Masters Theses

Date of Award

5-1990

Degree Type

Thesis

Degree Name

Master of Science

Major

Mechanical Engineering

Major Professor

Roy J. Schulz

Committee Members

Roger Crawford, R.L. Young

Abstract

The calibration of the axisymmetrical Navier-Stokes computational fluid dynamics (CFD) program PARC2D for predicting steady state diffuser performance is presented. The calibration effort used data from available literature and was limited to an axisymmetrical diffuser system that used single specie, single stream, nonreactive flow. The calibration range included a convergent/divergent nozzle with an exit area to throat area ratio of 9, a diffuser with a diffuser area to nozzle throat area ratio of 36 with a diffuser length to diameter ratio of 4, and a diffuser with a diffuser area to nozzle throat area ratio of 16 with a diffuser length to diameter ratio of 12. Calibration parameters selected for the study were diffuser pressure rise ratio and diffuser wall static pres sure. The calibration effort was divided into three parts. First, a parametric study was conducted to observe the relative effects of various pertubations in the solution technique on the calculated flow field. Variations in the cell volume, the grid resolution in the diffuser, the initial conditions used to start the solution, and the turbulence mixing level were studied. Secondly, the PARC2D program results were compared to data for the selected diffuser configurations. Finally, a modification to the program which based the local intensity of turbulence in the flow on the local Mach number was made and the results were compared to data. The parametric study results demonstrated the effects of variations in the solution technique, and provided parameter sensitivity information that was used in the remainder of the research. The cell volume study showed that the computational cell volume could be reduced to half of the actual test hardware volume with no adverse effects. Reducing the cell volume to a minimum, however, caused a lower cell pressure to be obtained. Increasing the resolution of the grid in the diffuser by including 7,000 additional grid points did not provide any significant benefits, but increased the computer memory requirements by 50 percent. The initial condition study concluded that using a cell pressure in the initial conditions lower than the expected result provided a faster convergence rate than a higher cell pressure, and that filling the diffuser with supersonic flow provided approximately the same convergence rate as specifing that there was not any flow at all in the diffuser. The turbulence mixing level study showed the effect of varying the level of the turbulence viscosity on the solution. As the level of turbulent viscosity was increased, the final value of the calculated cell pressure was lowered. In the actual calibration of the PARC2D program to test data, the PARC2D program predicted diffuser pressure rise ratio to within 5 percent of the test data for started diffuser operation. The wall static pressures, however, indicated that the PARC2D program overpredicted the shock strength in the diffuser. For conditions where the diffuser was unstarted and the nozzle flow was separated from the nozzle wall, the PARC2D cell pressure predictions were approximately 30 to 60 percent higher than the test data. The qualitative trend of the diffuser wall static pressures was correct, but quantitatively the pressures were too high because of the incorrect cell pressure. For unstarted diffuser operation with the nozzle flowing full, a stable solution was not obtained. The calculated cell pressure varied in a cyclic nature about the test data with a deviation of ±30 percent. Because of the large number of iterations required to complete the convergence, final resolution of the predicted cell pressure was not possible. The modification to the algebraic turbulence model was promising in that the largest time step size that permitted a stable solution increased by a factor of five for the cases considered. The modification did not, however, provide the correct diffuser pressure rise ratio. For the two cases operating with the diffuser in the started mode, the modified algebraic turbulence model resulted in a cell pressure as much as 80 percent higher than the test data. The diffuser wall static pressures showed that the shock strength in the diffuser was overpredicted. The one unstarted diffuser case considered did not reach a stable solution and the cell pressure was as much as 60 percent higher than the test data.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS