A Data-Driven Equivalent Modeling Approach for Large-Amplitude Liquid Sloshing Under Microgravity Environment

The establishment of an equivalent model of liquid sloshing in spacecraft tanks is of great importance for the stabilization of spacecraft attitude motions and the design of attitude control system. In this paper, a composite moving pulsating ball equivalent mechanical model (MPBM) with parameter identification and neural-network based error correction is proposed. In this model, the MPBM parameters are identified using a genetic algorithm combined with a particle swarm optimization algorithm (GA-PSO). The associated model error is predicted using a gate recurrent unit (GRU) neural network and compensated. Numerical experiments have been conducted and simulation results have confirmed the accuracy and reliability of the composite model, as well as the fast estimation of sloshing force and moment compared with the original MPBM.