Cooling performance analysis and structural parameter optimization of X-type truss array channel based on neural networks and genetic algorithm

• Published in: International journal of heat and mass transfer Vol. 186; p. 122452
• Main Authors: Xi, Lei, Xu, Liang, Gao, Jianmin, Zhao, Zhen, Li, Yunlong
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.05.2022; Elsevier BV
• Subjects: Arrays; Channels; Coefficient of friction; Cooling; Cooling channel; Design of experiments; Design optimization; Diameters; Flow and heat transfer; Fluid dynamics; Fluid flow; Gas turbine engines; Genetic algorithms; Maximization; Neural network; Neural networks; Nusselt number; Optimization; Parameters; Reynolds number; Truss array; Trusses; Turbine blade; Turbine blades; Truss array; Turbine blade; Cooling channel; Flow and heat transfer; Neural network; Optimization
• ISSN: 0017-9310; 1879-2189
• DOI: 10.1016/j.ijheatmasstransfer.2021.122452

•An innovative cooling channel filled with a novel X-type truss array structure was suggested to improve the cooling effect of turbine blade mid-chord region.•The comprehensive effects of each structural parameter and reynolds number on the cooling performance of X-type truss array channels were discussed.•A multi-input and multi-output neural network model with prediction deviations less than 3% for the X-type truss array channel was established.•The optimal values of d/H, Xs/C and Zs/C increase from 0.186 to 0.248, 1.696 to 2.482 and 1.0 to 1.945, respectively.•The heat transfer coefficient of the optimized X-type truss array channel increases by 44.50% to 145.49% and the comprehensive thermal coefficient increases by 4.16–9.56%. This study aims to optimize the structural parameters of an innovative cooling channel filled with a novel X-type truss array structure, so as to improve the cooling performance of the mid-chord region of gas turbine blades. In this study, the design of experiment (DOE) for influence parameters of X-type truss array channels was carried out, the variation ranges of influence parameters are as follows: Reynolds number (ReH, 10,000 to 60,000), truss rod diameter ratio (d/H, 0.1 to 0.3), transverse spacing ratio (Xs/C, 0.1 to 0.3) and streamwise spacing ratio (Zs/C, 0.1 to 0.3). The comprehensive effects of each structural parameter and Reynolds number on the flow and heat transfer performance of X-type truss array channels were analyzed. Then the multi-input and multi-output neural network model with prediction deviations less than 3% for the X-type truss array channels was established. Finally, the optimal structure parameters and arrangement of the X-type truss array channels were obtained by optimization design based on the genetic algorithm method, and have been verified by experimental measurement. The results show that the average Nusselt numbers and the friction coefficients of the X-type truss array cooling channel both rise with increasing d/H, drop with increasing Xs/C, and first rise then drop with increasing Zs/C; the comprehensive thermal coefficients first rise and then drop with increasing d/H, Xs/C, and Zs/C. Based on the maximization of average Nusselt number, the optimal values of d/H and Xs/C are 0.3 and 1, the optimal values of Zs/C increase from 1.179 to 2.201, as ReH increases from 10,000 to 60,000. According to the optimization results of maximization of comprehensive thermal coefficient, the optimal values of d/H, Xs/C and Zs/C increase from 0.186 to 0.248, 1.696 to 2.482 and 1.0 to 1.945 respectively with increasing ReH. Compared with the reference channel, the average Nusselt number of the optimized X-type truss array channel increases by 44.50 to 145.49% and the comprehensive thermal coefficient increases by 4.16–9.56% at different Reynolds numbers. The results may provide new ideas for the design of internal cooling structure in future advanced gas turbine blades.