XGBoost-Based Model for Heat Transfer Coefficient Prediction

This technical report presents a model developed using the extreme gradient boosting (XGBoost) algorithm to predict heat transfer coefficients in liquid cold plates (CPs) that are influenced by surface roughness. The study utilizes a computational fluid dynamics (CFD) approach to prepare the input dataset, which involves solving three-dimensional fluid flow and heat transfer equations under turbulent conditions. The model's performance is evaluated, revealing that it accurately predicts 63% and 90% of heat transfer coefficients within ±10% and ±20% of true values, respectively. The findings suggest that XGBoost is an effective tool for analyzing thermal management solutions, especially when performance data is limited. The report also discusses the implications of surface roughness resulting from additive manufacturing on the hydrothermal performance of thermal management systems. The results underscore the potential of machine learning techniques in enhancing the design and analysis of engineering applications.