Application of machine learning in heat transfer from correlations to design

Over the last two decades, there has been significant progress in using machine learning (ML) techniques for modeling heat transfer and thermal systems. ML techniques allow researchers to model the highly nonlinear and dynamic relationships between numerous variables which have been extremely challenging tasks when using traditional empirical correlations. This chapter briefly reviews several popular ML algorithms that have been adopted by heat transfer researchers including artificial neural networks, decision trees, convolutional neural networks, generative adversarial networks, and physics informed neural networks. Also, this chapter introduces several example studies where these ML techniques were used for modeling single-phase/two-phase convection heat transfer, heat transfer in composite materials and where the ML models were used for expanding the design space of heat sinks. These example studies demonstrate the capability of ML models to rapidly predict with hundreds of input variables or to design complex heat transfer systems which would take more than several hours or days with the conventional numerical simulations.