Kriging-assisted design optimization of the impeller geometry for an automotive torque converter

Abstract

To reduce the total design and optimization time, numerical analysis with Kriging surrogate model coupled with a multi-objective optimization method based on non-dominated sorting genetic algorithm II (NSGA-II) is used to design optimum impellers for an automotive torque converter. Design parameters used for optimization are determined based on a new impeller parametric model which is established by means of Creo software and validated by experimental tests. In this work, the objectives are to maximize the stall torque ratio and peak efficiency, which are always used to represent dynamic and economic characteristics of torque converters, respectively. Kriging surrogate models with different polynomial orders and correlation functions are evaluated in this work. Results indicate that for the impeller design optimization problem, the Gaussian model with zero order polynomial is best for stall torque ratio prediction while the Gaussian model with first order polynomial is more suitable for peak efficiency prediction. A comparison is also made among NSGA-II and other optimization methods studied. The results show that NSGA-II has broader and more evenly distributed solution points in the Pareto front space. Lastly, two extreme optimal designs are analyzed using computational fluid dynamics (CFD) runs. Results demonstrate that the presented method can provide an effective solution to geometric design of impellers for improving torque converter performance.