BAYESIAN CORRECTION OF SST MODEL FOR SHOCK WAVE-BOUNDARY LAYER INTERACTION FLOWS

Widely used turbulence models have uncertainties in the prediction of physical quantities such as force/heat due to the variability of their own closure parameters. The uncertainty quantitative analysis and improvement of turbulence models can effectively improve their prediction accuracy and credibility. For the common problem of column skirt surge/boundary layer interference flow in engineering, the shear stress transport (SST) turbulence model and the shear stress transport model modified by adding Quadratic Constitutive Relation (QCR) are selected as the research objects to quantify the uncertainty of the closure parameter, and the uncertainty of the closing parameter is investigated. The uncertainty of the turbulence model and the shear stress transport model with the addition of the Quadratic Constitutive Relation (QCR) correction are investigated to quantify the uncertainty of their closure parameters, and ultimately to realize the improvement of the turbulence model's aerothermal prediction. The specific steps include: firstly, the a priori samples are obtained through sampling to construct the proxy model, secondly, the sensitivity analysis of the model parameters is carried out to screen the key parameters, and finally, the correction of the model parameters is realized through the Bayesian inference method, and the applicability of the modified model is verified under similar working conditions. The results show that the turbulence model, after correction, has significantly improved the prediction ability for heat flux and pressure, while the two models have the same key parameters. In addition, Bayesian inference of the turbulence model results in significant changes in the calculated physical quantities, which affects the prediction of the separation zone and makes the heat flux and pressure predictions closer to the experimental values. Therefore, calibrating the turbulence model closure parameters using Bayesian inference and incorporating experimental data can effectively improve the turbulence model's ability to predict aerodynamic heat.