基于三分类支持向量机的船用逆变器故障诊断

针对电力推进船舶逆变器存在的开关器件开路故障诊断问题，提出一种基于三分类支持向量机的故障诊断方法。利用对称分量分析方法获得逆变器输出正序瞬时值分量，通过对信号进行小波包分解，得到不同开关元件故障下的小波能量，规范化后作为对应开关器件故障特征。根据开关器件位置和逆变器输出波形特点对开关器件进行分组，利用三分类支持向量机实现故障分类。仿真分析结果表明，该三分类支持向量机故障分类正确率94.29%，诊断方法有效。

Tri-Class Support Vector Machines Based Fault Diagnosis of Marine Inverter

Aimed at fault diagnosis of power switch with the marine inverter of electric propulsion ship，the paper presents a tri class support vector machines（SVM） based fault diagnosis algorithm.Instantaneous positive symmetrical component is obtained by using theory of symmetrical components，wavelet energy of faulty power switch is gotten by using wavelet packet decomposition.After normalizing these energies，the fault features are obtained.According to the power switch position in the inverter and the characteristics of the output waveform，power switch are grouped，and the power switch faults are isolated by using tri-class SVM.The simulation results show that the accuracy of fault classification is 94.29% using the method proposed in the paper，it also shows that the method is effective.