Florent Duchaine

An overset grid method for large eddy simulation of turbomachinery stages

A coupling method based on the overset grid approach has been successfully developed to couple multi-copies of a massively-parallel unstructured compressible LES solver AVBP for turbomachinery applications. As proper LES predictions require minimizing artificial dissipation as well as dispersion of turbulent structures, the numerical treatment of the moving interface between stationary and rotating components has been thoroughly tested on cases involving acoustical wave propagation, vortex propagation through a translating interface and a cylinder wake through a rotating interface. Convergence and stability of the coupled schemes show that a minimum number of overlapping points are required for a given scheme. The current accuracy limitation is locally given by the interpolation scheme at the interface, but with a limited and localized error. For rotor-stator type applications, the moving interface only introduces a spurious weak tone at the rotational frequency provided the latter is correctly sampled. The approach has then been applied to the QinetiQ MT1 high-pressure transonic experimental turbine to illustrate the potential of rotor/stator LES in complex, high Reynolds-number industrial turbomachinery configurations. Both wave propagation and generation are considered. Mean LES statistics agree well with experimental data and bring improvement over previous RANS or URANS results.

Computational-Fluid-Dynamics-Based Kriging Optimization Tool for Aeronautical Combustion Chambers

The current state of the art in computational fluid dynamics provides reasonable reacting-flow predictions and is already used in industry to evaluate new concepts of gas turbine engines. In parallel, optimization techniques have reached maturity and several industrial activities benefit from enhanced search algorithms. However, coupling a physical computational fluid dynamics model with an optimization algorithm to yield a decision-making tool needs to be undertaken with care to take advantage of the current computing power while satisfying the gas turbine industrial constraints. Among the many delicate issues for such tools to contribute efficiently to the gas turbine industry, combustion is probably the most challenging, and optimization algorithms are not easily applicable to such problems. In our study, a fully encapsulated algorithm addresses the issue by making use of a new multiobjective optimization strategy based on an iteratively enhanced metamodel (kriging) coupled to a design-of-experiments method and a fully parallel three-dimensional computational fluid dynamics solver to model turbulent reacting flows. With this approach, the computer cost needed for thousands of computational fluid dynamics computations is greatly reduced while ensuring an automatic error reduction of the approximated response function. Preliminary assessments of the search algorithm against simple analytical test functions prove the strategy to be efficient and robust. Application to a three-dimensional industrial aeronautical combustion chamber demonstrates the approach to be feasible with currently available computing power. One result of the optimization is that possible design changes can improve performance and durability of the studied engine. With the advent of massively parallel architectures, the intersection between these two advanced techniques seems a logical path to yield fully automated decision-making tools for the design of gas turbine engines.

Large-Eddy Simulation of Heat Transfer Around a Square Cylinder Using Unstructured Grids

This paper presents a method of large-eddy simulation on unstructured grids designed to predict the wall heat transfer in typical aeronautical applications featuring turbulent flows and complex geometries. Two types of wall treatments are considered: a wall-function model using a full tetrahedral grid and a wall-resolved method computed on a hybrid tetrahedral–prismatic grid. These two approaches are tested against the square-cylinder case at moderate Reynolds number (Re=22,050), in which many reference data are available for flow dynamics and heat transfer. Both predict accurately the unsteady flow around the cylinder and in its near wake, but only the wall-resolved approach reproduces the Nusselt-number global value and its spatial distribution around the cylinder wall. This latter method is used to investigate the coupling between periodic vortex shedding and wall heat transfer using a phase-averaged analysis.

Steady/Unsteady Reynolds-Averaged Navier–Stokes and Large Eddy Simulations of a Turbine Blade at High Subsonic Outlet Mach Number

Reynolds-averaged Navier–Stokes (RANS), unsteady RANS (URANS), and large eddy simulation (LES) numerical approaches are clear candidates for the understanding of turbine blade flows. For such blades, the flow unsteady nature appears critical in certain situations and URANS or LES should provide more physical understanding as illustrated here for a laboratory high outlet subsonic Mach blade specifically designed to ease numerical validation. Although RANS offers good estimates of the mean isentropic Mach number and boundary layer thickness, LES and URANS are the only approaches that reproduce the trailing edge flow. URANS predicts the mean trailing edge wake but only LES offers a detailed view of the flow. Indeed, LESs identify flow phenomena in agreement with the experiment, with sound waves emitted from the trailing edge separation point that propagate upstream and interact with the lower blade suction side. [DOI: 10.1115/1.4028493]

