Abstract
<title>Abstract</title> <p>To address the limitations of traditional methods for obtaining the equivalent parameters of axial symmetric liquid storage tanks in aerospace engineering, such as limited applicability, low computational efficiency, and high R&D costs, this paper proposes a rapid modeling method for equivalent parameters of oscillation based on random forests. First, a fully parametric description system for axisymmetric tanks is constructed. Through seven geometric parameters with clear physical significance, a unified characterization of commonly used engineering tank configurations such as cylindrical, capsule, and spherical is achieved; Second, the SLOSH-ML boundary element simulation tool, verified with engineering accuracy, was used to generate a large-scale standardized sample dataset. On this basis, a random forest proxy model was trained to achieve millisecond-level high-precision prediction of equivalent swaying parameters; Finally, combined with Sobol global sensitivity analysis, the influence patterns and interaction effects of tank geometric parameters on the equivalent sway parameters were quantified. The results show that the agent model achieves an average coefficient of determination R² of 0.946 for eight target variables, with five parameters exceeding 0.98, successfully capturing strong nonlinear mapping relationships between tank geometric features and equivalent sway parameters of various levels. Sobol analysis results show that the fill ratio and length-to-slenderness ratio are the most critical design parameters affecting the equivalent model parameters of tank swaying, providing a clear theoretical basis for tank optimization design. The agent model proposed in this paper provides efficient support for rapid design and multi-scheme iteration of spacecraft storage containers, significantly shortening design cycles and reducing R&D costs.</p>