1.

1 INTRODUCTION
The auto industry has seen a continual increase in the level of global competition.
The growth in the complexity of vehicle design and content has led to expensive and time –
consuming development processes. The large cost and long gestation implies very
significant risk for the automaker. At the same time, it is clear that technology will play an
ever increasing role as the basis of this global competition, requiring high quality products
that are safe to use, and economical to design and manufacture. The climate has been
favorable for the increasing use of Computer-Aided Engineering (CAE) tools for math-
based analysis of candidate designs and product features. The objectives of math-based
analysis include:
1. Shortening the product development process and reduce some hardware testing.
2. Developing high quality products through the evaluation of more design alternatives.
Aerodynamic development of motor vehicles is expensive. Much capital must be
invested in testing facilities such as wind tunnels and climatic tunnels. Secondly,
considerable costs result from the work itself. Finally, the development time may be
lengthened by aerodynamic work. The efforts to improve the aerodynamics of vehicles are
witnessed by the large numbers of wind tunnels constructed specifically for this purpose.
Nearly all major manufacturers have such facilities at their disposal or are currently building
them. Generally, the demands upon the quality of a wind tunnel increase with the
expectations placed upon the quality and reliability of the results. Similarly the development
costs increase steeply with the quality of the intended results.
The availability of a reliable numerical prediction method could greatly reduce
design costs by reducing the amount of wind-tunnel testing required.
Computational Fluid Dynamics, as one of the CAE tools, has been adopted to serve
this role for an increasing number of applications. When one examines the aerodynamic
development of vehicles, there are many reasons why CFD is expected to play an even more
important role in future years. Traditional aerodynamic development employs the use of
partial-scale or full-scale clay models of the proposed vehicle configuration. To ensure
geometric fidelity, these models include much of the vehicle’s details. As a result, they are
expensive to build and take considerable time to complete. Extensive use of large wind
tunnels for testing them is expensive and requires planned scheduling. Furthermore, certain
desired data may not be obtained from such models. For example, early assessment of the
potential of a shape for aerodynamic noise cannot be determined very accurately due to the

1
need for model construction. Finally in the process of detailed development of a model in a
wind tunnel, very elaborate flow measurement techniques may be required so that diagnosis
(e.g. for drag reduction) is possible. CFD, with its ability to display flow properties in great
detail, offers this additional capability.
Extensive research is going on in the field of automobile aerodynamics, especially of
cars, and there is a clear indication of the ever increasing popularity of numerical techniques
and CAE tools such as CFD to predict automobile characteristics in comparison to the
typical wind tunnel test methods. With the rapid progress of computer hardware and
software components, these simulation methods will become more predictable and accurate
and will completely replace wind tunnel testing in future.

2
 2.1 Introduction to Aerodynamics of Vehicles and its Importance
The performance, handling and comfort of an automobile are significantly affected
by its aerodynamic properties. A low drag is a decisive prerequisite for good fuel economy.
Increasing fuel prices and stringent legal regulations ensure that this long established
relationship becomes more widely acknowledged. But the other aspects of vehicle
aerodynamics are no less important for the quality of an automobile: side wind stability,
wind noise, soiling of the body, the lights and the windows, cooling of the engine, the gear
box and the brakes, and finally heating and ventilating of the passenger compartment all
depend on the flow field around and through the vehicle.

The flow processes to which a moving vehicle is subjected fall into three categories:
• Flow of air around the vehicle;
• Flow of air through the body;
• Flow processes within the machinery.
The first two flow fields are closely related. For example, the flow of air through the engine
compartment is directly dependent upon the flow field around the vehicle. Both fields must
be considered together. On the other hand, the flow processes within the engine and
transmission are not directly connected with the first two, and are not treated here.
The external flow subjects the vehicle to forces and moments which greatly influence the
vehicle’s performance and directional stability.
The aerodynamic drag D, as well as the other force components and moments, increases
with the square of the vehicle speed V:
D~ V2
With a medium size European car, aerodynamic drag accounts for nearly 80 percent pf the
total road resistance at 100 km/h. There is therefore much scope for improving economy by
reducing aerodynamic drag. For this reason drag remains the focal point of vehicle
aerodynamics, whether the objective is speed or fuel economy.
The complete expression for drag force is:
D = 0.5CD A V2
Where CD is the non-dimensional drag coefficient; A is the projected frontal area of the
vehicle (refer fig 2.1); and is the density of the surrounding air.

3
 Fig 2.1 Definition of the frontal A of a vehicle

The drag D of a vehicle is therefore determined by its frontal area A, and by its shape, the
aerodynamic quality of which is described by the drag coefficient CD. Generally the vehicle
size, and hence the frontal area, is determined by the design requirements, and efforts to
reduce drag are concentrated on reducing the drag coefficient.
The pressure difference between the upper and lower sides of the vehicle produces a
resultant force, at right angles to the direction of motion, which is called lift. As a rule the
lift is in the upward direction, i.e. it tends to lift the vehicle and therefore reduce effective
wheel loads. It is coupled with a pitching moment, which differentially affects the wheel
loads at the front and rear. Below 100 km/h, lift and pitching moment have only a small
effect upon the vehicle, even in a cross-wind. They do change the attitude of the car in
relation to the road and therefore slightly affect the aerodynamic drag. The reduction of the
wheel loads, however, is small in relation to the static wheel load and the directional
stability is hardly affected by the lift.
With cross winds the air flow around the vehicle is asymmetric to the longitudinal centre
plane. The shape of the car must be such that the additional forces and moments remain so
small that the directional stability is not greatly affected. First, the need to react to a cross-
wind of varying intensity and direction is inconvenient, as the driver must continually apply
steering corrections. Secondly, in very rare cases there is danger of total loss of control; this
can only be countered by suitable aerodynamic design.

4
 Soiling of the rear of the vehicle can be studied from the flow in the wake region.
Dust or dirty water is whirled up by the wheels and, dust particles and water droplets
distributed throughout the entire wake region by turbulent mixing, and deposited on the rear
of the vehicle. Since the flow pattern at the rear has a significant influence upon the
aerodynamic drag, soiling at the rear cannot be considered in isolation.
Fig 2.2 shows how the external flow field relates to flow processes inside the
vehicle. The flow into the radiator is determined by the flow pattern in front of the vehicle.
It can be seen that the stagnation point is at the level of the bumper, and that the air flow is
oblique to the openings above and below the bumper. The grill should be designed to direct
this air to the radiator, which is generally vertical, while keeping the pressure as low as
possible.

