========================================================================= WIND Test Case ========================================================================= ------------------------------------------------------------------ Diffusing Pipe Flow ------------------------------------------------------------------ This test case computes incompressible turbulent flow in a conical diffuser, and is referred to as the Fraser (flow A) case from the 1968 AFOSR-IFP Stanford Conference on Computation of Turbulent Boundary Layers. The geometry consists of a straight section of pipe followed by a 5 degree half-angle conical diffuser. The core Mach number at the diffuser entrance is 0.15 (52 m/s), and the Reynolds number based on pipe diameter is approximately 500,000. (This is an NPARC model validation case, which can be found at http://info.arnold.af.mil/nparc/model/subsdif/.) ------------------------------------------------------------------ Input Parameters ------------------------------------------------------------------ We assumed the gas is ideal air (R = 1714.48 ft^2/s^2*deg R and a ratio of specific heats = 1.4). The following freestream static conditions were used: Mach Number = 0.15 Pressure = 14.7 psi Temperature = 530 R Using the viscosity obtained from Sutherland's law, the resulting Reynolds number is 1.034176E06/ft. The CFL number was set as follows: CFL = 0.5 for the first 2000 iterations CFL = 1.3 for the remaining iterations The Spalart-Allmaras turbulence model was used. The JMAX boundary was set to CONFINED OUTFLOW in GMAN. In the WIND input data file, the corresponding mass flow rate was set to 2.453256 lbm/sec; this value was computed from the experimental data. ------------------------------------------------------------------ Grid ------------------------------------------------------------------ The grid used is structured and has 121 axial points and 71 radial points. It is packed at the wall such that y+=1 at the first point from the wall. It is also packed axially at the inflow boundary to resolve large axial flow gradients. The grid was obtained from the NPARC validation archive (http://info.arnold.af.mil/nparc/model/subsdif/case01/plotx1.bin). The CFCNVT utility was used to convert plotx1.bin, which was in PLOT3D format (2D/multigrid/formatted), to a .cgd file (subsdif.cgd) which could be read by GMAN and WIND. ------------------------------------------------------------------ Boundary Conditions ------------------------------------------------------------------ The boundary conditions, which were set using GMAN are: (1) I1 - inflow (2) IMAX - confined outflow (3) J1 - inviscid wall (4) JMAX - viscous wall The units were set to feet. The resulting grid and boundary condition file is subsdif.cgd. The script file containing the above GMAN inputs is called gman.jou. ------------------------------------------------------------------ Output Files ------------------------------------------------------------------ Wind produced an output listing file (subsdif.lis) and a solution file (subsdif.cfl). These were used to produce a residual history file (subsdif.resl2) via the RESPLT utility. CFPOST was used to generate plot3d grid and solution files (subsdif.x.bin and subsdif.q.bin, respectively) and a file containing the axial mass flux distribution in the diffuser (subsdif.mflux). ------------------------------------------------------------------ Comparison of Data ------------------------------------------------------------------ The WIND computed results are compared with experimental data, and with an NPARC solution, computed using the Spalart-Allmaras turbulence model. The file velocity.ps contains velocity profiles at three locations in the diffusing pipe section: x = 0.117, 0.381 and 0.642 m, where x = 0.0 m is the start of the diffusing section. The file cf.ps contains a plot of the skin friction coefficient, and the file cp.ps contains a plot of the static pressure coefficient. For the velocity and static pressure coefficient profiles, the WIND and NPARC curves are essentially coincident. For the skin friction in the upstream portion of the pipe, WIND predicts lower values than NPARC. The WIND solution took about 60,000 iterations to reach convergence, where convergence is defined as the point where the skin friction coefficient stops changing. (This was an eyeball estimate, obtained by examining skin friction plots.) The NPARC solution was started from a converged k-omega solution: the k-omega calculation was run for 18,000 iterations, and the Spalart-Allmaras calculation was run for an additional 18,000 iterations. No real attempt was made to optimize convergence with either code. Below is a list describing the files in this directory README This file cf.ps Plot of skin friction coefficient (PostScript format) cfpost.jou Script file of inputs for running CFPOST cp.ps Plot of static pressure coefficient (PostScript format) gman.jou Script file of inputs for running GMAN resl2.ps Plot of WIND residual history (PostScript format) subsdif.cfl WIND flow field data file (common file 3 format *) subsdif.cgd Grid (common file 3 format *) subsdif.dat WIND input data file subsdif.lis WIND list output file subsdif.resl2 Residual history file (genplot format) subsdif.q.bin Plot3d solution file (IEEE unformatted) subsdif.x.bin Plot3d grid file (IEEE unformatted) subsdif.mflux Mass flux file (genplot format) vel.ps Plot of velocity profiles (PostScript format) * See the "Common File Programmers Guide", by McDonnell Douglas Aerospace, St. Louis, Mo 63166. If you have questions or comments about this case, please contact: Julianne Dudek, MS 5-11 NASA Lewis Research Center 21000 Brookpark Road Cleveland, OH 44135 (216) 433-2188 julie@lerc.nasa.gov