Pogo accumulator optimization based on multiphysics of liquid rockets and neural networks

Abstract

In this study, a numerical analysis of pogo instability in liquid propulsion rockets was conducted and an optimization of the pogo suppressor was attempted. Pogo analysis was carried out using numerical results obtained via the major models for fuselage structure, feedline, and propulsion systems. In the structural system, the fuselage vibration modes were obtained and the relevant meta-model was constructed using the modal superposition method. To obtain accurate results for the hydraulic transmission line modeling, cavitation effects were also taken into account. Thus, a numerical analysis was performed on a pump inducer to provide the quantitative information of the cavitation volume in the liquid-oxygen feedline. By employing the rocket combustion equations, it was confirmed that the dynamic response was fed back to the longitudinal characteristics of the fuselage structure. In addition, an accumulator was installed to suppress pogo instability. For design optimization, an artificial neural network was suggested by performing Latin hypercube sampling. The sampling verifies the convergence by the learning process. Finally, a multi-objective optimization for the pogo accumulator was achieved with the present meta-model.