Fig 2.2 External flow field around the vehicle

The flow is attached in the region of the concave space formed by engine hood and the
windscreen. Here there is a pressure build up, which, can be utilized for driving air through
the heating and ventilation system. On most vehicles the fresh air inlet opening is positioned
in the middle of this area. However, at this point the pressure is dependent upon the driving
speed, which results in an increase of the fresh air flow as the speed increases, making
maintenance of the steady conditions in the passenger compartment quite difficult. If the
inlet openings for the fresh air are moved to points on the body which is at ambient
pressure, it is possible to separate the internal and external flow fields, at least while the
oncoming flow is symmetrical (no side wind). The fresh air fan, which must be
correspondingly larger, then provides a flow which is independent of the driving speed

5
(though only when the exit vents in the body are located in the areas of ambient pressure as
well).
The objectives of the aerodynamic design work are influenced by the type of the
vehicle under consideration. For instance, during the aerodynamic design of a passenger car,
the main consideration is drag. On a high-speed minibus or van, reduction of sensitivity to
cross winds may be the primary goal. Various solutions are available depending upon the
type of vehicle. On a racing car the objective will be to improve the traction of the tyres,
using negative aerodynamic lift regardless of the styling; the wings at the front and the back
have even become the characteristic of the modern racing cars. On the other hand
minimizing the drag of a passenger vehicle must be accomplished with less conspicuous
methods which conform to current styles.
The trend in the aerodynamic development of cars is summarized below. Fig 2.3
shows how drag decreased between 1920 and the mid-1970s. Owing to the lack of statistical
data only a general tendency can be outlined.
The reduction of the drag coefficient from CD 0.8 for cars in the 1920s to an average value
of 0.45 for cars of the 1960s and 1970s occurred in two stages. In the first, the period
between the two world wars, the cars were stretched and body details were rounded while
maintaining significant characteristics such as projecting fenders and headlights. In addition
to a lower drag coefficient of approximately 0.55, frontal areas were decreased, resulting in
a considerable reduction of the total aerodynamic drag.
The second stage in reduction of drag was reached with the introduction of the
pontoon body with its variants, the notchback, fastback and square back. By incorporating
the fenders and headlights in a closed body shape, it was possible to improve significantly
the flow of air around the vehicle. Using this design, drag coefficients of 0.4 to 0.5 were
reached, depending upon the detail design. The scatter range has remained change since
about 1960. However it is difficult to determine whether the reduction of drag resulted from
the influence of aerodynamics, from styling or from more advanced manufacturing
techniques.

6
Fig 2.3 Trend in aerodynamic drag coefficient against time, from 1920 to the mid-seventies

The average drag coefficient began to drop in 1978. The range of data – the scatter – is still
enormous. Even some contemporary cars have drag coefficients worse than 0.5, while the
best, the Opel Omega, has CD = 0.28!
With concept cars there is still room for further drag reductions. Drag figures of 0.14
(GM Aero 2002) and 0.15 (Ford Probe IV) have been claimed for operational cars.
Klemperer’s value of 0.15, established in 1922, at last seems attainable. Today a drag
coefficient of 0.30 is possible for without major and expensive technical compromise. In the
long run 0.20 might be achieved with production cars.
Increasing fuel prices will also encourage aerodynamic development of commercial
vehicles. Drag coefficients for box vans cover the range of 0.4 to 0.5. A value of 0.40 can be
realized without loss of transport space. Today the drag coefficients for heavy trucks lie
between 0.6 and 1.0. Considerable drag reduction can be achieved through the design of the
cab and the use of air deflectors. A value of 0.40 is a realistic goal for trucks and buses.
Aerodynamic development of motor vehicles is expensive. Much capital must be invested in
testing facilities such as wind tunnels and climatic tunnels. Secondly, considerable costs
result from the work itself. Finally, the development time may be lengthened by
aerodynamic work. The efforts to improve the aerodynamics of vehicles are witnessed by
the large numbers of wind tunnels constructed specifically for this purpose. Nearly all major
manufacturers have such facilities at their disposal or are currently building them.
Generally, the demands upon the quality of a wind tunnel increase with the expectations

7
placed upon the quality and reliability of the results. Similarly the development costs
increase steeply with the quality of the intended results.
The increase in the development costs resulting from higher development goals must
be counteracted by the availability of greater in-depth knowledge and the application of
theoretically sound development procedures. Numerical methods, which allow the
calculation of the external flow field or parts thereof, are under development. In the future,
improved predictions can be expected from new calculation procedures. Even so, they will
not replace testing and at most will facilitate test preparation and evaluation, and thus will
allow the test expenditure – in terms of costs as well as time – to be kept in check.

2.2 Flow Phenomena Related to Vehicles

The various flow phenomena related to vehicles can be divided into two groups.
These are (a) the external flow around a vehicle, including all details of its surface, and (b)
the internal flow through different systems such as carburetor, engine, exhaust system,
cooling system as well as the flow through the passenger cabin itself.
External Flow
The external flow around a vehicle is shown in fig. 2.4. In still air, the undisturbed velocity
V is the speed of the car. Provided no flow separation takes place, the viscous effects in the
fluid are restricted to a thin layer of a few millimeters thickness, called the boundary layer.
Beyond this layer the flow is inviscid and its pressure is imposed on the boundary layer

Fig 2.4 Flow around a vehicle (schematic)

Within the boundary layer, the velocity decreases from the value of the inviscid external
flow at the outer edge of the boundary layer to zero at the wall, where the fluid fulfills a no-
slip condition. When the flow separates at the rear part of the vehicle, the boundary layer is
‘dispersed’, and the flow is entirely governed by viscous effects. Such regions are quite
significant compared with the characteristic length of the vehicle. At some distance from the

