Computational Fluid Dynamics (CFD)
CFD RESEARCH & DEVELOPMENT
Starting with potential flow solvers, increasingly complex and accurate inviscid, viscous and real-gas flow solvers have been developed for both multi-zone structured (the USA code) and unstructured (the ICAT code) grid topologies. Multidisciplinary simulations involving the coupling between aerodynamic loads and structural response have been performed using a modal approach for structural response. The CFD applications that the group has been involved in range from almost incompressible flows to hypersonic and flows with chemical reactions. In terms of complexity in geometry, the whole range of configurations from simple flat plates to mated space shuttle with plume (fig.1), B1-B (fig.2), X-33 (fig.3), and Turbomachinery systems (fig.4) have been dealt with.
In the area of preprocessing that involves geometry creation, grid generation, specification of boundary conditions, and GUI for solver input, the CFD group has developed the user-friendly tools UNISG and XUNISBC. UNISG is capable of generating multi-zone-structured grids as well as unstructured, hybrid grids. UNISG is the only tool known to us that has the ability to employ the template approach for automating the grid generation process using hybrid grids. Geometry creation and repair capabilities implemented in UNISG have helped Boeing engineers reduce turnaround time for complex flow simulations by orders of magnitude. XUNISBC simplifies the boundary specification process. In the case of multi-zone-structured grids, XUNISBC automatically generates the zonal connectivity information. Boundary conditions for other surfaces are easily specified using the interactive GUI. XUNISBC may also be used to convert a multi-zone-structured grid to the ICAT unstructured-grid format.
ICAT, a hybrid, unstructured-grid CFD solver, solves numerically the Reynolds's Averaged Navier-Stokes (RANS) equations on a computational domain consisting of any combination of tetrahedral, hexahedral, prismatic, pyramidal, or any other type of grid cells. A suite of turbulence models is available for simulating the effects of flow turbulence. Equilibrium-air and multi-species real-gas flows with chemical reactions may be simulated using ICAT. ICAT is one of the very few unstructured-grid solvers capable of simulating flow past bodies in relative motion. This special capability of ICAT has been quite extensively made use of by the Turbomachinery group at Boeing, Rocketdyne. Many unsteady load calculations for multi-stage Turbomachinery systems have been performed with ICAT. ICAT was instrumental in understanding certain unexpected dynamic behavior of a supersonic turbine.
Limited conjugate heat transfer simulations have been performed using ICAT. One example of such simulations is the flow induced by a heated vertical plate (fig.5). The RANS equations including the buoyancy terms were solved in the fluid region, with the heat transfer at the solid-fluid interface obtained from the temperature distribution and conductivity in the solid. In the solid, only the energy equation was solved, with the heat transfer at the interface obtained from the temperature distribution and conductivity in the fluid. Since the conductivity in the solid is orders of magnitude larger than the conductivity in the fluid, special averaging techniques had to be employed in the solid to accelerate convergence. Note that the procedure described above may be classified under the category of a "loosely coupled" approach. The solutions obtained for heated glass and ceramic plates are compared with measurements in fig.5. We plan to use this capability of ICAT to perform multidisciplinary, aerothermal simulation for a Turbomachinery system.
Velocity vectors colored by temperature.
Like conjugate heat transfer, aeroelasticity, the study of the interaction between aerodynamic loads and structural response, is another popular topic in multidisciplinary analysis. The CFD group at HyPerComp has had the opportunity to study this problem for the transonic flow over a wing with control surface deflection (fig.6). While the full potential approximation was employed for aerodynamic simulation, the structural response was modeled using a modal approach. It was shown that flutter-suppression might be achieved by applying a suitable control law.
Efficiency and feasibility of automating the simulation process for fluid flow analysis have already been demonstrated using UNISG for the optimization of a rotor blade in 3D (figs.6-8) and computation of unsteady loads for a multistage Turbomachinery system in 2.5D (fig.9-11). In a 2.5D approximation, the computational domain consists of only one cell in the transverse direction, with the cell volume varied in the flow direction in such a way as to approximate some 3 dimensional effects.
The template approach employed in the optimization of a 3D rotor-blade is shown in figs.6&7. Pressure contours computed from the ICAT solution for a rotor blade are shown in fig.8. Shape optimization was performed using the response surface approach. The geometry of the rotor blade was defined using 65 parameters. BAT, a module developed at Boeing, Rocketdyne, was employed in generating the geometry. BAT was coupled to UNISG to automate computational geometry creation, grid generation, and boundary-condition specification. Only hexahedral cells were employed in generating the grid. ICAT, was used to compute the cost function used in the optimization procedure. Input data for ICAT was developed manually. Flow-turbulence was simulated using a pointwise 1-equation turbulence model. Boeing developed optimization software, DACE, which was used to compute the optimum blade shape, employed a response surface approach. The values of the parameters that were modified to generate the response surface were predetermined and the process of generating cost functions was automated. Parallel processing was employed to accelerate the whole process.
Computation of unsteady loads for a 2.5D multistage Turbomachinery system is in some respect much more complex than a simple single component 3D rotor blade optimization. In this case the geometry involves 3 different components, namely, nozzle, rotor, and stator, and many stages. The process starts with two-dimensional geometry of the components (fig. 9a) consisting of only one element per stage. The first task is to make several copies of the elements and place them in the right locations to obtain the computational geometry for the multistage system (fig.9b). A hybrid-grid consisting of quadrilateral elements in the viscous region and triangular elements in the inviscid region is then generated automatically (fig.9c). The two-dimensional grid is then extruded in the third direction according to a user-specified length distribution in the flow direction to obtain the final computational domain. This step is followed by automatic assignment of boundary conditions (fig. 9d). The hybrid grid topology employed is shown in fig.10. Mach number contours at some instance in time, computed from the ICAT solution, are displayed in fig.11.
Note that the simulation of unsteady loads for a Turbomachinery system involves relative motion between the nozzles and rotors, and rotors and stators. While the nozzles and the stators are stationary, different stages of rotors may be rotating at different speeds. Data required to handle such complex boundary conditions are specified by the user before the boundary conditions are defined. Two files, one containing the (x,y,z) coordinates of the vertices of the grid cells, and the other containing the boundary condition information, are created by UNISG. A stand-alone tool is used to combine these two files, perform domain decomposition for parallel processing, and generate the grid-file required by the solver ICAT. It has been determined from our comparison of the manual process that was employed in the past and the automated process described above that the template approach has enabled us to decrease the turn-around time for preprocessing from 2 to 3 weeks to a few hours. The template approach has thus enabled us to obtain unsteady loads for many configurations and flow conditions in a timely manner to be of great value to the designers of Turbomachinery system.
© 2003 HyPerComp, Incorporated. Last Modified: June 3, 2010