Park C., Choi Y., Kim Y.

A hub bearing in a wheel is connected to suspension system of an automobile. Angular contact rolling bearings are usually used for automobile hub bearing because of resisting axial and radial load simultaneously. Excessive loads and many other reasons can cause faults to be created and grown at each component in the bearing. Noise and vibration from the faults can diminish trust of the vehicle and also reduce comfort and delightfulness. Therefore, we need to find the faults in the hub bearing as early as possible for preventing from noise and vibration. How early we can detect the faults has to do with how the detection algorithm finds the faults information from measured signal. The incipient faults size is so small that periodic signal is buried by operating noise in measured signal. To detect early faults signal in noise, we have to distinguish different characteristics between them. The bearing faults signal has deterministic feature in contrast with random one of noise. This information allows us to find the early faults regardless how much noise contaminates the signal. Especially, what we have interested is not a system characteristic information but periodicity itself of faults. Therefore, we observe the periodicity in quefrency domain for disregarding the system information. Minimum variance cepstrum(MVC), which is a type of power cepstrum using minimum variance algorithm, is introduced for observing periodicity. This paper shows how the method is good for detecting small faults from experiment which is carried out in the condition of a hub bearing being equipped with an automobile. From the experiment, we found small faults at some bearings passed acceptance test. Moreover, we compared the result with other signal processing methods¡¯ one- moving window, continuous wavelet analysis- for inspecting how effective the method is.