Contents

Introduction

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-1

Design Optimization and Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1-1 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1-2 Design Optimization Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1-3 Sensitivity Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1-3 Features of the Design Optimization and Sensitivity Module (OPTSTAR) .1-4 Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1-8

Elements of Optimization and Sensitivity

. . . . . . . . . . . . 2-1

Design Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-1 Design Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-2 Objective Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-5 Behavior Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-5 Sensitivity Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-6 Sensitivity Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2-8

Procedures and Examples

. . . . . . . . . . . . . . . . . . . . . . . 3-1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3-1 Overview of Process for Design Optimization and Sensitivity . . . . . . . . . . .3-2 Overview of Commands for Design Optimization and Sensitivity . . . . . . . .3-3 Procedures for Performing Design Optimization . . . . . . . . . . . . . . . . . . . . .3-5 Procedures for Performing Sensitivity Studies . . . . . . . . . . . . . . . . . . . . . .3-12 Special Features for Optimization and Sensitivity . . . . . . . . . . . . . . . . . . .3-14

In de x

COSMOSM Advanced Modules

Contents

Numerical Aspects

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4-1 Basic Statements of Optimization Problems . . . . . . . . . . . . . . . . . . . . . . . . .4-2 Function Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4-2 Singular Value Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4-4 The Modified Feasible Direction Method . . . . . . . . . . . . . . . . . . . . . . . . . . .4-4 Overall Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4-4 Search Direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4-5 Convergence to the Optimum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4-6 Satisfaction of Kuhn-Tucker Conditions . . . . . . . . . . . . . . . . . . . . . . . . . .4-7 The Sequential Linear Programming Method . . . . . . . . . . . . . . . . . . . . . . . .4-9 Move Limits of Design Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4-10 Constraint Trimming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4-11 Convergence Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4-12 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4-14

Additional Problems

. . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5-1

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-1

In de x

ii

COSMOS/M Advanced Modules

Introduction

Design Optimization and Sensitivity

Before starting the topic of design optimization and sensitivity, it is important to distinguish between analysis and design. Analysis is the process of determining the response of a specified system to its environment. Design, on the other hand, is the actual process of defining the system. Analysis is therefore a subset of design. Engineering design in general, is an iterative process as shown in Figure 1-1. The design is continuously modified until it meets evaluation and acceptance criteria set by the engineer. Mathematical and empirical formulas aided by years of engineering judgment and experience have been useful in the traditional design processes to verify the adequacy of designs. However, a fully automated design optimization and sensitivity is used when engineers are trying to modify a design which level of complexity exceeds their ability to make appropriate changes. It is not surprising that even what might appear as extremely simple design task may easily be a real challenge to the designer during the decision-making process. The design optimization and sensitivity capability provides many design options. Whether you wish to design a simple truss or a complicated three dimensional solid model, COSMOSM or COSMOSFFE will modify both the size and geometrical shape in search for an improved design.

In de x

COSMOSM Advanced Modules

1-1

Chapter 1 Introduction Figure 1-1. Iterative Process of Engineering Design

Initial Design

Requirements Satisfied ? Yes Any Room for Improvments ? No

No

Change Design

Yes

Final Design

The following sections provide more information on the design optimization and sensitivity module (OPTSTAR). They include brief explanations of terminology, the optimization process, and sensitivity studies. There is also a summary of the important features of the OPTSTAR module.

Terminology
The terminology frequently used in design optimization and sensitivity study are: design variables, objective function, behavior constraints, response quantities, feasible design, optimum design, and sensitivity type. Chapter 2, Elements of Design Optimization and Sensitivity, explains these terminology in more detail.

In de x

1-2

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Design Optimization Process

The process of design optimization can be pictorially represented as shown in the following figure.
Figure 1-2. Design Optimization Process

GEOMETRY, MESHING, AND ANALYSIS

DEFINE OPTIMIZATION PARAMETERS

GEOMETRY MESHING

ANALYSIS

APPROXIMATION AND OPTIMIZATION

POSTPROCESSING

Yes

Is Convergence Achieved ?

No

O P TIMIZATIO N LO O P

Refer to Chapter 3, Procedures and Examples, for general guidelines to performing design optimization.

Sensitivity Studies
The process of sensitivity study is similar in principle to the design optimization process illustrated previously. The procedure is summarized in Figure 1-3.

In de x

COSMOSM Advanced Modules

1-3

Chapter 1 Introduction Figure 1-3. Sensitivity Process

GEOMETRY, MESHING, AND ANALYSIS

DEFINE SENSITIVITY PARAMETERS

GEOMETRY MESHING

ANALYSIS

POSTPROCESSING

Yes

Is Required Number of Runs Executed ?

No

S E NS ITIV ITY LO O P

Chapter 3, Procedures and Examples, describes more details about performing sensitivity studies.

Features of the Design Optimization and Sensitivity Module (OPTSTAR)

The process of finite element analysis, evaluation of analysis results and design changes, and modifications for yet another solution phase are performed automatically in COSMOSM. OPTSTAR performs two-dimensional and threedimensional sizing and shape optimization and sensitivity for structural and thermal applications. The following are some of the module's capabilities:

Full interaction with GEOSTAR for model creation, results manipulation and
display (pre- and postprocessing)

Access to COSMOSM and COSMOSFFE solvers, element and material

libraries

In de x

1-4

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Type of analyses:
Linear Static (including multiple load cases) Linear Dynamic (natural frequencies and mode shapes) Linearized Buckling Heat Transfer Nonlinear Fatigue Advanced Dynamic Dynamic Stress

Design variables:
Side constraints (upper and lower limits of design variables) Move limits control Shape Applications - Dimensions and parameters used in building the model's geometry Sizing Applications - Parameters used to define the model other than the shape parameters - For linear static analysis, predefined sizing options include:

Cross-sectional area of truss elements Thickness of 2D continuum elements Thickness of shell elements Width and height of beam elements with rectangular cross-sections Thickness and radius of pipe elements

Optimization behavior constraints:

Trimming control Different sets (with lower and upper limits) of: - Displacements - Relative displacements - Stresses - Strains - Reaction forces - Fatigue usage factors
In de x

COSMOSM Advanced Modules

1-5

Chapter 1 Introduction

- Natural frequencies - Linearized buckling load factors - Velocities - Accelerations - Temperatures - Temperature gradients - Heat Fluxes - Weight - Volume - User-defined quantities

Optimization objective function:

Minimization and maximization of one type composed of different sets with user-specified weight factors. - Volume - Weight - Displacement - Relative displacement - Stress - Strain - Reaction force - Fatigue usage factors - Velocity - Acceleration - Natural frequency - Linearized buckling load factor - Temperature - Temperature gradient - Heat flux - User-defined quantity

In de x

1-6

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Sensitivity options:
Global, local and offset pre-optimization sensitivity studies, in addition to optimization sensitivity results. Sensitivity response quantities include: - Displacements - Relative displacements - Stresses - Strains - Reaction forces - Fatigue usage factors - Velocities - Accelerations - Natural frequencies - Linearized buckling load factors - Temperatures - Temperature gradients - Heat fluxes - Volume - Weight - User-defined quantities

Numerical techniques:
Modified Feasible Directions Singular Value Decomposition technique Linear, quadratic and cubic approximations Restart and restore options

Results:
Output file X-Y convergence and sensitivity plots Color filled, colored line contour plots, and vector plots of displacement, stress, strain, temperature, temperature gradient, and heat flux for the current model.

In de x

COSMOSM Advanced Modules

1-7

Chapter 1 Introduction

Animation and plots of deformed shapes for linear static analysis and mode shapes for frequency and buckling analyses. Tabular data reports

Limits
25 design variables 60 constraint sets 100 objective function sets 60 sensitivity response quantities 75 design sets for optimization 20 increments for global sensitivity 20 sets for offset sensitivity 32,000 and 64,000 nodes and elements on PCs 3000 nodes and 3000 elements for EXPLORER 64,000 nodes and 64,000 elements on EWS (Unix workstations)

In de x

1-8

COSMOSM Advanced Modules

Elements of Optimization and Sensitivity

Design Optimization
Design optimization refers to the automated redesign process that attempts to minimize or maximize a specific quantity (objective function) subject to limits or constraints on the response by using a rational mathematical approach to yield improved designs. Figure 2-1 shows minimum weight design of a structure.
Figure 2-1. Minimum Weight Design of a Structure

Remov able Material Neutral Axis

Initia l D e sign
t3 t1 d2 d1 t2 t4

In de x

Fina l D e sign

COSMOSM Advanced Modules

2-1

Chapter 2 Elements of Optimization and Sensitivity

A feasible design is a design that satisfies all of the constraints. A feasible design may not be optimal. An optimum design is defined as a point in the design space for which the objective function is minimized or maximized and the design is feasible. If relative minima exist in the design space, other optimal designs can exist. Basic terminology in design optimization are: Design variables, objective function, and behavior constraints. They are explained in the following sections.

Design Variables
Design variables are the parameters (independent quantities) that users seek to find their values for an optimum design. Figure 2-2 shows a structure having four geometry dimensions defined as design variables. Upper and lower bounds are specified for each design variable. Lower and upper bounds are also referred to as side constraints. For example:
10 T1 25 Lower Bound Upper Bound

Figure 2-2. A Structure with Four Design Variables

t1

t2

t4

t3

In de x

2-2

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Depending on design variables, there are two types of optimization applications: sizing optimization and shape optimization. Sizing optimization refers to the class of problems where a change in design variables does not change the problem's geometry or mesh as shown in Figure 2-3.
Figure 2-3. Initial and Final Geometry and Mesh for a Sizing Optimization Problem

Initial Ge ome try and Me sh

Final Ge ome try and Me sh

For linear static analysis, predefined sizing options are summarized in Table 2-1.
Table 2-1. Predefined Sizing Options COSMOSM Element Type and Name Truss Beam (rectangular cross-sections) 2D Continuum Shell Pipe TRUSS2D, TRUSS3D BEAM2D, BEAM3D TRIANG, PLANE2D SHELLAX, SHELL3, SHELL3T, SHELL4, SHELL4T, SHELL6, SHELL9 PIPE Design Variable Cross-Sectional Area Width, Height Thickness Thickness Thickness, Radius

Shape optimization refers to the class of problems where any change in design variables causes change in the problem's geometry or mesh as shown in Figure 2-4.

In de x

COSMOSM Advanced Modules

2-3

Chapter 2 Elements of Optimization and Sensitivity Figure 2-4. Initial and Final Geometry and Mesh for a Shape Optimization Problem

Initial Geometry and Mesh

Final Geometry

Besides purely sizing optimization and shape optimization mentioned above, there is a class of problems where both sizing and shape parameters are defined as design variables as shown in Figure 2-5.
Figure 2-5. Initial and Final Geometry for Sizing/Shape Optimization Problems

Initial Geometry

Initial Geometry

Final Geometry

Final Geometry
Sizing Parameter: Thickness

Sizing Parameter: Cross-Section Area

Truss Elements

Shell Elements or Continuum Elements

In de x

2-4

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Objective Function
Objective function is a single quantity that the optimizer seeks to minimize or maximize. The objective function must be a continuous function of the design variables. The weight (or volume) of a structure is an example of the commonly used objective functions. Other quantities are:

Stress, Strain, Displacement, Reaction Force, Velocity, Acceleration, Natural Frequency, Linearized Buckling Load Factor, Temperature, Temperature Gradient, Heat Flux, Fatigue Usage Factor, and

Figure 2-6. A Structure Composed of Three Regions

Rg

Rg

Rg

User-Defined Functions.
The objective function can be composed of different sets of the same type, and can reflect different weight (importance) factors for different portions of the model as shown in Figure 2-6.

Behavior Constraints
A behavior constraint is defined as an inequality that must be satisfied in order to have a feasible design. The behavior constraints are typically response quantities that are functions of the design variables. Von Mises stress is a typical example in structural problems:
In de x

von Mises stress allowed stress

COSMOSM Advanced Modules

2-5

Chapter 2 Elements of Optimization and Sensitivity

Other quantities are:

Volume, Weight, Stress, Strain, Displacement, Reaction Force, Velocity, Acceleration, Natural Frequency,

Figure 2-7. An Optimization Problem with Stress and Displacement Constraint

Deflection

Linearized Buckling Load Factor, Temperature, Temperature Gradient, Heat Flux, Fatigue Usage Factor, and User-Defined Functions.

Multiple constraint sets of different types can also be specified. In COSMOSM, users have to specify lower and upper limits for behavior constraints. For example: 0 von Mises stress allowed stress

Sensitivity Study
A sensitivity study is the procedure that determines the changes in a response quantity for a change in a design variable. Figure 2-8 shows a sensitivity study of a control arm bracket and Figure 2-9 shows its result.

In de x

2-6

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization Figure 2-8. Sensitivity Study of a Control Arm Bracket in Frequency Analysis

Figure 2-9. Fundamental Frequency versus Design Variable-1, t1

In de x

Basic terminology in sensitivity study are: Design variables and response quantities. The definition of design variables is the same as that in design optimization. Response quantities are functions of the design variables. All the postprocessing quantities which are suitable for the objective function and behavior constraints are also suitable for the sensitivity response quantities.

COSMOSM Advanced Modules

2-7

Chapter 2 Elements of Optimization and Sensitivity

Sensitivity Types
There are four types of sensitivity study, namely, global sensitivity, offset sensitivity, local sensitivity, and optimization sensitivity results. They are explained in the following paragraphs. Global sensitivity - where design variables are changed between their lower and upper bounds in a user-specified number of steps. The number of steps is the same for all the design variables. Under this type of sensitivity, the user can change all the design variables simultaneously or one at a time. Have the frequency analysis of a control arm bracket as an example where: 0.5 design variable-1 2.5 and 1.5 design variable-2 3.5 The plots of response quantity versus design variable are shown in Figure 2-10 through Figure 2-12.
Figure 2-10. Global Sensitivity - One at a Time: Fundamental Frequency versus Design Variable-1, t1

In de x

2-8

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization Figure 2-11. Global Sensitivity - One at a Time: Fundamental Frequency versus Design Variable-2, t2

Figure 2-12. Global Sensitivity - Simultaneously: Fundamental Frequency versus Normalized Design Variable-1and -2

Offset sensitivity - where users specify the values of a series of design variables in a user-defined sets. The design variables are defined either by the actual values or by a perturbation ratio with respect to the initial value.

In de x

COSMOSM Advanced Modules

2-9

Chapter 2 Elements of Optimization and Sensitivity

Have the frequency analysis of a control arm bracket as an example where the series of design variables are:
Sensitivity Set Number 1 2 3 4 5 Design Variable-1 0.5 1.0 1.5 2.0 2.5 Design Variable-2 3.5 3.0 2.5 2.0 1.5

The plot of response quantity versus sensitivity set is shown in Figure 2-13.
Figure 2-13. Offset Sensitivity: Fundamental Frequency versus Sensitivity Set Number

Local sensitivity - where a design variable is perturbed at a time by a userspecified value while the rest of the design variables are kept unchanged. The perturbed design variables are defined either by the actual values or by a perturbation ratio with respect to the initial value. The gradients of the response quantities with respect to the design variables are computed based on the finite difference method. Have the frequency analysis of a control arm bracket as an example where:
In de x

2-10

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Initial Value:

design variable-1=2.5, design variable-2=3.5 design variable-1=+0.1, design variable-2=+0.1

Perturbation Ratio:

The plot of gradient of the response quantity versus design variable set is shown in Figure 2-14.
Figure 2-14. Local Sensitivity: Gradient of Fundamental Frequency versus Design Variable Number

Optimization sensitivity results - where gradients of behavior constraints and objective function are computed during the optimization process. The gradients are obtained by taking the derivatives of the approximation functions with respect to the design variables. This type of sensitivity study is available only when the design optimization is to be performed.

In de x

COSMOSM Advanced Modules

2-11

In de x

2-12

COSMOSM Advanced Modules

Procedures and Examples

Introduction
This chapter presents detailed examples that fully describes the procedures for performing design optimization and sensitivity in COSMOSM. The descriptions include: selection and definition of appropriate parameters required for geometry creation, generation of the finite element mesh parametrically or otherwise, applying loads and boundary conditions, optimization constraint definitions, defining the objective function, defining sensitivity response quantities, specifying the optimization and sensitivity options, performing the optimization and sensitivity loops, and postprocessing of optimization, sensitivity and analysis results.
Table 3-1. Examples Shape Optimization of a Slotted Control Arm in Static Analysis (See page 3-17.) Sensitivity Study of a Control Arm Bracket in Frequency (See page 3-37.)

In de x

COSMOSM Advanced Modules

3-1

Chapter 3 Procedures and Examples

Overview of Process for Design Optimization and Sensitivity

The general process for design optimization and sensitivity displayed in Chapter 1 is shown in the following figures in more detail.
Figure 3-1. Overview of Process for Design Optimization

P re proce ssing Build Geometry and Mesh in Terms of Design Parameters

O ptimiz ation Loop

Update Geometry and Mesh (If Needed)

P e rform Analysis

Static Frequency Buckling Thermal Nonlinear Post Dynamic Dynamic Stress Fatigue

De fine O ptimiz ation P arame te rs Design Parameters (Variables) Design Objective (Objective Function) Design Constraints (Behavior Constraints)

Perform Analysis

Extract Critical Constraints

P ostproce ssing

Approximate Objective Function and Constraints

Deformed Shapes Animation Mode Shapes Contour and Vector Plots X-Y Plots

Improved Design

Final Design

Yes

Requirements Achieved?

No

In de x

3-2

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization Figure 3-2. Overview of Process for Sensitivity Study

P re proce ssing Build Geometry and Mesh in Terms of Design Parameters S e nsitivity Loop P e rform Analysis De fine S e nsitivity P arame te rs Design Parameters (Variables)

Static Frequency Buckling Thermal Nonlinear Post Dynamic Dynamic Stress Fatigue

Update Geometry and Mesh (If Needed)

Response Quantities P ostproce ssing

Perform Analysis

Deformed Shapes Animation Mode Shapes Contour and Vector Plots X-Y Plots

Sensitivity Types

Yes

Is required number of runs executed?

No

Overview of Commands for Design Optimization and Sensitivity

The following figure provides an overview of commands required for defining design variables, objective function, constraints, response quantities, and optimization/sensitivity options in COSMOSM. For more information on the commands, please refer to the COSMOSM Command Reference manual (Volume 2).
In de x

COSMOSM Advanced Modules

3-3

Chapter 3 Procedures and Examples Figure 3-3. Overview of Commands for Design Optimization and Sensitivity

DE S IGN O P TIMIZATIO N AND S E NS ITIV ITY

Design Variables Optimization Objective Optimization Constraints Sensitivity Response Optimization Loops Sensitivity Runs

ANALY S I S > O P TI MI ZE / S E NS I TI V I TY ME NU TRE E

OP_OBJDEF OP_OBJSET OP_OBJLIST OP_OBJDEL

OP_CONDEF OP_CONLIST OP_CONDEL OP_CONTRIM

OP_CONTROL OP_RESTORE A_OPTIMIZE R_OPTIMIZE

A_SENSITIV R_SENSITIV

SN_RESPDEF SN_RESPLIST SN_RESPDEL

DVARDEF DVARLIST DVARVDEL OP_DVMOVE SN_SETDEF SN_SETLIST SN_SETDEL

CONVERGENCE AND SENSITIVITY PLOTS

DI S P LAY > X Y _ P LO TS ME NU TRE E

INITXYPLOT ACTXYPOST SETXYPLOT XYRANGE XYREFLINE XYIDENTIFY XYLIST XYPTLIST XYPLOT

In de x

3-4

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Procedures for Performing Design Optimization

The following steps are recommended to be followed for performing design optimization studies using OPTSTAR. These guidelines are not in any sense complete, and are intended to be complementary to your own optimization knowledge and expertise. Step 1. Build the model parametrically

For shape optimization problems, build the model geometry parametrically in

places where it is necessary. For sizing optimization, the section constants (e.g., cross-sectional area of a truss) and design aspects to be optimized will be defined as parameters. These parameters should then be defined as design variables using the DVARDEF (Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define) command. You may choose to use all or some of the defined parameters in the optimization process. Note that the DVARDEF (Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define) command needs to be applied to each design variable separately.

For design optimization and sensitivity, you need to use the COSMOSM
command language to parametrically model the design geometry and/or physical properties. The command language essentially facilitates you to describe the design variables in GEOSTAR for a fully automated design optimization and sensitivity processes. Some of the capabilities of this parametric language are: - the use of single parameters, arrays and functions, - construction of arithmetic expressions, - generating macros, - control structure commands, and - logical expressions. In most cases only the PARASSIGN (Control > PARAMETER > Assign Parameter) command needs to be used. Refer to COSMOSM User Guide, Volume 1, Appendix E for more details.

You must exercise caution when using parameters to describe the model
geometry. The chosen parameters have to define the model completely so that when their values are modified during the optimization loops, the geometry creation and meshing will not fail.
In de x

COSMOSM Advanced Modules

3-5

Chapter 3 Procedures and Examples

Step 2. Execute required initial analyses

Execute the initial analysis as usual in COSMOSM. The types of analyses

currently supported for design optimization are: - linear static stress analysis (including multiple load cases), - linearized buckling analysis, - analysis of natural frequencies and mode shapes, - heat transfer analysis, - nonlinear structural analysis, - post dynamic analysis, - dynamic stress analysis, and - fatigue analysis.

For multidisciplinary design optimization, you can execute the analyses in any
order before proceeding with optimization loops except the following cases.

It should be noted that natural frequency analysis and buckling analysis cannot
be combined in a multidisciplinary optimization application since they share the same database locations in the program unless user-defined postprocessing functions are used for constraints and/or objective functions.

If you are performing design optimization on a heat transfer - linear static

problem in which the temperatures are computed using either HSTAR or FFE Thermal, then you need to follow the procedure listed below in order to input the heat transfer results as thermal loads in static analysis: - Use R_THERMAL (Analysis > HEAT TRANSFER > Run Thermal Analysis) command to execute heat transfer analysis. - Use TEMPREAD (LoadsBC > LOAD OPTIONS > Read Temp as Load) command to read temperatures from heat transfer analysis. - Use A_STATIC (Analysis > STATIC > Static Analysis Options) command with flag T to include thermal loading in static analysis. - Use R_STATIC (Analysis > STATIC > Run Static Analysis) command to execute linear static analysis.

Post dynamic and fatigue analyses cannot be executed alone instead they must
follow other types of analysis. You need to follow the procedures listed below to get correct results. - Post dynamic analysis: 1. Use R_FREQUENCY (Analysis > FREQUENCY/BUCKLING > Run Frequency) command to execute analysis of natural frequencies and mode shapes,

In de x

3-6

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

2. Use R_DYNAMIC (Analysis > POST-DYNAMIC > Run Post Dynamic) command to execute post dynamic analysis, 3. Use R_STRESS (Analysis > STATIC > Run Stress Analysis) command to execute dynamic stress analysis if desired. - Fatigue analysis: 1. Use R_STATIC (Analysis > STATIC > Run Static Analysis) or R_NONLINEAR (Analysis > NONLINEAR > Run NonL Analysis) command to execute linear or nonlinear structural analysis respectively or R_FREQUENCY (Analysis > FREQUENCY/BUCKLING > Run Frequency), R_DYNAMIC (Analysis > POST-DYNAMIC > Run Post Dynamic), and R_STRESS (Analysis > STATIC > Run Stress Analysis) commands to execute frequency, post dynamic and dynamic stress analysis, 2. Use R_FATIGUE (Analysis > FATIGUE > Run Fatigue Analysis) command to execute fatigue analysis. Step 3. Perform postprocessing of initial analysis results Perform postprocessing of the initial executed analyses as usual. For multidisciplinary analysis, you need to first activate the required type of analysis. Users will have access to all existing GEOSTAR'S postprocessing features. Please refer to User Guide (Vol. 1) and Basic FEA System Manual (Vol. 3) for more information. Step 4. Begin design optimization procedures by defining design variables

First, define the design variables using the command DVARDEF (Analysis >
OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define). Note that each design variable, whether in shape or sizing optimization, must have already been defined as a parameter [using the PARASSIGN (Control > PARAMETER > Assign Parameter) command]. The command controls the following information: - Type of variable (shape or sizing) and its parametric name, - Lower and upper bounds, - Convergence tolerance (see Chapter 4), Default values = 1/100 | upper bound- to lower bound | - Method of choosing the pre-optimization design variable values (perturbation or random evaluation),
In de x

COSMOSM Advanced Modules

3-7

Chapter 3 Procedures and Examples

- Sizing options and element type (only for linear static analysis) where a distinction has to be made between 2D and 3D beam elements and either membrane or bending dominant behavior has to be indicated for shell elements.

Use the commands DVARDEL (Analysis > OPTIMIZE/SENSITIVITY >

DESIGN VARIABLES > Delete) and DVARLIST (Analysis > OPTIMIZE/ SENSITIVITY > DESIGN VARIABLES > List) to delete and list the design variables respectively.

