An Effective Mixed Annealing/Heuristic Algorithm for Problems in Mechanical Design

Abstract

The wide application of stochastic optimization methods in mechanical design has been partially hindered due to (a) the relatively long computation time required, and (b) discretization of the design space at the onset of the optimization process. This work proposes a new stochastic algorithm, the Mixed Annealing/Heuristic Algorithm (MAH), which addresses both these issues. It is based on the Simulated Annealing algorithm (SA) and the Heuristic Optimization Technique (HOT). Both these algorithms have been successfully applied to problems in mechanical design and up to now have been considered as competing algorithms. MAH capitalizes on each of their individual strengths and addresses their weaknesses, thereby considerably reducing the computational effort required to attain the final solution. A pseudo-continuous approach for configuration generation is employed, making the discretization of the design space no longer necessary. The effectiveness of MAH is demonstrated via three problems in kinematic synthesis. Comparison of the results with other stochastic optimization methods illustrates the potential of this technique.