Parallel Simulated Annealing Using Simplex Method
Ya-Zhong Luo, Guo-Jin Tang · AIAA Journal · 2006
By combining the advantages of Simulated Annealing with that of Simplex Method, a new kind of parallel Simulated Annealing is proposed in this paper. The details of the parallel Simulated Annealing are described. Experiment results are presented for several mathematical example problems and a 10-bar truss structural optimization. Just using a very simple penalty function, the parallel Simulated Annealing obtains the best-know solutions for two nonlinear constrained optimization problems including Golinski’s speed reducer problem, besides its computational cost is the lowest. The results demonstrate that our algorithm is effective and efficient is solving both functional and structural optimization problems, and it is superior to genetic algorithm, evolutionary programming, chaos optimization algorithm, and sequential simulated annealing, etc. The theoretic analysis on the speedup of the parallel Simulated Annealing is made, which shows that speedup is much larger with the more complex optimization problem. The master/client parallel computation is implemented using MPI, and the experiment testifies the theoretic result.