687E-02 1.4559E-03
COLUMN 115 ROWS 1 THRU 3 --------------------------------------------------
ROW
1) 1.2167E-02 8.5035E-02 1.8582E-03
THE NUMBER OF NON-ZERO TERMS IN THE DENSEST COLUMN = 3
THE DENSITY OF THIS MATRIX IS 98.84 PERCENT.
***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************
NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 1
NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 0
OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY
13-30

---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
CYCLE APPROXIMATE EXACT OF OF
NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------
INITIAL 2.005000E+03 1.460522E+01
---------------------------------------------------------------------------------------------------------------
DESIGN VARIABLE HISTORY
----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 10 | PBAR101 | 1.0000E+00 :
2 | 20 | PBAR102 | 1.0000E+00 :
3 | 30 | PBAR103 | 1.0000E-01 :
----------------------------------------------------------------------------------------------------------------------------------
*** USER INFORMATION MESSAGE 6464 (DOM12E)
RUN TERMINATED DUE TO PARAMETER OPTEXIT = 4.
 HA200B SELECTED RESULTS

EXAMPLE HA200B: 30 DEG FWD SWEPT WING WITH CANARD AND FIN PAGE 159
DEMONSTRATION OF AEROELASTIC OPTIMIZATION
SUPERSONIC FLUTTER ANALYSIS SUBCASE 7
THIS IS THE FIRST ANALYSIS - NO CONVERGENCE CHECK
-------------------------------------------------------
MAXIMUM VALUE OF CONSTRAINTS : 1.4605E+01
-------------------------------------------------------

********************************************************************
* *
* *
* D E S I G N O P T I M I Z A T I O N *
* *
* *
********************************************************************
13-31

*******************************************
* *
* D E S I G N C Y C L E 1 *
* *
*******************************************

***** OPTIMIZATION RESULTS BASED ON THE APPROXIMATE MODEL *****

----- DESIGN OBJECTIVE -----
---------------------------------------------------------------------------------------------------------------------------
INTERNAL TYPE MINIMIZE
RESPONSE OF OR SUPERELEMENT SUBCASE INPUT OUTPUT
ID RESPONSE LABEL MAXIMIZE ID ID VALUE VALUE
---------------------------------------------------------------------------------------------------------------------------
1 DRESP1 WEIGHT MINIMIZE 0 1 2.0050E+03 3.5277E+03
 HA200B SELECTED RESULTS (Cont.)

----- DESIGN VARIABLES -----

---------------------------------------------------------------------------------------------------------------------------
INTERNAL DESVAR LOWER INPUT OUTPUT UPPER
ID ID LABEL BOUND VALUE VALUE BOUND
---------------------------------------------------------------------------------------------------------------------------
1 10 PBAR101 1.0000E-03 1.0000E+00 2.0000E+00 1.0000E+02
2 20 PBAR102 1.0000E-03 1.0000E+00 1.6390E+00 1.0000E+02
3 30 PBAR103 1.0000E-03 1.0000E-01 2.0000E-01 1.0000E+02

----- DESIGNED PROPERTIES -----

---------------------------------------------------------------------------------------------------------------------------
PROPERTY PROPERTY FIELD TYPE OF LOWER INPUT OUTPUT UPPER
TYPE ID ID PROPERTY BOUND VALUE VALUE BOUND
13-32

---------------------------------------------------------------------------------------------------------------------------
PBAR 101 4 DVPREL1 1.0000E-03 1.5000E+00 3.0000E+00 1.0000E+20
PBAR 101 5 DVPREL1 1.0000E-03 1.7361E-01 3.4722E-01 1.0000E+20
PBAR 101 6 DVPREL1 1.0000E-03 2.0000E+00 4.0000E+00 1.0000E+20
PBAR 101 7 DVPREL1 1.0000E-03 4.6296E-01 9.2593E-01 1.0000E+20
PBAR 102 4 DVPREL1 1.0000E-03 1.5000E+00 2.4586E+00 1.0000E+20
PBAR 102 5 DVPREL1 1.0000E-03 1.7361E-01 2.8455E-01 1.0000E+20
PBAR 102 6 DVPREL1 1.0000E-03 2.0000E+00 3.2781E+00 1.0000E+20
PBAR 102 7 DVPREL1 1.0000E-03 4.6296E-01 7.5881E-01 1.0000E+20
PBAR 103 4 DVPREL1 1.0000E-03 1.5000E-01 3.0000E-01 1.0000E+20
PBAR 103 5 DVPREL1 1.0000E-03 1.7361E-02 3.4722E-02 1.0000E+20
PBAR 103 6 DVPREL1 1.0000E-03 2.0000E-01 4.0000E-01 1.0000E+20
PBAR 103 7 DVPREL1 1.0000E-03 4.6296E-02 9.2593E-02 1.0000E+20
 HA200B SELECTED RESULTS (Cont.)

----- DESIGN CONSTRAINTS ON RESPONSES -----

(MAXIMUM RESPONSE CONSTRAINTS MARKED WITH **)

---------------------------------------------------------------------------------------------------------------------------
INTERNAL INTERNAL
INTERNAL DCONSTR RESPONSE RESPONSE L/U REGION SUBCASE INPUT OUTPUT
ID ID ID TYPE FLAG ID ID VALUE VALUE
---------------------------------------------------------------------------------------------------------------------------
1 50 2 EQUA UPPER 6 1 -3.9543E-01 -6.6547E-01
2 50 1 EQUA UPPER 5 1 -3.9543E-01 -6.6440E-01
3 50 3 EQUA LOWER 2401 1 1.7848E-02 -3.4503E-02
4 60 4 EQUA LOWER 2401 2 1.1241E-02 -3.9465E-02
5 6 5 EQUA UPPER 4 6 -4.3507E-01 -3.6965E-01
6 6 6 EQUA UPPER 4 6 1.4605E+01** 1.3839E+00**
13-33

7 7 7 EQUA UPPER 21 7 8.1231E+00 -8.8540E+00

8 7 8 EQUA UPPER 21 7 -4.7149E-01 -4.1885E-01

----- CONSTRAINTS ON DESIGNED PROPERTIES -----

---------------------------------------------------------------------------------------------------------------------------
INTERNAL PROPERTY FIELD L/U CYCLE INPUT OUTPUT
ID ID FLAG FLAG LIMIT VALUE VALUE
---------------------------------------------------------------------------------------------------------------------------
9 101 4 LOWER 1.2000E+00 -2.5000E-01 -1.5000E+00
10 101 5 LOWER 1.3889E-01 -2.5000E-01 -1.5000E+00
11 101 6 LOWER 1.6000E+00 -2.5000E-01 -1.5000E+00
12 101 7 LOWER 3.7037E-01 -2.5000E-01 -1.5000E+00
13 102 4 LOWER 1.2000E+00 -2.5000E-01 -1.0488E+00
 HA200B SELECTED RESULTS (Cont.)

----- CONSTRAINTS ON DESIGNED PROPERTIES -----

---------------------------------------------------------------------------------------------------------------------------
INTERNAL PROPERTY FIELD L/U CYCLE INPUT OUTPUT
ID ID FLAG FLAG LIMIT VALUE VALUE
---------------------------------------------------------------------------------------------------------------------------
14 102 5 LOWER 1.3889E-01 -2.5000E-01 -1.0488E+00
15 102 6 LOWER 1.6000E+00 -2.5000E-01 -1.0488E+00
16 102 7 LOWER 3.7037E-01 -2.5000E-01 -1.0488E+00
17 103 4 LOWER 1.2000E-01 -2.5000E-01 -1.5000E+00
18 103 5 LOWER 7.3611E-03 -1.3585E+00 -3.7170E+00
19 103 6 LOWER 1.6000E-01 -2.5000E-01 -1.5000E+00
20 103 7 LOWER 3.6296E-02 -2.7551E-01 -1.5510E+00
21 101 4 UPPER 1.8000E+00 -1.6667E-01 6.6667E-01
22 101 5 UPPER 2.0833E-01 -1.6667E-01 6.6667E-01
13-34

23 101 6 UPPER 2.4000E+00 -1.6667E-01 6.6667E-01

24 101 7 UPPER 5.5556E-01 -1.6667E-01 6.6667E-01
25 102 4 UPPER 1.8000E+00 -1.6667E-01 3.6586E-01
26 102 5 UPPER 2.0833E-01 -1.6667E-01 3.6586E-01
27 102 6 UPPER 2.4000E+00 -1.6667E-01 3.6586E-01
28 102 7 UPPER 5.5556E-01 -1.6667E-01 3.6586E-01
29 103 4 UPPER 1.8000E-01 -1.6667E-01 6.6667E-01
30 103 5 UPPER 2.7361E-02 -3.6548E-01 2.6904E-01
31 103 6 UPPER 2.4000E-01 -1.6667E-01 6.6667E-01
32 103 7 UPPER 5.6296E-02 -1.7763E-01 6.4474E-01
 HA200B SELECTED RESULTS (Cont.)

