Abstract Recent success of machine learning (ML) and deep learning (DL) has significantly improved the simulation of complex heat-transfer systems, with strong nonlinearities, multiphysics interactions and large parameter spaces, that are hostile to conventional numerical simulation methods. This review is dedicated to the application of ML/DL techniques and their critical assessment that have been published within the past 10 years (2015–2025) concerning the key heat-transfer processes, including conduction, convection, radiation, nanofluid heat transfer, and multiphase systems. Performance measures used to implement identification of predominant trends in the modeling include coefficient of determination (R 2 ), root mean square error (RMSE) and mean absolute error (MAE). The values of problem R 2 are often measured to be larger than 0.95 in ensemble and kernel-based models of convection-dominated problems and larger than 0.98 in deep learning models such as convolutional neural networks (CNNs) and long short-term memory (LSTM) networks in transient and highly nonlinear heat-transfer problems. Physics-informed neural networks have also been found to exhibit superior generalization and physical consistency in conduction and multiphase systems and achieves errors on the order of the 3% range on benchmark problems. The key problems of the absence of data, the applicability of models, interpretability, and the integration of physics are also discussed in details to define the vision of the further usage of credible and scalable ML/DL-based heat-transfer modeling.
Alisha et al. (Tue,) studied this question.