Sie sind auf Seite 1von 5

Recent Researches in Mechanics

Direct computation of evaporation rate at the


surface of swimming pool
Nikolay A. Vinnichenko, Alexander V. Uvarov, Dmitry A. Vetukov, and Yulia Y. Plaksina.

reviews in [7] and [8]). All of them account for dependence of


Abstract Numerical model is proposed for simulating water evaporation rate on vapor pressure or absolute humidity
evaporation at the surface of swimming pool. It incorporates only difference between the air near the water surface and the
simple physical boundary conditions without prescribing mass flow ambient air. Some take into account wind velocity. Only a few
at the surface. 2D-simulations are performed. Correlation is analyzed
consider dependence on pool size. The main drawback is that
in terms of Sherwood-Rayleigh number relation. Evaporation rates
calculated numerically are shown to fall within the range of existing their applicability is usually limited to the range of parameters
empirical correlations for positive Rayleigh numbers. For negative studied in particular experiment.
Rayleigh numbers universal curve still exists but Sherwood number The development of modern computational tools did not
is less than estimated according to known correlations. Possible change the situation. In spite of complicated hydrodynamics
dependence of the evaporation rate on swimming pool size is modeling empirical correlations are often used in
discussed. Also, the model is useful in solving unstationary problems
computational studies in order to specify boundary condition
e.g. finding the temperature difference between air and water near the
surface of water reservoir. Temperature difference is shown to for the mass flux at air-water interface. For example, [9]
depend linearly on relative humidity of the ambient air. employed empirical correlation [4] and [10] used
phenomenological and analytical correlations [5]. Commercial
Keywords evaporation rate, humidity, convection, CFD. CFD codes were used to solve Navier-Stokes equations with
turbulence modeling. Then, in both studies, calculated
I. INTRODUCTION velocity, temperature and humidity fields were compared to
experimental measurements which are local in nature. Thus,
P REDICTING evaporation rate for various indoors and
outdoors water reservoirs has been attracting research
activity for many years [1]-[15]. It is important for planning
evaporation rate was not calculated directly. Instead, it was
provided by existing empirical correlations and verified by
energy consumption in swimming pool facilities, evaluating of comparison of the resultant fields for different empirical
cooling effect in air-conditioning systems, and predicting correlations.
liquid mass loss in reservoirs. Besides these industrial Another related problem is temperature and humidity profile
applications, there is also fundamental physical problem of near the surface of water reservoir in geophysics [16]. Under
explaining the temperature difference and profile near water various conditions different profiles were measured with thin
surface, which is presumably determined by evaporation. In films of cold or hot water at the surface, strong gradients near
1918 an empirical correlation was proposed [3] relating the surface and temperature difference up to 4 degrees between
evaporation rate with vapor pressure difference and wind air and water surface. As shown below, this difference can be
velocity. It is still recommended by ASHRAE as useful obtained numerically by solving unstationary problem for
approximate formula. Since, many empirical correlations have given air temperature and humidity far from the water surface.
been developed for evaporation in still and moving air, for The goal of the present study is to elaborate physical
different ranges of water and air temperatures, for unoccupied boundary conditions which allow to calculate evaporation rate
and occupied pools with various kinds of activity (see e.g. the directly without making use of empirical correlations. They are
simple, local and can be incorporated into any existing CFD
code with minor modifications. Yet, they differ from fixing
Manuscript received May 13, 2011. This work was supported in part by some value for mass or heat flux at the interface, so their
the Russian Foundation of Basic Research under Grant 09-08-00961.
N. A. Vinnichenko is with Faculty of Physics, Moscow M.V. Lomonosov
implementation by user of commercial package is not
State University, Leninskiye Gory, 1/2, Moscow, Russia (phone: straightforward.
+74959392741; fax: +74959328820; e-mail: nickvinn@ yandex.ru). The rest of the paper is organized as follows: Sec. II
A. V. Uvarov is with Faculty of Physics, Moscow M.V. Lomonosov State
University, Leninskiye Gory, 1/2, Moscow, Russia (e-mail: uvarov@phys.
describes the proposed boundary conditions coupled with
msu.ru). entire problem formulation for both problems. In Sec. III and
D. A. Vetukov is with Faculty of Physics, Moscow M.V. Lomonosov State IV we discuss the results of 2D-simulations for hot and cold
University, Leninskiye Gory, 1/2, Moscow, Russia (e-mail: water cases and compare them to known empirical
vetukov_1971@mail.ru).
Y. Y. Plaksina is with Faculty of Physics, Moscow M.V. Lomonosov State correlations. Unstationary problem of modeling the
University, Leninskiye Gory, 1/2, Moscow, Russia (e-mail: temperature difference between air and water, which results
yuplaksina@mail.ru).

