深度支持向量机在齿轮故障诊断中的应用

针对齿轮箱故障诊断中存在的早期非平稳微弱故障信号特征提取困难,易受强背景噪声干扰,故障诊断精度较低等问题,提出了一种基于变分模态分解(Variational Mode Decomposition,VMD)和深度支持向量机(Deep Support Vector Machine,DSVM)的齿轮箱故障诊断方法。首先,利用VMD将原始振动信号分解成若干个频率尺度的本征模态(Intrinsic Mode Function,IMF)分量,并根据峭度最大准则选取IMF分量对信号进行重构;构建多层支持向量机结构,在输入层利用支持向量机对信号进行训练,学习信号的浅层特征,利用"特征提取公式"生成样本新的表示,并作为隐藏层的输入,逐层利用深层SVM对新样本训练并学习信号的深层特征,最终由输出层输出诊断结果。最后,通过齿轮箱故障诊断实验验证了该方法的有效性。

Application of Deep Support Vector Machine in Gear Fault Diagnosis

Gearbox fault diagnosis has problems in early feature extraction of non-stationary weak faultsignals, vulnerability to strong background noise, and low accuracy of fault diagnosis. A gearbox fault diagnosismethod based on Variational Mode Decomposition(VMD)and Deep Support Vector Machine(DSVM) is pro-posed. Firstly, the original vibration signal is decomposed into several frequency scale Intrinsic Mode Function(IMF) components by VMD, and the IMF component is selected according to the maximum kurtosis criterion toreconstruct the signal. Secondly, the multi-layer support vector is constructed. The SVM is used to train thetraining sample on the input layer, and it learns the shallow features of the data. The feature extraction formula isused to generate a new expression of the sample, which is used as input of the hidden layer. The hidden layer ofthe SVM trains on the new sample, and it extracts and learns the deep features of the signal layer by layer, even-tually, it outputs the diagnostic results on the output layer. The effectiveness of the proposed method is verifiedby the gearbox fault diagnosis experiment.