Ocon et al CFD Letters Vol.

4(3) 2012

www.cfdl.issres.net Vol. 4 (3) – September 2012

Theory-derived law of the wall for parallel flat-

plates turbulent flow
Joey D. Ocon1,2,c , Ludek Jirkovsky3, Rizalinda L. de Leon1, Amador C. Muriel4,5
1
Department of Chemical Engineering, College of Engineering,
University of the Philippines Diliman, PHILIPPINES
2
Electrochemical Reaction and Technology Laboratory (ERTL), School of Environmental
Science and Engineering, Gwangju Institute of Science and Technology, REP. OF KOREA
3
Department of Informatics and Geoinformatics, Fakulta Zivotniho Prostredi, FZP,
University of J.E. Purkyne, CZECH REPUBLIC
4
Department of Electrical Engineering,
Columbia University, New York City, U.S.A
5
Department of Philosophy,
Harvard University, 25 Quincy Street, Cambridge, MA 02138, U.S.A

Received: 20/11/2012 – Revised 25/07/2012 – Accepted 31/07/2012

Abstract

It is well known that in a turbulent flow between two parallel flat plates, the horizontal
mean velocity varies logarithmically with height (the so-called 'logarithmic-law-of-the-
wall'). The law of the wall is a description of the mean velocity profile in wall bounded
flows and has been regarded as one of the underpinning doctrine in the turbulence
community for more than half a century. Much of our understanding in wall turbulence
has been based from the continuum Navier-Stokes Equation (NSE). More recently,
following studies of a modified Navier Stokes Equation, we applied a modified
incompressible NSE to the flow of turbulent fluid between two parallel flat plates and
solved it analytically [1]. We extended the analysis to the turbulent flow along a single
wall and compared the results with the established controversial von Karman logarithmic
law of the wall [2]. We found velocity profiles and velocity time evolution of a turbulent
system, through simple numerical simulations, that cannot be reproduced from the
classical NSE

Keywords: post-Navier Stokes equation; law of the wall; analytical solution; numerical
simulations; flat-plate system

1. Introduction

For more than a hundred years, fluid turbulence has been has been one of the greatest
mysteries of science. It has been an age-old topic of discussion among fluid dynamicists.
Turbulence, as a state of fluid motion, has been understood before as something that is governed by
dynamical laws such as the Navier-Stokes equation. Most fluid dynamicists agree that not only the
problem of turbulence is still far from being solved, but also it is extremely difficult to agree on
what is the problem to be solved [3].
The traditional method in standard hydrodynamics is to solve Navier- Stokes equation for
stationary velocity profiles, which are parabolic for Poisseuille-Hagen flow in laminar regime and
in good agreement with experiments. However, any agreement breaks down at the onset of

93
 Theory-derived law of the wall for parallel flat-plates turbulent flow

turbulence as velocity profiles flatten and become non-stationary [4-5]. An explanation for the
flattened velocity profiles was given by Prandtl and von Karman using the transverse component of
the fluctuation of the velocity [5]. The faster molecules of the central region of the pipe show up in
the boundary layer mix with slower molecules, and the velocity profile becomes roughly uniform
except in the boundary layers.
For the description of turbulent flows, many renormalized perturbation theories all based on
Reynolds equation were developed in the past [6]. The Reynolds equation, which aside from the
mean velocity considers its arbitrary time fluctuations, like Navier-Stokes equation, ignores the
structure of the molecules. Limitations of the Reynolds equation in the laminar-turbulent transition
and its difficulty in explaining the origin of turbulence suggest study of the phenomenon directly
from the Liouville equation.
It is known that difficulties arise when turbulence is studied from the standard Navier-Stokes
equation alone. That was first observed by D. Ruelle [7]. One such recent effort modifies the Navier
Stokes equation by a term resulting from a microscopic molecular consideration ignored by
conventional hydrodynamics [1]. A quantum kick imparts a constant unit of momentum with certain
probability per unit time to the molecules of the gas. This is a way of injecting energy into a system
to see if turbulence results.
The law of the wall for the interior part of a wall-bounded turbulent shear flow is the
cornerstone of fluid dynamics, and one of the very few pieces of turbulence theory whose outcome
includes a simple analytic function for the mean velocity distribution, the logarithmic law. For wall-
bounded flows, the so-called 'log law' is widely held to describe most turbulent wall-bounded flows,
and lies at the heart of the most commonly used engineering computational models concerning
turbulent flow in close proximity to surfaces. While there are several forms of the log law, the most
1
common is the mean velocity profile normalized in variables given by U   ln y   Bi where
K
 