8
vehicle, there exists no velocity difference between the free stream and the ground.
Therefore in vehicle-fixed coordinates, the ground plane is a stream surface with constant
velocity V and at this surface no boundary layer is present. This fact is very important for
the simulation of flow around vehicles in wind tunnels.
Mechanics of Air Flow
The flow around a vehicle is governed by two basic equations. The first equation is “Law of
Conservation of Mass” according to which:
w.s = constant
where; s denotes the local cross section of a small stream tube and w is the local velocity
which is assumed to be constant across s.
The second equation is “Newton’s Law of Momentum Conservation”. If this law is applied
to an inviscid flow it turns out that inertia forces and pressure forces are balanced. The
integration of the momentum equation along a streamline for incompressible flow leads to:
g = p + w2/2
Above equation is Bernoulli’s equation, which relates the pressure p and velocity w along a
streamline (p is static pressure, w2/2 is dynamic pressure, and g is total pressure).
In inviscid flow, the sum of static pressure and dynamic pressure is constant along a
streamline. Bernoulli’s equation indicates low pressures in regions of high local velocities
and vice-versa. If the flow comes to rest, w = 0, a so called ‘stagnation point’, as on the nose
of the vehicle (Fig. 2.4), the static pressure there will be equal to the total pressure, and this
is the highest possible pressure in the flow field.
The fundamental equations for inviscid flow may be applied to simple examples related to
vehicle aerodynamics and experimental techniques.
The two dimensional flow around a vehicle is shown in Fig 2.5. This flow is a
considerable simplification of the three dimensional flow around a vehicle. At the lower
surface of the vehicle, the pressure is higher than the free stream pressure, but for very small
ground distances, even suction may be present. At the upper surface, high pressures are
observed in

9
 Fig 2.5 Two dimensional flow around a vehicle
the region of the bonnet and windscreen, where as high suction(negative pressure) is found
at the cabin roof. Negative pressure means high local velocity at cabin roof. On the rear part
of the vehicle’s upper surface a steep pressure rise occurs, and it is the region where
considerable differences exist between the real flow of a viscous fluid and the inviscid flow.
The pressure level on the upper side of the vehicle is much lower than on the lower side.
This means that a net upward lift force acts on the vehicle.
Effects of Viscosity and Boundary Layer Development
The occurrence of drag in two dimensional incompressible flows can be explained by the
viscous effects. The flow in a boundary layer along a thin plate is shown in Fig. 2.6. The
corresponding external flow has parallel stream lines and constant velocity V and pressure
p . The viscous flow within the boundary layer fulfills the no-slip boundary condition along
the wall. In the front part of the plate the boundary layer flow is steady and (almost) parallel

Fig 2.6 Development of a boundary layer

10
to the wall. This state of the flow is called laminar. The thickness of the boundary layer
increase downstream according to

   

With increasing distance x, and kinematic viscosity and with decreasing free stream
velocity V , the boundary layer thickness increases. The laminar state of the boundary layer
flow is stable against disturbances for certain conditions only. At a distance x = xtr from the
leading edge of the plate, a transition to the so-called turbulent state of the boundary layer
takes place. The transition between the two states of the boundary layer flow is largely
governed by the value of the Reynolds number. For the flat plate transition occurs around
Rextr = 5 × 105
But this value applies only for negligible pressure gradient in the external flow. In case with
a pressure gradient, a pressure decrease in the flow direction leads to a stabilization of the
boundary layer, whereas an adverse pressure gradient cause an earlier transition to the
turbulent state. In general, for medium Reynold’s numbers transition from laminar to
turbulent occurs in the region of minimum pressure, and with increasing Reynold’s number
the transition point moves upstream.
Separation
Laminar and turbulent boundary layer flows depend strongly on the pressure deistribution
which is imposed by the external flow. For a pressure increase in flow direction the
boundary layer flow is retarded, especially near the wall and even reversed flow may occur.
This behavior is shown schematically in Fig. 2.7. It can be seen that, between forward and
reverse flow, a dividing streamline leaves the wall. This phenomenon is called separation.

Fig 2.7 Flow separation in adverse pressure gradient

11
There are two components of drag; friction drag and pressure drag. If a velocity gradient
du/dy is present in a viscous fluid at the wall, due to molecular friction a shear stress w acts

everywhere on the surface of a body. The integration of the corresponding force

components in the free stream direction leads to the “friction drag”. Integrating the force
components in the free stream direction, resulting from the pressure distribution, gives the
so called “pressure drag”. For blunt bodies the pressure drag is predominant. In turbulent
boundary layer flow, the friction drag is much higher than in the laminar case. This is
because the turbulent mixing process leads to velocity profiles with much steeper velocity
gradient at the wall than in laminar case.
The drag of bodies with finite thickness mainly consists of friction drag which is
small in all cases in which no flow separation occurs. This can be achieved by slender
shapes on the rear part of the body which produce only a weak pressure rise in the flow
direction. Shapes of this kind are aero foils and streamlined bodies.
Blunt bodies, such as circular cylinder, a sphere or a flat plate normal to the flow,
show quite different drag characteristics. On the rear part of such bodies in inviscid external
flow, extremely steep pressure gradients occur which lead to flow separation (Fig. 2.8).

Fig 2.8 Wake of a VW Golf I, smoke introduced into the wake

The pressure distribution is therefore considerably altered when compared with the
theoretical case of inviscid flow. As an example, Fig. 2.9 shows the pressure distribution for
a circular cylinder. In front part the pressure distribution is similar to that in inviscid flow,
where as on the rear part the flow separation leads to considerable suction.

12
Fig2.9 Pressure distribution and streamline pattern for a circular cylinder at different
Reynolds number (a) inviscid flow; (b) sub-critical flow, boundary layer laminar; (c)
supercritical flow, boundary layer turbulent

The pressure distribution is therefore asymmetrical with respect to the y-axis. Friction drag
also results from the wall shear stresses, but for blunt bodies the pressure drag is
predominant.
In general, the drag of a body may be written as
D = Df + Dp
Generally, a sudden change of the drag coefficient of a vehicle as a function of its Reynolds
number should be avoided. For this purpose, flow separation is fixed at certain points, for
instance at the upper edge of the rear sloping window, up to this point the shape of the body
should be designed so that the flow remains attached and the pressure rise is as large as
possible for various free stream conditions. The resulting wake should be as small as
possible to obtain low drag. The drag coefficients achieved by present day European cars
range from 0.30 to 0.52 (excluding sports and racing cars). In general, the dependence of
these drag coefficients on Reynolds number is very small and sudden changes do not occur.
This demonstrates that the predominant part of the drag of these vehicles is the pressure
drag. For some unconventional ‘streamlined’ body shapes, drag coefficients have been
measured in the region 0.15 to 0.27. For bodies of this type the portion of pressure drag is
relatively small. These drag coefficients thus contain a large proportion of friction drag and
therefore they depend noticeably on the Reynolds number.

13
 The flow separations that lead to a pressure drag can be divided into two different
types. As shown in Fig 2.10, the separation line may be located perpendicular to the flow
direction.

