12 gbm code optimum design

Upload: malika0931

Post on 06-Apr-2018

225 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/3/2019 12 GBM Code Optimum Design

    1/25

    11

    Center for Advanced eCenter for Advanced e--System Integration TechnologySystem Integration Technology

    Introduction to Optimum Design Programand ApplicationsIntroduction to Optimum Design Programand Applications

    [Contents]

    Selection of Good Optimization Algorithm

    DOT/VisualDOCOther Optimum Design Codes

    Optimum Design Examples using Optimum

    Design program

  • 8/3/2019 12 GBM Code Optimum Design

    2/25

    Center for Advanced eCenter for Advanced e--System Integration TechnologySystem Integration Technology

    1. Selection of Good Optimization Algorithm. Selection of Good Optimization Algorithm

  • 8/3/2019 12 GBM Code Optimum Design

    3/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    Attributes of a Good Optimization AlgorithmAttributes of a Good Optimization Algorithm

    Aspects need to be considered Robustness, efficiency, generality, ease to use

    Reliability or Robustness

    Reliabil ity of an algorithm is guaranteed if it is theoretically proven to

    converge

    Guarantees same optim ization results starting from different initial estimates.

    Has higher prior ity over efficiency

    Generality

    The algorithm should be able to treat equality as well as inequality constraints.

    Ease of use

    Ease to use by the experienced as well as inexperienced designer

    Efficiency Faster rate of convergence to the minimum point

    Least number of calculations w ithin one design iteration

    Efficiency w ithin an iteration implies the minimum of calculations for the

    search direction and the step size

  • 8/3/2019 12 GBM Code Optimum Design

    4/25Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    QuestionsBefore Selecting an Optimization Algorithm

    QuestionsBefore Selecting an Optimization Algorithm

    Does the algorithms have proof of convergence? Is it theoretically guaranteed to converge to an optimum point starting

    from any initial design estimate?

    Can the starting design be infeasible?

    Can the algorithm solve a general optimization problem without any

    restrictions on the constraint functions?

    Can it treat equality as well as inequality constraints?

    Is the algorithm ease to use?

    Does it require tuning for each problem?

    Does the algorithm incorporate a potential constraint strategy?

    Does it provide well defined GUI?

  • 8/3/2019 12 GBM Code Optimum Design

    5/25

    Center for Advanced eCenter for Advanced e--System Integration TechnologySystem Integration Technology

    2. DOT/VisualDOC. DOT/VisualDOC

  • 8/3/2019 12 GBM Code Optimum Design

    6/25Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    DOT (Design Optimization Tools)DOT (Design Optimization Tools)

    Description

    A general purpose numerical optim ization software (Fortran-based

    program) package created by Vanderplaats Research &

    Development(VR&D)

    DOT can be used to solve a w ide variety of nonlinear optimization

    problems.

    The user provides a main program for calling DOT, and an analysis

    program to evaluate the necessary function. Algorithms

    For constrained optimization :

    Modified Method of Feasible Direction (MMFD) : option 1

    Sequential Linear Programming(SLP) with adjustable move limits : option 2

    Sequential Quadratic Programming (SQP) : option 3

    For Unconstrained optimization :

    Broydon-Fletcher-Goldfarb-Shanno (BFGS) algorithm : option 1

    Fletcher-Reeves (FR) Algorithm : option 2

  • 8/3/2019 12 GBM Code Optimum Design

    7/25

  • 8/3/2019 12 GBM Code Optimum Design

    8/25Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    VisualDOC : DescriptionVisualDOC : Description

    VisualDOC is a software system created by VR&D

    It simplifies the addition of optimization to almost any design task.

    VisualDOC is uses a graphical user interface(GUI) along with optimization

    algorithms to setup and solve design problem

    VisualDOCs optimization library is based on the general purpose optimization

    software Design Optimization Control(DOC) and Design Optimization Tools(DOT).

    Option of VisualDOC

    Direct optimization by the VisualDOC optimizer

    Optimization through the use of response surface techniques

    (approximation technique) .

    To calculate responses user can either supply a computer routine written in C/C++

    or Fortran, or user can couple an existing program with VisualDOC. VisualDOC solves design problems by iteratively calling the optimizer to modify

    the design variables and then calculating the resulting responses.

