摘要： 奇异值分解(SVD)在信号分析时需限定主特征值的数量,影响了故障识别的准确性。为此,提出一种新的故障诊断方法。利用奇异值曲率谱自适应选择有效的奇异值进行信号重构,对重构信号实现二次SVD 处理,产生相同数量的正交分量,然后求解各正交分量的能量矩,构造特征向量,并采用变量预测模型的分类识别方法分析特征向量,从而建立故障识别模型。将该方法应用于实际轴承的故障诊断,实验结果表明,轴承在正常和故障状态下,该方法的综合识别精度达到97. 5% ,高于常规基于SVD 和支持向量机的方法8. 75% 。

关键词: 二次奇异值分解, 能量矩, 自适应, 故障诊断, 信噪比

Abstract: The number of feature values can be limited by using Singular Value Decomposition(SVD),which affects the accuracy of fault identification. A fault intelligent diagnosis method based on quadratic SVD and Variable Predictive Model-based Class Discriminate(VPMCD) is put forward,which can adaptively choose effective singular values firstly by using the curvature spectrum of singular values for reconstructing a signal. The same number of orthogonal components is acquired by SVD again,and the feature vector can be constructed by calculating its energy moment. The model of fault identification can be established by analyzing the feature vectors of using the VPMCD. This method is applied to the bearing fault diagnosis. Experimental results show that,in the normal and fault condition of bearing,the comprehensive identification precision of this method is 97. 5% ,and is 8. 75% higher than the conventional method based on SVD and Support Vector Machine(SVM).

Key words: quadratic Singular Value Decomposition(SVD), energy moment, self-adaptation, fault diagnosis, Signal to Noise Ratio(SNR)

中图分类号: 

• TP277