Computational Fluid Dynamics is a powerful tool used on a daily basis by designers and researchers in the advancement of propulsion technology. As hardware and software technology continue to evolve, the impact on propulsion systems has the potential to be disruptive. With continued development of High Performance Computing, Large Eddy Simulation, High-Order unstructured grid algorithms, optimization and uncertainty quantification it is conceivable that a new frontier of simulation based research, analysis and design capability is in the foreseeable future. In order to accelerate this development collaboration between industry, academia and government labs is required. NomenclatureBR = Bypass Ratio c = Chord Length CFD = Computational Fluid Dynamics DNS = Direct Numerical Simulation HLES = Hybrid Large Eddy Simulation HO = High-Order HPC = High Performance Computing HPC = High Pressure Compressor HPT = High Pressure Turbine IDDES = Improved Delayed Detached Eddy Simulation IDDES-T = Improved Delayed Detached Eddy Simulation with Transition LES = Large Eddy Simulation LPC = Low Pressure Compressor LPT = Low Pressure Turbine Ma = Mach Number 1 Principal Engineer high-fidelity CFD, Advanced Design Tools, 2 MDAO = Multi-Disciplinary Analysis and Optimization OPR = Overall Pressure Ratio Pt = Total Pressure RANS = Reynolds Averaged Navier Stokes Equations Re = Reynolds number Ro = Rossby Number S = Slot Height S1B = Stage 1 Blade S2B = Stage 2 Blade S1N = Stage 1 Nozzle S2N = Stage 2 Nozzle SFC = Specific Fuel Consumption St = Strouhal Number TIT = Turbine Inlet Temperature Tt = Total Temperature TVD = Total Variation Diminishing UQ = Uncertainty Quantification URANS = Unsteady Reynolds Averaged Navier Stokes Equations VKI = von Karman Institute WMLES = Wall-Modeled Large Eddy Simulation WRLES = Wall-Resolved Large Eddy Simulation

No abstract

Conventionally meridian, RANS or URANS simulations are used in design. However, their accuracy is not always satisfactory. Higher order methods like direct and large eddy simulation have shown improved results for various case. While traditionally many direct numerical simulation (DNS) and large eddy simulation (LES) studies have been conducted using highly tailored research codes. These codes often use high order methods with both low numerical dissipation and dispersion often in conjunction with structured grids. While these features allow for high efficiency and accuracy they require a considerable effort to obtain appropriate grids. The ongoing rapid increases in available computer power have allowed commercial software packages to also include LES using lower order methods and unstructured grids. These are more flexible and robust to apply but it is not clear how they compare to the higher accuracy research solvers. In this work an initial comparison between a structured research code and an unstructured commercial code is presented. To that end, the paper compares compressible LES conducted with a high-resolution research code and with a commercial code of a statistically 2D linear transonic high pressure turbine vane cascade. The geometry investigated is that of Arts and Rouvroit [1] at isentropic exit Reynolds number approximately 590,000 and isentropic exit Mach number 0.93 with a target inlet turbulence intensity of 4%. The results are also compared to Direct Numerical Simulation results, published by Wheeler et al. [2], for boundary layer and wake development for low levels of freestream inlet turbulence. In the research code inlet turbulence is generated synthetically while the turbulence grid comprised of bars is simulated in the commercial code. While the flow around the blade is in good agreement differences are present in the wake data.

A series of systematic computational studies have been conducted for transonic HPT nozzles and blades to evaluate the impact of free stream turbulence on boundary layer growth and downstream wake mixing. Transition modeling is first compared to measurements for an uncooled nozzle. The computational results are compared against measurements of loading, HTC, and wake predictions. The approach is then applied to a cooled trail-edge nozzle. The added complexity of cooling flow injection at the trail-edge showed an increase in deviation between the SST Transition model predictions and wake measurement. By applying a scale resolve model (WALE LES), wake mixing predictions are found within 1% of measurement. Finally, the computational approach is extended to a coupled uncooled nozzle/blade stage analysis. Given the favorable results for an uncooled nozzle using the SST transition model, preliminary assessment of boundary layer impact on HTC and the overall stage loss was made for the nozzle/blade stage design.