改进型小波神经网络在变频系统故障诊断中的应用

Application of Improved Wavelet Neural Network in Fault Diagnosis in Frequency Conversion System

目前,变频系统故障诊断还一直依赖于对电压和电流信号的监测与分析,现提出用改进小波包提取变频器故障信号的特征量,并将该特征向量作为改进型网络的学习样本进行训练的方法。仿真结果表明,改进型小波神经网络与常规神经网络诊断方法相比,具有准确度高、诊断速度快和泛化能力强等优点。

Currently, fault diagnosis for the frequency conversion system depends on the monitoring and analyzing of voltage and electric current signals. This paper presented a method that the eigenvectors of the fault signals of the frequency converser were extracted using the improved wavelet packet, and then the eigenvectors were used as learning samples for the training. The diagnosis results show that in comparison with the conventional neural network diagnosis method, the improved wavelet neural network diagnosis method has the advantages of high accuracy, fast diagnosis and good generalization ability.