支持向量机分类法在异步电机故障诊断中的应用

Application of support vector machine classification in asynchronous motor fault diagnosis

近些年来,支持向量机算法开始应用于电机控制和故障诊断,取得了较好的效果。与以往的此类论文不同,笔者对支持向量机核函数进行了深入的理论分析和性能预测,在此基础上,选择了基于径向基核函数(RadialBasisFunction简称RBF)的支持向量机进行电机故障诊断,并对基于RBF的支持向量机核函数进行了特性分析和参数优化,从理论上证明了采用基于径向基核函数的支持向量机在故障诊断中的优势。此外,针对现有的检测方法所能检测的故障种类单一,不能对几种故障同时检测的弊端,采用阈值设定法和样本补偿法,进行了两种以上故障的神经网络分类研究。实验及仿真结果证实了该方法的有效性。

In recent years, the application of support vector machine(SVM) algorithms to motor control and fault diagnosis achieved good results. Different from the previous studies in this field, this paper makes in-depth theoretical analysis and performance prediction of SVM kernel function. On this basis, a SVM algorithm based on Radial Basis Function(RBF) is selected for the motor fault diagnosis. The characteristic analysis and parameter optimization of SVM kernel function based on RBF present the advantage of this method in fault diagnosis theoretically. In addition, aiming at the disadvantage that the existing detection methods can detect single fault types and can not detect several faults at the same time, the threshold setting method and sample compensation method are used to classify more than two kinds of faults. Experiments and simulation results verify the effectiveness of this method.