Optimization design of modular-to-annular rocket-based combined cycle engine supersonic isolator based on Co-kriging model

Abstract

To address the challenges posed by the complex design space and the high computational cost of high-fidelity optimization for the multi-module turning isolation section of rocket-based combined-cycle (RBCC) engines, this paper establishes a Bayesian optimization framework based on the Co-Kriging model. The study employs a spine-rib curve collaborative parameterization method to characterize the duct shape and spatial turning features. By fusing low- and high-fidelity data sources, efficient global optimization is achieved under the drive of the expected improvement (EI) criterion. Results indicate that the optimization process converges with only 37 high-fidelity and 60 low-fidelity computational fluid dynamics (CFD) evaluations. The Co-Kriging model achieves a prediction accuracy of R2 = 0.9753, and the total pressure recovery coefficient of the optimized configuration improves by 9.5% compared to the baseline. Flow field analysis reveals that the aerodynamic gains stem from the smoothing of the longitudinal turning mode (L1 = L2 = 1.43 Rr) and the rationalization of the cross-section transition location (α = 0.65), which effectively suppresses wall shock wave/boundary layer interference and corner horseshoe vortex accumulation.