See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/200016703

Computational Simulation of Heavy Trucks

Conference Paper January 2007

DOI: 10.2514/6.2007-1087

CITATIONS READS

2 6

7 authors, including:

Kidambi Sreenivas Lafayette K. Taylor

University of Tennessee at Chattanooga 88 PUBLICATIONS 409 CITATIONS
71 PUBLICATIONS 179 CITATIONS
SEE PROFILE

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Kidambi Sreenivas
letting you access and read them immediately. Retrieved on: 21 April 2016
 45th AIAA Aerospace Sciences Meeting and Exhibit AIAA 2007-1087
8 - 11 January 2007, Reno, Nevada

Computational Simulation of Heavy Trucks

Kidambi Sreenivas*, D. Stephen Nichols, Daniel G. Hyams, Brent Mitchell, Shane Sawyer**, Lafayette K.
Taylor and David L. Whitfield
UT SimCenter at Chattanooga, University of Tennessee at Chattanooga, Chattanooga, TN 37403

Aerodynamic simulations were carried out for the Generic Conventional Model, a 1/8th
scale tractor-trailer model, that was tested in the NASA Ames 7x10 tunnel. The computed
forces and pressure coefficients are compared to experiment for the zero as well as the ten
degree yaw cases. DES versions of the one-equation Menter SAS and the two-equation k-
/k-
hybrid turbulence model are used in the simulations. Significant unsteadiness is
observed in the zero degree case which prevented a symmetry plane option from being used.
Overall forces as well as pressure distributions matched well for the zero degree case. For
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

the ten degree case, the forces were in reasonable agreement while the pressure distributions
indicated a need for tighter grid resolution as well as possibly advanced turbulence models.

I. Introduction
N 1997, fuel consumption among Class 8 trucks was 18 billion gallons1. At typical highway speeds (70 mph),
I65% of the overall output of the engine is used for overcoming aerodynamic drag2. Consequently, a reduction
in aerodynamic drag will result in substantial fuel savings. Given the large number of tractor/trailers on the road in
the United States, this could translate into a significant reduction in domestic fuel consumption as well as a
reduction in emissions that contribute to pollution.
The main contributors to the aerodynamic drag are the gap between tractor and trailer, the vehicle underbody and
the base flow region of the trailer3. Significant flow structures exist in these regions and several experimental
studies have been carried out to characterize them. One of the experimental studies used a 1/8th scale Ground
Transportation System (GTS) model that consisted of simplified tractor trailer geometry with a cab-over-engine
design and no tractor-trailer gap. This geometry was tested both at the Texas A&M University Low Speed Wind
Tunnel4 as well as the NASA Ames 7x10 tunnel5 and extensive data was collected for validation of computational
simulations. Modifications to the GTS geometry to include a tractor trailer gap were incorporated in tests carried
out at USC6 and the influence of the gap on the overall flowfield assessed. Detailed experimental studies were also
carried out on a realistic tractor trailer combination, the Generic Conventional Model (GCM), in the NASA Ames
7x10 as well as the 12 wind tunnel7.
Computational studies aimed at evaluating the capabilities of current flow solvers for the prediction of heavy
vehicle aerodynamics were carried out by a number of researchers8-13. Salari et al. 8 used the data from the GTS
experiments and showed that with an appropriate choice of turbulence model, the overall drag coefficient could be
predicted with reasonable accuracy. However, one of their key findings was that the details of the flow field,
especially in the base flow region, were not being captured accurately. Pointer8 used a commercial CFD flow solver
to study the GCM model in the 7x10 tunnel and the results indicated that the overall drag coefficient could be
predicted with reasonable accuracy. However, no detailed flow field comparisons were presented. Maddox et al.10
used another commercial CFD solver to simulate the flow around GTS model. They employed a detached eddy
simulation (DES) approach and showed that an improvement in the predicted pressure (especially toward the base of
the model) can be achieved. RANS simulations using the one-equation Spalart-Allmaras model and the two-
equation Menter k- model were reported by Roy et al.11. LES simulations of a truncated GTS model were carried

*
Associate Research Professor, Senior Member AIAA

Assistant Research Professor, Member AIAA

Associate Research Professor, Member AIAA

Research Associate
**
Graduate Research Assistant

Research Professor, Senior Member AIAA

Professor and Director, Graduate School of Computational Engineering,, Fellow AIAA