Use the commands OP_DVMOVE (Analysis > OPTIMIZE/SENSITIVITY >

DESIGN VARIABLES > Move Limits) to control the move limits of design variables during the optimization loops. This command is seldom necessary since its default options will suffice in most cases. Chapter 4, Numerical Aspects explains in more detail how the move limits of design variables function in each optimization loop. Step 5. Define objective function

Define the objective function using the OP_OBJDEF (Analysis > OPTIMIZE/
SENSITIVITY > OBJECTIVE FUNCTION > Define Function) command. The command controls the following information: - Types of objective functions, - Layer and face numbers (for composite shells only), - Analysis type (for multidisciplinary optimization), - Type of application (minimization or maximization), - Criterion and approximation type (see Chapter 4), - Convergence tolerance (see Chapter 4), - Reference keypoint (relative displacement), - Load case (multiple load cases only) or time step number (nonlinear or post dynamic only).

You can use the OP_OBJSET (Analysis > OPTIMIZE/SENSITIVITY >

OBJECTIVE FUNCTION > Define Function Set) command to declare portions (sets) of the model for objective function computations. It is also possible to assign weight factors to different parts of the model using the same command. This command is most useful for volume and weight objective functions (cost). However, it cannot be used with frequency and buckling functions.
In de x

3-8

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Use the OP_OBJDEL (Analysis > OPTIMIZE/SENSITIVITY > OBJECTIVE

FUNCTION > Del Function Set) command to delete sets defined by the OP_OBJSET (Analysis > OPTIMIZE/SENSITIVITY > OBJECTIVE FUNCTION > Define Function Set) command.

To delete the objective function defined using the OP_OBJDEF (Analysis >
OPTIMIZE/SENSITIVITY > OBJECTIVE FUNCTION > Define Function) command, you need to use the same command again to overwrite the old information.

Use the OP_OBJLIST (Analysis > OPTIMIZE/SENSITIVITY > OBJECTIVE

FUNCTION > List Function) command to list information defined by OP_OBJDEF (Analysis > OPTIMIZE/SENSITIVITY > OBJECTIVE FUNCTION > Define Function) and OP_OBJSET (Analysis > OPTIMIZE/ SENSITIVITY > OBJECTIVE FUNCTION > Define Function Set) commands. Step 6. Define constraints

Define behavior constraints using the OP_CONDEF (Analysis > OPTIMIZE/

SENSITIVITY > BEHAVIOR CONSTRAINT > Define) command. The command controls the following information: - Types of behavior constraints, - Layer and face numbers (for composite shells only), - Analysis type (for multidisciplinary optimization), - Geometry association, - Lower and upper limits (bounds), - Feasibility tolerance: Default values = 1/100 | upper bound- to lower bound | - Reference keypoint (relative displacement), - Criterion and approximation type (see Chapter 4), - Load case (multiple load cases only) or time step number (nonlinear or post dynamic only).

Use OP_CONLIST (Analysis > OPTIMIZE/SENSITIVITY > BEHAVIOR

CONSTRAINT > List Behavior Const) and OP_CONDEL (Analysis > OPTIMIZE/SENSITIVITY > BEHAVIOR CONSTRAINT > Del Behavior Const) commands to list and delete constraints, respectively.

In de x

COSMOSM Advanced Modules

3-9

Chapter 3 Procedures and Examples

By default, OPTSTAR considers only the violated and potentially critical

constraints during calculations. In order to control this step, use the command OP_CONTRIM (Analysis > OPTIMIZE/SENSITIVITY > BEHAVIOR CONSTRAINT > Truncate Constraint) which allows you to input trimming (truncation) factors for the unviolated constraints. If the normalized value of a particular constraint is above the negative value of the truncation factor, then that constraint is appended to the critical list. For more detailed explanation, refer to Chapter 4, Numerical Aspects. Step 7. Specify parameters for optimization

Specify the parameters for the optimization loops using the A_OPTIMIZE
(Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP > Optimize Analysis Options) command. The important input for this command are: Maximum number of optimization loops (nloops flag) Number of stages to check convergence (loop_conv flag) Type of analyses. It should be mentioned that an optimization loop will execute the analyses in the same sequence specified here. For heat transfer linear static problems requiring data transfer from heat transfer to static analysis, you must specify THERMAL followed by STATIC. Number of consecutive infeasible design sets. You are recommended to always start with a feasible initial design. Otherwise, the program will prompt you with a choice to continue or stop. If you choose to continue, the optimization loops will be terminated if a feasible solution is not reached after five consecutive attempts. To change this number (five), use the A_OPTIMIZE (Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP > Optimize Analysis Options) command and specify the appropriate input for the infeas flag.

Start the optimization process using the R_OPTIMIZE (Analysis > OPTIMIZE/
SENSITIVITY > OPTIMIZE LOOP > Run Optimize Analysis) command. Step 8. Restart options prior to convergence In cases where the maximum number of optimization loops are exceeded, you can restart the process by activating the restart flag under the A_OPTIMIZE (Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP > Optimize Analysis Options) command, followed by the R_OPTIMIZE (Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP > Run Optimize Analysis) command. If you choose to use this option, only the following commands can be reissued (if needed): DVARLIST, OP_DVMOVE, OP_CONLIST, OP_CONTRIM, OP_OBJLIST, OP_CONTROL.

In de x

3-10

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Step 9. Postprocess optimization results

You can display convergence plots of objective function (component name

OP_OBJ), behavior constraints (component name OP_CON), and design variables (component name OP_DVAR) against number of loops using the ACTXYPOST (Display > XY PLOTS > Activate Post-Proc) and XYPLOT (Display > XY PLOTS > Plot Curves) commands from the Display-XY Plots menu tree. It is also possible to list the activated result component on-line using the XYPTLIST (Display > XY PLOTS > List Points) command. It should be mentioned that the ACTPOST (Results > SET UP > Set PostProcess Type) command has to be used prior to using the ACTXYPOST (Display > XY PLOTS > Activate Post-Proc) command.

The optimization results for each loop are summarized in the output file
jobname.OPT. Step 10. Postprocess converged analyses results

Perform postprocessing of the converged analyses as usual. Please refer to User

Guide (Vol. 1) and Basic FEA System Manual (Vol. 3) for more information.

One of the good features of the optimization module is that you will have the
optimum product in terms of its geometric dimensions (not giving only the final coordinates of the mesh) which is an aspect favored in the manufacturing process. Using the standard methods of transferring geometry (such as IGES and DXF formats), the users can transfer the final geometry to NC machines. Step 11. Restore an interim design set

It is possible to examine the design configuration at an interim step even after

the convergence of the optimization loops. The command OP_RESTORE (Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP > Restore Design Set) can be used to restore a design set (corresponding to a specified loop number) so that the entire database can be reconstructed for the specified design set and the required analyses run automatically.

If you use the OP_RESTORE (Analysis > OPTIMIZE/SENSITIVITY >

OPTIMIZE LOOP > Restore Design Set) command, note that the converged solution obtained will be lost. You need to therefore save the database of the converged solution before applying this command.

In de x

COSMOSM Advanced Modules

3-11

Chapter 3 Procedures and Examples

Procedures for Performing Sensitivity Studies

The following steps are recommended to be followed for performing sensitivity studies using OPTSTAR. These guidelines are not in any sense complete, and are intended to be complementary to your own knowledge and experience. Step 1. Initial model and analysis The first step is similar to steps 1 through 3 of the procedures recommended for performing design optimization: Build the model parametrically Execute required initial analyses Perform postprocessing of initial analysis results

Step 2. Begin sensitivity study by defining design variables

Define the design variables using the DVARDEF (Analysis > OPTIMIZE/
SENSITIVITY > DESIGN VARIABLES > Define) command. Note that each design variable, whether of shape of sizing type, must have been already defined as a parameter (using the PARASSIGN (Control > PARAMETER > Assign Parameter) command). It should be noted that the DVARDEF (Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define) command is used for both optimization and sensitivity applications where some options of the command are needed only for optimization.

Use the commands DVARDEF (Analysis > OPTIMIZE/SENSITIVITY >

DESIGN VARIABLES > Define) and DVARLIST (Analysis > OPTIMIZE/ SENSITIVITY > DESIGN VARIABLES > > List) to delete and list the design variables respectively. Step 3. Define the sensitivity response quantities

Define the response quantity using the SN_RESPDEF (Analysis > OPTIMIZE/
SENSITIVITY > RESPONSE QUANTITY > Define) command. The commands controls the following information: - Type of response quantities, - Layer and face numbers (for composite shells only), - Analysis type (for multidisciplinary optimization), - Geometry association,
In de x

3-12

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

- Reference keypoint (relative displacement), - Criterion type (see Chapter 4), - Load case (multiple load cases only) or time step number (nonlinear or post dynamic only).

Use SN_RESPDEL (Analysis > OPTIMIZE/SENSITIVITY > RESPONSE

QUANTITY > Delete) and SN_RESPLIST (Analysis > OPTIMIZE/ SENSITIVITY > RESPONSE QUANTITY > List) to delete and list response quantities respectively. Step 4. Specify type of sensitivity

Specify the type and parameters of the sensitivity study using the A_SENSITIV
(Analysis > OPTIMIZE/SENSITIVITY > SENSITIVITY RUN > Options) command. The important input for this command are: Type of sensitivity:

Global sensitivity, offset sensitivity, or local sensitivity. For offset and local sensitivity studies, the command SN_SETDEF (Analysis > OPTIMIZE/ SENSITIVITY > RESPONSE QUANTITY > Define) should be used to define the user-specified design sets.

Type of analyses:

It should be mentioned that a sensitivity run will execute the analyses in the
same sequence specified here. For instance if you want to find the fundamental frequency of the model based on the deformed configuration calculated by the nonlinear program, you must specify NONLINEAR followed by FREQUENCY.

Start the sensitivity study by using the R_SENSITIV (Analysis > OPTIMIZE/
SENSITIVITY > SENSITIVITY RUN > Run Analysis) command. Step 5. Restart option In cases where the sensitivity study is terminated by the user or due to an error, you can restart the process by activating the restart flag under the A_SENSITIV (Analysis > OPTIMIZE/SENSITIVITY > SENSITIVITY RUN > Options) command, followed by the R_SENSITIV (Analysis > OPTIMIZE/SENSITIVITY > SENSITIVITY RUN > Run Analysis) command.

In de x

COSMOSM Advanced Modules

3-13

Chapter 3 Procedures and Examples

Step 6. Postprocess sensitivity results

You can display plots of response quantities against value or label of design
variables using the ACTXYPOST (Display > XY PLOTS > Activate Post-Proc) and XYPLOT (Display > XY PLOTS > Plot Curves) commands from the Display-XY Plots menu tree. It is also possible to list the activated result component on-line using the XYPTLIST (Display > XY PLOTS > List Points) command. It should be mentioned that the ACTPOST (Results > SET UP > Set PostProcess Type) command has to be used prior to using the ACTXYPOST (Display > XY PLOTS > Activate Post-Proc) command.

The sensitivity results are summarized in the output file jobname.OPT.

Special Features for Optimization and Sensitivity

Dynamic change of parameters The user can specify the flags controlling the optimization and sensitivity features in terms of parametric expressions. For example, the upper and lower bounds of a design variable may not always be a constant value but depends on some boundary conditions and geometric features that change from one optimization or sensitivity run to another. Restart option

For optimization, users can restart the process from the last successful design
sets in the following cases: Iterations exceed the maximum allowed number. Number of trials to find a feasible solution exceeds the allowed number. Optimization process terminated by the user either during the optimization screen or during regenerating the model's geometry or mesh using the <Esc> key Optimization terminated because of failure to regenerate the model or crash of optimization procedure.

For sensitivity studies, you can restart the process from the last successful run in
the following cases: - Terminating the process during the sensitivity screen or generating the model's geometry or mesh using the <Esc> key.
In de x

- Failure to regenerate the model.

3-14

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

It should be mentioned that you can restart the optimization process (as a fresh
start) from the current design variable values as new initial values by choosing the OFF option in the restart flag. In this case, optimization will begin its first iteration all over again. For example, if your model converged to an optimum selection in eight iterations, and you chose to restore the fifth design set to inspect it. If you choose fresh start at this point, the optimizer will consider the design variables, constraints and objective function values of the fifth design set as initial values in a new optimization run. User-defined constraints, objective function and sensitivity response quantities

Situations may arise where you need to customize your own objective function,
behavior constraints, or sensitivity response quantities. A user-defined feature is included in the OP_OBJDEF (Analysis > OPTIMIZE/SENSITIVITY > OBJECTIVE FUNCTION >Define Function), OP_CONDEF (Analysis > OPTIMIZE/SENSITIVITY > BEHAVIOR CONSTRAINT > Define Behavior Constraint) and SN_RESPDEF (Analysis > OPTIMIZE/SENSITIVITY > RESPONSE QUANTITY > Define) commands. The user-defined quantity has to be declared by a PARASSIGN (Control > PARAMETER > Assign Parameter) command prior to issuing the OP_OBJDEF (Analysis > OPTIMIZE/SENSITIVITY > OBJECTIVE FUNCTION > Define Function), OP_CONDEF (Analysis > OPTIMIZE/SENSITIVITY > BEHAVIOR CONSTRAINTS > Define Behavior Constraint) or SN_RESPDEF (Analysis > OPTIMIZE/SENSITIVITY > RESPONSE QUANTITY > Define) commands. These quantities can be calculated using the extensive options provided by the COSMOSM command language (User Guide, Vol. 1, Appendix E).

Regenerating the model through GEOSTAR requires reading all the steps you
followed as stored in the session file in every optimization or sensitivity loop. There are however some commands (action commands) in the session file that are ignored by GEOSTAR (e.g. R_ or R... commands). The reason is that it would be very time consuming to follow the user's steps of running analysis modules since you might have executed these commands many times to check and modify the initial model. Instead, the control of this step is left to the A_OPTIMIZE (Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP > Optimize Analysis Options) and A_SENSITIV (Analysis > OPTIMIZE/ SENSITIVITY > SENSITIVITY RUN > Options) commands.

In de x

COSMOSM Advanced Modules

3-15

Chapter 3 Procedures and Examples

User-defined objective function, behavior constraint or a response quantity may

be a preprocessing or a postprocessing quantity. An example of a preprocessing quantity is the volume or weight of some elements. Once the mesh is generated you will be able to use its information to calculate these quantities (using VOLUME and WEIGHT functions explained in Appendix E of the User Guide.) If the user-defined feature is a postprocessing-function dependent quantity, it must be defined after running the analysis. For example, in order to use the frequency function (FREQ) you have to execute R_FREQUENCY (Analysis > FREQUENCY/BUCKLING > Run Frequency) first. Another example is the use of secondary load cases using the LCCOMB (Results > Combine Load Case) command. You need to run R_STATIC (Analysis > STATIC > Run Static Analysis first followed by the LCCOMB (Results > Combine Load Case) command.

In order to respect the sequence (or order) of issuing some of GEOSTAR

commands which is essential in some user-defined cases, a separate option is included in the type of analysis flag defined by the A_OPTIMIZE (Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP > Optimize Analysis Options) and A_SENSITIV (Analysis > OPTIMIZE/SENSITIVITY > SENSITIVITY RUN > Options) command. This option is called FILE, and it means that the type of analysis (R_ commands.) and other GEOSTAR commands are included in a separate file prepared by the user. The file name should be GEOFILE.FIL and it has to be located in the local directory. This file include the R_ commands and other action commands given in the sequence necessary for calculating the user-defined quantities.

In de x

3-16

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Shape Optimization of a Slotted Control Arm in Static Analysis

In this example, you are required to find the thickness of the two shafts (TR1 and TR2), and the size and location of the cutout (TW). The outer arm thickness is 20 mm, modulus of elasticity is 2 x 105 MPa, and Poissons ratio is 0.3. The control arm is subjected to a pressure loading of 4 MPa as shown in the following figure. The only constraint on the control arm is that the von Mises stress due to the applied loading should neither exceed 225 MPa nor fall below 10 MPa. The objective function for minimization is the volume of the control arm with a tolerance of 1%.
Figure 3-4. Initial Geometry, Boundary Conditions and Loads
Y

(See page 3-1.)

Y
r2 = 7 10 tr 2 5 5

20 tr1 5 r1 = 30

tw

Py

10 20 Fixed = 140

6 - Node TRI ANG Thickness = 20 mm E = 2 x 105 N/mm2 = 0.3 Py = 4 N/ mm 2 (Y direction) Note: All dimensions in millimeters.

The design variables for this problem as seen from the above figure are designated as TR1, TR2 and TW. Since their value will change with each optimization cycle, they will be defined as parameters using the PARASSIGN (Control > PARAMETER > Assign Parameter) command. The bounds, initial values and tolerances for the design variables are as shown below:
In de x

COSMOSM Advanced Modules

3-17

Chapter 3 Procedures and Examples

Bounds 8 TR1 25 8 TR2 20 3 TW 8

Initial Value 25 20 8

Tolerance 1 1 0.5

Using the design variables as well as other geometric dimensions, you will first build the initial geometry of the model parametrically. In the next step, the finite element mesh of the initial geometry will be subjected to loads and boundary conditions. The linear static stress analysis is performed as usual. After successful completion, you need to specify the input for design optimization, and solve for the optimal shape of the control arm. During the optimization cycle, the program will automatically change the design variable values as required and perform linear static stress analysis to satisfy the constraints. The following paragraphs describe all relevant steps in detail with illustrations.
1.

To start with, set a working plane by executing the following command:

Geo Panel: Geometry > GRID > Plane (PLANE) Rotation/sweep axis > 2 Offset on axis > 0.0 Grid line style > Solid

2.

Initialize all parameters used to build the model using the PARASSIGN (Control > PARAMETER > Assign Parameter) command. Let us start with the design variables (TR1, TR2 and TW):
Geo Panel: Control > PARAMETER > Assign Parameter (PARASSIGN) Parameter name > TR1 Data type > REAL Parametric real value > 25 Geo Panel: Control > PARAMETER > Assign Parameter (PARASSIGN) Parameter name > TR2 Data type > Real Parameter real value > 20

In addition to the required design variables, you can define some other dimensions of the model as parameters. These dimensions include the length between the centers of the shafts and their radii (L, R1 and R2).
In de x

3-18

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Geo Panel: Control > PARAMETER > Assign Parameter (PARASSIGN) Parameter name > R1 Data type > Real Parameter real value > 30 Geo Panel: Control > PARAMETER > Assign Parameter (PARASSIGN) Parameter name > R2 Data type > Real Parameter real value > 7 Geo Panel: Control > PARAMETER > Assign Parameter (PARASSIGN) Parameter name > L Data type > Real Parameter real value > 140

The parameters defined above can be listed on-screen using the PARLIST (Control > PARAMETER > List Parameter) command which provides a summary as shown below:
Num 1 2 3 4 5 6 3. Name TR1 TR2 TW R1 R2 L Type REAL REAL REAL REAL REAL REAL Value 25.000000 20.000000 8.000000 30.000000 7.000000 140.000000

For this problem, you need to increase the closure tolerances at keypoints. Establish keypoints for the centers of shafts as follows (note the parametric input for the center of the smaller shaft):
Geo Panel: Geometry > POINTS > Merge Tolerance (PTTOL) Tolerance > 0.001 Geo Panel: Geometry > POINTS > Define (PT) Keypoint > 1 XYZ-coordinate value > 0.0, 0.0, 0.0 Geo Panel: Geometry > POINTS > Define (PT) Keypoint > 2 XYZ-coordinate value > L, 0, 0

In de x

COSMOSM Advanced Modules

3-19

Chapter 3 Procedures and Examples 4.

Use the Auto scale icon to adjust the view on the screen. Next, draw circles at the points created above using the CRSCIRCLE (Geometry > CURVES > CIRCLES > By Center/Edge) command to model the two shafts. The shafts have a radius of R1 and R2 with thickness of TR1 and TR2 (defined as parameters). The prompts and input for creating the shafts are as below [ACTNUM (Control > ACTIVATE > Entity Label) command is used to activate labels of curves]:
Geo Panel: Geometry > CURVES > CIRCLES > By Center/Edge
(CRSCIRCLE) Curve > 1

XYZ-coordinate value of center of circle > 0.0, 0.0, 0.0 XYZ-coordinate value of center of circle > R1, 0.0, 0.0 Number of segments > 4 Geo Panel: Geometry > CURVES > CIRCLES > By Center/Edge
(CRSCIRCLE) Curve > 5

XYZ-coordinate value of center of circle > L, 0.0, 0.0 XYZ-coordinate value of center of circle > L+R2, 0.0, 0.0 Number of segments > 4 Geo Panel: Geometry > CURVES > CIRCLES > By Center/Edge
(CRSCIRCLE) Curve > 9

XYZ-coordinate value of center of circle > 0.0, 0.0, 0.0 XYZ-coordinate value of center of circle > R1+TR1, 0.0, 0.0 Number of segments > 4 Geo Panel: Geometry > CURVES > CIRCLES > By Center/Edge
(CRSCIRCLE) Curve > 13

XYZ-coordinate value of center of circle > L XYZ-coordinate value of center of circle > L+R2+TR2 Number of segments > 4
5.

Use the Auto scale icon to re-scale the screen view. The circles created above will be connected together by straight lines tangential to the inner circles of the shafts using the CRTANLIN (Geometry > CURVES > MANIPULATION MENU > Tangent btwn 2 Cr) command as illustrated below.

In de x

3-20

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Geo Panel:

Geometry > CURVES > MANIPULATION MENU > Tangent

btwn 2 Cr (CRTANLIN) Curve > 17

Curve 1 > 1 Curve 2 > 5 Break flag > Do not break Geo Panel: Geometry > CURVES > MANIPULATION MENU > Tangent
btwn 2 Cr (CRTANLIN) Curve > 18

Curve 1 > 4 Curve 2 > 8 Break flag > Do not break

The figure below shows a view of the two shafts connected together by tangential lines. Note that the lines protruding into the shaft thicknesses will be trimmed later.
Figure 3-5 Construction of Tangential Lines Connecting the Shafts

6.

Activate the display of point tables using the Status 1 icon. In the next step, you will construct two more lines parallel to the tangents created earlier representing the slotted portion of the model. In order to draw these lines, you can define a coordinate system along the tangential lines so that the new curves generated will be parallel to the tangential lines. Use the CSYS (Geometry > COORD SYS > By 3 Points) command as shown below:
Geo Panel: Geometry > COORD SYS > By 3 Points (CSYS) Coordinate system > 3 Coordinate system type > Cartesian Keypoint at origin > 19 Keypoint on the X-axis > 20 Keypoint on the X-Y plane > 1

In de x

COSMOSM Advanced Modules

3-21

Chapter 3 Procedures and Examples

Geo Panel: Geometry > COORD SYS > By 3 Points (CSYS) Generation number > 1 Beginning curve > 18 Ending curve > 18 Increment < 1 Generation flag > Translation X-displacement > 0.0 Y-displacement > -TW Z displacement > 0.0

Next, apply the CRGEN (Geometry > CURVES > GENERATION MENU > Generate) command at the top tangent line and input an offset of -TW (with respect to the new coordinate system) which represents the thickness of the slotted portion.
Geo Panel:
(CRGEN)

Geometry > CURVES > GENERATION MENU > Generate

Generation number > 1 Beginning curve > 17 Ending curve > 17 Increment > 1 Generation flag > Translation X-displacement > 0.0 Y-displacement > -TW Z-displacement > 0.0

Similarly apply the CSYS (Geometry > COORD SYS > By 3 Points) and CRGEN (Geometry > CURVES > GENERATION MENU > Generate) commands at the bottom tangent line as shown below:
Geo Panel: Geometry > COORD SYS > By 3 Points (CSYS) Coordinate system > 4 Coordinate system type > Cartesian Keypoint at origin > 21 Keypoint on the X-axis > 22 Keypoint on the X-Y plane > 1
7.

In de x

You now need to remove the lines protruding into the shaft thicknesses. This can be achieved by finding the intersection of the straight lines with the outer circles of the shafts and then subdividing the lines at the points of intersection. The lines inside the shafts can be then easily removed. The commands below illustrate these tasks:

3-22

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Geo Panel:

Geometry > POINTS > GENERATION MENU > Cr/Cr

Intersect (PTINTCC) Primary curve > 17

Beginning curve > 9 Ending curve > 14 Increment > 5 Tolerance > 0.000050 Geo Panel: Geometry > CURVES > MANIPULATION MENU > Break
Near Pt (CRPTBRK)

Curve to be broken > 17 Reference keypoint > 27 Original curve keeping flag > Do not keep Geo Panel: Geometry > CURVES > MANIPULATION MENU > Break
Near Pt (CRPTBRK)

Curve to be broken > 21 Reference keypoint > 28 Original curve keeping flag > Do not keep Geo Panel: Edit > DELETE > Curves (CRDEL) Beginning curve > 17 Ending curve > 22 Increment > 5

You need to repeat the above procedure for the remaining three lines of the slotted portion as outlined below: (Cryptic commands need to be typed in the command window; however, the command paths are also shown.)
Geo Panel: Geometry > POINTS > GENERATION MENU >
Cr/Cr Intersect (PTINTCC)

PTINTCC,18,12,15,3; Geo Panel: Geometry > CURVES > MANIPULATION MENU > Break
Near Pt (CRPTBRK)

CRPTBRK,18,29,0; CRPTBRK,22,30,0; Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,18,18,1; CRDEL,23,23,1; Geo Panel:
In de x

Geometry > POINTS > GENERATION MENU >

Cr/Cr Intersect (PTINTCC)

PTINTCC,19,9,14,5;

COSMOSM Advanced Modules

3-23

Chapter 3 Procedures and Examples

Geo Panel:

Geometry > CURVES > MANIPULATION MENU > Break

Near Pt (CRPTBRK)

CRPTBRK,19,31,0; CRPTBRK,23,32,0; Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,19,19,1; CRDEL,24,24,1; Geo Panel: Geometry > POINTS > GENERATION MENU >
Cr/Cr Intersect (PTINTCC)

PTINTCC,20,12,15,3; Geo Panel: Geometry > CURVES > MANIPULATION MENU > Break
Near Pt (CRPTBRK)

CRPTBRK,20,33,0; CRPTBRK,24,34,0; Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,20,20,1; CRDEL,25,25,1;
8.