Fig 2.10 Flow separation on a bluff body

In this case, vortices are generated- the axes of which are also perpendicular to the outer
flow. Thus the velocity components parallel to the vortex axes are very small. A
symmetrical flow in the separated region as shown in Fig 2.10 exists only for small
Reynolds numbers. For larger Reynolds numbers, periodic vortex shedding occurs and flow
in the separated region is basically unsteady. The kinetic energy of the vortex field is
rapidly dissipated by turbulent mixing and irreversibly converted into frictional heat. This
leads to a considerable total pressure loss in the region behind the body and the
corresponding deficit in kinetic energy is equal to the work which is necessary to overcome
the pressure drag. Behind the body a wake is formed in which time-averaged, relatively
uniform suction and very low flow velocities are present.
The other type of flow separation is characterized by a separation line inclined with respect
to the oncoming flow, see Fig. 2.11. In this case, vortices are shed, the axes of which are
roughly parallel to the separation line. A considerable velocity component, parallel to the

14
Fig 2.11 Flow separation on a body with oblique blunt base (separation line at an angle to the
flow direction)

separation line and therefore in direction of the vortex axes, is present. Thus, a well ordered,
steady three dimensional flow separation is found. On the rearward surface of the body this
separated flow induces suction which leads to a pressure drag. On the inclined base of the
body the flow is attached. In the vicinity of the vortices the pressure distribution is
characterized by suction peaks. The flow field of the concentrated vortices, however,
contains a lot of kinetic energy which corresponds to the work necessary to overcome the
pressure drag.
Fig 2.12 shows the three dimensional flow separation at the rear of a vehicle.

Fig 2.12 Three dimensional flow separation on the rear of a vehicle

15
 On the drag problem of a body, it might be mentioned finally that the shape of a
body in front of the largest cross-section has only minor influence on the total drag. The
main contributions to the drag force originate from the rear part of the body. It is not
important to find a proper shape to divide the oncoming flow but it is very important to
design a rear body surface which brings the divided streamlines smoothly together.
Optimum shapes are ‘streamlined’ bodies having a very slender rear part.
Overall Forces and Moments
In addition to the drag discussed so far, other forces and moments occur on vehicles which
are shown schematically in Fig. 2.13. In symmetrical flow ( = 0) the drag D is
accompanied by a lift force L. Furthermore, a pitching moment M, with respect to the
lateral axis (y-axis) is present. The three components L, D and M completely determine the
vector of the resulting air force.

Fig 2.13 Forces and moments acting on a vehicle (c.g. = centre of gravity)

In cross wind conditions ( 0) an asymmetrical flow field is present around the vehicle. In
this case, in addition to the forces and moments mentioned so far, a side force Y is
observed. Furthermore, there occur a rolling moment R with respect to the longitudinal axis
(x - axis) and a yawing moment N with respect to the vertical axis (z - axis). Thus six

16
components L, D, M and Y, R, N determine the vector of the total force. Expressions for
these are given below:


(1)  = 




(2)  = 




(3) = 


 
(4)  = 




(5)
= 


 
(6)  = 


Here is the total length of the vehicle. is the air density, V is the free stream velocity, A
is the projected frontal area. CD, CL, CY, CM, CR, CN are corresponding coefficients.

2.3 Flow Field around a Passenger Car

The flow field around a vehicle is not yet fully understood, so a picture must be built
up from pressure distribution measurements, velocity field measurements and flow
observations on the vehicle surface. There are two types of separation. The first type has a
quasi-two-dimensional character. In this case the line of separation tends to run
perpendicular to the local flow direction. If reattachment occurs, so-called separation
bubbles are formed. Of course the flow inside the bubble, which is shed from a three-
dimensional body, is three-dimensional in nature. However, since the separation itself is
mainly two-dimensional with separation line normal to the flow and vortex axes parallel to
the separation line, it is designated ‘quasi-two-dimensional’. This type of flow can occur at
the leading edge of the front hood, at the sides on the fenders, on the cowl and on the front
spoiler, and possibly in the notch of a notchback (see Fig 2.14).

17
 Fig 2.14 Quasi-two-dimensional separation (shaded and hatched regions)

Wakes also form on the blunt rear of a square back. Depending on the outer flow field, long
wakes are formed, which extend far downstream, or the wakes are short and closed (see Fig
2.15).

Fig 2.15 Large, long, open wake of a square back and small, short, closed wake of a fastback

Fig 2.16 shows the counter-rotating vortex pair of a notch back, a fastback and a square
back. The lower vortex rotates counterclockwise and is responsible for carrying the
contamination to the rear of the vehicle. The upper vortex rotates in the opposite direction.
After the separation bubble closes, a pair of counter-rotating longitudinal vortices forms in
the trailing wake. This produces an upwash in the case of a square back, and induces a
downwash in the trailing wake flow on a notchback or fastback.

18
Fig 2.16 Counter rotating transverse vortices in the wake of cars with three typical rear end
configurations
The vector diagrams in Fig 2.17 clearly show these vortices. On a square back, the
vortex pair rises in the flow direction and wanders toward the plane of symmetry. On
fastbacks and notchbacks the vortices approach the road downstream and move to the
outside. It can be postulated that these longitudinal vortices are the continuation of the
lateral vortices described above. There is a velocity decrease toward the centre of the
vortex. The longitudinal vortices are slowly exhausted downstream by dissipation.

19
Fig 2.17 Continued on next page

20
Fig 2.17 Transverse velocity vector diagrams for notchback, fastback and square back cars

The second type of separation is three dimensional in nature. Vortex trains are
formed at sharp edges where the flow is oblique, as with a delta wing. Such a vortex pair is
formed on the two A-pillars and is bent back towards the roof at the upper end of the A-
pillars. Its effect on the rear end flow is still unknown. A strong vortex pair forms at the rear
of the vehicle, depending upon the inclination of the rear end (Fig 2.18.)

Fig 2.18 Three dimensional flow separation

These rear vortices interact with the external flow field and with the quasi-two-dimensional
wake.

21
2.4 Wind Forces due to Steady Side Winds
A special case of side wind consists of a steady airflow which has a constant
velocity profile regardless of height. Although hardly possible in nature, this case is in fact
provided in the wind tunnel with the exclusion of a thin ground-floor boundary layer (Fig
2.19).

