基于WPD和LPP的设备故障诊断方法研究  

Machine fault diagnosis based on WPD and LPP

作　　者：丁晓喜 何清波[1] 

机构地区：[1]中国科学技术大学精密机械与精密仪器系,合肥230026

出　　处：《振动与冲击》2014年第3期89-93,共5页Journal of Vibration and Shock

基　　金：国家自然科学基金(51005221);教育部高等学校博士学科点专项科研基金(20103402120017)

摘　　要：小波包分解(WPD)能够将非平稳信号在低频和高频上同时分解以有效反映信号潜在的特征信息,而局部保留投影法(LPP)在降维的同时保留了信号的局部特征信息。结合上述特点,给出了选取信号小波包分解后形成全部节点的谱能量,作为表征信号的特征,采用LPP提取降维特征进行模式识别的方法进行设备故障分类研究。在多组不同轴承故障及同故障不同损伤程度的多类别数据集上进行了实验,实验结果验证了这种方法的有效性。Wavelet packet decomposition （WPD）can be used to decompose a non-stationary signal within low and high frequency fields to reflect effectively the potential feature information of the signal.Locality preserving projection （LPP）can be used to retain local features of an analyzed signal during its dimension reduction.Here,combining these features,the spectral energy of all nodes after WPD was given as a characterization of the analyzed signal and LPP was used to extract characters after dimension reduction for pattern recognition of machine faults.The effectiveness of the proposed method was verified by using several multi-class data sets of bearing faults with different fault types and damage levels.

关 键 词：故障诊断 特征提取 小波包分解 局部保留投影 高斯混合模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]