Next, it is necessary to remove arcs of the outer circles at the connection with the slotted portion so that model domain is continuous. There are four such arcs and the CRPTBRK (Geometry > CURVES > MANIPULATION MENU > Break Near Pt) and CRDEL (Edit > DELETE > Curves) commands are applied to remove them as illustrated below:
Geo Panel: Geometry > CURVES > MANIPULATION MENU > Break
Near Pt (CRPTBRK)

CRPTBRK,9,27,0; CRPTBRK,25,31,0; Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,25,25,1; Geo Panel: Geometry > CURVES > MANIPULATION MENU > Break
Near Pt (CRPTBRK)

CRPTBRK,12,29,0; CRPTBRK,12,33,0; Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,28,28,1;
In de x

3-24

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Geo Panel:

Geometry > CURVES > MANIPULATION MENU > Break

Near Pt (CRPTBRK)

CRPTBRK,14,28,0; CRPTBRK,14,32,0; Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,29,29,1; Geo Panel: Geometry > CURVES > MANIPULATION MENU > Break
Near Pt (CRPTBRK)

CRPTBRK,15,30,0; CRPTBRK,29,34,0; Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,29,29,1;

The figure below shows the geometry created so far, with curve labels activated. These labels will be helpful in defining fillets at sharp corners, described next.
Figure 3-6. Interim Model Geometry with Curve Labels

9.

It is necessary to smooth the sharp corners at the intersection points of straight lines with the shafts. Fillets at these points can be defined using the CRFILLET (Geometry > CURVES > MANIPULATION MENU > Fillet) command. The labels of adjacent curves can be seen from the above figure for this command.
Geo Panel: Geometry > CURVES > MANIPULATION MENU > Fillet
(CRFILLET) Curve > 31

In de x

Curve 1 > 21 Curve 2 > 9 Radius of fillet > 5 Trim flag > Original curve keeping flag > Tolerance > 0.000001

COSMOSM Advanced Modules

3-25

Chapter 3 Procedures and Examples

Repeat the above command at other locations as shown below (as before, the cryptic commands need to be typed in the command window; command paths on the menu are also shown):
Geo Panel: Geometry > CURVES > MANIPULATION MENU > Fillet
(CRFILLET)

CRFILLET,33,22,27,5,1,0; CRFILLET,35,21,28,5,1,0; CRFILLET,37,22,15,5,1,0; CRFILLET,39,23,26,20,1,0; CRFILLET,40,24,12,20,1,0; CRFILLET,41,23,14,10,1,0; CRFILLET,42,24,30,10,1,0;

The figure below shows the initial geometry of the control arm for finite element model development. In order to generate the finite element mesh, you need to convert the geometry to a region entity and use the automatic meshing feature for regions.
Figure 3-7. Initial Geometry of the Control Arm for Finite Element Modeling

10.

There are four contours constituting one region entity for this model. The outer contour will be designated as the first contour and the inner ones will be numbered 2 through 4. The design variables TR1, TR2, and TW will alter the profiles of the outer contour and the middle inner contour in the slotted portion. The average element size is specified as half of the value of TW, the thickness of the slotted part. You also need to switch to the global coordinate system at this point. The contour and region definitions are illustrated below:
Geo Panel: Control > ACTIVATE > Set Entity (ACTSET) Set Label > CS

In de x

Coordinate system > 0

3-26

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Geo Panel: Geometry > CONTOURS > Define (CT) Contour > 1 Mesh flag > Esize Average element size >TW/2 Number of reference boundary curves > 1 Curve 1 > 23 Use selection set > No Geo Panel: Geometry > CONTOURS > Define (CT) CT,2,0,TW/2,1,21,0; CT,3,0,TW/2,1,1,0; CT,4,0,TW/2,1,5,0; Geo Panel: Geometry > REGIONS > Define (RG) Region > 1 Number of contours > 1 Outer contour > 1 Inner contour 1 > 2 Inner contour 2 > 3 Inner contour 3 > 4 Underlying surface > 0
11.

Generate the finite element mesh and define the element group, material properties, section constants, and higher order elements as illustrated below:
Geo Panel: Propsets > Element Group (EGROUP) Element group > 1 Element Category > Area Element Type (for area) > TRIANG Accept defaults .... Geo Panel: Propsets > Real Constant (RCONST) Associated element group > 1 Real constant set > 1 Start location of the real constants > 1 No. of real constants to be entered > 2 RC1: thickness > 20 RC2: material angle (Beta) > 0.0

In de x

COSMOSM Advanced Modules

3-27

Chapter 3 Procedures and Examples

Geo Panel: Propsets > Material Property (MPROP) Material property set > 1 Material property name > EX Property value > 200000 Material property name NUXY Property value > 0.3 Geo Panel: Meshing > AUTO MESH > Regions (MA_RG) Beginning region > 1 Ending region > 1 Increment > 1 Number of smoothing iterations > 0 Method of Sweeping Geo Panel: Meshing > AUTO MESH > Region Mesh Type (MARGCH) Beginning region > 1 Ending region > 1 Increment > 1 Element type > T Total element nodes > 6 Push flag > Yes Associate element group > 1

Define the boundary conditions and loads.

Geo Panel: LoadsBC > STRUCTURAL > DISPLACEMENT > Define
Curves (DCR)

Beginning curve > 1 Displacement label > All Value > 0.0 Ending curve > 4 Increment > 1 Geo Panel:
(PCR)

LoadsBC > STRUCTURAL > PRESSURE > Define Curves

Beginning curve > 5 Pressure magnitude > 4 Ending curve > 6 Increment > 1 Pressure at the end of direction 1 > 4 Pressure direction > Normal direction
In de x

3-28

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization 12.

Proceed with linear static stress analysis by executing the R_STATIC (Analysis > STATIC > Run Static Analysis) command. It is a good practice to postprocess the results from the preliminary design before starting the optimization loops due to several reasons. First, you can eliminate any modeling errors by inspecting the results. More importantly, you can make sure that the behavior constraints imposed on the model for optimization are realistic. For this example, the von Mises stress constraint has bounds of 10 and 225, whereas the computed von Mises stress for the preliminary design ranges from 0 to 90 (you can either use the STRMAX (Results > EXTREMES > Min/Max Stress) or STRPLOT (Results > PLOT > Stress) commands to process the von Mises stress). Therefore, the constraints specified for this example are indeed realistic.

13.

After the linear static analysis is successfully completed, you can proceed to define the input for shape optimization. The following commands for specifying optimization analysis options are found in the Analysis > OPTIMIZE/ SENSITIVITY menu tree.
14.

Define the design variables. There are three design variables (TR1, TR2, and TW) for this problem that will be applied in obtaining an optimal shape of the model under linear static analysis. Each of the design variable needs to be defined separately using the DVARDEF (Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define) command. You also need to specify the upper and lower bounds of the design variable under this command.
Geo Panel: Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define (DVARDEF) Design variable set number > 1 Design variable type > Shape Design variable parametric name > TR1 Design variable lower bound > 8 Design variable upper bound > 25 Design variable conv. tol. for optimization > 1 Preopt process: zero=random nonzero=perturb_ratio > 0.0 Geo Panel: Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define (DVARDEF) Design variable set number > 2 Design variable type > Shape Design variable parametric name > TR2 Design variable lower bound > 8

In de x

COSMOSM Advanced Modules

3-29

Chapter 3 Procedures and Examples

Design variable upper bound > 20 Design variable conv. tol. for optimization > 1 Preopt process: zero=random nonzero=perturb_ratio > 0.0 Geo Panel: Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define (DVARDEF) Design variable set number > 3 Design variable type > Shape Design variable parametric name > TW Design variable lower bound > 3 Design variable upper bound > 8 Design variable conv. tol. for optimization > 0.5 Preopt process: zero=random nonzero=perturb_ratio > 0.0

The DVARLIST (Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > List) command can be used to obtain an on-screen listing of design variables defined above. The listing for this problem is as shown below:
Set 1 2 3 Type Shape Shape Shape DvName TR1 TR2 TW Value 2.5000e+001 2.0000e+001 8.0000e+000 Lower/ Upper Bound 8.0000e+000 2.5000e+001 8.0000e+000 2.0000e+001 3.0000e+000 8.0000e+000 15. Tolerance/ Perturb Ratio 1.0000e+000 0.0000e+000 1.0000e+000 0.0000e+000 5.0000e-001 0.0000e+000

The objective function to be minimized (or maximized) and the associated parameters for nodal or elemental type functions is defined using the OP_OBJDEF (Analysis > OPTIMIZE/SENSITIVITY > OBJECTIVE FUNCTION > Define Function) command. For this example, you will be minimizing volume as an elemental quantity using linear approximation to start with. Quadratic approximations with cross terms as well as cubic terms will be used for subsequent approximations of the objective function (approximation type 3).
Geo Panel: Analysis > OPTIMIZE/SENSITIVITY> OBJECTIVE FUNCTION > Define Function (OP_OBJDEF) Type > Elemental Objective > Minimize

In de x

Elemental objective function name > Volume

3-30

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Analysis type > Linear Unused option > Unused option > Objective function convergence tolerance (ratio) > 0.01 Objective function approximation type > Linear + quadratic

The objective function input can be listed on-screen using the OP_OBJLIST (Analysis > OPTIMIZE/SENSITIVITY > OBJECTIVE FUNCTION > List Function) command which provides the following information:
Objtyp Objname Objective Layer Analysis 16. = Elemental = VOLUME : Minimize =1 = Static Tolerance ApprxType Value Face = 1.000000e-002 =1 = 2.003479e+005 =1

The remaining input for performing shape optimization of the control arm is the constraint definition. The OP_CONDEF (Analysis > OPTIMIZE/ SENSITIVITY > BEHAVIOR CONSTRAINT > Define) command is used to define the constraints as illustrated below:
Geo Panel: Analysis > OPTIMIZE/SENSITIVITY > BEHAVIOR CONSTRAINT > Define (OP_CONDEF) Constraint set number > 1 Type > Nodal Nodal constraint name > VON Analysis type > Linear Criterion flag > Max Abs Layer number > 1 Face flag (shell) > Top Load case/time step > 1 Entity type assoc. with constraint > Nodes Beginning node > 1 Ending node > NDMAX See paragraph below Increment of nodes > 1 Constraint lower bound > 10 Constraint upper bound > 225 Constraint feasibility tolerance > 3 Constraint approximation type > Linear + Quad

In de x

COSMOSM Advanced Modules

3-31

Chapter 3 Procedures and Examples

In the above, the second prompt seeks your input for the type of constraint, in this case, nodal. The criterion flag specifies whether the absolute maximum, or algebraic maximum, etc., is applied for evaluating the constraints. The entity type associated with the constraint are nodes for this example. Since the label of the ending node is not known in advance, and will keep changing for different optimization loops, you can input a parametric equivalent (NDMAX) for that prompt. The constraint upper and lower bounds as well as the tolerance are input for the next three prompts. The last prompt seeks your input for constraint approximation type. The default option of 1 for this prompt will use linear and quadratic terms for approximations of the constraint function. The constraint input can be listed on-screen using the OP_CONLIST (Analysis > OPTIMIZE/SENSITIVITY > BEHAVIOR CONSTRAINT > List Behavior Const) command.
17.

The last set of input you need to specify before beginning the optimization loops is furnished using the A_OPTIMIZE (Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP > Optimize Analysis Options) command. The command and its input are shown below:
Geo Panel: Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP >
Optimize Analysis Options (A_OPTIMIZE) Maximum number of optimization loops > 10

Convergence check stages > To the previous loop No. of consec. infeasible designs > 5 Output print flag > On Echo option flag > On Restart flag > Off Type of analysis > Static

In de x

In the above, the default option of 1 for the second prompt means that the convergence is achieved if the change in the objective function and design variables compared to the previous loop and the best design so far is less than the tolerance. If the initial design is infeasible and your model goes through five consecutive infeasible designs, the optimization process will be halted by default. The type of analysis to be run in the optimization loop is static stress analysis by default, and the program will prompt you further if you plan to run multidisciplinary analyses. The options you specified under the A_OPTIMIZE (Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP > Optimize Analysis Options) command can be listed on-screen using the A_LIST (Analysis > List Analysis Option) command (from the ANALYSIS submenu) as shown below:

3-32

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Geo Panel: Analysis > List Analysis Option (A_LIST) Component > Optimization
18.

You can now proceed to perform the shape optimization analysis of the slotted control arm. Use the R_OPTIMIZE (Analysis > OPTIMIZE/SENSITIVITY > OPTIMIZE LOOP > Run Optimize Analysis) command to start the optimization loops. The results of the static stress analysis performed will be written as usual to the jobname.OUT file. Note that this file will only contain results corresponding to the final design. If you want to store analysis results of all loops. you need to use the PRINT_OPS (Analysis > OUTPUT OPTIONS > Set Print Options) command to append all subsequent results to this file. For optimization iteration summary of all loops, you can view the jobname.OPT file. Lets create four windows using the New Win icon. Move to the first window by clicking inside he window. In main window, we will plot the von Mises stresses using the ACTSTR and STRPLOT commands (Results > PLOT > Stress). In window 1, we will study the variation of objective functions versus the optimization loops. Move to window 2 and execute the ACTXYPOST (Display > XY PLOTS > Activate Post-Proc) and XYPLOT (Display > XY PLOTS > Plot Curves) commands (Display > WINDOWS menu tree) as illustrated below:
Geo Panel: Results > SET UP > Set to Post-Proc (ACTPOST) Select optimization Geo Panel: Display > XY PLOTS > Activate Post-Proc (ACTXYPOST) Graph number > 1 Y_variable > OP_OBJ Type of results > FEA Graph color > 12 Graph line style > Solid Graph symbol style > 1 Graph id > OP_OBJ

19.

20.

21.

Adjust the X- and Y-axis intervals of the X-Y plot (for a clear display) by making use of the SETXYPLOT (Display > XY PLOTS > Set Plot Parameter) command as illustrated below:
In de x

COSMOSM Advanced Modules

3-33

Chapter 3 Procedures and Examples

Geo Panel: Display > XY PLOTS > Set Plot Parameter (SETXYPLOT) X logarithmic > No Y logarithmic > No Number of X intervals > 7 Number of Y intervals > 4 Accept Defaults Geo Panel: Display > XY PLOTS > Plot Curves (XYPLOT) Plot graph 1 > Yes
22.

In window 2, we will study the variation of the design variables versus the optimization loops. Since there are three design variables in this example, you will use set numbers 1, 2, and 3 for TR1 TR2 and TW respectively. Move to window 3 and execute the ACTXYPOST (Display > XY PLOTS > Activate Post-Proc) and XYPLOT (Display > XY PLOTS > Plot Curves) commands as illustrated below (remember to use the SETXYPLOT (Display > XY PLOTS > Set Plot Parameter) command to adjust the X-axis and Y-axis intervals):
Geo Panel: Display > XY PLOTS > Activate Post-Proc (ACTXYPOST) Graph number > 1 Y_variable > OP_DVAR Set number > 1 Graph color > 12 Graph line style > Solid Graph symbol style > 1 Graph id > OP_DVAR-1 Geo Panel: Display > XY PLOTS > Activate Post-Proc (ACTXYPOST) Graph number > 2 Y_variable > OP_DVAR Set number > 2 Graph color > 14 Graph line style > Solid Graph symbol style > 1 Graph id > OP_DVAR-2 Geo Panel: Display > XY PLOTS > Activate Post-Proc (ACTXYPOST) Graph number > 3 Y_variable > OP_DVAR Set number > 3 Graph color > 8 Graph line style > Solid

In de x

3-34

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Graph symbol style > 1 Graph id > OP_DVAR-3 Geo Panel: ... Display > XY PLOTS > Set Plot Parameter (SETXYPLOT)

Geo Panel: Display > XY PLOTS > Plot Curves (XYPLOT) Plot graph 1 > Yes Plot graph 2 > Yes Plot graph 3 > Yes
23.

In window 3, we will study the variation of the behavior constraint versus the optimization loops. Move to window 3 and execute the ACTXYPOST (Display > XY PLOTS > Activate Post-Proc) and XYPLOT (Display > XY PLOTS > Plot Curves) commands as illustrated below:
Geo Panel: Display > XY PLOTS > Activate Post-Proc (ACTXYPOST) Graph number > 1 Y_variable > OP_CON Set number > 1 Type of results > FEA Graph color > 12 Graph line style > Solid Graph symbol style > 1 Graph id > OP_CON-1 Geo Panel: ... Display > XY PLOTS > Set Plot Parameter (SETXYPLOT)

Geo Panel: Display > XY PLOTS > Plot Curves (XYPLOT) XYPLOT Plot graph 1 > Yes

You will obtain a plot as shown in the following figure. Note that you can use the XYPTLIST (Display > SET UP > List Points) command to list the values of the optimization results for each loop on-screen.

In de x

COSMOSM Advanced Modules

3-35

Chapter 3 Procedures and Examples Figure 3-8. Final Stress Distribution and Convergence Plots

In de x

3-36

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Sensitivity Study of a Control Arm Bracket in Frequency

In this example, you are required to find the size and location of the cutout under the first (out-of-plane) mode of free vibration. The bracket thickness is 0.3 cm, modulus of elasticity is 2 x 107 N/cm2, and Poissons ratio is 0.3. The material mass density is 0.0075 Kg/cm3. The following figure shows the initial geometry of the bracket.
Figure 3-9. Initial Geometry of the Control Arm Bracket
r = 1.0 r=

(See page 3-1.)

1.5

1.5

S h ell3
Thickness E = = = = 0.3 cm 2 x 107 N/cm2 0.30 0.0075 Kg/cm2

15.0

t1 0.5 1.0 t2 X

t1 1.0

Note: All dimensions in centimeters.

5.0

5.0

The design variables for this problem as seen from the above figure are designated as T1 and T2. Since their value will change with each sensitivity run, they will be defined as parameters using the PARASSIGN (Control > PARAMETER > Assign Parameter) command. The initial values and bounds for the design variables are as shown below:
Design Variable T1 T2
In de x

Initial Value 2.5 3.5

Lower Bound 0.5 1.5

Upper Bound 2.5 3.5

COSMOSM Advanced Modules

3-37

Chapter 3 Procedures and Examples

Using the design variables as well as other geometric dimensions, you will first build the initial geometry of the model parametrically. In the next step, the finite element mesh of the initial geometry will be subjected to boundary conditions. The frequency analysis is then performed as usual. After successful completion, you need to specify the input for sensitivity study. During he sensitivity runs, the program will automatically change the design variable values as required and perform frequency analysis. The following steps describe all relevant procedures in detail with illustration.
1.

To start with, set the working plane and the view, and initialize the design variables T1 and T2 by executing the following commands:
Geo Panel: PLANE; Geometry > GRID > Plane (PLANE)

Use the VIEW icon VIEW; Geo Panel: Control > PARAMETER > Assign Parameter (PARASSIGN) PARASSIGN,T1,REAL,2.5, PARASSIGN,T2,REAL,3.5,
2.

Establish the following keypoints for use in geometry creation:

Geo Panel: Geometry > POINTS > Define (PT) PT,1,0,0,0, PT,2,10,0,0, PT,3,5,15,0, Use Auto scaling icon SCALE,0,

3.

Scale the view using the Auto scaling icon. Construct a triangle connecting the three keypoints created above using the CRLINE (Geometry > CURVES > Line with 2 Pts) command as shown below ACTNUM (Control > ACTIVATE > Entity Label) command is used to activate labels of curves):
Geo Panel: Control > ACTIVATE > Entity Label (ACTNUM) ACTNUM,CR,1; Geo Panel: Geometry > CURVES > Line with 2 Pts (CRLINE) CRLINE,1,1,3, CRLINE,2,1,2, CRLINE,3,2,3,

In de x

3-38

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization 4.

Next, we will generate a line parallel to the base of the triangle which aligns along the X-axis of the default Cartesian coordinate system (label 0). The CRGEN (Geometry > CURVES > GENERATION MENU > Generate) command will be used with a Y-axis offset of T2 as illustrated below. If you activate the SNP and PIC icons, you can select entities from the screen using the left button of the mouse. Note that you can still use the keyboard.
Geo Panel:
(CRGEN)

Geometry > CURVES > GENERATION MENU > Generate

Generation number [1] > Pick/Input Beginning Curve > 2 Pick/Input Ending Curve > 2 Increment > 1 Generation flag > Translation X-Displacement > 0.0 Y-Displacement > T2 Z-Displacement > 0.0
5.

In order to create lines parallel to the inclined sides of the triangle, you need to make use of the local coordinate systems. First, define a local Cartesian coordinate system (label 3) with its X-axis aligned with the left inclined side and repeat the CRGEN (Geometry > CURVES > GENERATION MENU > Generate) command as shown below:
Geo Panel: Geometry > COORD SYS > By 3 Points (CSYS) CSYS,3,0,3,1,2, Geo Panel:
(CRGEN)

Geometry > CURVES > GENERATION MENU > Generate

CRGEN,1,1,1,1,0,0,T1,0,

Next, define another local Cartesian coordinate system (label 4) with its X-axis aligned with the right inclined side and repeat the CRGEN (Geometry > CURVES > GENERATION MENU > Generate) command as shown below:
Geo Panel: Geometry > COORD SYS > By 3 Points (CSYS) CSYS,4,0,2,3,1, Geo Panel:
(CRGEN)

Geometry > CURVES > GENERATION MENU > Generate

CRGEN,1,3,3,1,0,0,T1,0,

In de x

COSMOSM Advanced Modules

3-39

Chapter 3 Procedures and Examples

The following figure shows a plot of keypoints and curves generated so far, along with the coordinate systems. In the next step, you will find the points of intersection of curves 4 through 6 and delete the unwanted segments.
6.

Find the points of intersection of the straight lines bounding the inner triangle using the CRINTCC (Geometry > CURVES > GENERATION MENU > Cr/Cr Intersect) command and delete the unwanted segments using the CRDEL (Edit > DELETE > Curves) command:
Geo Panel: Geometry > CURVES > GENERATION MENU > Cr/Cr
Intersect (CRINTCC)

Figure 3-10. Construction of the Control Arm Geometry

CRINTCC,4,5,6,1,2,0.00005, CRINTCC,7,8,8,1,2,0.00005, Geo Panel: Edit > DELETE > Curves CRDEL,5,6,1, CRDEL,4,10,6, CRDEL,11,12,1,

The geometry of the control arm is now as shown below. Note that the corners of the inner triangle will be rounded off using fillets. At the corners of the outer triangle, you will construct circles and connect them to the rest of the geometry using procedures similar to the above.
7.

Figure 3-11. Construction of the Control Arm Geometry

Next, you need to create fillets at the sharp corners created by the intersection of straight lines with the circles. The CRFILLET (Geometry > CURVES > MANIPULATION MENU > Fillet) command is applied as shown below at the sharp corners:

In de x

3-40

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Geo Panel:

Geometry > CURVES > MANIPULATION MENU > Fillet

(CRFILLET)

CRFILLET,10,9,7,0.5,1,0,1E-006, CRFILLET,12,7,8,0.5,1,0,1E-006, CRFILLET,14,9,8,0.5,1,0,1E-006,

8.

Define concentric circles at the three corners as shown below:

Geo Panel: Geometry > CURVES > CIRCLES > Circle in Plane
(CRPCIRCLE)

CRPCIRCLE,16,1,2,1,360,4, CRPCIRCLE,20,2,1,1,360,4, CRPCIRCLE,24,3,20,1,360,4, Use Auto scale icon SCALE,0, Geo Panel: Geometry > CURVES > CIRCLES > Circle in Plane
(CRPCIRCLE)

CRPCIRCLE,28,1,2,2,360,6, CRPCIRCLE,34,2,1,2,360,6, CRPCIRCLE,40,3,20,2,360,6, Use Auto scale icon SCALE,0,

Figure 3-12. Construction of the Con

The geometry of the slotted control arm you constructed up to this stage is as shown below. You need to find the intersection points of the straight lines with the circles and remove the unwanted segments.
9.