U  U / u* and y  yu* / v . U is the mean velocity, y is the distance from the wall, v is kinematic
viscosity and u* is the friction velocity defined from the wall skin friction,  W , as u*2   W / 
where  is the fluid density. The von Karman constant, K , and the additive constant Bi is widely
thought to be universal constants. However, there seems to be little consensus on the values of these
constants. For several decades, K was believed to be 0.41, but a few years ago Nagib estimated it
to be 0.38 and K can be as high as 0.45 [9].
The history and theory supporting the universality of the log law for turbulent wall-bounded
flows was examined by W. George. George articulated that the idea of a universal log law for wall-
bounded flows is not supported either the theory or the data [9]. Several modifications of the log
law of the wall for turbulent flow in smooth pipes are proposed already which are based from
empirical data, one being a law consisting of three terms: a logarithmic term, a sine-square term and
a cubic term [10]. Virtually most turbulence models are calibrated to reproduce the law of the wall
in simple flows, such that when the law of the wall fails, current Reynolds-averaged turbulence
models are not susceptible to failure [11].
Not long ago, attempts were made by several researchers such as Chorin and Barenblatt to
describe the turbulent flow along a single plate by a power law [12]. They proposed that the
velocity profile is not universal but a weakly varying power law with coefficients that vary with
Reynolds Number, in general, of the form U  Cy  where C and  are Reynolds number dependent
empirical constants, while von Karman log law is Reynolds number independent and is a good
approximation for distances far enough from the wall only, the so called intermediate or overlap
region. It fails in the regions near the wall, (viscous and buffer sublayers), and for very large
distances from the wall. In the region near the wall, the viscous sublayer, the power law holds
approximately for   1 , (linear dependency). Their analysis of the new experimental data adduces
additional arguments against the von Karman-Prandtl universal logarithmic law and in favor of a

94
 Ocon et al CFD Letters Vol. 4(3) 2012

specific power law. However, measurements done by Zanoun and Durst for high Reynolds numbers
support the approximate validity of the log law in the intermediate or overlap region (inertial
sublayer) [13].
This paper will present the equations derived from the modified Navier Stokes Equation
which will be compared with known flow characteristics. We have applied flat plate boundary
conditions to generate the respective velocity fields and corresponding velocity time evolution. The
generated velocity profiles in the flat plate geometry will be explored and compared with
experimental data. All of the flows we considered in these studies will be assumed to have
undergone transition and is fully developed. The treatment of the equations will be paying attention
entirely on smooth walls for which surface roughness have no quantifiable influence. The model
reproduced the velocity profiles of turbulent flow between two flat plates. Extending the analysis to
the flow along a single wall by letting the distance between two flat plates to go to infinity, we
arrive at a result qualitatively similar to known results, nevertheless fundamentally different from
the logarithmic law known to have discrepancies with empirical data by even 65% [2].

