Experimental-Based Fault Diagnosis of Rolling Bearings Using Artificial Neural Network

Abstract

In this paper, an innovative system for condition-based monitoring (CBM) using model-based estimation (MBE) and artificial neural network (ANN) is proposed. Fault diagnosis of deep groove ball bearings (DGBB) is a key machine element for stability of rotating machinery. MBE model is proposed to demonstrate and estimate the vibration characteristics of bearings. It is realized that it may be worth mentioning that the vibration analysis of damaged bearings at all the positions of a structure is difficult to obtain. For this purpose, methods have been discussed to get the utmost information to notify bearing faults. The ANN approach enables us to determine the effects of various parameters of the vibrations by conducting the experiments. The results point out that defect size, speed, load, unbalance, and clearance influence the vibrations significantly. Experimental simulated data using the MBE and ANN models of rotor–bearing are used to identify the damage diagnosis at a reasonable level of accuracy. The results of the experiments consist in constantly evaluating the performance of the bearing and thereby detecting the faults and vibration characteristics successfully. The effects of faults and vibration characteristics obtained using the experimental MBE and ANN are studied.