Prediction of bearing fault size by using model of adaptive neuro-fuzzy inference system

Abstract

Condition monitoring of bearings faults which have vital importance in machines and detection of faults earlier have very big importance in terms of disruption of process. In this study, certain sizes artificial faults are generated by the laser beam on inner rings of bearing and vibration signals are obtained from these bearings in a shaftbearing setup. It is aimed to diagnose the size of the defects occurring in the bearings by using adaptive neuro-fuzzy inference system (ANFIS) model in the study. After extracting the real-time features of obtained vibration data, they are multiplied by the specific weight and they are given as input to the generated classification model. It has been observed difference of features extracted from of 0.15 cm, 0.5 cm, 0.9 cm diameter inner ring faulty bearings created by the laser depending on size of faults. ANFIS classification model is developed by using these features and the size of the faults occurring in these bearings were calculated with an actual error 2.40 %. Then a error band are created with 0.1mm threshold value and it is observed that all the predicted values are inside this error band.