ISBN: 978-1-61804-020-6 120


Recent Researches in Mechanics

from evaporative cooling, is discussed in Sec. V. Possible of state. Here csat (T ) is saturated vapor density, is relative
extensions of the present model and directions for further humidity of air far from the pool.
studies are outlined in Sec. VI. v x / x = v y / x = T / x = c / x = 0 at the left and right
boundaries, v x / y = v y / y = T / y = c / y = 0 at upper
II. NUMERICAL MODEL
boundary, v x = v y = T / x = c / x = 0 at lower boundary
A. Problem Formulation except for the pool surface. Computational domain is extended
We consider natural thermal and concentration-induced using increasing-grid-step layer more than 200 pool lengths to
convection in the air above water reservoir. Water temperature the left and to the right and more than 600 pool lengths
and parameters (temperature and relative humidity) of the air upwards in order to avoid interaction between boundary
far from the water surface are maintained constant. Only air- conditions and the developing flow.
side is simulated, water is represented by boundary conditions Problem formulation for finding the temperature difference
described below. Air dynamics is governed by Navier-Stokes between air and water is almost the same except that
equations complemented by transport equation for vapor temperature is uniform in the initial state. Water surface
temperature evolution is controlled by evaporation mass and
density. Compressibility effects are taken into account in
heat balance at the surface, which is described by boundary
simplified form, neglecting dp / dt in energy equation. This
conditions. Hereinafter we will refer to problem with fixed
allows to use numerical methods developed for incompressible water temperature as case A and to problem with water
fluid without considering the sound waves. Correspondingly, cooling due to evaporation as case B for brevity.
time step limitation becomes associated with local gas velocity
B. Boundary Conditions at Air-Water Interface
instead of speed of sound, which is important for slow
convective flows simulations. Both experimental data and numerical computations [9]
show that convection velocity in water is 10-20 times less than
air-side velocity. Moreover, water heat conductivity is
+ div( v ) = 0,
t relatively large, so temperature gradient is negligible in water.
dv Thus, it is possible to use no-slip condition v = 0 at the
= grad p + (v ) + grad div v + g,
dt 3 boundary and to calculate air-side only. Slow decrease of the
dT v v 2 v l
2 (1) water level due to evaporation is neglected.
c p = ( a T ) + i + k ik + g v,
dt 2 x k x i 3 x l According to Hertz-Knudsen relation [17], evaporative mass
dc
flux from water to the vapor near the surface equals
= (Dc ),
dt
p ( y) a a w p (T ) p (T )
= 0 1 c. m = sat sat + , (2)
RT w 2R T T+

Here , v , p and T are density, velocity, pressure and


where T and T+ are temperatures below and above the water
temperature of air/vapor mixture, c is vapor density, , a
surface (at both sides of Knudsen layer), psat (T ) is saturated
and c p are air viscosity, heat conductivity and specific heat, vapor pressure. Each of the terms in (2) is about 2-7 kg/(m2s).
a and w are molecular masses of air and water vapor, D is However, evaporation rate is usually less than 10-3 kg/(m2s).
diffusion coefficient for water vapor in air, R is the gas This means that two terms in (2) are nearly balanced. One can
constant, g is acceleration of gravity, p0 ( y ) is the base accept that T+ T and the vapor near the water surface is
pressure, corresponding to barometric formula. Turbulence close to saturation. This is due to the fact that vapor removal
modeling is not included so far. Typical velocities of from the surface by diffusion is much slower than vapor
convective flow are small and there is no obstacle to create generation by evaporation.
high-turbulence regions. However, Rayleigh number is large For case A, when we are interested in finding evaporation
and turbulence can be generated by buoyancy. Hence, present rate for given air and water temperatures and air humidity,
simulations should be considered as compulsory laminar, with temperature of water surface is fixed: T ( y = 0) = Tw , and
further verification including turbulence models. humidity at lower boundary corresponds to saturated vapor at
Computations are performed in 2D-configuration. The pool this temperature: c( y = 0) = csat (Tw ) . If we consider
surface occupies part of the lower border ( y = 0 , unstationary problem (case B), water temperature is allowed to
0.5 x 0.5 ). The rest of the lower border is treated as evolve according to heat balance equation
adiabatic wall. Following initial conditions are used:
v x = v y = 0 , p = p0 ( y ) , T and c take values Tw and Ta , T
0 = a J h (T Ta ) + ql qr , (3)
cw = csat (Tw ) and ca = csat (Ta ) at the water surface and far y
from the pool with smooth transition, follows from equation

