Professional Documents
Culture Documents
Mark R. Anderson
1. Introduction
The importance of validation in confirming any modeling approach is well recognized, and CFD is
no exception. Obviously, the more pertinent it is to the relevant problem, the better suited it is for
building confidence in the approach. Fortunately, a significant body of work is available through
the open literature for turbomachinery validation. The majority of these data are carefully
structured tests that emphasize a few particular aspects of the flow physics, such as two-
dimensional performance. Less common, but still important, are complex problems involving full
system performance.
To accomplish effective validation from a code development point of view, it is essential that the
process start simply and become progressively more complex. Otherwise, issues such as problems
in the implementation and modeling are lost in the complexity of the solution. For example, a
CFD validation that shows a 4% difference in efficiency from the test results for a full radial
compressor stage would rarely give the analyst the insight needed to identify the root cause of the
problem, such as a boundary condition implementation or equation of state error.
Obviously, consistency in setting up the model problem with the test is essential. This goes much
further than making sure the geometry and basic flow conditions are the same. It is important that
the data be reduced in a consistent fashion to that quoted in the test. Even a relatively basic
quantity such as efficiency can vary according to averaging technique, location of the
measurement, and thermodynamic assumptions used. More subtle parameters, such as the
measurement of turbulence intensity and length scale, can affect results as well. Many earlier test
programs in turbomachinery dismissed this measurement as too esoteric to note, but flow behavior
can sometimes be significantly affected by it. Harder still to come by are test results with solid
data on the error bars of the measurements. It is important to note that differences in model and
test results are the products of both model and test error.
This paper presents some results of a systematic validation study conducted at Concepts NREC for
the purposes of validation and refinement of a Full Navier-Stokes CFD code. The solver, known
as the “MultiBlock” solver in the Pushbutton CFD®1 package, is an advanced, structured grid
solver developed specifically for turbomachinery. Some included features are: a 3rd order AUSM
TVD scheme (Liou, 2006), full real equation of state, two-phase equilibrium capability, five state-
of-the-art turbulence models, and highly automated pre- and post-processing.
All results shown are for 3rd order scheme settings with the Spalart-Allmaras (Spalart-Allmaras,
1992) turbulence model, unless otherwise noted. The general convention used in the graphs is that
the test results shown in symbols and CFD results are symbols connected with solid lines.
Perhaps the simplest turbulent flow solution that can be generated is the standard flat plate
solution. Despite its simplicity, a surprising number of issues come up when these cases are
scrutinized. The reason for this is the implementation of the wall functions or whether they are
even implemented at all. In fact, CFD solution results can show more sensitivity to the details of
the wall function than to the turbulence model itself. Strictly speaking, wall functions are not
necessary in a CFD code, but most practical problems require them. Without them, grid
resolutions must reach deep into the viscous sub-layer and require a huge number of grid cells.
Below is an implementation of the Spalart-Allmaras turbulence model using a standard wall
function (straight line) that is modified in the sub-layer region (curved line). The results show a
fairly consistent replication of the dimensionless velocity distribution, and as such, free the user
from overly serious dependency on grid resolutions.
Because the flat plate solution has little or no pressure gradient, one would not expect any
significant variation in results from one turbulence model to the next, as is seen below.
1
Pushbutton CFD is a registered trademark of Concepts ETI, Inc.
3. One-dimensionalized Flow
Several challenging examples of CFD solutions can be found in one-dimensionalized flow fields.
Although they are one-dimensional in nature, the solutions are typically two- or three-dimensional
CFD implementations that may or may not include boundary layer effects. One distinct advantage
of looking at problems such as these is that there are often closed form solutions to the flow field
or well-established empirical observations that can validate key models in the code.
The first example shown below is the rough duct example (Schlichting, 1951). Here, the
dimensionless pressure drop through the long pipe-like duct is plotted as a function of Reynolds
number. It is interesting to note that the left-hand point is a duct section about the size of a typical
coffee stirrer. The point on the right is one kilometer in length and has the diameter of a large
pipeline. The result shows good agreement with the empirically derived model. The roughness
effect is accounted for in an adjustment of the wall function. Despite this simple implementation,
the results are quite acceptable.
Another example specifically tests a two-phase equation of state model used in the CFD code. The
case is a converging-diverging supersonic nozzle case using steam. As the flow passes through the
nozzle, the kinetic energy increases and the thermal energy decreases, passing into the two-phase
domain. Results are shown for previously published experimental and numerical results (Morre et
al., 1973, and Kermani et al., 2003). The CFD follows the test quite well up to about the x=0.1
point, where there is a slight blip in the experimental results. The reason for this is the finite rate
nature of the droplet formation in the test, whereas the CFD results are implemented in an
equilibrium formulation.
Many, many cases relevant to turbomachinery can be solved using a fundamentally two-
dimensional approach. Blade-to-blade solvers and throughflow solvers are just two examples of
solution methods using a two-dimensional formulation that have been used industry wide for many
years. These solvers may account for three-dimensional effects (at least partially) through source
terms in the conservation equations. As mentioned previously, these cases can be invaluable for
model validation, since key flow phenomena can be present (such as turbulence, adverse pressure
gradients, and shock structures) that would be impossible to quantify in more complicated three-
dimensional solutions.
