Calibration and sensitivity analysis of under-expanded hydrogen jet CFD simulation based on surrogate modeling

Abstract

Accidental leaks from high-pressure hydrogen storage containers can lead to under-expanded jet flows, posing risks of fires or explosions. Practitioners traditionally harness Computational Fluid Dynamics (CFD) software to assess the hazards of such hydrogen releases. The Reynolds-Averaged Navier-Stokes (RANS) equations, often chosen by their efficiency and reliability and significantly influences CFD results. However, the closure coefficients within RANS harbor uncertainties. These coefficients, often defaulting to values ascertained from foundational flow experiments, may not generalize across diverse flow contexts, especially for intricate flow challenges like under-expanded hydrogen jets with high velocities and pressures. This study employs surrogate modelling to recalibrate these closure coefficients for CFD simulations of under-expanded hydrogen jets. Sensitivity analysis with Sobol indices quantifies their impact. The recalibration resulted in significantly improved accuracy, compared to their default values, of the CFD against literature reported experimental data. The results can enhance safety assessments and measures for hydrogen applications.