ISBN: 978-1-61804-020-6 121


Recent Researches in Mechanics

where J = Dc / y is evaporation rate, h is latent heat of


vaporization, is efficient heat transfer coefficient for heat
transfer by radiation and through walls of the pool,
q = w T / x is heat flux along the water surface, indices l
and r denote left and right approximations, w is water heat
conductivity. The first term in the right-hand side of (3)
describes heat exchange between air and water surface, the
second one represents cooling due to evaporation. Heat
transfer by radiation and through the walls is taken into
account by the third term (radiation heat flux can be linearized Fig. 1. Snapshot (a) and time-averaged field (b) of temperature for
because temperature difference is small), and heat conductivity evaporation from hot water surface.
in water is also considered. Boundary condition (3) is
approximated implicitly and solved along water surface. Numerous empirical correlations have been proposed for
Humidity derivative is averaged over three recent time steps in evaporation rate at fixed water and air temperature and
order to avoid stability problems. humidity. Most of them take the form (see [7] for review)
Heat balance equation (3) (or fixed temperature of water in
case A), coupled with c = csat (Tw ) and no-slip condition for J = (a + bV )( pw pa ), (4)
velocity, constitute the complete set of boundary conditions for
air-water interface. Note that no analytical expression is where V is horizontal wind velocity, p w is saturated vapor
prescribed for mass flux. Evaporation rate is calculated
directly from physical modeling. pressure at water surface temperature, p a is partial vapor
pressure far from the water surface, a and b are empirical
C. Computational details
coefficients. Fig. 2 shows some of the calculated results in J -
The problem is solved using semi-implicit scheme for fluid ( pw pa ) coordinates for different pool sizes L (in meters).
with simplified description of compressibility effects [18].
Though the curve for each value of L can be approximated
Convective terms are approximated with 4-points upwind
with linear dependence, it is obvious that the slope depends
differences. Crank-Nicolson scheme is used for diffusive
strongly on the pool size. The empirical formulae given in
terms. Time integration is performed by 3rd-order Runge-Kutta
form (4) are valid only for pools similar to those studied in
method. Pressure is obtained by solving Poisson equation of
experiments.
fractional-step method. Computations are performed on
200160-points grid. Minimal non-dimensional grid step is
2 10 3 in vertical direction and 0.025 in horizontal direction.
Calculations are performed until quasi-stationary regime is
obtained. Then flow variables and evaporation rate are
recorded and averaged over a long time period (e.g., 2000 s
for a pool size 1 m). It should be noted that the present model
does not include volume condensation, which can take place
under certain conditions. Therefore, only the results for
computational runs, during which maximal relative humidity
did not exceed 1, are presented.
Heat conductivity and viscosity are supposed to be constant
since temperature differences are not large. However, power-
law temperature dependence is used for diffusion coefficient
D = (2.05 10 5 m 2 / s ) (T / 273) 2.072 . Parabolic fit is used for
saturated vapor pressure data in the range 0-50C. Fig. 2. Simulation results for evaporation rate vs. vapor pressure
difference.
III. FIXED WATER TEMPERATURE: POSITIVE RAYLEIGH
NUMBERS
However, there are several correlations, written in
If water is hot, concentration-induced convection is assisted nondimensional form of relation between Sherwood number
by thermal convection. Plumes of hot humid air are irregularly Sh = L J /( D (c w c a )) and Rayleigh number
produced in different locations at water surface. The final state
Ra = gL3 Pr / 2 , where Pr = c p / a is Prandtl number
is not stationary. Snapshots of temperature and humidity fields
are similar to those observed in flames. Snapshot and time- and is the difference between air/vapor mixture densities
averaged temperature field are presented in Fig. 1. Pool size is far from the pool and at water surface. Note that Rayleigh
0.5 m, = 0 , Ta = 300 K, Tw = 302 K. White rectangle number takes into account both the effects of thermal and
represents pool surface. concentration-induced convection. These correlations have the

