一种新的神经网络故障诊断方法

Neural network in fault diagnosis based on data fusion

基于数据融合思想 ,提出一种新的神经网络故障诊断方法。利用系统故障征兆的分散性和复杂性 ,采用多个神经网络分别对每一类故障进行诊断 ,网络输入为与输出故障相关联的监测信号的特征值 ,将各网络输出进行融合 ,给出最后诊断结果。将该方法应用于斜轴式无铰柱塞液压泵故障诊断 ,结果表明能够充分利用各种特征信息 ,提高诊断速度和精确度。

Basing on the thought of data fusion, we propose a new approach of neural network fault diagnosis. Using the decentralization and complexity of system fault symptom, we diagnose every system fault by multiple network whose inputs are eigenvalue related to measuring signals of output fault, then fuse every output of every neural network and get the results. At last, we apply the approach to hydraulic pump fault diagnosis, results indicate that the approach can take full advantage of diversified characterisitic information, and improves diagnosis rate.