    VisualDOC provides API and third party interface.

    More options and capabilities included in VisualDOC 2.0

  • 8/3/2019 12 GBM Code Optimum Design

    9/25Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    VisualDOC : OptimizerVisualDOC : Optimizer

    For constrained optimization :

    Modified Method of Feasible Direction (MMFD)

    Sequential Linear Programming(SLP) w ith adjustable move l imits

    Sequential Quadratic Programming(SQP)

    For Unconstrained optimization :

    Broydon-Fletcher-Goldfarb-Shanno (BFGS) algorithm

    Fletcher-Reeves (FR) Algorithm

    For problems of fewer than 10 design variables, user can use response

    surfaces (approximation technique).

    There are no implicit limits on problem size.

    Design variables may be continuous, discrete, or any combination. Linear or nonlinear design variable linking is supported.

    An arbitrary number of constraints may be specified, and multiple

    optimization is supported.

  • 8/3/2019 12 GBM Code Optimum Design

    10/25Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    VisualDOC : ComponentsVisualDOC : Components

  • 8/3/2019 12 GBM Code Optimum Design

    11/25Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    VisualDOC : Graphical User InterfaceVisualDOC : Graphical User Interface

    Using the graphical user interface (GUI), user define a catalog ofvariables and responses.

    The graphical user interface lets user enter and modify all optimization

    components of a design problem.

    User can supply initial values and limits for the variables.

    User can also supply limits on some responses, which are called

    constraints, and specify one or more responses as design objectives.

    The design objectives can be minimized, maximized, or driven to target

    values by a simple click of the mouse.

    Different windows organize users design data, providing a clear,

    concise description of users design problems.

  • 8/3/2019 12 GBM Code Optimum Design

    12/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    GUI Example(Windows) in VisualDOCGUI Example(Windows) in VisualDOC

  • 8/3/2019 12 GBM Code Optimum Design

    13/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    How VisualDOC Interface Works?How VisualDOC Interface Works?

    The primary function of the VisualDOCinterface is to set up the optimization problem.

    The data is passed between these modules

    using a global data structure

    Optimizer adjusts the values of theindependent design variables.

    The modified design variable values are sent

    back to the analysis routine.

    The analysis code returns the direct response

    values to the optimizer.

    Values of objective and constraint functions

    are calculated.

    The objective and constraint values are

    passed back to the optimizer, which then

    adjusts the design variables once again.

    This process continues in the same fashion

    until optimum has been found.

    VisualDOCGraphical User Interface

    VisualDOC

    Optimizer

    AnalysisRoutine

  • 8/3/2019 12 GBM Code Optimum Design

    14/25

    Center for Advanced eCenter for Advanced e--System Integration TechnologySystem Integration Technology

    3. Other Optimum Design Codes. Other Optimum Design Codes

    GBM d liGBM d li t

  • 8/3/2019 12 GBM Code Optimum Design

    15/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    GBM code listsGBM code lists

    Several GBM codes for general design optimization purpose CFSQP - nonl inear and minmax optimizat ion.

    CONOPT - nonlinear programming.DOT - Design Optimization Tools.

    FSQP - nonl inear and minmax optimizat ion.GINO - nonl inear programming.GRG2 - nonl inear programming.LANCELOT - large-scale problems.LSGRG2 - nonlinear programming.

    MINOS - l inear programming and nonlinear optimization.NLPQL - nonl inear programming.NLPQLB - nonlinear programming with constraints.NLPSPR - nonl inear programming.

    NPSOL - nonl inear programming.OPSYC - OPt imisation de SYstmes Creux.OPTIMA Library - optimization and sensitivity analysis.OPTPACK- constrained and unconstrained optimization.SQP - nonl inear programming.

    MINOS 5 1MINOS 5 1

  • 8/3/2019 12 GBM Code Optimum Design

    16/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    MINOS 5.1MINOS 5.1

    MINOS is a Fortran-based computer system designed to solve large-

    scale optimization problems.

    Algorithms

    Linear problem : the primal Simplex method

    Nonlinear objective : a linearly constrained nonlinear program.

    a reduced gradient algorithm in conjunction with a quasi-Newton algorithm

    (BFGS)

    Nonlinear Constraints :

    a projected augmented Lagrangian algorithm

    MINOS is designed to find solutions that are locally optimum.

