Reynolds-averaged Navier-

Stokes Equations

Dr. Soumendra Nath Kuiry

DEPARTMENT OF CIVIL ENGINEERING, IIT MADRAS, INDIA
E-mail: [email protected]
Contents

o Navier-Stokes Equations
• Conservation of Mass
• Conservation of Momentum
o Turbulent Flow
o Reynolds-averaged Navier – Stokes Equations
• Conservation of Mass
• Conservation of Momentum
 Conservation of Mass

For conservation of mass,

u v w
  0
x y z

Figure 1: Differential control volume in

Cartesian coordinate system
Conservation of Momentum

 u u u u  p    yx  zx 
x-momentum equation:    u  v  w    g x    xx    (26)
 t x y z  x  x y z 

 v v v v  p   xy  yy  zy 
y-momentum equation:    u  v  w    g y       (27)
 t x y z  y  x y z 

 w w w w  p   xz  yz  zz 
z-momentum equation:  u v w    gz       (28)
 t x y z  z  x y z 
 Constitutive equations
For a fluid at rest the stress tensor When a fluid is moving, pressure still acts inwardly
reduces to normal, but viscous stresses may also exist
  xx  xy  xz    p 0 0    xx  xy  xz    p 0 0    xx  xy  xz 
       

 yx  yy  yz    0  p 0  (30)  ij    yx  yy  yz    0  p 0    yx  yy  yz  (31)
   0  p     0  p    zx  zy  zz 
 zx  zy  zz   0  zx  zy  zz   0

For or incompressible and Newtonian fluids

 u  u v   u w  
 2        
 x  y x   z x  
p 0 0  
    v u  v  v w  
 ij   0  p 0       2    (46)
 0     x y  y  z y  
 0 p   
 w u   w v  w 
 
  x z       2  
  y z  z 
 Navier-Stokes equations
Substituting: 1. Continuity equation
  w    w 
       (u ) (v) ( w)
z  x  x  z    0
x y z
and for incompressible flow 2. Momentum equations
u v w
  0 x:
 u u u u  p   2 u  2 u  2u 
   u  v  w    gx     2  2  2 
x y z  t x y z  x  x y z 

after applying above two

conditions in Eqs. (47-49),  v v v v  p   2v  2v  2v 
y:    u  v  w   gy     2  2  2 
we get the Navier-Stokes  t x y z  y  x y z 
equations as following.
 w w w w  p  2w 2w 2w 
z:    u  v  w    gz     2  2  2 
 t x y z  z  x y z 
Vector form of Navier-Stokes Equations
Continuity equation:

 V  0

Navier stokes equation:

V  2
t
  1
 V  V  g  p   V
 
 Turbulent flow
 The most fluid flows are turbulent
 Turbulent flows are characterized by the chaotic and random behaviour of the various fluid
parameters (velocity, pressure, shear stress, temperature and any variable that has a field
description).
 Reynolds (1980) systematically investigated the transition from laminar to turbulent flow in his
experimental setup.
 The Navier-Stokes equations embody the physics of all fluid flows including turbulent flow.
 Practically complex turbulent flow is almost impossible to solve even using numerical schemes.
 Engineers are usually not interested in the complete description of fluctuating motion of turbulent
flows but are interested in studying the effect of turbulence on the mean flow behaviour.
 A statistical approach is usually taken and the equations are averaged over a time scale. The
time scale is long compared to the time scale of turbulent motion but small compared to the
time scale of the unsteady mean flow.
 Time averaged & fluctuating velocity components
 Hydrodynamic quantities of turbulent flow are separated into:
 time averaged values
 fluctuations
 This decomposition of an instantaneous hydrodynamic value → Reynold’s Decomposition