2. Application to a Flat-Plate System


We postulate a modified Navier-Stokes equation for the macroscopic velocity U of the form

U i 1 
 U j  jU i   U i    j Pij  ; i, j  1, 2, 3 (1)
x  m
where the last term of both sides of the equation are adopted from [1] to represent quantized kicks
to the system, with  as the probability per unit time that a particle of mass m is imparted a
momentum kick  . It can be interpreted as forcing by a paddle wheel which increases the velocity
of a particle in a fluid. This is no longer a purely continuum model. We shall justify this ad-hoc
approach from the apparent success of the model in explaining empirical data heretofore
unexplained by continuum mechanics [4, 14]. All other terms follow conventional definitions. The
justification of the strange terms appearing in the Equation 1 comes, besides its utility, as a result of
inelastic interactions among the molecules of the fluid. Thus the contribution from the collision
integral in Boltzmann transport equation used in their derivation does not reduce to zero unlike in
the derivation of standard incompressible NS equation. From the hypothesis about the quantum
nature of turbulence, dissipative effects are present from excitation to higher energy state of the
particles due to inelastic collisions. It is well known that inelastic interactions among the molecules
of the fluid result in deterministic chaos we associate with turbulence.
If one uses a decomposition of the pressure tensor Pij into its diagonal and off-diagonal
parts, p ij   ij , i, j  1, 2,3 , where p is the pressure and  ij   ( iU j   jU i ) is the component of
shear stress tensor and applying incompressible fluid condition  jU j  0 , Equation 1 simplifies to
U i  1
 U j  jU i   U i    ( i p   j 2U i ) (2)
t m 
where  is the kinematic viscosity. For the flat plates configuration the mean velocity vector is

U  (U  z , t  , 0, 0) and we can further reduce it to Equation 3.
 U  2U
 kU  2  g (3)
 t z

The parameter k  is proportional to the probability of kicking particles to different momentum

1 k
and g    x p  represents the constant effective pressure head per unit kinematic viscosity
 m
that drives the motion of the fluid. The solutions to the Equation 3 did not require numerical

95
 Theory-derived law of the wall for parallel flat-plates turbulent flow

computations since it is easily solved analytically upon application of the necessary flow conditions.
All plots were generated by substituting parameter values on the governing equation and adjusting
the scaling of the plots to match with widely known flow profiles.

3. Results and discussion

3.1 New Law of the Wall

We find the stationary solution of Equation 3 using non-slip boundary conditions,

U  0, t   U  L, t   0 and static initial condition, U  z , 0   0 and subject to flat plate
boundary conditions, U  0   U ( L) , with plates at a distance L apart,
g  sinh k  z  L   sinh k z 
U z  1  , k  0 (4)
k  sinh k L 

The results for the flat plate system are plotted in figure 1. Fixed parameters used for the
numeric computations are: g  100 and L  1 . For the numerical results for the two-plate
system, it exhibits broad flattening of the velocity profiles, for large k values, for example
with probability parameter k  250 . It can be observed that for increasing k values, the
velocity profile looks more like the predicted logarithmic curve from empirical studies. On
the other hand, velocity profiles for lower values of the control parameter k are nearly
parabolic or circular which are usually associated with laminar fluid flow. The flattening of
the velocity profiles can be simply explained by increase in number of inelastic molecular
interactions. With increased inelastic collisions, velocity profile exhibits flattening at the
center which is characteristic of turbulence.
The velocity profile of the flow along a single flat plate, z  0 , can be obtained simply
by letting the distance, L , approach infinity. Then for the mean velocity of the flow we get

g  sinh k  z  L   sinh k z 
U w  z   lim 1   (5)
L k sinh k L
 
g
Uw  z  
k

1  exp  k z 
  (6)
The result for the wall turbulence for a single wall flow is plotted in Figure 2, with
the same parameters, except L   . For small distances z from the wall (boundary viscous
sublayer), the exponential velocity profile reduces approximately to linear in good
agreement with observations. The exponential description of the velocity profile deviates
from the logarithmic law for small distances z , and is somehow consistent with an
exponential profile derived using a field theoretical method in [8].