----------------------------------------------------------------------
| R E S P O N S E S IN D E S I G N M O D E L |
----------------------------------------------------------------------

(N/A - BOUND NOT ACTIVE OR AVAILABLE)

----- WEIGHT RESPONSE -----

---------------------------------------------------------------------------------------------------------------------------
INTERNAL DRESP1 RESPONSE LOWER INPUT OUTPUT UPPER
ID ID LABEL BOUND VALUE VALUE BOUND
---------------------------------------------------------------------------------------------------------------------------
1 10 WEIGHT N/A 2.0050E+03 3.5277E+03 N/A

D E S I G N C Y C L E = 1 S U B C A S E = 1
13-35

----- DISPLACEMENT RESPONSES -----

---------------------------------------------------------------------------------------------------------------------------
INTERNAL DRESP1 RESPONSE GRID COMPONENT LOWER INPUT OUTPUT UPPER
ID ID LABEL ID NO. BOUND VALUE VALUE BOUND
---------------------------------------------------------------------------------------------------------------------------
2 100 RTROT 100 5 N/A 1.1291E-02 1.1403E-02 N/A
3 101 RTIPROT 120 5 N/A 2.1841E-02 1.7259E-02 N/A
4 201 LTIPROT 220 5 N/A 2.1841E-02 1.7240E-02 N/A

----- STABILITY DERIVATIVE RESPONSES -----

---------------------------------------------------------------------------------------------------------------------------
RESTRAINED / AESTAT /
INTERNAL DRESP1 RESPONSE UNRESTRAINED COMPONENT AESURF LOWER INPUT OUTPUT UPPER
ID ID LABEL FLAG NO. ID BOUND VALUE VALUE BOUND
---------------------------------------------------------------------------------------------------------------------------
5 1401 CLDELTA 0 4 517 N/A 2.6340E-01 2.6922E-01 N/A
6 1402 CLP 0 4 513 N/A -4.4697E-01 -4.3374E-01 N/A
 HA200B SELECTED RESULTS (Cont.)

D E S I G N C Y C L E = 1 S U B C A S E = 2

----- STABILITY DERIVATIVE RESPONSES -----

---------------------------------------------------------------------------------------------------------------------------
RESTRAINED / AESTAT /
INTERNAL DRESP1 RESPONSE UNRESTRAINED COMPONENT AESURF LOWER INPUT OUTPUT UPPER
ID ID LABEL FLAG NO. ID BOUND VALUE VALUE BOUND
---------------------------------------------------------------------------------------------------------------------------
7 1401 CLDELTA 0 4 517 N/A 2.1655E-01 2.2531E-01 N/A
8 1402 CLP 0 4 513 N/A -5.0933E-01 -5.0408E-01 N/A

D E S I G N C Y C L E = 1 S U B C A S E = 6

----- FLUTTER RESPONSES ----- -----

13-36

------------------------------------------------------------------------------------------------------------------------
INTERNAL DRESP1 RESPONSE MODE MACH LOWER INPUT OUTPUT UPPER
ID ID LABEL NO. VELOCITY NO. BOUND VALUE VALUE BOUND
---------------------------------------------------------------------------------------------------------------------------
9 1 FLUTTER 6 1.0000E+03 9.0000E-01 N/A -1.3052E-02 -1.1089E-02 N/A
10 1 FLUTTER 6 1.5000E+03 9.0000E-01 N/A 4.3816E-01 4.1516E-02 N/A

D E S I G N C Y C L E = 1 S U B C A S E = 7

----- FLUTTER RESPONSES -----

---------------------------------------------------------------------------------------------------------------------------
INTERNAL DRESP1 RESPONSE MODE MACH LOWER INPUT OUTPUT UPPER
ID ID LABEL NO. VELOCITY NO. BOUND VALUE VALUE BOUND
---------------------------------------------------------------------------------------------------------------------------
11 11 FLUTTER 5 1.5000E+03 1.2000E+00 N/A 2.4369E-01 -2.6562E-01 N/A
12 11 FLUTTER 6 1.0000E+03 1.2000E+00 N/A -1.4145E-02 -1.2566E-02 N/A
 HA200B SELECTED RESULTS (Cont.)

---- RETAINED DRESP2 RESPONSES ----

---------------------------------------------------------------------------------------------------------------------------
INTERNAL DRESP2 RESPONSE EQUATION LOWER INPUT OUTPUT UPPER
ID ID LABEL ID BOUND VALUE VALUE BOUND
---------------------------------------------------------------------------------------------------------------------------
1 5 RHSTWIST 5 N/A 1.0550E-02 5.8562E-03 1.7450E-02
2 6 LHSTWIST 5 N/A 1.0550E-02 5.8376E-03 1.7450E-02
3 2401 ROLLEFF 103 6.0000E-01 5.8929E-01 6.2070E-01 N/A
4 2401 ROLLEFF 103 4.3000E-01 4.2517E-01 4.4697E-01 N/A
5 4 GDAMP 4 N/A -4.3052E-01 -4.1089E-01 -3.0000E-01
6 4 GDAMP 4 N/A 4.0816E+00 1.1516E-01 -3.0000E-01
7 21 GDAMP 4 N/A 2.1369E+00 -2.9562E+00 -3.0000E-01
8 21 GDAMP 4 N/A -4.4145E-01 -4.2566E-01 -3.0000E-01
****************************************************************************************
13-37

INSPECTION OF CONVERGENCE DATA FOR THE OPTIMAL DESIGN WITH RESPECT TO APPROXIMATE MODELS
(SOFT CONVERGENCE DECISION LOGIC)
****************************************************************************************

RELATIVE CHANGE IN OBJECTIVE 7.5947E-01 MUST BE LESS THAN 1.0000E-03

OR ABSOLUTE CHANGE IN OBJECTIVE 1.5227E+03 MUST BE LESS THAN 1.0000E-02
AND MAX OF RELATIVE PROP.CHANGES 1.0000E+00 MUST BE LESS THAN 1.0000E-03
--- AND ---
MAXIMUM CONSTRAINT VALUE 1.3839E+00 MUST BE LESS THAN 5.0000E-03
OR MAX OF RELATIVE D.V. CHANGES 1.0000E+00 MUST BE LESS THAN 1.0000E-03
****************************************************************************************
 HA200B SELECTED RESULTS (Cont.)

***** NORMAL CONVERGENCE CRITERIA SATISFIED ***** (HARD CONVERGENCE DECISION LOGIC)

**************************************************************************************
CONVERGENCE ACHIEVED BASED ON THE FOLLOWING CRITERIA
(HARD CONVERGENCE DECISION LOGIC)

RELATIVE CHANGE IN OBJECTIVE 0.0000E+00 MUST BE LESS THAN 1.0000E-03

OR ABSOLUTE CHANGE IN OBJECTIVE 0.0000E+00 MUST BE LESS THAN 1.0000E-02
--- AND ---
MAXIMUM CONSTRAINT VALUE -2.5652E-03 MUST BE LESS THAN 5.0000E-03
(CONVERGENCE TO A FEASIBLE DESIGN)
--- OR ---
MAXIMUM OF RELATIVE PROP. CHANGES 0.0000E+00 MUST BE LESS THAN 1.0000E-03
AND MAXIMUM OF RELATIVE D.V. CHANGES 0.0000E+00 MUST BE LESS THAN 1.0000E-03
(CONVERGENCE TO A BEST COMPROMISE INFEASIBLE DESIGN)
13-38

**************************************************************************************
 HA200B SELECTED RESULTS (Cont.)

***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************

(HARD CONVERGENCE ACHIEVED)

(SOFT CONVERGENCE ACHIEVED)

NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 8

NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 7

OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY

---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
13-39

CYCLE APPROXIMATE EXACT OF OF

NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------
INITIAL 2.005000E+03 1.460522E+01
1 3.527728E+03 3.528712E+03 -2.787540E-04 -3.269444E-02
2 2.823344E+03 2.822969E+03 1.327524E-04 -1.713157E-02
3 2.627507E+03 2.627666E+03 -6.067127E-05 7.991234E-02
4 2.591437E+03 2.591402E+03 1.318965E-05 1.907182E-01
5 2.595658E+03 2.595581E+03 2.962894E-05 1.129563E-01
6 2.601817E+03 2.601818E+03 -4.691731E-07 -2.565185E-03
7 2.601818E+03 2.601818E+03 0.000000E+00 -2.565185E-03
---------------------------------------------------------------------------------------------------------------
 HA200B SELECTED RESULTS (Cont.)

DESIGN VARIABLE HISTORY

----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 4 : 5 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 10 | PBAR101 | 1.0000E+00 : 2.0000E+00 : 1.6000E+00 : 1.8339E+00 : 1.9884E+00 : 2.0878E+00 :
2 | 20 | PBAR102 | 1.0000E+00 : 1.6390E+00 : 1.3112E+00 : 1.0490E+00 : 9.4408E-01 : 8.9742E-01 :
3 | 30 | PBAR103 | 1.0000E-01 : 2.0000E-01 : 1.6000E-01 : 1.2795E-01 : 1.4074E-01 : 1.4312E-01 :
----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | 6 : 7 : 8 : 9 : 10 : 11 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 10 | PBAR101 | 2.0927E+00 : 2.0927E+00 :
2 | 20 | PBAR102 | 8.9963E-01 : 8.9963E-01 :
3 | 30 | PBAR103 | 1.4322E-01 : 1.4322E-01 :
----------------------------------------------------------------------------------------------------------------------------------
*** USER INFORMATION MESSAGE 6464 (DOM12E)
13-40

RUN TERMINATED DUE TO HARD CONVERGENCE TO AN OPTIMUM AT CYCLE NUMBER = 7.

 APPENDIX A

CASE CONTROL COMMANDS AND BULK

DATA ENTRIES

DESIGN MODELING INPUT DATA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-1

CASE CONTROL COMMANDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-3

AUXCASE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-4

AUXMODEL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-5

DESGLB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-6

DESOBJ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-7

DESSUB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-8

MODTRAK. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-9

BULK DATA ENTRIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-10

BNDGRID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-11

DCONADD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-12

DCONSTR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-13

DEQATN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-15

DESVAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-19

DLINK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-20

DOPTPRM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-21

DRESP1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-25

DRESP2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-31

DSAPRT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-34
 APPENDIX A

CASE CONTROL COMMANDS AND BULK DATA

ENTRIES (Cont.)

