Hybrid-driven model for temperature prediction in wet clutches

A hybrid-driven model for predicting the temperature at the friction interface of wet clutch friction components was developed, employing a radial basis function neural network algorithm. This model uses temperature data obtained from wet clutch test bench experiments to compensate for the results from a thermo-mechanical coupled numerical simulation model of the friction components. The hybrid model inputs include load pressure, relative speed, radial position of the friction components, and numerically simulated temperatures at the friction interface, to output predicted temperatures under specific operating conditions. The results indicate that the predicted temperatures of the friction components under various conditions align well with experimental values, with a maximum mean relative error of 0.033 and all prediction intervals having a coefficient of determination higher than 0.97. Thus, the hybrid-driven model demonstrates high accuracy and stability in predictions, and possesses strong adaptability and predictive capabilities for different temperature distributions at the friction interface.