Instantaneous velocity components

(𝑢, 𝑣, 𝑤) & pressure intensity (𝑝):

u  u  u
v  v  v
w  w  w
where:
𝑢, 𝑣, 𝑤, 𝑝: time-average 𝑢, 𝑣, 𝑤 & 𝑝
𝑢′ , 𝑣 ′ , 𝑤 ′ , 𝑝′: fluctuations of 𝑢, 𝑣, 𝑤 & 𝑝

Fig: Time – averaged component & fluctuating velocity components

 If 𝑢 = 𝑢(𝑥, 𝑦, 𝑧, 𝑡) is x- component of instantaneous velocity
 its time- averaged value (𝑢):
t0  T
1
u
T  u  x , y , z , t dt
t0
0 0 0

where T = time interval; & T > period of longest fluctuations

T < any unsteadiness of average velocity

 its fluctuating value (𝑢′): the time varying portion that differs from average value (𝑢)
 Time – average of fluctuations:

1
t0  T
1  t0  T t0  T
 1
u    u  u dt    udt   udt   uT  uT   0
T t0 T  t0 t0  T
 u   v  w  p  0
 Similarly time- averaged values of derivatives of velocity fluctuations also vanish:

u   2 u  uu 
 2  2  ......  0
x x x
 Quadratic terms resulting from the product of cross velocity fluctuations:

u v
u u , u v, 0
x 2

 Velocity fluctuations (𝑢′ , 𝑣 ′ , 𝑤 ′ ) influences time-averaged velocity components

 hence 𝑢, 𝑣, 𝑤 exhibits apparent increase in the resistance to deformation
Reynolds Conditions
 Following relationships are defined as Reynolds conditions & written for 2 quantities (E&G)

E G  E G
constant  E  constant  E
constant  constant
E E

s1 s1
;  Eds   Eds
1 1

E G  E G
EE
E  0
E G  E G
E  G  E  G  0
Reynolds-averaged Navier-Stokes Equations
Continuity Equation: incorporating time-averaging of velocity components:

 (u  u )  (v  v)  ( w  w)
  0
x y z

u  v  w u   v  w
      0
x y z x y z
0

Time averaged velocity components

u  v  w
   0 (A)
x y z
Reynolds-averaged Navier-Stokes Equations
Continuity Equation:

 (u  u )  (v  v)  ( w  w)
  0
x y z

u  v  w  u  v w 
      0
x y z  x y z 
 0 From (A)

Fluctuating velocity components

u  v w
   0
x y z
Observation
 Time-averaged & fluctuating velocity components satisfy continuity equation
 2D flow wherein turbulent components are independent of z- direction:

u 
'

v '


x y

 Increase in 𝒖′ in x-direction will be followed by increase in 𝒗′ in negative y-

direction
 𝒖′ 𝒗′ ≠ 0 and is negative
 Momentum Equations
Time-averaging on momentum equation

  (u  u )  (u  u ) 2  (u  u )(v  v)  (u  u )( w  w) 

    
 t x y z 
 ( p  p)   2 (u  u )  2 (u  u )  2 (u  u ) 
   gx      
x  x
2
y 2
z 2

  (u 2
 2uu   u 2 )  (u v  uv u v  u v)  (u w  uw u w u w) 
 u 
  0
 0 0
 0 0
 t x y z 
 
 2 
p p  u  2
u   2
u  2
u   2
u  2
u 
    gx    2  2  2  2  2  2 
x x  x x y y z z 
0  0 0 0 
Momentum Equations
Time-averaging on momentum equation

 u  (u 2  u 2 )  (u v  u v)  (u w  u w) 
    
 t x y z 
p   2u  2u  2u 
    gx    2  2  2 
x  x y z 

 u u u u  p   2u  2u  2u 
   u v  w      gx    2  2  2 
 t x y z  x  x y z 
  (u 2 )  (u v)  (u w) 
   
 x y z 
Momentum Equations
Similarly for y and z momentum equations:

 v v v v  p   2v  2v  2v 
   u v  w     gy    2  2  2 
 t x y z  y  x y z 
  (u v)  (v2 )  (vw) 
   
