A NSGA-II, web-enabled, parallel optimization framework for NLP and MINLP

Engineering design increasingly uses computer simulation models coupled with optimization algorithms to find the best design that meets the customer constraints within a time constrained deadline. The continued application of Moore’s law combined with linear speedups of coarse grained parallelization will allow more designs to be evaluated in shorter periods of time. This paper presents a scalable, standards based framework that uses web services and grid services with a multiple objective genetic algorithm to solve continuous, mixed integer, single objective or multiple objective nonlinear, constrained design problems. Test data is provided to validate a linear speedup based on the number of processors and to show the robustness of the genetic algorithm on a set of 10 design problems. Categories and Subject Descriptors D.3.3 [Programming Languages]: Language Constructs and Features—Frameworks, concurrent programming structures; G.4 [Mathematical Software]: User Interfaces; I.6 [...

• Type: Conference
• Year: 2007
• Where: GECCO
• Authors: David J. Powell, Joel K. Hollingsworth
• Keywords: GECCO 2007, Genetic Algorithm, Linear Speedup, Objective Genetic Algorithm