    No GUI and APIs are available (Consol Application)

  • 8/3/2019 12 GBM Code Optimum Design

    17/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    GENESIS-Structural Optimization SoftwareENESIS-Structural Optimization Software

    Genesis GENESIS is a fully integrated structural analysis and design optimization

    software. Analysis is based on the finite element analysis (FEA) for static,

    normal modes, direct and modal analysis, heat transfer calculations, and

    buckling analysis.

    Design optimization is based on the advanced approximation concepts

    approach to find an optimum design efficiently and reliably. Actual

    optimization is performed by the well established DOT and BIGDOT

    (excess of 2 mill ion design variables) optim izers, also from VR&D.

    GENESIS performs both sizing (member dimension) and shape

    (geometry) optimization. Typical problem sizes involve well over 100

    design variables and many thousands of constraints. Almost any

    calculated response (or nonlinear function of variables and responses)

    can be chosen as the objective or may be constrained. These include mass,volume, eigenvalues, stresses, and deflections.

    Pre- and post-processing is done by the PDA/ PATRAN and SDRC/ IDEAS

    programs.

  • 8/3/2019 12 GBM Code Optimum Design

    18/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    Excel Solverxcel Solver

    Excel Solver The Solver option in EXCEL 2000 (and earlier versions) may be

    used to solve linear and nonlinear optimization problems.

    Integer restrictions may be placed on the decision variables.

    Solver may be used to solve problems w ith up to 200 decision

    variables (design variables), 100 explicit constraints and 400

    simple constraints (lower and upper bounds and/ or integer

    restrictions on the decision variables).

    Algorithms

    Newtons method

    Conjugate gradient method

  • 8/3/2019 12 GBM Code Optimum Design

    19/25

    Center for Advanced eCenter for Advanced e--System Integration TechnologySystem Integration Technology

    4. Optimum Design Examples usingOptimum Design program4. Optimum Design Examples usingOptimum Design program

    Space Launch Vehicle DesignSpace Launch Vehicle Design

  • 8/3/2019 12 GBM Code Optimum Design

    20/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    Space Launch Vehicle DesignSpace Launch Vehicle Design

    Design problem

    Minimize CD : Drag Coefficient

    Subject to Ch a : Surface heat t ransfer rateNose Pairing Volume bFineness ratio = c, where a, b, c are constant

    Number of design variables : Xi, i= 1, 4

    Linearization of objective function and design constraints

    Sensitivity Analysis

    Update the design

    44

    43

    3

    32

    3

    21

    1

    1 XX

    CXX

    CXX

    CXX

    CCCDDDD

    nomialDD

    ++

    ++

    1

    432143211

    1

    1 ),,,(),,,(

    X

    XXXXCXXXXXC

    X

    CDDD

    =

    XXXnomialnew

    44

    43

    3

    32

    3

    21

    1

    1 XX

    CX

    X

    CX

    X

    CX

    X

    CCC

    hhhh

    nomialhh

    ++

    ++

    Updated Design variables

  • 8/3/2019 12 GBM Code Optimum Design

    21/25

    DOT Main program (main for)DOT Main program (main for)

  • 8/3/2019 12 GBM Code Optimum Design

    22/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    DOT Main program (main.for)DOT Main program (main.for)

    DIMENSION X(100),XL(100),XU(100),G(100),

    *WK(800),IWK(200),RPRM(20),IPRM(20)

    C DEFINE NRWK, NRIWK.

    NRWK=800

    NRIWK=200

    C ZERO RPRM AND IPRM.DO 10 I=1,20

    RPRM(I)=0.0

    10 IPRM(I)=0

    c

    RPRM(1)=-0.0005

    RPRM(2)=0.00005

    C DEFINE METHOD,NDV,NCON.METHOD=2

    NDV=4

    NCON=2

    C DEFINE BOUNDS AND INITIALDESIGN .DO 20 I=1,NDV

    X(I) = 0.0

    XL(I)=-0.2

    20 XU(I)= 0.2

    100 CALL DOT (INFO,METHOD,IPRINT,NDV,NCON,X,XL,XU,

    *OBJ,MINMAX,G,RPRM,IPRM,WK,NRWK,IWK,NRIWK)

    C FINISHED?

