A performance optimization method based on the flow field structure of the vortex tubes with ANN

Optimization of vortex tube performance based on the internal flow field structure is considered as an effective optimization strategy, yet limited by the complex flow process inside the vortex tube, and the unknown mutual influence mechanism of different geometrical and operational parameters on the performance, there still has few quantitative flow field calculation model and corresponding optimization method. Herein, the direct mapping relationship between different parameters and the dimensionless reverse flow radius, an important indicator of the internal flow field structure was dug out by constructing an artificial neural network (ANN) in this work. Eight variables, cold mass fraction (mu c), dimensionless axial position (Z), length to tube diameter ratio (Lvt/Dvt), inlet to cold outlet pressure ratio (pin/pc), cold orifice to tube diameter ratio (Dc/Dvt), Reynolds number (Re), geometrical swirl number (Sgeo) as well as vortex tube diameter (Dvt) were selected as the inputs. The MSE and the correlation coefficient of this ANN model were 0.00132 and 0.981, respectively. By comparing and corre-lating the flow structure calculated by this ANN model with the optimization work in the literature, it was showed that the boundary conditions where secondary circulation and stagnation point occur were in good agreement with the optimal parameters for maximum energy separation performance, then a multi-parameter co-optimization strategy of the vortex tubes was proposed. Further, mutual impact of different parameters to the flow structure was investigated, and the synergy of factors' effects was analyzed. This research provided an effective model and a new method for exploring the flow structure and improving the performance of vortex tubes.