基于自适应紧框架学习的轴承故障诊断  

Bearing fault diagnosis based on adaptive tight frame learning

作　　者：柏壮壮 卢一相[1] 高清维[1] 孙冬[1] BAI Zhuangzhuang;LU Yixiang;GAO Qingwei;SUN Dong(School of Electrical Engineering and Automation,Anhui University,Hefei 230601,China)

机构地区：[1]安徽大学电气工程与自动化学院,合肥230601

出　　处：《振动与冲击》2021年第10期296-303,共8页Journal of Vibration and Shock

基　　金：安徽省自然科学基金(2008085MF183);安徽省教育厅重点项目(KJ2018A0012);国家自然科学基金(61402004,61370110)。

摘　　要：轴承是工业生产设备中最关键的零部件之一,其运行状态直接影响到整个机器或系统的运行。针对传统多尺度变换特征无法自适应解决故障诊断问题,提出一种基于自适应紧框架学习的故障诊断方法。针对过完备字典各原子具有明显的相关性,影响故障信号的重建效果,在含噪条件下通过学习构造紧框架,它既可以对轴承故障信号进行有效的自适应描述,又能完全重构故障特征信号。各故障模态有着不同的特征频率,在紧框架下,根据故障特征频率在不同滤波器下具有不同的频率响应构造故障特征;利用基于遗传算法优化的支持向量机(SVM)进行分类训练与测试。实验表明,提出的算法对轴承的故障分类有很好的效果,而且可以针对不同的故障有很好的诊断能力。A bearing fault diagnosis method based on adaptive compact frame learning was proposed to solve the problem that by the method of traditional multi-scale transformation,it is difficult to do the diagnosis adaptively.In view of the obvious correlation between the atoms of an over-complete dictionary,which affects the reconstruction effect of the fault signal,a compact framework in the condition of noise was constructed by learning.The method,could not only effectively describe the bearing fault signal adaptively,but also completely reconstruct the fault characteristic signal.Considering each fault mode having different characteristic frequencies,by making use of the tight frame,different frequency responses were generated by different filters,and according to the fault characteristic frequencies,fault features were constructed.A support vector machine(SVM)based on genetic algorithm optimization was used for classification training and testing.The experimental results show that the proposed algorithm is effective in fault classification of bearings and has good diagnostic capability for different faults.

关 键 词：故障诊断 紧框架 遗传算法 支持向量机(SVM) 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]