Abstract
Real-time monitoring of the friction coefficient of the moving parts of a machine system is a challenging problem. The development of intelligent perception and data technology provides the possibility to use tribological correlation information to predict the friction coefficient. This paper uses multi-source friction information such as sound during the friction and wear test to form a time-sectioned friction information data set, establishes a K-fold cross-validation double-stacked regression integration model, defines the evaluation indicators for scope evaluation, and the model was tested with a variety of load test data. The results showed that the model can effectively refine the correlation characteristics of friction information, so as to accurately fit and predict the friction coefficient, and has universality for data under different load conditions.