基于D-S证据理论和支持向量机的液压泵故障诊断技术

Fault Diagnosis of Hydraulic Pump Based on D-S Evidence Theory and SVM

针对D-S证据理论中难以确定基本概率分配的问题,提出利用"一对一"支持向量机分类解决基本概率分配的方法。首先,通过试验采集了液压泵各种故障状态下的振动信号和压力信号;其次,运用小波包分解提取各频带能量作为特征向量;最后,用所提出方法解决了基本概率分配,并将D-S证据理论和支持向量机相结合对液压泵进行了故障诊断。实验结果验证了该方法的有效性。

Because D- S evidence theory can hardly detemine basic probability assignment（ BPA）, the method used the one against one support vector machine（SVM） class to solve this problem. It collected the vibration signal and pressure in hydraulic pump of a certain type of equipment under a variety of fault conditions. Then it used wavelet packet to decompose energy of fre- quency area in order to get fault feature vectors. The fault was diagnosed by using D - S evidence theory and SVM method. The ex- periment result shows that the method is effective.