does not change appreciably with changes in temperature or pressure, however the

Basic Fluid Mechanics in relation to wind density of gases is highly dependent on pressure and temperature in accordance
with p = ρ ⋅ R ⋅ T ; RAIR = 288 J/kg/K, and p and T is absolute pressure and
Basic definitions
temperature respectively. At sea level air ρAIR = 1.2 kg/m3. Atmospheric pressure
Air is a fluid and therefore obeys the laws of fluid mechanics. The study of fluids
employs the fundamental laws of Physics: Newton’s laws of motion, conservation varies with height according to dp = −ρ ⋅ g ⋅ dz . In the tropopause (lower 10 km of
of mass, and the laws of thermodynamics. Therefore, in many respects fluid the atmosphere) the temperature decreases with height. The rate of temperature
mechanics is similar to rigid body mechanics. Under normal forces a fluid acts drop is not constant as it depends on topography and time of year, but a typical
similarly to a solid body with both deforming elastically. Under shear forces a temperature decrease is between 0.65 and 1°C/100 m elevation.
solid will deform elastically with a finite force producing a finite distortion,
whereas a fluid is a substance that will continually distort under the action of The bulk modulus defines how easily the volume (hence the density) of a given
shearing forces even though these may be exceptionally small in magnitude. change in pressure δp
mass of fluid can be changed. Bulk modulus K = =
volumetric strain δ
− V
Air is compressible; with density changing with pressure in accordance with the V
equation of state. However, for the wind flow around structures pressure changes where V is the volume of fluid and the negative sign indicates a decrease in
are relatively small and the air can be considered incompressible. volume with an increase in pressure. Since the mass of the fluid can not change,
dp
m = ρ ⋅ V , then K = ρ ⋅ , units N·m-2, Pa. However for gases, which are easily
In most applications it is usual to treat a fluid as a continuum: a substance dρ
continuous in space with compressed, compressibility must be taken into account.
Pressure 1, positive
continuous properties Standard Gauge Gauge
omitting the complex motion Atmospheric Density describes the ‘heaviness’ of the fluid, but this
Pressure Pressure u+δu
of discrete molecules. Like Pressure does not define how the fluid flows. For example n δn
other continua normal and 101.3 kPa = ds
water and oil can have approximately the same u
shear stresses are developed. 1 bar dt
10.3 m H2O Local Absolute density, but flow quite differently. Viscosity is a s
Normal stresses are termed
0.762 m Mercury Atmospheric Pressure Pressure 2, measure of the resistance in the fluid to shear forces. Solid boundary
pressure and in air, which negative,
For all fluids in contact with a solid surface, adhesion
14.7 psi Pressure (vacuum)
cannot develop tension, are 8 569 m Air of the fluid molecules to the surface occurs. If two Figure 2
always positive measured at 1.205 kg·m-3 Absolute zero, complete vacuum adjacent fluid layers separated by distance δn are in
relative to absolute zero of
Figure 1 steady parallel motion at speeds u and u + δu , Figure 2, the shear stress
pressure. Negative or suction
pressures are measured ⎛ δu ⎞ ⎛ δu ⎞
between the two layers τ = fn ⎜ ⎟ , where ⎜ ⎟ is the velocity gradient.
relative to some positive reference pressure such as atmospheric, Figure 1. At a ⎝ δn ⎠ ⎝ δn ⎠
point in space, pressure has the same magnitude in all directions; isotropic. In a ⎛ du ⎞ ⎛ d ds ⎞
fluid shear stresses, τ, can only be developed if the fluid is in motion through As δn → 0, τ = fn⎜ ⎟ = fn⎜ ⋅ ⎟ . For the majority of fluids this function
dn
⎝ ⎠ ⎝ dt dn ⎠
viscous forces.
⎛ δu ⎞
reduces to a simple proportionality τ = μ ⋅ ⎜ ⎟ , where the constant of
The density of a fluid, designated by ρ (rho), is defined as its mass per unit ⎝ δn ⎠
volume. It is used to characterise the mass of a fluid system and has units kg·m-3. proportionality μ (mu), units N·s·m-2, is called the dynamic viscosity of the fluid.
Density varies widely from one fluid to another. For liquids the value of density
Fluids with such a constant of proportionality are called Newtonian. Dynamic Pathline – the path of an individual particle in the flow over a period v
viscosity is highly sensitive to temperature. The dynamic viscosity of liquids tend of time.
to decrease with an increase in temperature, the converse is true for gases. Streakline – inject a visible tracer into the flow and the resulting v
For fluid mechanic problems dynamic viscosity is often expressed as a function of streak is an instantaneous picture of all particles which have passed
μ through a point.
density, ν = . This is called the kinematic viscosity, ν, (nu) units m2·s-1. viscous
ρ Streamline – an imaginary line in the flow across which there is no Streamline
flow, Figure 3. Velocity is tangential to the streamline at all points. Figure 3
effects become important in regions of high velocity gradients. The dynamic
Solid boundaries are streamlines as there is no flow across them.
viscosity of air at atmospheric pressure is 1.81 x 10 -5 N·s·m-2. v2
Streamtube – if a series of streamlines are drawn
v1
around the perimeter of an elemental area, they will δA2
Fluid Motions and Flow Patterns form a streamtube, Figure 4. As there is no flow δA1 ρ
across a streamtube, the fluid acts as if it were in an ρ1 2
This section considers the motion of air (or any fluid) as distinct from the forces imaginary pipe. For steady flow the rate of fluid Streamtube
generated. entering the streamtube must equal the rate of fluid Figure 4
leaving the streamtube.
Compared to rigid body mechanics, the study of fluid motion is complicated by For steady flow a pathline, streakline, and streamline will all coincide and remain
the introduction of viscosity dependent shear forces. When a real fluid flows over constant. For unsteady flow either a series of instantaneous streamlines or an
a solid boundary, the velocity of the fluid in contact with the boundary is the same average streamline pattern can be drawn.
as the boundary, and a velocity gradient is created at right angles to the boundary.
The change in velocity from layer to layer gives rise to shear stresses in the fluid.
Continuity
Fluid motion is driven by forces created by differences in pressure or elevation,
For fluid flow, conservation of mass must be maintained. In steady flow this can
and governed by inertia and viscous shear stresses. The resulting motion is not
be illustrated using the streamtube concept, therefore:
easy to define theoretically and generally requires experimental verification.
Mass of fluid = Mass of fluid + Increase of mass in the
entering per unit exiting per unit control volume per unit time
Flow can be either steady or unsteady, depending whether the flow changes with
time time ∂
time at a given location or, uniform or non-uniform depending whether the flow + ∫
ρ ⋅ ∂V
changes with distance. Flow may be considered compressible or incompressible.
Flow can be rotational or irrotational depending whether the particles can rotate as
∫ ρ ⋅ v in ⋅ dA ∫ ρ ⋅ v out ⋅ dA ∂ t

they flow. The most common flow in practice is unsteady, non-uniform flow, but
Consider a streamtube with a small cross sectional area so the velocity is constant
for many problems steady, non-uniform flow can be assumed.
over any cross-section, Figure 4. Since there can be no flow across the streamtube:
Whether flow is steady or unsteady will depend upon the observer. If a wave
mass entering section 1 = mass leaving section 2, therefore
travels along a channel, the flow is unsteady to someone on the bank, but steady to
ρ1 ⋅ v 1 ⋅ δA 1 = ρ 2 ⋅ v 2 ⋅ δA 2 = const, m
& . For real flow the velocity may not be
someone travelling with the wave. For steady flow a fixed set of co-ordinate axes
are employed. For unsteady flow it is sometimes useful to use a moving set of constant across the conduit. However, using the mean velocity, the mass flowrate
axes, reducing the unsteady flow to a steady flow problem. Wind is unsteady as it m& = ρ ⋅ v ⋅ A = const . If the fluid is incompressible, which is true for most wind
is continually changing speed and direction with time, the wind can be separated engineering applications, the density remains constant and the volume flowrate,
into a mean or average component, which does not fluctuate with time and a v1 ⋅ δA1 = v 2 ⋅ δA 2 = const, Q . Thus if streamlines are closely spaced the velocity
fluctuating component. will be high and widely spaced streamlines indicated areas of low velocity.
Laminar and Turbulent flow For aerodynamically smooth surfaces the roughness elements are covered by the
Consider a tank as shown in viscous sub-layer while for aerodynamically
Tank
Figure 5, at low velocities, rough surfaces the roughness protrudes through
fluid layers flow parallel to one Dye the viscous sub-layer and the turbulent flow in