DSCREEN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-37

DTABLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-39

DVBSHAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-40

DVGRID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-41

DVPREL1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-43

DVPREL2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-46

DVSHAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-48

MODTRAK. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A-49
 DESIGN MODELING INPUT DATA

• Design variables

• DESVAR Design variable definition

• DLINK Definition of dependent design variable

• Relation between design variables and analysis model parameters

• DVPREL1 Linear relations

• DVPREL2 Nonlinear relations

• Definition of structural responses

• DRESP1 Responses computed directly by analysis

• DRESP2 Synthesized responses for design problems

A-1
 DESIGN MODELING INPUT DATA (Cont.)

• Definition of objective and constraint functions

• DESOBJ Case Objective function definition

Control Ccommand

• DCONSTR Constraint functions

• DCONADD Constraint set combinations

• DESSUB Case Selection of subcase-dependent constraints

Control command

• DESGLB Case Selection of global constraints

Control command

• Optimization control parameters and constants

• DSCREEN Measures of contraint screening

• DOPTPRM Optimization process control

• DTABLE Constants

• User equation input

• DEQATN User-definition equation

A-2
CASE CONTROL COMMANDS

A-3
AUXCASE Auxiliary Model Case Control Delimiter

AUXCASE Case Control Commands for an Auxiliary Model in SOL 200.

Delimits Auxiliary Model Case Control Delimiter

Format:
AUXCASE

Examples:
AUXCAS
AUXC

Remarks:
1. AUXCASE indicates the beginning of Case Control commands for an auxiliary model.
AUXCASE must follow the primary model Case Control commands.

2. All Case Control commands following this entry are applicable until the next AUXCASE or
BEGIN BULK command. Commands from preceding Case Control Sections are ignored.

3. Each auxiliary model Case Control must be delimited with the AUXCASE command.

4. The AUXMODEL command is used to associate the auxiliary model Case Control with a
particular auxiliary model.

5.

6.

A-4
Auxiliary Model Identification Number AUXMODEL
AUXMODEL an auxiliary model for generation of boundary shapes in Auxiliary
References shape optimization.
Model Identification Number

Format:
AUXMODEL = n

Examples:
AUXMODEL = 4
AUXM = 4

Describer Meaning
n Auxiliary model identification number. (Integer > 0)

Remarks:
1. AUXMODEL references a particular auxiliary model for analysis and may only be specified
in the auxiliary model Case Control Section.
2. See the BEGIN BULK command for the Bulk Data definition of an auxiliary model.

A-5
DESGLB Request Design Constraints at the Global Level

DESGLB the design constraints to be applied at the global level

Selects in a design
Request optimization
Design Constraints task.
at the Global Level

Format:
DESGLB = n

Examples:
DESGLB = 10
DESG = 25

Describer Meaning
n Set identification of a DCONSTR or DCONADD Bulk Data entry identification
number. (Integer > 0)

Remarks:
1. A DESGLB command is optional and invokes constraints that are to be applied indepen-
dent of a particular subcase. These constraints could be based on responses that are
independent of subcases (e.g., WEIGHT or VOLUME).

2. The DESGLB command can be used to invoke constraints that are not a function of
DRESP1 entries; e.g., DRESP2 responses that are not functions of DRESP1 responses
are subcase independent.

A-6
Design Objective DESOBJ
DESOBJ the DRESP1 or DRESP2 entry to be used as the design objective.
Selects Design Objective

Format:

MAX
DESOBJ ( ) = N
MIN

Examples:
DESOBJ = 10
DESO = 25

Describer Meaning
MIN Specifies that the objective is to be minimized.

MAX Specifies that the objective is to be maximized.

n Set identification of a DRESP1 or DRESP2 Bulk Data entry. (Integer > 0)

Remarks:
1. A DESOBJ command is required for a design optimization task and is optional for a sensi-
tivity task.

2. If the DESOBJ command is specified within a SUBCASE, the identified DRESPi Bulk Data
entry use a response only from that subcase. If DESOBJ appears above all SUBCASE
commands and there are multiple subcases, it uses a global response.

3. The referenced DRESPi entry must define a scalar response.

A-7
DESSUB Design Constraints Request at the Subcase Level

DESSUBthe design constraints to be used in a design optimization

Select task for the
Design Constraints current
Request subcase.
at the Subcase Level

Format:
DESSUB = n

Examples:
DESSUB = 10
DESS = 25

Describer Meaning
n Set identification of a DCONSTR or DCONADD Bulk Data entry identification
number. (Integer ≥ 0)

Remark:
1. A DESSUB command is required for every subcase for which constraints are to be
applied.

A-8
MODTRAK Mode Tracking Request

MODTRAKmode tracking options in design optimization (SOL 200).

Selects Mode Tracking Request

Format:
MODTRAK = n

Example:
MODTRAK = 100

Describer Meaning
n Set identification of a MODTRAK Bulk Data entry. (Integer > 0)

Remark:
1. Selection of a MODTRAK Bulk Data entry with the MODTRAK Case Control command
activates mode tracking for the current subcase. This request is limited to normal modes
subcases (ANALYSIS = MODES) in design optimization (SOL 200).

A-9
BULK DATA ENTRIES

A-10
Boundary Grid Points BNDGRID
BNDGRIDa list of grid point identification numbers on design boundaries or surfaces
Specifies for shape
Boundary opti-
Grid Points
mization (SOL 200).

Format:

1 2 3 4 5 6 7 8 9 10
BNDGRID C GP1 GP2 GP3 GP4 GP5 GP6 GP7

GP8 -etc.-

Example:

BNDGRID 123 41 42 43 44 45 46 47

49

Alternate Format and Example:

BNDGRID C GP1 "THRU" GP2

BNDGRID 123 41 THRU 49

Field Contents
C Component number (any unique combination of integers 1 through 6 with no
embedded blanks). See Remark 1.

GPi Shape boundary grid point identification number. (0 < Integer < 1000000; For
THRU option, GP1< GP2)

Remarks:
1. C specifies the components for the listed grid points for which boundary motion is
prescribed.

2. Multiple BNDGRID entries may be used to specify the shape boundary grid point identifi-
cation numbers.

3. Both fixed and free shape boundary grid point identification numbers are listed on this
entry.

4. The degrees of freedom specified on BNDGRID entries must be sufficient to statically con-
strain the model.

5. Degrees of freedom specified on this entry form members of the mutually exclusive s-set.
They may not be specified on other entries that define mutually exclusive sets. See the
MSC/NASTRAN Quick Reference Guide, Appendix B for a list of these entries.

A-11
DCONADD Design Constraint Set Combination

DCONADDthe design constraints for a subcase as a union of DCONSTRDesign

Defines entries.
Constraint Set Combination

Format:

1 2 3 4 5 6 7 8 9 10
DCONADD DCID DC1 DC2 DC3 DC4 DC5 DC6 DC7

DC8 -etc.-

Example:

DCONADD 10 4 12

Field Contents
DCID Design constraint set identification number. (Integer > 0)

DCi DCONSTR entry identification number. (Integer > 0)

Remarks:
1. The DCONADD entry is selected by a DESSUB or DESGLB Case Control command.

2. All DCi must be unique from other DCi and DCID.

A-12
Design Constraints DCONSTR
DCONSTR
Define design constraints. Design Constraints

Format:

1 2 3 4 5 6 7 8 9 10
DCONSTR DCID RID LALLOW UALLOW

Example:

DCONSTR 10 4 1.25

Field Contents
DCID Design constraint set identification number. (Integer > 0)

RID DRESPi entry identification number. (Integer > 0)

LALLOW Lower bound on the response quantity. (Real, Default = –1.0E20)

UALLOW Upper bound on the response quantity. (Real, Default = 1.0E20)

Remarks:
1. The DCONSTR entry may be selected in the Case Control Section by the DESSUB or
DESGLB command.

2. DCID may be referenced by the DCONADD Bulk Data entry.

3. For a given DCID, the associated RID can be referenced only once.

4. The units of LALLOW and UALLOW must be consistent with the referenced response de-
fined on the DRESPi entry. If RID refers to an “EIGN” response, then the imposed bounds
must be expressed in units of eigenvalue, (radian/time)2. If RID refers to a “FREQ” re-
sponse, then the imposed bounds must be expressed in cycles/time.

5. LALLOW and UALLOW are unrelated to the stress limits specified on the MATi entry.

6. Constraints are computed as follows:

LALLOW – r
g = -------------------------------- for lower bound constraints
GNORM

r – UALLOW
g = --------------------------------- for upper bound constraints
GNORM

where r is the response defined on the DRESPi entry and if the RID refers to a “FREQ”
response, then the imposed bounds must be expressed in cycles/time.

(Continued)

A-13
DCONSTR Design Constraints

LALLOW for lower bounds if LALLOW > GSCAL

GNORM = UALLOW for upper bounds if UALLOW > GSCAL
GSCAL otherwise

GSCAL is specified on the DOPTPRM entry (Default = 0.001)

7. As Remark 6 indicates, small values of UALLOW and LALLOW require special processing
and should be avoided. Bounds of exactly zero are particularly troublesome. This can be
avoided by using a DRESP2 entry that offsets the constrained response from zero.

A-14
Design Equation Definition DEQATN
DEQATN one or more equations for use in design sensitivity or p-element analysis.
Defines Design Equation Definition

Format:

1 2 3 4 5 6 7 8 9 10
DEQATN EQID EQUATION

EQUATION (Cont.)

