Magnetic Resonance Velocimetry Measurements of Internal Blade Cooling Flow and Computational Fluid Dynamic Validation by Data Matching With the Experimental Data

Abstract

Abstract The optimal Reynolds-averaged Navier–Stokes (RANS) turbulence model to be used in a Computational Fluid Dynamics (CFD) simulation varies depending on the application. Conventionally, the model is selected from benchmark tests and experience, but its performance is difficult to predict. For this reason, this study presents a cost-effective CFD validation routine, which uses three-dimensional experimental velocity data obtained in replicas of the specific flow system. Magnetic Resonance Velocimetry is used as the measurement technique. Since the objective is only the validation of the turbulence model, the experiment and the simulation are performed with simplified flow conditions, hence stationary isothermal isovolumetric flow without inertial forces. The routine applies a data-matching routine to align the two three-dimensional data sets before they are interpolated on a common grid. Various error metrics are presented, which provide the degree of the CFD modeling error and indicate its source. For demonstration, the validation routine is used to evaluate RANS-CFD results of a three-pass internal cooling system of a high-pressure turbine airfoil used in a small industrial gas turbine. The simulations are performed with the eddy-viscosity-based turbulence model k–ω shear stress transport (SST), the Reynolds-stress Speziale, Sarkar and Gatski (SSG), and baseline-Explicit algebraic Reynolds stress model turbulence (BSL-EARSM) models. The results indicate strong local errors in the examined turbulence models. None of the models performed well enough, underlining that every RANS-CFD application needs to be validated.