Neural-Network-Assisted Optimization of Rectangular Channels with Intersecting Ribs for Enhanced Thermal Performance

Abstract

Optimization of rectangular cooling channels with intersecting ribs was performed to determine optimal rib configurations for thermal performance enhancement using Reynolds averaged Navier-stokes equations at Reynolds number of 10,000. The study aims to achieve two-fold objectives: (1) to maximize the thermal performance in three different rectangular channels with aspect ratios equal to 1, 2, and 4, respectively at a fixed hydraulic diameter and (2) analyze the trend in optimum design variables with change in aspect ratio. Angle of attack, position of the intersecting rib relative to channel center-line, and height of rib were selected as design variables for optimization. The objective function viz. thermal efficiency was formulated using radial basis neural network and particle swam optimization was used to determine the optimal point. The thermal performance of the channel was found to be sensitive to the chosen design variables, and strongly governed by rib-induced secondary motion. As a result of optimization, the optimal design showed an enhancement of 14.4, 12.8, and 16.2% in overall thermal efficiency for channels with aspect ratios 1, 2, and 4, respectively. Finally, to achieve maximum thermal efficiency, shift in optimum design variables was observed with change in aspect ratios of rectangular channels.