Example:

DEQATN 14 F1(A, B, C, D, R) = A + B ∗ C – (D ∗ ∗ 3 + 10.0) + sin(PI(1) ∗ R)

+ A ∗ ∗ 2/(B – C); F = A + B – F1 ∗ D

Field Contents
EQID Unique equation identification number. (Integer > 0)

EQUATION Equation(s). See Remarks. (Character)

Remarks:
1. EQUATION is a single equation or a set of nested equations and is specified in fields 3
through 9 on the first entry and may be continued on fields 2 through 9 on the continuation
entries. On the continuation entries, no commas can appear in columns 1 through 8. All
data in fields 2 through 9 must be specified in columns 9 through 72. The large-field for-
mat is not allowed.

A single equation has the following format:

variable-1 (x1, x2, ..., xn)+expression-1

A set of nested equations is separated by semicolons and has the format:

variable-1 (x1, x2, ..., xn) = expression-1;

variable-2 = expression-2;
variable-3 = expression-3;
etc...................
variable-m = expression-m

Expression-i is a collection of constants, real variables, and real functions, separated by

operators, and must produce a single real value. (x1, x2, ..., xn) is the list of all the vari-
able names (except variable-i) that appear in all expressions. Variable-i may be used in
subsequent expressions. The last equation, variable-m = expression-m, provides the val-
ue that is returned to the Bulk Data entry that references EQID; e.g., DRESP2. The
example above represents the following mathematical equations:

(Continued)

A-15
DEQATN Design Equation Definition

2
3 A
1 = A + B ∗ C – ( D + 10 ) + sin ( PI ( 1 ) ∗ R ) + --------------
B–C
F = A + B – F1 ∗ D

where SIN and PI are intrinsic functions. See Remark 4.

2. EQUATION may contain embedded blanks. EQUATION must contain less than 12,500
nonblank characters. This is equivalent to approximately 195 continuation entries.

3. The syntax of the expressions follows FORTRAN language standards. The allowable
arithmetic operations are shown in Table 1 in the order of execution precedence. Paren-
thesis are used to change the order of precedence. Operations within parentheses are
performed first with the usual order of precedence being maintained within the
parentheses.

Table 1. DEQATN Entry Operators.

Operator Operation Sample Expressions Interpreted As

–, + Negative or Positive immediately X∗∗–Y X∗∗(–Y)
preceded by exponentiation

∗∗ Exponentiation –X∗∗Y (–X∗∗Y)

–, + Negative or Positive –X–Y (–X)–Y

∗, / Multiplication or Division X∗Y+Z (X∗Y)+Z

+, – Addition or Subtraction X+Y X+Y

4. The expressions may contain intrinsic functions. Table 2 contains the format and descrip-
tions of functions that may appear in the expressions. The use of functions that may be
discontinuous must be used with caution because they can cause discontinuous deriva-
tives. These are ABS, DIM, MAX, MIN, and MOD. For examples and further details see
the MSC/NASTRAN DMAP Module Dictionary.

(Continued)

A-16
Design Equation Definition DEQATN
Table 2. DEQATN Entry Functions.

Format Description Mathematical Expression

ABS(x) absolute value |x|

ACOS(x) arccosine cos-1 x

ACOSH(x) hyperbolic arccosine cosh-1 x
ASIN(x) arcsine sin-1 x
ASINH(x) hyperbolic arcsine sinh-1 x
ATAN(x) arctangent tan-1 x
ATAN2(x,y) arctangent of quotient tan-1 (x/y)
ATANH(x) hyperbolic arctangent tanh-1 x
ATANH2(x,y) hyperbolic arctangent of quotient tanh-1 (x/y)
COS(x) cosine cos x

COSH(x) hyperbolic cosine cosh x

DIM(x,y) positive difference x – MIN(x,y)

EXP(x) exponential ex
LOG(x) natural logarithm loge x

LOG10(x) common logarithm log10 x

LOGX(x,y) base x logarithm logx y

MAX(x1, x2, ...) maximum maximum of x1, etc.

MIN(x1, x2, ...) minimum minimum of x1, etc.

MOD(x,y) remainder (modulo) x – y ∗ (INT(x/y))

PI(x) multiples of pi (p) x∗π

SIN(x) sine sin x

SINH(x) hyperbolic sine sinh x

SQRT(x) square root x

TAN(x) tangent tan x

TANH(x) hyperbolic tangent tanh x

(Continued)

A-17
DEQATN Design Equation Definition

5. If the DEQATN entry is referenced by the:

a. DVPREL2 entry, then xi represents the DVIDj and LABLk fields.

b. DRESP2 entry, then xi represents the DVIDj, LABLk, NRm, and Gp fields in that
order.

c. GMLOAD, GMBC, or TEMPF entries, then

x1 represents x in the basic coordinate system,
x2 represents y in the basic coordinate system, and
x3 represents z in the basic coordinate system.

d. GMCURV entry, then

x1 represents line parameter u.

e. GMSURF entry, then

x1 represents surface parameter u and
x2 represents surface parameter v.

6. If the DEQATN entry is referenced by the GMLOAD, GMBC, TEMPF, GMCURV, or

GMSURF entries and your computer has a short word length (e.g., 32 bits/word), then
EQUATION is processed with double precision and constants may be specified in double
precision; e.g., 1.2D0. If your machine has a long word length (e.g., 64 bits/word) then
EQUATION is processed in single precision and constants must be specified in single pre-
cision; e.g., 1.2.

7. The DMAP logical operators NOT, AND, OR < XOR, and XQV cannot be used as Xi
names.

8. Input errors on the DEQATN entry often result in poor messages. Substituting a “[” for a
parenthesis or violating the restriction against large field format are examples. Known
messages are UFM 215, SFM 233 and UFM 5199. If any of these messages are encoun-
tered then review the DEQATN entry input.

A-18
Design Variable DESVAR
Defines
DESVAR a design variable for design optimization. Design Variable

Format:

1 2 3 4 5 6 7 8 9 10
DESVAR ID LABEL XINIT XLB XUB DELXV

Example:

DESVAR 2 BARA1 35.0 10. 100. 0.2

Field Contents
ID Unique design variable identification number. (Integer > 0)

LABEL User-supplied name for printing purposes. (Character)

XINIT Initial value. (Real, XLB ≤ XINIT ≤ XUB)

XLB Lower bound. (Real, Default = –1.0E+20)

XUB Upper bound. (Real, Default= +1.0E+20)

DELXV Fractional change allowed for the design variable during approximate
optimization. (Real > 0.0, for Default see Remark 2.)

Remarks:
1. DELXV can be used to control the change in the design variable during one optimization
cycle.

2. If DELXV is blank, the default is taken from the specification of the DELX parameter on
the DOPTPRM entry. If DELX is not specified, then the default is 1.0.

A-19
DLINK Multiple Design Variable Linking

Relates
DLINK one design variable to one or more other design variables. Multiple Design Variable Linking

Format:

1 2 3 4 5 6 7 8 9 10
DLINK ID DDVID C0 CMULT IDV1 C1 IDV2 C2

IDV3 C3 -etc.-

Example:

DLINK 10 2 0.1 0.33 2 2.0 6 –1.0

8 7.0

Field Contents
ID Unique identification number. (Integer > 0)

DDVID Dependent design variable identification number. (Integer > 0)

C0 Constant term. (Real; Default = 0.0)

CMULT Constant multiplier. (Real; Default = 1.0)

IDVi Independent design variable identification number. (Integer > 0)

Ci Coefficient i corresponding to IDVi. (Real)

Remarks:
1. DLINK defines the relationship

DDVID = C0 + CMULT ∑i Ci ∗ IDVi

2. This capability provides a means of linking physical design variables such as element
thicknesses to nonphysical design variables such as the coefficients of interpolating
functions.

3. CMULT provides a simple means of scaling the Ci. For example if Ci = 1/7, 2/7, 4/7, etc.
is desired, then CMULT = 1/7 and Ci = 1, 2, 4, etc., may be input.

4. An independent IDVi must not occur on the same DLINK entry more than once.

5. ID is for user reference only.

A-20
Design Optimization Parameters DOPTPRM
DOPTPRM default values of parameters used in design optimization.
Overrides Design Optimization Parameters

Format:

1 2 3 4 5 6 7 8 9 10
DOPTPRM PARAM1 VAL1 PARAM2 VAL2 PARAM3 VAL3 PARAM4 VAL4

PARAM5 VAL5 -etc.-

Example:

DOPTPRM IPRINT 5 DESMAX 10

Field Contents
PARAMi Name of the design optimization parameter. Allowable names are given in
Table 1. (Character)
VALi Value of the parameter. (Real or Integer, see Table 1.)

Remarks:
1. Only one DOPTPRM entry is allowed in the Bulk Data Section.

Table 1. PARAMi Names and Descriptions.

Name Description, Type, and Default Value

APRCOD Approximation method to be used. 1 = Direct Linearization; 2 = Mixed Method

based on response type; 3 = Convex Linearization. APRCOD = 1 is recommended
for shape optimization problems. (Integer 1, 2, or 3; Default = 2)

CONV1 Relative criterion to detect convergence. If the relative change in objective

between two optimization cycles is less than CONV1, then optimization is termi-
nated. (Real > 0.0; Default = 0.001)

CONV2 Absolute criterion to detect convergence. If the absolute change in objective

between two optimization cycles is less than CONV2, then optimization is termi-
nated. (Real > 0.0; Default = 1.0E-20)

CONVDV Relative convergence criterion on design variables. (Real > 0.0; Default = 0.001)

CONVPR Relative convergence criterion on properties. (Real > 0.0; Default = 0.001)

(Continued)

A-21
DOPTPRM Design Optimization Parameters

Name Description, Type, and Default Value

CT Constraint tolerance. Constraint is considered active if current value is greater

than CT. (Real < 0.0; Default = –0.03)

CTMIN Constraint is considered violated if current value is greater than CTMIN.