Fig 2.19 Comparison of various side-wind profiles

Under natural side-wind conditions, the relationship shown in the right-hand figure is
present, whereby the whole side-wind profile has boundary layer characteristics. When
combined with vehicle speed the resultant air floe profile is strongly twisted. For
comparison, the centre parts of Fig 2.19 shows the wind profile of a side-wind simulator,
where it can be seen seriously. For low side wind speeds there is apparently no increase in
lift force. But at somewhat higher side winds, the lift force may become manifold.

2.5 Numerical Methods for Computation of Flow around Road Vehicles

The traditional predictive tools used in the automobile industry to evaluate
aerodynamic performance are the wind tunnel and road tests. Full-scale wind tunnel tests
are expensive to build and operate whereas scale model test results are subjected to
numerous doubts associated with realistic simulation of Reynolds number, surface and
underbody details, engine cooling and passenger compartment flows, tunnel wall boundary
layer and model support interference effects, model and wake blockage effects, effect of
flow-intrusive probes etc.

22
 Road tests represent the most realistic simulation of the environment in which a
vehicle operates. However, the difficulties associated with the ever-changing environment
often make the results obtained open to debate. Great care is needed to make the results
meaningful and conclusive.
Computers, and with them computational fluid dynamics (CFD), are slowly
emerging as additional basic tools in aerodynamic design. Wind tunnels and computers are
both simulators-wind tunnels analogue, computers digital. There characteristic differences
make them complimentary rather than completely competitive. The relative role of the two
simulation techniques is however changing.
In future, wind tunnels may be used for validation and refinement of theoretical
predictions or global simulation of the entire flow field rather than for extensive parameter
studies as in the past.
Numerical simulation is well suited to the analysis of a wide range of shape options-
for example during an early design stage-thus increasing thus increasing the prospect that an
optimum shape will be identified. Sometimes a numerical simulation permits the
investigation of situations that cannot be realistically duplicated in a wind tunnel. The
aerodynamics of two vehicles in the passing or overtaking mode, for example, poses a
difficult problem for wind tunnel tests.
Numerical simulations are most useful in predicting trends of how shape changes
will affect flow field features. Absolute performance prediction is usually poor. Computer
size and speed limitations, and the lack of information about the physics involved, often
limit the predictive capacity of numerical methods.
An interesting application of numerical methods is the effectively enhancement of wind-
tunnel tests through pre-test planning, on-line test diagnosis and post-test validation.
Although this may not lead to a reduction in wind tunnel testing, it can help to ensure that
the time spent is used more intelligently.
All numerical methods to compute fluid flow are based on approximations to the full
Navier-Stokes equations. These are second order non-linear partial differential equations
which govern all fluid motion. Except for the simplest approximations, they are solved by
techniques such as finite element, finite volume or finite difference to achieve the spatial
and temporal detail needed.
In these techniques the physical region of interest is divided up (or discretized) by a
two- or three-dimensional grid. Such grids are in practice complicated orthogonal or non-

23
orthogonal networks that may originate in the body contour envelope and have a flow
physics oriented spacing. The vast amount of detail needed to analyze the flow around a real
vehicle will limit the use of computational methods for quite some time to come.

Navier – Stokes Equations

The Navier-Stokes equations are the fundamental partial differentials equations that
describe the flow of incompressible fluids. Using the rate of stress and rate of strain tensors,
it can be shown that the components Fi of a viscous force F in a non rotating frame are
given by
Fi ∂ ∂
∂
 + + 

∇
 (1)
V ∂ ∂ ∂

∂ ∂  ∂   
+  − ∂ ∇
 +      ∇
 ,
∂ ∂ ∂ 
(2)

(Tritton 1988, Faber 1995), where is the dynamic viscosity, is the second viscosity
coefficient, ij is the Kronecker delta, ∇  is the divergence, B is the bulk viscosity,
and Einstein summation has been used to sum over j = 1, 2, and 3. Now, for an
incompressible fluid, the divergence ∇  = 0, so the term drops out. Taking to be
constant in space and writing the remainder of (2) in vector form then gives
 
= ∇
,

(3)



where ∇
is the vector Laplacian.

There are two additional forces acting on fluid parcels, namely the pressure force

  !!#"
= − ∇
, (4)


where P is the pressure, and the so-called body force

Fbody
F= . (5)
V

24
Adding the three forces (3), (4), and (5) and equating them to Newton'
s law for fluids yields
the equation

∂
 + ∇=∇ + ∇
 + F,
(6)
∂

and dividing through by the density gives

∂
∇

+
 ∇
 =  + ν∇
+ ,
$
(7)
∂  

where the kinematic viscosity is defined by


ν  (8)


The vector equations (7) are the (irrotational) Navier-Stokes equations. When combined
with the continuity equation of fluid flow, the Navier-Stokes equations yield four equations
in four unknowns (namely the scalar and vector u). However, except in degenerate cases
in very simple geometries (such as Stokes flow, these equations cannot be solved exactly, so
approximations are commonly made to allow the equations to be solved approximately. As
it must, the Navier-Stokes equations satisfy conservation of mass, momentum, and energy.
Mass conservation is included implicitly through the continuity equation,


    φ %'&)((                 . (9)

So, for an incompressible fluid,


     . (10)



Conservation of momentum requires




 [
 ]
[     ]
[        ]
[      ] (11)



25
so

 
*
      
  ν . (12)



Finally, conservation of energy follows from

 
     
(13)

 

where s is the entropy per unit mass, Q is the heat transferred, and T is the temperature.
Now consider the irrotational Navier-Stokes equations in particular coordinate systems. In
Cartesian coordinates with the components of the velocity vector given by,


 

the continuity equation is



  (14)





and the Navier-Stokes equations are given by









+ + +

        
+  +    (15)









+ ,









- - -

        
-  -    (16)









- .









