System and method for bearing defect auto-detection

A method for performing bearing defect auto-detection comprising:

• providing a wheel rail set comprising a rotatable shaft and at least one bearing,
• positioning at least one vibration sensor on the at least one bearing, the at least one bearing having an inner ring and an outer ring,
• measuring vibrations of the wheel rail set in motion by the at least one vibration sensor, the measured vibrations providing vibration data,
• sending the vibration data from the at least one vibration sensor to a processor, the vibration data including vibration harmonics of the at least one bearing;
• wherein the at least one bearing is coupled to the rotatable shaft, the method further including;
• a) transforming the vibration data via the processor into the frequency domain, the transformed vibration data being provided as magnitude and frequency arrays, along with an approximate shaft speed in revolutions per minute (RPM) or revolutions per second (RPS),
• b) applying a peaks determination method to the transformed vibration data that determines individual peaks from background noise, the peak true amplitudes and their exact peak center frequencies being determined, wherein the exact frequency and amplitude of each possible peak is stored into a peaks determination array,
• c) identifying whole number sub-orders of each peak up to a predefined order number including any peak that falls into a predetermined range of target fundamental frequencies ranges, wherein each peak has a defined defect type and stores those peaks along with a harmonic number and theoretical fundamental frequencies as a frequency array,
• d) clustering the theoretical fundamental frequencies that fall within the predetermined range of target fundamental frequencies ranges, clustered theoretical fundamental frequencies being within a predefined cluster dimension limit, the predefined cluster dimension limit being defined by a number peaks in the theoretical fundamental frequencies, the clustered theoretical fundamental frequencies forming the tightest cluster dimensions are organized into groups and all others as outliers, and the groups stored in cluster arrays, and
• e) Identifying all peaks from the peaks determination array that could be sidebands components of the peaks identified in step (d), having a base delta frequency within a predetermined range of target sideband frequencies ranges, wherein each peak represents a corresponding defined defect type, and up to a predefined sideband+/−delta orders to determine all theoretical base delta sideband frequencies and storing those sideband peaks along with the sideband number, the theoretical delta sideband frequency and the center peak ID index, that fall within each of the sideband target ranges as sideband arrays, and
• f) clustering the base delta frequencies of the sideband components within a predefined cluster dimension limit, the predefined cluster dimension limit being defined by a number peaks in the base delta frequencies, the clustered base delta frequencies forming the smallest cluster dimensions are organized into sideband groups and all others as outliers, and the sideband groups stored in arrays with references to each respective center frequency peak from an orders array.