1
American Institute of Aeronautics and Astronautics

Copyright 2007 by the authors. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
 out by Ortega et al.12. Sreenivas et al.13 simulated the flow field around the GTS model using an unstructured flow
solver. They employed the DES version of the two-equation k-/k- turbulence model and showed that a single
vortex structure that was observed experimentally could be reproduced by time-averaging an unsteady RANS
simulation.
The present research effort focuses on Tenasi, a family of structured and unstructured flow solvers that have
been developed at the University of Tennessee SimCenter at Chattanooga. The unstructured flow solver is used to
simulate the flowfield around the GCM model inside the NASA 7x10 tunnel at zero and ten degrees yaw. Given
the high freestream Mach number that is present in the GCM experiments, a single parameter based global
preconditioned formulation termed the variable Mach number formulation is used.
The remainder of the paper is organized as follows: Section II contains a brief overview of the governing
equations. The details of the algorithm implemented in the flow solvers are presented in Section III. Results
including detailed comparisons of the flow fields are presented in Section IV and some conclusions are drawn in
Section V.

II. Governing Equations

Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

The governing equations for the variable Mach number formulation are the compressible RANS equations. The
incompressible governing equations are derived from the compressible equations by assuming that the flow is iso-
energetic and by introducing artificial compressibility. The final form of the governing equations can be written as
1
t Re



Q dV + F .
n dA = G . n dA (1.1)

where Q = [ , u , v, w, p ] for the variable Mach number case. In Eq.(1.1), F and G are the inviscid and viscous
 

fluxes respectively and n is the outward pointing unit normal to the control volume . The specifics of the
governing equations, including non-dimensionalizations, are in Sreenivas et al.15 and will not be repeated here.

III. Numerical Approach

The baseline flow solver in Tenasi employs a finite volume, implicit scheme with high resolution fluxes based
on Roe averaging and a Newton subiteration procedure for time accuracy. The linear system at each time step is
solved using a Symmetric Gauss-Seidel algorithm. Some of the features of the unstructured solver are highlighted
below. The details of the numerical algorithm are available in Sreenivas et al.15.

A. Residual Computation
The structured flow solver utilizes a cell centered formulation while the unstructured solver is based on a node-
centered formulation. Higher order spatial accuracy is achieved using a linear or quadratic reconstruction of the
dependent variables at the control volume faces and using these reconstructed values to evaluate the fluxes. Viscous
terms are evaluated using a directional derivative based approach. The gradients required for the variable
reconstruction are computed using an unweighted least squares approach while the gradients for the viscous terms
are computed using a weighted least squares approach.

B. Time Evolution
The flow solver employs a discrete Newton relaxation approach to solve the unsteady RANS equations.
Newtons method is used to drive the right-hand-side to zero. The flux Jacobians arising from this linearization can
be evaluated using numerical derivatives or the complex Taylor series method. They can also be replaced by
approximations which result in substantial savings in computational time. The resulting linear system is solved
using a Symmetric Gauss Seidel algorithm (point relaxation). For deforming grids, the Geometric Conservation
Law (GCL) has to be satisfied in order to prevent the occurrence of spurious sources in the solutions. This leads to
an additional contribution to the residual.

C. Turbulence Modeling
The turbulence models are implemented in a loosely coupled manner. The flow solver has the Spalart-Allmaras
model, the Menter SAS model, the k-/k- hybrid model (with and without SST), and the Wilcox Reynolds Stress
model. In addition, DES modes are available for the Spalart-Allmaras, the Menter SAS and the k-/k- hybrid
models19. No wall functions or transition models were used in either solver with integration being performed to the
wall with grids designed to give y+ values less than unity.

2
American Institute of Aeronautics and Astronautics
 D. Parallel Implementation
The parallel solution procedure consists of a scalable solution algorithm implemented to run efficiently on
subdomains distributed across multiple processes and communicating through MPI. The algorithm has multiple
nested kernels viz. time step, Newton iteration, LU/SGS iteration etc., and the subdomain coupling is at the
innermost level, i.e., in the solution of the linear system. A block-Jacobi type updating of the subdomain boundaries
ensures efficient parallelization with a small incremental cost incurred in terms of sub-iterations required to recover
the convergence rate of the sequential algorithm. Details about the parallel algorithm can be found in Hyams14.

