PROPOSING A METHOD TO ELIMINATE SIGNAL NOISE OF VIBRATION FOR ENGINE FAULT DIAGNOSIS

Abstract

Deep Learning (DL) has lately emerged as the secret to success in the
industrial sector. A current trend in the scientific community is the identification
of motor defects based on vibration data, one of the deep learning applications
in the contemporary manufacturing model. As a result of the vibration data's
great sensitivity to various disturbances. The information input for the
acceleration sensor may be negatively impacted by background movements that
are unneeded. For this reason, cleansing vibration signals may be thought of as
the initial step in diagnosing a bearing machine's issue. In order to enhance the
effectiveness of the motor defect detection, a new denoising approach based on
Fast Fourier Transform (FFT) and K-means clustering is first suggested in this
study. In this paper, a new denoising method based on Fast Fourier Transform
(FFT) and K-means clustering is firstly proposed to improve the performance of
the motor fault diagnosis. Convolutional Neural Network (CNN) is then applied to
classify the motor faults. To validate the performance of the proposed approach,
the open-source Case Western Reserve University (CWRU) data-set is considered.
The experimental results confirm the advantages of the proposed denoising
method when compared to the other existing methods.