ISBN: 978-1-61804-020-6 122


Recent Researches in Mechanics

form (see [8]) and stationary. Numerical results are shown in Fig. 4 as
Sherwood number relation to absolute value of Rayleigh
Sh = B Sc1 / 3 Ra n , (5) number. There is still universal curve in this case, but it lies
considerably lower than known correlations. Note that all
correlations except [13] were obtained for hot water case and
where Sc = /( D ) is Schmidt number. Fig. 3 presents the
are presented in Fig. 4 for qualitative comparison only. Power-
comparison between simulation results and known correlations
law approximation yields Sh = 0.981 | Ra |0.154 with even lower
for positive Rayleigh numbers. This implies positive, zero or
slightly negative temperature difference between water and air exponent.
as thermal convection usually dominates over concentration-
induced convection. Universal curve is obtained with good
accuracy in contrast to J - ( p w p a ) plane. Numerical results
lie within the range of existing correlations though Sherwood
numbers are less than predicted by most of them, especially for
high Rayleigh numbers. This can be associated either with the
absence of turbulence modeling or with two-dimensional
geometry of simulations. Note that both for analytical [5] and
empirical [15] correlations turbulent regime yields higher Sh .
Obtained power-law approximation Sh = 1.11 Ra 0.197 has
exponent lower than any of existing correlations (from 0.205
[13] to 0.327 [15]). The exponent in Sherwood-Rayleigh
relation determines evaporation rate dependence on pool size:
J ~ L3n 1 . If exponent is smaller than 1/3, evaporation rate Fig. 4. Sherwood-Rayleigh relation for negative Rayleigh numbers.
decreases with pool size increasing. Most of existing
correlations suggest that this decrease is slow. However, in V. WATER SURFACE COOLING DUE TO EVAPORATION
1882 Stefan derived that for circular and elliptical pans
evaporation rate is inversely proportional to pan radius, which
corresponds to n = 0 . Thus, the question about dependence on
pool size seems to be still open.

Fig. 5. Average temperature difference between air and water


surface as function of relative humidity far from the pool for
T=300, 293 and 283 K. L=0.5 m. Linear approximations are
shown with solid lines.

Fig. 3. Sherwood-Rayleigh relation for positive Rayleigh As mentioned above, using boundary condition (3) and
numbers: simulation results and known correlations. assuming vapor saturation at the water surface enables one to
calculate temperature difference between air and water in
IV. FIXED WATER TEMPERATURE: NEGATIVE RAYLEIGH quasi-stationary state. Water is gradually cooled until heat loss
NUMBERS due to evaporation is balanced by gain from conductive
Cold water situation when concentration-induced exchange with air, pool walls and from radiation. Note that
convection is suppressed by thermally stable stratification is radiation and heat gain by water through the walls are
less popular. Correlation by Sparrow et al. [13] is the only one described in simplified form using efficient heat transfer
known for this case. Nevertheless, this situation is common coefficient = 13.7 W/(m2K) as evaluated from experiments
e.g. in oceanography in upwelling phenomenon. No stochastic [16]. The final state is not stationary because convection
plume generation occurs in this case, the final state is laminar proceeds in chaotic bursts as in the case A for hot water. If a

