基于VMD-DBN的滚动轴承故障诊断方法

为揭示滚动轴承故障振动信号的典型特征规律，结合变分模态分解(VMD)与深度置信网络(DBN)的优势，提出轴承振动信号特征的提取方法.将信号先进行基于VMD的分解，根据各模态分量频谱图确定其模态参数，得到若干个模态分量.然后，基于DBN强大的特征提取能力，采用DBN无监督特征提取方法，将得到的模态分量映射到一维，并融合各分量的DBN特征形成特征向量，将其作为粒子群优化支持向量机(PSO-SVM)的输入进行故障诊断.实验验证与对比分析证明了VMD-DBN方法的可行性与优越性.

Fault Diagnosis Method of Rolling Bearing Based on VMD-DBN

In order to identify the vibration signal features of faulty bearing， a feature extraction method of bearing vibration signals based on the variational mode decomposition (VMD) and deep belief network (DBN) is proposed. First， the signal is decomposed based on VMD and the parameters of each modal component are determined by the modal component spectrogram， thus several modal components being obtained. Then an unsupervised feature extraction method based on DBN， which has powerful feature extraction ability， is used to map the modal components obtained to one dimension， and the DBN features of each component are merged to form feature vectors and input into particle swarm optimization support vector machine (PSO-SVM) for fault diagnosis. Experimental verification and comparative analysis show the feasibility and superiority of the VMD-DBN method proposed.