CFD analysis and design of bypass dual throat nozzle for high-performance fluidic thrust vectoring

Abstract

The purpose of the study is to investigate detailed flow properties of the bypass dual throat nozzle (BDTN) for fluidic thrust vectoring, and to find an optimal geometry to maximize its performance. The performance metrics of the BDTN are defined as the thrust efficiency and flow deflection angle at the nozzle exit. Given the nozzle pressure ratio (NPR), secondary flows injected from the bypass duct of the nozzle create circulatory flows in the nozzle cavity, produce complex interactions of shock and expansion waves, and deflect the directions of the exit flows. To identify key parameters for the BDTN performance, a sensitivity study is carried out using the traditional finite difference method as well as the AI-assisted Shapley additive explanation methods with respect to geometric variables of the BDTN. For the design optimization, a total of eight geometric parameters were chosen including an upstream convergent angle (θ1), a bypass injection angle (θ2), cavity divergence and convergence angles (θ3 and θ4), upstream and downstream throat diameters (d2 and d3), bypass channel diameter (d4), and cavity divergence length (l1). Those parameters were varied by 10∼20 % of the baseline values to create more than 100 random BDTN geometries which were solved by the full CFD analysis. The multi-variate Gaussian process regression (GPR) model was developed by training the data as a surrogate model to the CFD analysis of arbitrary BDTN shape during the design iteration. Multi-objective optimization was conducted to generate the Pareto optimal front of multiple design candidates for maximum deflection angle and thrust values. The optimum BDTN geometry produced a deflection angle increased up to 13 %, while thrust value was slightly increased from that of the baseline by less than 1%. The approach provides a foundation for future research into adaptive nozzle designs responsive to real-time flow conditions, potentially expanding the applications of fluidic thrust vectoring. © 2024 Elsevier Ltd