A Novel Hybrid Aeroengine Modeling Method for Combining Data-Driven Modules

Abstract

Aeroengine models are widely used to identify aerodynamic parameters of components and play an important role in many applications. Due to differences in gas path characteristics, traditional physics-based models do not typically match actual engines. This paper proposes a new hybrid modeling framework that combines data-driven modules. The proposed method adds a steady-state correction module and a dynamic compensation module on the basis of the physics-based model, reducing the differences in gas path characteristics through two stages of modeling, and obtaining the final hybrid model. The steady-state correction module uses a particle swarm optimization (PSO) algorithm, and the dynamic compensation module uses a long short-term memory (LSTM) neural network. The modeling method is validated using actual data from different engine individuals, and the proposed method demonstrates better performance than traditional methods.