Distance from boundary

another at the same or slightly the outer boundary layer impinges directly on
different velocities and there is them creating drag forces instead of shear. The
no mixing of the fluid Valve flow over the surface of the earth is always Laminar
particles, the dye filament aerodynamically rough. Outside the boundary
would be straight. This is layer the effects of shear stresses can be ignored
called laminar or viscous flow, and the fluid can be treated as an ideal fluid, i.e. Turbulent
and viscous forces dominate it has no viscosity, or shear stresses therefore the
particle motion. As the flow Figure 5 fluid particles will not be distorted. If the fluid Mean speed

increases there comes a point velocity is high and the viscosity low, the
when the dye filament oscillates then breaks up, indicating velocity changing in boundary layer will be thin and the fluid can be Figure 7
all directions and a mixing of the fluid particles. This is called turbulent flow and treated as an ideal fluid. Even in problems in
accounts for about 99% of real flow situations including all structural wind which the effects of viscosity and turbulence can not be neglected, it is often
engineering. When the flow is disturbed the inertia of the particles tend to carry convenient to correct theoretical calculations with experimental results. Typical
them off in new directions, but the viscous force due to the surrounding fluid boundary layer
tends to make it conform to the rest of the flow. In laminar flow, the viscous thicknesses are of the n
1
forces are sufficient to overcome small inertial forces. The primary criterion on order of mm for flow turbulent flow
0.8
whether the flow will be laminar or turbulent is the ratio of inertial to viscous over a flat plate and km actual logarithmic profile
ρ⋅v⋅D v⋅D for wind flow over the 0.6
forces, called the Reynolds number, RN = = . surface of the earth. theory transition flow
μ ν Fluid normally flows in
0.4
viscous viscous flow
When real fluid flows past a boundary, the fluid in contact with the boundary has 3-D, but sometimes -∞ 0.2 sublayer parabolic profile
the same velocity as the boundary. For a stationary boundary the flow next to the changes in fluid 0
boundary is brought to rest by surface shear stresses, or friction, which arises due -2 0 2 4 6 u8
properties can be
to the viscosity of the fluid. The velocity of successive layers of fluid will increase simplified to one or two
with distance from the boundary associated with dimensions without Figure 8
decreasing shear stress. Eventually a steady state distance vo introducing major errors.
velocity will be reached, where the drag (friction) of the One-dimensional flow is assumed to occur if the fluid properties (velocity,
boundary has no effect, Figure 6, this is called the pressure, and elevation) only change in the direction of flow and not across the
boundary layer and must be smooth with no 0.99vo cross-section. The one-dimension is taken along the centre streamline although
discontinuities. A laminar boundary layer will form a boundary this may curve in space. The applicable values of velocity, pressure, and elevation
parabolic distribution and a turbulent boundary layer will layer along the streamline are averaged across the cross-section normal to the
form a logarithmic boundary layer, Figure 7. A turbulent streamline. An example of 1-D flow is pipe flow, but since the real velocity
boundary layer must have a very thin layer next to the profile across the section is not uniform, a correction has to be applied. In two
boundary in which the velocity profile is dominated by velocity dimensional flow, the flow parameters are the same in parallel planes, e.g. flow
viscosity called the viscous sub-layer, Figure 8. over a weir.
Figure 6
Any particle or element of fluid will obey the normal laws of mechanics, as B D
predicted by Newton: ∑ Δq = ∫ v ⋅ dn =∫ v ⋅ dn = constant = Ψ where v is the velocity at any point and
A C
1. A body will remain at rest, or in a state of uniform motion in a straight line, dn is a small distance orthogonal to the streamline. The units of ψ are volume per
until acted upon by an external force. unit width per unit time, or a volume discharge per unit width; streamlines are
2. The rate of change of momentum is proportional to the force applied and
generally drawn at equal intervals of Ψ.
takes place in the direction of that force.
3. Action and reaction are equal and opposite.
Another method of marking the flow, which becomes useful, is by plotting
velocity potential lines, φ. Potential is a fictitious quantity
The analysis of fluid flow requires us to define the section of flow that is of
and cannot be directly measured. Along any streamline
interest, and what is to be considered as the surroundings acting on the system. It
velocity potential increases with distance, e.g. starting at L
is normal to consider either a closed or an open system. A closed system Ψ=2a
C D H
constitutes a fixed mass of fluid: therefore, the shape of the boundary may change
with time, but the mass of fluid inside the boundary remains constant. At any
A, the velocity potential, φ at C = v ⋅ ds . [note: the
A ∫ C G
K
Ψ=a
instant a free-body diagram can be drawn showing the forces being exerted on the similarity with the streamline, n across the streamline, s B F J
element by the surrounding fluid, or a solid boundary. An open system constitutes along the streamline]. The faster the velocity, the shorter
A E I
a fixed region, known as the control volume, through which the fluid flows. The the distance to develop a certain amount of velocity Ψ=0
boundary is chosen for convenience, called a control surface, does not change potential. Points on any streamline can be marked off the φ=0 φ=a φ=2a
with time, normally follows solid boundaries where present, and cuts the flow at same way and equipotential lines drawn between points of
equal potential. By definition potential is a property that Figure 10
right angles.
increases along the streamline, therefore equipotential
Two Dimensional Steady Ideal Flow lines are always normal to streamlines. Any number of equipotential lines can be
Streamlines, although purely theoretical, are exceptionally useful for visualising drawn, each of which will have a value of φ. Consider EFGH, Figure 10, so that
E B
fluid flow, especially at the outset of a design project. As they are defined as being
tangential to instantaneous velocity vectors, there can be no flow across them.
φ= ∫
A ∫
v ⋅ ds = v ⋅ dn = a . If the streamlines have been drawn close together then
A
Streamlines cannot cross, for if they did so, then the fluid at the intersection would v is nearly uniform in ABFE, thus as a first approximation, v, is independent of
have two velocities. As there is no flow n E B
across a solid boundary, it is evident that Δq B both s and n and ∫
A
ds = ∫
A
dn , or AE = AB.
any solid boundary must also be a Δq
s ABFE is therefore a shape bounded by curves which intersect at 90° and has two
streamline. Because the flow is always D
Δq sides of equal length, conveniently called a square. Similarly BCGF and CDHG
along a streamline, no fluid can pass across
are squares, but of different size. If the streamlines are close enough they will be
it, so that the flow is same between any two A v true squares. If the interval of φ was chosen to ensure the formation of squares at
streamlines. The velocity of flow therefore dn one portion of the flow the entire flow net will consist of squares. Thus when
varies inversely with distance between Δq=v·dn C forming a flow net, the correct solution is achieved through a process of trial and
streamlines.
Figure 9 error, by plotting streamlines and velocity potential lines ensuring that they form
Consider two streamlines AC and BD,
squares. From a flow net it is possible to extract fluid velocities therefore
Figure 9, since there is no flow across them
accelerations and forces.
the same quantity of fluid per unit time must cross AB and CD:
Equations associated with 2D flow nets The solution of the irrotational flow equation is to define a function, φ, called the
There are four distinct types of motion: ∂φ ∂φ
velocity potential where v x = and v y = . Substituting into the continuity
translation, rotation, and deformation, (or a ∂x ∂y
combination), Figure 11. Only the first two are
applicable to flow nets as there is no ∂ 2φ ∂ 2φ
equation above yields: + = 0 , which is again a Laplace equation. If the
deformation of the particles for ideal flow. dx 2 dy 2
From continuity of mass the flow into an streamline equation for the flow can be defined then the velocity at any point can
element of fluid must equal the flow out of the Translation Rotation
be determined.
element. Considering a small element,
∂v x ∂v y Velocities
continuity yields: + = 0.
∂x ∂y If the flow pattern can be described mathematically using an appropriate stream
Consider the motion of a particle in which Angular deformation Linear deformation function, a series of numbered streamlines can be plotted. Convention is to start at
translation, rotation, and deformation occur, Figure 11 an arbitrary streamline and number the streamlines positive to the left as looking
Figure 12. In time δt, the angles of rotation in the direction of flow. The numerical value of the streamline can be related to
∂v y y the flowrate in the system being described.
∂v
α= ⋅ δt and β = x ⋅ δt will ∂vx
∂x ∂y δyδt
vx δt ∂y Accelerations
be different and the average rate of Knowing the velocities in the field it is possible to calculate the accelerations.
angular rotation is: β Generally, velocity will change from point to point, and may vary in time. If a
∂
⎛ α − β ⎞ 1 1 ⎛⎜ y ∂v x ⎞⎟
v particle moves from A to C the
ω=⎜ ⎟ = − . ∂vy
δxδt
⎝ 2 ⎠ δt 2 ⎜⎝ ∂x ∂y ⎟⎠ δy α ∂x total change in = difference in velocity + change in velocity at
vyδt velocity between A and C C with time
For irrotational flow with no δx
∂v ∂v
shearing stresses causing time t time t+δt dv = ⋅ δs + ⋅ δt
deformation of the fluid element, ∂s ∂t
the nett rotation must be zero. x Differentiating velocity in direction of flow with respect to time gives
Translation, rotation
The solution of the continuity and deformation
dv ∂v ds ∂v ∂v ∂v
acceleration in direction of flow, = ⋅ + = v⋅ + . The first
equation is to define a function, Ψ, dt ∂s dt ∂t ∂s ∂t
Figure 12
called the stream function where term is called the convective acceleration due to the movement of the particle
∂Ψ ∂Ψ from point to point. The second term is called the temporal acceleration due to the
vx = and v y = − . Substitution in the irrotational flow expression yields ∂v
∂y ∂x change in velocity at every point with time. For steady flow = 0 , and for
∂t
∂ 2Ψ ∂ 2Ψ ∂v
2
+ = 0 , which is a Laplace equation with a unique solution for a set of uniform flow = 0 . If the particle accelerates in more than one direction:
dx dy 2 ∂s
boundary conditions.
 ∂v x ∂v ∂v ∂v Therefore, integrating along the streamline with respect to s gives
a x = vx ⋅ + vy ⋅ x + vz ⋅ x + x
∂x ∂y ∂z ∂t ⎛ 2 ⎞
⎜ h + v ⎟ = constant = H s . Total energy per unit weight is constant along a
∂v y ∂v y ∂v y ∂v y ⎜ 2g ⎟⎠
⎝
a y = vx ⋅ + vy ⋅ + vz ⋅ +
∂x ∂y ∂z ∂t v2
streamline, h is piezometric head (c.f. potential energy), and is kinetic head
∂v ∂v ∂v ∂v 2g
a z = vx ⋅ z + vy ⋅ z + vz ⋅ z + z
∂x ∂y ∂z ∂t (c.f. kinetic energy).
∂v s ∂v n
Relationship between acceleration and head For steady irrotational flow = , thus in the n direction (along a velocity
