.Rolling element bearings are the one of most critical component of rotating machinery and common mechanical system. Thus accurate prediction of bearing faults become of great importance to minimizing unexpected failure of machine. It is well known for that the main failure mechanism of bearings is rolling contact fatigue under normal operating condition. Once an incipient fault is generated by rolling contact fatigue, rolling contact wear is the most frequently seen phenomenon. Many research studies have modeled the wear process of the bearing; however, each prior study has limitations which that come arise from the non-linearity of the wear process. To solve this limitations, therefore, we propose a descriptive degradation model that uses using vibration signals and adds understanding of the physical status of the bearing during the wear process. To attain the actual wear process of the bearing, we constructed and performed an experiment that represented in-situ bearing faults. From the vibration signal measured in the experiments, we extracted time domain and frequency domain features. As a result, we examined the process of bearing degradation as series of stages which that relate to the conditions of the physical bearing conditions. This study will benefit to enhance physical understanding for bearing faults in various engineering systems. Also using the proposed method, we expect to be able to predict the condition of the bearing condition without collecting a lot of condition monitored data.