    IF(INFO.EQ.0) STOP

    C EVALUATE OBJECTIVE AND CONSTRAINT.

    CALL EVAL(OBJ,X,G)

    C GO CONTINUE WITH OPTIMIZATION.

    GO TO 100

    END

    c-----OBJECTIVE & CONSTRAINTS------------------------------------------

    SUBROUTINE EVAL (OBJ,X,G)

    implicit real*8 (a-h, o-z)

    IMENSION X(*),G(*), d(5), h(5),aold(4)

    OBJ = d1 + d2* X(1) + d3* X(2) + d4* X(3) + d5* X(4)

    G(1)=h1 + h2* X(1) + h3*X(2) + h4* X(3) + h5*X(4)-

    0.012453

    G(2)= (-1.)* vol + 0.198919

    RETURN

    END

    Optimization method : SLP

    No. of design variables : 4

    No. of constraints : 2

    OBJ:Objectivefunction

    G(1),G(2): Constraintfunction

    X(i):DesignvariableSide Constraints

    2-Bar Truss Design : Structural Optimization2-Bar Truss Design : Structural Optimization

  • 8/3/2019 12 GBM Code Optimum Design

    23/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    2 Bar Truss Design : Structural Optimization2 Bar Truss Design : Structural Optimization

    Description The symm etric 2-bar truss design shown in below has been studied by several

    researchers

    Ball ing and Clark(1992), Schmit, (1981), Sobieszczanski-Sobieski et al(1982)

    The objective of this optimum problem is to min imize the weight of truss systemsubject to behavioral constraints

    Related parameters

    - B = 30 in

    - t = 0.1 in

    - = 0.3 lbs/ in3- y = 60,000 psi- E = 30E6 ps i

    2-Bar Truss Design : Structural Optimization2-Bar Truss Design : Structural Optimization

  • 8/3/2019 12 GBM Code Optimum Design

    24/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    2 Bar Truss Design : Structural Optimization2 Bar Truss Design : Structural Optimization

    Formulation for Optimization Minimize W(X)=2Dt(B2+H2)1/ 2

    Minimize the weight of truss system

    Subject to

    g1(X) = e - 0 the first constraint prevents failure due to Euler buckling

    g2(X) = y - 0 the second constraint prevents failure due to yield stress

    Where,0.5D5.0 (in) X(1) : mean tube diameter5.0H50.0 (in) X(2) : height of the truss

    The resulting optimum value from (Schmit, 1981) for W(x) is 19.8lbs.

    - W * = 19 .8 lbs (at D* = 2.47 in , H* = 30.15 in )

    2-Bar Truss Design : Structural Optimization2-Bar Truss Design : Structural Optimization

  • 8/3/2019 12 GBM Code Optimum Design

    25/25

    Center for Advanced eCenter for Advanced e--System Integration Technology,System Integration Technology,KonkukKonkuk Univ.Univ.

    2 Bar Truss Design : Structural Optimization2 Bar Truss Design : Structural Optimization

    Comparisons with Algorithm and initial Valueinitial values X(1) =0. , X(2)= 0.

    DOTProgram

    Constrained Unconstrained

    Algorithm MMFD SLP SQP BFGS F-R -X1 (D) 2.481 2.476 2.476 4.558 4.558 2.47

    X2 (H) 29.870 29.992 30.00 15.031 15.031 30.15

    Opt. val. (W) 19.800 19.800 19.800 28.828 28.828 19.8

    Num. of Iter. 9 20 11 3 3 -

    Num. of fun. eval. 77 69 45 24 24 -

    true

    initial values X(1) =1. , X(2)= 5.

    -2626406626Num. of fun. eval.

    -3310205Num. of Iter.

    19.823.60123.60119.80019.80022.690Opt. val. (W)

    30.1517.82017.82030.04129.99017.588X2 (H)

    2.473.5883.5882.4742.4803.462X1 (D)

    -F-RBFGSSQPSLPMMFDAlgorithm

    UnconstrainedConstrainedtrue

    DOTProgram