Professional Documents
Culture Documents
Esgesgs
Esgesgs
To cite this article: N D Fowkes & A P Bassom (2009) Piped Water Cooling of Concrete:
An Exercise in Scaling, Australasian Journal of Engineering Education, 15:2, 51-58, DOI:
10.1080/22054952.2009.11464026
Article views: 3
problem, but few understood the need to combine the temperatures of the circulating water; and rw and
two, and the need for intuition as a driving force. This cw are the density and specific heat of water. This
of course requires experience, and early experience provides us with an estimate for the Qv required to
is invaluable. maintain an acceptable DT; we get Qv ≈ 1.4 × 10−6 s−1,
which is again in accord with practice.
Another unfortunate fact of life is that too often
students immediately and mindlessly resort to It is noted that without any sophisticated mathematics
numerics due to the ready availability of powerful we have hopefully identified the major features of the
programs. Also many students seek “complexity” in problem. One might call this crude engineering (and
a problem, rather than an understanding of it. Thus perhaps there ought to be more of it), but we need
simple models/tests are dispensed with and there is to further check out and refine the results. In what
often no reality check; very accurate results may be follows we must keep the above observations in mind;
obtained to the wrong problem. While for standard the scalings we use must accord with these simple
engineering problems, reality tests may not be predictions, and the mathematics must be tailored
necessary; for new problems (where understanding to highlight the observations or identify deficiencies.
the physics is essential) such simple models are a Hopefully the results extracted will be consistent with
necessary beginning. Our present concrete cooling the above observations, if not then either there is some
problem is one such example, and rather than unexpected subtlety that has not been appreciated or
leap immediately to equations, we begin with a a very embarrassing misjudgement.
consideration of the important issues. It is to be expected that determining solutions
corresponding to all possible winding paths of
2 The crux of the problem pipes through a slab is probably both impossible
and pointless, so a simplified geometric model is
The conductivity of concrete is relatively small so essential. “Keep it simple” is the catch-cry here. We
check
that very little of the internalout and refine
heat generated can bethe results.
thus examine a In what
model follows
consisting we must
of a single water keep the
expelled through the surface during thein construction bearing pipe of radius a encased in an insulated
observations mind; the scalings we use must accord with these simp
period. The time scale for conduction is of the order hydrating concrete sleeve of radius R, see figure 1.
of L2/kc, where kc isdictions, and
the diffusitivity the mathematics
of concrete and One may thinkmustof be tailored
this as modellingto highlight
a section of the the observ
L is a characteristicor length scale. With pipe network embedded in the concrete; the
L = 10 m, a Hopefully the results extracted will be consisten radius
identify deficiencies.
typical slab size, and kc = 6.6 × 10−7 m2/s, this suggests of the concrete sleeve corresponds to approximately
a conduction time the above
scale of years; clearly half
observations;
about 4.7 the separation
if not distance
then either between
there isneighbouring
some unexpected su
increased surface cooling is not going to remove the pipes (where heat flux levels almost vanish). This
which has not been appreciated or a very embarrassing misjudgment.
excess heat in a reasonable timeframe, and hence model has the benefit of removing complications
the need for piped water. A key question is: what
spacing S between neighbouring pipes is required
to remove the hydration heat? Over a time scale z=L
heat conducts a distance of order c , which, with
t equal to two weeks (the typical time span between
pourings), gives S ≈ 0.89 m as an effective length
scale for heat removal. This compares well with the
recommended spacing of about 40 cm, see design
code ACI 207.4R‑05 (American Concrete Institute, ∂Tc
2009). Such codes are developed empirically and ∂r
=0
suffer the weakness of not being able to account for
differing construction conditions and the differing
thermal characteristics of the concrete used. In the
present context, however, it is encouraging that our
very rough calculations give values that are accord
with those used in practice.
