ANFIS system for prognosis of dynamometer high-speed ball bearing based on frequency domain acoustic emission signals

Bearing faults account for approximately half of all electric machine failures. Bearing condition monitoring is of practical importance. Until now, there is not any complete research on application of frequency domain signal features on bearing natural fault prediction and also application of ANFIS fuzzy systems for natural bearing fault classification, especially using acoustic emission signals. So, this paper focuses on the study of natural fault detection of angular contact ball bearing using frequency domain signal processing based on acoustic emission signals. This study consists of three stages. At the first stage, after recording the acoustic emission waveforms from the experimental test rig, 45 frequency domain features are introduced and calculated. Next, principal components of features are computed using the PCA and FDA methods. Finally, Adaptive Neuro-Fuzzy Inference System (ANFIS) is used to classify the healthy and naturally defected bearings based on principal components of PCA and FDA methods and the accuracy of these methods are compared with each other. The results show that the classification using the ANFIS network based on the FDA features has less error compared to the features extracted from the PCA method. The classification error using the FDA features has its lowest value using the input pimf membership function and the hybrid optimization method for the constant output membership function, which is 9.0352 x 10(-9). The ANFIS accuracy for the first principal component is 100%. The Antis accuracy for second and third principal component for PCA and FDA methods are 67.6%, 59.15%, 64.8% and 56.3%, respectively. (C) 2020 Elsevier Ltd. All rights reserved.