IV. Results
The results presented here include the GCM at zero and ten degree yaw. These were two of the test cases chosen
from the available experimental data. The simulations were carried out to include the NASA Ames 7x10 tunnel as
well as the GCM geometry. The reason for this was the experimental pressure coefficient was referenced to a
specific location on the test section wall and including the tunnel was an easy way of comparing computed and
experimental pressure distributions. An initial simulation for the zero yaw case considered here was carried out
using a symmetry plane assumption. However, the flow field around the GCM turned out to be highly unsteady and
had significant vortex shedding, thereby destroying any flow symmetry. Consequently, the symmetry assumption
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

was abandoned and the entire geometry was simulated. For the ten degree yaw case, there is no symmetry and
therefore the entire geometry was simulated. A new grid had to be built for the ten degree case as the tunnel was
present. The grids for both the cases were generated using a combination of Gridgen (for the tetrahedral portion)
and HUGG20 (for inserting the viscous layers). The salient features of the grids used here are described in Table 1.
A three-view of the GCM geometry is shown in Figure 1. The relative locations of the various pressure taps on the
GCM tractor and trailer are shown in Figure 2.
All the cases presented here were run on 200 processors at a Reynolds number of 1.15x106 and a tunnel test section
Mach number of 0.15. The average y+ for all viscous surfaces was less than unity. Furthermore, all physical
dimensions are scaled by the GCM model trailer width (w = 12.75 inches).

Case Nodes Tetrahedra Prism Hex

Zero degree yaw 20,636,180 40,366,718 25,870,856 680,800
Ten degree yaw 18,937,191 41,828,550 23,253,152 0

Table 1: Grid Characteristics for the zero and ten degree yaw cases for the GCM inside the 7'x10' Tunnel

Figure 1: Three-view Drawing of the GCM Tractor-Trailer Configuration (measurements in cm) [Ref. 5]

3
American Institute of Aeronautics and Astronautics
 (a) Tractor Locations
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

(b) Trailer Locations

Figure 2: Relative Pressure Tap Locations on the GCM [Ref. 5]

A. Zero Degree Yaw

The solution for this case was
obtained using the one equation
Menter-SAS turbulence model that
was operated in the DES mode. An
initial solution was obtained by
running for 5000 time steps using a
fixed CFL of 25.0. After the initial
solution had reached a quasi-steady
state, the solver was switched to an
unsteady mode and the run was
continued for a further 47,500 time
steps. In the unsteady mode, a non-
dimensional time step of 10-3 was
used together with three Newton
subiterations to ensure time accuracy.
Second order accuracy in time and a
quadratic reconstruction in space
were used to ensure higher order
accuracy in time and space. The
Barth limiter was used to ensure that
the solution was well behaved. Figure 3: GCM Surface Geometry Shaded by Pressure Coefficient
Figure 3 shows the surface geometry of the GCM shaded by pressure coefficient. Figure 4 shows the unsteady axial

4
American Institute of Aeronautics and Astronautics
 force history as a function of time step. Initially it appeared as if the axial force was diverging, but upon continuing
the run, the axial force appears to have attained an almost periodic state. Indeed, a time average of the axial force
coefficient over the last 10000 time steps yields an average value of 0.4203 as opposed to an experimental value of
0.4060.
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

Figure 4: Unsteady Axial Force Coefficient History

The sources of the unsteadiness are illustrated by Figure 5, which shows axial velocity contours on a cutting
plane about half-way up the trailer. The main sources are the gap between the trailer and the tractor (tractor base
region) and the base region of the trailer. Furthermore, the GCM model is mounted on cylindrical posts which,
when placed in a crossflow, will create unsteadiness in the flow field. As can be seen from the figure, there is a
distinct pattern of vortex shedding originating from the tractor-trailer gap and the base flow of the trailer is
asymmetric too. The asymmetry on the shed vortex pattern between the two sides of the trailer as well as the
asymmetry of the base flow clearly indicates that imposing an arbitrary symmetry condition on the flow field is not
quite correct.

Figure 5: Instantaneous Axial Velocity Contours

In an effort to filter out the effects of the unsteadiness, the solution was time-averaged over 1000 time steps. This
frequency was picked arbitrarily, but it does match the frequency in the axial force history fairly closely. In the next

5
American Institute of Aeronautics and Astronautics
 set of figures, time-averaged and steady pressure coefficients are compared to experimental data. Figure 6 shows a
comparison of pressure distribution along the centerline of the tractor. The two different views are for the centerline
with the difference being that one is plotted against the vertical coordinate while the other is plotted against the
horizontal coordinate. As can be seen from the figure, the agreement is very good between the experimental data
and computations. The effect of unsteadiness is not very significant along the centerline of the tractor.
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