Find the points of intersection of the straight lines with the circles using the CRINTCC (Geometry > CURVES > GENERATION MENU Cr/Cr Intersect) command and delete the unwanted segments using the CRDEL (Edit > DELETE > Curves) command:

In de x

COSMOSM Advanced Modules

3-41

Chapter 3 Procedures and Examples

Geo Panel:

Geometry > CURVES > GENERATION MENU >

Cr/Cr Intersect (CRINTCC)

CRINTCC,1,29,45,16,2,0.00005, Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,1,49,48, Geo Panel: Geometry > CURVES > GENERATION MENU >
Cr/Cr Intersect (CRINTCC)

CRINTCC,3,38,40,2,2,0.00005, Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,3,52,49, Geo Panel: Geometry > CURVES > MANIPULATION MENU > Break
Near Pt (CRPTBRK)

CRPTBRK,2,37,0, CRPTBRK,52,43,0, Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,2,53,51,
10.

Smoothen the sharp apices of the control arm using fillets as illustrated below:
Geo Panel: Geometry > CURVES > MANIPULATION MENU > Fillet
(CRFILLET)

CRFILLET,53,33,52,1.0,1,0,1E-006, CRFILLET,54,34,52,1.0,1,0,1E-006, CRFILLET,55,38,51,1.0,1,0,1E-006, CRFILLET,56,46,48,1.0,1,0,1E-006, CRFILLET,57,50,51,1.5,1,0,1E-006, CRFILLET,58,45,48,1.5,1,0,1E-006,

11.

Delete the remaining unwanted segments as illustrated below:

Geo Panel: Edit > DELETE > Curves (CRDEL) CRDEL,28,29,1, CRDEL,39,49,10, CRDEL,40,47,7,

With the steps executed so far, the initial geometry of the control arm is now fully constructed. The following figure shows a view of the initial geometry. You can next proceed to generate the finite element model.
In de x

3-42

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization 12.

In order to prepare the initial geometry for meshing, execute the CT (Geometry > CONTOURS > Define) and RG (Geometry > CONTOURS > Regions) commands as illustrated below:
Geo Panel: Geometry > CONTOURS > Define (CT) CT,1,0,1,1,51,0, CT,2,0,1,1,8,0, CT,3,0,1,1,22,0, CT,4,0,1,1,16,0, CT,5,0,1,1,25,0, Geo Panel: Geometry > REGIONS > Regions (RG) RG,1,5,1,2,3,4,5,0,

Figure 3-13. Completed Initial Geometry of the Control Arm

13.

You will be using triangular 3-node shell elements with an average element size of 1 units. The control arm will be held in place at the two bottom openings against rotations and translations in all directions (we will also be switching back to the default Cartesian global coordinate system at this stage):
Geo Panel: Control > ACTIVATE > Set Entity (ACTSET) ACTSET,CS,0, Geo Panel: Propsets > Element Group (EGROUP) EGROUP,1,SHELL3,0,0,0,0,0,0,0, Geo Panel: Propsets > Real Constant (RCONST) RCONST,1,1,1,6,0.3,0,0,0,0,0, Geo Panel: Propsets > Material Property (MPROP) MPROP,1,EX,20.E6,DENS,0.0075, Geo Panel: Meshing > AUTO MESH > Regions (MA_RG) MA_RG,1,1,1,0,1, Geo Panel: LoadsBC > STRUCTURAL > DISPLACEMENT >
Define Contours (DCT)

DCT,3,ALL,0,4,1,

In de x

COSMOSM Advanced Modules

3-43

Chapter 3 Procedures and Examples

The following figure shows the completed finite element mesh of the initial design. You need to next execute the analysis to compute the first out-of-plane free vibration mode.
14.

Figure 3-14. Finite Element Mesh of the Initial Design

Before starting the sensitivity study, you need to run the analysis once to make sure that you did not make any errors in modeling or in the required features of analysis. Execute the R_FREQUENCY (Analysis > FREQUENCY/ BUCKLING > Run Frequency) command and use the default options from the A_FREQUENCY (Analysis > FREQUENCY/BUCKLING > Frequency Options) command which include the computation of the first mode of vibration using the subspace iteration algorithm. Note that it is possible to study the sensitivity of the models frequency of different modes. It is a good practice to postprocess the results from the preliminary design before starting the sensitivity study so that you can eliminate any modeling errors by inspecting the results. The figure below was obtained using the DEFPLOT (Results > PLOT > Deformed Shape) command illustrated as follows:
Use Viewing icon to set isometric view (XYZ) Geo Panel: Results > PLOT >
Deformed Shape (DEFPLOT) Mode shape number > 1 Figure 3-15. Deformed Shape Plot of the Initial Design (First Mode)

15.

In de x

3-44

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization 16.

With the successful completion of the initial model, you can now proceed to define the sensitivity data. As required, you will define frequency as the response quantity:
Geo Panel: Analysis> OPTIMIZE/SENSITIVITY> RESPONSE QUANTITY> Define (SN_RESPDEF) Response set number > 1 Type > Freq Mode shape number > 1

Define the design variables. There are two design variables (T1 and T2) for this problem. Each of the design variable needs to be defined separately using the DVARDEF (Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define) command. You also need to specify the upper and lower bounds of the design variable under this command.
Geo Panel: Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define (DVARDEF) Design variable set number > 1 Design variable type > Shape Design variable parametric name > T1 Design variable lower bound > 0.5 Design variable upper bound > 2.5 Accept defaults ... Geo Panel: Analysis > OPTIMIZE/SENSITIVITY > DESIGN VARIABLES > Define (DVARDEF) Design variable set number > 2 Design variable type > Shape Design variable parametric name > T2 Design variable lower bound > 1.5 Design variable upper bound > 3.5 Accept defaults ...

Lastly, the A_SENSITIV (Analysis > OPTIMIZE/SENSITIVITY > SENSITIVITY RUN > Options) command is used to specify the sensitivity study options and the analysis type:
Geo Panel: Analysis > OPTIMIZE/SENSITIVITY > SENSITIVITY RUN >
Options (A_SENSITIV) Sensitivity type > 1 by 1

In de x

Number of increment > 5

COSMOSM Advanced Modules

3-45

Chapter 3 Procedures and Examples

Output print flag > On Echo option flag > On Restart flag > Off Type of analysis > Frequency

Now, you can use R_SENSITIV (Analysis > OPTIMIZE/SENSITIVITY > SENSITIVITY RUN > Run Analysis) command to start the sensitivity study. After ten runs, a message sensitivity study completed will be displayed.
17.

Use the ACTPOST (Results > SET UP > Set to Post-Proc), ACTXYPOST Display > XY PLOTS > Activate Post-Proc) and XYPLOT (Display > XY PLOT > Plot Curves) commands to view the variation of the fundamental frequency versus the design variable values.

In de x

3-46

COSMOSM Advanced Modules

Numerical Aspects

Introduction
The basic features and procedures for design optimization and sensitivity were introduced in Chapter 1 to Chapter 3. These materials are sufficient in most situations to perform the design problems. This chapter is intended for the users who wish to learn more about the implementation of the optimizer in OPTSTAR. Many of the methods used in the optimizer use numerical parameters to control the convergence, tolerance, and so on. Default values are provided for those parameters and in most situations they are working perfectly. Nevertheless, the users are allowed to override these constants by using appropriate commands. The material covered in this chapter includes: basic statements of optimization problems, function approximation, singular value decomposition, the modified feasible direction method, the sequential linear programming method, move limits of design variables, constraint trimming, and convergence criteria.

In de x

COSMOSM Advanced Modules

4-1

Chapter 4 Numerical Aspects

Basic Statements of Optimization Problems

The basic problem that we consider in OPTSTAR is the minimization of a function subject to inequality constraints. Minimize: Subject to: Objective function Side constraints Behavior constraints where: X = ith design variable i
Figure 4-1. Objective Function and Constraints in the Design Space with Two Design Variables

Figure 4-1 shows the objective function and constraints in the design space with two design variables. It is noted that for a maximization problem we can always transform it to be a minimization one by multiplying the objective function by -1.

X2 g1 (X1,X 2) F(X1,X 2) = Constant

X2 = X2

g (X1,X 2) 2 X2 = X2 X1 X1 = XL 1 X1 = X U 1
L

Function Approximation
The main idea of design optimization presented herein relies on finding a mathematical relationship between the objective function or constraints and design variables. Such a relationship is generally not known in advance. OPTSTAR makes use of the existing response (objective function or constraints) at a number of points in the design space to construct a polynomial approximation to the response at other points. The optimization process is then applied to the approximate problem represented by the polynomial approximation.

In de x

4-2

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Linear, quadratic, cubic, or quadratic cross-terms may be selected for the polynomial approximation depending on the approximation type. They are as follows:

where:

Nd Xi ai, bi, cij, di

= Number of design variables = ith design variable = Coefficients to be determined

OPTSTAR (Advanced Modules Manual, Part 2, page 4-3)

Table 2-1. Approximation Types

Flag
-1 -2 -3 0 +1 +2 +3 Only linear terms (
1 , 2

Type
) )

Only quadratic terms ( Only cubic terms (

1 ,

1 , 3 5

Automatic determination of approximation type Start with linear and add quadratic terms if needed (
1 , 2 , 3

Start with linear and add quadratic and quadratic cross terms if needed

1 , 2 ,

3 ,

Start with linear and add quadratic, quadratic cross terms and cubic terms if needed ( 1 , 2 , 3 , 4 , 5 )

For example, if we want to fit the response by the approximation type +1, we need Nd+2 (extra one for quality factor) design sets to start with a linear approximation. As the optimization loop number exceeds this value, new design sets are added to the linear approximation until it reaches 2Nd+2. After then, a quadratic approximation is adopted. The coefficients of the polynomial function are determined by a least squares regression.

In de x

COSMOSM Advanced Modules

4-3

Chapter 4 Numerical Aspects

Singular Value Decomposition

Singular value decomposition (SVD) is used for regression analysis. It is advantageous to use SVD because it can handle:

the mathematical irony that least-squares problems are both overdetermined

(number of data points greater than number of parameters) and underdetermined (ambiguous combinations of parameters exist),

sets of equations that are either singular or very close to singular.

The svd_thr of the OP_CONTROL command sets the threshold for singular values allowed to be non-zero. it is a ratio of the maximum singular value of the matrix. The svd_iter of the OP_CONTROL command controls the maximum number of iterations allowed to detect singular values.

The Modified Feasible Direction Method

Overall Process
After the objective function and constraints are approximated and their gradients with respect to the design variables are calculated based on the approximation, we are able to solve the approximate optimization problem. One of the algorithms used in the optimizer is called the Modified Feasible Direction method (MFD). The solving process is iterated until convergence is achieved. It is important to distinguish the iteration inside the approximate optimization from the loop in the overall optimization process. Figure 4-2 shows the iterative process within each optimization loop.

In de x

4-4

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization Figure 4-2. The Modified Feasible Direction Algorithm

Ge ne ra l Optimiza tion
Parametric Geometry and Mesh X (1) i m=1 Update Geometry and Mesh (m) (if needed) X i

MF D 1. q = 0, X q = X m 2. q = q + 1 3. Evaluate objective function F(X i ) and behavior constraints g j (X i ) 0 where j = 1, 2, . . ., N c 4. Identify critical and potentially critical constraints, N c

Initial Analysis

Define Design Variables Objective Function Behavior Constraints

Perform Analysis

Postprocessing

Approximate Objective Function and Constraints

Improved (m+1) Design X i

m = m +1

Yes

Requirements Achieved? Optimization Loop

No

5. Calculate gradient of objective function F(X i) and behavior constraints g k (X i ) where k = 1, 2, . . ., N c 6. Find a usable-feasible search direction S q 7. Perform a one-dimensional search X q = X q-1 + S q 8. Check convergence. If satisfied, go to 9. Otherwise, go to 2. 9. X m +1 = X q

Search Direction
In order to make any further improvement in an optimization loop, a new search direction must be found that continues to reduce the objective function but keeps the design feasible. We seek a usable-feasible search direction, in which:

A usable direction is the one that reduces the objective function, and A feasible direction is the one that a small move in this direction will not violate
In de x

the constraints.

COSMOSM Advanced Modules

4-5

Chapter 4 Numerical Aspects

This situation is shown in Figure 4-3. To find the search direction, active and violated constraints have to be identified. A constraint is active if its value lies between mfd_viol (+ve value) and mfd_active (-ve value) as defined by the command OP_CONTROL. A small positive value, mfd_viol, is allowed before categorizing a constraint as violated. If a constraint is less than mfd_active, it is then inactive. These conditions are displayed in Figure 4-4.

Figure 4-3. Usable-Feasible Search Direction

X2
Feasible Sector F(X1 , X 2 ) = Constant Usable Feasible Sector

g2 (X1 , X 2 ) = 0

S
Usable Sector

g1 (X1 , X 2 ) = 0 X1

Figure 4-4. Tolerances for a Constraint in a Two Design Variable Space

X2 g j (X) = mfd_ active g j (X) < 0 g j (X) = 0 g j (X) > 0 X1

g j (X) = mfd_ viol

Convergence to the Optimum

The optimizer uses several criteria to decide when to end the iterative search process. The criteria are described as the following. Maximum iterations - The maximum number of iterations (search directions) is defined by mfd_ifsrch in the command OP_CONTROL. It is intended to avoid excessive computations and the default value, 100, is usually more than enough for finding an optimum. Changes of objective function - To measure the progress made in the successive iterations, one of the following criteria is to be satisfied:

In de x

4-6

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

or

where mfd_relobj and mdf_absobj are the specified tolerances defined in the command OP_CONTROL. The first criterion, relative change, is an indication of convergence if the objective function is large. However, the convergence is controlled by the second criterion, absolute change, if the objective function is small. The number of successive iterations is defined by mfd_conv in the command OP_CONTROL.

Satisfaction of Kuhn-Tucker Conditions

Besides the previously mentioned criteria, the Kuhn-Tucker conditions necessary for optimality must be satisfied. Unconstrained problems - The conditions degenerate to the case where the gradient of the objective function vanishes:

In de x

It is noted that this Figure 4-5. Relative and Global Minima in the Design Space condition is necessary but not sufficient for F(X) optimality. To ensure a function to be a 2 minimum, the 1 Hessian matrix * * (second derivatives with respect to design variables) must be positive-definite. X Also, the optimum is in a sense of relative optimum rather than global one. In general, the conditions to ensure a global minimum can rarely be demonstrated. If a global minimum is intended, the designers must restart the optimization process from different initial points to check if other solutions are possible. Figure 4-5 shows the relative and global minima in the design space.

COSMOSM Advanced Modules

4-7

Chapter 4 Numerical Aspects

Constrained problems - The conditions of optimality are more complex. By using the Lagrangian multiplier method, we define the Lagrangian function as the following:

where tj is a slack variable which measures how far the jth constraint is from being critical. Differentiating the Lagrangian function with respect to all variables we obtain the Kuhn-Tucker conditions which are summarized as follows:
1.

2.

The corresponding j is zero if a constraint is not active.

The physical interpretation of these conditions is that the sum of the gradient of the objective function and the scalars j times the associated gradients of the active constraints must vectorally add to zero as shown in Figure 4-6.
Figure 4-6. Kuhn-Tucker Conditions at a Constrained Optimum
X2 2 g 2 (X) g2 (X) = 0 F (X)

1 g 1(X)

(X) g 2 (X) g 1 (X)

F (X) g 1 (X) = 0

In de x

The Kuhn-Tucker conditions are also sufficient for optimality when the number of active constraints is equal to the number of design variables. Otherwise, sufficient conditions require the second derivatives of the objective function and constraints (Hessian matrix) similar to the unconstrained one. If the objective function and all of the constraints are convex, the Kuhn-Tucker conditions are also sufficient for global optimality.

4-8

COSMOSM Advanced Modules

X1

Part 2 OPSTAR / Optimization

The Sequential Linear Programming Method

In OPTSTAR, the other algorithm for solving the approximate optimization problem is called the Sequential Linear Programming method (SLP). The iterative process within each optimization loop is shown in Figure 4-7. In Figure 4-8, linear approximations to the objective function and constraints are displayed.
Figure 4-7. The Sequential Linear Programming Method

Ge ne ra l Optimiza tion
Parametric Geometry and Mesh X (1) i m=1 Update Geometry and Mesh (m) (if needed) X i

Initial Analysis

SLP
p m 1. p = 0, X = X
Postprocessing

Define Design Variables Objective Function Behavior Constraints

Perform Analysis

Approximate Objective Function and Constraints

2. p = p + 1 3. Linearize the problem at X p-1 by creating a first order Taylor Series expansion of the objective function and retained constraints: F(X) = F(X p-1) + F(X p-1) (X - X p-1) g j (X) = g j (X p-1) + g j (X p-1) (X - X p-1) 4. Use this approximation of optimization instead of the original nonlinear functions: Minimize: F(X) L U Subject to: g j (X) 0 and X i X i X i 5. Find an improved design X p (using the MFD algorithm) 6. Check feasibility and convergence. If both of them are satisfied, go to 7. Otherwise, go to step 2. 7. X m +1 = X p

Improved (m+1) Design X i

m = m +1

Yes

Requirements Achieved?

No

Optimization Loop

In de x

COSMOSM Advanced Modules

4-9

Chapter 4 Numerical Aspects Figure 4-8. Linear Approximation to Objective Function and Constraints
X2 g (X)
1

Linear Approximation to F (X p)

Linear Approximation to g (X p)
1

F(X) True Oprimum

g (X)
2

Approximate Optimum

Linear Approximation to g (X p)
2

X1

The flag slp_iter controls the maximum number of iterations (repeated linearization process) and slp_conv controls the number of successive iterations for convergence check. Both flags are specified in the command OP_CONTROL.

Move Limits of Design Variables

During the optimization process, each design variable is bounded by its global lower and upper limits as shown below:

Within each optimization loop, a percentage change, move limit ML, is temporarily applied to the current value of design variable such that local bounds are created as:

The physical interpretation of move limits is that the optimizer creates a temporary box around the current value of design variable as shown in Figure 4-9.
In de x

4-10

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization Figure 4-9. Move Limits of Two Design Variables
X2

Objective Function

X1
Behavior Constraint

X1

X1

X1 X1

X1

The command OP_DVMOVE controls the initial move limit, its lower and upper bounds, MlL and MuL, and multiplier, . With these parameters, the move limit for a subsequent optimization loop is computed by the following formula:

Constraint Trimming
Most structural optimization problems contain more constraints than what are necessary to adequately guide the design. The constraints are filtered such that only violated and potentially critical ones are considered in the optimizer to increase the computational efficiency. To identify what is meant by potentially critical, a trimming (truncation) factor is applied to the non-violated constraints by using the command OP_CONTRIM. If the normalized value of a particular constraint is below the negative value of the trimming factor, then this constraint is temporarily deleted from the critical list. It is
In de x

COSMOSM Advanced Modules

4-11

Chapter 4 Numerical Aspects

noted that those constraints which are temporarily deleted may become active during the subsequent optimization loop, thus may be retained in the critical list later. Constraints are normalized by using their lower and upper bounds specified by the command OP_CONDEF. The original constraint is bounded by:

A pair of normalized constraints with respect to upper and lower bounds are bounded by:

This normalization provides a clear indication for the trimming. For example, a normalized constraint with a value +0.4 has violated its bound by 40%; another constraint with a value -0.3 is within its bound 30%. If the trimming factor is 0.2, then the first constraint will be retained in the critical list, however the second one will be deleted.

Convergence Criteria
Convergence or termination checks are performed at the end of each optimization loop. The optimization process continues until either convergence or termination occurs. The process may be terminated before convergence in two cases:

The number of design sets so far exceeds the maximum number of optimization
loops specified in the command A_OPTIMIZE,

If the initial design is infeasible and the allowed number of consecutive

In de x

infeasible designs has been exceeded.

4-12

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

The optimization problem is considered converged if all of the following conditions are satisfied:

The current design is feasible, Changes in the objective function F:

- The difference between the current value and the best design so far is less than the tolerance F specified in the command OP_OBJDEF,

- The difference between the current value and the previous design is less than the tolerance,

- (Optional) The differences between the current value and two previous designs are less than the tolerance,

and

Changes in the design variables Xi:

- The difference between the current value of each design variable and the best design so far is less than the respective tolerance i specified in the command DVARDEF,

- The difference between the current value of each design variable and the previous design is less than the respective tolerance,

- (Optional) The differences between the current value of each design variable and two previous designs are less than the respective tolerance,

and
In de x

COSMOSM Advanced Modules

4-13

Chapter 4 Numerical Aspects

Figure 4-10 shows the convergence plots.

Figure 4-10. Convergence Plots

Objective Function (F) Fbe st Design 10 Set

Infeasible Constraint Value

F easible

Infeasible 1 2 3 4 5 6 7 8 9

Design 10 Set

References
1.

R. T. Haftka and Z. Grdal Elements of Structural Optimization, Third Edition, Kluwer Academic Publishers, 1992. W. H. Press, B. P. Flannery, S. A. Teukolsky and W. T. Vetterling, Numerical Recipes, Cambridge University Press, 1986. G. N. Vanderplaats, Numerical Optimization techniques for Engineering Design with Applications, McGraw Hill, Inc., 1984.

2.

3.

In de x

4-14

COSMOSM Advanced Modules

Additional Problems

Introduction
This chapter presents additional examples for performing shape and sizing optimization and sensitivity in COSMOSM. The input files for these problems are available in the ...\Vprobs\Optimization folder. Where ... denotes the COSMOSM directory. The prefix alphabets for the shape optimization problems are OPS, whereas for sizing, they are OPZ. The prefix alphabets for the shape sensitivity problems are SNS, whereas for sizing, they are SNZ. Static stress analysis is denoted by letters ST, natural frequency analysis by FQ, buckling analysis by BK, heat transfer analysis by HT, nonlinear by the letter N, post dynamic analysis by D, and fatigue analysis by FT. Multidisciplinary analysis examples use a combination of the first alphabets from the above. Note that post dynamic and fatigue analyses are categorized to multidisciplinary since they cannot run alone. For example, the problem OPSST5 discusses shape optimization analysis of a beam under linear static stress analysis; OPZBK2 discusses minimum weight of a cantilever subject to buckling constraints; OPSTSB1 addresses thermal, static and buckling analysis of a fixed channel for the optimal shape. The following tables provide a listing of shape and sizing optimization examples and sensitivity with respect to the type of analysis. It is noted that the results may vary in a certain range from one platform to another.

In de x

COSMOSM Advanced Modules

5-1

Chapter 5 Additional Problems

Table 3-1. Dominantly Shape Optimization Problems OPSST1 OPSST2 Static OPSST3 OPSST4 OPSST5 OPSST6 Natural Frequency Linearized Buckling OPSFQ1 OPSBK1 OPSHT1 Heat Transfer OPSHT2 OPSHT3 OPSHT4 Nonlinear Structural OPSN1 OPSN2 OPSFS1 OPSBS1 OPSTS1 OPSTSF1 Multidisciplinary OPSTSB1 OPSTN1 OPSFDS1 OPSFDS2 OPSFDS3 OPSTNFT1 OPSFDSFT1 Stress analysis of a cantilever bracket (see page 5-6). Stress analysis of a Steering Control Arm (see page 5-8). Stress analysis of a slotted control arm (see page 5-9). Stress analysis of a simply supported beam - line curves (see page 5-11). Stress analysis of a simply supported beam - Bezier curves (see page 5-12). Stress analysis for an engine bearing cap under multiple load cases (see page 5-13). Frequency analysis of a control arm bracket (see page 5-14). Buckling analysis of a C-shape column (see page 5-15). Thermal analysis of a circular disk (see page 5-17). Thermal analysis of a pipe cooling system (see page 5-19). Thermal analysis of a simplified mechanical part (see page 5-21). Transient temperature - dependent heat conduction of a slab (see page 5-22). Nonlinear analysis of a thick-walled pipe (see page 5-23). Nonlinear analysis of a rubber circular ring (see page 5-24). Stress and frequency analysis of a bracket (see page 5-26). Stress and buckling analysis of a C-shape column (see page 5-28). Stress and thermal analysis of a mechanical part (see page 5-30). Thermal, static and frequency analysis of a circular disk (see page 5-31). Thermal, Static and buckling analysis of a fixed channel (see page 5-33). Transient thermal and nonlinear analysis of a cylinder (see page 5-35). Harmonic response analysis of a culvert (see page 5-37). Random vibration analysis of a lever arm (see page 5-39). Response spectrum analysis of a trophy setting on a table (see page 5-42). Fatigue analysis of a nozzle under a cyclic temperature loading (see page 5-45). Fatigue analysis of a curved pipe under a cyclic pressure loading (see page 5-48).

Table 3-2. Sizing Optimization Problems OPZST1 OPZST2 Static OPZST3 OPZST4 Minimum volume of a 1-bar truss subject to stress constraint (see page 5-52). Minimum volume of a 1-bar truss subject to displacement constraint (see page 5-53). Minimum volume of a 3-bar statically determinate truss subject to stress constraints (see page 5-54). Minimum weight of a 3-bar statically indeterminate truss - multiple load cases (see page 5-55).

