In the context of global device miniaturization trend, there is an urgent need for heat dissipation in electronic devices, and microchannel heat sinks have attracted much attention due to their advantages of small size and high heat transfer capacity. This paper firstly introduces the application of Computational fluid dynamics simulation methods in microchannel research and various design cases, then elaborates the knowledge related to Machine Learning, including its development history, main tasks and common algorithms, and analyzes its performance in predicting the performance coefficients of microchannels. Finally, we discuss the feature importance analysis technique, point out that there are differences in the evaluation of feature importance in different models, and introduce the new SHapley Additive exPlanations technique and its limitations, and finally conclude: Computational fluid dynamics simulation is an important tool for microchannel simulation research, and then when predicting the performance parameters based on the obtained data, it is necessary to select appropriate Machine learning algorithms and extend the data set to improve the prediction accuracy, while the feature importance analysis helps to understand the input-output relationship, which should be combined with physics and ML to improve the analysis method for the optimal design of microchannel heat sinks.
Wang et al. (Thu,) studied this question.