基于稀疏表示的轴承早期故障特征提取

发布时间：2018-02-16 08:49

本文关键词： 稀疏表示 K-SVD Batch-OMP 峭度值冲击成分轴承故障诊断　出处：《北京理工大学学报》2016年04期 　论文类型：期刊论文

【摘要】：低速重载机械设备中的滚动轴承由于承受巨大载荷,极易出现内外环故障.在故障早期阶段,反映故障特征的冲击成分很微弱,极易被噪声覆盖而难以识别.为准确诊断轴承早期故障,提出基于稀疏表示的故障特征提取方法.该方法利用K-SVD字典训练算法构造出能准确匹配冲击成分的字典,克服了参数化字典缺乏自适应性的问题;稀疏编码过程中,采用批处理正交匹配追踪算法(batch orthogonal matching pursuit,Batch-OMP)对振动信号进行分解,以逼近信号的峭度值最大原则作为分解结束条件,自适应确定出分解次数;最后,通过对重构的特征成分进行包络谱分析得出故障类型.对仿真信号和轴承振动信号进行故障特征提取,结果表明所提方法能准确提取出冲击成分,验证了其有效性和实用性.
[Abstract]:The rolling bearing in low speed and heavy load mechanical equipment is prone to internal and external ring failure due to the huge load. In the early stage of the fault, the impact component reflecting the fault characteristics is very weak. It is easy to be covered by noise and difficult to identify. In order to accurately diagnose the early fault of bearing, a method of fault feature extraction based on sparse representation is proposed. This method uses K-SVD dictionary training algorithm to construct a dictionary that can accurately match the impact component. It overcomes the lack of adaptability of parameterized dictionaries, and in the sparse coding process, the batch orthogonal matching pursuitt Batch-OMP) is used to decompose the vibration signal, and the maximum kurtosis value of the approximate signal is taken as the end condition of the decomposition. Finally, the fault types are obtained by the envelope spectrum analysis of the reconstructed feature components, and the fault features are extracted from the simulation signals and the bearing vibration signals. The results show that the proposed method can extract the impact components accurately and verify its validity and practicability.
【作者单位】：武汉科技大学信息科学与工程学院;中原工学院信息商务学院;
【基金】：国家自然科学基金资助项目(61174106)
【分类号】：TH133.3

【参考文献】