Bearing faults analysis using modified ANFIS architecture in AC motors implementation

Bearing is a mechanical element of machine to reduce friction between two rotating objects. Bearings have an optimal lifetime but in real application, not all of them are able to reach their lifetime and most of them are damaged in a relatively short time. Any flaw on bearing that is not addressed earlier can lead to operation failure on mechanical equipment or machines. Usually type of bearing damage can be used to identify the real cause of the failure. In order to do a correct measure, accurate information about the type of bearing damage is needed. In this research, the bearing fault analysis is done using modified ANFIS method. Training data and test data used in this research were taken from Case Western Reserved University. Training data was observed in time and frequency domains. Observation results in time domain are root mean square (RMS) and kurtosis values. In frequency domain, data was filtered using Hilbert transformation to obtain the maximum value of amplitude and frequency. The RMS, kurtosis, amplitude, and frequency were used as input parameters on modified ANFIS. The results has shown that the system can identify bearing damage in accordance with the type and level of the damage with success rate is 99.61 percent.