A novel extension model for predicting the friction coefficient of fluorinated ethylene propylene based on temporal convolutional networks expansion algorithms

Fluorinated ethylene propylene (FEP) polymers have very low friction coefficients and are widely used in industrial applications. Therefore, establishing a model that correlates the friction behavior of polymer FEP with the friction environment is crucial for studying the friction mechanism of polymers. This study collected characteristics of 10 friction pairs to construct an extended time series dataset of friction behavior. PCA dimensionality reduction was employed to reduce the complexity of the friction behavior data, followed by the construction of models for all 10 pairs simultaneously using the TCN algorithm and TCN-GRU algorithm, to build an extension model capable of simultaneously predicting the friction coefficient of different friction pairs. The method for constructing the extension model was selected by comparing the modeling results to construct the extension model. Experiments on different test sets of 10 corresponding friction pairs showed that the model can achieve high-precision, long-term (600 s) universal prediction of the friction coefficient under different working conditions and different counterface metals of FEP. At the same time, the experiment also showed that the TCN algorithm is more suitable for building an extension model than the TCN-GRU algorithm.