Aero-engine Performance Optimization Using MOEA with Selective Ensemble Learning

Aero-engine performance optimization is crucial for pursuing reliability and security during the operation of an aero-engine. However, optimization of aero-engine performance is a multiobjective, computationally expensive programming problem. Ordinarily, such a problem is assumed that analytic expressions of the objective functions are available. However, only historical data could be obtained in practice, which is impossible to apply in the existing optimization algorithm. Therefore, a data-driven multiobjective evolutionary algorithm is proposed to address these difficulties, which employs NSGA-II as the fundamental element assisted by selective ensemble learning, model management strategy, and a final solution set generation mechanism. The numerical results demonstrate that the proposed algorithm can approximate the Pareto front of the aero-engine performance optimization problem.