基于BP神经网络的航空发动机故障检测技术研究  

Research on Aeroengine Fault Detection Technology Based on BP Neural Network

作　　者：殷锴 钟诗胜[2] 那媛 李臻[2] 

机构地区：[1]中国航发商用航空发动机有限责任公司,上海200241 [2]哈尔滨工业大学机电工程学院,哈尔滨150001

出　　处：《航空发动机》2017年第1期53-57,共5页Aeroengine

基　　金：国家自然基金重点项目(U1533202);民航科技项目(MHRD20150104);山东省自主创新及成果转化专项(2014CGZH1101)资助

摘　　要：为了提高航空发动机故障检测正确率,将BP神经网络应用于航空发动机故障检测中。从某航空公司使用的CFM56-7B系列发动机的实际飞行历史数据中选取研究样本,对比了6种训练方法的效果并最终选择弹性BP法对网络加以训练并进行测试。结果表明:该方法对CFM56-7B系列发动机的排气温度指示故障、进口总温指示故障和可调放气活门故障的检测正确率高达83.33%。BP神经网络能够很好地应用于航空发动机的实际故障检测,其学习记忆稳定、网络收敛速度快,具有一定的工程实用价值。The BP neural network was applied to fault detection of aeroengine to improve the accuracy rate. The research samples were selected from the historical data of the CFM56-7B series engine provided by an aviation company, and then choosing the elastic BP training algorithm by comparing the performance of six training methods to test. The results show that the accuracy rate of fault detection for the EGT_F, TAT_F and VBV_F is high as up to 83.33%. BP neural network can be well applied to the practical fault detection of aeroengine on account of its fast convergence speed and the stability of learning and memory, which guarantee practical engineering value.

关 键 词：故障检测 BP神经网络 航空发动机 

分 类 号：V233.7[航空宇航科学与技术—航空宇航推进理论与工程]