- - -

        
-  -    (17)









- /

In cylindrical coordinates with the components of the velocity vector given by,

 (
0 
13

2 )

26
the continuity equation is

 4  4 
φ
 5
     , (18)

  
φ


and the Navier-Stokes equations are given by


7
φ 
7 
7
φ
6

7
ρ 
7  
8 
      φ   


 9 
 9  
 9
 
 φ
- - -

 
   - -   
9 9
(19)


 
  
φ
 
φ
- - - - 9


  9 


 φ
  φ
 φ
 φ φ
  φ


  
φ

9 /

φ 
φ φ

φ
φ 
 ;
: : :


     :  : :  :   (20)


φ







φ


φ
: :
φ


< 
<
φ 
< 
<

 
=  
<
      φ  


 ? 
 ? 
?
?
> > >

     > >    ? (21)

 
 
φ

> >

In spherical coordinates with the components of the velocity vector given by,

 (
@B

A 
φ )

the continuity equation is

 D   D 
C C     
φ

 (22)

  
     
φ

27
and the Navier-Stokes equations are given by

E E
J I J I I J I I J I I I
H L H L G H L F H M G
K I M φ
JON H JQP PRJTS PUQVXWYSZJ P P
φ
]Qb ]_[ ^ ` ]T^ `a^ c ]_[ ^ dOegfh ]Y^
ikj lnm \ l \ j \ l \ l \
]Yo ]To [ oZ]To o [ [
oB ]ph [ o [ ]ph

} txq u v tYu r vauxr wOy{z| v tYu

~ s    φ €~k‚
q
ƒ…„g† ‡
q |ˆt q
ƒ
q tT| ƒ
q q
ƒ…„g† ‡ |‰t
s (23)
φ φ
” ” ” ” Š ’ “
‹ ‹ …
Œ ‹ ‹ ‹ ‹
• – – – – φ — φ
”˜ “ Œ  ™
”T …™   R
™ ”˜ “ ™šT › œTZ
“ ”  Ž‘
Ž‘™ 
 φ
  ¡ £  T¡
¡ ¤  ¢ ¡ ¥‘¦Q§#¨  T¡ ¤  ¢ ¡ £  T¡
ž © ª ž ž © ž © ž © ž © Ÿ
¤  Y´  T«  «¬ T«­«¯ ®T° ±  ¨ 
«²   ¨  «    ¨ «¯ ®T° ±  ¨    
«³   ¨
µ ª © ¶ φ ©¸·
«¹  ¨ £g¥‘¦Q§#¨  T¡ ž (24)
ª φ
«º ®p° ± ¨  
φ

φ
φ  » φ ¼
φ  ¼ φ   φ
φ

  »    




  
φ
À½ Á Å ÀÃÁ
Á Æ À½ Á ǁÈÊÉË ÀÄÁ Æ À½ Á Å ÀÄÁ ¾
φ Ì φ φ Í Ì φ Ì φ Ì φ Ì
Æ ÀÃÔ ÀÃÎ ½ ΏÀÃÎÏθÐÃÑ Ò ½ Ë
¿
½ ½ À Ë ½
ÎÓ Î ½ À Ë ¿
½ ÐÃÑ Ò ½ Ë
θ À ½ ½ ÐÃÑ Ò Ë
θ Õ Í Ì Ö φ φ ÌØ×
ÎÐÃÑ Ò Ë À őǁÈÊÉË ÀÃÁ φ (25)
φ Ì φ
½ ÐÃÑ Ò Ë
θ φ

The Navier-Stokes equations with no body force (i.e. F = 0)


        
 
Ù
(26)



28
can be put into dimensionless form using the definitions

(27)

Ú 
Ú


 (28)
 
Ú


 (29)
 
Ú


 (30)
 
Ú

Ü (31)


Û

 
  
Ý Ý Ý
(32)
    
 
Û

   

Û Û Û

 (33)
  
Û


Here, U and L are a characteristic velocity and a characteristic length. Then



 ( Þ
)   
  

Þ

 

Þ Þ

Þ
( 
ßRÞ
) 

ß 
Þß
( ) Þ
(34)


Þ

Assuming constant and multiplying both sides by L/ U2 gives

 à
à


à          
àáàZà âãâ â äåâ
(35)

  


   
ç
(36)
æãæ   æ æ

where Re is a dimensionless parameter known as the Reynolds number. Pressure is a

parameter fixed by the observer, so it follows that the only other force is inertia force.
Furthermore, the relative magnitudes of the pressure and inertial forces are describe by the
Reynolds number, defined as

(37)

29
For irrotational, incompressible flow with , the Navier-Stokes equation then
simplifies to


         
  
ç
(38)



For low Reynolds number, the inertia term is smaller than the viscous term and can
therefore be ignored, leaving the equation of creeping motion


       
è
(39)

In this regime, viscous interactions have an influence over large distances from an obstacle.
For low Reynolds number flow at low pressure, the Navier-Stokes equation becomes a
diffusion equation


 
  
ç
(40)



For high Reynolds number flow, the viscous force is small compared to the inertia force, so
it can be neglected, leaving Euler'
s equation of inviscid motion


         (41)

In the absence of a pressure force,


        
é
(42)

which can be written as


      (  )  ν  (  ) (43)


ê

30
For steady incompressible flow,


   (44)



At low Reynolds number,

  

ë
(45)

At low Reynolds number and low pressure

 
ì
(46)

At high Reynolds number

     

 (47)

For small pressure forces,



    
í
(48)

which can be written as

    (   )  ν  î (   ) (49)

The set of equations for incompressible, viscous, 2D unsteady flow is:

∂u ∂v
+ =0
∂x ∂y

du ∂u ∂u ∂P ∂ 2u ∂ 2u
ρ +u +v = X − +µ +
dt ∂x ∂y ∂x ∂x 2 ∂y 2

dv ∂v ∂v ∂P ∂ 2v ∂ 2 v
ρ +u +v =Y − +µ +
dt ∂x ∂y ∂y ∂x 2 ∂y 2

31
Unknowns are u, v, P which are to be solved in terms of x and y - i.e. across flow domain.

Solutions are typically plots of velocity vectors, streamlines, pressure contours (and

temperature contours if energy equation is added). These may be processed to produce such

data as forces (eg lift and drag on a foil) or pressure loss in pipes and fittings.

Since analytical solution is available only in simplest of cases, numerical techniques are

required; thus a grid across flow domain needs to be defined.

