DEEP-LEARNING ALGORITHM-BASED SELF-ADAPTIVE CORRECTION METHOD FOR FULL-ENVELOPE MODEL OF AERO-ENGINE

摘要

A deep-learning algorithm-based self-adaptive correction method for a full-envelope model of an aero-engine, wherein same falls within the field of the modeling and simulation of aero-engines. A recurrent neural network-based dynamic parallel compensator is used to compensate for an original non-linear model error within a full-envelope range in a state where the performance of an aero-engine is not impaired; and at the same time, a genetic algorithm-based corrector is used to self-adaptively adjust a correction factor of a health parameter to be corrected in an original non-linear component-level model, wherein the health parameter to be corrected is determined by means of an integrated evaluation-based multi-index decision-making algorithm, and the sum of an output of the corrected non-linear component-level model and an output of the compensator is consistent with running trial output data of the aero-engine, thereby improving the full-envelope modeling precision of the aero-engine. By means of the method, strong support is provided for the design of an aero-engine control system and a fault diagnosis system, thereby facilitating the improvement of the reliability of aero-engine hardware in loop and semi-physical experiment verification.