一种改进的TVD电机轴承故障特征提取方法  

An improved TVD fault feature extraction method for motor bearing

作　　者：王凡 马军[1,2] 王晓东[1,2] 朱江艳 WANG Fan;MA Jun;WANG Xiaodong;ZHU Jiangyan(Faculty of Information Engineering&Automation,Kunming University of Science and Technology,Kunming 650500,China;Yunnan Key Laboratory of Artificial Intelligence,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,昆明650500 [2]昆明理工大学云南省人工智能重点实验室,昆明650500

出　　处：《振动与冲击》2023年第10期203-214,共12页Journal of Vibration and Shock

基　　金：国家自然科学基金(62163020,62173168);云南省基础研究计划项目(202101BE070001-055,202102AD080007)。

摘　　要：针对全变分去噪(total variation denoising,TVD)方法在恢复信号特征的过程中使用L1范数会导致信号振幅降低和正则化参数λ难以选取的问题,提出了一种基于非凸全变分去噪(no convex total variation denoising,NCTVD)和天牛须搜索(beetle antennae search,BAS)算法的电机轴承故障特征提取方法。首先,引入反正切非凸惩罚函数定义二阶TVD中的正则化项,增强信号的冲击特征并诱导稀疏性;其次,利用BAS算法对NCTVD中的正则化参数λ和凸性参数a进行寻优并选取最佳参数组合来增强所构造模型的降噪性能,并给予参数约束来保证模型严格凸的性质;然后,通过最小优化算法求解新的NCTVD模型,实现振动信号的降噪和特征增强;最后,结合Teager能量算子(Teager-kaiser energy operator,TKEO)方法对降噪后的信号进行频谱分析,实现对电机轴承故障特征提取的应用验证。公开数据和实测数据的试验结果表明,该方法不仅有效地抑制噪声干扰和表征故障信息,还改善了传统TVD模型在提取故障特征过程中产生的脉冲能量衰减和稀疏效果欠佳的问题。In order to solve the problem that the use of L1 parametrization in the process of signal feature recovery by total variation denoising(TVD) method leads to the reduction of signal amplitude and the difficulty of selecting the regularization parameter λ,a method of motor bearing fault feature extraction based on no convex total variation denoising(NCTVD) and beetle antennae search(BAS) was proposed.Firstly,the regularization term in the second-order TVD was defined by introducing an inverse tangent nonconvex penalty function to enhance the impact characteristics of the signal and induce sparsity.Secondly,the BAS algorithm was used to optimize the regularization parameter λ and the convexity parameter a in the NCTVD and select the best combination of parameters to enhance the noise reduction performance of the constructed model,and parameter constraints were given to ensure the strict convexity of the model.Then,the new NCTVD model was solved by the minimum optimization algorithm.Finally,the Teager-kaiser energy operator(TKEO) method was used to analyze the spectrum of the noise-reduced signal and to validate the application of motor bearing fault feature extraction.The results of public and measured data show that the proposed method can effectively suppress noise interference and characterize fault information,and relieve the problems of pulse energy attenuation and poor sparse effect caused by traditional TVD model in the process of fault feature extraction.

关 键 词：电机轴承 故障诊断 全变分去噪(TVD) 天牛须搜索(BAS) 

分 类 号：TN911.7[电子电信—通信与信息系统] TH165.3[电子电信—信息与通信工程]