Sequential ensemble multi-objective optimization algorithm and its application on the multi-working performance optimization of the variable cycle engine

Abstract

The numerical design optimization is becoming more and more popular as the rapid development of computer technology. The efficient global optimization (EGO) algorithm is one of the most popular optimization algorithm used in the investigation of aerospace engineering systems in recent years. Many researches optimize several conflicting performance objectives by using the improved version of EGO in recent years. The EGO algorithm is a Kriging-assisted optimization algorithm. Wherein, the Kriging surrogate model is used to predict the performance objectives. Thus the accuracy of the Kriging surrogate model affects the EGO algorithm performance. The multi-surrogate technology can build an ensemble surrogate model by combining different individual surrogates. The ensemble model shows better accuracy than the other individual surrogate model. Thus, this paper proposes a sequential ensemble multi-objective optimization (SEMO). In the proposed algorithm, the ensemble model is used to generate the new sample points during the optimization process. Also, a new Pareto conserve strategy is proposed to enhance the accuracy of the optimal solution set. To verify the performance of SEMO, four well-known multi-objective optimization mathematical functions is considered in this paper. Further, SEMO algorithm is applied to the multi-working performance optimization of the variable cycle engine.