A Novel Wavelet Thresholding Method for Vibration Data Denoising and Diagnostic Feature Enhancement in Condition Monitoring

Khalid Rabeyee, Yuandong Xu, Fengshou Gu, Andrew D. Ball

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Data denoising is essential to extract meaningful information and enhance diagnostic features in condition monitoring (CM). Hence, many techniques have been proposed for vibration data denoising. However, manual techniques require well-skilled labourers who will increase the cost of the monitoring process and may not always be available at the site. Denoising in the transformation domain using wavelets by thresholding has attracted researchers in recent years and been shown to be more effective than conventional filtering methods. However, estimation of the threshold value that can effectively suppress the noise has always been a hot research topic. One of the most promising approaches is to automate the process using adaptive and optimization methodologies. A new denoising method is proposed based on adaptive wavelet coefficients thresholding: a thresholding function is constructed, it adopts an adaptive procedure to estimate the initial threshold value based on the energy distribution of the signal; then the optimization procedure is adopted and the optimal threshold is selected through an iterative procedure by maximizing the harmonic-to-signal ratio (HSR). The new method is applied to both simulated data and real datasets collected from a defective tapered roller bearing and it is found that the proposed method increases the signal-to-noise ratio (SNR) significantly without distorting the signal when compared to benchmark thresholding methods.

Conference

Conference25th IEEE International Conference on Automation and Computing
Abbreviated titleICAC 2019
CountryUnited Kingdom
CityLancaster
Period5/09/197/09/19
Internet address

Fingerprint Dive into the research topics of 'A Novel Wavelet Thresholding Method for Vibration Data Denoising and Diagnostic Feature Enhancement in Condition Monitoring'. Together they form a unique fingerprint.

Cite this