(a) Pressure distribution as function of vertical coordinate

(b) Pressure distribution as function of horizontal coordinate

Figure 6: Comparison of Tractor Centerline Pressure Distribution for 0 Yaw

6
American Institute of Aeronautics and Astronautics
 Figure 7 shows a comparison of pressure distributions for the trailer. Figure 7(a) is a comparison along the
centerline of the trailer. As can be from the figure, the effect of unsteadiness is not significant (especially over the
top of the trailer), although it has an effect on the bottom side of the trailer. This is very likely being caused by the
presence of the cylindrical posts that were used to mount the model to the floor of the wind tunnel. Figure 7(b)
shows a comparison for the side of the trailer at a location about half way up the trailer. Again, the pressure
distribution is not affected substantially by the time averaging. As can be seen from the figure, the comparisons
between the experimental and computed results are pretty good.
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

(a) Trailer Centerline

(b) Trailer Side (y/w = 0.9137)

Figure 7: Comparison of Trailer Pressure Distribution for 0 Yaw

7
American Institute of Aeronautics and Astronautics
 Comparisons of base pressures for the tractor as well as the trailer are shown in Figure 8. As can be seen, the effect
of unsteadiness is more visible here; however, one has to take account of the scale of the pressure distributions. In
the previous figures, they had a typical range of 1.0 or higher. In this case, the range of the pressure coefficient is
only 0.2, thereby potentially exaggerating some of the visual effect. In spite of this scale effect, the time averaged
values do seem to track the experimental pressure distributions better.
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

(a) Tractor

(b) Trailer

Figure 8: Comparison of Base Pressure Distributions for 0 Yaw

8
American Institute of Aeronautics and Astronautics
 B. Ten Degree Yaw
The solution for this case was obtained using the two-equation k-/k- turbulence model that was operated in the
DES mode. An initial solution was obtained by running for 20,000 time steps using a fixed CFL of 25.0 in a non-
DES mode. Then the DES mode was activated and the solution continued for a further 10,000 time steps at a fixed
CFL of 25.0. The solver was then switched to an unsteady mode and the run was continued for a further 30,000
time steps. In the unsteady mode, a non-dimensional time step of 5x10-4 was used together with three Newton
subiterations to ensure time accuracy. Second order accuracy in time and a quadratic reconstruction in space were
used to ensure higher order accuracy in time and space. The Barth limiter was used to ensure that the solution was
well behaved. A geometric overview of the GCM inside the tunnel is shown in Figure 9. Contours of instantaneous
(at time step 57,500; 27,500 time step unsteady) helicity and vorticity magnitude about half way up the trailer are
shown in Figure 10. As can be seen from the figure, there is a significant level of unsteadiness in the flowfield.
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

Figure 9: Two Views of the GCM Model in the 7'x10' Tunnel at 10 Yaw (Tunnel in Yellow)

(a) Helicity Contours (b) Vorticity Magnitude Contours

Figure 10: Instantaneous Helicity and Vorticity Magnitude Contours at y/w = 0.9137

Unlike the zero degree yaw case, the force history at the end of 30,000 unsteady time steps did not indicate any kind
of periodic behavior. For the results presented here, the solutions were averaged over 1000 and 5000 time steps, an
arbitrary choice. Furthermore, the experiments were run so that yaw sweeps were conducted. This resulted in data
being obtained for +10 as well as -10 yaw. Both of these experimental data are shown in the comparisons. This
serves two purposes: (1) it gives a measure of the error (side-to-side) in the experimental measurements and (2) The
trailer was not instrumented on both the windward and leeward sides to the same extent. Therefore, using data from
the positive and negative angles of yaw fills in the gap between the measurements.

9
American Institute of Aeronautics and Astronautics
 Figure 11 shows a comparison of pressure distribution along the centerline of the tractor. Both sets of experimental
data are shown along with the time averaged computed solutions. As can be seen from the figure, the agreement is
very good between the experimental data and computations. The effect of the choice in averaging period can be
seen in both the plots. The 5000 time-step averaged solution appears to match the experimental data slightly better
than the 1000 time-step averaged solution. There also appears to be a significant discrepancy between the
experimental data around the x/w = 1.5 or so (close to the base of the tractor). The computations match the +10
yaw results very closely in this area. Overall, the computations seem to lie between the experimental data over the
bulk of the centerline of the tractor.
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

