Coupling mechanism and data-driven approaches for Hydro-viscous Drive torque characteristics under impact condition

Abstract

A hybrid model for Hydro-Viscous Drive(HVD) by combining a mechanism-based torque balance equation for clutch engagement and a Particle Swarm Optimization-Long Short-Term Memory(PSO-LSTM) neural network model is proposed to simulate the abrupt torque change accurately. The model accounts for sudden torque fluctuations under impact conditions, enhancing the accuracy of predicting the time and torque of the mutation node using data from a comprehensive test bench. The experimental validation demonstrates that the hybrid model outperforms the traditional mechanism model, reducing the average error by 20.04%. Systematic assessment highlights the stability and robustness of the model performance, showing substantial improvements over conventional neural networks. The parametric study reveals the crucial influence of rotational speed and oil temperature on torque characteristics.