The next concern is how much water should be
z=0 a R
pumped through the pipes to remove the hydration
heat? If q denotes the heat production rate per
unit volume of concrete, then under steady-state
conditions all the heat generated by hydration must Water flux Q
be removed by the water so that:
associated with dealing with the water network h(z, t) = gw(Tc(a, z, t) − Tw(z, t)) (5)
geometry, but is deficient for the same reason;
such a model cannot address the network issues. where the heat transfer parameter
Nevertheless one might hope to obtain useful a
w 2k p ' (6)
estimates for the appropriate pipe length and s
spacing between pipes within the network.
is composed of two parts. The first term represents
The concrete produces hydration heat at a prescribed
the conduction of the pipe; here s is the pipe thickness
rate q per unit volume (assumed constant), while cool
and kp its conductivity, see Carslaw & Jaeger (1959).
water at temperature of T0 enters the pipe at z = 0 and
The second term g ‘ is a turbulent heat exchange
removes hydration heat from the slab. The volume
term, which is dependent on the flux through the
flux of water through the pipe is Q. For any choice
pipe; experimental values can be obtained from
of parameters this situation is physically realisable
engineering handbooks, for example Rohsenow et al,
and we ought to be able to determine the resultant
(1998). Equation (5), together with the requirement
temperature change in the concrete brought about by
that heat flux across the external pipe surface be
the piped water over any prescribed length L. Using
continuous, gives:
such results we should be able to derive the optimal
heat transfer arrangement. Tc a, z , t
kc w Tc a, z , t Tw z , t (7)
r
3 The governing equations
The heat flux into the water causes an increase in
Heat conservation within the concrete requires that temperature along the pipe described by:
the temperature distribution Tc(r, z, t) satisfies the Tw z , t T z , t
heat equation: w cw a2 w cw Q w
t t (8)
T
c cc c kc 2Tc q in a r R, 0 z L (2) 2a w Tc a, z , t Tw z , t
t
where Q is the volume flux of water through the
and for t > 0, where: pipe, and rw and cw are the density and specific heat
2Tc r , z , t 1 Tc r , z , t 2Tc r , z , t of water. This conservation result is obtained by
2Tc equating the change in heat content in an element of
r 2 r r z 2
water-filled pipe to the net heat transfer rate into the
expressed in cylindrical geometry; and where rc, kc and element due to conduction across the interior of the
cc denote the density, thermal conductivity and specific pipe (as detailed above), and advection into and out
heat of the concrete, respectively. The cylindrical of the element. The radial heat transfer rate term in
concrete sleeve is insulated at r = R, which requires: this equation is simplified using equations (5) and (7).
if the enhanced set is solvable), but experience equation (11) with the eventual temperature rise
suggests otherwise. At best such unnecessary terms in the concrete determined by a balance between
obscure the physics, but often their inclusion yields hydration heat input and radial diffusive transport.
a system that is not well behaved and thus not By equating to unity the associated square bracketed
easily solved. Thus, while incorrect scaling is not terms in equation (11) we can ensure the balance
necessarily disastrous, a correct choice can greatly described above is correctly represented in the scaled
facilitate the analysis and focus the investigation. equations and any additional terms are correctly
In any problem there are some natural scales that are assessed. The scales for T and tc are thus determined,
imposed by the problem formulation (here R and T0) see below.
and others that are unprescribed, which need to be So what determines z0? Under steady-state conditions,
deduced from the information contained in the field temperature variations along the pipe will be
equations. We introduce the scaling: determined by a balance between the radial heat
flux into the water and the net advective transport
r = Rr′, z = z0z′, t = tct′,
(10) along the pipe, so that the scale z0 should be chosen
Tc = T0 + TT′c, Tw = T0 + TT′w
from the water heat transport equation (12) so that
where the scales z0, T and tc remain to be fixed. In this balance is explicitly exhibited. This is achieved
terms of the new (primed) variables, the conservation by putting [(2pagwz0)/((rwcwQ)] = 1. We are thus led
equations (2) and (8) become: to the choices:
Table 1: Typical parameter values. equation is second order in r with two flux boundary
conditions (at r = e and r = 1), while the Tw equation
rc 2350 kg/m3 is first order in z with the one boundary condition
rw 1000 kg/m3 at the entry point z = 0. The system evolves with
time and the reduced equations therefore appear
kc 1.37 W/m °C
to be mathematically well-defined. It is remarked
q 300 W/m3 that the longitudinal conduction terms disappears
Q 2 × 10–4 m3/s from the concrete equation, indicating that the
longitudinal temperature variation is much smaller
kw 0.59 kg/m3 (of relative order e) than the radial variation. Given
cc 880 J/kg °C the cylindrical geometry and rapid heat transfer
along the pipe one might expect this result, which
cw 4200 J/kg °C
also is consistent with the long thin geometry used
R 0.5 m in practice.
