Study on data fusion of multi-dimensional sensors for health monitoring of rolling bearings

The sensor fusion of multi-sensory measurements is believed to improve the defect detection ability for machinery condition monitoring. A new fault diagnosis method for rolling bearings based on the sensor fusion of oil analysis data, microscopic debris analysis data and vibration analysis data is proposed in this paper. Multi-dimensional sensors were used to record the tribological and vibration data of rolling bearings in typical fault experiments. Oil and microscopic debris analysis was applied to obtain the wear particle number and size distribution, chemical compositions and particle textures, etc. Wavelet transform (WT) and empirical mode decomposition (EMD) were employed to attain the distinguishing features of the vibration data. Then, an intelligent data fusion method based on principal component analysis (PCA) and a genetic algorithm fuzzy neural network (GAFNN) was employed to identify the rolling bearing conditions. Experimental tests have been carried out to evaluate and verify the proposed method. The analysis results show that the fault detection model using the sensor fusion technique produces superior results to those using single measurements and thus it has application importance.