级联双稳随机共振和基于 Hermite 插值的局部均值分解方法在齿轮故障诊断中应用
2016-01-12李永波,徐敏强,赵海洋等
第一作者李永波男,博士生,1986年8月生
通信作者徐敏强男,教授,博士生导师,1960年生
级联双稳随机共振和基于Hermite插值的局部均值分解方法在齿轮故障诊断中应用
李永波,徐敏强,赵海洋,张思杨,黄文虎
(哈尔滨工业大学深空探测基础研究中心,哈尔滨150001)
摘要:针对于弱信号在齿轮故障中难以提取问题,提出了一种基于级联双稳随机共振 (Cascaded Bistable Stochastic Resonance,CBSR)降噪和局部均值分解(Local Mean Decomposition,LMD)齿轮故障的诊断方法。随机共振可有效削弱信号中的噪声,利用噪声增强故障信号的微弱特征;LMD方法可自适应将复杂信号分解为若干个具有一定物理意义上PF分量之和,适合处理多分量调幅调频信号。首先将振动信号进行CBSR消噪处理,然后对消噪信号进行LMD分解,通过PF分量的幅值谱找到齿轮的故障频率。通过齿轮磨损故障诊断的工程应用,表明该方法可以有效提取齿轮故障微弱特征,实现齿轮箱的早期故障诊断。
关键词:级联双稳随机共振 局部均值分解 故障诊断 齿轮
基金项目:国家自然科学基金资助项目(10772061)
收稿日期:2013-09-09修改稿收到日期:2014-02-20
中图分类号:TH165.3文献标志码:A
Application of cascaded bistable stochastic resonance and Hermite interpolation local mean decomposition method in gear fault diagnosis
LIYong-bo,XUMin-qiang,ZHAOHai-yang,ZHANGSi-yang,HUANGWen-hu(Deep Space Exploration Research Center, Harbin Institute of Technology, Harbin 150080, China)
Abstract:Aiming at the difficulty of extracting the weak signal in gear fault diagnosis, a method for gear fault diagnosis based on cascaded bistable stochastic resonance(CBSR)denoising and local mean decomposition(LMD)was proposed. The technique of stochastic resonance can remove noise in signals effectively and make use of noise to strengthen the weak fault feature; LMD can decompose a complicated signal into several stationary PF (product function) components with reality meanings, so it is very suitable to analyze the multi-component amplitude-modulated and frequency-modulated signals. Here, the CBSR was employed in the pretreatment to remove noise in vibration signals, the denoised signal was decomposed with LMD, and then the fault frequency of gear was found by inspecting the amplitude spectra of PF components. The engineering application of the method in fault diagnosis of gear wear demonstrated that it can extract the weak feature of gear fault effectively and realize the early gear fault diagnosis.
Key words:cascaded bistable stochastic resonance; local mean decomposition; fault diagnosis; gear
在实际的旋转设备中,由于现场环境恶劣,例如振动传递路径的多变、强背景噪声以及多分量的振动源激励等,导致测点信号的信噪比低、特征微弱且噪声污染信号严重[1]。因此在设备早期故障阶段,特征信号往往会被淹没,难以实现早期故障诊断。微弱特征提取方法较多,多是以尽量消除或抑制噪声为主,然而对于特殊非线性系统,此方法在降低噪声同时也削弱了特征信号。……
