Roller Element Bearing Fault Size Estimation Using Adaptive Neurofuzzy Inference System

Abstract

The vibration level is a function of the defects in the bearing. By identifying a change in vibration level, one can predict the dynamic behavior and fault in the rotor-bearing system. An imminent bearing fault detection can reduce downtime or avoid the failure of rotating machinery. The condition monitoring or maintenance schedule can be set well if the diagnosis estimate bearing fault size accurately. In view of this, the adaptive neurofuzzy inference system (ANFIS) and dimension analysis (DA) were utilized to detect the bearing fault size. Several experiments were performed at different rotating speeds on the rotor-bearing system. Localized defects were simulated on bearing races artificially using electrode discharge machining (EDM) and the vibration responses are acquainted by accelerometer and fast Fourier techniques (FFT). With a 0.1 mm error band to fix minor bugs, a two-performance indicator evaluated the model accuracy. A comparison of the performance of models with experimental results and artificial neural network (ANN) validated the potential of the present approach. The results showed that an ANFIS performs better over DA and ANN. This contributes to detecting bearing fault effectively and accuracy improvement in the estimation of the bearing fault size.