Aeroengine Performance Prediction Based on Double-Extremum Learning Particle Swarm Optimization

摘要

To predict the performance parameter changing trend of an aeroengine, a novel double-extremum learning particle swarm optimization (DELPSO) algorithm is proposed. Inspired by human learning behavior, this algorithm simulates this behavior, including the strategies of collective learning, private tutoring, and research behavior, so that obtained final solutions would be in the global optimal area or its neighbor area as close as possible. Meanwhile, to improve the prediction performance, a nonlinear mapping function is designed to describe the feature relationship between inputs and outputs of historical data. Based on the DELPSO, the fitness function synthetically considers the changing trend and the prediction error and can adaptively select optimal parameters of the nonlinear mapping function. The experimental results demonstrate that the DELPSO has globally stable and reliable performance. To validate the prediction performance of the proposed DELPSO, it is also applied to an aeroengine. Its good prediction performance indicates that the proposed DELPSO is an important reference for maintenance decision-making of aeroengines.