(a) Pressure distribution as function of horizontal coordinate

(b) Pressure distribution as function of vertical coordinate

Figure 11: Comparison of Tractor Centerline Pressure Distribution for 10 Yaw

10
American Institute of Aeronautics and Astronautics
 A comparison of the trailer centerline pressure distribution is shown in Figure 12. The data is presented in two
different graphs for clarity. The computational results are repeated between the Figure 12(a) and (b), while only the
top of the trailer data is shown in Figure 12(a) and only the bottom of the trailer is shown in Figure 12(b). As can be
seen from Figure 12(a), the trend towards the base of the trailer is not captured correctly by the computed results.
There is also no significant difference between the 1000 and 5000 time-step averaged solutions. It is possible that
the base flow region has a very strong influence on the pressure distribution in the neighborhood of the trailer base
(some evidence of this can be seen in the zero degree yaw case too; Figure 7(a)). The comparison for the bottom of
the trailer fares much better with the trend as well as the magnitude of the pressures being captured reasonably well
(especially the 5000 time-step averaged solution).
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

(a) Top of trailer

(b) Bottom of trailer

Figure 12: Comparison of Trailer Centerline Pressure Distribution for 10 Yaw

11
American Institute of Aeronautics and Astronautics
 The next comparison shown is for a line that is halfway up the side of the trailer (Figure 13). As can be seen from
the figure, the experimental data from the 10 yaw provides a better picture of the overall pressure distribution on
the trailer. The computations are reasonably close to the experimental data over the bulk of the trailer. The 5000
time-step averaged solution has smoothed out a lot of the fluctuations that are visible in the 1000 time-step averaged
solution. The biggest discrepancy between the computed results and experimental data is at the front corner of the
trailer. The experimental data seems to indicate a larger low pressure region (possibly caused by a large-scale
vortex) while the computations indicate a much smaller low pressure region. Again, the 5000 time-step averaged
solution compares much better to the experimental data than the 1000 time-step averaged solution. One of the
reasons for this discrepancy could be that the grid resolution is inadequate in this area. Indeed, the surface grids for
the zero and ten degree yaw cases were the same. In the ten degree case, there is significant flow through the
tractor-trailer gap while that was not the case for the zero degree case and it could be necessary to resolve the flow
features (both in terms of surface and volumetric resolution) in order to capture the flow physics correctly.
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

Figure 13: Comparison of Trailer Pressure Distribution at y/w = 0.9137 for 10 Yaw

In addition to the lack of sufficient grid resolution, it is also possible that a more advanced turbulence model or
perhaps an LES approach could provide a better resolution of flow features. Furthermore, the time-averaging period
was chosen arbitrarily and that could have a potential impact on the comparisons. A comparison of the axial and
side forces is shown in Table 2. The numbers reported for the computed results are time-averaged over the unsteady
part of the run. As can be seen, there is quite a bit of variation between the two sets of experimental data, with the
computation being within 10% of the experiment.

Case Axial Force Coefficient Side Force Coefficient

Experiment (+10 yaw) 0.70980 1.19873
Experiment (-10 yaw) 0.73849 -1.22730
Computed 0.80906 1.31911
Table 2: Comparison of Forces for the 10 Yaw Case

12
American Institute of Aeronautics and Astronautics
 V. Conclusion
Computed forces and pressure coefficient distributions were compared to experimental data for the GCM tractor-
trailer configuration for zero and ten degrees yaw. Good agreement was obtained for the zero degree case while the
ten degree case was in fair agreement with experimental data. The base pressures were predicted reasonably well
for the zero degree case. The ten degree case pointed to the need for grid refinement in specific regions of the
computational domain. It is also possible that the overall agreement for both the test cases could be improved by
employing advanced turbulence models.

Acknowledgments
This work was supported in part by the US Department of Energy and the Oak Ridge National Laboratory
through UT-Batelle contract number 4000035270 with Dr. W. Keith Kahl as technical monitor. This support is
gratefully acknowledged. Thanks are also due to Dr. Rose McCallen, Dr. Jules Routbort and the late Dr. Sid
Diamond for their generous guidance and encouragement. This work was also supported by the THEC Center of
Excellence in Applied Computational Science and Engineering with Dr. Harry McDonald as technical monitor.
Their support is greatly appreciated.
Downloaded by Kidambi Sreenivas on March 22, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2007-1087

