Machine learning regression-CFD models for the nanofluid heat transfer of a microchannel heat sink with double synthetic jets

Abstract

A comprehensive analysis consisting of computational fluid dynamics (CFD) and machine learning algorithms (MLAs) is conducted to study the effect of geometrical and operational parameters on nanofluid heat transfer in a microchannel heat sink (MCHS) with double synthetic jets (SJs). A parametric CFD study is initially carried out on insert types (dimple/protrusion), insert arrangements (inline/staggered), insert diameters, and jet phase actuation (in-phase/180° out-of-phase actuation). Four popular regression models used to save the computational cost are k-nearest neighbor (k−NN), random forest (RF), Gaussian process regression (GPR), and Multi-layer perceptron (MLP). In the MCHS with inactive jets, the heat transfer coefficient (HTC) is enhanced by 104.8% by adding 5% alumina particles and inline protrusions (D = 0.26 mm). Staggered arrangements also show a higher heat transfer rate and pressure drop ratio. In the case of active jets, adding nanoparticles and staggered dimples significantly reduces the maximum temperature. The k-NN regression model shows more accurate predictions than the other MLAs. The conductive heat transfer rate is maximized by 53.99% in the in-phase actuation cases. The optimum results are obtained at the inline dimple set with D = 0.265 mm in the in-phase actuation case.