Pintle Injector Spray Study via Physics-Based Spray Model and Generative Adversarial Networks

Pintle injectors are increasingly used in liquid rocket engines due to their favorable combustion stability and thrust modulation capabilities. This study presents a multi-stage numerical investigation of spray formation in gas-liquid discrete orifice pintle injectors, aimed at enabling high-fidelity yet computationally efficient simulations. A physics-based modification of the plain orifice atomizer model is proposed to characterize droplet velocity, angle, and size distributions at the injector exit, forming the basis for a Lagrangian spray injection model. Validation through CFD simulations in Ansys Fluent confirms the model’s consistency with experimental spray morphology and angle, using a Rosin-Rammler droplet distribution and a breakup model suited for subsonic crossflow conditions. The validated simulation data are further used to train a Physics-Informed Generative Adversarial Network (GAN) that integrates a spray angle evaluator as a physical loss function. The resulting lightweight GAN model shows promising accuracy in generating instantaneous flowfields while maintaining low computational cost, demonstrating the viability of combining physics-driven modeling with data-driven generative techniques for pintle injector flowfield prediction