In de x

5-2

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Table 3-2. Sizing Optimization Problems (Continued) OPZST5 OPZST6 OPZST7 OPZST8 OPZST9 OPZST10 OPZST11 OPZST12 Static OPZST13 OPZST14 OPZST15 OPZST16 OPZST17 OPZST18 OPZST19 OPZST20 OPZST21 OPZFQ1 Natural Frequency OPZFQ2 OPZFQ3 Minimum weight of a 4-bar statically indeterminate truss subject to stress constraints (see page 5-56). Minimum displacement of a cantilever subject to weight constraint - beam elements (see page 5-57). Minimum weight of a cantilever subject to stress constraint - beam elements (see page 5-58). Minimum weight of a cantilever subject to stress constraint - beam elements (see page 5-59). Minimum weight of a cantilever subject to displacement constraint - beam elements (see page 5-60). Minimum stress of a simply supported rectangular plate subject to weight constraint shell elements (see page 5-61). Minimum stress of a simply supported rectangular plate subject to stress constraint shell elements (see page 5-62). Minimum stress of a simply supported rectangular plate subject to displ.t constraint shell elements (see page 5-63). Minimum weight of a cantilever plate subject to stress constraint - quad. continuum elements (see page 5-64). Minimum volume of a plate subject to displacement constraint - triangular continuum elements (see page 5-65). Minimum volume of a plate subject to stress constraint - triangular continuum elements (see page 5-66). Minimum volume of a plate subject to stress constraint - quadrilateral continuum elements (see page 5-67). Minimum volume of a plate subject to stress constraint - quadrilateral continuum elements (see page 5-68). Minimum weight of a simply supported rectangular plate subject to effective strain constraint - shell elements (see page 5-69). Minimum weight of a simply supported rectangular plate subject to strain energy density constraint - shell elements (see page 5-70). Minimum volume of a cantilever pipe subject to a stress constraint - pipe radius (see page 5-71). Minimum volume of a cantilever pipe subject to a stress constraint - pipe thickness (see page 5-71). Minimum weight of a cantilever subject to frequency constraint - beam elements (see page 5-72). Minimum weight of a cantilever subject to frequency constraint - shell elements (see page 5-73). Minimum weight of a pipe cantilever subject to frequency constraint - pipe elements (see page 5-74).

In de x

COSMOSM Advanced Modules

5-3

Chapter 5 Additional Problems

Table 3-2. Sizing Optimization Problems (continued) OPZBK1 OPZBK2 Linearized Buckling OPZBK3 OPZBK4 OPZBK5 Multidisciplinary OPZNB1 OPZFDS1 Minimum weight of a cantilever subject to buckling load factor constraint - beam elements (see page 5-76). Minimum weight of a cantilever subject to buckling constraint - shell elements (see page 5-77). Minimum weight of a cantilever plate subject to buckling constraint - quadrilateral continuum elements (see page 5-78). Minimum weight of a pipe cantilever subject to buckling load factor - pipe elements (see page 5-79). Maximum buckling load design of a graphite-epoxy laminate (see page 5-80). Snap buckling of a thin hinged cylindrical shell under a central point load (see page 5-82). Modal time history analysis of a simply supported shell structure (see page 5-83).

Table 3-3. Dominantly Shape Sensitivity Problems SNSST1 Static Natural Frequency Linearized Buckling Heat Transfer Nonlinear SNSST2 SNSFQ1 SNSBK1 SNSHT1 SNSN1 SNSTSF1 SNSTSB1 Multidisciplinary SNSTN1 SNSFDS1 SNSFDS2 SNSFDS3 Sensitivity study of a steering control arm in stress analysis (see page 5-86). Sensitivity study of an engine bearing cap in stress analysis under multiple load cases (see page 5-87). Sensitivity study of a control arm bracket in frequency analysis (see page 5-88). Sensitivity study of a C-shape column in buckling analysis (see page 5-89). Sensitivity study of a circular disk in heat transfer analysis (see page 5-91). Sensitivity study of a thick-walled pipe in nonlinear analysis (see page 5-93). Sensitivity study of a circular disk in thermal, stress and frequency analysis (see page 5-94). Sensitivity study of C-shape column in thermal, stress and buckling analysis (see page 5-96). Sensitivity study of a cylinder in transient thermal and nonlinear analyses (see page 5-98). Sensitivity study of a culvert in harmonic response analysis (see page 5-98). Sensitivity study of a lever arm in random vibration analysis (see page 5-102). Sensitivity study of a trophy in response spectrum analysis (see page 5-105).

Table 3-4. Sizing Sensitivity Problems SNZST1 SNZST2 Static SNZST3 SNZST4 SNZST5 Nonlinear SNZN1 MultiSNZNF1A disciplinary SNZNF1B Global sensitivity of a 2-bar truss: (incremented simultaneously) (see page 5-108). Global sensitivity of a 2-bar truss: design variables incremented one at a time (see page 5-109). Offset sensitivity of a 2-bar truss (see page 5-110). Local sensitivity of a 5-bar truss (see page 5-111). Sensitivity study of a graphite-epoxy laminate (see page 5-112). Sensitivity of a rubber ring to coefficient of friction in nonlinear analysis (see page 5-114). Sensitivity study of a cantilever beam in linear, nonlinear and frequency analyses (see page 5-116).

In de x

5-4

COSMOSM Advanced Modules

Dominantly Shape Optimization Problems

In de x

COSMOSM Advanced Modules

5-5

Chapter 5 Additional Problems

OPSST1: Stress Analysis of a Cantilever Bracket

KEYWORDS: Shape, static analysis, 8-node PLANE2D, minimum volume, and stress constraint.
Pressure

(see page 5-2).

12.5

r=5 6 100.0 6

dv1

6 dv2

r=5
X

12.5

8 - Node P LANE 2 D E le me nt s ( S iz e = 4 ) Thickness = 10 mm 2 Pressure = 5 N/mm E = 200,000 N/mm 2 = 0.3 Note: All dimensions in milimeters. 175.0 I nit ia l P roble m G e ome t ry

12.5 25 dv3

I nit ia l Finit e E le me nt Me s h

In de x

5-6

COSMOSM Advanced Modules

25

Part 2 OPSTAR / Optimization

PROBLEM: Find size and location of the cutout. The bracket thickness is 10 mm, modulus of elasticity is 200,000 N/mm2 and Poisson's ratio is 0.3. A pressure of 5 N/mm2 is applied to the top edge of the bracket. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters DV1 DV2 DV3 Volume von Mises stress Initial Value(s) and Bounds 10 25 25 10 25 25 20 50 50 106,794 0 94 300 Final Value(s) 19.814 10.0 20.0 67,600 311.08 Tolerance 2.0 2.0 2.0 0.05 (Ratio) 15.0

Design Variables Objective Function Constraints

In de x

COSMOSM Advanced Modules

5-7

Chapter 5 Additional Problems

OPSST2: Stress Analysis of a Steering Control Arm

KEYWORDS: Shape, static analysis, TETRA4, minimum volume, and stress constraint.
Y TE TRA4 E LE ME NTS E Py = 2 x 105 N/mm 2 = 0.3 = 31 N/mm 2 20 30 X Internal surfaces fixed in all directions 20

(see page 5-2).

Note: All dimensions in milimeters.

7 tw tr 2 dint

tr1

140

Py

I nit ia l P roble m G e ome t ry I nit ia l Finit e E le me nt Me s h

PROBLEM: Find thickness of the two shafts and size and location of the cutout. The arm outer thickness is 20 mm, modulus of elasticity is 200,000 N/mm2, and Poisson's ratio is 0.3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters TR1 TR2 TW DINT Volume von Mises stress Initial Value(s) and Bounds 8 24 25 8 19 20 488 4 6 17 218,870 0 105.35 225 Final Value(s) 8.0 8.0 4.6981 17.00 76,046 218.11 Tolerance 1.0 1.0 1.0 1.0 0.01 (Ratio) 3.0

Design Variables

Objective Function Constraints

In de x

5-8

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSST3: Stress Analysis of a Slotted Control Arm

KEYWORDS: Shape, static analysis, 6-node TRIANG, minimum volume, and stress constraint.
Y

(see page 5-2).

Y Py
tr2

20 Fixed tr1 tw r1 = 30 5 5 5 10

r2 = 7

10 20 = 140 I nit ia l P roble m G e ome t ry

6 - Node Tria ngula r Me s h Thickness = 20 mm 5 2 E = 2 x 10 N/mm = 0.3 = 4 N/mm2 Py (Y direction) Note: All dimensions in milimeters.

I nit ia l Finit e E le me nt Me s h

In de x

COSMOSM Advanced Modules

5-9

Chapter 5 Additional Problems

PROBLEM: Find thickness of the two shafts and size and location of the cutout. The arm outer thickness is 20 mm, modulus of elasticity is 200,000 N/mm2, and Poisson's ratio is 0.3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters TR1 TR2 TW Volume von Mises stress Initial Value(s) and Bounds 8 25 25 8 20 20 3.5 8 8 200,347.9 10 90.64 225 Final Value(s) 8.0 8.0 5.8192 72,047.13 227.70 Tolerance 1.0 1.0 0.5 0.00125 (Ratio) 3.0

Design Variables Objective Function Constraints

In de x

5-10

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSST4: Stress Analysis of a Simply Supported Beam Line Curves

KEYWORDS: Shape, static analysis, 8-node SOLID, minimum volume, and stress constraint.
4" 100 lbs/Node

(see page 5-2).

D1

D2 10" I nit ia l P roble m G e ome t ry Cros s S e c t ion

45"

45"

I nit ia l Finit e E le me nt Me s h

PROBLEM: Find height of the beam at support and middle part. The beam width is 4 in, modulus of elasticity is 1E7 psi, and Poisson's ratio is 0.3. Nodal forces of 100 lb are applied to the middle part of the beam. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints D1 D2 Volume von Mises 1 von Mises 2 Initial Value(s) and Bounds 0.2 6 10 0.2 6 10 2400 1 1575.6 5000 1 1648.2 5000 Final Value(s) 1.4442 3.4309 1014.8 4949.7 5033.3 Tolerance 0.098 0.098 0.03 (Ratio) 49.99 49.99

In de x

COSMOSM Advanced Modules

5-11

Chapter 5 Additional Problems

OPSST5: Stress Analysis of a Simply Supported Beam Bezier Curves

KEYWORDS: Shape, static analysis, 8-node SOLID, minimum volume, and stress constraint.
1,000 lbs/node 4"

(see page 5-2).

D1

D2

D3

D4 Cros s S e c t ion

20"

20" init ia l P roble m G e ome t ry

20"

init ia l Finit e E le me nt Me s h

PROBLEM: Find heights along the beam span. The beam width is 4 in, modulus of elasticity is 1E7 psi, and Poisson's ratio is 0.3. Nodal forces of 1000 lb are applied at the midsection of the beam. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters D1 D2 D3 D4 Volume von Mises 1 von Mises 2 von Mises 3 von Mises 4 von Mises 5 Initial Value(s) and Bounds 0.1 6 10 0.1 6 10 0.1 6 10 0.1 6 10 1,440.0 1 7401.2 15000 1 5255.1 15000 1 2759.7 15000 1 3409.7 15000 1 7378.1 15000 Final Value(s) 4.2882 4.1135 1.8693 1.8858 729.42 15,124 14,035 15,052 2681.5 15,110 Tolerance 0.099 0.099 0.099 0.099 0.03125 (Ratio) 149.99 149.99 149.99 149.99 149.99

Design Variables

Objective Function

Constraints

In de x

5-12

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSST6: Stress Analysis of an Engine Bearing Cap Under Multiple Load Cases
KEYWORDS: Shape, static analysis, TETRA10, multiple load cases, minimum volume, and stress constraint.
TWEB
1 0 -Node Te tra Me sh

(see page 5-2).

HUMP Symmetry Boundary Conditions Fixed

= 2 x 10 5 N/mm 2 = 0.30

Height

Initial P roble m Ge ome try

ECCENT

Initial Finite E le me nt Me sh

PROBLEM: Find size and geometric dimensions of the bearing cap. The bearing cap modulus of elasticity is 200,000 N/mm2 and Poisson's ratio is 0.30. A horizontal pressure (as load case 1) and a vertical pressure (as load case 2) are applied to the internal surfaces of the cylindrical hole. The initial values and bounds of design variables, constraints and objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters ECCENT TWEB HEIGHT HUMP Volume von Mises stress Initial Value(s) and Bounds 50 75 75 6 10 10 40 70 70 1 15 15 69,844.78 STAR: 0 117.4 FFE: 113.6 250 Final Value(s) STAR 50 6.6044 40 6.7424 21,767 259.21 FFE 50 6 40 9.7574 21,450 261.02 Tolerance 0.25 0.04 0.30 0.14 0.05 (Ratio) 12.50

Design Variables

Objective Function Constraints

In de x

COSMOSM Advanced Modules

5-13

Chapter 5 Additional Problems

OPSFQ1: Frequency Analysis of a Control Arm Bracket

KEYWORDS: Shape, frequency analysis, SHELL3, minimum volume, and frequency constraint.
r = 1.0 r = 2.0 3 - Node Tria ngula r Me s h ( S iz e = 1 ) Thickness E = = = = 0.3 cm 2 x 107 N/cm2 0.30 0.0075 Kg/cm3

(see page 5-2).

1.5

1.5

Note: All dimensions in centimeters.

15.0

t1 0.5 1.0 t2 X

t1 1.0

Fixed

Fixed

5.0

5.0 I nit ia l Finit e E le me nt Me s h

I nit ia l P roble m G e ome t ry

PROBLEM: Find size and location of the cutout. The bracket thickness is 0.3 cm, modulus of elasticity is 2E7 N/cm2, and Poisson's ratio is 0.3. The material mass density is 0.0075 Kg/cm3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T1 T2 Volume Fundamental frequency Initial Value(s) and Bounds 0.5 2.5 2.5 1.5 3.5 3.5 28.473 8 13.65 14 Final Value(s) 0.6256 1.5 16.491 7.9992 Tolerance 0.02 0.02 0.008 (Ratio) 0.06

In de x

5-14

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSBK1: Buckling Analysis of a C-shape Column

KEYWORDS: Shape, buckling analysis, SHELL3, minimum volume, and buckling load factor constraint.
Fixed T1 Fillet (T2 /4) Element Size Thickness Material E T1 S HE LL3 = = = = 3 in 0. 25 in A_Steel 3 x 10 7 psi

(see page 5-2).

B = 20" T2

= 0.28 -3 2 4 = 0.73 x 10 lbf sec /in

0.18h

T2

T1

A = 6"

0.32h

T2

C = 3"

h = 120"

0.32h

I nit ia l P roble m G e ome t ry I nit ia l Finit e E le me nt Me s h

0.18h

p= 5,000 psi

In de x

COSMOSM Advanced Modules

5-15

Chapter 5 Additional Problems

PROBLEM: Find size of the column cutouts. The cross section thickness is 0.25 in., modulus of elasticity is 3E7 psi, and Poisson's ratio is 0.28. A pressure of 5000 psi is applied to the column's face. The initial values and bounds of design variables, constraints and the objective function are shown below. Summary of Results:
Optimization Parameters Design Variables Objective Function Constraints T1 T2 Volume Buckling load factor Initial Value(s) and Bounds 1 5 35 1 5 12 1018.15 1.6 2.87 10 Final Value(s) 35.0 11.261 745.91 1.5843 Tolerance 0.34 0.11 0.01 (Ratio) 0.02

In de x

5-16

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSHT1: Thermal Analysis of a Circular Disk

KEYWORDS: Shape, thermal analysis, SHELL3, minimum volume, and temperature constraint.
Heat Flux 0.1 W/mm 2
5

(see page 5-2).

10 5 5 10 Convection 0.0005 W/mm2 -C Ambient Temperature 50C

Radius

5 5

25 10

Initial P roble m Ge ome try

25 Heat Flux 0.1 W/mm 2 Convection 0.0 W/mm 2 -C Ambient Temperature 50C

3 - Node S he ll Me s h ( S iz e = 2 ) Thickness = 2 mm

Initial Finite E le me nt Me sh

In de x

COSMOSM Advanced Modules

5-17

Chapter 5 Additional Problems

PROBLEM: Find radius of the disk. The disk thickness is 2 mm, and conductivity is 0.57 W/mm-C. A convection of 0.0005 W/mm-C with an ambient temperature of 50 C is applied to the entire model except for the heat sources. The heat source regions (heat flux of 0.1 W/mm2) is assumed to have a convection of 0 W/mm2-C with an ambient temperature of 50C. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints Radius Volume Temperature Initial Value(s) and Bounds 30 70 70 33,375.56 50 77.63 140 Final Value(s) 30.849 8566.86 140.15 Tolerance 1.0 0.0015 (Ratio) 1.0

In de x

5-18

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSHT2: Thermal Analysis of a Pipe Cooling System

KEYWORDS: Shape, thermal analysis, 8-node SOLID, SHELL4, minimum volume, and temperature constraint.
R
2

(see page 5-2).

RADD

R1 = 5

T1

H1 = 5 H2 =7

T3 T3

T2
H3 =7

T3

I nit ia l P roble m G e ome t ry R4 (thickness of pipes = 0.2 mm)

T2
H3 =7

T3

T2
H3 =7

T3

T2
H2 =7 H1 = 5

T3

Cros s S e c t ion

I nit ia l Finit e E le me nt Me s h

In de x

COSMOSM Advanced Modules

5-19

Chapter 5 Additional Problems

PROBLEM: Find radii and thicknesses of plates and pipes of the cooling system. The material conductivity is 0.57 W/mm-C. A convection of 0.0015 W/mm2-C with an ambient temperature of 50C is applied to the entire model except for the heat source. The heat source region (heat flux of 0.1 W/mm2) is assumed to have a convection of 0 W/mm2-C with an ambient temperature of 50C The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters R2 RADD R4 T1 T2 T3 Initial Value(s) and Bounds 13 25 25 8 50 50 133 122 122 122 HSTAR 196.436 50 77.33 100 50 76.36 100 50 50.55 100 FFE 196.436 50 77.34 100 50 76.36 100 50 50.56 70 Final Value(s) HSTAR 13.0 8.0 1.0 1.3246 1.0 1.0 HSTAR 37.0046 100.18 98.995 68.317 FFE 13.0 8.0 1.0 1.3671 1.0 1.0 FFE 37.588 99.754 98.602 68.264 Tolerance 1.0 1.0 0.1 0.1 0.1 0.1 Tolerance 0.001 (Ratio) 1.0 1.0 1.0

Design Variables

Optimization Parameters Objective Function Constraints Weight Temp 1 Temp 2 Temp 3

In de x

5-20

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSHT3: Thermal Analysis of a Simplified Mechanical Part

KEYWORDS: Shape, thermal analysis, TETRA10, minimum volume, and temperature constraint.
T

(see page 5-2).

TW

P la n

H TW I nit ia l P roble m G e ome t ry I nit ia l Finit e E le me nt Mode l

PROBLEM: Find dimensions and thicknesses of the mechanical part. The material conductivity is 0.57 W/mm-C. A convection of 0.0003 W/mm2-C with an ambient temperature of 50 C and a volume heat of 0.02 W/mm3 are applied for the entire model. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables TW H T Initial Value(s) and Bounds 266 55 100 100 3 15 15 HSTAR FFE 69405.3 Final Value(s) HSTAR 2.434 55.0 3.0 HSTAR FFE 2.4367 55.0 3.0 FFE Tolerance 0.04 0.45 0.12 Tolerance 0.001 (Ratio) 1.0

Optimization Parameters Objective Function Constraints

Volume 69405.3 Temp

17030.75 17046.7 123.44

125 222.5 250 125 222.27 250 123.57

In de x

COSMOSM Advanced Modules

5-21

Chapter 5 Additional Problems

OPSHT4: Transient Temperature Dependent Heat Conduction of a Slab

KEYWORDS: Shape, transient thermal analysis, temperature-dependent conductivity, PLANE2D, minimum volume, and temperature constraint.
1.0" H Initial P roble m Ge ome try

(see page 5-2).

Initial Finite E le me nt Me sh

PROBLEM: Find the thickness of an infinitely large slab. The thermal conductivity K is assumed to vary linearly with temperature, T, (K = 2 + 0.01 T BTU/in-s F). The specific heat is constant (C = 8 BTU in/lb-s2 F). The temperature of the left side is suddenly raised to 200 F and returns to the initial temperature of 100 F after 10 seconds. The initial values and bounds of design variables, constraints and objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints H Volume Temp Initial Value(s) and Bounds 5 20 20 20 0 104.71 150 Final Value(s) 5.6253 5.62531 151.47 Tolerance 0.15 0.01 (Ratio) 1.50

In de x

5-22

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSN1: Nonlinear Analysis of a Thick-walled Pipe

KEYWORDS: Shape, nonlinear analysis, von Mises plasticity, automatic time stepping, PLANE2D, minimum volume, and stress constraint.

(see page 5-2).

Rint Rout Initial P roble m Ge ome try

Initial Finite E le me nt Me sh

PROBLEM: Find the pipe's outer diameter. The modulus of elasticity is 86,666 psi, Poisson's ration is 0.3, yield stress is 17.32 psi and tangential modulus is 866 psi. The initial values and bounds of design variables, constraints and objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints ROUT Volume von Mises stress Initial Value(s) and Bounds 1.5 2 2.5 2.34716 0 17.396 23 Final Value(s) 1.5427 1.0795 23.007 Tolerance 0.01 0.01 (Ratio) 0.23

In de x

COSMOSM Advanced Modules

5-23

Chapter 5 Additional Problems

OPSN2: Nonlinear Analysis of a Rubber Circular Ring

KEYWORDS: Shape, nonlinear analysis, rubber, Mooney model, contact, prescribed displacement, automatic time stepping, axisymmetric PLANE2D, minimum volume, and stress constraint.
Top Steel Plate Rubber Ring Bottom Steel Plate 0.5615" 0.5615" 0.278" 0.3475" 0.3" 0.3"

(see page 5-2).

Initial P roble m Ge ome try

2 R Cross Initial Finite E le me nt Me sh

In de x

5-24

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

PROBLEM: Find the cross-section radius of a circular rubber ring squeezed between two parallel steel plates. For rubber, the Mooney's constants are 175 and 10 psi and Poisson's ratio is 0.49. For steal plates, the Young's modulus is 30 x 106 psi, and Poisson's ratio is 0.30. The coefficient of friction is 0.01. The initial values and bounds of design variables, constraints and objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints RCROSS Volume (rubber) von Mises stress Initial Value(s) and Bounds 0.05 0.139 0.15 0.03 0 209.218 450 Final Value(s) 0.069462 0.00749278 453.37 Tolerance 0.001 0.01 (Ratio) 4.5

In de x

COSMOSM Advanced Modules

5-25

Chapter 5 Additional Problems

OPSFS1: Stress and Frequency Analysis of a Bracket

KEYWORDS: Shape, static analysis, frequency analysis, multidisciplinary optimization, 8-node PLANE2D, minimum volume, stress constraint, and frequency constraint.
Pressure

(see page 5-2).

12.5

r=5 6

dv1

6 dv2

100.0

r=5 X 12.5 25 dv3 175.0

8 - Node P LANE 2 D Me s h ( S iz e = 4 ) Thickness Pressure E Density = 10 mm = 5 N/mm 2 = 200,000 N/mm 2 = 0.3 = 0.00785 gm/mm 3

12.5

Note: All dimensions in milimeters.

I nit ia l P roble m G e ome t ry

I nit ia l Finit e E le me nt Me s h

In de x

5-26

COSMOSM Advanced Modules

25

Part 2 OPSTAR / Optimization

PROBLEM: Find size and location of the cutout. The bracket thickness is 10 mm, modulus of elasticity is 200,000 N/mm2 and Poisson's ratio is 0.3. A pressure of 5 N/mm2 is applied to the top edge of the bracket. The material density is 0.00785 gm/mm3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters DV1 DV2 DV3 Volume von Mises stress Fundamental frequency Initial Value(s) and Bounds 10 25 25 10 25 25 20 50 50 106,794 0 94.12 300 1.8 2.8613 3 Final Value(s) 25.0 10.0 22.0 76,063.64 241.52 1.7978 Tolerance 2.0 2.0 2.0 0.05 (Ratio) 15.0 0.012

Design Variables Objective Function Constraints

In de x

COSMOSM Advanced Modules

5-27

Chapter 5 Additional Problems

OPSBS1: Stress and Buckling Analysis of a C-shape Column

KEYWORDS: Shape, stress analysis, buckling analysis, multidisciplinary optimization, SHELL3, minimum volume, buckling load factor constraint, and stress constraint.
Fixed T1 Fillet (T2 /4) Element Size Thickness Material E T1 S HE LL3 = = = = 3 in 0. 25 in A_Steel 3 x 107 psi

(see page 5-2).