3.2 Velocity Time Evolution

A complete non-stationary solution of equation 3, the time development of the velocity
profiles, can be found. If one writes a solution of equation 3 with g  0 as
 vkt   vkt 
uG1  Aexp    2
  By  z  and another in the form uG  Cexp    y  z  where A , B
     
and C are constants to be determined. The solution of Equation 3 for g  0 , satisfying
g
static and non-slip boundary conditions can be constructed as, U  z , t    uG1  uG 2 . With
k

96
Ocon et al CFD Letters Vol. 4(3) 2012

 sinh k z  sinh k ( z  L)  g
y  z     and constants A  B  C   the time evolution of
 sinh k L  k
the profile is,

g   vkt   g  vkt 
U  z, t   1  exp     y ( z )   y  z  exp    (7)
k      k   

 vkt 
The exponential term exp    can be neglected as t   , for equation 7 to
  
simplify into the equation 4 form. The same fixed parameters were used for the velocity
profile progression plots in figure 3. It shows how the velocity profile approaches that for
equation 4 for increasing t values.

Figure 1. Stationary velocity profiles for turbulent flow between two flat-plates for
k  1,50,100,150, 200, 250 .

97
 Theory-derived law of the wall for parallel flat-plates turbulent flow

Figure 2. Wall Turbulence in a single plate for k  1,50,100,150, 200, 250 .

Figure 3. Velocity time evolution for turbulent flow in a two-plate system for various t values

98
 Ocon et al CFD Letters Vol. 4(3) 2012

3.3 Model fitness with experimental data

In Figure 4, we compared the fit of our equation with experimental data at low
Reynolds numbers for turbulent flow in flat plates [16]. The exact data points used by Shiyi
Chen et al. to validate the results of their viscous Camassa-Holm equations as a closure
approximation for the Reynolds averaged equations of the incompressible Navier-Stokes
fluid [15]. The velocity profile was plotted as well for the von Karman log law predictions
using 0.41 and 5.5 for the von Karman constant and the additive constant, respectively [13].

Figure 4. Mean Velocity Profile in the Channel for the new Law of the Wall and von
Karman log law compared with the experimental data.

As the Reynolds numbers for these experiments are small, g , k , and L somehow
have not reached their theoretical asymptotic values. Our new law of the wall used the
abovementioned adjustable parameters for the fitting. We obtained the best data fit for the
experimental data when we assumed that the probability parameter k is slowly changing
with the wall distance z . By definition, k is proportional to the probability of kicking
particles to different momentum. Hence, k can be a function of the wall distance due to the
varying levels of turbulence along the wall. As seen in Figs. 1 and 2, the shape of the profile
is greatly affected by the parameters k and g which for the optimized fitting were found to
be k  z   0.0001  0.000046 z and g  0.00522 . Parameter L has minimal effect on the
velocity curves for values greater than 1000. This assumption is justified since it is well
established that the fluid is more turbulent near the wall. To provide proper scaling of the
experimental data, g has been chose to have a small value. It is also known that the von
Karman log law fails in the regions near and far from the wall. The new exponential curve
on the other hand explains the linear behaviour near the wall. As predicted from our
equation, at the region very near the wall, equation 6 simplifies to a linear form,
U w  gz / k . We emphasize here that our predictions are consistent with the von Karman
and experimental data at a wide range of distances from the wall, as seen in Figure 4.

4. Conclusion

In the present paper, a new theory of the law of the wall, regarding the nature of wall-bounded
flow turbulence, is outlined in brief. In particular, the the analytical solutions of the modified NSE

99
 Theory-derived law of the wall for parallel flat-plates turbulent flow

in stationary and non-stationary conditions and the validity of the law derived is focused upon. A
self-consistent theoretical formulation for the law is derived which matches the experimental data at
the log layer. Upon examination of the resulting velocity profiles in a fully developed two-plate
system flows indicates that it can be used to model channel flows. The present observations
emphasize that the above results cannot be obtained from the classical NSE.
Further data must be used to test the new law of the wall equation developed from the modified
NSE especially the data from the measurements done on the turbulent velocity profiles in channel
and circular pipe flows for a wide range of Reynolds numbers. To get the solution of the Equation 3
in cylindrical coordinates it suffice to replace in flat plate non-stationary solution the hyperbolic
sine functions by first modified Bessel function I 0 . Qualitatively, similar result was obtained and
tested against experimental data by Chen et al. [15]. Their rather complicated result for wall
turbulence velocity profiles was in terms of hyperbolic cosine functions and some power law terms
obtained from a closure of Reynolds-averaged equation (Camass Holm equations). For the turbulent
pipe flow the cosh functions could be just replaced by first modified Bessel functions in agreement
with our result except the power law terms [17]. We presented a more simple result. Also, the
fundamental assumption requires further investigations, both experimentally and theoretically, i.e.
on basis of the quantum nature of turbulence.

