Skip to Main Content
In this paper, we present a simple and efficient machine performance assessment approach based on Gaussian mixture model (GMM). By only utilizing the machine performance signatures generated from normal machine operation, a GMM can be trained to model the underlying density distribution of the training data. Machine performance assessment can be accomplished by quantifying the distance between the GMM for the most recent observed machine condition and that for normal machine operation. Experimental results based on real industrial run-to-failure bearing tests have shown that GMM can efficiently assess the performance of test bearings. The proposed approach has a great potential for a variety of machine performance assessment applications.