SOM神经网络对永磁同步电机的故障诊断

Fault Diagnosis of Permanent Magnet Synchronous Motor Based on SOM Neural Network

为实现永磁同步电机的故障类别的诊断,采用小波函数根据不同频段进行故障特征提取,对提取样本数据进行归一化处理,对奇异样本。SOM(Self Organizing Map)的领域函数选取小波函数,形成次兴奋神经元不断对权值更新,避免SOM的产生局部最优解。采用实验提取的故障数据作为SOM神经网络的输入样本进行网络训练,从而得出产生特定故障时所激发的相应神经元索引。实验结果验证了该方法的可行性和实用性。

In order to diagnose the faults of PMSM,the wavelet function is used to extract the fault features ac-cording to different frequency bands, and the normalized data samples are processed to eliminate the singular sam-ples. The domain functions of SOM （Self, Organizing, Map） are constructed by using wavelet function, and the weights of the sub excited neurons are updated to avoid the local optimum of SOM. The experimental data is trained as the input sample of SOM neural network, and the correspondin lg neuron index is generated when the fault is generated. The experimental results verify the feasibility and practicability of the method.