Multi-objective optimization of symmetric sinusoidal wavy microchannel with multiple pairs of dual-rib using neural network-assisted NSGA-II algorithm

Microchannel heat sinks provide an effective solution to meet the increasingly demanding thermal management requirements of microelectronic components. This study proposes the integration of multiple dual-rib structures in symmetric wavy microchannels with hydraulic diameter of 0.2 mm. Under the inlet Reynolds number Re = 496, three design variables are selected: the transverse distance of dual-rib (D $\in$ [0, 0.12] mm), the longitudinal distance of dual-rib (S $\in$ [0, 0.12] mm), and the height of ribs (H_r $\in$ [0, 0.2] mm). Eighty-one samples are generated using the Latin Hypercube Sampling method for computational fluid dynamics simulations. Grey relational analysis is conducted based on the simulation results, followed by the construction of backpropagation neural networks (BPNN) and genetic algorithm-optimized BPNN (GA-BP) prediction model. The Non-dominated Sorting Genetic Algorithm II (NSGA-II) is employed for multi-objective optimization of the Nusselt number (Nu) and friction factor (f). The results show that all three structural parameters have strong correlations with Nu and f. Moreover, GA-BP outperforms BPNN in terms of generalization ability and prediction accuracy. The Pareto optimal solution set obtained through NSGA-II indicates that when the performance evaluation criterion PEC > 1.6, the optimal parameter ranges are: D $\in$ [0.0735, 0.0841] mm, S $\in$ [0.0117, 0.0255] mm, H_r $\in$ [0.0356, 0.0704] mm. Furthermore, the maximum PEC of the optimized channel structure reaches 1.7263.