∂n ∂s
Consider the forces acting on an element, Figure 13, in the streamline s-direction: ∂v ∂v ∂ ⎛⎜ v s2 ⎞⎟ ∂h
⎛ ∂p ⎞ potential) a n = v s ⋅ n = v s ⋅ s = ⋅ = −g ⋅ . Integrating along the
ΔFs = p ⋅ dt ⋅ dn − ⎜ p + ⋅ ds ⎟ ⋅ dt ⋅ dn − ρ ⋅ g ⋅ ds ⋅ dt ⋅ dn ⋅ cos α ∂s ∂n ∂n ⎜⎝ 2 ⎟⎠ ∂n
⎝ ∂ s ⎠
⎛ v 2 ⎞⎟
⎛ ∂p ∂z ⎞ velocity potential with respect to. n gives ⎜⎜ h + = constant = H n = H s .
= ⎜− − ρ ⋅ g ⋅ ⎟ ⋅ ds ⋅ dt ⋅ dn 2g ⎟⎠
⎝ ds ds ⎠ ⎝
= mass x acceleration = ρ ⋅ ds ⋅ dt ⋅ dn ⋅ a s Therefore at all points in steady irrotational flow the total energy remains
1 d dh constant. This equation is called the Bernoulli equation:
Therefore acceleration a s = − ⋅ (p + ρ ⋅ g ⋅ z ) = −g ⋅ , where h is
ρ ds ds Total Head = Piezometric head + Kinetic head
p ⎛ p ⎞ v2
piezometric head = z + . This equation holds for all directions e.g. H = h ⎜⎜ = z + ⎟ +
ρ⋅g ρ ⋅ g ⎟⎠ 2g
⎝
dh Or in terms of energy:
a n = −g ⋅ ; i.e. if know piezometric
dn ∂p Total Energy = Potential + pressure + kinetic
head change in any direction, the p+ ·ds dt·dn
∂s p m ⋅ v2
acceleration in that direction can be E = m ⋅g ⋅z + +
calculated. α V 2
p·dt·dn
For any direction accelerations can be The Bernoulli equation is one of the most important equations in fluid mechanics.
dn From the Bernoulli equation, it can be seen that a change in fluid velocity can
combined, e.g. for steady flow in the s
direction along the streamline: dt come about by a change in mass or pressure. From this equation changes in
n ds pressure, velocity, or elevation can be calculated. For wind engineering
∂v s ∂v s z
a s = vs ⋅ + . For steady flow t applications the elevation term can generally be neglected.
∂s ∂t s
∂v s ρ·g·ds·dn·dt The Bernoulli equation is strictly applicable to steady, irrotational, incompressible
= 0,
dt flow. No real flows are frictionless, but in many applications including wind the
∂v ∂ ⎛ v2 ⎞ ∂h effects of friction are negligible. The Bernoulli equation can be used giving
a s = v s ⋅ s = ⎜ s ⎟ = -g ⋅ . Figure 13 satisfactory accuracy throughout the majority of the flow region around buildings.
∂s ∂s ⎜⎝ 2 ⎟⎠ ∂s The Bernoulli equation cannot be applied in boundary layers and shear layers.
 pressure in the boundary layer is required to determine the wind forces on
Application of the Bernoulli Equation - Pitot Tube structures. Fortunately, the pressure in the boundary layer does not change
The Pitot tube is used to measure the velocity of a liquid stream and consists of a significantly and the pressure outside the boundary layer can be calculated using
simple L-shaped tube facing directly into the oncoming flow. At point B, Figure the Bernoulli equation and applied to the surface of the structure.
14, the fluid is brought to rest and a stagnation point occurs (all the kinetic energy
is converted to potential energy). Bernoulli’s equation can be applied between In studies of turbulent boundary layers in pipes and open channels, the surface
point B and point A upstream, roughness is characterised by the size of the material projections. This is
h found to be inappropriate for flow over the earth’s surface and a convenient
assuming there is no loss in energy: total
z ‘aerodynamic roughness’ parameter zo is used to define the surface roughness.
pA v 2A pB v 2B v v =0 The specifics of this parameter are beyond the scope of this course. Another
z+ + =z+ + A B
static
ρ ⋅ g 2g ρ ⋅ g 2g important parameter is the surface shear stress, τo, which is defines the
A B
2
pA v p τo
+ A = B abstract shear or friction velocity, v * = . These parameters can be used
ρ ⋅ g 2g ρ ⋅ g ρ
hole 7 holes in
Thus, pB must be greater than pA; to define the mean velocity profile, such as that for flow adjacent to a rough
in tip outer skin
and v 1
1 boundary = ⋅ ln (n ) + C ., where κ and C are constants and n is the
2
p A − p B = Δp = ρ ⋅ v A = ρ ⋅ g ⋅ h v* κ
2 D D
2 distance from the boundary.
2 ⋅ Δp
vA = = 2g ⋅ h
ρ 6D In summary wind flow overt the earth’s surface creates a turbulent boundary
layer, and most of the flow in a can be treated as frictionless or inviscid. The
The Pitot-static tube, Figure 14 is a Figure 14
local boundary layer on buildings and structures is likely to determine the
development of the Pitot tube used
flow field resulting in the pressure and force distribution.
and can be used for measuring the velocity in gases as well as liquids. The Pitot-
static tube has a separate outer and inner tube to measure the piezometric head at
the tip. The hole at the tip of the instrument forms the inner tube, as with the Pitot- Dimensional Analysis and similarity
tube the velocity at this point is zero. Further back along the horizontal shaft
(where the 7 holes are located) the streamlines are again parallel with a velocity Dimensional analysis is an analytical tool that is used to find the most appropriate
equal to that upstream. Thus the piezometric head at this point is lower than at the way of presenting complicated data as there is seldom an analytical solution to
tip as the fluid particles have some kinetic head. The velocity for the Pitot-Static fluid mechanics problems. The application of the results from one set of
experimental tests to another depends on the understanding of the principles of
2 ⋅ (pT − p A ) 2 ⋅ (pT − pS )
tube is given by v A = = . geometric and dynamic similarity. Similarity is the appropriate scaling between
ρ ρ model and prototype (full-scale); geometric scaling deals with the physical
properties of the model, and dynamic scaling refers to the scaling of the fluid.
Bernoulli and boundary layers Similarity is commonly associated with model testing in air or water. The basic
As mentioned previously boundary layers form next to the boundary. Since shear principles rely on grouping variables formed by dimensional analysis.
stress is proportional to viscosity and velocity gradient, shear stresses will always
be significant within the boundary layer. One outcome of this fact is that the
Bernoulli Equation cannot be used within the boundary layer. However, the
Dimensional Analysis is the number of variables, n, minus the number of dimensions, m, involved.
Any physical situation can be described in terms of a number of recognisable These are called the Π dimensionless groups.
properties, which an object possesses, e.g. for a moving body could measure mass,
dimensions, volume, velocity, acceleration, weather conditions, electrical The drag force on the body has 8 variables (F, l, b, h, k, v, ρ, μ), and 3 dimensions
properties, density of fluid medium…. These measurable properties are all called (M, L, T). Therefore the number of Π groups will be 8 − 3 = 5 .
dimensions. It is necessary to know the magnitude of each dimension. For this an
agreed set of units of measurement are used, e.g. length – m, ft, km, in, yd, mile, The solution will be of the form Π 1 = fn(Π 2 , Π 3 , Π 4 , Π 5 ) where Π1 is the
furlong, ell, chain…. The units employed do not affect the physical properties of dependent dimensionless group and is a function of the independent
the object, only the numerical value. Dimensions are properties that can be dimensionless groups. There is no single or correct set of dimensionless groups.
measured in units.
Π 1 = Fb a v b ρ c Π 2 = μb d v e ρ f , setting μ 1
[−] = MLT −2 La (LT −1 ) (ML ) [−] = ML−1T −1 Ld (LT −1 ) (ML )
b −3 c e −3 f
Dimensional reasoning is necessary to decide which factors are important, to try
to determine a qualitative relationship, and thus enables a selection of the least
number of experiments required. This is done through observation, experiment, or Mass, 0 = 1 + c ∴ c = −1 Mass, 0 = 1 + f ∴ f = −1
intuition. In fluids, general mathematical models can not be constructed, as the Time, 0 = −2 − b ∴ b = −2 Time, 0 = −1 − e ∴ e = −1
interaction between factors, or exact conditions, can never be specified precisely. Length, 0 = 1 + a + b − 3c ∴ a = −2 Length, 0 = −1 + d + e − 3f ∴ a = −1
Qualitative rather than quantitative techniques are employed – empiricism and −2 − 2 −1 −1 −1 −1
therefore dimensions not the numerical values are of concern. The fundamental Π 1 = Fb v ρ Π 2 = μb v ρ
dimensions include mass [M], length [L], time [T], and temperature [K]. It must F ρvb
be ensured that consistent units are employed across the relationship: = Drag coefficient, C D if μ -1 , Π 2 = = Reynolds number
ρv b2 2
μ
For wind flow around buildings, the relevant variables are: geometric length l h k
Π3 = Π4 = Π5 =
(x,y,z) the size of the body, fluid properties (density and dynamic viscosity), fluid b b b
velocity, and dependent variable, response (pressure, p, force, F, frequency, f). It
is the aim of dimensional reasoning to determine how the dependent variables rely Similitude
on the independent variables e.g. How does pressure vary as velocity, or viscosity The dimensionless groups, or numbers, are given specific names. Π3 and Π4 are
of the fluid changes? To fully explore all variables is impractical and dimensional length scales and determine geometric similarity between model and prototype.
analysis aims to reduce the magnitude of the problem by grouping variables. Geometric similarity exists when the ratio of all the linear dimensions in the
L m (odel ) l h
Consider the drag force experienced by a 3-D object moving through a fluid. model and prototype are equal. Length ratio, L r = = =
Force, L p(rototype) b b
F = fn(length , l , breadth , b , height , h , roughness, k, velocity, v, density, ρ , viscosity, μ ) Following from this comes: area ratio, A r = L2r ; volume ratio, Vr = L3r .
Kinematic similarity exists when the ratio of the particle velocities are equal.
Lm
From knowledge of the important variables required to define the problem, a set v m (odel ) Tm L r L
of independent dimensionless groups can be formed. This is the Buckingham Pi Velocity ratio, v r = = = , Acceleration ratio, Ac r = 2r
v p (rototype) Lp Tr Tr
theorm. The number of dimensionless groups required to specify the relationship Tp
Dynamic similarity exists between similar geometric and kinematic systems when prototype as long as the Reynolds number in both systems is the same and
the ratios of all the forces are equal. dynamic similarity exists.
Pressure Force m Viscous Force m Inertial Force m
= = Reynolds number, RN, describes two fundamental types of flow; laminar (low
Pressure Force p Viscous Force p Inertial Force p
RN) and turbulent (high RN). Viscous or laminar flow is where the viscous
Pressure Force m Pressure Force p forces dominate producing a flow where the particles move in smooth regular
Alternatively, = , etc.
Viscous Force m Viscous Force p layers with little mixing. Turbulent flow is where the inertial forces dominate
resulting in irregular particle motion with vigorous mixing between layers. Energy
dissipation is greater for turbulent flow.
Π2 described above is the Reynolds number, and is the ratio of inertial forces to
viscous forces.
For wind flow around buildings, the Reynolds number is typically very large
p resulting in turbulent flow and viscous effects can be neglected.
Dependent variables such as Pressure coefficient C p = , or Force
0.5ρ ⋅ V 2
F
coefficient, C F = , are a measure of the pressure force divided by
0.5ρ ⋅ V 2 ⋅ b 2
inertial force.
Although similarity can be divided into the above three conditions, dynamic
similarity is frequently used to describe flow systems that satisfy geometric,
kinematic, and dynamic similarity. When similarity is achieved, the dependent
variable will be scaled according to the independent variables.

