基于多小波能量谱与SVM的导弹惯性器件故障预报

针对利用时域信号进行故障预报精度低的问题,提出一种基于多小波能量谱与支持向量机(SVM)的故障预报方法.该方法以信号在多小波域上分解形成的能量谱作为故障的诊断特征,通过距离评测准则选取最优多小波能量谱特征子集.最后将最优特征作为样本训练支持向量机.利用训练后的SVM进行故障预报,试验结果表明多小波能量谱能更有效地反映惯性器件故障特征,利用SVM与多小波能量谱结合可以得到更好的预测精度.

Inertia Devices Fault Prediction Based on Multiwavelet Energy Spectrum and Support Vector Machine

To improve fault predicting precision,the signal energy spectrum in multi-wavelet domain was used as fault diagnosis characteristics,support vector machine (SVM) was used for fault prediction of inertia devices of missile. With the distance evaluation technique,the sub-filed with optimal features was obtained. The optimal features were input into the SVM to identify different fault cases. The experimental results show that the fault predicting precision is improved by the proposed method.