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：《新型工业化》2013年第7期40-49,共10页The Journal of New Industrialization

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

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形成特征信息,有效的反映出信号潜在的特征,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,本文给出了选取信号4层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提取特征降维进行模式识别这种方法。在多组轴承的不同故障、同故障不同损伤程度的多类别数据集上进行了实验,比较识别精度,实验结果验证了这种方法的有效性。Wavelet packet decomposition(WPD) can effectively reflect the potential signal characteristics by decomposing the non-stationary signal into the low and high frequencies. Locality preserving projection(LPP) can retain the local features of the analyzed signal in dimensionality reduction. Combining these two benefits, this paper selects the spectral energy of all nodes with WPD as a characterization of the analyzed signal and uses LPP feature extraction to reduce dimensions for pattern recognition of machine faults. Experimental results show the effectiveness of the proposed method by using multi-class dataset of multiple sets of bearing faults with different fault types and defect severities.

关 键 词：故障诊断 特征提取 识别精度 小波包分解 局部保留投影法 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]