Time varying flows, for example a cylindrical structure such as a chimney stack
experiences a lateral force perpendicular to the mean approaching wind, due to the
periodic shedding of vortices, this will be described in more detail later. The
f ⋅b
dependent variable in this case is the Strouhal number, St = where f is the
V
excitation frequency cause by vortices being sed from the body.

This is a lengthy explanation and can generally be simplified; for example the
drag force coefficient measured on a sphere
of diameter, D, can be simplified to: CD
F ⎛ ρvD ⎞
C D = 2 2 = fn⎜⎜ ⎟⎟ = fn (RN ) and the
ρv D ⎝ μ ⎠
results for all spheres passing through all
fluid will collapse onto a single line, Figure
15. This tells us that the drag force
coefficient will be the same in the model and
RN
Figure 15
Turbulent Flow v function (pdf), f(u), is defined as the amount of time spent at each wind speed, u,
∞
The velocity at any point in time V(t) can
be split in to mean component, U , and
z V(t)
w for a particular record. The integral under the pdf must equal unity ∫ f (u)du = 1 .
orthogonal fluctuating components; in the 0
mean longitudinal direction, u, in the y x The probability density function of a steady wind can generally be expressed as a
lateral direction, v, and in the vertical
u normal or Gaussian distribution:
direction, w, Figure 17. The combined 1 ⎛ (u − U )2 ⎞
f (u ) = ⋅ exp⎜ − ⎟
effect of the fluctuating components is
turbulence. Turbulence can be thought of
U
Figure 17
(
2πσ u 2
)
⎜
⎝ 2 σ u
2 ⎟
⎠
as a large number of eddies, varying in where U and σ u are the mean and standard deviation of the velocity time series
U
size and orientation, being transported by the
mean flow, Figure 16. in the direction required, Figure 18.

Consider turbulent flow along a streamtube in

10 Mean
the mean direction of flow, from the L 9
max
The mean wind speed U is the
8
continuity equation there is no nett transfer of standard average of the fluctuating velocity
7 deviation
V(t) over an recorded period T, or

Wind speed /m/s

mass across the boundary as u = v = w = 0 . 6
There is a nett transfer of momentum; eddies the first moment of area of the
5
momentum per unit area mean probability density function:
Figure 16 4
T
mass 1
= ⋅ velocity = ρ ⋅ v 2 ⋅ dt . The mean V(t )dt .
∫
3
U=
unit area 2 T
1 0
turbulent momentum with respect to time min The mean velocity is therefore a
0
⎛ 2 2 2 2 2 2 ⎞
= ρ⎜ U + V + W + u + v + w + uv + uw + vw ⎟ and the mean turbulent 1:45:00 1:47:00 1:49:00 1:51:00 1:53:00 1:55:00 function of the record length;
⎝ ⎠ PM PM PM PM PM PM standard averaging times are 1
minute, 2 minute (USA), 10 minutes
= 0.5ρ⎛⎜ U + V + W + u 2 + v 2 + w 2 ⎞⎟ . The terms of the
2 2 2 Time
dynamic pressure is
⎝ ⎠ (most countries including Australia),
10 or an hour (UK). These are rarely
form ρuv are cross-products of the fluctuating components and are turbulent
8
continuous samples, but measured in
shear stress or Reynolds stress terms and do not exist in viscous flow.
Wind speed /m/s
the time preceding standard
6
observation times. The mean
Turbulence cannot be satisfactorily defined without recourse to complex 4 velocity in Figure 18 is 4.89 m/s.
deterministic functions due to its random nature. Such descriptions will be
2
discussed in the following sections. Standard deviation
0 The standard deviation gives an
0 0.01 0.02 0.03 0.04
indication of the level of fluctuation,
Probability density Probability or turbulence, from the mean value
Wind speed fluctuations are random in nature caused by randomly sized vortices and can be considered to be caused
which travel along with the mean flow, Figure 18. The probability density Figure 18
by different sized eddies changing the amplitude and frequency of the velocity
fluctuation. The variance is the square of the standard deviation or the second Correlation
T The above descriptors give us an indication of the magnitude at a point, but no
∫ (V(t ) − U ) dt
1 2
moment of area of the probability density function: σ 2 = and information on the spatial and temporal distribution of the wind speed which is of
T more importance for the wind loading of structures. For large structures the spatial
0
T
correlation is important, and for dynamically sensitive structures the temporal

∫ (V(t ) − U ) dt .
1 2 correlation is important. Fluctuating
standard deviation σ = In wind engineering the standard
T A covariance function is the mean product of wind speed
0
the fluctuating components (means removed) at τ
deviation is often (incorrectly) called the root-mean-square (rms) velocity; this is one (autocovariance) or more points in space
only true if the mean wind speed is zero, it should more correctly be called the either simultaneously or with a time lag, Figure
rms of the fluctuating part of velocity. The standard deviation of each fluctuating 19. If only one point is used the autocovariance
time
T will give information on the extent of gusts in a
1
velocity component can be estimated: e.g. σu =
T ∫
u 2 dt . The standard
0
temporal sense, if two locations are used the
covariance will give information on the Figure 19

deviation of the time series above is 1.49 m/s (variance 2.20 m2/s2). turbulence in both a temporal and spatial sense;
A useful non-dimensional parameter for any random variable is the coefficient of not included here.
A covariance function is calculated using:
σ
variability defined as the standard deviation divided by the mean: e.g. I u = u , T
1
U C ij (τ ) = i(t ) ⋅ j(t + τ ) = lim T → ∞ i(t ) ⋅ j(t + τ )dt
∫
σv σw T
Iv = , and I w = . In wind engineering this is called turbulence intensity 0
U U Since there are 3 components of the fluctuating wind there are nine possible
and is important for interpreting the wind loads on structures. It typically varies covariance functions. The most important are the autocovariance functions and the
from about 0.05 at gradient height to 50% near the ground in city centres. In the uw covariance.
figure above Iu=0.30.
Third and fourth moments of the probability density function (skewness and
kurtosis) can also be calculated for more detailed statistics on the velocity
fluctuations.

Gust wind speeds

Usually for design it is the 3 s gust wind speed and not the mean wind speed
which is important for design. The difficulty of wind being a random process is
that the peak gust within an averaging period is a random variable. However using
the assumption that the velocity distribution is Gaussian, an expected peak value,
Û , within the averaging period can be estimated from Û = U + g ⋅ σ u , where g is
a peak factor. The peak factor is approximately 3.5 and 3.7 for a 10 minute and
hourly mean respectively in strong wind conditions. In the time series above g =
2.8 from a 10 minute mean to a 3 s peak.
When the time lag structures where resonance (excitation at the structures natural frequency) is a
is zero the possibility. Correlation tells us something about the frequency components in
autocovariance the wind, but it is not necessarily in a convenient form for engineering purposes,
function is the therefore the spectral density function, or spectrum for short, was developed to
variance of the operate in the frequency domain. The spectral density function provides the
original signal and distribution of vibrational energy (turbulence) with respect to frequency for a
therefore it is useful fluctuating signal and forms a Fourier pair with the covariance function. As
to normalise the previously the most important parameters are the one-dimensional power
autocovariance spectral density functions, S ii (n ) , the Fourier transforms of the autocovariance
function by the function Cii(τ). Any structures in which two dimensional loading is important
variance to form a (e.g. bridge decks, vertical advertising hoardings, etc.) the cross-spectral density
correlation functions may be required.
function: Each fluctuating velocity signal can be regarded as the sum of a large number of
C ij eddies of different sizes or a series of sine and cosine functions of varying
ρ ij = , which
σi ⋅ σ j frequency and amplitude, called a Fourier series of the form:
∞ ∞ k
must lie between +1 Figure 20 a0 ⎛ k ⎞ ⎛ k ⎞ −1 2 π t
fully correlated and x (t ) =
2
+
k =1
∑
a k cos⎜ 2π t ⎟ +b k sin⎜ 2π t ⎟ or x (t ) =
⎝ T ⎠ ⎝ T ⎠ k = −∞
Ake ∑ T .
-1 fully uncorrelated. For small values of τ the autocorrelation function would be
expected to be close to unity, since the velocities would be expected to be in Examples of Fourier series can be found in numerous books.
phase. As τ increases the value of the autocorrelation function will decrease For a finite time series the spectral density function becomes:
2
approaching zero at large time offsets, Figure 20. Cross correlation values would T
S ii (n ) =
2 ( )dt
∫ x(t )⋅ e
− −1 2 πnt
be expected to be low when the spacing between the two measuring location is
large. T
0
The power spectral density function can also be calculated from the
Length and time scale of turbulence ∞
These are useful for characterising the average length and duration of a gust. The autocovariance function via a Fourier transform: S ii (n ) = 2 C ii (τ ) ⋅ e − ( ∫
−1 2 πnτ )dτ
time scale, Ti, can be calculated by integrating the autocorrelation function:
−∞
∞
One of the most important features of a spectrum is that the area under the
Ti = ρ ii (τ )dτ . The turbulence length scale (e.g. l u ) can be estimated from a
∫ spectrum is equal to the variance of the original signal; i.e.
0 ∞
point measurement by assuming U >> u (t ) , and therefore l u = Ti ⋅ U . For the (n )dn = C ii (0) = σ i2 ,
above time trace the time and length scales are 18.5 s and 90 m respectively.
∫S
0
ii Figure 21. For this reason power spectra are often

n ⋅ S ii (n )
Wind spectral density expressed in a normalised form of plotted against ln(n) where the area
σ i2
The probability density function tells us something about the magnitude of the
wind speed, but nothing about the frequency content. From a civil engineering under the graph will be unity.
point of view, information on the frequency content is important for flexible
 1

