Health Monitoring Approach of Bearing Application of Adaptive Neuro Fuzzy Inference System (ANFIS) for RUL-estimation and Autogram Analysis for Fault-localization

Bearings usually operate under harsh conditions which result in a dynamic behavior generating non-stationary vibration signals and overwhelmed by noise. Therefore, bearing fault diagnosis and prognosis become difficult since the purpose is to extract robust features able to detect the appearance of faults, monitoring the degradation of health state and to predict the remaining useful life (RUL) of bearing. The aim of this paper, is to propose a method for bearing faults feature-extraction using adaptive neuro fuzzy inference system (ANFIS) and autogram analysis. First, times domain features are applied for the raw vibration signal. Then, the selected features are computed to will be analyzed as one of the characteristics that describes the degradation of state system. After that, the curve fitting (smoothing) is applied to normalize the amplitude of the irregular values relatively to others feature values. The calculated value of acquired signal cannot be smoothed or calculated three or more times, hence ANFIS intervenes for modeling the transfer from an indeterminate input to a more relevant value for monitoring the fault evolution. Then, the output of ANFIS estimates the days of acquisition and predict the RUL of bearing. Finally, the autogram analysis is used to identify the degraded element in the bearing.