The Application of Improved Cyclical Spectrum Density Method in Fault Diagnosis of Rolling Bearing

Abstract

The characteristic of cyclical impact is reflected on the signal of rolling bearing in fault condition. The carrier frequency is modulated by times of the failure frequencies. When the traditional cyclical spectrum density (CSD) method is used to analyze the signal, all the modulation frequencies will be demodulated in the cyclic frequency spectrum. In this case, it is difficult to recognize the fault type of the bearing. Therefore, a new cyclical spectrum density method based on the kurtosis energy (CSDK) is proposed. The kurtosis of every cyclic frequency’s slice is used as the weight coefficient of the cyclic frequency’s energy accumulation to extract fault feature effectively. The proposed method has greatly reduced times of the harmonic frequencies’ effect in traditional CSD method. The analysis of the signal gathered from the outer rolling bearing of blast furnace belt cylinder shows that the fault feature extracted by the new method is more clear and accurate than CSD method.