Health monitoring of rolling element bearings using improved wavelet cross spectrum technique and support vector machines

An improved wavelet cross spectrum (IWCS) scheme is proposed in this paper for the health monitoring of the rolling element bearings. The parameters under investigation are the axial preload, the lubricant condition and the surface roughness of the contact surfaces. These parameters determine the performance and the life of ball-bearings, especially in the space application which demands low torque noise during the mission life and operating under varying conditions of temperature and speed. The developed method takes the advantages of the wavelet cross spectrum technique for the feature extraction from the non-stationary vibration signatures. The vibration signals of the Rolling Element Bearings (REB) are first analysed by a continuous wavelet transform (CWT) over the selected scales corresponding to the bearing fundamental fault frequencies. Further, cross-correlation is utilised to enhance the defect-related periodic features. In this improved scheme, the contributive bandwidth selection from the Jarque-Bera (JB) statistic index is carried out with the assistance of an outlier technique. This method removes any outliers in the JB index data by using the linear interpolation and thereby enhancing the index value of the other cross-correlated scales. Experiments are conducted to verify the validity of the IWCS and found to be very effective in diagnosing the bearing health conditions. Using the support vector machines (SVM), the classification of the health conditions is obtained based on the novel improved wavelet cross spectrum analysis. •Deals with the condition monitoring of rolling element bearings.•The wavelet cross-spectrum method for extracting defect related signatures from vibration data is enhanced.•An outlier removing scheme is used in determining the contributing bandwidth selection.•The classification of different health conditions is obtained using support vector machines.