Acknowledgement

We acknowledge the assistance of the Balik-Scientist Program of the Department of Science

and Technology (DOST), Republic of the Philippines, DOST-UP ERDT Faculty Development
Program, and the Department of Chemical Engineering at the University of the Philippines Diliman.

Nomenclature

U Mean velocity
v Kinematic viscosity
u* Friction velocity
W Wall skin friction
 Fluid density
K von Karman constant
Bi Additive constant in the von Karman logarithmic law
 Coefficient in Chorin and Barenblatt power law
 Probability per unit time that a particle of mass m is imparted a momentum
kick 
 Momentum kick
Pij Pressure tensor
 ij Components of the shear stress
g Constant effective pressure head per unit kinematic viscosity that drives the
motion of the fluid

References

[1] Getreuer, P., A. Albano, A. Muriel. Possibility of turbulence from a post-Navier-Stokes

equation. Physics Letters A, 2007. 366(1-2): p. 101-104.
[2] Cipra, B., A new theory of turbulence causes a stir among experts. Science, 1996. 272(5264):
p. 951
[3] Aly, H.S. Towards a Better Understanding of Turbulence. CFD Letters, 2009. 1(1): p. 2-3

100
 Ocon et al CFD Letters Vol. 4(3) 2012

[4] Landau, L.D., and Lifschitz, E.M. Hydrodynamics (Nauka, Moscow, 1988).
[5] Fox, G. and Germano, F. Fluid mechanics, Van Nostrand Co. Inc., Princeton, USA. 1961.
[6] McComb, W.D. The physics of fluid turbulence. Oxford, Clarendon Press. 1992.
[7] Ruelle, D. The Turbulent Flow as a Dynamical System, New Perspectives in Turbulence,
edited by L. Sirovich L. Springer, New York, USA. 1991
[8] Oberlack, M. A unified approach for symmetries in plane parallel turbulent shear flows.
Journal of Fluid Mechanics, 2001. 427(1): p. 299-328.
[9] George, W. K. Is there a universal log law for turbulent wall-bounded flows? Philosophical
Transactions of the Royal Society A, Mathematical, Physical and Engineering Sciences. 2007,
365(1852): p. 789-806.
[10] Guo, J. and P.Y. Julien. Modified log-wake law for turbulent flow in smooth pipes. Journal of
Hydraulic Research, 2003. 41(5): p. 493-501.
[11] Bradshaw, P. and G.P. Huang. The Law of the Wall in Turbulent Flow. Proceeding of the
Royal Society of London A, 1995. 451(1941): p. 165-188.
[12] Chorin, A., and G. Barenblatt. Scaling of the intermediate region in wall bounded turbulence:
The power law, Phys. Fluids, 1998, 10: p. 1043-1044.
[13] Zanoun, E., F. Durst, H. Nagib. Evaluating the law of the wall in two-dimensional fully
developed turbulent channel. Phys. Fluids, 2003. 15(10) 3079-89.
[14] Fox, G. and F. Germano. Fluid mechanics, Van Nostrand Co. Inc., Princeton, USA. 1961.
[15] Chen, S., C. Foias, C. Holm, E. Olson, E. Titi, S. Wynne. Camassa Holm equations as a
closure model for turbulent channel and pipe flow. Physical Review Letters, 1998. 81(24): p.
5338–5341
[16] Wei, T. and Willmarth, W.W. Reynolds number effects on the structure of a turbulent channel
flow. J. Fluid Mechanics, 1989. 204: p. 57-95.
[17] Abramowitz, M. and Stegun, I.A., Handbook of Mathematical Functions. Dover, New York,
USA. 1972, 9th ed.

101