Optimization of a Lobed Mixer with BP Neural Network and Genetic Algorithm
Yukuan Song,Zhijun Lei,Xin-Gen Lu,Gang Xu,Junqiang Zhu
DOI: https://doi.org/10.1007/s11630-022-1766-7
IF: 2.013
2022-12-23
Journal of Thermal Science
Abstract:A Sequential Approximate Optimization framework (SAO) for the multi-objective optimization of lobed mixer is established by using the BP neural network and Genetic Algorithm: the ratio of lobe wavelength to height ( η ) and the rise angle ( α ) are selected as the design parameters, and the mixing efficiency, thrust and total pressure loss are the optimization objectives. The CFX commercial solver coupled with the SST turbulence model is employed to simulate the flow field of lobed mixer. A tetrahedral unstructured grid with 5.6 million cells can achieve the similar global results. Based on the response surface approximation model of the lobed mixer, it is necessary to avoid increasing or decreasing α and η at the same time. Instead, the α should be reduced while the η is appropriately increased, which is conducive to achieving the goal of increasing thrust and reducing losses at the expense of a small decrease in the mixing efficiency. Compared with the normalized method, the non-normalized method with better global optimization accuracy is more suitable for solving the multi-objective optimization problem of the lobed mixer, and its optimal solution ( α =8.54°, η =1.165) is the optimal solution of the lobed mixer optimization problem studied in this paper. Compared with the reference lobed mixer, the α, β (the fall angle) and H (lobe height) of the optimal solution are reduced by 0.14°, 1.34° and 3.97 mm, respectively, and the η is increased by 0.074; its mixing efficiency is decreased by 4.46%, but the thrust is increased by 2.29% and the total pressure loss is decreased by 0.64%. Downstream of the optimized lobed mixer, the radial scale and peak vorticity of the streamwise voritices decrease with the decreasing lobe height, thereby reducing the mixing efficiency. For the optimized lobed mixer, its low mixing efficiency is the main factor for the decrease of the total pressure loss, but the improvement of the geometric curvature is also conducive to reducing its profile loss. Within the scope of this study, the lobed mixer has an optimal mixing efficiency ( ε =74.14%) that maximizes its thrust without excessively increasing the mixing loss.
engineering, mechanical,thermodynamics