Acoustical Semi-blind Deconvolution for Bearing Defect Detection

• Main Authors: Wang Yu, Zhong Wen, Yin Yang, Liu Xiaoyin, Song Chunhua
• Format: Article
• Language: English
• Published: EDP Sciences 2016-01-01
• Series: MATEC Web of Conferences
• Online Access: http://dx.doi.org/10.1051/matecconf/20166301012

Acoustical machine monitoring is frequently complicated by noisy environments at a production site. This paper presents a semi-blind deconvolution algorithm to extract only one desired acoustic source signal from different sources which are convoluted and mixed by mechanical systems before being measured. The method is based on blind model transformation, robust independent component analysis, reference signal and spectral distance. The new algorithm is tested on simulation and experimental cases. Results demonstrate that blind separation of acoustic signals is possible even when measurements are distanced from vibration exciting sources of faulty bearings. Furthermore, the method can eliminate the effect of structural resonances and large reverberation time of mixtures, which often causes severe problems in classical acoustical diagnostic methods of rolling element bearings.