电机故障模式识别与诊断

Pattern recognition and diagnosis of faults in electrical machines

为了实现对电机故障模式的自动识别与诊断，通过对人工神经网络分类功能及传统电机故障诊断技术的分析，提出了一种利用人工神经网络进行模式识别的方法。针对电机故障特征在实际中可能是非线性可分的情况，利用φ函数可以将非线性可分的模式转化到线性空间并实现分类，基于这种思想提出了利用径向基函数（ＲＢＦ）网络实现对这种复杂故障模式的分类。以感应电机转子故障分类的实验结果表明，这种神经网络模式识别方法是有效的，并且可以通过自动调节径向基函数中心提高网络分类的正确率。

In order to automatically recognize and diagnose the fault patterns of electrical machines, a pattern recogition method by using an artificial neural network is developed based on the analysis of both the classification characteristics of artificial neural networks and the traditional technology of fault diagnosis of electrical machines. For cases that the fault patterns are nonlinear separable, a radial basis function (RBF) network is adopted as by using a φ function a nonlinear separable pattern can be transformed into a linear one, and achieve classification. Test results on the rotor faults of induction motor show that this fault pattern recognition method by using the RBF network can not only be effective, but also improve the average probability of correct classification by a supervised selection of centers of the RBF network.