B = 20" T2

= 0.28 -3 2 4 = 0.73 x 10 lbf sec /in

0.18h

T2

T1

A = 6"

0.32h

T2

C = 3"

h = 20"

0.32h

I nit ia l P roble m G e ome t ry I nit ia l Finit e E le me nt Me s h

0.18h

p= 5,000 psi

In de x

5-28

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

PROBLEM: Find size of the column cutouts. The cross section thickness is 0.25 in., modulus of elasticity is 3E7 psi, and Poisson's ratio is 0.28. A pressure of 5000 psi is applied to the column's face. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T1 T2 Volume von Mises stress Buckling load factor Initial Value(s) and Bounds 1 5 35 1 5 12 1018.15 0 8,729 22,000 1.8 2.87 10 Final Value(s) 35.0 9.8635 780.98 19,125 1.8012 Tolerance 0.34 0.11 0.01 (Ratio) 220 0.02

In de x

COSMOSM Advanced Modules

5-29

Chapter 5 Additional Problems

OPSTS1: Stress and Thermal Analysis of a Mechanical Part

KEYWORDS: Shape, static analysis, thermal analysis, multidisciplinary optimization, TETRA4R, minimum volume, temperature constraint, and stress constraint.
T

(see page 5-2).

TW

P la n

H TW

Initial P roble m Ge ome try

Initial Finite E le me nt Me sh

PROBLEM: Find dimensions and thicknesses of the mechanical part. The material conductivity is 0.57 W/mm-C. A convection of 0.0003 W/mm2-C with an ambient temperature of 50 C and a volume heat of 0.02 W/mm3-C are applied for the entire model. The modulus of elasticity is 1E07 N/mm2 and Poisson's ratio is 0.30. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters TW H T Volume von Mises stress Temperature Initial Value(s) and Bounds 266 55 100 100 3 15 15 69,599.5 0 8752.45 9000 125 222.658 250 Final Value(s) 2.4053 55.0 3.0 17,017.8 6133.8 123.18 Tolerance 0.04 0.45 0.12 0.001 (Ratio) 20.0 2.0

Design Variables Objective Function Constraints

In de x

5-30

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSTSF1: Thermal, Static and Frequency Analysis of a Circular Disk

KEYWORDS: Shape, thermal analysis, static analysis, frequency analysis, multidisciplinary optimization, SHELL3, TEMPREAD command, minimum volume, temperature constraint, displacement constraint, and frequency constraint.
Heat Flux 0.1 W/mm 2
5

(see page 5-2).

10 5 5 10 Convection 0.0005 W/mm2 -C Ambient Temperature 50C

Radius

5 5

25 10

Initial P roble m Ge ome try

25 Heat Flux 0.1 W/mm 2 Convection 0.0 W/mm 2 -C Ambient Temperature 50C

3 - Node S he ll Me s h ( S iz e = 2 ) Thickness = 2 mm

Initial Finite E le me nt Me sh

In de x

COSMOSM Advanced Modules

5-31

Chapter 5 Additional Problems

PROBLEM: Find radius of the disk. The disk thickness is 2 mm, and conductivity is 0.57 W/mmC. A convection of 0.0005 W/mm2-C with an ambient temperature of 50 C is applied to the entire model except for the heat sources. The heat source regions (heat flux of 0.1 W/mm2) is assumed to have a convection of 0 W/mm2-C with an ambient temperature of 50 C. The modulus of elasticity is 200000 N/mm2, Poisson's ratio is 0.30, coefficient of thermal expansion is 0.13E-4/C and material density is 0.00785 gm/mm3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints Radius Volume Temperature Displ. frequency Displ. frequency Initial Value(s) and Bounds 30 70 70 33,251.96 50 76.46 140 0.09 0.166 0.17 1.0 1.139 8 Final Value(s) 38.859 11,955.40 107.39 0.08974 5.28985 Tolerance 1.0 0.0015 (Ratio) 1.0 0.0008 0.07

In de x

5-32

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSTSB1: Thermal, Static and Buckling Analysis of a Fixed Channel

KEYWORDS: Shape, thermal analysis, static analysis, buckling analysis, multidisciplinary optimization, SHELL3, TEMPREAD command, minimum volume, buckling load factor constraint, and displacement constraint.
Fixed T1 Fillet (T2 /4) Element Size Thickness Material E T1 S HE LL3 = = = = 3 in 0. 25 in A_Steel 3 x 107 psi

(see page 5-2).

B = 20" T2

= 0.28 -3 2 4 = 0.73 x 10 lbf sec /in

0.18h

T2

T1

A = 6"

0.32h

T2

C = 3"

Fixed h = 120" 0.32h

I nit ia l P roble m G e ome t ry I nit ia l Finit e E le me nt Me s h

0.18h

In de x

COSMOSM Advanced Modules

5-33

Chapter 5 Additional Problems

PROBLEM: Find size of the channel cutouts. The cross section thickness is 0.25 in, modulus of elasticity is 3E7 psi, and Poisson's ratio is 0.28. The material conductivity is 6.7E-4 BTU/in/s/F. A convection of 0.0001 BTU/sec in2-F with an ambient temperature of 50 F and a volume heat of 0.005 BTU/sec in3 are applied for the entire model. The material's coefficient of thermal expansion is 7.4E-6/F The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T1 T2 Volume Buckling factor Displacement Initial Value(s) and Bounds 1 5 35 1 5 12 1018.15 1.6 1.8432 10 0.0 0.01102 0.018 Final Value(s) 17.257 12.00 887.157 3.981 0.018179 Tolerance 0.34 0.11 0.01 (Ratio) 0.02 0.00018

In de x

5-34

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSTN1: Transient Thermal Nonlinear Analysis of a Cylinder

KEYWORDS: Shape, transient thermal, nonlinear analysis, radiation, convection, heat flux; element heat, prescribed temperature, multi-disciplinary optimization, von Mises plasticity, Axisymmetric PLANE2D, minimum volume, stress, strain and temperature constraints.
Convection _ Ambient Temperature = 200 F Firm Coefficient = 0.1 BTU / (sec in inF) Thick Temperature = 70 F

(see page 5-2).

Element Heat = 0.140625 BTU / (sec in in in)

Radiation Source = 500 F Emissivity = 0.9 View Factor = 0.8

Thick Heat Flux = 0.0125 BTU / (sec in in) Initia l Ge ome try, Loa ds a nd Bounda ry C onditions

ROUT Initia l Finite Ele me nt Me sh

PROBLEM: Find the radius and thickness of a cylinder subject to thermal loads and boundary conditions. Steel alloy and Aluminum materials are used. The initial values and bounds of design variables, constraints and objective function are shown below.

In de x

COSMOSM Advanced Modules

5-35

Chapter 5 Additional Problems

SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function ROUT THICK Volume von Mises 1 von Mises 2 Temp Effective Strain 1 Effective Strain 2 Initial Value(s) and Bounds 5 12 12 0.5 2 2 39.999 0 44,045 50,000 0 30,213 30,000 350. 413.63 450 0 0.003896 0.01 0 0.0021032 0.01 Final Value(s) 5 0.66144 9.4841 38954.81 30152.49 381.429 0.003445 0.001796 Tolerance 0.07 0.015 0.01 (Ratio) 500 300 1.0 1 x 10-4 1 x 10-4

Constraints

In de x

5-36

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSFDS1: Harmonic Response Analysis of a Culvert

KEYWORDS: Frequency analysis, post-dynamic analysis (harmonic response), dynamic stress analysis, 4-noded PLANE2D elements, shape optimization, multidisciplinary optimization, minimum volume design, frequency, displacement, and stress constraints.

(see page 5-2).

Initial Problem Geometry

Initial Finite Element Mesh

In de x

COSMOSM Advanced Modules

5-37

Chapter 5 Additional Problems

PROBLEM: Find the radius R and slope s of a culvert. The material constants of the culvert are given as: Youngs modulus E = 30E6 psi, Poissons ratio = 0.3, and Density = 1 lb*sec2/in4. A harmonic pressure loading with constant amplitude 500 psi within the desired range of frequency (1 rad/sec - 400 rad/sec) is applied to the top of the culvert. A modal damping 0.015 is assumed for the first 10 modes. The input data regarding optimization as well as the converged results are listed in the following table. Note that the displacement and stress constraints are defined as the extreme values within the desired range of frequency. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints DV1 DV2 Volume Frequency (1st) Displacement (Uy) Stress (von Mises) Initial Value(s) and Bounds 20 20 35 1.3333 2 2 2575.5 5 16.179 20 0 0.0429 0.05 0 4.6893E4 5E4 Final Value(s) 23.268 1.3333 1954.8 15.470 0.05021 4.7637E4 Tolerance 0.15 6.6667E-3 0.01 (ratio) 0.15 5.E-4 500.

In de x

5-38

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSFDS2: Random Vibration Analysis of a Lever Arm

KEYWORDS: Frequency analysis, post-dynamic analysis (random vibration), dynamic stress analysis, 6-noded TRIANG elements, shape optimization, multidisciplinary optimization, minimum volume design, frequency, displacement, and stress constraints.

(see page 5-2).

Initial Finite Element Mesh Initial Problem Geometry

In de x

COSMOSM Advanced Modules

5-39

Chapter 5 Additional Problems

Base Excitation versus Frequency Curve

Pressure Loading versus Frequency Curve

In de x

5-40

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

PROBLEM: Find the hyperbolic arc parameter RATIO and thickness T2 of a lever arm. The arm is made of A_STEEL and has a uniform thickness 1.0 in. Both harmonic pressure loading and base excitation (acceleration) in the y-direction are applied to the structure as shown in the figure. A modal damping 3% is assumed for the first 5 modes. The input data regarding optimization as well as the converged results are listed in the following table. Note that the displacement constraint is defined as the extreme value of PSD within the desired range of frequency and the stress constraint is the extreme value of RMS. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints DV1 DV2 Volume Frequency (1st) Displacement (Uy) Stress (von Mises) Initial Value(s) and Bounds 0.5 0.5 0.75 15 40 50 2.008E4 75 92.377 100 0 3.1874E-5 3.5E-5 0 1.1893E4 1.25E4 Final Value(s) 0.6325 29.782 1.6078 82.089 2.1262E-5 1.2624E4 Tolerance 2.5E-3 0.35 0.01 (ratio) 0.25 3.5.E-7 125.

In de x

COSMOSM Advanced Modules

5-41

Chapter 5 Additional Problems

OPSFDS3: Response Spectrum Analysis of a Trophy Setting on a Table

KEYWORDS: Frequency analysis, post-dynamic analysis (time history analysis, response spectrum generation, and response spectrum analysis), dynamic stress analysis, SHELL4 elements, shape optimization, multidisciplinary optimization, maximum volume design, frequency, displacement, and stress constraints.

(see page 5-2).

Initial Problem Geometry

Initial Finite Element Mesh

In de x

5-42

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Geometry of a Table

Impulsive Base Excitation (for Table)

In de x

COSMOSM Advanced Modules

5-43

Chapter 5 Additional Problems

PROBLEM: Find the radii (R2 and R3) and heights (H3 and H4) of a trophy which is setting on the top level of a table at the point P. Both structures are made of A_STEEP. The trophy has a 5% critical damping. This problem is solved in three steps:
1. Maximum Response Spectrum Generation (for Trophy) A time history analysis of the table is performed where an impulsive displacement base excitation in the xdirection is applied.

2. 3.

A maximum acceleration response spectrum is generated from the previous results. A design optimization of the trophy in the response spectrum analysis is performed where the spectrum generated previously is used to excite the trophy in the form of acceleration base excitation.

The input data regarding optimization as well as the converged results are listed in the following table. Note that the displacement and stress constraints are defined as the extreme values by using the SRSS mode combination method. SUMMARY OF RESULTS:
Optimization Parameters DV1 DV2 DV3 DV4 Volume Frequency (1st) Displacement (Ux) Stress (von Mises) Initial Value(s) and Bounds 458 8 10 20 8 10 15 15 20 25 377.49 10 27.043 40 0 0.3549 0.5 0 1.849E4 2.5E4 Final Value(s) 4.994 20 8 25 1229.75 10.935 0.5019 1.9645E4 Tolerance 0.04 0.12 0.07 0.1 0.01 (ratio) 0.3 0.005 250

Design Variables

Objective Function Constraints

In de x

5-44

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPSTNFT1: Fatigue Analysis of a Nozzle Under a Cyclic Temperature Loading

KEYWORDS: Transient thermal analysis, nonlinear elastoplastic analysis, fatigue analysis, 4noded PLANE2D elements, shape optimization, multidisciplinary optimization, minimum volume design, fatigue usage factor, displacement, and stress constraints.

(see page 5-2).

Initial Finite Element Mesh Initial Problem Geometry

In de x

COSMOSM Advanced Modules

5-45

Chapter 5 Additional Problems

Temperature Variation Cycle

Fatigue Design Curve (S-N Curve)

In de x

5-46

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

PROBLEM: Find the thickness parameters T1, T2, and T3 of a circular nozzle. The material constants of the nozzle are given as: Youngs modulus E = 3E7 psi, Poissons ratio =0.3, density = 3E-4 lb*hr2/in4, yield stress y = 1E4 psi, tangent modulus ET = 3E6 psi, coefficient of thermal expansion = 8E-6 1/F, thermal conductivity Kx = 0.1 BTU/(in*hr*F), and specific heat C = 40 BTU*in/(lb*hr2*F). The convection coefficient h and the adjacent ambient temperature Ta for the exterior and interior surfaces of the nozzle are 1 and 5 BTU/(in2*hr*F) and 60 and 1 F, respectively. Assuming that the nozzle is exposed to a fluid heat-up condition which is expected to occur 5000 times during its service life. One complete cycle of this heat-up condition is shown in the figure. The problem is solved in three steps:
1.

A transient thermal analysis is performed with the initial temperature T0 equal to 60 F. The nodal temperatures of the structure are evaluated at each time interval 0.01 hours for a total time 0.2 hours. Applying the nodal temperatures obtained from the transient thermal analysis to the structure, a nonlinear elastoplastic analysis is performed. The reference temperature Tref is 60 F. With the stresses obtained from the nonlinear analysis, a fatigue analysis is performed. A specified fatigue design curve (S-N curve) is shown in the figure.

2.

3.

The input data regarding optimization as well as the converged results are listed in the following table. Note that the constraints, fatigue usage factor and von Mises stress, are estimated at node N on the exterior surface of the nozzle and the resultant displacement is defined as the extreme value within the desired range of time (0 - 0.2 hours). SUMMARY OF RESULTS:
Optimization Parameters DV1 DV2 DV3 Volume Fatigue Usage Factor Displacement (Ures) Stress (von Mises) Initial Value(s) and Bounds 3<5<5 1<2<2 0.25 < 1 < 1 23.219 0 < 0.1030 < 0.15 0 < 0.01057 < 0.012 0 < 1.1523E4 < 1.5E4 Final Value(s) 3.0 1.1708 0.25 12.824 0.1491 0.01144 1.5116E4 Tolerance 0.02 0.01 7.5E-3 0.01 (ratio) 1.5E-3 1.2E-4 150

Design Variable Objective Function Constraints

In de x

COSMOSM Advanced Modules

5-47

Chapter 5 Additional Problems

OPSFDSFT1: Fatigue Analysis of a Curved Pipe Under a Cyclic Pressure Loading

KEYWORDS: Frequency analysis, post-dynamic analysis (modal time history), dynamic stress analysis, fatigue analysis, SHELL4 elements, shape/sizing optimization, multidisciplinary optimization, minimum volume design, frequency, fatigue usage factor, displacement, and stress constraints.

(see page 5-2).

Initial Problem Geometry

Initial Finite Element Mesh

In de x

5-48

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Pressure Variation Cycle

Fatigue Design Curve (S-N Curve)

In de x

COSMOSM Advanced Modules

5-49

Chapter 5 Additional Problems

PROBLEM: Find the dimension parameters CURL, RADIUS, and thickness THICK of a curved pipe. The pipe is made of A_STEEL. A varying internal pressure loading is applied to the pipe which is expected to occur 10000 and 2000 times during its service life. One complete cycle of this loading is shown in the figure. The problem is solved in two steps:
1.

A modal time history analysis of the pipe is performed with the prescribed pressure loading. A modal damping 0.05 is assumed for the first 5 modes. The responses during a total range of time 0.05 sec are recorded. With the stresses obtained from the dynamic stress analysis, a fatigue analysis is performed. A specified fatigue design curve (S-N curve) is shown in the figure.

2.

The input data regarding optimization as well as the converged results are listed in the following table. Note that the constraint, fatigue usage factor, is the extreme value among all the nodes and the resultant displacement and von Mises stress are defined as the extreme value within the desired range of time (0 - 0.05 sec). SUMMARY OF RESULTS:
Optimization Parameters DV1 DV2 DV3 Volume Frequency (1st) Fatigue Usage Factor Displacement (Ures) Stress (von Mises) Initial Value(s) and Bounds 1.5 2.5 2.5 0.25 1 1 0.05 0.1 0.1 14.930 50 146.28 150 0 0.07713 0.15 0 0.1122 0.3 0 2.7786E4 5E4 Final Value(s) 1.5 0.25 0.0739 2.5789 50.882 0.1085 0.3012 2.8889E4 Tolerance 0.01 7.5E-3 5.E-4 0.01 (ratio) 1 1.5E-3 3.E-3 500

Design Variable Objective Function

Constraints

In de x

5-50

COSMOSM Advanced Modules

Sizing Optimization Problems

In de x

COSMOSM Advanced Modules

5-51

Chapter 5 Additional Problems

OPZST1: Minimum Volume of a 1-bar Truss Subject to Stress Constraint

KEYWORDS: Sizing, static analysis, TRUSS2D, minimum volume, and stress constraint.
E = 5 psi F = 10 lbs Area = A 1

(see page 5-2).

1" P roble m G e ome t ry

Finit e E le me nt Me s h

PROBLEM: Find the minimum volume of a 1-bar truss subject to a concentrated force of 10 lb. The length of the bar is 1 in. and modulus of elasticity is 5 psi. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints A1 Volume Max stress Initial Value(s) and Bounds 0.25 1.0 1.0 1.0 0.5 10 20 Final Value(s) 0.50005 0.50005 19.998 Tolerance 0.0075 0.001 (Ratio) 0.00001

In de x

5-52

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZST2: Minimum Volume of a 1-bar Truss Subject to Displacement Constraint

KEYWORDS: Sizing, static analysis, TRUSS2D, minimum volume, and displacement constraint.
E = 5 psi F = 10 lbs Area = A 1 1" P roble m G e ome t ry

(see page 5-2).

Finit e E le me nt Me s h

PROBLEM: Find the minimum volume of a 1-truss element subject to a concentrated force of 10 lb. The length of the bar is 1 in. and modulus of elasticity is 5 psi. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints A1 Volume Max displacement Initial Value(s) and Bounds 0.25 1.0 1.0 1.0 124 Final Value(s) 0.49776 0.49776 4.018 Tolerance 0.0075 0.001 (Ratio) 0.03

In de x

COSMOSM Advanced Modules

5-53

Chapter 5 Additional Problems

OPZST3: Minimum Volume of a 3-bar Statically Determinate Truss Subject to Stress Constraints
KEYWORDS: Sizing, static analysis, TRUSS2D, minimum volume, and stress constraint. PROBLEM: Find the minimum volume of a 3bar truss subject to a concentrated force of 20,000 lb. The modulus of elasticity is 1E7 psi. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints A1 = A2 A3 Volume x 1, 2 x 3 Initial Value(s) and Bounds 0.50 2.0 4.0 0.50 1.0 4.0 76.5685 -15,000 7,071 20,000 -15,000 -10,000 20,000 Final Value(s) 0.70394 0.66342 33.179 20,090 -15,073 Tolerance 0.001 0.001 0.025 (Ratio) 150 150
A1 20"

(see page 5-2).

A3

A2

10"

20,000 lbs P roble m G e ome t ry

In de x

5-54

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZST4: Minimum Weight of a 3-bar Statically Indeterminate Truss Subject to Stress Constraints Multiple Load Cases
KEYWORDS: Sizing, static analysis, TRUSS2D, minimum weight, and stress constraint, and multiple load cases. PROBLEM: Find the minimum weight of a 3-bar truss subject to concentrated forces of 20,000 lb applied in two distinct load cases. The modulus of elasticity is 1E7 psi. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints A1 A2 Weight x 1 x 2 x 3 Initial Value(s) and Bounds 0.10 1.0 100 0.10 2.0 100 4.8284 -15000 12,612 20000 -15000 5,224 20000 -15000 12,612 20000 Final Value(s) 0.7147 0.6665 2.6881 20,025 12,068 20,025 Tolerance 0.01 0.01 0.001 (Ratio) 150 150 150
10" A1 A2 A3 = A1 10" 10"

(see page 5-2).

20,000 lbs (Load Case 2)

20,000 lbs (Load Case 1)

P roble m G e ome t ry

In de x

COSMOSM Advanced Modules

5-55

Chapter 5 Additional Problems

OPZST5: Minimum Weight of a 4-bar Statically Indeterminate Truss Subject To Stress Constraints
KEYWORDS: Sizing, static analysis, TRUSS3D, minimum weight, and stress constraint. PROBLEM: Find the minimum weight of a 4-bar truss subject to concentrated forces in X, Y, and Z directions. The modulus of elasticity is 1E4 ksi and the material density is 0.10 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters A1 A2 A3 A4 Weight x 1 x 2 x 3 x 4 Initial Value(s) and Bounds 0.001 1.0 2 0.001 2.0 2 0.001 2.0 2 0.001 1.0 2 101.879 -25 -12.165 25 -25 -21.360 25 -25 -16.530 25 -25 -11.89 25 Final Value(s) 0.4708 1.7078 1.2976 0.4873 65.2703 -25.18 -25.163 -25.205 -25.207 Tolerance 0.001 0.001 0.001 0.001 0.001 (Ratio) 0.25 0.25 0.25 0.25
60 K 20 K 10 K A3 A2 A4 A1 Y 120" Z P roble m G e ome t ry X 60" 144" 96"

(see page 5-3).

72"

Design Variables

Objective Function

Constraints

In de x

5-56

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZST6: Minimum Displacement of a Cantilever Subject to Weight Constraint Beam Elements

KEYWORDS: Sizing, static analysis, BEAM2D, minimum displacement, and weight constraint.
Y

(see page 5-3).

B1

100 5" 5"

50" Finit e E le me nt Me s h

50" B2 P roble m G e ome t ry

PROBLEM: A cantilever beam is subject to a concentrated load at the tip. Find the beam widths (in each half) maintaining a uniform height of the beam. The beam length is 100 in., height is 5 in., modulus of elasticity is 1E07 psi, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints B1 B2 Displacement UY Weight Initial Value(s) and Bounds 0.01 1.0 5 0.01 0.25 5 0.44 1 31.25 50 Final Value(s) 1.4614 0.55234 0.2640145 50.344 Tolerance 0.001 0.001 0.001 (Ratio) 0.5

In de x

COSMOSM Advanced Modules

5-57

Chapter 5 Additional Problems

OPZST7: Minimum Weight of a Cantilever Subject to Stress Constraint Beam Elements

KEYWORDS: Sizing, static analysis, BEAM2D, beam width, minimum weight, and stress constraint.
B1

(see page 5-3).

B2 5" 5"

50" Finit e E le me nt Me s h

50"

100

P roble m G e ome t ry

PROBLEM: A cantilever beam is subject to a concentrated load at the tip. Find the beam widths (in each half) maintaining a uniform height of the beam. The beam length is 100 in., height is 5 in., modulus of elasticity is 1E07 psi, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints B1 B2 Weight x 1 x 2 Initial Value(s) and Bounds 0.01 3 5 0.01 3 5 150 -10000 800 10000 -10000 400 10000 Final Value(s) 0.23833 0.11917 8.93745 10,070 10,070 Tolerance 0.001 0.001 0.01 (Ratio) 100 100

In de x

5-58

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZST8: Minimum Weight of a Cantilever Subject to Stress Constraint Beam Elements

KEYWORDS: Sizing, static analysis, BEAM2D, beam height, minimum weight, and stress constraint.
Y

(see page 5-3).

5"

1 2

5"

H1 50" Finit e E le me nt Me s h 50"

H2

100

P roble m G e ome t ry

PROBLEM: A cantilever beam is subject to a concentrated load at the tip. Find the beam heights (in each half) maintaining a uniform width of the beam. The beam length is 100 in., width is 5 in, modulus of elasticity is 1E07 psi, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints H1 H2 Weight x 1 x 2 Initial Value(s) and Bounds 0.01 2 10 0.01 2 10 100 -10000 3000 10000 -10000 1500 10000 Final Value(s) 1.0916 0.7719 46.588 10,070 10,070 Tolerance 0.01 0.01 0.001 (Ratio) 100 100

In de x

COSMOSM Advanced Modules

5-59

Chapter 5 Additional Problems

OPZST9: Minimum Weight of a Frame Subject to Displacement Constraint Beam Elements

KEYWORDS: Sizing, static analysis, BEAM3D, minimum weight, and displacement constraint.

(see page 5-3).

