William Z. Strang

Cobalt: a parallel, implicit, unstructured Euler/Navier-Stokes solver

Abstract A parallel, implicit, unstructured Euler/Navier-Stokes solver has been developed that is accurate, robust and user-friendly. The parallel performance of the code is excellent, showing linear, and at times better than linear, scalability on the current parallel architectures. This, combined with the time savings afforded by the implicit algorithm, permits the routine analysis of large viscous problems on workstation-class processors.

Large-eddy simulation of turbulent mixing layers

Large eddy simulations of turbulent mixing layers starting from a laminar profile have been performed using an incompressible psuedo-spectral code and a compressible finite-volume code. The simulations were done at both moderateand high-Reynolds numbers (final full-width Reynolds number of w 11,000, and sa 10) at low Mach number (M = 0.1). The results of the simulations are compared with direct numerical simulations and experimental data.

A solution on the F-18C for store separation simulation using Cobalt(60)

A demonstration is presented of the ability of Computational Fluid Dynamics (CFD) methods to predict store carriage loads and support store trajectory generation. A complete, complex aircraft, the F/A- 18C, was modeled with actual stores in their carriage positions. Cobalt,, a parallel, implicit unstructured flow solver was used to calculate the flow field and resultant aerodynamic loads on grids composed of tetrahedral cells. Three grids were used to simulate three different flow field approximations. The first grid was a purely inviscid grid containing 3.15 million cells. The second grid was made up of 3.96 million cells clustered to capture viscous effects on only the store components. The third grid was a full viscous grid containing 6.62 million cells. Store carriage loads for two flight conditions were calculated and compared with wind-tunnel measurements and flight-test data for each of the above grids. The resulting carriage loads were used in a separate six degree-of-freedom (6DOF) rigid-body motion code to generate store trajectories. All CFD solutions were second-order accurate and run to steady-state with CFL numbers of one million. Turnaround times ranged from 6 to 21 hours, depending on the number of processors used.

Computation of Compressible Flows Through a Chemical Laser Device with Crossflow Injection

The Air Force Research Laboratory has an ongoing effort to develop an accurate and efe cient computational tool to support the development of advanced chemical oxygen/iodine laser (COIL)devices. In this study, a series of computational simulations have been performed to provide a better understanding of e uid dynamic phenomena withingeometriesassociatedwithCOILe owe elds.Theparallel,implicitunstructuredNavier ‐StokescodeCobalt 60 was used to compute laminar, turbulent, and unsteady e ows of helium within the research assessment and device improvement chemicallaser (RADICL)nozzle. Computational results showing details of thejetmixing interaction and topological structure are presented. The laminar and turbulent results obtained with Cobalt 60 are in excellent agreementwithmeasuredmasse owratesandsurfacepressuredataobtainedfromrecentcold-e owtestsperformed with the RADICL device. Insufe cient experimental measurement prevents the determination of whether or not transition occurs within the injector region. The laminar time-accurate results indicate small-scale unsteadiness in the frequency range of 200 kHz downstream of the nozzle throat.

Three-Dimensional Simulations of Buoyancy-Driven Flows Within High Pressure Compressor Bore Cavities

The quantitative prediction of buoyancy-driven conv ection within compressor rotor disk cavities has not previously been clearly demonstrat ed, limiting the ability of gas turbine engineers to efficiently design compressor rotors t hat are optimized for weight, cost, and durability. The standard practice is to rely on sca ling previous rotor designs whenever possible, and to validate the predicted thermals a posteriori through expensive instrumented engine tests. This process often leads to sub-optimal designs that fall short of their intended design requirements. This study p rovides an approach and the tools to overcome this shortcoming. Specifically, we developed a three-dimensional Computational Fluid Dynamics (CFD) methodology for buoyancy-dominated convective flows. We then successfully applied the methodology to a compressor/turbine bore cavity, capturing buoyancy-driven convective cells and predicting a bore cooling temperature rise in excellent agreement with engine test measurements. We next applied this methodology to an entire 14-stage compressor b ore cavity detailing the flow structures and thermal evolution of bore cooling fr om forward to aft stages. This not only provided new insight into the thermal environment within rotor cavities, but also yielded a valuable database to correlate pumped flow and co mpressor rotor rim heat transfer to the available buoyant force. Non-dimensional relati onships between pumped flows into enclosed rotor cavities, compressor rim heat transf er, and buoyancy force are presented, laying the foundation for an improved and more effi cient thermal modeling methodology to predict compressor rotor temperatures.