A simple diverging nozzle series of experiments was performed in the mid-1960s (Reneau et al.,
1967) that provides an excellent basis for examining the performance of turbulence models using
controlled, adverse pressure gradients. Figure 5 shows one of the simple geometric shapes used in
the study and the pressure recovery as a function of divergence angle of the side walls. The results
plotted are for the Spalart-Allmaras turbulence model. Generally, good agreement was shown,
with the trailing off of pressure recovery due to separation captured reasonably well.
A shape more directly relevant to turbomachinery is shown below in Figure 6. The case is of a
moderate Mach number and moderately loaded compressor profile tested in a wind tunnel (Stark,
Hoheisel, 1981, AGARD 1990).
Another compressor cascade tested in the mid-1990s (Steinert et al., 1996, and Steinert, Starken,
1996) shows very good agreement with the pressure distribution for two different incidence
angles.
Next is an example of a turbine blade tested in four separate wind tunnels in Europe in the 1990s
(Kiock et al., 1986). The results are extremely valuable for validation purposes, since they clearly
show the error bar in the testing series. The variation in results is particularly evident in the
adverse pressure region of the suction side, as seen in Figure 12 below. The CFD results pass
nicely through the test data “cloud”. Loss and flow angle shown in Figure 13 compare nicely as
well.
Fig. 12. European turbine cascade test (dots) and CFD results (solid lines)
Fig. 13. European turbine cascade test (dots) and CFD results with various
turbulence model solutions (solid lines)
A condensing steam profile was tested at the University of Birmingham that provided an excellent
test case for validating condensing flow cases (Bakhtar, 1994). The profile is that of a rotor or
stator tip that was tested in the subsonic to supersonic range. The results of pressure ratio around
the blade are shown in Figure 14. Note the blip in pressure similar to that seen on the converging-
diverging nozzle case that shows the effect of finite rate droplet kinetics in the test.
A study conducted at Cambridge University provided some high level validation data for a fully
three-dimensional case specifically related to turbine secondary flow (Hodson, Dominy, 1987).
The experiment consisted of an untwisted profile with a well-defined velocity distribution at the
inlet. Figure 15 shows the general shape of the complex flow field resulting from the turning of
the rotational flow.
The results from CFD show good quantitative and qualitative agreement despite the fact that this
type of solution is quite sensitive to grid resolution, since a high degree of refinement is needed
throughout the whole flow volume, not just the boundary layer region.
The Eckardt O compressor was a famous test done in Germany with very detailed measurements
for performance and internal flow distribution measured via a laser (Eckardt, 1975, 1976, 1980).
The series is one of the best documented compressor tests available in the open literature.
A radial turbine case solution, extensively tested and documented starting in the 1950s, is shown
below (Hiett, 1956, 1958, 1962, 1963).
The plot on the right-hand side shows the flow angles calculated at the inlet nozzles. The second
figure is the efficiency versus U/Ct ratio for several rotational speeds and pressure ratios. This is
interesting in terms of accuracy of the solution, as well as the extraordinary tendency for radial
turbine performance to collapse on this term. The right-hand plot shows the non-dimensional
mass flow versus pressure ratio.
6. Conclusion
The cases reviewed in this publication represent a small, but still significant, set of interesting
examples for CFD validation of turbomachinary problems. Extra care was taken to eliminate
superfluous errors that can result from inadequate grid resolution, poor grid quaility, or improper
boundary condition setup.
For validation to be truly useful, it must present the user with three things:
1) It must represent a statistically significant number of cases. Only then can the user be
sure that dumb luck did not merely give canceling errors.
2) It must reflect what a skilled user can reasonably expect to achieve without knowing the
answer beforehand. If a number of factors are tweaked to match the results, then the
exercise is really one of curve fitting and not validation.
3) It must consist of examples that are pertinent to the user’s problem.
These results generated through Concepts NREC’s Pushbutton CFD product represent some of the
best examples available for general turbomachinery validation. The results show very good
agreement overall, with specific cases highlighted to demonstrate interesting challenges that still
remain in CFD modeling.
7. Reference
1. Bakhtar, F., M., Ebrahimi, M., and Webb, R.A., 1994, On the Performance of a Cascade
of Turbine Rotor Tip Section Blading in Nucleating Steam Part 1: Surface Pressure
Distributions, Proc Instn Mech Engrs, Vol. 209, pp. 115-124.
2. Bardina, J.E., Huang, P.G., Coakley, T.J., 1997, Turbulence Modeling Validation, AIAA
97-2121.
3. Hodson, H.P., Dominy, R.G., 1987, Three Dimensional Flow in a Low-Pressure Turbine
Cascade at Its Design Condition, Journal of Turbomachinery, Vol. 109, April 1987.
4. Hodson, H.P., Dominy, R.G., 1987, Off Design Performance of a Low-Pressure Turbine
Cascade, Journal of Turbomachinery, Vol. 109, April 1987.