gw 500 W/m2 °C The efficiency of the heat transfer into the pipe is
a 0.025 m essentially set by the value of x, and since this is the
lynchpin of the problem, it is not surprising that the
z0 ∼ 10.7 m, tc ∼ 3.7 × 105 s ∼ 4.4 days,1 and T ∼ 54.7 K, parameter x is of unit order or even larger. We remark
which suggests a typical temperature rise of about that the value 182 is perhaps greater than might
55 K in the concrete and typical pipe lengths of about have been expected; it turns out this occurs due to a
10 m. The values are consistent with engineering logarithmic solution behaviour close to r = 0.
practice. The dimensionless length scale parameters, It is possible, but somewhat tedious, to solve our
e ≈ 0.05 and d ≈ 0.046, are both small, as was supposed, simplified system using Laplace transforms, but
while the dimensionless water speed parameter it is much easier to find solutions appropriate to
u ≈ 3.6 × 103 is large, so that the advective transport the steady-state limit. One might excuse this short
term in the pipe water equation (15) dominates – we cut by noting that for the proposed network to be
have consistency so far. The scaled dimensionless practicable it must extract most of the released heat
heat transfer coefficient x ≈ 182. of hydration. Thus, while steady-state might not be
If we ignore the terms that appear to be sub- an achievable ideal, one might expect the associated
dominant, what remains (see equations (14) to (17)) is steady-state solution to be close enough to the
the simplified set consisting of the concrete equations: actual solution so that design information can be
inferred. Under steady-state conditions, the concrete
Tc r , z , t 2Tc r , z , t 1 Tc r , z , t
1, r 1 equation (19) reduces to a second-order ordinary
t r 2 r r differential equation in r, and after imposing the
(19)
e < r < 1 zero flux condition at r = 1 (see equation (20)), one
is left with an arbitrary function of z that needs to
and t > 0, subject to: determined so that the results match the piped water
results. The piped water equation (21) is a first-order
Tc 1, z , t
0 (20) linear ordinary differential equation in z, and after
r imposing the entry condition, one ends up with a
linear temperature variation along the pipe. The flux
together with the piped water equation:
continuity condition (22) can now be imposed to give:
Tw z , t
Tc , z , t Tw z , t , z 0 (21) Tc r , z Tw z
z
1 (23)
for t > 0 with Tw(0, t) = 0. These equations are coupled
2
1 2 14 2 r 2 12 ln r , r
via the flux continuity boundary condition (across
z
r = e) given by: Tw z
2
1 2 (24)
Tc , z , t
Tc , z , t Tw z , t (22) Note that the temperature profile in the concrete is
r
logarithmic in r close to the pipe, thus exhibiting a
Additionally we need to require that Tc and Tw “sink-like” behaviour due to the heat removal by
vanish at t = 0. Note that the two field equations the water, and quadratic in form away from the pipe
in two unknowns Tc(r, z, t) and Tw(z, t) are linked due to the hydration heat input. Thus temperature
via a Newtonian-type flux condition (22). The Tc levels increase radially away from the pipe with
the maximum level reached at the insulated surface
1
Also the time scale for water heating is τw ∼ 105 s ≈ 2 minutes, (r = 1) of the sleeve at the location most remote from
which is negligible compared with the concrete cooling time. the water entry point (z = L/z0).
