Thermal-hydraulic performance and multi-objective optimization using ANN and GA in microchannels with double delta-winglet vortex generators

• Published in: International journal of thermal sciences Vol. 193; p. 108489
• Main Authors: Li, Zhenzhou, Feng, Zhenfei, Zhang, Qingyuan, Zhou, Jianyang, Zhang, Jinxin, Guo, Fangwen
• Format: Journal Article
• Language: English
• Published: Elsevier Masson SAS 01.11.2023
• Subjects: Delta-winglet vortex generator; Heat transfer enhancement; Machine learning; Microchannel; Thermal-hydraulic characteristics; Heat transfer enhancement; Delta-winglet vortex generator; Microchannel; Thermal-hydraulic characteristics; Machine learning
• ISSN: 1290-0729; 1778-4166
• DOI: 10.1016/j.ijthermalsci.2023.108489

This study presents a multi-objective optimization approach to design microchannels with double delta-winglet vortex generators for obtaining the best hydrothermal performance under high heat flux. Grey relational analysis, neural networks, and Non-dominated Sorting Genetic Algorithm (NSGA-II) are utilized in the optimization process. Back-propagation neural network improved by genetic algorithm (GA-BP) is found to have superior generalization and prediction abilities for Nusselt number (Nu) and friction factor (f) compared to Back-propagation neural network (BPNN). The design parameters include the angle between vortex generator and bottom plane (θ1), the angle between the baseline of vortex generator and water incoming flow direction (θ2), the distance from the apex of the vortex generator to the baseline (h), and inlet velocity (vin). NSGA-II is employed to obtain the optimal design point (θ1 = 90°, θ2 = 45°, h = 0.3 mm, vin = 0.8 m/s). The hydrothermal performance under this design point is significantly improved, as compared to smooth microchannel, with a 128.1% increase in Nu, a 72.0% increase in f, and a performance evaluation criterion (PEC) of 1.89. Furthermore, flow and temperature fields, entropy production, local Nu, local f and axial vorticity are analyzed. The results highlight the potential of double delta-winglet vortex generators to improve overall thermal performance in high heat flux applications. [Display omitted] •Hydrothermal behavior of delta-winglet vortex generator in microchannel is studied.•The prediction performances of two neural networks are compared.•The neural network optimized by genetic algorithm are established.•Multi-objective optimization of thermo-hydrodynamic performance is conducted.•The maximum value of performance evaluation criterion reaches to 1.89.