Normalised spectral density

The Harris von Karman spectrum is an empirical formulation of the longitudinal

Normalised spectral density

production inertial dissipation
⎛ n⋅lu ⎞
4⎜ ⎟
n ⋅ S u (n )
0.1
turbulence spectrum: = ⎝ U ⎠ , Figure 22. Three
σ 2u ⎡
5
2⎤ 6 0.01
⎛ n⋅lu ⎞
⎢1 + 70.8⎜ ⎟ ⎥
⎢⎣ ⎝ U ⎠ ⎥⎦
0.001 measured
distinct regions of turbulence can be identified as characterised by the size of the Harris von Karman
eddy, Figure 23: 0.0001
• Production – turbulence generated in the mean flow from meteorological 0.001 0.01 0.1 1 Frequency /Hz
disturbances causes very large eddies.
• Inertial – the large eddies break apart due to inertial forces forming smaller Frequency /Hz Figure 23
eddies
• Dissipation – the eddies become small enough that viscous effects become Figure 22
important with the energy finally being dissipated as heat.

Figure 21
 external flow Boundary layer limit, V~0.99Vo
Form Resistance Vo

Flow of a real fluid around a body

For a real fluid, viscous shear stresses occur adjacent
shear layer
to the boundary giving rise to surface resistance, or
surface drag. The shear forces are transmitted
through the fluid shear layers giving rise to the
boundary layer often called a shear flow. At the
stagnation point where the fluid is brought to rest the Wake
Positive Adverse
boundary layer is thin and generally increases in pressure pressure Separation
thickness with distance along the surface, Figure 24, gradient gradient point Flow reversal
Stagnation point High energy dissipation
except in areas of high acceleration when the
boundary layer thickness can decrease. Lower pressure
Figure 24
In a flow where the pressure decreases in the
direction of flow (area decreases), the fluid is accelerated and the boundary layer motion, and the dissipation of mechanical energy through turbulence and viscous
can thin. In such a contracted flow the pressure has such a negative gradient that shear.
the flow stabilises while the turbulence generally increases.
As discussed previously, in the boundary layer the frictional and viscous forces
Flows with a positive pressure gradient (divergent flow, area increases with act to decelerate the fluid so at the boundary the velocity is the same as the body,
distance), Figure 24, the fluid decelerates in the direction of motion, and the i.e. no slip. When the boundary layer thickness is small compared to the
pressure increases in the direction of flow. Fluid far from the wall has a large dimension of the body, which is true is most wind engineering applications, the
velocity and high inertia; therefore the flow can overcome the higher pressure pressure variation across the boundary layer can be neglected. Thus, in a real
downstream. Fluid near the wall with low velocity has a low inertia and can not fluid, although the Bernoulli equation can strictly only be applied outside the
overcome the pressure gradient. Thus the flow velocity decreases until the boundary layer, the pressure calculated at the edge of the boundary layer is the
velocity gradient becomes zero. At this point the viscous shear forces must also be same as the pressure at the surface of the body. However, downstream of the
zero, indicating that the boundary layer can no longer progress along the length of separation point the flow outside the wake does not follow the shape of the body.
the body and separates from the surface. This point is called the separation point. The wake region consists of a highly turbulent flow that bears little resemblance
Beyond this point the velocity gradient is negative and flow reversal occurs. The to the free stream.
location of the separation point depends on the shape and roughness of the solid
boundary and the Reynolds number of the flow; this will be discussed in more By definition a streamlined body is one
detail later. Low pressure prevails downstream of the point of separation in what where the major component of drag
or
is called the wake region. After separation the boundary layer retains its coherence force in the mean wind direction results
and the detached boundary layer is called a shear layer and separates the from surface friction caused by surface
undisturbed external flow from the wake. The nett pressure acting on the solid shear stresses. A non-streamlined or
body in the direction of flow when separation has occurred is called the form drag, bluff body is when the major component
Figure 25
compare with surface drag. Thus, flow separation increases both the resistance to of drag arises from separation of the
boundary layer from the surface of the body. For example circular or square
prismatic bodies are bluff, and a flat plate orientated parallel to the direction of Ideal flow round a circular cylinder
Laminar Flow
flow is streamlined. Almost all civil engineering bodies are bluff due to either the For steady irrotational flow of an ideal
shape of the body or high Reynolds number. As will be seen circular prismatic Turbulent Flow
fluid, there is assumed no viscous effects
bodies do not have well defined separation points, whereas sharp edged bodies Ideal Flow
and therefore there is no resistance to
have constant separation points as the fluid cannot resist the high shear stresses at motion regardless of the degree of fluid p
the discontinuity, Figure 25. <0
deformation. According to the 1 2
ρv
shear mathematical equation for ideal flow 2
Features of flow around bluff bodies layer around a circular cylinder, and Bernoulli,
the pressure distribution around a cylinder
Flow always separates at sharp corners. The p
is symmetric about the mid-section and the >0
separated boundary layer is a shear layer resultant force in the direction of motion is 1 2
ρv
dividing the flow into two regions; the wake zero, Figure 27. 2
and the free stream flow, Figure 26. wake
v
Laminar flow round a circular cylinder
The flow outside the shear layer is essentially Boundary layer flow starts at the stagnation
inviscid. The wake region is highly turbulent stagnation point and energy is dissipated to overcome
with a low mean velocity and flow reversal point separation
point surface resistance. Considering the
near the separation point. The pressure on the parabolic velocity profile of laminar flow,
leeward face of the body is low compared to the momentum and kinetic energy of fluid
the approach flow and Bernoulli cannot be particles in the boundary layer are
applied in this region. However if the stagnation separation
points significantly less than in the ideal flow p 1 0 -1 -2 -3
streamline pattern outside the wake can be point situation. The particles decelerate until at 1
predicted these pressures propagate across the some point they stop and separate from the ρv 2