(Real > 0.0; Default = 0.003)

DABOBJ Maximum absolute change in objective between ITRMOP consecutive iterations

(see ITRMOP) to indicate convergence at optimizer level. F0 is the initial objective
function value. (Real > 0.0; Default = MAX[0.00 ∗ ABS(F0), 0.0001])

DELB Relative finite difference move parameter. (Real > 0.0; Default = 0.0001)

DELOBJ Maximum relative change in objective between ITRMOP consecutive iterations to

indicate convergence at optimizer level. (Real > 0.0; Default = 0.001)

DELP Fractional change allowed in each property during any optimization design cycle.
This provides constraints on property moves. (Real > 0.0; Default = 0.2)

DELX Fractional change allowed in each design variable during any optimization cycle.
(Real > 0.0; Default = 1.0)

DESMAX Maximum number of design cycles to be performed. (Integer > 0; Default = 5)

DOBJ1 Relative change in objective attempted on the first optimization iteration. Used to
estimate initial move in the one-dimensional search. Updated as the optimization
progresses. (Real > 0.0; Default = 0.1)

DOBJ2 Absolute change in objective attempted on the first optimization iteration. (Real
> 0.0; Default = 0.2 ∗ (F0))

DPMIN Minimum move limit imposed. (Real > 0.0; Default = 0.01)

DX1 Maximum relative change in a design variable attempted on the first optimization
iteration. Used to estimate the initial move in the one dimensional search. Updated
as the optimization progresses. (Real > 0.0; Default = 0.01)

DX2 Absolute change in a design variable attempted on the first optimization iteration.
(Real > 0.0; Default = 0.2 ∗ MAX[X(I)])

DXMIN Minimum design variable move limit (Real > 0.0; Default = 0.05).

GMAX Maximum constraint violation allowed at the converged optimum. (Real > 0.0;
Default = 0.005)

GSCAL Constraint normalization factor. See Remarks under the DSCREEN and
DCONSTR entries. (Real > 0.0; Default = 0.001)

(Continued)

A-22
Design Optimization Parameters DOPTPRM

Name Description, Type, and Default Value

IGMAX If IGMAX = 0, only gradients of active and violated constraints are calculated. If
IGMAX > 0, up to NCOLA gradients are calculated including active, violated, and
near active constraints. (Integer > 0; Default = 0)

IPRINT Print control during approximate optimization phase. Increasing values represent
increasing levels of optimizer information. (0 ≤ Integer ≤ 7; Default = 0)
0 no output (Default)
1 internal optimization parameters, initial information, and results
2 same, plus objective function and design variables at each iterations
3 same, plus constraint values and identification of critical constraints
4 same, plus gradients
5 same, plus search direction
6 same, plus scaling factors and miscellaneous search information
7 same, plus one dimensional search information

IPRNT1 If IPRNT1 = 1, print scaling factors for design variable vector. (Integer 0 or 1;
Default = 0)

IPRNT2 If IPRNT2 = 1, print miscellaneous search information. If IPRNT2 = 2, turn on print

during one-dimensional search process. (Warning: This may lead to excessive
output.) (Integer 0, 1, or 2; Default = 0)

ISCAL Design variables are rescaled every ISCAL iterations. Set ISCAL= –1 to turn off
scaling. (Integer; Default = NDV (number of design variables))

ITMAX Maximum number of iterations allowed at optimizer level during each design cycle.
(Integer; Default = 40)

ITRMOP Number of consecutive iterations for which convergence criteria must be satisfied
to indicate convergence at the optimizer level. (Integer; Default = 2)

ITRMST Number of consecutive iterations for which convergence criteria must be met at
the optimizer level to indicate convergence in the Sequential Linear Programming
Method. (Integer > 0; Default = 2)

JTMAX Maximum number of iterations allowed at the optimizer level for the Sequential
Linear Programming Method. This is the number of linearized subproblems
solved. (Integer ≥ 0; Default = 20)

JPRINT Sequential Linear Programming subproblem print. If JPRINT > 0, IPRINT is turned
on during the approximate linear subproblem. (Default = 0)

JWRITE If JWRITE > 0, file number on which iteration history will be written. (Integer > 0;
Default = 0)

(Continued)

A-23
DOPTPRM Design Optimization Parameters

Name Description, Type, and Default Value

METHOD Optimization Method: (Integer 1, 2, or 3; Default = 1)

1: Modified Method of Feasible Directions. (Default)
2: Sequential Linear Programming
3: Sequential Quadratic Programming

P1 Print control items specified for P2. (Integer ≥ 0; Default = 0) Initial results are
always printed prior to the first approximate optimization. If an optimization task
is performed, final results are always printed for the final analysis unless
PARAM,SOFTEXIT,YES is specified. These two sets of print are not controllable.
n: Print at every n-th design cycle.

P2 Items to be printed according to P1: (Integer; Default = 1)

0: No print.
1: Print objective and design variables. (Default)
2: Print properties.
4: Print constraints.
8: Print responses.
16: Print weight as a function of a material ID (note that there is not a design
quantity so that only inputs to the approximate design are available).
n: Sum of desired items. For example, P2 = 10 means print properties and
responses.

PTOL Maximum tolerance on differences allowed between the property values on

property entries and the property values calculated from the design variable values
on the DESVAR entry (through DVPRELi relations). PTOL is provided to trap ill-
posed design models. (The minimum tolerance may be specified on user
parameter DPEPS. See MSC/NASTRAN Quick Reference Guide, Section 6.)
(Real > 0.0; Default = 1.0E+35)

STPSCL Scaling factor for shape finite difference step sizes, to be applied to all shape
design variables. (Real > 0.0; Default = 1.0)

A-24
Design Sensitivity Response Quantities DRESP1
Defines a set of structural responses that is used in the designDesign
DRESP1 either as constraints
Sensitivity Responseor as an
Quantities
objective.

Format:

1 2 3 4 5 6 7 8 9 10
DRESP1 ID LABEL RTYPE PTYPE REGION ATTA ATTB ATT1

ATT2 -etc.-

Example:

DRESP1 1 DX1 STRESS PROD 2 3 102

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. See Remark 10 for defaults.
(Integer > 0)

ATTA, ATTB, Response attributes. See Table 1. (Integer > 0 or Real or blank)
ATTi

(Continued)

A-25
DRESP1 Design Sensitivity Response Quantities

Table 1. Design Sensitivity Response Attributes.

Response Attributes
Response
Type ATTB (Integer > 0 or
(RTYPE) ATTA (Integer > 0) Real > 0.0) ATTi (Integer > 0)

WEIGHT Blank Blank SEIDi or ALL

VOLUME Blank Blank SEIDi or ALL

EIGN Normal Modes Mode Approximation Code. Blank

Number See Remark 19.

FREQ Normal Modes Mode Approximation Code. Blank

Number See Remark 19.
See Remark 18.

LAMA Buckling Mode Number Approximation Code. Blank

See Remark 19.

DISP Displacement Component Blank Grid ID

STRAIN Strain Item Code Blank Property entry (PID)

STRESS Stress Item Code Blank Property entry (PID)

FORCE Force Item Code Blank Property entry (PID)

CSTRAIN Strain Item Code LAMINA Number Property entry (PID)

(Integer; Default = 1)

CSTRESS Stress Item Code LAMINA Number Property entry (PID)

(Integer; Default = 1)

CFAILURE Failure Criterion Item LAMINA Number Property entry (PID)

Code (Integer; Default = 1)

FRDISP Displacement Component Frequency Value. (Real > Grid ID

0.0) See Remark 15.
FRVELO Velocity Component Frequency Value. (Real > Grid ID
0.0) See Remark 15.

FRACCL Acceleration Component Frequency Value. (Real > Grid ID

0.0) See Remark 15.
FRSPCF SPC Force Component Frequency Value. (Real > Grid ID
0.0) See Remark 15.

FRSTRE Stress Item Code Frequency Value. (Real > Property entry (PID)
0.0) See Remark 15..

(Continued)

A-26
Design Sensitivity Response Quantities DRESP1

Response Attributes
Response
Type ATTB (Integer > 0 or
(RTYPE) ATTA (Integer > 0) Real > 0.0) ATTi (Integer > 0)

FRFORC Force Item Code Frequency Value. (Real > Property entry (PID)
0.0) See Remark 15.

TDISP Displacement Component Time Value. (Real) See Grid ID

Remark 16.

TVELO Velocity Component Time Value. (Real) See Grid ID

Remark 16.

TACCL Acceleration Component Time Value. (Real) See Grid ID

Remark 16.

TSPCF SPC Force Component Time Value. (Real) See Grid ID

Remark 16.

TSTRE Stress Item Code Time Value. (Real) See Property entry (PID)
Remark 16.

TFORC Force Item Code Time Value. (Real) See Property entry (PID)
Remark 16.

TRIM AESTAT or AESURF Blank Blank

Entry ID

STABDER AESTAT or AESURF Restraint Flag. (Integer 0 Component

Entry ID or 1) See Remark 13.

FLUTTER Blank Method (PK or PKNL, See Remark 14.

Default = PK).

Remarks:
1. Stress, strain, and force item codes can be found in the MSC/NASTRAN Quick Reference
Guide, Appendix A. For stress or strain item codes that have dual meanings, such as von
Mises or maximum shear, the option specified in the Case Control Section will be used;
i.e., STRESS(VONM) or STRESS(MAXS).

