Generative adversarial networks with physical evaluators for spray simulation of pintle injector

Due to the adjustable geometry, pintle injectors are especially suitable for liquid rocket engines, which require a widely throttleable range. However, applying conventional computational fluid dynamics approaches simulate complex spray phenomenon in whole range still remains great challenge. In this paper, novel deep learning approach used instantaneous fields under continuous operating conditions is explored. Based on one specific type of neural network and idea physics constraint, Generative Adversarial Networks with Physics Evaluators framework proposed. The geometry design mass flux information embedded as inputs. After adversarial training between generator discriminator, generated field solutions fed into two evaluators. framework, conversation evaluator designed improve robustness convergence. A angle evaluator, composed down-sampling Convolutional Neural Network theoretical model, guides networks generate more closely according injection conditions. characterization simulated spray, including morphology, droplet distribution, angle, well predicted. This work suggests potential prior knowledge employment simulation flow fields.