shear layer and can be applied to the surface of boundary. This point signifies the
the body. The effect of the wake region is to wake Figure 27
beginning of the wake region. For laminar
increase the drag force on the body; the major shear boundary layer flow (RN>100) separation
boundary
contributor of which is the low pressure on the layers occurs at approximately 80° from the stagnation point.
layers
leeward face, therefore the wider the wake the
higher the drag force Figure 26
Turbulent flow around a circular cylinder
The shear layers are unstable and tend to form discrete vortices. There are usually At high Reynolds number flows, the growth of the laminar boundary layer is
two shear layers, one from either side of the bluff body, which interact with one curtailed by instabilities in the boundary layer and a turbulent boundary layer is
another and shed from alternate sides of the body. established. If the incident flow is turbulent then this promotes the early
In certain circumstances the flow can reattach to the surface, this generally occurs establishment of a turbulent boundary layer. As the velocity profile for turbulent
when there is a long afterbody. When reattachment occurs the separated region is flow is logarithmic, the velocity of particles close to the solid boundary is faster
called a separation bubble. The pressure inside a separation bubble can be very than for the laminar boundary layer, and therefore the particles have a higher
low due to the high rotation. For long bodies reattachment occurs at low angles of momentum. This higher momentum allows them to overcome the adverse
attack. The effect of reattachment is to narrow the wake, therefore reduce the drag pressure gradient and maintain attached flow further downstream until separation
force. finally occurs at approximately 130° from the stagnation point.
 Mean pressures and forces acting on a bluff body
For both laminar and turbulent flow of a real fluid around a circular cylinder, the
mean pressure distribution is asymmetric about the central axis normal to the Figure 29 shows the time averaged pattern of flow around a bluff body. The bluff
flow, in contrast to the symmetric pressure distribution for an ideal fluid. As a body changes the velocity pattern which in turn changes the pressure distribution
result the cylinder experiences a nett pressure force in around the body in accordance with the Bernoulli equation. Along any streamline:
the direction of flow. It is noted that the mean pressure 1 1
distribution for highly turbulent flow is similar to that p o + ρ ⋅ Vo2 = p x + ρ ⋅ Vx2
for ideal flow up to the point of separation, after which 2 2
the pressures change significantly. Roughening the where subscript x is any point in the system where Bernoulli can be applied. The
following can be deduced from the Bernoulli
boundary surface promotes an early onset of the
streamline pattern in Figure 29: equation
turbulent boundary layer and has become a common is valid
2
experimental means of decreasing separation effects by Va = 0, p a = p o + 0.5ρ ⋅ Vo
Streamlines
delaying separation. Vb < Vo , p b > p o d Bernoulli
c equation
For solid boundaries that are angular in form, there are Vc >> Vo , p c << p o is not valid
b e
well defined points of separation, which are fixed in Vd > Vo , p d < p o line of
po, Vo a
position and independent of Reynolds number. symmetry
pe ≈ pd
Eddy pattern in two dimensional flow
It is exceptionally useful to express
The pattern of fluid motion in the wake region behind a
the pressure on the body in pressure Pressures
circular cylinder consists of a series of vortices,
coefficient format:
commonly referred to as the von Karman street, which
alternately form and detach themselves from opposite p − p ref Figure 29
Figure 28 Cp = x 2
. The reference
sides of the cylinder, Figure 28. The zone of low 0.5ρ ⋅ Vref
pressure at the downstream side of the cylinder shifts from side to side, thereby location is generally chosen for convenience; generally taken upstream of the
producing a time varying side thrust as well as longitudinal drag. The pressure body in the free stream flow. Thus applying Bernoulli around a bluff body gives
distributions shown earlier indicate only the mean component and therefore the 2
pressure associated with these fluctuations are not shown. The frequency of the ⎛V ⎞
C p = 1 − ⎜⎜ x ⎟ . The main outcomes of this expression are:
⎟
vortices being shed from a body is characterised by the Strouhal number, ⎝ Vo ⎠
f ⋅D
St = , where f is the frequency of vortex shedding (Hz), D is a characteristic • At stagnation point Vx = 0, C p = 1 which for incompressible fluid is the
v
dimension of the solid body (m), and v is the mean velocity of the flow (m/s). maximum value of Cp as Vx ≥ 0
• For Vx < Vo , C p > 0 , positive pressures
The Strouhal number depends on the shape of the body and the Reynolds number
of the flow. Values of Strouhal number for common shapes can be found in most • For Vx > Vo , C p < 0 , negative pressures
texts on fluid mechanics; at high Reynolds number the value for a square prism is
It should be noted that there is theoretically no lower limit for Cp. Physically
approximately 0.1 and for a circular prism approximately 0.2.
however there is a lower limit when compressibility effects will influence the
pressure distribution. Low pressures exist where the velocity is high which
generally occurs at building corners; this is important for cladding considerations. Dimensional Analysis
Peak localised pressure coefficients of about -15 have been measured full scale. Flow around bluff bodies is highly complex and depends on many variables.
Pressure coefficients and drag forces are generally measured through experiments
Nett force acting on a body at model scale. The analysis of experimental data lends itself to dimensional
As mentioned previously as fluid flows around a body it will develop normal analysis to obtain a rational for predicting the pressure and forces acting on bluff
stresses (pressures) and shear stresses. Shear stresses develop from the viscosity bodies. Typically force coefficients are a function of:
of the fluid, which develops a boundary layer at the surface. The total force acting • Reynolds number
on an immersed body is an integration of these stresses over the surface area of ρ ⋅ Vo ⋅ D
the body. The total force is made up from both the surface and form drag, both of Reynolds number is defined as RN = where D is a characteristic length
μ
which can be determined through knowledge of the surface resistance to flow and
the pressure distribution around the body. for the body, typically the width of the body normal to the flow. Since the inertia
Cross-wind Along-wind
The surface area integration of surface stresses Lift FL Drag FD forces acting on the body are proportional to ρ ⋅ Vo2 ⋅ D 2 and the viscous forces is
around a body produces a resultant force acting ⎛V ⎞ ρ ⋅ Vo2 ⋅ D 2 ρ ⋅ Vo ⋅ D Inertia forces
on the solid body. This resultant force can be of the order μ ⋅ ⎜⎜ o ⎟⎟ ⋅ D 2 then RN = = =
resolved into the direction of flow, (drag) and D
⎝ ⎠ ⎛ V ⎞ μ Viscous forces
μ ⋅ ⎜⎜ o ⎟⎟ ⋅ D 2
normal to the direction of flow (lift), Figure 30. ⎝ D⎠
For symmetric bodies normal to the flow, the Reynolds number along with pressure gradient, roughness, and turbulence
mean lift force will be zero. prescribes the separation point. Separation is highly dependent on whether the
Mean wind U0
For streamlined bodies, such as aeroplane wings, flow is laminar or turbulent and hence is a function of Reynolds number. If the
at high Reynolds number flow the drag is almost Figure 30 flow around a body is either predominantly laminar, or turbulent flow is well
entirely due to surface resistance (as there is a established, the drag force on a body tends to become Reynolds number
very small wake, therefore minimal form drag). For prismatic bodies, unless very independent. This is due to the fact that at either extreme the separation point
long, the drag is almost entirely due to form drag caused by separation leading to remains reasonably constant. Thus if the separation point remains relatively
a large pressure difference between the front and rear of the object. The mean lift constant, such as for a sharp-edged bluff body the effect of Reynolds number on
force on bluff bodies is generally small compared to the lift that can be generated the drag force will be small especially in highly turbulent flow. This is evidently
on streamlined bodies. not true for streamlined bodies, or in some Reynolds number ranges for curved
edged bodies.
The drag and lift forces can be more conveniently expressed in coefficient form,
ρ ⋅ V2 ρ ⋅ V2 • Mach number
FD = C D ⋅ ⋅ A p and FL = C L ⋅ ⋅ A p respectively, where, ρ is the
2 2 V
Mach number is defined as Ma = and is the ratio of the local speed to the
density of the fluid (kg/m3), V is the mean velocity of flow (m/s), Ap is a c
reference area (normally either the projected area or the surface area) of the body k⋅p K
local speed of sound, c, where c 2 = = , where k is the specific heat ratio,
normal to the flow (m2), and CX is the drag or lift coefficient for the body relative ρ ρ