2. RTYPE=“CSTRESS”, “CSTRAIN”, and “CFAILURE” are used only with the PCOMP entry.
“CSTRESS” and “CSTRAIN” item codes are described under Table 1. (Element
Stress/Strain Item Codes) in the MSC/NASTRAN Quick Reference Guide, Appendix A.
“CFAILURE” item codes are described under Table Table 2. (Element Force Item Codes)
in the MSC/NASTRAN Quick Reference Guide, Appendix A. Only force item codes that
refer to failure indices of direct stress and interlaminar shear stress are valid.

(Continued)

A-27
DRESP1 Design Sensitivity Response Quantities

The CFAILURE response type requires the following specifications on the applicable
entries:

a. Failure theory in the FT field on PCOMP entry.

b. Allowable bonding shear stress in the SB field on PCOMP entry.

c. Stress limits in the ST, SC, and SS fields on all MATi entries.

3. ATTB is used only for responses of composite laminae, dynamics, and stability deriva-
tives. For other responses, this field must be blank.

4. All grids associated with a DRESP1 entry are considered to be in the same region for
screening purposes. Only up to NSTR displacement constraints (see DSCREEN entry)
per group per load case will be retained in the design optimization phase.

5. DRESP1 identification numbers must be unique with respect to DRESP2 identification

numbers.

6. If PTYPE = “ELEM”, the ATTi correspond to element identification numbers.

7. If RTYPE = “DISP”, “TDISP”, “TVELO”, “TACCL” or “TSPCF”, multiple component num-

bers (any unique combination of the digits 1 through 6 with no embedded blanks) may be
specified on a single entry. Multiple response components may not be used on any other
response types.

8. If RTYPE = “FRDISP”, “FRVELO”, “FRACCL”, or “FRSPCF” only one component number

may be specified in the ATTA field. Numbers 1 through 6 correspond to real (or magni-
tude) components and 7 through 12 imaginary (or phase) components. If more than one
component for the same grid is desired, then a separate entry is required.

9. Real/imaginary representation is the default for complex response types. Magni-

tude/phase representation must be requested by the corresponding Case Control
command; e.g., DlSP(PHASE) = ALL.

10. REGION is used for constraint screening. The NSTR field on DSCREEN entries gives the
maximum number of constraints retained for each region per load case.

IF RTYPE = “WEIGHT”, “VOLUME”, “LAMA”, “EIGN” or “FREQ”, no REGION identification

number should be specified. For all other responses, if the REGION field is left blank, the
default specified in Table 2 is used. Usually, the default value is appropriate.

If the REGION field is not blank, all the responses on this entry as well as all responses on
other DRESP1 entries that have the same RTYPE and REGION identification number will
be grouped into the same region.

(Continued)

A-28
Design Sensitivity Response Quantities DRESP1
Table 2. Default Regions for Design Sensitivity Response Types.

Response Type Default Region

WEIGHT No region

VOLUME No region

LAMA No region

EIGN No region

FREQ No region

DISP One region per DRESP1 entry

FRDISP One region per DRESP1 entry

FRVELO One region per DRESP1 entry

FRACCL One region per DRESP1 entry

FRSPCF One region per DRESP1 entry

TDISP One region per DRESP1 entry

TVELO One region per DRESP1 entry

TACCL One region per DRESP1 entry

TSPCF One region per DRESP1 entry

FLUTTER One region per DRESP1 entry

OTHER One region per PROPERTY entry. If PTYPE = “ELEM”,

then one region per DRESP1 entry

11. REGION is valid only among the same type of responses. Responses of different types
will never be grouped into the same region, even if they are assigned the same REGION
identification number by the user.

12. If RTYPE = “WEIGHT” or “VOLUME”, field ATTi = “ALL” implies total weight/volume of all
superelements except external superelements.

13. RTYPE = “STABDER” identifies a stability derivative response. ATTB is the restraint flag
for the stability derivative. ATTB = 0 means unrestrained, and ATTB = 1 means re-
strained. For example, ATTA = 4000, ATTB = 0, and ATT1 = 3 reference the unrestrained
Cz derivative for the AESTAT (or AESURF) entry ID = 4000.

(Continued)

A-29
DRESP1 Design Sensitivity Response Quantities

14. RTYPE = “FLUTTER” identifies a set of damping responses. The set is specified by ATTi:

ATT1 = Identification number of a SET1 entry that specifies a set of modes.

ATT2 = Identification number of an FLFACT entry that specifies a list of densities.

ATT3 = Identification number of an FLFACT entry that specifies a list of Mach numbers.

ATT4 = Identification number of an FLFACT entry that specifies a list of velocities.

15. For RTYP = “FRDISP”, “FRVELO”, “FRACCL”, “FRSPCF”, “FRFORC”, and “FRSTRE”,
ATTB specifies a frequency value in cycles per unit time. If ATTB is specified, then the
responses are evaluated at the closest frequency selected by the OFREQ command. The
default for ATTB is all frequencies selected by the OFREQ command.

16. For RTYPE = “TDISP”, “TVELO”, “TACCL”, “TSPCF”, “TFORC”, and “TSTRE”, ATTB
specifies a time value. If ATTB is specified, then the responses are evaluated at the clos-
est time selected by the OTIME command. The default for ATTB is all time steps selected
by the OTIME command.

17. Intermediate station responses on CBAR elements due to PLOAD1 and/or CBARAO en-
tries may not be defined on the DRESP1 entry.

18. RTYPE = “EIGN” refers to normal modes response in terms of eigenvalue (radi-
an/time)∗∗2 while RTYPE = “FREQ” refers to normal modes response in terms of natural
frequency or units of cycles per unit time.

19. For RTYPE = LAMA, EIGN or FREQ, the response approximation used for optimization
can be individually selected. (Approximation Code = 1 = direct linearization, = 2 = Inverse
Linearization).

A-30
Design Sensitivity Equation Response Quantities DRESP2
DRESP2 equation responses that are used in the design, either
Defines asSensitivity
Design constraints or asResponse
Equation an objective.
Quantities

Format:

1 2 3 4 5 6 7 8 9 10
DRESP2 ID LABEL EQID REGION

“DESVAR” DVID1 DVID2 DVID3 DVID4 DVID5 DVID6 DVID7

DVID8 -etc.-

“DTABLE” LABL1 LABL2 LABL3 LABL4 LABL5 LABL6 LABL7

LABL8 -etc.-

“DRESP1” NR1 NR2 NR3 NR4 NR5 NR6 NR7

NR8 -etc.-

“DNODE” G1 C1 G2 C2 G3 C3

G4 C4 -etc.-

“DVPREL1” DPIP1 DPIP2 DPIP3 DPIP4 DPIP5 DPIP6 DPIP7

DPIP8 DPIP9 -etc.-

Example:

DRESP2 1 LBUCK 5 3

DESVAR 101 3 4 5 1 205 209

201

DTABLE PI YM L

DRESP1 14 1 4 22 6 33 2

DNODE 14 1 4 1 22 3

2 1 43 1

DVPREL1 101 102

Field Contents
ID Unique identification number. (Integer > 0)

LABEL User-defined label. (Character)

(Continued)

A-31
DRESP2 Design Sensitivity Equation Response Quantities

EQID DEQATN entry identification number. (Integer > 0)

REGION Region identifier for constraint screening. See Remark 5. (Integer > 0)

“DESVAR” Flag indicating DESVAR entry identification numbers. (Character)

DVIDi DESVAR entry identification number. (Integer > 0)

“DTABLE” Flag indicating that the labels for the constants in a DTABLE entry follow.
(Character)

LABLj Label for a constant in the DTABLE entry. (Character)

“DRESP1” Flag indicating DRESP1 entry identification numbers. (Character)

NRk DRESP1 entry identification number. (Integer > 0)

“DNODE” Flag signifying that the following fields are “designed grid points.” See Remark 7.
(Character)

Gm Grid point identification number. (Integer > 0)

Cm Degree of freedom number of grid point Gm. (1 ≤ Integer ≤ 3)

DVPREL1 Flag indicating DVPREL1 entry identification number. (Character)

DPIPi DVPREL1 entry identification number. (Integer > 0)

Remarks:
1. DRESP2 entries may only reference DESVAR, DTABLE, DRESP1, and DVPREL1 en-
tries. They may not reference other DRESP2 entries.

2. Referenced DRESP1 entries cannot span analysis types or superelements.

3. DRESP2 entries must have unique identification numbers with respect to DRESP1
entries.

4. The “DESVAR”, “DTABLE”, “DRESP1”, “DNODE”, and “DVPREL1” flags in field 2 must
appear in the order given above. Any of these words, along with the identification num-
bers associated with them, may be omitted if they are not involved in this DRESP2
relationship. However, at least one of these four types of arguments must exist.

5. The REGION field follows the same rules as for the DRESP1 entries. DRESP1 and
DRESP2 responses will never be contained in the same region, even if they are assigned
the same REGION identification number. The default is to put all responses referenced
by one DRESP2 entry in the same region.

(Continued)

A-32
Design Sensitivity Equation Response Quantities DRESP2
6. The variables identified by DVIDi, LABLj, NRk, the Gm, Cm pairs and DPIPi are assigned
(in that order) to the variable names (x1, x2, x3, etc.) specified in the left-hand side of the
first equation on the DEQATN entry referenced by EQID. The variable names x1 through
xN (N = m + n + p + q) are assigned in the order DVID1, ..., DVIDm, LABL1, ..., LABLn,
NR1, ..., NRp, G1, ..., Gq, DPIP1, ..., DPIPr. In the example below,

DESVARs 101 and 3 are assigned to arguments A and B.

DTABLEs PI and YM are assigned to arguments C and D.
Grid 14, Component 1 is assigned to argument R.