5. Eckardt, D., 1976, Detailed Flow Investigations Within a High-Speed Centrifugal
Compressor Impeller, Journal of Fluids Engineering, Sept 1976.
6. Eckardt, D., 1976, Detailed Flow Investigations Within a High-Speed Centrifugal
Compressor Impeller, ASME 76-FE-13.
7. Eckardt, D., 1975, Instantaneous Measurements in the Jet-Wake Discharge Flow of a
Centrifugal Compressor Impeller, Journal of Engineering for Power, July 1975.
8. Eckardt, D., 1980, Flow Field Analysis of Radial and Backswept Centrifugal Compressor
Impellers, Part 1: Flow Measurements Using a Laser Velocimeter, “Performance
Prediction of Centrifugal Pumps and Compressors”, ASME 1980.
9. Eckardt, D., private communication.
10. Hiett , 1962, First Report on Aerodynamic Performance Tests on Cold Rig, M.I.R.A.
Radial Inflow Turbine, Engineering Report DP.6845, Ricardo & Co Engineers.
11. Hiett, 1956, D.I.G.T. Radial Inflow Turbine, Scroll and Nozzle Performance, Part I,
Engineering Report 1365A, Ricardo & Co Engineers.
12. Hiett and Palmer, 1956, D.I.G.T. Radial Inflow Turbine, Cold Tests on Turbine B, Scroll
and Nozzle Performance, Part II, 1956, Engineering Report 1368, Ricardo & Co
Engineers.
13. Hiett and Palmer, 1956, D.I.G.T. Radial Inflow Turbine, Cold Tests on Turbine B,
Turbine A and Turbine D, Engineering Report 1381, Ricardo & Co Engineers.
14. Hiett, 1956, D.I.G.T. Radial Inflow Turbine, Mechanical Design and Operating
Experience of Cold Test Rig, Engineering Report 1382, Ricardo & Co Engineers, Sept.
1956.
15. Hiett, 1958, D.I.G.T. Radial Inflow Turbine , Cold Tests on Turbine B, and Turbine A
and Turbine D, Engineering Report 1412, Ricardo & Co Engineers, Dec. 1958.
16. Hiett, D.I.G.T. Radial Inflow Turbine, Cold Tests on Turbine C and Turbine A, Hot Tests
on Turbine A, Engineering Report 1436, Ricardo & Co Engineers.
17. Hiett, 1963, D.I.G.T. Radial Inflow Turbine, Cold Tests on Turbine BA, and Turbine C,
and Turbine A., Engineering Report 1467, Ricardo & Co Engineers.
18. Kermani, M.J. , Gerber, A.J., and Stockie, J.M., 2003, Thermodynamically based
moisture prediction using Roe’s scheme, Univ. of New Brunswick, Fredericton, Canada.
19. Kiock, R., Lehthaus, F., Baines, N., Sieverding, C., 1986, The Transonic Flow Through a
Plane Turbine Cascade as Measured in Four European Wind Tunnels, Journal of
Engineering for Gas Turbines and Power, April 1986, Vol. 108.
20. Liou, Meng-Sing, 2006, A sequel to AUSM, Part II: AUSM+-Up for All Speeds, Journal
of Computational Physics, 214, 137-170.
21. Mentor, F.R., 1993, Zonal Two Equation k-w Turbulence Models for Aerodynamic
Flows, AIAA Paper 93-2906.
22. Moore, M.J., Walters, P.T., Crane, R.I., and Davidson, B.J., 1973, Predicting the Fog
Drop Size in Wet Steam Turbines, Inst. of Mechanical Engineers (UK), Wet Steam Conf.
4, Univ. of Warwick, Paper C37/73.
23. Reneau, L.R., Johnston, J.P., and Kline, S.J., 1967, Performance and Design of Straight,
Two-Dimensional Diffusers, Journal of Basic Engineering, March, pp. 141-150.
24. Schuster, P., Schmidt-Eisenlohr, 1980, U., Flow Field Analysis of Radial and Backswept
Centrifugal Compressor Impellers, Part 2: Comparison of Potential Flow Calculation and
Measurements, Performance Prediction of Centrifugal Pumps and Compressors”, ASME
1980.
25. Schlichting, H., 1951, Boundary-Layer Theory, Springer Berlin / Heidelberg.
26. Stark, U., Hoheisel, H., 1981, The Combined Effects of Axial Velocity Density Ratio on
Compressor Cascade Performance, Transactions of the ASME, Journal of Engineering
for Power, Vol. 107, January 1981, pp. 247-255.
27. Spalart, P.R., Allmaras, S.R., 1992, A One-Equation Turbulence Model for Aerodynamic
Flows, AIAA 92-0439.
28. Steinert, Eisenburg, Starken, 1991, Design and Testing of a Controlled Diffusion Airfoil
Cascade for Industrial Axial Flow Compressor Application”, Journal of Turbomachinery,
Vol. 113.
29. Steinert, Starken, 1996, Off Design Transition and Separation Behavior of a CDA
Cascade, Journal of Turbomachinery, Vol. 118.
30. Test Cases for Computation of Internal Flows in Aero Engine Components, 1990,
AGARD Advisory Report No. 275.