基于SVD降噪和STFT解调方法的轴承故障检测

Rolling Bearing Fault Detection Based on SVD Denoising and STFT Demodulation Method

针对目前振动信号STFT解调方法在轴承故障检测应用中存在的不足,提出奇异值分解降噪与STFT解调相结合的轴承故障检测新方法。利用振动测试响应信号重构系统的相空间,得到吸引子轨迹矩阵,然后对轨迹矩阵进行奇异值分解以剔除噪声,并利用STFT技术对降噪信号进行解调分析。对于轴承故障自动检测中如何准确选出1组含有丰富故障信息的解调信号序列用于解调谱分析这一难题,提出按修正的归一化峭度(MKv)最大化准则进行选择。仿真和实际轴承故障信号分析结果表明,与传统的基于STFT解调的检测方法相比,提出的改进方法可以更早地发现轴承早期故障,实用性更强。

Aimed at the deficiency of current STFT based vibration signal demodulation method applied to rolling bearing fault detection, a new bearing fault detection method based on singular value decomposition （SVD） denoising and STFT demodulation techniques is proposed. Firstly, the trajectory matrix of an attractor is constructed by the time series of measured response. And then the SVD is performed to eliminate the noise and obtain useful signal. Finally, STFT demodulation technique is applied to denoised signal to extract fault feature. About the issue on how to select a group of demodulated signal series, which contains abundant fault information, for envelop spectrum analysis in the process of bearing fault automatic detection, a new method in terms of the maximum criteria of modified normalized kurtosis factor （MKv） to select is proposed. The simulation and real bearing fault signal analysis results show that compared with current STFT demodulation based detection method, the new improved one has better performance on early bearing fault detection.