A surrogate-based flow-field prediction and optimization strategy for hypersonic thrust nozzle

The thrust nozzle is a component that directly generates thrust, which is of great importance in the design of a hypersonic propulsion system. In this study, a surrogate-based flow-field prediction and optimization strategy for hypersonic thrust nozzles is proposed based on reduced-order modeling of the flow field. The effects of various methods on flow field prediction accuracy are studied, including domain decomposition-based proper orthogonal decomposition and Kriging-based surrogate models. A high-precision prediction of the viscid flow field can be achieved when an arbitrary nozzle contour parametrized by non-uniform rational B-spline is given. It is revealed that domain decomposition methods and improvements to the orthogonal basis effectively enhance the accuracy of flow field reconstruction. Compared to the full flow field surrogates, the near-wall region surrogates predict wall pressure more effectively and yield higher thrust accuracy. A hybrid infill sampling criterion is also proposed to improve the prediction accuracy of the model. Combining that with the surrogate-based flow-field prediction model and the intelligent optimization algorithms, an optimized nozzle contour is obtained, whose performance is 0.5% higher than that designed using the Rao method.