关键词: 航空发动机;故障诊断;人工神经元网络;发展趋势

Abstract: Introduced several methodologies for usage of neural networks in fault diagnosis for aero engines. First, qualitative diagnosis for a single fault was discussed. The methodology was verified by experiments on the test cell and demonstrated a relatively high success rate. Followed by a discussion of a measurement preprocessor that can perform sensor data validation and reduce the effect of data deviation. A quantitative diagnosis for multi fault was discussed. In the simulation investigation, all major component performance parameters that reflect deterioration of engine, such as airflow capacities and efficiencies, can be accurately calculated by neural networks. Comparing approaches based on a mathematical thermal dynamics engine model, neural networks as a potential tool can be applied to engine condition monitoring and modular fault isolation.

Key words: Aircraft engine;Fault diagnosis;Artificial neural network;Development trent