A hybrid analytical and data driven framework for optimizing radially grooved wet clutch geometry

Abstract

Abstract In wet clutch systems, the drag torque is generated by the relative motion between rotating disks in the presence of the oil film. This study aims to optimize the geometry of radially grooved wet clutches to minimize drag torque. A recently developed analytical model was used to efficiently generate the required dataset in both the single- and multiphase flow conditions, as numerical simulations are computationally expensive. The accuracy of the model was first validated with CFD, showing strong agreement with a maximum deviation of 8%. Two artificial neural networks were then trained on the generated dataset and exhibited reliable performance. In the following, these data-driven models were used in an optimization study across four distinct design cases using the genetic algorithm. Numerical simulation of the fluid flow in optimized geometries confirmed significant drag torque reductions of at least 70% across all cases, with the most substantial improvement observed in Case 3 where peak drag torque decreased from 0.92 to 0.022, representing a 97% reduction. Ultimately, a parametric study was conducted to interpret the optimization results. The results showed that changing the distance between the two disks had the most significant impact on drag torque, reducing the peak value by 86%, while the groove angle had the smallest effect, with only a 2% reduction.