to the reference area (-). It is important that the correct reference area is used with and K is the bulk modulus of the fluid. Mach number is a measure of the effect of
the corresponding coefficient. The force coefficient is a function of the shape of the compressibility of the flow and is important at speeds near the speed of sound
the body and the Reynolds number and is normally determined experimentally. in the fluid (for air c ≈ 330 m/s); when Ma << 1 the fluid may be regarded as
Values of drag coefficient for common shapes can be found in most fluid
mechanics textbooks.
incompressible, which is the majority of wind engineering applications for regular vortex shedding process is stopped and only general turbulence is
calculating wind loads on structures. produced.
Vortex shedding is most coherent for circular cylinders and it is found that
• Froude number Strouhal number varies with Reynolds number in a similar manner to drag
V coefficient. It is found that the Strouhal number is approximately 0.2 away from
Froude number is defined as Fr = and is the ratio of inertial to the critical Reynolds number. At critical Reynolds number the vortex shedding is
g⋅D random due to the fluctuating nature of separation.
gravitational forces. The effects of gravity on the wind loading of structures is The effect of oscillating vortices results in an oscillating pressure distribution
generally unimportant as the density of air is small. As the structures becomes which in turn imparts a lateral lift force onto the body. The frequency of the
lighter and more flexible, such as cable nets, porous shade cloth, architectural oscillating force must be the same as the shedding frequency. The magnitude of
features, etc., where the movement of the object is significant, Froude number the fluctuating lateral force can be of the same order as the mean drag force. The
needs to be considered. dynamic effect on a body that is free to move with low damping and a natural
frequency near the shedding frequency can in an extreme case cause failure.
• Strouhal number
f ⋅D For flow around a three dimensional body, trailing vortices are generated at the
Strouhal number is defined as St = where f is the vortex shedding free end of a finite length body generating lift (or side) force normal to the free-
V
frequency, and D is a convenient dimension (normally taken as the surface width stream direction and the major body axis. Consider a cylinder which is generating
normal to the fluid flow. Strouhal number is often called the reduced frequency of lift, this means that the upper surface has a lower pressure than the lower surface.
vortex shedding, or the inverse of reduced velocity. Since there can be no discontinuity in the pressure distribution, there must be a
As discussed the separation process from a bluff body is associated with the flow from the lower to the upper surface. This lateral movement combined with
formation of discrete vortices and a large energy loss in wide wake. At normal the forward movement of the free stream induces a vortex at the end of the
operating Reynolds numbers the flow in the wake is characterised by the Von cylinder, the vorticity of which is strongest at the cylinder ends. As the vortex
Karman vortex street. The boundary layers that separate from the two sides of the sheet moves downstream, it tends to form discrete vortices at the free ends. The
body are separated by approximately the width of the body. The boundary layers change in flow pattern around the body evidently changes the pressure distribution
tend to roll up n this region producing large vortices upon reaching a certain size and therefore the drag force on the body.
detach from the body and move downstream. These detached vortices are
approximately the size of the body, and produce fluctuation of large scale and low For flow around a infinitely long cylinder the trailing vortex is not generated. For
frequency in the downstream flow. The shedding mechanism is the interaction streamlined bodies where the flow does not separate it is possible to theoretically
between the boundary layers on either side of the body. The boundary layer from estimate the drag associated with the trailing vortices. However, for a bluff body
one side of the body grows forcing the weaker vortex towards the body. The where there is a significant amount of separated flow, theoretical treatment is not
weaker vortex grows under the dominant vortex and eventually reaches the shear possible and the induced drag is included in the total drag.
layer feeding the dominant vortex. Detached from its source of vorticity, the
dominant vortex is separated from the body and convected downstream. The
weaker vortex is now dominant and a new vortex forms in place of the convected
vortex and the process repeats. The steady narrow-band shedding frequency, f, is
used in the determination of Strouhal number. Vortex shedding can occur for
almost any cross-sectional shape provided the approach flow is uniform and
contains low turbulence, and the afterbody length does not allow full
reattachment. If the interaction is prevented, by physical or fluid means, then the
• Turbulence space and time are necessary to fully describe turbulence. These statistical
As discussed previously, turbulence intensity is the degree of fluctuation in the properties have been defined previously.
incident flow. The higher the turbulence in the approach flow the sooner the Increasing turbulence tends to have little effect on the central vortex shedding
transition to a turbulent boundary layer on the body surface. This turbulent frequency, but increases the bandwidth of frequencies.
boundary layer will delay separation, resulting in a narrower wake and lower drag.
Turbulence is significant in terms of the wind loading on structures, not only in • Surface roughness
terms of the mean and fluctuating forces on the body, but also on local pressures The effect of surface roughness is to cause 2.2
r/b=0.021
near the point of separation, which are critical for cladding design. Generally the transition form a laminar to turbulent 1.8
speaking increasing the turbulence decreases the radius of curvature of the shear boundary layer at a lower Reynolds roughness
1.4 increasing r b
layer because turbulence promotes a better mixing of the shear layers; this will number than if the surface were smooth. CD
1.0 r/b=0.167
cause the flow to reattach earlier. Therefore depending on the geometry increasing The effect of this is to generally increase
turbulence can increase or decrease the drag force, Figure 31. The longer the body the drag coefficient, but in certain 0.6 r/b=0.5
circular Smooth
and higher the turbulence, the more likely reattachment will occur. Turbulence in circumstances, particularly with curved 0.2
the upper left sketch indicates that the increase curvature will cause a lower edged bodies, the drag coefficient will 104 105 10
6
107
pressure on the leeward side of the object resulting in an increase in drag with decrease at certain values of Reynolds RN
Figure 32
number, Figure 32. In this case of a
curved edged body the laminar boundary layer separates early, the wake is wide
3 Smooth flow and the drag high. However, for the rougher body at the same Reynolds number
5
10 <RN<106 the transition to turbulent flow comes before separation of the laminar boundary
d
2 layer. The flow will remain attached for longer around the body, due to the larger
CD b velocity near the boundary, and the wake will be narrower and the drag reduced.
d/b=0.1 d/b=0.5 In practice the surface roughness depends on the material used, the extent of
1
corrosion, and the build up of surface deposits. For estimating the drag coefficient
Turbulence RN<10,6 and corner radius it is convenient to define an equivalent surface roughness height, ε. The equivalent
0 1 surface roughness height relates to the uniform size of particles distributed over a
d/b=2 2 d/b 3 4
smooth body, which gives the same drag results as the naturally rough body. The
Low turbulence equivalent surface roughness height is generally independent of Reynolds number,
High turbulence therefore the ratio ε/D is a dimensionless parameter influencing the drag
coefficient.
Figure 31
turbulence. However, the middle upper sketch shows that with an increase in • Geometric shape and orientation to the flow.
turbulence the flow will partially reattach to the body. This reattachment results in The method of allowing for cylinder
a narrower wake and therefore a decrease in drag. inclination, Figure 33, depends on whether the
Knowledge of the energy contained in the turbulence structure is important in flow is subcritical or supercritical. For
determining both the mean and fluctuating forces on a body. If the body is large subcritical flow the simple cross-flow theory
compared to the scale of turbulence the gusts produced by individual eddies will applies, for supercritical flow the evidence is
not act simultaneously over the entire body, thereby reducing the total force. that this simple cross-flow theory
Knowledge of both the power spectrum and the correlation functions in both underestimates the drag for relatively smooth