Massively parallel conjugate heat transfer methods relying on large eddy simulation applied to an aeronautical combustor

Optimizing gas turbines is a complex multi-physical and multi-component problem that has long been based on expensive experiments. Today, computer simulation can reduce design process costs and is acknowledged as a promising path for optimization. However, performing such computations using high-fidelity methods such as a large eddy simulation (LES) on gas turbines is challenging. Nevertheless, such simulations become accessible for specific components of gas turbines. These stand-alone simulations face a new challenge: to improve the quality of the results, new physics must be introduced. Therefore, an efficient massively parallel coupling methodology is investigated. The flow solver modeling relies on the LES code AVBP which has already been ported on massively parallel architectures. The conduction solver is based on the same data structure and thus shares its scalability. Accurately coupling these solvers while maintaining their scalability is challenging and is the actual objective of this work. To obtain such goals, a methodology is proposed and different key issues to code the coupling are addressed: convergence, stability, parallel geometry mapping, transfers and interpolation. This methodology is then applied to a real burner configuration, hence demonstrating the possibilities and limitations of the solution.

Towards Massively Parallel Large Eddy Simulation of Turbine Stages

The context of integrated numerical simulations of gas turbine engines by use of high-fidelity Computational Fluid Dynamic (CFD) tools recently emerged as a promising path to improve engines design and understanding. Relying on massively parallel super-computing such propositions still have to prove feasibility to efficiently take advantage of the ever increasing computing power made available worldwide. Although Large Eddy Simulation (LES) has recently proven its superiority in the context of the combustion chamber of gas turbine, methodologies need to be developed and start addressing the problem of the turbomachinery stages, if integrated simulations based on LES are to be foreseen. In the proposed work an in-house code and strategy, called TurboAVBP, is developed for turbomachinery LES thanks to the coupling of multi-copies of the unstructured compressible reacting LES solver AVBP, designed to run efficiently on high performance massively parallel architectures. Aside from the specificity of such wall bounded flows, rotor/stator LES type simulations require specific attention and the interface should not interfere with the numeric scheme to preserve proper representation of the unsteady physics crossing this interface. A tentative LES compliant solution based on moving overset grids method is proposed and evaluated in this work for high-fidelity simulation of the rotor/stator interactions. Simple test cases of increasing difficulty with reference numerical are detailed and prove the solution in handling acoustics, vortices and turbulence. The approach is then applied to the QinetiQ MT1 high-pressure transonic turbine for comparison with experimental data. Two configurations are computed: the first one is composed of 1 scaled stator section and 2 rotors while the second computation considers the geometrically accurate periodic quarter of the machine, i.e. 8 stators and 15 rotors to test scalability issues of such applications. Although under-resolved, the LES pressure profiles on the stator and rotor blades appear to be in good agreement with experimental data and are quite competitive compared to the traditional (Unsteady) Reynolds-Averaged Navier-Stokes (RANS or URANS) modeling approach. Unsteady features inherently present in these LES underline the complexity of the flow in a turbine stage and clearly demand additional diagnostics to be properly validated.Copyright

Analysis of high performance conjugate heat transfer with the OpenPALM coupler

In many communities such as climate science or industrial design, to solve complex coupled problems with high fidelity external coupling of legacy solvers puts a lot of pressure on the tool used for the coupling. The precision of such predictions not only largely depends on simulation resolutions and the use of huge meshes but also on high performance computing to reduce restitution times. In this context, the current work aims at studying the scalability of code coupling on high performance computing architectures for a conjugate heat transfer problem. The flow solver is a Large Eddy Simulation code that has been already ported on massively parallel architectures. The conduction solver is based on the same data structure and thus shares the flow solver scalability properties. Accurately coupling solvers on massively parallel architectures while maintaining their scalability is challenging. It requires exchanging and treating information based on two different computational grids that are partitioned differently on a different number of cores. Such transfers have to be thought to maintain code scalabilities while maintaining numerical accuracy. This raises communication and high performance computing issues: transferring data from a distributed interface to another distributed interface in a parallel way and on a very large number of processors is not straightforward and solutions are not clear. Performance tests have been carried out up to 12 288 cores on the CURIE supercomputer (TGCC/CEA). Results show a good behavior of the coupled model when increasing the number of cores thanks to the fully distributed exchange process implemented in the coupler. Advanced analyses are carried out to draw new paths for future developments for coupled simulations: i.e. optimization of the data transfer protocols through asynchronous communications or coupling-aware preprocessing of the coupled models (mesh partitioning phase).

