مسار بولونيا
EN

ايمان علي محمد

السنة عنوان البحث نشر البحث
2026 Thermo-Hydrodynamic Performance and CFD–ANN Surrogate Modeling of a Backward-Facing Step Channel with Circular-Arc Baffles Journal of thermal analysis and Calorimetry
encountered in thermal systems, but the separated recirculation region downstream of the step can weaken near-wall mixing and produce non-uniform heat transfer. This study aims to improve the thermo-hydrodynamic performance of a BFS channel using a smooth circular-arc baffle configuration and to develop a rapid prediction framework for design evaluation. The main novelty of the work is the combined CFD-ANN evaluation of a smooth circular-arc baffle as a passive flow-control element in the expanded section of a BFS channel for both detailed thermo-hydrodynamic analysis and rapid performance prediction. CFD simulations were performed for 5000 ≤ Re ≤ 20,000 and 0 ≤ h ≤ 5 mm under constant wall heat flux, producing a 96-case database. A compact multilayer perceptron ANN with a 2-16-8-3 architecture was trained to predict the Nusselt number (Nu), pressure drop (Δ p), and performance evaluation criterion (PEC). The ANN achieved R² values of 0.99892, 0.99875, and 0.99256 for Nu, Δ p, and PEC, respectively. The largest baffle height (h = 5 mm) provided the strongest heat-transfer enhancement, increasing the average Nusselt number and PEC by approximately 99.9% and 48.7%, respectively, relative to the smooth BFS case, while maintaining PEC > 1. The normalized log-polynomial correlations predicted all responses with MAPE values below 4%, supporting rapid engineering estimation.Keywords: Computational fluid dynamics; Artificial neural network; Backward-facing step; Circular-arc baffle; Heat transfer enhancement; Design correlation.