DRESP2 1 LBUCK 5 3

DESVAR 101 3

DTABLE PI YM

DNODE 14 1

DEQATN 5 F1(A, B, C, D, R) = A + B ∗ C – (D ∗ ∗ 3 + 10.0) + sin(C ∗ R)

7. (Gm, Cm) refer to a designed grid component. Depending on the scheme used in gener-
ating basis vectors, a designed grid component can be one of the following:

a. A component defined by a DVGRID entry if manual input of DVGRID entries is used.

b. A grid component that is free to move in a full auxiliary model when the external dis-
placement fields are used.

c. When the auxiliary model boundary shape method is used, a designed grid compo-
nent can be either a loaded grid component in the auxiliary model or a grid
component with motion that is obtained from interpolation of boundary shapes.

d. When the interface using geometric boundary shapes is used, a designed grid com-
ponent is either a grid component defined on a DVGRID entry or one with motion that
is obtained from interpolation of boundary shapes.

A-33
DSAPRT Design Sensitivity Output Parameters

Specifies
DSAPRT design sensitivity output parameters. Design Sensitivity Output Parameters

Format:

FORMATTED  ALL 
NOEXPORT  
DSAPRT ( UNFORMATTED , , [ START = i ], [ BY = j ], [ END = k ] ) =  n 
EXPORT  
NOPRINT  NONE 

Examples:
DSAPRT(FORMATTED,EXPORT)
DSAPRT(FORMATTED,START = FIRST,BY = 3,END = LAST) = 101
DSAPRT(UNFORMATTED,START = FIRST)
DSAPRT(UNFORMATTED,EXPORT)
DSAPRT(FORMATTED,END = 4) = ALL
DSAPRT(UNFORMATTED,END = SENS) = ALL
DSAPRT(NOPRINT,EXPORT)

Describers Meaning
FORMATTED Output will be presented with headings and labels.

UNFORMATTED Output will be printed as a matrix print (see description of the MATPRN module
in the MSC/NASTRAN DMAP Module Dictionary).

NOPRINT No output will be printed.

EXPORT Output will be exported to an external binary file specified by PARAM,IUNIT.

NOEXPORT Output will not be exported to an external binary file.

START = i Specifies the first design cycle for output. (Integer > 0 or Character: “FIRST” or
“LAST”; Default = 1 or “FIRST”)

BY = j Specifies the design cycle interval for output. (Integer ≤ 1, or > 0; Default = 0)
See Remark 2.

END = k Specifies the last design cycle for output. (Integer > 0 or Character: “FIRST”,
“LAST”, or “SENS”; Default = “LAST”)

ALL All design responses (defined in DRESP1 and DRESP2 entries) will be output.

n Set identification of a previously appearing SET command. Only sensitivities of

responses with identification numbers that appear on this SET command will be
output. (Integer > 0)

(Continued)

A-34
Design Sensitivity Output Parameters DSAPRT
Remarks:
1. Only one DSAPRT may appear in the Case Control Section and should appear above all
SUBCASE commands.

2. Sensitivity data will be output at design cycles i, i + j, i + 2j, ..., k. Note that the BY = 0
default implies no sensitivity analysis at the intermediate design cycles.

3. END = SENS requests design sensitivity analysis, and no optimization will be performed.

4. Prior to Version 69, PARAM,OPTEXIT,4, –4, or 7 requested design sensitivity output and
may still be used in Version 69. However, if both DSAPRT and PARAM,OPTEXIT,4, –4,
or 7 are specified, then DSAPRT overrides PARAM,OPTEXIT,4, –4, or 7.
PARAM,OPTEXIT values and the equivalent DSAPRT commands are as follows:

OPTEXIT Equivalent DSAPRT Command

4 DSAPRT(UNFORMATTED, END = SENS)

–4 DSAPRT(NOPRINT,EXPORT,END = SENS)

7 DSAPRT(UNFORMATTED,START = LAST)

A-35
A-36
DSCREEN Design Constraint Screening Data

DSCREENscreening data for constraint deletion.

Defines Design Constraint Screening Data

Format:

1 2 3 4 5 6 7 8 9 10
DSCREEN RTYPE TRS NSTR

Example:

DSCREEN STRESS –0.7 2

Field Contents
RTYPE Response type for which the screening criteria apply. See Table 1. (Character)

TRS Truncation threshold. (Real; Default = –0.5)

NSTR Maximum number of constraints to be retained per region per load case. See
Remark 3. (Integer > 0; Default = 20)

Remarks:
1. Displacement and displacement derivative constraints associated with one particular load
case are grouped by the specification of DRESP1 entries. From each group, a maximum
of NSTR constraints are retained per load case.

Table 1. Default Region Specifications for RTYPE.

Default Region Default Region

RTYPE Specification RTYPE Specification

DISP DRESP1 FRVELO DRESP1

STRESS Property ID FRACCL DRESP1
STRAIN Property ID FRSTRE Property ID
FORCE Property ID FRFORC Property ID
EQUA DRESP2 FRSPC DRESP1
CSTRESS Property ID TACCL DRESP1
CSTRAIN Property ID TDISP DRESP1
CFAILURE Property ID TVELO DRESP1
FLUTTER DRESP1 TSTRE Property ID
FRDISP DRESP1 TFORC Property ID
TSPC DRESP1

(Continued)

A-37
Design Constraint Screening Data DSCREEN
2. Stress-strain constraints are grouped by the property; i.e., all elements belonging to the
set of PIDs specified under ATTi on a DRESPi entry are regarded as belonging to the
same region. In superelement sensitivity analysis, if the property (PID) is defined in more
than one superelement, then separate regions are defined. A particular stress constraint
specification may be applied to many elements in a region generating many stress con-
straints, but only up to NSTR constraints per load case will be retained.

3. If RTYPE = “WEIGHT”, “VOLUME”, “EIGN”, “FREQ”, “LAMA”, “TRIM”, or “STABDER”

then NSTR is not applicable.

4. If a certain type of constraint exists but no corresponding DSCREEN entry is specified, all
the screening criteria used for this type of constraint will be furnished by the default
values.

5. Constraints can be retained only if they are greater than TRS. See the Remarks under
the DCONSTR entry for a definition of constraint value.

6. Constraint screening is applied to each superelement.

A-38
DTABLE Table Constants

DTABLE a table of real constants that are used in equations (see DEQATN entry).
Defines Table Constants

Format:

1 2 3 4 5 6 7 8 9 10
DTABLE LABL1 VALU1 LABL2 VALU2 LABL3 VALU3 LABL4 VALU4

LABL5 VALU5 LABL6 VALU6 -etc.-

Example:

DTABLE PI 3.142 H 10.1 E 1.0E6 T 0.1

G 5.5E5 B 100.

Field Contents
LABLi Label for the constant. (Character)

VALUi Value of the constant. (Real)

Remarks:
1. Only one DTABLE entry may be specified in the Bulk Data Section.

2. LABLi are referenced by the LABi on the DVPREL2 or DRESP2 entries.

A-39
Design Variable to Boundary Shapes DVBSHAP
Associates a design variable identification number to a linear combination
DVBSHAP of to
Design Variable boundary shape
Boundary Shapes
vectors from a particular auxiliary model.

Format:

1 2 3 4 5 6 7 8 9 10
DVBSHAP DVID AUXMOD COL1 SF1 COL2 SF2 COL3 SF3

Example:

DVBSHAP 4 1 1 1.6

Field Contents
DVID Design variable identification number of a DESVAR entry. (Integer > 0)

AUXMOD Auxiliary model identification number. (Integer > 0)

COLi Load sequence identification number from AUXMODEL Case Control command.
(Integer > 0)

SFi Scaling factor for load sequence identification number. (Real; Default = 1.0)

Remarks:
1. Design variable DVID must be defined on a DESVAR entry.

2. Multiple references to the same DVID and/or COLi will result in the vector addition of the
referenced boundary shape vectors.

3. Multiple DVBSHAP entries may be specified.

A-40
DVGRID Design Variable to Grid Point Relation

DVGRID the relationship between design variables and grid point locations.
Defines Design Variable to Grid Point Relation

Format:

1 2 3 4 5 6 7 8 9 10
DVGRID DVID GID CID COEFF N1 N2 N3

Example:

DVGRID 3 108 5 0.2 0.5 0.3 1.0

Field Contents
DVID DESVAR entry identification number. (Integer > 0)

GID Grid point (GRID) or geometric point (POINT) identification number. (Integer > 0)

CID Coordinate system identification number. (Integer ≥ 0; Default = 0)

COEFF Multiplier of the vector defined by Ni. (Real; Default = 0.0)

Ni Components of the vector measured in the coordinate system defined by CID.

(Real; at least one Ni ≠ 0.0.)

Remarks:
1. A CID of zero or blank (the default) references the basic coordinate system.

2. Multiple references to the same grid ID and design variable result in vectorial addition of
the participation vectors defined by CID, COEFF, and Ni. There is no restriction on the
number of DVGRID entries that may reference a given grid (GID) or design variable
(DVID).

3. The coordinate update equation is given as

∑i COEFF j ( X DVIDj – X DVIDj ){ N } j

0 0
{ g }i – { g }i =

T
where { g } i is the location of the i-th grid, [ g x g y g z ] .

T
The vector { N } = [ N x N y N z ] is determined from CID and Ni. Note that it is a change
0
in a design variable from its initial value X , and not the absolute value of the design
0
variable itself, that represents a change in a grid point location, { g } i – { g } i .

