Abstract: Aiming at the problem that multiple sensors collected multiple data synchronously for same vibration source,a combination denoising method based on Minimum-Entropy Deconvolution (MED) and Independent Components Analysis (ICA) was proposed and was used to extract cyclic impact feature of rolling bearing.MED was exploited to preprocess the vibrations from each sensor prior to eliminate the effect of vibration transfer path and enhance the cyclic shocks induced by bearing fault.The signals preprocessed by MED were fed to ICA for furthermore denoising where the resulted ICA component with biggest Kurtosis value was chosen as the finally filtered signal.From the envelop spectrum of the finally filtered signal obtained by a sequent use of MED and ICA,the health condition identification or fault localization could be achieved.The experiments on simulated and laboratorial data verified the conjunct application of MED and ICA within the context of multi-sensor condition monitoring of manufacturing equipment.

Key words: rolling bearing, minimum entropy deconvolution, independent components analysis, fault feature recognition

摘要： 针对强噪声背景下多个传感器对同一振动源同步采集多组数据的情况,提出一种基于最小熵解卷积和独立成分分析的联合降噪方法,并用于滚动轴承循环冲击故障特征的提取。利用最小熵解卷积对各传感器的信号分别进行盲解卷滤波,消除信号传递路径的影响,从噪声信号中初步提取出故障冲击特征;对各传感器的滤波信号进行独立成分分析处理,将信号进行重组后得到重构分量,进一步消除噪声成分,使故障冲击特征成分得到二次增强;选取峭度最大的最优独立成分分析重构分量并进行包络谱分析,得到诊断结果。通过仿真数据和实验室数据分析验证了该方法能够增强滚动轴承的循环冲击特征,便于识别故障类型。

关键词: 滚动轴承, 最小熵解卷积, 独立成分分析, 故障特征识别