Using artificial intelligence algorithms to reconstruct the heat transfer coefficient during heat conduction modeling

Elzbieta Gawronska; Maria Zych; Robert Dyja; Grzegorz Domek

doi:10.1038/s41598-023-42536-w

Using artificial intelligence algorithms to reconstruct the heat transfer coefficient during heat conduction modeling

Sci Rep. 2023 Sep 15;13(1):15343. doi: 10.1038/s41598-023-42536-w.

Authors

Elzbieta Gawronska¹, Maria Zych², Robert Dyja¹, Grzegorz Domek³

Affiliations

¹ Faculty of Mechanical Engineering and Computer Science, Czestochowa University of Technology, Dabrowskiego 69, 42-201, Czestochowa, Poland.
² Faculty of Mechanical Engineering and Computer Science, Czestochowa University of Technology, Dabrowskiego 69, 42-201, Czestochowa, Poland. maria.zych@icis.pcz.pl.
³ Faculty of Mechatronics, Kazimierz Wielki University, Jana Karola Chodkiewicza 30, 85-064, Bydgoszcz, Poland.

Abstract

The article shows the usage of swarming algorithms for reconstructing the heat transfer coefficient regarding the continuity boundary condition. Numerical calculations were performed using the authors' own application software with classical forms of swarm algorithms implemented. A functional determining error of the approximate solution was used during the numerical calculations. It was minimized using the artificial bee colony algorithm (ABC) and ant colony optimization algorithm (ACO). The considered in paper geometry comprised a square (the cast) in a square (the casting mold) separated by a heat-conducting layer with the coefficient [Formula: see text]. Due to the symmetry of that geometry, for calculations, only a quarter of the cast-mold system was considered. A Robin's boundary condition was assumed outside the casting mold. Both regions' inside boundaries were insulated, but between the regions, a continuity boundary condition with nonideal contact was assumed. The coefficient of the thermally conductive layer was restored using the swarm algorithms in the interval [Formula: see text]] and compared with a reference value. Calculations were carried out using two finite element meshes, one with 111 nodes and the other with 576 nodes. Simulations were conducted using 15, 17, and 20 individuals in a population with 2 and 6 iterations, respectively. In addition, each scenario also considered disturbances at 0[Formula: see text], 1[Formula: see text], 2[Formula: see text], and 5[Formula: see text] of the reference values. The tables and figures present the reconstructed value of the [Formula: see text] coefficient for ABC and ACO algorithms, respectively. The results show high satisfaction and close agreement with the predicted values of the [Formula: see text] coefficient. The numerical experiment results indicate significant potential for using artificial intelligence algorithms in the context of optimization production processes, analyze data, and make data-driven decisions.