Framework for sequential approximate optimization

J.H. Jacobs, L.F.P. Etman, F. Keulen, van, J.E. Rooda

    Research output: Contribution to journalArticleAcademicpeer-review

    36 Citations (Scopus)
    1 Downloads (Pure)


    An object-oriented framework for Sequential Approximate Optimization (SAO) isproposed. The framework aims to provide an open environment for thespecification and implementation of SAO strategies. The framework is based onthe Python programming language and contains a toolbox of Python classes,methods, and interfaces to external software. The framework distinguishesmodules related to the optimization problem, the SAO sequence, and thenumerical routines used in the SAO approach. The problem-related modulesspecify the optimization problem, including the simulation model for theevaluation of the objective function and constraints. The sequence-relatedmodules specify the sequence of steps in the SAO approach. The routine-relatedmodules represent numerical routines used by the SAO approach as 'black-box'functions with predefined input and output, e.g. from external softwarelibraries. The routines carry out specific computational tasks, such as thedetermination of a design of experiments, a linear regression analysis, or thesolution of a nonlinear programming problem. The framework enables the user to(re-) specify or extend the SAO dependent modules, which is generally notpossible in most available SAO implementations. This is highly advantageoussince many SAO approaches are application-domain specific due to the type ofapproximation functions used. The paper describes the layout and data classesof the developed framework. The ten-bar truss design problem with fixed loadsas well as uncertain loads is used as an illustration and demonstrates theflexibility of the framework.
    Original languageEnglish
    Pages (from-to)384-400
    JournalStructural and Multidisciplinary Optimization
    Issue number5
    Publication statusPublished - 2004


    Dive into the research topics of 'Framework for sequential approximate optimization'. Together they form a unique fingerprint.

    Cite this