航空发动机气路改进神经网络故障诊断研究  

Research on the Fault Diagnostic Based on Improved Neural Network for Gas Path of Aeroengine

作　　者：卢俊文[1] 吴瑞[1] 常虎山[1] 王威风 尚泽译 

机构地区：[1]中国民用航空飞行学院,四川广汉618307

出　　处：《自动化仪表》2015年第1期5-8,12,共5页Process Automation Instrumentation

基　　金：中国民用航空飞行学院面上基金资助项目(编号:J2012-06);中国民用航空飞行学院研究生创新基金资助项目(编号:X2012-10)

摘　　要：针对航空发动机气路故障,建立了基于改进BP神经网络的航空发动机气路故障诊断模型。采用该模型寻找发动机状态参数与不同故障模式之间的映射关系,并利用收集的JT9D发动机的气路故障样本数据对诊断模型进行训练与检验。试验结果表明,改进的BP神经网络故障诊断模型对JT9D发动机气路部件故障模式的识别具有较高的准确率,能够为基于状态的维修决策提供有效的指导与建议,进而提高航空发动机的可靠性。In accordance with the faults of gas path of aeroengine, the fault diagnostic model based on improved neural network is built up for gas path of aeroengine, to find out the mapping relation between condition parameters of engine and different fault modes. Then the diagnostic model is trained and tested by adopting the sample data of gas path collected from JT9D engine. The experimental results indicate that the fault diagnostic model based on improved neural network possesses higher accuracy for recognizing the fault mode of parts in gas path ; it can provide effective guidance and advice for maintenance decision based on the status, thus the reliability of the aeroengine can be enhanced.

关 键 词：航空发动机 气路故障诊断 BP神经网络 BP算法 LEVENBERG-MARQUARDT算法 

分 类 号：V263.6[航空宇航科学与技术—航空宇航制造工程]