摘要
用 BP网络实现闸门典型故障的诊断。信号依次经历小波除噪、归一化和去趋 3个预处理过程 ,从处理前和处理后的信号中提取能有效描述信号特征的特征量作为网络输入。同时 ,为 4种典型信号设立了相对应的网络输出 ,形成输入 -输出样本对。网络用这些样本对训练自身。网络性能测试表明 ,它能有效识别典型故障 ,而没有训练过的模式 。
Backpropagation is used to recognize the typical vibrating patterns of hydraulic gate. To improve the performance of network, some preprocessing measures are applied to the signal, such as using wavelet to denoise the signal, then scaling it to fall within a specified range [-1,1] and removing the signals trend. The selection of network's inputs and outputs also influences its performance. Some indexes are used to indicate the properties and the inputs of the network. The trained network can recognize not only the typical vibrating patterns, but also an unknown one.
出处
《解放军理工大学学报(自然科学版)》
EI
2004年第4期90-94,共5页
Journal of PLA University of Science and Technology(Natural Science Edition)
关键词
闸门
BP网络
故障诊断
小波除噪
预处理
hydraulic gate
Backpropagation
fault diagnosis
wavelet denoise
preprocessing