THE USE OF A NAIVE BAYES CLASSIFIER IN CONTINUOUS MONITORING SYSTEMS FOR ROTATING MACHINERY

Abstract

This paper presents a method for diagnosing rotating machines operating under varying conditions based on order analysis, Gaussian mixture models (GMMs) and Bayesian inference. A classifier was constructed based on the values of the amplitudes of the order spectrum and the value of the rotational speed changing due to the variable load of the machine. An analysis was conducted on the functionality of the method in diagnosing misalignment and unbalance of a drive system consisting of a drive motor and planetary gearbox. The drive train was subjected to a variable load of the main gear of a bucket wheel excavator at a variable oil temperature. In the diagnostic experiment, the method was shown to be highly effective in diagnosing preset system faults. The implementability of the method in embedded systems was also investigated. The method was implemented in a system with a real-time operating system and an FPGA. The method was tested in continuous monitoring mode on a laboratory bench.