基于孪生支持向量机的齿轮箱故障诊断

Gear Box Fault Diagnosis Based on Twin Support Vector Machine

针对齿轮箱振动信号非平稳特性以及故障样本数据处理困难的特点,提出了基于小波包分解和孪生支持向量机的故障诊断方法。首先采集信号通过Mallat塔式算法对信号进行小波分解再重构从而获得频带能量谱,然后通过归一化的方法再提取各频带的故障诊断特征向量。并将它送入孪生支持向量机进行训练。实验表明,该方法有效提高了分类精度和鲁棒性,而且具有较高的诊断效率。

Aiming at the non-stationary characteristics of gearbox vibration signal and the difficulty of processing fault data,a fault diagnosis method based on wavelet packet decomposition and twin support vector machine is proposed.Firstly,the signal is acquired by wavelet decomposition and reconstruction by Mallat tower algorithm to obtain the band energy spectrum,and then the fault diagnosis feature vector of each frequency band is extracted by the normalization method.And send it to the twin support vector machine for training.Experiments show that the method effectively improves classification accuracy and robustness,furthermore has high diagnostic workpiece ratio.