Fig 2.20 Grid across a flow domain

Unknowns are determined at each grid point. Concept may be extended into time domain:

Fig 2.21 Grid extended into time domain

32
Typical grid notation:

Fig 2.22 Typical grid notation

Numerical methods to solve the Navier-Stokes equations can be classified into the
following four categories, depending upon the degree of approximation made:
1. Linearized inviscid flow methods
2. Non-linear inviscid flow methods
3. Methods based on Reynolds-averaged Navier-Stokes equations
4. Solutions of full Navier-Stokes equations
1. Linearized inviscid flow methods
These are used routinely in aircraft design and have reached maturity. They are applicable
to subsonic, contour-attached flow. Vortex-lattice and panel method codes belong to this
category. Use of these methods to compute the flow field around the cars, whose
predominant feature is the large separation region at the vehicle rear, remains severely
restricted.
2. Non-linear inviscid flow methods
The non-linear inviscis flow methods based on the solution of Euler equations have
established themselves as accurate design tools for the prediction of trans-sonic flow around
a class of aircraft components, e.g. wings. The ‘automatic’ simulation of flow kinematics in
the subsonic separated flow computations claimed by developers of these codes needs
further substantiation. Provided this simulation capability turns out to be general, a coupling

33
of these methods with boundary layer approaches may means a significant advance also of
benefit to vehicle aerodynamicists.
3. Methods based on Reynolds-averaged Navier-Stokes equations
These methods are still undergoing extensive research and development. These equations
need a turbulent model for closure. The difficulty of modeling turbulence with sufficient
generality and the complex mesh generation needed to resolve flows such as around road
vehicles are the principle difficulty that have to be overcome.
4. Solutions of full Navier-Stokes equations
Methods to solve the full Navier-Stokes equations, which belong to the last category named
above, are practically non-existent. Only very preliminary research is underway here.
Some Difficulties of Numerical Simulation of the Road Vehicle Flow Field
Regions of separated flow being the key features of a road vehicle flow field, an analytical
approach is extremely difficult. Even simplified, basic vehicle-like configurations free of all
appendages and having smooth surfaces create ‘closed’ separation regions and a large wake.
One of the main difficulties encountered in modeling such flows is the lack of generally
applicable information about three-dimensional separated flows. The variety of separation
phenomena that can occur in three-dimensional flows is a subject of continuing research.
Factors governing the initiation of different types of three-dimensional flow separation,
kinematics of the structures in separated flow, unsteady behavior of bluff body wakes,
turbulence, etc., are all phenomena not well understood. Modular or sequential approaches
similar to those used in aircraft applications remain inadequate since computational methods
to treat three-dimensional boundary layers, in an adverse pressure gradient and strong cross-
flow environment, which is typical of road vehicle flows, are not yet available.
As noted earlier, use of Reynolds-averaged Navier-Stokes equations need a turbulence
model to close the system of equations and make them amenable to solution. Standard
mixing length and eddy viscosity concepts cannot be used to model complex real turbulent
flows. Higher order turbulence models, currently not available, are therefore needed.
Methods Based on Solution of Navier-Stokes Equations
The Navier-Stokes equations for a homogeneous, incompressible medium, together with the
continuity equation, can be used to describe adequately the laminar flow around a road
vehicle. As these equations represent, in principle, all the physics involved, no additional
assumptions and modeling are needed.

34
 However, the flow around road vehicles is mainly turbulent and Navier-Stokes
equations for turbulent flows need a turbulence ‘model’ , to make the system of equations
amenable to numerical analysis. The basic equations employed are averaged over a time
interval. This interval is chosen so as to make the equations independent of the random eddy
fluctuations, yet permit a resolution of the unsteady macro-structures which may be present.
The industry standard turbulence model used most commonly is the k- turbulence model.
The Standard k- Model
The standard k- model is used in the prediction of most turbulent flow calculations because
of its robustness, economy, and reasonable accuracy for a wide range of flows. However,
the model performs poorly when faced with non-equilibrium boundary layers. It tends to
predict the onset of separation too late and to under-predict the amount of separation.
Separation influences the overall performance of many devices, such as diffusers, turbine
blades and aerodynamic bodies. It also has a strong influence on other effects, such as wall
heat transfer and multi-phase phenomena. In some applications, this can have dangerous
consequences, a notable example being the prediction of wing stall on airplanes. The
standard k- model is a semi-empirical model based on model transport equations for the
turbulence kinetic energy (k) and its dissipation rate ( ). The model transport equation for
k is derived from the exact equation, while the model transport equation for was obtained
using physical reasoning and bears little resemblance to its mathematically exact
counterpart. In the derivation of the k- model, it was assumed that the flow is fully
turbulent, and the effects of molecular viscosity are negligible. The standard k- model is
therefore valid only for fully turbulent flows.

Transport Equations for the Standard k- Model

The turbulence kinetic energy, k, and its rate of dissipation, are obtained from the
following transport equations:

(1)

35
and

(2)

In these equations, G k represents the generation of turbulence kinetic energy due to the
mean velocity gradients. G b is the generation of turbulence kinetic energy due to buoyancy.
Y M represents the contribution of the fluctuating dilatation in compressible turbulence to the
overall dissipation rate.C1 , C2 , and C3 are constants. k and are the turbulent Prandtl
numbers for k and , respectively. S k and S are user-defined source terms.

Modeling the Turbulent Viscosity

The turbulent (or eddy) viscosity, t, is computed by combining k and as follows:

(3)

where C is a constant.

Model Constants

The model constants C1 , C2 , C , k and have the following default values:

These default values have been determined from experiments with air and water for
fundamental turbulent shear flows including homogeneous shear flows and decaying
isotropic grid turbulence. They have been found to work fairly well for a wide range of
wall-bounded and free shear flows.

36
The RNG k- Model

The RNG k- model is an alternative to the standard k- model. In general it offers little
improvement compared to the standard k- model. The RNG-based k- turbulence model is
derived from the instantaneous Navier-Stokes equations, using a mathematical technique
called ``renormalization group'
'(RNG) methods. The analytical derivation results in a
model with constants different from those in the standard k- model, and additional terms
and functions in the transport equations for k and .

The Standard k- Model

One of the main problems in turbulence modeling is the accurate prediction of flow
separation from a smooth surface. Standard two-equation turbulence models often fail to
predict the onset and the amount of flow separation under adverse pressure gradient
conditions. This is an important phenomenon in many technical applications, particularly
for airplane aerodynamics since the stall characteristics of a plane are controlled by the flow
separation from the wing. For this reason, the aerodynamic community has developed a
number of advanced turbulence models for this application. In general, turbulence models
based on the
-equation predict the onset of separation too late and under-predict the
amount of separation later on. This is problematic, as this behavior gives an overly
optimistic performance characteristic for an airfoil. The prediction is therefore not on the
conservative side from an engineering stand-point. The models developed to solve this
problem have shown a significantly more accurate prediction of separation in a number of
test cases and in industrial applications. Separation prediction is important in many
technical applications both for internal and external flows. Currently, the most prominent
two-equation models in this area are the k- based models of Menter.