Compatibility of Characteristic Boundary Conditions with Radial Equilibrium in Turbomachinery Simulations

Setting up outlet boundary conditions in configurations that have a strong rotating motion is a crucial issue for turbomachinery simulations. This is usually done using the so-called radial equilibrium assumption, which is used before the simulation and provides an approximate expression for the pressure profile to impose in the outlet plane. This paper shows that recent methods developed for compressible flows, based on characteristic methods, including the effects of transverse terms, can capture the radial equilibrium naturally without having to impose a precomputed pressure profile. In addition, these methods are also designed to control acoustic reflections on boundaries, and the present work suggests that they could replace classical radial equilibrium assumption approximations when nonreflecting boundary conditions are required at the outlet of a turbomachine simulation, for example, in large eddy simulation. This is demonstrated in two cases: 1) a simple annulus flow with a swirl imposed at the inlet and 2) a transonic turbine vane.

Comparative study of coupling coefficients in Dirichlet-Robin procedure for fluid-structure aerothermal simulations

This paper tests the performance of coupling coefficients of a Dirichlet-Robin transmission procedure in the context of steady conjugate heat transfer (CHT). Particular emphasis is put on the optimal coefficients highlighted recently in a theoretical study based on a normal mode stability analysis. This work can be seen as the logical continuation of that study in order to assess the relevance of the coefficients provided by the model problem in a realistic aerothermal computation. First, the numerical and physical CHT modeling methodologies are presented. Then, the optimal procedure applied to a Dirichlet-Robin algorithm (one-coefficient method) is briefly described. In order to gauge the ability of this model to predict the stability and convergence properties of a realistic case, it is compared on a heated cylinder in a flowfield test case. A series of five coupling coefficients and three Fourier numbers are considered. These parameters are introduced into the model problem as data to compute the amplification factor and the stability limits. The stability and convergence properties predicted by the model problem are then compared to those obtained in the CHT computation. This comparison shows an excellent overall agreement. Moreover, for all the Fourier numbers considered, the numerical solution is stable and oscillation-free when the optimal coefficient of the model problem is used. This would suggest that the one-dimensional normal mode analysis can provide relevant coefficients directly applicable to real CHT problems.

Large Eddy Simulation of a High Pressure Turbine Stage: Effects of Sub-Grid Scale Modeling and Mesh Resolution

The use of Computational Fluid Dynamics (CFD) tools for integrated simulations of gas turbine components has emerged as a promising way to predict undesired component interactions thereby giving access to potentially better engine designs and higher efficiency. In this context, the ever-increasing computational power available worldwide makes it possible to envision integrated massively parallel combustion chamber-turbomachinery simulations based on Large-Eddy Simulations (LES). While LES have proven their superiority for combustor simulations, few studies have employed this approach in complete turbomachinery stages. The main reason for this is the known weaknesses of near wall flow modeling in CFD. Two approaches exist: the wall-modeled LES, where wall flow physics is modeled by a law-of-the-wall, and the wall-resolved LES where all the relevant near wall physics is to be captured by the grid leading to massive computational cost increases. This work investigates the sensitivity of wall-modeled LES of a high-pressure turbine stage. The code employed, called TurboAVBP, is an in-house LES code capable of handling turbomachinery configurations. This is possible through an LES-compatible approach with the rotor/stator interface treated based on an overset moving grids method. It is designed to avoid any interference with the numerical scheme, allow the proper representation of turbulent structures crossing it and run on massively parallel platforms. The simulations focus on the engine-representative MT1 transonic high-pressure turbine, tested by QinetiQ. To control the computational cost, the configuration employed is composed of 1 scaled stator section and 2 rotors. The main issues investigated are the effect of mesh resolution and the effect of sub-grid scale models in conjunction with wall modeling. The pressure profiles across the stator and rotor blades are in good agreement with the experimental data for all cases. Radial profiles at the rotor exit (in the near and far field) show improvement over RANS predictions. Unsteady flow features, inherently present in LES, are, however, found to be affected by the modeling parameters as evidenced by the obtained shock strengths and structures or turbulence content of the different simulations.Copyright