Predictive method of pressure loss in wet clutch caused by seal wear based on stacking ensemble learning

Abstract

The harsh operating conditions of heavy-duty vehicles accelerates the wear of the sealing ring in the transmission, leading to increased oil leakage and a reduction of the operating pressure in the piston cylinder of wet clutch. This impairs the proper functioning of the transmission in the heavy-duty vehicle. Therefore, it is necessary to predict the pressure loss inside the transmission quickly and effectively after the wear of the sealing ring. The wear of the sealing ring under different operating conditions is calculated through the modified Archard model. The relationship between the oil leakage and pressure loss after the wear of the sealing ring is analyzed using Fluent software. The analysis involves the effects of different wear levels of the sealing ring. The simulation results are validated through a high-speed oil cylinder performance test rig. Based on the validated simulation data and test data, a prediction model for pressure loss is established by using stacking ensemble learning with MLR (multiple linear regression), DTR (decision tree regression), and SVR (support vector regression) as the base learners and RF (random forest) as the meta-learner. The risk of model overfitting is reduced through k-fold cross-validation. The research results indicate that the fused stacking ensemble learning algorithm fully utilizes the advantages of each base learner and can effectively predict the pressure loss after the wear of the sealing ring, and achieve a higher accuracy. The establishment of this model provides theoretical support for real-time prediction of pressure loss after the wear of the sealing ring in actual heavy-duty vehicles.