Study on heat transfer performance of steam-cooled ribbed channel using neural networks and genetic algorithms

This work aims to optimize the steam-cooled ribbed channels to achieve the best heat transfer performance. The combined effects of channel aspect ratio (W/H = 0.25-4), rib angle (α = 30-90°) and Reynolds number (Re = 10,000-100,000) on the heat transfer characteristics of steam-cooled ribbed channels were analyzed. The semi-empirical heat transfer correlation related to W/H, α and Re was developed. The back propagation neural network (BPNN) combined with genetic algorithm (GA) was used to predict the heat transfer coefficients and optimize the structural parameters of steam-cooled ribbed channels based on 90 groups experimental data, and an excellent BPNN model with a maximum prediction error of 1.9% was obtained. Flow fields in the steam-cooled ribbed channels were numerically calculated to explore the heat transfer enhancement mechanism of optimized channels. The results show that the average heat transfer coefficients of steam-cooled ribbed channels increase at first and then decrease with the increase of W/H and α. The optimized neural network has better prediction accuracy than that of the fitted empirical correlation. Reynolds number has a great influence on the optimal aspect ratio and rib angle of the steam-cooled ribbed channel. The optimal W/H and the optimal α increase from 2.23 to 3.35 and 41.12° to 60.89°, respectively, with the increase of Re. Large α within the range of 41.12-60.89° should be selected for cooling channels with relatively larger W/H in the range of 0.25-4. The enhancement of longitudinal secondary flows and the suppression of main secondary flows result in the heat transfer enhancement of the optimized channels.