Research on Aero-engine Fault Diagnosis Based on Meta- learning

Abstract

Fault diagnosis is a key technology of aero-engine health management,and it has great significance for aircraft safety.Aiming at the problem of poor generalization of fault diagnosis model, a combined classification algorithm model based on meta-learning is proposed.The learning results of multiple models are learned again to reduce the tendency caused by inductive bias in single algorithm model. Then, the ground idle speed state data of a turboshaft engine test bed is used for simulation verification. The results show that compared with the single algorithm model, the combination classification model has better generalization ability and accuracy. And an improved particle swarm optimization algorithm is studied in this paper. Aiming at the problem of selecting the maximum particle velocity, the maximum velocity coefficient is defined, and the maximum speed is related to the search range. Through analysis and verification, the conclusion that the improved algorithm has faster convergence speed in the early stage and higher search accuracy in the later stage is obtained.