A New Algorithm for Combinatorial Optimization: Application to Tolerance Synthesis with Optimum Process Selection

Abstract

This paper describes a new algorithm for combinatorial search optimization. The method is general and allows the designer to optimize in more than one domain. It represents an integration of classical optimization and experimental design techniques. The combinatorial problems are approximated by planning the search within the algorithm. The proposed algorithm assumes the availability of cost-tolerance data for various alternative manufacturing processes and a stackup-tolerance model. The method does not depend on the form of objective function and/or contraints (linear vs. nonlinear) as it does not require any functional derivatives. A formulation is presented for modelling two search domains; as an application. The algorithm is used to deal with the problem of least cost tolerance allocation with optimum process selection. The algorithm, which can be classified as a heuristic technique, is tested for 11 example problems with excellent results compared with both local and global methods. A designer can obtain efficient solutions for discrete and multi-search domain problems using this optimization tool.