Jixie chuandong (Jan 2017)
Health Assessment of Rolling Bearing based on Self-organizing Map and Restricted Boltzmann Machine
Abstract
In order to conduct dynamic health assessment for the roller bearing and describe the dynamic process of its performance degradation accurately,by using the method of combining the Self-organizing map（ SOM） and restricted Boltzmann machine（ RBM） to implement the health assessment of rolling bearing. With the consideration of degradation induced changes for response features,unsupervised learning scheme based on SOM,with the help of multi-domain feature sets consist of features in time domain,frequency domain and time-frequency domain,the optimal feature domain is constructed by sorting the diverse features adopting the sequential forward selection（ SFS） regulation,the mapping relationship between the selected feature vectors and bearing health status is obtained. To avoid the problems of being easy to be run into local optimum,parameter adjustment difficulties and a long training process,when conducting the learning process using traditional neural network algorithms,the model for health assessment is constructed by using the RBM as an alternative.Experimental results demonstrate that the proposed method can identify the health status of rolling bearing during the dynamic process of performance degradation with a good engineering applicability.