(Continued)

A-41
Design Variable to Grid Point Relation DVGRID
4. The DVGRID entry defines the participation coefficients (basis vectors) of each design
variable for each of the coordinates affected by the design process in the relationship

{ ∆g } i = ∑j { T }ij ∗ ∆X j
5. DVGRID entries that reference grid points on MPCs or RSSCON entries produce incorrect
sensitivities. Often the sensitivities are 0.0 which may result in a warning message indi-
cating zero gradients which may be followed by UFM 6499. Other rigid elements produce
correct results.

A-42
DVPREL1 Design Variable to Property Relation

DVPREL1 the relation between an analysis model property and design variables.
Defines Design Variable to Property Relation

Format:

1 2 3 4 5 6 7 8 9 10
DVPREL1 ID TYPE PID FID PMIN PMAX C0

DVID1 COEF1 DVID2 COEF2 DVID3 -etc.-

Example:

DVPREL1 12 PBAR 612 6 0.2 3.0

4 0.25 20 20.0 5 0.3

Field Contents
ID Unique identification number. (Integer > 0)

TYPE Name of a property entry, such as “PBAR”, “PBEAM”, etc. (Character)

PID Property entry identification number. (Integer > 0)

FlD Field position of the property entry, or word position in the element property table
of the analysis model. (Integer ≠ 0)

PMIN Minimum value allowed for this property. If FID references a stress recovery
location, then the default value for PMIN is –1.0+35. PMIN must be explicitly set
to a negative number for properties that may be less than zero (for example, field
ZO on the PCOMP entry). (Real; Default = 1.0E–20)

PMAX Maximum value allowed for this property. (Real; Default = 1.0E–20)

C0 Constant term of relation. (Real; Default = 0.0)

DVIDi DESVAR entry identification number. (Integer > 0)

COEFi Coefficient of linear relation. (Real)

Remarks:
1. The relationship between the analysis model property and design variables is given by:

P i = C0 + ∑i COEFi ∗ DVIDi
2. The continuation entry is required.

(Continued)

A-43
Design Variable to Property Relation DVPREL1
3. PTYPE = “PBEND” is not supported, either directly through FIDs or indirectly via word po-
sitions in the element property table.

4. FID may be either a positive or a negative number. If FID > 0, it identifies the field position
on a property entry. If FID < 0, it identifies the word position of an entry in the element
property table. For example, to specify the area of a PBAR, either FID = +4 or FID = –3
can be used. However, if PTYPE = “PBEAM”, FID must be negative. See the following
element property table for the word positions for PBEAM.

Table 2. EPT Section for PTYPE = “PBEAM”.

Word Type Item

1 I Property ID
2 I Material ID
3 I Number of segments
4 I Constant cross section flag (1 = yes, 0 = no)
5 I Unused
6 I Stress output request flag, SO (1 = yes, 0 = no)
7 R X/XB ratio; at end A, X/XB = 0.0
8 through 13 R A, I1, I2, I12, J, NSM
14 through 21 R C1, C2, D1, D2, E1, E2, F1, F2
22 I
23 R Repeat of words 6 through 21 for the 1st intermediate
24 through 29 R station
30 through 37 R
38 I
39 R
40 through 45 R 2nd intermediate station
46 through 53 R
54 I
55 R
56 through 61 R 3rd intermediate station
62 through 69 R
70 I
71 R
72 through 77 R 4th intermediate station
78 through 85 R
86 I
87 R
88 through 93 R 5th intermediate station
94 through 101 R

(Continued)

A-44
DVPREL1 Design Variable to Property Relation

Word Type Item

102 I
103 R
104 through 109 R 6th intermediate station
110 through 117 R
118 I
119 R
120 through 125 R 7th intermediate station
126 through 133 R
134 I
135 R
136 through 141 R 8th intermediate station
142 through 149 R
150 I
151 R
152 through 157 R 9th intermediate station
158 through 165 R
166 I
167 R
168 through 173 R End B
174 through 181 R
182 through 189 R K1, K2, S1, S2, NSI(A), NSI(B), CW(A), CW(B)
190 through 197 R M1(A), M2(A), M1(B), M2(B), N1(A), N2(A), N1(B),
N2(B)

A-45
Design Variable to Property Relation DVPREL2
Defines the relation between an analysis model property and design Design
DVPREL2 variables withtoaProperty
Variable user-supplied
Relation
equation.

Format:

1 2 3 4 5 6 7 8 9 10
DVPREL2 ID TYPE PID FID PMIN PMAX EQID

“DESVAR” DVID1 DVID2 DVID3 -etc.-

“DTABLE” LABL1 LABL2 LABL3 -etc.-

Example:

DVPREL2 13 PBAR 712 5 0.2 4

DESVAR 4 11 13 5

DTABLE PI YM

Field Contents
ID Unique identification number. (Integer > 0)

TYPE Name of a property entry, such as PBAR, PBEAM, etc. (Character)

PID Property entry identification number. (Integer > 0)

FID Field position of the property in the analysis model entry. (Integer ≠ 0)

PMIN Minimum value allowed for this property. If FID references a stress recovery
location field, then the default value for PMIN is –1.0+35. PMIN must be
explicitly set to a negative number for properties that may be less than zero (for
example, field ZO on the PCOMP entry). (Real; Default = 1.E–20)

PMAX Maximum value allowed for this property. (Real; Default = 1.0E20)

EQID DEQATN entry identification number. (Integer > 0)

“DESVAR” DESVAR flag. Indicates that the IDs of DESVAR entries follow. (Character)

DVIDi DESVAR entry identification number. (Integer > 0)

“DTABLE” DTABLE flag. Indicates that the IDs for the constants in a DTABLE entry follow.
This field may be omitted if there are no constants involved in this relation.
(Character)

LABLi Label for a constant on the DTABLE entry. (Integer > 0)

(Continued)

A-46
DVPREL2 Design Variable to Property Relation

Remarks:
1. The variables identified by DVIDi and LABLi correspond to variable names (x1, x2, etc.)
listed in the left-hand side of the first equation on the DEQATN entry identified by EQID.
The variable names x1 through xN (where N = m + n) are assigned in the order DVID1,
DVID2, ..., DVIDn, LABL1, LABL2, ..., LABLm.

2. If both “DESVAR” and “DTABLE” are specified in field 2, “DESVAR” must appear first.

3. FID may be either a positive or a negative number. If FID > 0, it identifies the field position
on a property entry. If FID < 0, it identifies the word position of an entry in EPT. For
example, to specify the area of a PBAR, either FID = +4 or FID = –3 may be used. How-
ever, if PTYPE = “PBEAM”, FID must be negative. See Remark 4 on the DVPREL1 entry
description for specification of the PBEAM element property tables.

4. PTYPE = “PBEND” is not supported, either directly through FIDs or indirectly via word po-
sitions in the element property table.

A-47
DVSHAP Design Variable to Basis Vector(s)

Defines a shape basis vector by relating a design variable identificationDesign

DVSHAP number (DVID)
Variable to columns
to Basis Vector(s)
of a displacement matrix.

Format:

1 2 3 4 5 6 7 8 9 10
DVSHAP DVID COL1 SF1 COL2 SF2 COL3 SF3

Example:

DVSHAP 2 1 2.0 4 1.0

Field Contents
DVID Design variable identification number on the DESVAR entry. (Integer > 0)

COLi Column number of the displacement matrix. See Remark 2. (1 ≤ Integer ≤

maximum column number in the displacement matrix.)

SFi Scaling factor applied to the COLi-th column of the displacement matrix. (Real;
Default = 1.0)

Remarks:
1. DVID must be defined on a DESVAR entry.

2. COLi must be a valid column number in the displacement matrix.

3. Multiple references to the same DVID and/or COLi will result in a linear combination of
displacement vectors. In the example above, the shape basis vector is a linear combina-
tion of the fourth column and twice the second column.

4. The displacement matrix must have been created by MSC/NASTRAN and be available on
a database, which is attached via the DBLOCATE FMS statement shown below:

ASSIGN DISPMAT=’ physical filename of MASTER DBset ’

DBLOCATE DATABLK=(UG/UGD,GEOM1/GEOM1D,GEOM2/GEOM2D) ,
LOGICAL=DISPMAT

A-48
Mode Tracking Parameters MODTRAK
MODTRAK parameters for mode tracking in design optimization (SOL 200).
Specifies Mode Tracking Parameters

Format:

1 2 3 4 5 6 7 8 9 10
MODTRAK SID LOWRNG HIGHRNG MTFILTER

Example:

MODTRAK 100 1 26 0.80

Field Contents
SID Sets identification number that is selected in the Case Control Section with the
MODTRAK command. (Integer; No Default) See Remark 1.

LOWRNG Lowest mode number in range to search. See Remark 2. (Integer ≥ 0,

Default = 0. If nonzero, LOWRNG < HIGHRNG.)

HIGHRNG Highest mode number in range to search. See Remark 2. (Integer > 0,
Default = number of eigenvalues extracted. If nonzero, LOWRNG < HIGHRNG.)

MTFILTER Filtering parameter used in mode cross-orthogonality check. See Remark 3.

(Real, Default = 0.9)

Remarks:
1. Only the designed modes for the subcase will be tracked. A designed mode is one that is
used in the design model (in connection with either objective or constraints) and, there-
fore, identified on a DRESP1 entry.

2. The range of modes LOWRNG through HIGHRNG, inclusive, will be used to track the de-
signed modes. If LOWRNG and HIGHRNG are both blank, then all computed modes will
be used to search for the designed modes. Since large numbers of computed modes will
result in higher computational costs, limiting the search range with LOWRNG and HIGH-
RNG is recommended.

3. Modes are considered to correlate if their mass normalized cross orthogonalities are
greater than MTFILTER.

A-49