Condition Monitoring of Rolling Bearings for Automotive Water Pumps

The paper presents a study aimed at comparing the results of the classification of the operating conditions of rolling bearings for automotive water pumps based on different methods, all of which are experimental in nature and based on different features and different classification and clustering techniques. Various operating conditions of the bearings were taken into account. Vibration measurements on rolling bearing housings were experimentally recorded using a dedicated test rig and also analysed in the time and frequency domain. The statistical indices (Standard Deviation, Skewness, Kurtosis, Mean, Peak, and Crest Factor) were calculated for the analysis in the time domain, while the FFT was used for the frequency analysis. The indicators Standard Deviation and Skewness were used as input for a clustering procedure to perform an automatic identification of the health status of the rolling element bearing. A self-organizing-map is used for non-supervised clustering. A feed-forward back-propagation neural network and the DTW technique are used for supervised classification. The proposed approaches are compared with each other based on the results and their different characteristics.