A study of recognising the friction state of revolute pairs based on the motor current signature analysis

The recognition of the friction state of revolute pairs can help with the monitoring of their operating conditions. A new method for recognising the friction state of revolute pairs based on motor current signature analysis was proposed in this paper. Current signals were decomposed and multidimensional features were extracted by variational mode decomposition. The support vector machine was trained by ensemble learning, and a combined classifier was built. The friction state of revolute pairs at this moment could be evaluated precisely after their friction feature signal was recognised by multiple types of classifiers with different classification intervals, and the time before dry friction could be preliminarily estimated. The results indicate that, the change of the friction state of revolute pairs could be represented completely by analysing the stator current features of the motor driving the revolute pairs and extracting the multi-domain and multi-dimensional friction features.