Thermal Performance Analysis in a Lid‐Driven Square Enclosure With Various Shapes of Inner Block: A Comparative Study Using Artificial Neural Network and Lattice Boltzmann Method

Abstract

ABSTRACT This study coupling the lattice Boltzmann method (LBM) with Artificial Neural Networks (ANNs) to explore combined convection within a moving‐wall square cavity containing variably shaped heated blocks, targeting applications in thermal management of electronic components such as lithium‐ion batteries. This hybrid approach significantly reduces computational cost and time while enabling accurate prediction of thermal behavior for different geometries. The enclosure is stuffed with air ( Pr = 0.71) and cooled from its vertical sides by a constant cold temperature, , while the horizontal walls are thermally insulated, with the top wall is driven at a uniform velocity, . For Richardson numbers between 0.01 and 100, the investigation has been conducted for various geometrical shapes of the heated block under consideration, including square, rhombus, circular, horizontal, and vertical ellipses. The ANN is employed to forecast new cases, thereby reducing computational effort while also serving to validate the numerical results obtained. The findings are graphically displayed as stream function contours, temperature contours, mean fluid temperature, and average heat transfer (HT) in relation to the aforementioned controlling parameters. The results indicate that, across all Ri values, the horizontal elliptical block (HEB) achieves the highest HT rate, while the circular block (CB) exhibits the lowest. More precisely, changing the block shape from CB to the horizontal elliptical one (HEB) increases the HT rate to about 20.43% at Ri = 0.01. Finally, the ANN predictions showed excellent accuracy, with results closely aligning with numerical simulations, thereby confirming the reliability and robustness of the hybrid ANN–LBM approach.