摘要
针对当前应用于航空发动机传感器故障诊断中的基于梯度的传统学习算法多存在参数选择困难、容易陷入局部最小化、过拟合等问题,提出了基于极限学习机(ELM)的航空发动机传感器故障诊断方法。算法只需设置隐含层神经元的个数,能够较好地避免上述问题,缩短故障诊断时间、提升诊断精度。通过仿真试验表明:基于ELM算法所建的航空发动机传感器故障诊断模型要比基于BP神经网络算法所建的模型耗时短且精度高。
Aiming at problems of traditional gradient-based learning algorithm used for aircraft engine sensor fault diagnosis always have currently, such as difficulties with multi-parameter selection, easy to fall into local minimum,over-fitting, and so on, propose fault diagnosis method for aircraft engine sensor based on extreme learning machine(ELM). ELM algorithm can avoid above problems,and further reduce fault diagnostic time and improve diagnostic precision, since it only need to set a parameter,i, e. the number of hidden layer. Simulation test shows that aircraft engine sensor fault diagnosis model based on ELM algorithm has higher precision and shorter time than that based on BP neural network algorithm.
出处
《传感器与微系统》
CSCD
北大核心
2014年第8期23-26,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金委员会和中国民用航空局联合研究基金重点资助项目(U1233201)
天津市科技支撑计划重点资助项目(11ZCKFGX04000)
中央高校基本科研基金资助项目(ZXH2012B002
3122013P005)
关键词
极限学习机
航空发动机
传感器
故障诊断
extreme learning machine (ELM)
aircraft engine
sensor
fault diagnosis