基于正交邻域保持嵌入特征约简的故障诊断模型.pdfVIP

学兔兔 第32卷 第3期 仪 器 仪 表 学 报 V01．32 No．3 2011年3月 Chinese Joumal of Scientific Instrument Mar．2010 基于正交邻域保持嵌入特征约简的故障诊断模型 李 锋，汤宝平，董绍江 (重庆大学机械传动国家重点实验室 重庆 400030) 摘 要：提出一种基于正交邻域保持嵌入(orthogonal neighborhood preserving embedding，ONPE)特征约简的故障诊断模型。首 先将原振动信号经验模式分解(empirical mode decomposition，EMD)并构造Shannon熵得到高维特征向量，再利用ONPE将高维 特征向量约简为低维特征向量，并输入到最近邻分类器(k-nearest neighbors classifier，KNNC)中进行故障识别。本模型充分利 用了EMD分解在故障特征提取、ONPE在信息压缩和KNNC在分类决策方面的优势，实现了旋转机械故障特征提取到故障诊 断的伞程自动化，并提高了诊断精度，为旋转机械故障诊断提供了一种新的模型分析方法。一个滚动轴承故障诊断实例验证了 该模型的有效性。 关键词：正交邻域保持嵌入；流形学习；特征约简；最近邻分类器；经验模式分解；故障诊断 中图分类号：TH165．3 TN911．2 文献标识码：A 国家标准学科分类代码：510．40 Fault diagnosis model based on feature compression with orthogonal neighborhood preserving embedding Li Feng，Tang Baoping，Dong Shaojiang (The State Key Laboratory ofMechanical Trarhsmission，Chongqing University，Chongqing 400030，China) Abstract：A new fault diagnosis model is proposed based on feature compression with Orthogonal Neighborhood Pre— serving Embedding(ONPE)．Firstly，the fault vibration signals are decomposed using EMD and Shannon entropy is constructed to get high—dimensional eigenvectors：then the high—dimensional eigenvectors are compressed to low—di— mensional eigenvectors with ONPE；finally，the low—dimensional eigenvectors are inputted into K—nearest neighbor classifier(KNNC)f0r fault classification．Making full use of the advantages of EMD decomposition in fault feature extraction，ONPE in information compression and KNNC in classification decision—making，the proposed model not only realizes complete automation from fault feature ex