基于分数低阶Stockwell变换时频的机械轴承故障特征提取  

Feature Extraction of Bearing Fault Signals Based on Fractional Lower Order Stockwell Transform Time Frequency Representation

作　　者：龙俊波 汪海滨[1] LONG Jun-bo;WANG Hai-bin(Electronic and Engineering College ,Jiujiang University,Jiujiang 332005,China;Information Science and Technology College ,Jiujiang University,Jiujiang 332005,China)

机构地区：[1]九江学院信息科学与技术学院

出　　处：《科学技术与工程》2019年第18期210-217,共8页Science Technology and Engineering

基　　金：国家自然科学基金(61261046);江西省自然科学基金(20142BAB207006);江西省教育厅科技项目(GJJ170954);九江学院校级课题(2014SKYB009)资助

摘　　要：时频分布是机械滚动轴承故障信号的有效分析方法,特殊情况下的机械故障信号或噪声属于非高斯 Alpha(α)稳定分布,传统的 Stockwell 变换( S 变换)时频方法性能退化甚至失效。基于 S 变换时频和分数低阶矩提出了一种分数低阶 S 变换时频分布算法,为了减少计算量及在线及时分析信号,提出了一种快速分数低阶 S 变换时频算法。仿真结果表明,所提出的分数低阶 S 变换时频算法及其快速算法能很好地工作在高斯噪声和α稳定分布噪声环境,性能优于已有的 S 变换时频。在实际应用中,所提出的时频算法能够较好的提取机械轴承故障信号的故障特征。Time frequency distribution is an effective method to analyze the mechanical bearings fault signals. The mechanical bearing fault signals or noise belong to non-Gaussian Alpha(α) stable distribution in some special cases,then the traditional Stockwell transform time-frequency methods will degrade,even fail. Hence,fractional lower order Stockwell transform ( FLOST) and a fast algorithm are proposed employing Stockwell transform time fre- quency distribution and fractional lower order moments. The simulation results show that the proposed FLOST time frequency distribution method and its fast algorithm can better work in normal Gaussian noise environment and α stable distribution environment,and their performance are better than the existing S transform. In reality,the pro- posed methods can better extract the fault feature of the bearing fault signals.

关 键 词：时频分布 轴承故障 Stockwell变换 特征提取 分数低阶矩 

分 类 号：TN911[电子电信—通信与信息系统]