基于Shannon小波支持矢量机二级决策的故障诊断

Fault Diagnosis Based on the Second Class Decision Making through Shannon Wavelet Support Vector Machine

提出一种基于Shannon小波支持矢量机(Shannon wavelet support vector machine,SWSVM)二级决策的故障诊断模型。先求出原信号的双谱相关值特征矩阵奇异值谱,并用BP神经网络对主分量分析(Principal component analysis,PCA)后的奇异值谱调维得到可分性更高的三维模式矢量,再将该三维模式矢量用SWSVM进行二级故障诊断。SWSVM可以对BP网络因陷入局部极值﹑欠/过学习输出的低分辨率进行校正,获得更高的故障识别精度和自适应识别能力。本模型实现了BP网络和SWSVM优势互补。一滚动轴承故障诊断实例验证了该模型的有效性。

A fault diagnosis model based on the second class decision-making through Shannon wavelet support vector machine（SWSVM） is proposed： firstly,singular value spectrum of original signals’ bispectrum correlative character matrix is solved,then singular value spectrum compressed by principal component analysis（PCA） is adjusted to become 3-dimensional pattern vector with better separability by back-propagation neural network,finally,the 3-dimensional pattern vector is input into SWSVM for the second class fault diagnosis. SWSVM can calibrate the low output resolution of BP neural network on account of lapsing into local extremum and insufficient training or over training,and obtain the higher fault recognition accuracy and adaptive recognition capacity. This model realizes advantage complementation between back-propagation neural network and SWSVM. An example of rolling bearing fault diagnosis proves the effectivity of the proposed model.