References
1
http://www.transportation.anl.gov/research/technology_analysis/truck_fuel_use.html
2
McCallen R., Couch, R., Hsu, J., Browand, F., Hammache, M., Leonard, A., Brady, M., Salari, K., Rutledge, W., Ross, J.,
Storms, B., Heineck, J.T., Driver, D., Bell, J., and G. Zilliac, Progress in Reducing Aerodynamics Drag for Higher
Efficiency of Heavy Duty Trucks( Class 7-8), SAE Paper 1999-01-2238.
3
Cooper, K.R., Truck Aerodynamics Reborn Lessons from the Past, SAE Paper 2003-01-3376, 2003.
4
Croll, R. H., Gutierrez, W. T., Hassan, B.,Suazo, J. E., and Riggins, A. J., Experimental Investigation of the Ground
Transportation Systems (GTS) Project for Heavy Vehicle Drag Reduction, .SAE Paper 960907, February 1996.
5
Storms, B.L., Ross, J.C., Heineck, J.T., Walker, S.M., Driver, D.M. and Zilliac, G.G., An Experimental Study of the Ground
Transportation System (GTS) Model in the NASA Ames 7-by 10-Ft Wind Tunnel, NASA/TM-2001-209621, February
2001.
6
Arcas, D., Browand, F., and Hammache, M., Flow Structure in the Gap Between Two Bluff Bodies, AIAA Paper 2004-
2250, June 2004.
7
Storms, B.L., Satran, D.R., Heineck, J.T., Walker, S.M., A Summary of the Experimental Results for a Generic Tractor-
Trailer in the Ames Research Center 7- by 10-Foot and 12-Foot Wind Tunnels, NASA TM-2006-213489, July 2006.
8
Salari,K., Ortega, J., and Castellucci, P., Computational Prediction of Aerodynamic Forces for a Simplified Integrated
Tractor-Trailer Geometry, AIAA Paper 2004-2253, 34th AIAA Fluid Dynamics Conference and Exhibit, Portland, Oregon,
June 2004.
9
Pointer, D.W., Evaluation of Commercial CFD Code Capabilities for Prediction of Heavy Vehicle Drag Coefficients,
AIAA Paper 2004-2254, June 2004.
10
Maddox, S., Squires, K.D., Wurtzler, K. E., and J.R. Forsythe, Detached-Eddy Simulation of the Ground Transportation
System, The Aerodynamics of Heavy Vehicles: Trucks, Buses and Trains, Springer, 2004, pp 89-104.
11
Roy, C., Payne, J., McWherter-Payne, M., and Salari, K., RANS Simulations of a Simplified Tractor/Trailer Geometry,
The Aerodynamics of Heavy Vehicles: Trucks, Buses and Trains, Springer, 2004, pp 207-218.
12
Ortega, J.M., Dunn, T., McCallen, R., and Salari, K., Computational Simulation of Heavy Vehicle Trailer Wake, The
Aerodynamics of Heavy Vehicles: Trucks, Buses and Trains, Springer, 2004, pp 219-233.
13
Sreenivas, K., Pankajakshan, R., Nichols, D.S., Mitchell, B., Taylor, L.K., and Whitfield, D.L., Aerodynamic Simulation
of Heavy Trucks with Rotating Wheels, AIAA Paper 2006-1394, January 2006.
14
Hyams, D.G., An Investigation of Parallel Implicit Solution Algorithms for Incompressible Flows on Unstructured
Topologies, Ph. D. Dissertation, Mississippi State University, May 2000.
15
Sreenivas, K., Hyams, D. G., Nichols, D. S., Mitchell, B., Taylor, L. K., Briley, W. R., and Whitfield, D. L., Development
of an Unstructured Parallel Flow Solver for Arbitrary Mach Numbers, AIAA Paper No. 2005-0325, January 2005.
16
Wilcox, D.C., Turbulence Modeling for CFD, 2nd ed., KNI, Inc., Anaheim, California, 2000, Chap. 6.
17
Spalart, P.R., and Allmaras, S.R., A One-Equation Turbulence Model for Aerodynamic Flows, AIAA Paper 92-0439,
January 1992.
18
Menter, F.R., Kuntz, M., and Bender, R., A Scale-Adaptive Simulation Model for Turbulent Flow Predictions, AIAA
Paper 2003-0767, January 2003.
19
Nichols, D.S., Hyams, D.G., Sreenivas, K., Mitchell, B., Taylor, L.K., and Whitfield, D.L., Aerosol Propagation in an
Urban Environment, AIAA Paper 2006-3726, July 2006.
20
Karman, S., Unstructured Viscous Layer Insertion Using Linear-Elastic Smoothing, AIAA Paper 2006-0531, January
2006.

13
American Institute of Aeronautics and Astronautics