Figure 33
two-dimensional cylinders and should not be used.
The critical flow velocity for an inclined cylinder is found to be lower than that
for the same cylinder normal to the flow. In practice the critical Reynolds number
is approximately independent of φ if expressed in terms of the streamwise
⎡ V ⋅ D ⋅ sec φ ⎤
components so that RN crit = ⎢ ∞ ⎥⎦
⎣ ν crit
For subcritical Reynolds numbers, force coefficients are dependent on the
component of free-stream velocity normal to the cylinder axis, VN = V∞ ⋅ cos φ ,
and on the streamwise component of Reynolds number. Thus for inclined
cylinders the normal force coefficient is given by C N = C Doφ ⋅ cos 2 φ .
The axial force is small and for a smooth cylinder can be represented by
CT π ⎛ π ⎞
= 0.033 ⋅ φ ⋅ ⎜2 − φ⋅ ⎟
C Do 180 ⎝ 180 ⎠
C Doφ C ⋅ sin φ C Loφ C ⋅ cos φ
= cos 3 φ + T and = cos 2 φ ⋅ sin φ − T
C Do C Do C Do C Do
For supercritical Reynolds numbers simple cross-flow theory described tends to
underestimate CN for relatively smooth, inclined circular cylindrical members.
The reason for this is that while for the laminar boundary layer (and the associated
pressure distribution) tends to depend only on the cross-flow velocity, when
transition to turbulent flow in the boundary layer has occurred the subsequent
development and separation of the boundary layer are adversely affected by the
three dimensional nature of the turbulent wake flow. This exerts a considerable
influence on the pressure distribution. For supercritical flow, the force coefficients
may be estimated form C N = C Do ⋅ fφ , C Doφ = C N ⋅ cos φ, C Loφ = C N ⋅ sin φ .
The value fφ ranges from 1 at φ = 0°, to 0 at φ = 90°.