Improvement Multidisciplinary Collaborate Optimization based on Simu- lated Annealing and Artificial Neural Networks

Multidisciplinary Design Optimization (MDO) is the most active fields in the current complex engineering system design. Forcing to the defects of traditional Collaborative Optimization, such as unable to convergence or falling into local optimum, we propose a Collaborate Optimization based on Simulated Annealing and Artificial Neural Networks, (SA-ANN-CO). The SA-ANN-CO algorithm inherit the parallel distribution strategy of tradition Collaborative Optimization, and then establish accuracy Artificial Neural Networks models by Latin Hypercube Experimental design replace the realistic model of sub-disciplines to reduce computing costs and smooth numerical noise. The possibility of falling into local solutions is reduced by using the Simulated Annealing algorithm in system-level. Two classic test examples results show that, SA-ANN-CO algorithm has good robustness and can quickly and effectively converge to the global optimum solution, which provides a effective way for complex engineering systems design.