ISBN: 978-1-61804-020-6 123


Recent Researches in Mechanics

convective plume emerges somewhere at the water surface, [3] W. H. Carrier, The temperature of evaporation, ASHVE Trans., vol.
24, pp. 25-50, 1918.
humidity gradient is decreased and evaporation heat loss [4] G. W. Himus, J. W. Hinchly, The effect of a current of air on the rate
diminishes. This leads to local water surface temperature of evaporation of water below the boiling point, Chemistry and
increase and correspondent decrease of temperature difference Industry, vol. 22, pp. 840845, August 1924.
with air. Heat gain from conductivity and radiation drops, [5] M.M. Shah, Prediction of evaporation from occupied indoor swimming
pools, Energy & Buildings, vol. 35, pp. 707-713, August 2003.
which results in water temperature decrease. These oscillations [6] ASHRAE Handbook HVAC Applications. Atlanta: American Society
take place all over the air-water interface, causing stochastic of Heating, Refrigerating and Air-Conditioning Engineers, Inc, 2003.
plume generation. Average water temperature is higher near [7] M. Al-Shammiri, Evaporation rate as a function of water salinity,
Desalination, vo1. 150, pp. 189-203, November 2002.
the pool center, which confirms Stefans result that [8] S.M. Bower and J.R. Saylor, A study of the Sherwood-Rayleigh
evaporation is more intense near the edge. relation for water undergoing natural convection-driven evaporation,
Average temperature difference between air and water is Int. J. Heat Mass Transfer, vol. 52, pp. 3055-3063, June 2009.
[9] S.J.K Bukhari and M.H.K. Siddiqui, Characteristics of air and water
shown in Fig. 5 as function of ambient air relative humidity for velocity fields during natural convection, Heat Mass Transfer, vol. 43,
different ambient temperatures. The dependence show linear pp. 415-425, March 2007.
behavior in all three cases. The values show good agreement [10] Z. Li and P. Heiselberg, CFD Simulations for water evaporation and
airflow movement in swimming baths indoor environmental
with experiment [16]. Investigation of terms in (3)
engineering, Aalborg University, Denmark, Report for the project
demonstrates that most important are the second and the third Optimization of ventilation system in swimming bath, April 2005.
terms i.e. evaporation and external heat from radiation and [11] L.M.K. Boelter, H.S. Gordon, B.F. Sharpley, Free evaporation into air
through the walls. Conductive heat flux and heat fluxes along of water from a free horizontal quiet surface, Ind. Eng. Chem., vol. 38,
pp. 596-600, 1946.
the surface are about 10 times less in quasi-stationary state. [12] B.F. Sharpley and L.M.K. Boelter, Evaporation of water into quiet air
from a one-foot diameter surface, Ind. Eng. Chem., vol. 30, pp. 1125-
VI. CONCLUSIONS 1131, 1938.
[13] E.M. Sparrow, G.K. Kratz, M.J. Schuerger, Evaporation of water from
Boundary conditions have been proposed which enable one a horizontal surface by natural convection, J. Heat Transfer, vol. 105,
to calculate evaporation rate directly, performing simulations pp. 469-475, August 1983.
[14] R.J. Goldstein, E.M. Sparrow, D.C. Jones, Natural convection mass
in either fixed water temperature formulation or taking into transfer adjacent to horizontal plates, Int. J. Heat Mass Transfer, vol.
account water cooling due to evaporation. Computations have 16, pp. 1025-1035, May 1973.
yielded universal curves for Sherwood-Rayleigh relation in [15] J.R. Lloyd and W.R. Moran, Natural convection adjacent to horizontal
both cases of negative and positive Rayleigh numbers. surface of various planforms, J. Heat Transfer, vol. 96, pp. 443-447,
November 1974.
However, these curves are characterized with lower power-law [16] Y.Y. Plaksina, V.N. Aksenov, E.G. Andreev, Study of cool and warm
exponent than the existing empirical correlations. Next step is surface skin formation in laboratory conditions, Moscow Univ. Phys.
to incorporate turbulence modeling and to explore turbulence Bulletin, vol. 64, pp. 355-357, June 2009.
[17] C. Cercignani, Rarefied gas dynamics. Cambridge University Press,
influence on Sherwood-Rayleigh relation at high Rayleigh 2000, pp. 273-285.
numbers. Another obvious extension is performing 3D- [18] A.I. Osipov, A.V. Uvarov, N.A. Vinnichenko, Influence of the initial
simulations. Also, volume condensation model should be nonequilibrium state of a medium on the structure of von Karman
vortex street, Phys. Fluids, vol. 18, 105106, October 2006.
included to extend the range of situations which can be
modeled. So far, water-side flow was not taken into account.
Its velocity is much lower than the velocity of air-side flow,
however it is important for exploration of temperature
gradients in water and more accurate modeling of water
heating from the pool walls. Hence, more detailed simulations
should include water consideration and v x / y = 0 instead of
v x = 0 at air-water interface. One more factor which can
affect evaporation rate is horizontal wind in the vicinity of
water surface. As in the case of forced convection heat
exchange, even weak wind can enhance humid air removal
from the surface, leading to evaporation rate increase. All
these factors are to be modeled and compared directly to the
experiment in the near future.

REFERENCES
[1] H. Hertz, Ueber die verdunstung der flssigkeiten, insbesondere des
quecksilbers, im luftleeren raume, (in German), Ann. Phys., vol. 17,
pp. 177-197, August 1882.
[2] M. Knudsen, Die maximale verdampfungsgeschwindigkeit des
quecksilbers, (in German), Ann. NatPhys., vol. 47, pp. 697-708,
August 1915.

ISBN: 978-1-61804-020-6 124

Das könnte Ihnen auch gefallen