B H = 2B

100" 50"

Finit e E le me nt Me s h

P roble m G e ome t ry

PROBLEM: A 3D frame is subject to a concentrated load at the tip. Find the frame height and width given that width-to-height ratio is equal to the initial ratio. The modulus of elasticity is 1E07 psi, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints B H = 2B Weight UY Initial Value(s) and Bounds 0.1 6 10 1080 -0.10 -0.01536 0.10 Final Value(s) 3.7370 7.4740 418.96 -0.1018 Tolerance 0.01 0.001 (Ratio) 0.002

In de x

5-60

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZST10: Minimum Stress of a Simply Supported Rectangular Plate Subject to Weight Constraint Shell Elements
KEYWORDS: Sizing, static analysis, SHELL3, bending, minimum stress, and weight constraint. PROBLEM: Find the thickness of a 100 x 50 inch plate subject to pressure of 1 psi. The modulus of elasticity is 1E07 psi, Poisson's ratio is 0.3, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T (Thickness) von Mises stress Weight Initial Value(s) and Bounds 0.01 0.1 1 132,432 0.01 50 100 Final Value(s) 0.2014 32,649 100.7 Tolerance 0.001 0.001 (Ratio) 1.0
50" P = 1 psi T = Thickness

(see page 5-3).

100" P roble m G e ome t ry

Finit e E le me nt Me s h

In de x

COSMOSM Advanced Modules

5-61

Chapter 5 Additional Problems

OPZST11: Minimum Weight of a Simply Supported Rectangular Plate Subject to Stress Constraint Shell Elements
KEYWORDS: Sizing, static analysis, SHELL3, minimum weight, and stress constraint.
50"

(see page 5-3).

PROBLEM: Find the thickness of a 100 x 50 inch plate subject to pressure of 1 psi. The modulus of elasticity is 1E07 psi, Poisson's ratio is 0.3, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:

P = 1 psi T = Thickness

100" P roble m G e ome t ry

Finit e E le me nt Me s h

Optimization Parameters Design Variables Objective Function Constraints T (Thickness) Weight von Mises stress

Initial Value(s) and Bounds 0.01 2 2 1000 10 331.08 1700

Final Value(s) 0.87877 439.388 1714.9

Tolerance 0.02 0.001 (Ratio) 20.0

In de x

5-62

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZST12: Minimum Weight of a Simply Supported Rectangular Plate Subject to Displacement Constraint Shell Elements
KEYWORDS: Sizing, static analysis, SHELL3, bending, minimum weight, and displacement constraint. PROBLEM: Find the thickness of a 100 x 50 inch plate subject to pressure of 1 psi. The modulus of elasticity is 1E07 psi, Poisson's ratio is 0.3, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T (Thickness) Weight Displacement UZ Initial Value(s) and Bounds 0.01 2 5 1000 -0.10 -0.0086 0.10 Final Value(s) 0.87757 438.78 -0.1017 Tolerance 0.01 0.001 (Ratio) 0.002

(see page 5-3).

50"

P = 1 psi T = Thickness

100" P roble m G e ome t ry

Finit e E le me nt Me s h

In de x

COSMOSM Advanced Modules

5-63

Chapter 5 Additional Problems

OPZST13: Minimum Weight of a Cantilever Plate Subject to Stress Constraint Quadrilateral Continuum Elements
KEYWORDS: Sizing, static analysis, 8-node PLANE2D, minimum weight, and stress constraint.
100 lbs

(see page 5-3).

5"

50" P roble m G e ome t ry

50"

Finit e E le me nt Me s h

PROBLEM: Find the cantilever plate thickness for each half. The cantilever length is 100 in., height is 5 in., modulus of elasticity is 1E07 psi, Poisson's ratio is 0, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints B1 B2 Weight von Mises 1 von Mises 2 Initial Value(s) and Bounds 0.01 2 2 0.01 2 2 100 -10000 1200 10,000 -10000 600 10,000 Final Value(s) 0.23850 0.11628 8.86944 10,070 10,069 Tolerance 0.001 0.001 0.001 (Ratio) 100 100

In de x

5-64

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZST14: Minimum Volume of a Plate Subject to Displacement Constraint Triangular Continuum Elements
KEYWORDS: Sizing, static analysis, 3-node TRIANG, minimum weight, and displacement constraint.

(see page 5-3).

1"

T = Thickness

1 psi

Finit e E le me nt Me s h 1"

P roble m G e ome t ry

PROBLEM: Find the plate thickness. The plate length is 1 inch, height is 1 inch, modulus of elasticity is 5 psi, and Poisson's ratio is 0. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T (Thickness) Volume Displacement UX Initial Value(s) and Bounds 0.1 1 2 1.0 0.01 0.2 0.4 Final Value(s) 0.50025 0.50025 0.3998 Tolerance 0.01 0.001 (Ratio) 0.001

In de x

COSMOSM Advanced Modules

5-65

Chapter 5 Additional Problems

OPZST15: Minimum Volume of a Plate Subject to Stress Constraint Triangular Continuum Elements
KEYWORDS: Sizing, static analysis, 3-node TRIANG, minimum weight, and stress constraint.

(see page 5-3).

1"

T = Thickness

1 psi

Finit e E le me nt Me s h 1"

P roble m G e ome t ry

PROBLEM: Find the plate thickness. The plate length is 1 inch, height is 1 inch, modulus of elasticity is 5 psi, and Poisson's ratio is 0. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T (Thickness) Volume Stress x Initial Value(s) and Bounds 0.1 1 2 1.0 0.1 1.0 2.0 Final Value(s) 0.499 0.499 2.004 Tolerance 0.01 0.001 (Ratio) 0.01

In de x

5-66

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZST16: Minimum Volume of a Plate Subject to Stress Constraint Quadrilateral Continuum Elements
KEYWORDS: Sizing, static analysis, 8-node PLANE2D, minimum volume, and stress constraint.

(see page 5-3).

Thickness T1

Thickness T2

1 psi

1"

Finit e E le me nt Me s h 0.5" 0.5"

P roble m G e ome t ry

PROBLEM: Find the plate thickness for each half. The plate length is 1 inch, height is 1 inch, modulus of elasticity is 5 psi, and Poisson's ratio is 0. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T1 (Thickness) T2 (Thickness) Volume Stress x 1 Stress x 2 Initial Value(s) and Bounds 0.1 2 2 0.1 1 2 1.5 -2 1.0 2 -2 1.0 2 Final Value(s) 1.0034 0.499 0.7512 1.9932 2.004 Tolerance 0.01 0.01 0.001 (Ratio) 0.01 0.01

In de x

COSMOSM Advanced Modules

5-67

Chapter 5 Additional Problems

OPZST17: Minimum Volume of a Plate Subject to Stress Constraint Quadrilateral Continuum Elements
KEYWORDS: Sizing, static analysis, 4-node PLANE2D, minimum volume, and stress constraint.

(see page 5-3).

T = Thickness

1 psi

1"

Finit e E le me nt Me s h 1"

P roble m G e ome t ry

PROBLEM: Find the plate thickness. The plate length is 1 inch, height is 1 inch, modulus of elasticity is 5 psi, and Poisson's ratio is 0. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T (Thickness) Volume von Mises stress Initial Value(s) and Bounds 0.1 1 1 1.0 0 1.0 2.0 Final Value(s) 0.50126 0.50126 1.995 Tolerance 0.01 0.001 (Ratio) 0.001

In de x

5-68

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZST18: Minimum Weight of a Simply Supported Rectangular Plate Subject to Effective Strain Constraint Shell Elements
KEYWORDS: Sizing, static analysis, SHELL3, bending, minimum weight, and effective strain constraint. PROBLEM: Find the thickness of a 100 x 50 inch plate subject to pressure of 1 psi. The modulus of elasticity is 1E07 psi, Poisson's ratio is 0.3, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T (Thickness) Weight Effective Strain Initial Value(s) and Bounds 0.01 2 5 1000 2.0E-5 2.47E-5 1.28E-4 Final Value(s) 0.87699 438.495 1.287E-4 Tolerance 0.01 0.001 (Ratio) 1.08E-6

(see page 5-3).

50"

P = 1 psi T = Thickness

100" P roble m G e ome t ry

Finit e E le me nt Me s h

In de x

COSMOSM Advanced Modules

5-69

Chapter 5 Additional Problems

OPZST19: Minimum Weight of a Simply Supported Rectangular Plate Subject to Strain Energy Density Constraint Shell Elements
KEYWORDS: Sizing, static analysis, SHELL3, bending, minimum weight, and strain energy density constraint. PROBLEM: Find the thickness of a 100 x 50 inch plate subject to pressure of 1 psi. The modulus of elasticity is 1E07 psi, Poisson's ratio is 0.3, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T (Thickness) Weight Strain Energy Density Initial Value(s) and Bounds 0.01 2 5 1000 0 0.0022 0.059 Final Value(s) 0.8771 438.550 0.05941 Tolerance 0.01 0.001 (Ratio) 0.00059

(see page 5-3).

50"

P = 1 psi T = Thickness

100" P roble m G e ome t ry

Finit e E le me nt Me s h

In de x

5-70

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZST20: Minimum Volume of a Cantilever Pipe Subject to Stress Constraint Pipe Radius
KEYWORDS: Sizing, static analysis, PIPE, radius of cross-section, minimum volume, and stress constraint.
Radius Average Radius 1 Average Radius 2 100

(see page 5-3).

50" Thickness P roble m Ge ome try

50"

Finite E le me nt Me sh

PROBLEM: A cantilever pipe is subject to a concentrated load at the tip. Find the pipe radii (in each half). The cantilever length is 100 inches, modules of elasticity is 1 x 107 psi, and Poisson's ratio is 0.30. The initial values and bounds of design variables, objectives function and constraints are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints RAVE1 RAVE2 Volume x1 x2 Initial Value(s) and Bounds 0.5 1.6 2 0.5 1.6 2 231.6925 -10,000 6,302 10,000 -10,000 3,151 10,000 Final Value(s) 1.2994 0.9650 158.21 10,100 10,140 Tolerance 0.015 0.015 0.01 (Ratio) 200 200

In de x

COSMOSM Advanced Modules

5-71

Chapter 5 Additional Problems

OPZST21: Minimum Volume of a Cantilever Pipe Subject to Stress Constraint Pipe Thickness
KEYWORDS: Sizing, static analysis, PIPE, thickness of cross section, minimum volume, and stress constraint.
Thickness 100 T1 T2

(see page 5-3).

50" P roble m Ge ome try

50"

Finite E le me nt Me sh

PROBLEM: A cantilever pipe is subject to a concentrated load at the tip. Find the pipe thickness (in each half). The cantilever length is 100 inches, modules of elasticity is 1 x 107 psi, and Poisson's ratio is 0.30. The initial values and bounds of design variables, objectives function and constraints are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T1 T2 Volume x1 x2 Initial Value(s) and Bounds 0.01 0.50 0.5 0.01 0.25 0.5 304.3418 -10,000 4,701 10,000 -10,000 3,643 10,000 Final Value(s) 0.1644 0.07497 107.68 10,149 10,170 Tolerance 0.005 0.005 0.01 (Ratio) 200 200

In de x

5-72

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZFQ1: Minimum Weight of a Cantilever Subject to Frequency Constraint Beam Elements

KEYWORDS: Sizing, frequency analysis, BEAM2D, minimum weight, and fundamental frequency constraint.
1"

(see page 5-3).

E = 10 7 psi Cros s S e c t ion 100" P roble m G e ome t ry

Finit e E le me nt Me s h

PROBLEM: Find the height of a cantilever. The beam's length is 100 inches, width is 1 inch, modulus of elasticity is 1 x 107 psi and material density is 0.10 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints H (Height) Weight Fundamental frequency Initial Value(s) and Bounds 0.1 10 20 100 0.8 1.6135 4 Final Value(s) 4.9109 49.109 0.7924 Tolerance 0.01 0.01 (Ratio) 0.01

In de x

COSMOSM Advanced Modules

5-73

Chapter 5 Additional Problems

OPZFQ2: Minimum Weight of a Cantilever Subject to Frequency Constraint Shell Elements

KEYWORDS: Sizing, frequency analysis, SHELL4, minimum weight, and frequency constraint.

(see page 5-3).

5"

100" P roble m G e ome t ry

Finit e E le me nt Me s h

PROBLEM: Find the cantilever plate thickness. The cantilever length is 100 in, height is 5 in, modulus of elasticity is 1 x 107 psi, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints DV1 Weight Fundamental frequency Initial Value(s) and Bounds 0.1 2 3 100 0.03 0.3243 1 Final Value(s) 0.18494 9.24676 0.029989 Tolerance 0.001 0.01 (Ratio) 0.0001

In de x

5-74

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZFQ3: Minimum Weight of a Pipe Cantilever Subject to Frequency Constraint Pipe Elements
KEYWORDS: Sizing, frequency analysis, PIPE, radius of cross-section, minimum weight, and frequency constraint.
R

(see page 5-3).

100" P roble m Ge ome try

Finite E le me nt Me sh

PROBLEM: Find the pipe radius. The cantilever length is 100 in., modulus of elasticity is 1 x 107 psi, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints R Weight Fundamental frequency Initial Value(s) and Bounds 0.1 1.5 2 21.59846 0.2 0.5432 2 Final Value(s) 0.57467 7.063443 0.18381 Tolerance 0.019 0.01 (Ratio) 0.018

In de x

COSMOSM Advanced Modules

5-75

Chapter 5 Additional Problems

OPZBK1: Minimum Weight of a Cantilever Subject to Buckling Load Factor Constraint Beam Elements
KEYWORDS: Sizing, buckling analysis, BEAM2D, minimum weight, and buckling load factor constraint.
1"

(see page 5-4).

E = 10 7 psi F = 1,000 lb Cros s S e c t ion 100"

P roble m Ge ome try

Finite E le me nt Me sh

PROBLEM: Find the height of a cantilever beam subject to a concentrated compressive force of 1000 lb The beam's length is 100 in, width is 5 in, modulus of elasticity is 1 x 107 psi and material density is 0.10 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints H (Height) Weight Buckling load factor Initial Value(s) and Bounds 0.1 5 10 50 2 25.702 50 Final Value(s) 2.101 21.007 1.9060 Tolerance 0.001 0.01 (Ratio) 0.1

In de x

5-76

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZBK2: Minimum Weight of a Cantilever Subject to Buckling Constraint Shell Elements

KEYWORDS: Sizing, buckling analysis, SHELL4, and minimum weight, buckling constraint.
250 lbs/ Node

(see page 5-4).

5"

100" P roble m G e ome t ry

Finit e E le me nt Me s h

PROBLEM: Find the cantilever plate thickness. The cantilever length is 100 in, height is 5 in, modulus of elasticity is 1 x 107 psi, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T (Thickness) Weight Buckling load factor Initial Value(s) and Bounds 0.1 2 3 100 2 8.2604 50 Final Value(s) 1.2279 61.393 1.9115 Tolerance 0.001 0.01 (Ratio) 0.01

In de x

COSMOSM Advanced Modules

5-77

Chapter 5 Additional Problems

OPZBK3: Minimum Weight of a Cantilever Plate Subject to Buckling Constraint Quadrilateral Continuum Elements
KEYWORDS: Sizing, buckling analysis, 8-node PLANE2D, minimum weight, and buckling constraint.
250 lbs/ Node 100"

(see page 5-4).

5"

P roble m Ge ome try

Finite E le me nt Me sh

PROBLEM: Find the cantilever plate thickness. The cantilever length is 100 inches, height is 5 inches, modulus of elasticity is 1 x 107 psi, and material density is 0.1 psi. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints T (Thickness) Weight Buckling load factor Initial Value(s) and Bounds 0.1 1 2 50 2 14.66 50 Final Value(s) 0.13 6.50 1.906 Tolerance 0.001 0.001 (Ratio) 0.1

In de x

5-78

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZBK4: Minimum Weight of a Pipe Cantilever Subject to Buckling Load Constraint Pipe Elements
KEYWORDS: Sizing, buckling analysis, PIPE, radius of cross-section, minimum weight, and linearized buckling constraint.
R 1,000 lbs

(see page 5-4).

100" P roble m Ge ome try

Finite E le me nt Me sh

PROBLEM: Find the pipe radius. The cantilever length is 100 in, modulus of elasticity is 1 x 107 psi, and material density is 0.1 lb/in3. The initial values and bounds of design variables, constraints and the objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints R Weight Buckling load factor Initial Value(s) and Bounds 0.1 1.5 2 21.598 2 5.0732 6 Final Value(s) 1.1242 15.696 1.9624 Tolerance 0.019 0.01 (Ratio) 0.04

In de x

COSMOSM Advanced Modules

5-79

Chapter 5 Additional Problems

OPZBK5: Maximum Buckling Load Design of a Graphite-epoxy Laminate

KEYWORDS: Buckling analysis, SHELL4L elements, sizing optimization (ply orientation), maximum buckling load factor cr. PROBLEM: Find the ply orientation of a 8-layer [ / - / / - ]s Graphite-Epoxy laminated composite plate which is subject to a uniaxial edge pressure loading 1 psi. The total thickness of 8 plies is 1 inch. The material constants are given as: modulus of elasticity in the first material direction Ex = 26.27E6 psi, modulus of elasticity in the second material direction Ey = Initial Problem Geometry and Finite Element Mesh 1.49E6 psi, Poissons ratio = 0.28, shear modulus in the material first and second plane Gxy = 1.04E6 psi, tensile and compressive strengths in the first material direction F1T = F1C = 2.17E5 psi, tensile and compressive strengths in the second material direction F2T = 5.81E3 psi and F2C = 3.57E4 psi, shear strength in the material first and second plane F12 = 9.87E3 psi. Four cases are studied in which the aspect ratios of the dimension A to B are equal to 1, 2, 3, and 4. The input data regarding optimization as well as the converged results are listed in the following table. Note that the optimum design is achieved without applying any constraint.

(see page 5-4).

In de x

5-80

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

SUMMARY OF RESULTS:
Optimization Parameters Design Variable, Objective Function, cr Initial Value(s) and Bounds 1 2 3 4 1.E-16 < 30 < 90 860.96 1.E-16 < 30 < 90 846.81 1.E-16 < 30 < 90 965.58 1.E-16 < 30 < 90 888.42 Final Value(s) 45.146 948.96 41.986 1193.0 44.985 1062.6 46.768 1126.1 0.9 0.01 (ratio) 0.9 0.01 (ratio) 0.9 0.01 (ratio) 0.9 0.01 (ratio)

A/B

Tolerance

In de x

COSMOSM Advanced Modules

5-81

Chapter 5 Additional Problems

OPZNB1: Snap Buckling of a Thin Hinged Cylindrical Shell Under a Central Point Load
KEYWORDS: Sizing, nonlinear analysis, linearized buckling, snapthrough/snapback, arc-length, limit point, SHELL4, minimum volume, and user-defined constraint. PROBLEM:
P = 1.0 Thickness b E = 3102.75 N/mm 2 = 0.30 = 2540 mm = 254 mm = 0.10 Rad

(see page 5-4).

R b

Finite Ele me nt

Find the thickness Me sh of a thin cylindrical shell. The curved edges of the shell are free while the straight edges are hinged and immovable. The modulus of elasticity is 3102.75 N/mm2 and Poisson's ratio is 0.3. The initial values and bounds of design variables, constraints and objective function are shown below. SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Thickness Volume Limit Point Load Factor or Linearized Buckling Load Factor (userdefined) Initial Value(s) and Bounds 5 10 10 6.451412 x 105 50 588.39 10,000 Final Value(s) 5.7611 3.7167E5 Tolerance 0.10 0.01 (Ratio)

Constraints

49.151

1.0

In de x

5-82

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

OPZFDS1: Modal Time History Analysis of a Simply Supported Shell Structure

KEYWORDS: Frequency analysis, post-dynamic analysis (time history analysis), dynamic stress analysis, SHELL4 elements, sizing optimization, multidisciplinary optimization, minimum strain design, frequency and mass constraints.

(see page 5-4).

Initial Problem Geometry

Initial Finite Element Mesh

In de x

COSMOSM Advanced Modules

5-83

Chapter 5 Additional Problems

Pressure Loading versus Time Curve

PROBLEM: Find the thickness t of a simply supported shell structure which is subject to an impulsive pressure loading 1 psi. The material constants of the shell are given as: Youngs modulus E = 1E7 psi, Poissons ratio = 0.3, and Density = 0.1 lb*sec2/ in4. The input data regarding optimization as well as the converged results are listed in the following table. Note that the objective function, effective strain, is defined as the extreme value within the desired range of time (0.0 - 0.6 sec). SUMMARY OF RESULTS:
Optimization Parameters Design Variables Objective Function Constraints DV1 Effective Strain Frequency (1st) Mass Initial Value(s) and Bounds 0.01 0.1 1 5.6057E-4 1 31560 10 0.01 52.255 100. Final Value(s) 0.1927 2.7323E-4 3.9598 100.70 Tolerance 0.001 0.01 (Ratio) 0.09 1.0

In de x

5-84

COSMOSM Advanced Modules

Dominantly Shape Sensitivity Problems

In de x

COSMOSM Advanced Modules

5-85

Chapter 5 Additional Problems

SNSST1: Sensitivity Study of a Steering Control Arm in Linear Stress Analysis

KEYWORDS: Shape, static analysis, TETRA4, local sensitivity, stress response quantity.
Y TE TRA4 E LE ME NTS E Py = 2 x 105 N/mm 2 = 0.3 = 31 N/mm 2 20 30 X Internal surfaces fixed in all directions 20

(see page 5-4).

Note: All dimensions in milimeters.

7 tw tr 2 dint

tr1

140

Py

I nit ia l P roble m G e ome t ry I nit ia l Finit e E le me nt Me s h

PROBLEM: Study the sensitivity of the steering control arm to changes in the thickness of the shafts, size and location of the cutout. Perform a local sensitivity study by perturbing the design variables by a ratio of 0.10. The arm outer thickness is 20 mm, modulus of elasticity is 200,000 N/mm2 and Poisson's ratio is 0.30 SUMMARY OF RESULTS:
Design Variables 1 2 3 4 Name TR1 TR2 TW DINT Initial Value 24 19 8 6 Perturbation Ratio -0.10 -0.10 -0.10 +0.10 Gradient of Maximum von Mises Stress -1.131 1.1946 -11.324 4.3820

In de x

5-86

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

SNSST2: Sensitivity Study of an Engine Bearing Cap in Stress Analysis Under Multiple Load Cases
KEYWORDS: Shape, static analysis, TETRA10, primary and secondary load cases, local sensitivity analysis, and stress response quantity.
TWEB 1 0 -Node Te tra Me sh HUMP Symmetry Boundary Conditions Fixed E = 2 x 10 5 N/mm 2 = 0.30

(see page 5-4).

Height

Initial P roble m Ge ome try

ECCENT

Initial Finite E le me nt Me sh

PROBLEM: Study the sensitivity of the bearing cap to changes in its dimensions. Perform a local sensitivity study by perturbing the design variables by a ratio of 0.10. The bearing cap modulus of elasticity is 200,000 N/mm2 and Poisson's ratio is 0.30. SUMMARY OF RESULTS
Design Variables Set No. 1 2 3 4 Name ECCENT TWEB HEIGHT HUMP Initial Value 75 10 70 15 Perturbation Ratio -0.10 -0.10 -0.10 -0.10 Gradient of Max von Mises Stress Load Case 1 0.6238 1.7975 0.0633 0.03275 2 1.4428 -12.318 -1.7607 0.4751 51 1.7547 -14.555 -2.1395 1.6491

In de x

COSMOSM Advanced Modules

5-87

Chapter 5 Additional Problems

SNSFQ1: Sensitivity Study of a Control Arm Bracket in Frequency Analysis

KEYWORDS: Shape, frequency analysis, SHELL3, global sensitivity, frequency response quantity.
r = 1.0 r = 2.0 3 - Node Tria ngula r Me s h ( S iz e = 1 ) Thickness E = = = = 0.3 cm 2 x 107 N/cm2 0.30 0.0075 Kg/cm3

(see page 5-4).

1.5

1.5

Note: All dimensions in centimeters.

15.0

t1 0.5 1.0 t2 X

t1 1.0

PROBLEM:

Fixed

Study the sensitivity Fixed 5.0 5.0 of the I nit ia l P roble m G e ome t ry I nit ia l Finit e E le me nt Me s h fundamental frequency of the control arm bracket to changes in design variables. Use the global sensitivity option by changing only one design variable at a time in 5 increments. The bracket thickness is 0.3 cm, modulus of elasticity is 2 x 107 N/cm2 Poisson's ratio is 0.3, and mass density is 0.0075 Kg/cm3. SUMMARY OR RESULTS:
Design Variables T1 0.5 1.0 1.5 2.0 2.5 T2 3.5 Fundamental Frequency (Hz) 8.6495 10.798 12.119 13.029 13.662 Design Variables T1 2.5 T2 1.5 2.0 2.5 3.0 3.5 Fundamental Frequency (Hz) 13.035 13.170 13.329 13.491 13.662

In de x

5-88

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

SNSBK1: Sensitivity Study of a C-shape Column in Linearized Buckling Analysis

KEYWORDS: Shape, linearized buckling analysis, SHELL3, global sensitivity, buckling response quantity.
Fixed T1 Fillet (T2 /4) Element Size Thickness Material E T1 S HE LL3 = = = = 3 in 0. 25 in A_Steel 3 x 107 psi

(see page 5-4).

