A novel approach for fault detection in bearings of rotary machineries at variable load conditions

Abstract

This thesis proposes a novel approach for machine fault detection from vibration data collected at variable load conditions of a system. Although load variation is a common phenomena in real industry, most of the traditional fault detection techniques fails to take this load variability into account while analyzing vibration data. Plant loads and machine rpm change have a significant influence on the vibration data and to address this fact accurately, amultivariate technique combiningMultiscale PCA (MSPCA) andMultiway PCA (MPCA) is presented here. The methodology takes the powerful data signature extraction feature of Wavelet Transform (WT) and strong fault detection ability of PCA and integrate them with the multiple conditions monitoring ability of MPCA. Another significant feature of this proposed multiscale MPCA technique is that it combines the process variables with the vibration analysis. An advanced simulation system of bearing fault at variable loads is presented and the methodology is used on the acquired simulated data. The results are presented along with a comparison with a conventional technique. The efficacy of the proposed methodology is demonstrated on a DC motor experimental setup.