Optimization design of cooling water pump for fuel cell electric vehicle based on machine learning and NSGA-II

With the increasing demand for efficient fluid machinery in cooling systems of fuel cell electric vehicle, the optimization of centrifugal pumps has become a key factor in improving hydraulic performance and heat dissipation efficiency. Due to the strong size limitations of vehicle-mounted pump, the traditional design methods may be inadequate, such as the failure of linear assumptions of empirical formulas. On this basis, an optimization design to achieve multi-objective performance enhancement of automotive cooling water pumps is carried out in this work. The collaborative framework based on the Extra Trees (ET) regression model and Non-dominated Sorting Genetic Algorithm II (NSGA-II) is adopted. The Hammersley sampling method is used to construct a high-dimensional parameter space sample set, and then the foundational dataset is generated by the computational fluid dynamics (CFD) simulations. Moreover, the data augmentation technology is employed to improve the generalization capability of the machine learning model. The results show that the Extra Trees model is significantly superior to other algorithms in predicting head and efficiency, with R 2 values of 0.9985 and 0.9978, respectively. The NSGA-II-based optimization scheme increased the head by 5.2% and efficiency by 7.7%. And the error of this result compared with CFD verification is less than 1%. The global sensitivity analysis further reveals the dominant role of blade outlet diameter and wrap angle in performance, as well as their interaction mechanisms. In addition, the flow field of optimized pump exhibits improved pressure distribution, velocity uniformity, and reduced turbulent kinetic energy. This study provides theoretical and practical insights into multi-objective optimization of centrifugal pumps under stringent geometric constraints, advancing the integration of machine learning and intelligent algorithms in fluid machinery design.