Discrimination of Multiple Faults in Bearings Using Density-Based Orthogonal Functions of the Time Response

This study investigates the use of the mapped density of time response using orthogonal functions to detect single and multiple faults in rolling element bearings. The method is based on constructing the density of a single time response of the system by using orthogonal functions. The coefficients of the orthogonal functions create the feature vector in order to discriminate between different rolling element bearing faults. The method does not require preprocessing of the data, noise reduction, or feature selection. This method has been applied to vibration data of different bearing conditions at rotational speeds ranging from 300 rpm to 3000 rpm. These conditions include a healthy bearing, and bearings with defects in inner race, outer race, combination of inner race and outer race and rolling element. The results have shown remarkable detection efficiency in the case of a single and two bearing fault configurations. In general, for all bearing configurations, the approach has high performance in detecting defective conditions. These results indicate that using the mapped density to characterize the system under different conditions has considerable potential in bearing diagnostics.