Enhancement of fault diagnosis of rolling element bearing using maximum kurtosis fast nonlocal means denoising

• Published in: Measurement : journal of the International Measurement Confederation Vol. 100; pp. 157 - 163
• Main Author: Laha, S.K.
• Format: Journal Article
• Language: English
• Published: London Elsevier Ltd 01.03.2017; Elsevier Science Ltd
• Subjects: Algorithms; Bearings; Fault diagnosis; Image processing; Kurtosis; Machinery; Noise levels; Noise reduction; Nonlocal means; Roller bearings; Rolling; Rolling element bearing; Signal processing; Vibration; Fault diagnosis; Nonlocal means; Rolling element bearing
• ISSN: 0263-2241; 1873-412X
• DOI: 10.1016/j.measurement.2016.12.058

•NL-means denoising method is applied for rolling element bearing fault diagnosis.•Parameters of the NL-means method maximize kurtosis of the estimated signal.•Denoised signals enhance the fault detection and diagnosis capability. In this paper, a modified nonlocal means denoising (NL-means) algorithm is proposed for rolling element bearing fault diagnosis. Although, nonlocal means denoising is widely used in image processing, this algorithm is rarely used in 1-D signal processing. The present work deals with application of 1-D nonlocal means denoising method for enhancement of fault signature in rolling element bearings. The parameters for the NL-means method are obtained by maximizing kurtosis value of bearing vibration signal. The proposed method is compared with minimum entropy deconvolution (MED) technique and the results indicate that the proposed method performs better for bearing fault diagnosis. The method is shown to be robust against various noise levels. Further, envelope spectrum of bearing vibration signal is also used to obtain characteristic bearing defect frequencies.