Calibration of the k– ω shear stress transport turbulence model for transitional boundary layer flows over flat plate and turbulent backward facing step flows

Reynolds Averaged Navier–Stokes turbulence models have been used over several decades in industrial and academic applications to predict turbulent flow fields. Adequacy of the predefined closure coefficients has received attention in the research community with increasing computational power in the last 20 years. In this study, the transitional and standard k–ω shear stress transport turbulence models are calibrated for flows over flat plate and backward facing step using six benchmark cases available in the literature. A systematical calibration methodology is presented, consisting of validation studies, global sensitivity analysis, reduced order modeling with Deep Neural Networks, and multi-objective optimization algorithm. Moreover, three different approaches are suggested to examine different aspects of the applicability of the calibrated closure coefficients. The results of the study demonstrate that the closure coefficients can be successfully calibrated using multi-objective optimization to improve the prediction of wall shear stresses and velocity profiles independent of the quantity of selected test cases. Comparison between different calibration approaches and their performance on different objective functions indicated their advantages and disadvantages. Calibrated values of the holistic calibration approach for the β1, β*, a1, b1, β2, and γ1 were found to be 6%, 25%, 24%, and 16% higher and 7% and 25% lower, respectively, than their default values.