B = 20" T2

= 0.28 -3 2 4 = 0.73 x 10 lbf sec /in

0.18h

T2

T1

A = 6"

0.32h

T2

C = 3"

h = 120"

0.32h

I nit ia l P roble m G e ome t ry I nit ia l Finit e E le me nt Me s h

0.18h

p= 5,000 psi

PROBLEM: Study the sensitivity of the column's buckling load factor to changes in design variables. Change the design variables one at a time in 5 increments. The column thickness is 0.25 inches, modulus of elasticity is 3 x 107 psi, Poisson's ratio is 0.28, and material density is 0.73 x 10-3 lb. sec2/in4.
In de x

COSMOSM Advanced Modules

5-89

Chapter 5 Additional Problems

SUMMARY OR RESULTS:
Design Variables T1 10.00 16.25 22.50 28.75 35.00 10 T2 5 5.00 6.75 8.50 10.25 12.00 Linearized Buckling Load Factor 2.8069 2.7425 2.6475 2.5432 2.4502 2.8069 2.7630 2.6785 2.5992 2.5103

In de x

5-90

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

SNSHT1: Sensitivity Study of a Circular Disk in Heat Transfer Analysis

KEYWORDS: Shape and sizing, thermal analysis, SHELL3, global sensitivity, temperature response quantity.
Heat Flux 0.1 W/mm 2
5

(see page 5-4).

10 5 5 10 Convection 0.0005 W/mm2 -C Ambient Temperature 50C

Radius

5 5

25 10

Initial P roble m Ge ome try

25 Heat Flux 0.1 W/mm 2 Convection 0.0 W/mm 2 -C Ambient Temperature 50C

3 - Node S he ll Me s h ( S iz e = 2 ) Thickness = 2 mm

Initial Finite E le me nt Me sh

In de x

COSMOSM Advanced Modules

5-91

Chapter 5 Additional Problems

PROBLEM: Study the sensitivity of the circular disk temperature to changes in thickness and radius of the model. Use the global sensitivity feature and change one design variable at a time. The disk conductivity is 0.57 W/mm-C. A convection of 0.0005 W/mm2 - C with an ambient temperature of 50C is applied to the entire model except for the heat sources. The heat source regions (heat flux of 0.1 W/mm2) is assumed to have a convection of 0 W/mm2-C with an ambient temperature of 50 C. SUMMARY OR RESULTS:
Design Variables Temperature Radius 30 40 50 60 70 70 Thickness 2 0.500 0.875 1.250 1.625 2.000 144.6651 108.324 91.6635 82.8095 77.6308 111.055 92.2412 84.5067 80.2878 77.6308

In de x

5-92

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

SNSN1: Sensitivity Study of a Thick Walled Pipe in Nonlinear Analysis

KEYWORDS: Shape, nonlinear analysis, von Mises plasticity, automatic time stepping, PLANE2D, global sensitivity, and stress response quantity.

(see page 5-4).

Rint Rout Initial P roble m Ge ome try

Initial Finite E le me nt Me sh

PROBLEM: Study the sensitivity of the pipe to changes in the outer radius. The modulus of elasticity is 86,666 psi, Poisson's ratio is 0.3, yield stress is 17.32 psi and tangential modulus is 866 psi. Use the global sensitivity feature for 5 increments. SUMMARY OF RESULTS:
Design Variable: ROUT 1.50 1.75 2.0 2.25 2.50 von Mises Stress 25.0516 17.4911 17.3962 17.3580 17.3420

In de x

COSMOSM Advanced Modules

5-93

Chapter 5 Additional Problems

SNSTSF1: Sensitivity Study of a Circular Disk in Thermal, Stress and Frequency Analyses
KEYWORDS: Shape, thermal analysis, static analysis, frequency analysis, SHELL3, TEMPREAD, local sensitivity, temperature, displacement and frequency response quantities.
Heat Flux 0.1 W/mm 2
5

(see page 5-4).

10 5 5 10 Convection 0.0005 W/mm2 -C Ambient Temperature 50C

Radius

5 5

25 10

Initial P roble m Ge ome try

25 Heat Flux 0.1 W/mm 2 Convection 0.0 W/mm 2 -C Ambient Temperature 50C

3 - Node S he ll Me s h ( S iz e = 2 ) Thickness = 2 mm

Initial Finite E le me nt Me sh

In de x

5-94

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

PROBLEM: Study the sensitivity of the circular disk response to changes is its thickness and radius. Perform a local sensitivity study by perturbing the design variables by a ratio of 0.1. The disk conductivity is 0.57 W/mm-C. A convection of 0.0005 W/mm2-C with an ambient temperature of 50C is applied to the entire model except for the heat sources. The heat source regions (heat flux of 0.1 W/mm-C) is assumed to have a convection of 0 W/mm2-C with an ambient temperature of 50C. SUMMARY OF RESULTS
Design Variables Set No. 1 2 Name Radius Thickness Initial Value 70 1.8 Perturbation Ratio -0.10 -0.10 Gradient of Response Quantities Temperature -0.4415 -6.3049 Resultant Displacement +0.003120 -0.01540 Fundamental Frequency -0.04806 +0.5076

In de x

COSMOSM Advanced Modules

5-95

Chapter 5 Additional Problems

SNSTSB1: Sensitivity Study of a C-shape Column in Thermal, Stress and Buckling Analyses
KEYWORDS: Shape, thermal analysis, static analysis, buckling analysis, SHELL3, TEMPREAD, global sensitivity, buckling load factor, displacement and temperature response quantities.
Fixed T1 Fillet (T2 /4) Element Size Thickness Material E T1 S HE LL3 = = = = 3 in 0. 25 in A_Steel 3 x 107 psi

(see page 5-4).

B = 20" T2

= 0.28 -3 2 4 = 0.73 x 10 lbf sec /in

0.18h

T2

T1

A = 6"

0.32h

T2

C = 3"

h = 120"

Fixed 0.32h

I nit ia l P roble m G e ome t ry I nit ia l Finit e E le me nt Me s h

0.18h

In de x

5-96

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

PROBLEM: Study the sensitivity of the C-Shape column to changes in the cutouts sizes. Change design variables one at a time in 5 increments. The cross section thickness is 0.25 inch, modulus of elasticity is 3 x 107 psi, and Poisson's ratio is 0.28. The material conductivity is 6.7E-4 BTU/in/s F. A convection of 0.0001 BTU/sec in2-F with an ambient temperature of 50F and a volume heat of 0.005 BTU/sec in3 are applied for the entire model. The material's coefficient of thermal expansion is 7.4E-6/F. SUMMARY OR RESULTS:
Design Variables T1 10.00 16.25 22.50 28.75 35.00 10.00 10.00 10.00 10.00 10.00 T2 5 5 5 5 5 5.00 6.75 8.50 10.25 12.00 Max Temp 88.316 89.275 90.325 90.713 90.939 88.316 89.402 89.321 88.905 89.674 Response Quantities Max Resultant Displ. 0.01299 0.01739 0.02204 0.02327 0.01941 0.01299 0.01294 0.01232 0.01235 0.01306 Buckling Load Factor 1.8271 1.7903 1.7771 1.8015 1.8406 1.8271 2.1557 2.5762 2.9213 3.1636

In de x

COSMOSM Advanced Modules

5-97

Chapter 5 Additional Problems

SNSTN1: Sensitivity Study of a Cylinder in Intransient Thermal and Nonlinear Analyses

KEYWORDS: Shape, transient thermal, nonlinear analysis, radiation, convection, heat flux, element heat, prescribed temperatures, multi-disciplinary global sensitivity study, von Mises plasticity, Axisymmetric PLANE2D, stress, strain and temperature response quantities.
Convection _ Ambient Temperature = 200 F Firm Coefficient = 0.1 BTU / (sec in in F) Element Heat = 0.140625 BTU / (sec in in in) Temperature = 70 F Thick

(see page 5-4).

Radiation Source = 500 F Emissivity = 0.9 View Factor = 0.8

Heat Flux = 0.0125 BTU / (sec in in) Initia l Ge ome try, Loa ds a nd Bounda ry C onditions

Thic k ROUT Initia l Finite Ele me nt Me sh

PROBLEM: Study the sensitivity of a cylinder to changes in its outer diameters and thickness. Change the design variables one at a time in 5 increments. Steel alloy and aluminum materials are used.

In de x

5-98

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

SUMMARY OF RESULTS:
Design Variables ROUT 5.00 6.75 8.50 10.25 12.00 12.00 Thick 2.000 0.500 0.875 1.250 1.625 2.000 von Mises 1 57,563 57,911 52,877 47,987 44,040 22,894 28,577 34,039 39,259 44,040 2 31,415 30,495 30,283 30,315 30,217 22,724 27,551 29,361 30,130 30,217 Response Quantities Temperature 570.292 509.935 470.377 439.297 413.634 260.386 304.249 345.801 382.817 413.634 Eff. Strain 1 0.5043E-2 0.5094E-2 0.4670E-2 0.4247E-2 0.3896E-2 0.2025E-2 0.2525E-2 0.3012E-2 0.3472E-2 0.3896E-2 2 0.4254E-2 0.3051E-2 0.2569E-2 0.2536E-2 0.2109E-2 0.0839E-2 0.1085E-2 0.1371E-2 0.1782E-2 0.2109E-2

In de x

COSMOSM Advanced Modules

5-99

Chapter 5 Additional Problems

SNSFDS1: Sensitivity Study of a Culvert in Harmonic Response Analysis

KEYWORDS: Frequency analysis, post-dynamic analysis (harmonic response), dynamic stress analysis, 4-noded PLANE2D elements, shape, offset sensitivity, frequency, displacement, and stress response quantities.

(see page 5-4).

Initial Problem Geometry

Initial Finite Element Mesh

In de x

5-100

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

PROBLEM: Study the sensitivity of a culvert due to changes in the design variables, radius R and slope s. The material constants of the culvert are given as: Youngs modulus E = 30E6 psi, Poissons ratio = 0.3, and Density = 1. lb*sec2/in4. A harmonic pressure loading with constant amplitude 500 psi within the desired range of frequency (1 rad/sec - 400 rad/sec) is applied to the top of the culvert. A modal damping 0.015 is assumed for the first 10 modes. Six sets of design variables and their respective results are listed in the following table. Note that the response quantities, displacement and stress, are defined as the extreme values within the desired range of frequency. SUMMARY OF RESULTS:
Design Variables Set Number 1 2 3 4 5 6 DV1 (R) 20 23 26 29 32 35 DV2 (s) 2.0 1.86 1.7 1.55 1.4 1.3333 Response Quantities Frequency (1st) 16.181 14.851 13.354 11.684 9.5110 6.8929 Displacement (Uy) 0.0429 0.0575 0.0803 0.1177 0.1948 0.4574 Stress (von Mises) 4.7188E4 5.3807E4 6.2323E4 8.0398E4 1.2790E5 2.3871E5

In de x

COSMOSM Advanced Modules

5-101

Chapter 5 Additional Problems

SNSFDS2: Sensitivity Study of a Lever Arm in Random Vibration Analysis

KEYWORDS: Frequency analysis, post-dynamic analysis (random vibration), dynamic stress analysis, 6-noded TRIANG elements, shape, global sensitivity (one-by-one), frequency, displacement, and stress response quantities.

(see page 5-4).

Initial Finite Element Mesh Initial Problem Geometry

In de x

5-102

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

Base Excitation versus Frequency Curve

Pressure Loading versus Frequency Curve

PROBLEM: Study the sensitivity of a lever arm due to changes in the design variables, hyperbolic arc parameter RATIO and thickness T2. The arm is made of A_STEEL and has a uniform thickness 1.0 in. Both harmonic pressure loading and base excitation (acceleration) in the y-direction are applied to the structure as shown in the figure. A modal damping 3% is assumed for the first 5 modes. Changing one design variable

In de x

COSMOSM Advanced Modules

5-103

Chapter 5 Additional Problems

at a time in 5 increments, their values and the respective results are listed in the following table. Note that the response quantity, displacement, is defined as the extreme value of PSD within the desired range of frequency and the stress is the extreme value of RMS. SUMMARY OF RESULTS:
Design Variables DV1 (RATIO) 0.5 0.5625 0.625 0.6875 0.75 0.5 DV2 (T2) 40.0 15.0 23.75 32.5 41.25 50.0 Response Quantities Frequency (1st) 92.563 87.351 81.806 75.813 69.242 85.956 89.917 92.059 92.524 91.206 Displacement (Uy) 3.1740E-4 3.2949E-4 3.4710E-4 3.7408E-4 4.1664E-4 2.3235E-4 1.6545E-4 2.1621E-4 3.1188E-4 2.2683E-4 Stress (von Mises) 1.1834E4 1.2498E4 1.3734E4 1.5190E4 1.7617E4 7710.6 9111.7 1.1198E4 1.1832E4 1.1321E4

In de x

5-104

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

SNSFDS3: Sensitivity Study of a Trophy in Response Spectrum Analysis

KEYWORDS: Frequency analysis, post-dynamic analysis (response spectrum analysis), dynamic stress analysis, SHELL4 elements, shape, local sensitivity, frequency, displacement, and stress response quantities.

(see page 5-4).

Initial Problem Geometry

Initial Finite Element Mesh

In de x

COSMOSM Advanced Modules

5-105

Chapter 5 Additional Problems

Acceleration Spectrum Excitation

PROBLEM: Study the sensitivity of a trophy due to changes in its dimensions, R2, R3, H3, and H4. The trophy is made of A_STEEL and has a 5% critical damping. The base of the trophy is experiencing an acceleration spectrum excitation as shown in the figure. Performing a local sensitivity study by perturbing all the design variables by a ratio +0.05, the gradients of the response quantities are listed in the following table. Note that the response quantities, displacement and stress, are defined as the extreme values by using the SRSS mode combination method. SUMMARY OF RESULTS:
Design Variables Set No. 1 2 3 4 Name R2 R3 H3 H4 Initial Value 5 10 10 21 Perturbation Ratio +0.05 +0.05 +0.05 +0.05 Gradients of Response Quantities Frequency (1st) 9.8388 -2.4151 -0.4585 -1.1676 Displacement (Ux) -0.0393 9.0626E-3 4.3791E-4 8.1406E-3 Stress (von Mises) -1241.5 475.12 116.50 219.92

In de x

5-106

COSMOSM Advanced Modules

Sizing Sensitivity Problems

In de x

COSMOSM Advanced Modules

5-107

Chapter 5 Additional Problems

SNZST1: Global Sensitivity of a 2-bar Truss: All Design Variables Incremented Simultaneously
KEYWORDS: Sizing, static analysis, TRUSS2D, global sensitivity.
U2 A1 1 2 A2 3 F = 10 lbs U3

(see page 5-4).

30"

30"

P roble m Ge ome try

Finite E le me nt Me sh

PROBLEM: Study the behavior of the 2-bar truss for different values of cross sectional areas A1, and A2. The length of each bar is 30 inches and modulus of elasticity is 5 psi. The design variables are to be changed simultaneously in 5 increments. SUMMARY OF RESULTS:
Run Number 1 2 3 4 5 Design Variables A1 1.000 3.250 5.500 7.750 10.000 A2 1.000 3.250 5.500 7.750 10.000 Response Quantities U2 60.000 18.462 10.909 7.7419 6.000 U3 120.000 36.923 21.818 15.484 12.000

In de x

5-108

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

SNZST2: Global Sensitivity of a 2-bar Truss: Design Variables Incremented One at a Time
KEYWORDS: Sizing, static analysis, TRUSS2D, global sensitivity.
1 U2 A1 2 A2 3 F = 10 lbs U3

(see page 5-4).

PROBLEM: Study the behavior of the 2-bar truss for different values of cross sectional areas A1, and A2. The length of each bar is 30 inches and modulus of elasticity is 5 psi. The design variables are to be changed one at a time in 5 increments. SUMMARY OF RESULTS:
Run Number 1 2 3 4 5 6 7 8 9 10 Design Variables A1 1.0 3.25 5.50 7.75 10.00 1.0 1.0 1.0 1.0 1.0 A2 1.0 1.0 1.0 1.0 1.0 1.0 3.25 5.50 7.75 10.00

30"

30"

P roble m Ge ome try

Finite E le me nt Me sh

Response Quantities U2 60.000 18.462 10.909 7.7419 6.000 60.000 60.000 60.000 60.000 60.000 U3 120.000 78.462 70.909 67.742 66.000 120.000 78.462 70.909 67.742 66.000

In de x

COSMOSM Advanced Modules

5-109

Chapter 5 Additional Problems

SNZST3: Offset Sensitivity of a 2-bar Truss

KEYWORDS: Sizing, static, TRUSS2D, offset sensitivity.
A1 U2 A2 2 3 F = 10 lbs 1 U3

(see page

PROBLEM: Study the behavior of the 2-bar truss for specified values of cross sectional areas A1, and A2. The length of each bar is 30 in and modulus of elasticity is 5 psi. Four different sets are used to specify design variables. SUMMARY OF RESULTS:
Set Number 1 2 3 4

30"

30"

P roble m Ge ome try

Finite E le me nt Me sh

Design Variables A1 1.0 10.0 10.0 5.0 A2 10.0 1.0 10.0 5.0

Response Quantities U2 60.0 6.0 6.0 12.0 U3 66.0 66.0 12.0 24.0

In de x

5-110

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

SNZST4: Local Sensitivity of a 5-bar Truss

KEYWORDS: Sizing, static, TRUSS2D, local sensitivity. PROBLEM: Study the effect of changing (perturbing) design variables on the response of a 5-bar truss. The cross-sectional area of each bar is 1.0 in, length is 10 in (each) and modulus of elasticity is 5 psi. Each design variable is perturbed by 10%. SUMMARY OF RESULTS:
Design Variable A1 A2 A3 A4 A5 U1* -18.1818 0.0 0.0 0.0 0.0 U2* -18.1818 -18.1818 0.0 0.0 0.0 U3* -18.1818 -18.1818 -18.1818 0.0 0.0 U4* -18.1818 -18.1818 -18.1818 -18.1818 0.0 U5* -18.1818 -18.1818 -18.1818 -18.1818 -18.1818
A1 1 2 U2 A2 3 U3 A3 4 U4 A4 5 U5 A5 6 F = 10 lbs U6

(see page

10"

10"

10"

10"

10"

P roble m Ge ome try

Finite E le me nt Me sh

*Derivative of Response Quantity with Respect to Design Variables

In de x

COSMOSM Advanced Modules

5-111

Chapter 5 Additional Problems

SNZST5: Sensitivity Study of a Graphite-epoxy Laminate

KEYWORDS: Linear static analysis, SHELL4L elements, size (ply orientation), global sensitivity, maximum failure index response quantity.

(see page 5-4).

Initial Problem Geometry and Finite Element Mesh

PROBLEM: Study the sensitivity of a 6-layer [- / 0 / ]s Graphite-Epoxy laminate due to changes in the ply orientation . The total thickness of 6 plies is 1 inch. The material constants are given as: modulus of elasticity in the first material direction Ex = 26.27E6 psi, modulus of elasticity in the second material direction Ey = 1.49E6 psi, Poissons ratio = 0.28, shear modulus in the material first-second plane Gxy = 1.04E6 psi, tensile and compressive strengths in the 1st material direction F1T = F1C = 2.17E5 psi, tensile and compressive strengths in the second material direction F2T = 5.81E3 psi and F2C = 3.57E4 psi, shear strength in the material first-second plane F12 = 9.87E3 psi. A biaxial tensile loading is applied to the edges of the structure with the ratios of the transverse load (Ny) to the longitudinal load (Nx) equal to 0, 0.5, and 1.0 where Nx = 100 psi. Changing the ply orientation in 10 increments, their values and the respective maximum failure indexes are listed in the following table.

In de x

5-112

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

SUMMARY OF RESULTS:
Design Variable, q 1E-16 (=0) 10 20 30 40 50 60 70 80 90 Response Quantity, Maximum Failure Index Ny/Nx = 0 2.1236E-7 -2.3987E-4 -8.2275E-4 -1.2468E-3 -1.0689E-3 7.8222E-4 1.4606E-3 1.8906E-3 2.1198E-3 2.1912E-3 Ny/Nx = 0.5 7.2160E-3 6.7899E-3 5.4160E-3 3.3621E-3 1.7210E-3 1.3938E-3 1.6914E-3 2.0304E-3 2.2559E-3 2.3327E-3 Ny/Nx = 1 0.01446 0.01384 0.01167 7.9831E-3 4.5160E-3 2.6199E-3 1.9231E-3 2.1705E-3 2.3923E-3 2.4743E-3

In de x

COSMOSM Advanced Modules

5-113

Chapter 5 Additional Problems

SNZN1: Sensitivity of a Rubber Circular Ring to Coefficient of Friction in Nonlinear Analysis

KEYWORDS: Shape, nonlinear analysis, rubber, Mooney model, contact, prescribed displacement, automatic time stepping, axisymmetric PLANE2D, friction, and stress response quantity.
Top Steel Plate Rubber Ring Bottom Steel Plate 0.5615" 0.5615" 0.278" 0.3475" 0.3" 0.3"

(see page 5-4).

Initial P roble m Ge ome try

2 R Cross Initial Finite E le me nt Me sh

In de x

5-114

COSMOSM Advanced Modules

Part 2 OPSTAR / Optimization

PROBLEM: Study the sensitivity of circular ring response to the friction coefficient value. Use the global sensitivity option. For rubber, the Mooney's constants are 175 and 10 psi, and the Poisson's ratio is 0.49. For steel plates, the Young's modulus is 30 x 106 psi, and Poisson's ratio is 0.3. SUMMARY OF RESULTS:
Design Variable: FRICTION 0.01 0.05 0.09 0.13 0.17 0.21 0.25 von Mises Stress in Rubber Ring 209.218 217.170 226.550 233.667 232.391 229.952 230.037

In de x

COSMOSM Advanced Modules

5-115

In de x

5-116

COSMOSM Advanced Modules

Index

A analysis 1-4 approximations 1-7 arc-length 5-82 automatic time stepping 5-23, 524, 5-93, 5-114

displacement constraint 5-31, 5displacement response quantity 5-94, 5-96, 5-100, 5-102,
5-105 33, 5-37, 5-39, 5-42, 5-45, 5-48, 553, 5-57, 5-60, 5-63, 5-65

2-11

H heat transfer 5-1 L limit point 5-82 limits 1-8 linearized buckling 5-79, 5-82, 5local sensitivity 5-86, 5-87, 5-94,
5-105, 5-111 89

B beam height 5-59 beam width 5-58 BEAM2D 5-57, 5-58, 5-59, 5-73, 5BEAM3D 5-60 behavior constraints 1-2, 1-5 bending 5-61, 5-63, 5-69, 5-70 buckling 5-1 buckling response quantity 5-89 C contact 5-24, 5-114 convection 5-35, 5-98 converged 3-11 convergence 1-7, 3-10, 3-11, 3-32 curves 3-20 D design optimization 2-1 design variables 1-2, 1-5, 1-8, 2-2
In de x

E effective strain constraint 5-69 element heat 5-98 F failure index response quantity 5frequency analysis 2-8 frequency constraint 5-14, 5-26, 531, 5-37, 5-39, 5-42, 5-48, 5-73, 574, 5-75, 5-83 frequency response quantity 5-88, 5-94, 5-100, 5-102, 5-105 friction 5-114 112

76

M mass constraint 5-83 modified feasible directions 1-7 Mooney model 5-24, 5-114 multidisciplinary design optimization 3-6 multidisciplinary optimization 526, 5-28, 5-31, 5-33, 5-37, 5-39, 542, 5-45, 5-48, 5-83 multiple load cases 5-13, 5-55

G global sensitivity 2-8, 5-88, 5-89,

5-91, 5-93, 5-96, 5-98, 5-102, 5-108, 5-109, 5-112

N natural frequency 5-1 nonlinear 5-1, 5-23, 5-24, 5-35, 5numerical techniques 1-7
45, 5-82, 5-93, 5-98, 5-114

COSMOSM Advanced Modules

I-1

Index

O objective function 1-2, 1-6, 1-8, 2-5 optimization loops 3-5, 3-6, 3-8, 310, 3-11, 3-29, 3-323-35

singular value decomposition 1-7 sizing 5-1 static 5-1 stress constraint 5-37, 5-39, 5-42, 5stress response quantity 5-86, 5-87,
5-93, 5-100, 5-102, 5-105, 5-114 61

P PLANE2D 5-6 prescribed displacement 5-24, 5114

R restart 1-7 restore 1-7 S sensitivity 1-7 shape optimization 5-1 shape sensitivity 5-1

T temperature response quantity 5trimming factors 3-10 truncation 3-10 types of sensitivity 2-8 W weight factors 3-8
91, 5-96

In de x

I-2

COSMOSM Advanced Modules