The standard k- model is an empirical model based on model transport equations for the
turbulence kinetic energy (k) and the specific dissipation rate ( ), which can also be thought
of as the ratio of to k.

As the k- model has been modified over the years, production terms have been added to
both the k and equations, which have improved the accuracy of the model for predicting
free shear flows.

37
The Shear Stress Transport Model
To solve the problems encountered by k- models, new models have been developed. One
of the most effective is the Shear Stress Transport (SST) model of Menter. The model
works by solving a turbulence/frequency-based model (k– ) at the wall and k- in the bulk
flow. A blending function ensures a smooth transition between the two models. The SST
model performance has been studied in a large number of cases. SST is rated as the most
accurate model for aerodynamic applications.

Transport Equations for the SST k- Model

The SST k- model has a similar form to the standard k- model:

(1)

and

(2)

In these equations, G k represents the generation of turbulence kinetic energy due to mean
velocity gradients. G represents the generation of . k and represent the effective
diffusivity of k and , respectively, which are calculated as described below. Y k and Y
represent the dissipation of k and due to turbulence. D represents the cross-diffusion
term, calculated as described below. S k and S are user-defined source terms.

Modeling the Effective Diffusivity

The effective diffusivities for the SST k- model are given by

= (3)

= (4)

where k and are the turbulent Prandtl numbers for k and , respectively. The turbulent

viscosity, t, is computed as follows:

38
 (5)

where

(6)

= (7)

= (8)

ij is the mean rate-of-rotation tensor and is defined as shown below:

The coefficient damps the turbulent viscosity causing a low-Reynolds-number

correction. It is given by

where
=

Rk = 6
=

= 0.072

Note that, in the high-Reynolds-number form of the k- model,

39
The blending functions, F 1 and F 2, are given by

F1 = (9)

= (10)

= (11)

F2 = (12)

= (13)

where y is the distance to the next surface and D+ is the positive portion of the cross-
diffusion term.

Modeling the Turbulence Production

Production of k

The term G k represents the production of turbulence kinetic energy, and is defined in the
same manner as in the standard k- model.

The term G represents the production of and is given by

(14)

Note that this formulation differs from the standard k- model. The difference between the
two models also exists in the way the term ï is evaluated. In the standard k- model, ï

is defined as a constant (0.52). For the SST k- model, ï is given by

(15)

40
where

= (16)

= (17)

where is 0.41. i,1 and i,2 are given by Equations (21) and (22), respectively.

Modeling the Turbulence Dissipation

Dissipation of k

The term Y k represents the dissipation of turbulence kinetic energy, and is defined in a
similar manner as in the standard k- model. The difference is in the way the term ƒ * is
evaluated. In the standard k- model, ƒ * is defined as a piecewise function. For the SST k-
model, ƒ * is a constant equal to 1. Thus,

(18)

Dissipation of

The term Y represents the dissipation of , and is defined in a similar manner as in the
standard k- model. The difference is in the way the terms i and ƒ are evaluated. In the
standard k- model, i is defined as a constant (0.072) and ƒ is defined as shown below:

For the SST k- model, ƒ is a constant equal to 1. Thus,

(19)

41
Instead of a having a constant value, i is given by

(20)

where

= 0.075 (21)

= 0.0828 (22)

and F 1 is obtained from Equation (9).

Cross-Diffusion Modification

The SST k- model is based on both the standard k- model and the standard k- model.
To blend these two models together, the standard k- model has been transformed into
equations based on k and , which leads to the introduction of a cross-diffusion term (D in
Equation 2). D is defined as

(23)

Model Constants

* *
All additional model constants ( ï , ï , o, ï , R , R k, R , *, and M t0) have the
same values as for the standard k- model.

The method often used to solve the Navier-Stokes equations in aeronautical practice
is the finite difference technique. According to this approach, the Navier-Stokes equations
are discretized in a domain around the body into a set of finite difference equations. The
flow region itself is partitioned by overlaying it with a certesian grid. In this grid the
solution of the discretized equations is sought iteratively through standard solution
techniques. The irregular shape of a road vehicle does not allow proper representation

42
through a rectangular grid, so that in the numerical representation the grid lines cannot be
made to coincide with the body contour and consequently the surface of the vehicle is
serrated. Obviously this is a drawback. As is well known, seemingly insignificant surface
details can trigger major changes in the overall vehicle flow field. To improve the body
surface representation, a special treatment of the near-wall flow region therefore is
necessary.

A simple illustration of the finite difference method is given below:

Differencing Formulae:

Taylor Expansion:

Subtracting:

43
Adding the Taylor series equations:

∂ 2u ∂ 4u ∆x 4
ui +1 + ui −1 = 2ui + ∆x +
2
+ ....
∂x 2 i ∂x 4 i
12
∂ 2u ui +1 − 2ui + ui −1
+ O(∆x )
2
∴ =
∂x 2 i
∆x 2

Thus if we take, say Navier-stokes equation in X-direction (steady for simplicity) i.e.

∂u ∂u ∂P ∂ 2u ∂ 2u
ρ u +v = +µ +
∂x ∂y ∂x ∂x 2 ∂y 2

becomes

ui +1, j − ui −1, j ui , j +1 − ui , j −1 Pi +1, j − Pi −1, j

ρ ui , j + vi , j =
2∆x 2∆y 2∆x
ui +1, j − 2ui , j + ui −1, j ui , j +1 − 2ui , j − ui , j −1
+µ +
∆x 2
∆y 2

If we set up this set of equations at each of n interior points in the domain, and we know the
boundary conditions (b) at the exterior points ….

…. then we will form 3n simultaneous equations in 3n unknowns.

Unfortunately, these are non-linear, so an iterative approach is usually employed – e.g.

44
The Pressure Correction Approach

The Navier-Stokes equations represent, in principle, the true simulation of the

physics of the viscous flow. The main drawback of the methods currently employed to solve
these equations is the cartesian grid, which allows the body contour to be only crudely
represented. Special treatment of the near wall region limits the generality and can introduce
incorrect physics in the computed flow. A body centred grid can bring improvements in the
discretization.
Finite volume techniques to solve the Navier-Stokes equations avoid the
disadvantages of a cartesian grid and are flexible in generation of a general body contour.
With the provision of a correct turbulence model now (the SST model), the Navier-
Stokes equations describe completely the physics of the viscous flow. Accuracy of
prediction is then basically dependent on the grid density. As the present numerical schemes
have to retain this grid density over a major portion of the computational domain, the
computational effort becomes very large.

45