 x y z 

 w w w w  p  2w 2w 2w 

   u v  w      gz    2  2  2 
 t x y z  z  x y z 
  (u w)  (vw)  ( w2 ) 
   
 x y z 
Observation
 Additional terms compared to original Navier-Stokes Eq.
 Component of stress tensor

 u u u u  p
   u v  w      g x   2u
 t x y z  x
  xx  yx  zx 
   
 x y z 
Similarly in y and z directions
In matrix form:
 u 2 u v u w
 xx  xy  xz   
     yz      uv v vw 
2

 yx yy
 
  zx  zz  zz  u w vw w  2

 
Observation
 Additional Stresses are known as Reynold’s Stresses
 Total stresses are expressed as:

 u 
 xx    p  2    u2
 x 
 u v 
 xy        uv
 y x 
 Apparent stresses >> viscous stresses
 Turbulence Intensity
 Structure & characteristics of turbulence vary from one flow situation to another & given by:

1. Turbulence intensity → level of the turbulence

 Turbulence intensity may be larger in a very gusty wind than it is in a relatively steady
wind

 Turbulence intensity → square root of the mean square of fluctuating velocity divided
by time - averaged velocity 1/2
1 t0  T

  u  dt 
2

u 2
 T 
Turbulence Intensity   
t0

u u
 larger the turbulence intensity, larger the fluctuations of velocity & other flow
parameters

2. Time scale of fluctuations - period of the fluctuations

 Laminar Shear Stresses
 Laminar flow – modelled as smooth flow of fluid particles in layers

 But, actually numerous molecules dart about in an almost random fashion

 motion is not entirely random — slight bias in one direction → flowrate associated to 𝑢

Across plane A– A:

 molecules moving upward come from area of

smaller average velocity (x-component)

 molecules moving downward come from an area of

larger velocity

 momentum flux (in x-dir) across A–A gives rise to

drag (to the left) of lower fluid on upper fluid &
Fig: Laminar flow shear stress by random motion equal but opposite effect of upper fluid on the lower
of molecules
Laminar Shear Stresses
Across plane A– A :
 sluggish molecules moving upward is accelerated by fluid above A –A

 Rate of change of momentum in this process produces a shear force (on macroscopic
scale)
 Similarly, more energetic molecules moving down are slowed down by fluid below A -A
 Shear Force is present only if there is a gradient in 𝑢 = 𝑢(𝑦)

 Combining the effects of shear force & attractive forces between molecules:
 Newton viscosity law:
u
 
y
 Turbulent Shear Stresses
 random motion of molecules also present in turbulent flow, but, consisting of a series of
random, 3D eddy type motions

 eddies range in size & move about randomly,

conveying mass with an average velocity, 𝑢 = 𝑢(𝑦)

 eddy structure promotes mixing within the fluid &

increases transport of x momentum across A– A

 finite fluid particles (not merely individual molecules as

in laminar flow) are randomly transported across A-A

 relatively large shear force when compared with

laminar flow

Fig: Turbulent flow as a series of random, 3D eddies

Turbulent Shear Stresses
 Random velocity components responsible for momentum transfer & shear force:
1. 𝑢′ → for x component of velocity
2. 𝑣 ′ → for the rate of mass transfer crossing A-A
 Apparent shear stress on plane A– A

u
    u v   laminar   turbulent
y
 In laminar flow, 𝑢′ = 𝑣 ′ = 0 → Eq. reduces to the customary random molecule-motion-
induced laminar shear stress (τlaminar)

 For turbulent flow, turbulent shear stress, is positive (τturbulent= −𝜌𝑢′ 𝑣′)

 Hence, shear stress is greater in turbulent flow than in laminar flow

 Terms of the form −𝜌𝑢′ 𝑣′ are called Reynolds stresses in honor of Osborne Reynolds
 Turbulent Shear Stresses
 In a very narrow region near the wall, within the viscous sublayer, τlaminar is dominant

