Log In    |    Join IASS    |    Online Journal    |    Proceedings    |    Directory    |    Events    |    News

Online IASS Proceedings

IASS-SLTE Symposium 2014: Shells, Membranes and Spatial Structures: Footprints


IASS Symposium 2014

SESSION: Computational Methods

Using database storage to improve explorative optimization of form critical structures

< Table of Contents for Computational Methods
  • Proceedings Name: IASS-SLTE Symposium 2014: Shells, Membranes and Spatial Structures: Footprints
  • ISSN: (Electronic Version) 2518-6582
  • Session: Computational Methods
  • Title: Using database storage to improve explorative optimization of form critical structures
  • Author(s): Peter VON BUELOW
  • Keywords: conceptual design, morphology, form finding, optimization, exploration, generative, performance.
Abstract
Heuristic methods, including genetic algorithms (GAs) have been used and improved for many years in the design and optimizations of complex structural systems. In recent years an additional modification has been developed at the University of Michigan which offers several advantages over traditional GA implementations. This paper outlines specific innovations to GA based design tools, and describes a method which includes both multi-objective optimization as well as more open exploration, with particular application in the early phases of a design project. The ParaGen, method has been presented previously in several articles. This paper looks in more detail at the advantages offered through a non-destructive dynamic population GA (NDDP GA) coupled with a relational database. These two components are combined with associative parametric software for form generation, and simulation and analysis software for performance evaluation in what we call the ParaGen cycle. By storing all generated solutions in a database, solutions can not only be retrieved on demand, but can be retrieved in a programed way based on structured queries. ParaGen formulates fitness or objective functions using a structured query language (SQL). SQL queries can describe very specific sets of data which can be extracted from the entire data set. In the ParaGen method, breeding populations are dynamically created using SQL sorts and queries at the moment each new child is formulated. Unlike traditional populations, which are generated based only on the preceding population, dynamic populations in ParaGen are assembled from the entire range of solutions evaluated up to that point in the process. Since the RDBMS can easily store all solutions, there is no need to ever drop a solution from the potential range of selections. So the NDDP GA is able to make breeding selections from an ever increasing base of potential parents. This paper describes in detail a list of advantages made possible by coupling an NDDP GA with a relational database. These include the ability to: * Store all solutions without duplicates;* Use multi-objective fitness functions with SQL queries;* Create dynamic parent populations;* Change search direction instantaneously;* Explore solution space with interactive search;* Graph parameters to determine Pareto trade-off sets;* Utilize parallel hardware for efficient computation. These points are particularly useful in determining good solutions early in the design process when there is still both the freedom to significantly alter the design direction and the potential to benefit most from the improved path.

Read the complete paper (pdf) (requires membership)
© IASS All rights reserved.
IASS · CEDEX - Laboratorio Central de Estructuras y Materiales · C/ Alfonso XII, 3 · 28014 Madrid, Spain     tel. 34 91 335-7409; fax: 335-7422
Site Design & Programming
by World Design Group, Inc.