Optimization of Cooling Jacket Geometry Based on Numerical Modeling and Machine Learning Algorithms

Cooling jackets are crucial components in electric vehicles, helping to regulate the temperature of inverters, which are crucial for managing electrical energy. In this research study, the design of cooling jackets with wavy fins was investigated, and the performance of these jackets was analyzed by varying the geometry of the wavy fins. Machine Learning (ML) techniques were employed to optimize fin design and reduce computational cost and time. The geometry of the wavy fins was parameterized, and the most sensitive parameters affecting performance parameters, such as pressure drop and temperature difference, were identified. Dimensionless parameters were used to ensure the applicability of the method to different sizes and designs of cooling jackets. Two ML models, the Global Regression Model (GRM) and Random Forest Regressor (RFR), were trained on Computational Fluid Dynamics (CFD) analysis results. The trained GRM and RFR models were then used to predict the optimal parameters of the wavy fins. Optimization results demonstrated a 36.03% reduction in pressure drop. The ML models exhibited high accuracy and efficiency in predicting cooling jacket performance, providing a fast and effective approach for parameter optimization. This methodology holds significant potential for future studies, offering insights and guidance for designing and optimizing cooling jackets with wavy fins.