 Away from the wall, in the outer layer, τturbulent is dominant

 Transition between these two regions occurs in the overlap layer.

 viscous sublayer is usually very thin,

adjacent to the wall
 no-slip condition & wall shear stress occur
in this layer
 turbulent pipe flow properties quite
dependent on roughness of pipe wall,
while laminar pipe flow is independent of
roughness
 scratches, rust, sand or dirt particles
easily disturb viscous sublayer & affects
flow

Fig: Turbulent flow in a pipe (a) Shear stress (b) Average velocity
 Eddy Viscosity
 Alternate form for shear stress for turbulent flow: in terms of the eddy viscosity, η

u
 turbulent  
y

 Unlike absolute viscosity, which is a known value for a given fluid, eddy viscosity is a
function of both fluid & flow conditions

 Eddy viscosity of water cannot be looked up in handbooks—its value changes from one
turbulent flow condition to another and from one point in a turbulent flow to another

 Several semi-empirical theories have been proposed to determine approximate η values

 L. Prandtl, a German physicist & aerodynamicist, proposed Prandtl Mixing Length Theory
 Prandtl Mixing Length Theory
 Consider fully developed turbulent flow, with main flow direction parallel to x- axis

 Time- averaged velocity components:

u  u ( y ); v  w  0
 Fluctuating component 𝑣′ transports mass & momentum across a plane at y, from the wall

 Total shear stress

 u v  u
        uv     u v
 y x  y
 0 

   laminar   turbulent
 turbulent    uv
 Prandtl Mixing Length Theory
 Lump of fluid which comes to layer 𝑦1from layer (𝑦1 − 𝑙) has positive 𝑣 ′

 If lump retains its original momentum, then velocity at its current location 𝑦1 is smaller than
the velocity prevailing there
 u 
 Difference in velocities: u  u ( y )  u ( y  l )  l  
 y  y1
1 1

 Prandtl proposed that the transverse displacement of

fluid particles is, on an average, 𝑙 and this length scale is
known as Prandtl’s mixing length

 Consider another lump of fluid with negative 𝑣 ′ arriving at

𝑦1from layer (𝑦1 + 𝑙)

 If lump retains its original momentum, mean velocity at

the current location 𝑦1 is somewhat more than original
mean velocity
 Prandtl Mixing Length Theory

 u 
 Difference in velocities: u2  u ( y1  l )  u ( y )  l  

  y1
y

 Velocity difference caused by transverse motion can be regarded as turbulent velocity

components at 𝑦1
1  u 
 Time- average of the absolute value of fluctuations: u   u2  u1   l  
2  y  y1
 Suppose these two lumps of fluid meet at 𝑦1

 Lumps will collide with a velocity 2𝑢′ and diverge

 This proposes the possible existence of transverse velocity component in both directions
with respect to the layer at 𝑦1
 Prandtl Mixing Length Theory
 Suppose these two lumps of fluid move away in a reverse order from 𝑦1 with velocity 2𝑢′

 Empty space will be filled from the surrounding fluid creating transverse velocity
components which will again collide at 𝑦1

v  u 
u
 v  const  u   const  l
y
 along with the condition that moment at which 𝑢′ is positive, 𝑣 ′ is more likely to be
negative and conversely when 𝑢′ is negative

2
2  u 
   
u v  C1 u v  C2l  
 y 
C1 & C2 are proportionality constants
 Prandtl Mixing Length Theory
 Including C2 in unknown mixing length

2 2
2  u  2  u  2 u  u  u
u v  l     turbulent    u v   l     l    
    y
t
 y  y y  y
where 𝜇𝑡 is the apparent viscosity

 Near a solid wall, 𝑙 is approximately proportional to distance from the wall and Karman
suggested
l   y;   0.4

where K is the Karman constant