As a general rule, students (especially mathematics will need to balance costs associated with network
students) tend to think that an investigation is construction with the expenses and risks associated
completed once a mathematical result is exhibited; a with thermally induced structural problems.
product of the question/answer format often adopted
in courses. In an industrial mathematics context, the
specific quantitative results obtained are generally 5 Conclusion
much less important than qualitative information
Our primary objective here has been to illustrate
contained in the formulae, and so interpretation is
by example a standard approach taken by applied
essential. In part this is due to the fact that solutions
mathematicians and engineers when addressing
are (necessarily) obtained for a problem that is only
continuum industrial modelling, and to indicate the
remotely related to the real problem; in our case the
central role of scaling in this process. In the available
geometry is greatly simplified. More importantly,
week, the PIMS graduate students (with little
however, engineers are normally more interested
assistance) managed to extract the scaled equations;
in knowing how best to tweak the system. For
itself a notable achievement in the short time, but
example, in the present context engineers would like
more importantly they learnt much about real world
to know about the relative effectiveness of various
modelling. This problem, which can be extended in
procedures for reducing the maximum temperatures
many ways, could either be presented in lecture form
in the concrete. If we (partially) convert back into the
to advanced students as an example of real world
original variables we have:
modelling or set as a project.
qR 2 r z t
Tc T0 T'c , ,
kc R z0 c References
qR 2
z t
Tw T0 T'w , American Concrete Institute, 2009, 207.4R-05:
kc z0 c Cooling and Insulating Systems for Mass Concrete, ACI
Committee 207.
with z 0 and t c as in equation (13), and where
T′c → Tc and T′w → Tw over a time scale of about Carslaw, H. S. & Jaeger, J. C. 1959, Conduction of heat
tc ∼ 4.4 days for the spacings of interest. Notice that in solids, Clarendon Press, Oxford.
the temperature levels in the concrete and water pipe
increases in proportion to R2, so that the choice of Charpin, J. P. F., Myers, T. G., Fitt, A. D. & Fowkes,
spacing between pipes is critical. If there is some pre- N. D. 2004, “Piped water cooling of concrete dams”,
set maximum temperature rise within the concrete Proc. Mathematics in Industry Study Group South Africa
that is to be allowed, the corresponding spacing can 2004, Mason, D. P. & Fowkes, N. D. (editors), pp. 69-
be derived using equation (23). The length scale 87, ISBN 0-620-33850-4.
for significant temperature changes along the pipe
and in the concrete is z0, which can be decreased by Fowkes, N. D. & Broadbridge, P. 2008, “A graduate
increasing the water flux (normally by using larger modelling workshop: Canadian style”, Aust M.S.
pipes). Evidently if the pipe length L is much less than Gazette, Vol. 35, No. 5, pp. 315-319.
z0, then the network is over-engineered (the network
will remove the heat, but the pipe length is greater Myers, T. G., Fowkes, N. D. & Ballim, Y. 2009,
than it need be), but if L is much greater than z0, then “Modelling the cooling of concrete by piped water”,
unacceptably large high concrete temperatures will ASCE Journal of Engineering Mechanics, posted March.
occur. Here L refers to the length of an individual
pipe winding through the concrete. Rohsenow, W., Hartnett, J. & Cho, Y. 1998, Handbook
of Heat Transfer, 3rd edition, McGraw-Hill Handbooks.
Our analysis here is very much a first look at the
problem and is by no means complete (for more Springenschmid, R. (editor), 1994, Thermal cracking in
information see Myers et al, 2005), but is enough concrete at early ages, E&FN Spon, London
to clarify the major issues. The network design
problem has not been addressed at all (and is an US Bureau of Reclamation, (n. d.), “Hoover Dam:
interesting problem that can be solved using graph Concrete”, www.usbr.gov/lc/ hooverdam/History/
theory), and we must note that an optimal network essays/concrete.html.
NEVILLE FOWKES
ANDREW BASSOM
Andrew Bassom started his PhD in boundary layer theory at the University of
Exeter (UK) in 1985. He stayed put for 20 years; first as a postdoc, then a lecturer
and finally a reader. In 2005 he moved to Perth to take up a chair in applied
mathematics at the University of Western Australia. Research interests include
boundary-layer theory, fluid and solid mechanics, and differential equations.
Outside work he is a keen scuba diver and dinghy sailor.