Sunteți pe pagina 1din 10

Rameys Wellbore Heat

Transmission Revisited
Jacques Hagoort, Hagoort & Assocs. B.V.

Summary
In this work we assess Rameys classic method for the calculation
of temperatures in injection and production wells. We show that
Rameys method is an excellent approximation, except for an early
transient period in which the calculated temperatures are significantly overestimated. We present a simple graphical correlation to
estimate the length of this early transient period. We further demonstrate that Rameys procedure for the estimation of overall heattransfer coefficients holds good only for large values of the Fourier
dimensionless time. Finally, we illustrate the results of this work
by an example calculation of wellhead temperature in a flowing
oil well.
Introduction
Estimation of temperatures in a wellbore during injection or production is a recurring problem in petroleum engineering. Examples
are the prediction of bottomhole temperatures of injection fluids
(water, gas, and steam) and of wellhead temperatures in oil and gas
wells. Almost all practical methods for the calculation of temperatures in wellbores go back to the classic paper by Ramey1 on
wellbore heat transmission published in the early 1960s. In that
paper, Ramey presented a simple analytical equation for wellbore
temperatures based on a grossly simplified heat balance. Apart
from this analytical temperature equation, Ramey also proposed a
simple procedure to estimate an overall heat-transfer coefficient
for wellbore heat losses comprising both transient heat resistance
in the formation and near-wellbore heat resistance.2 Despite its
simplifications, in practice Rameys approach seems to work remarkably well.
In 1990, Wu and Pruess3 presented an analytical solution for
wellbore heat transmission in a layered formation with different
thermal properties without introducing the simplifying assumptions of Ramey. From their example calculations, they observed
that the Ramey method is valid at long times but can generate large
errors at early times. However, quantification of the conditions
under which Rameys method could be applied was beyond the
scope of their paper.
The objective of this work is to assess the Ramey method and
to establish criteria for its applicability. To this end we have first
performed an inspection analysis of the basic wellbore heat transmission equations without the simplifying Ramey assumptions.
This entails formulation of the equations in dimensionless form
and identification of the governing dimensionless numbers. Subsequently, we have developed a rigorous solution of these dimensionless equations and have compared this solution with the
Ramey solution for various ranges of the dimensionless numbers.
We have also checked Rameys procedure for estimating overall
heat-transfer coefficients. To illustrate the results of the paper, we
conclude with an example calculation of the estimation of wellhead temperature in a flowing oil well.
Our focus is primarily on heat transmission between the wellbore tubing and the formation. We have therefore kept the heat
transmission within the tubing as simple as possible by assuming
a constant fluid density and heat capacity, no heat conduction in
the flow direction, and no frictional heating. In addition, we have

Copyright 2004 Society of Petroleum Engineers


Original SPE manuscript received for review 9 September 2003. Revised manuscript received 22 June 2004. Paper SPE 87305 peer approved 22 August 2004.

December 2004 SPE Journal

restricted ourselves to the temperature distribution in production


wells. Extension of the analysis to injection wells is straightforward.
Equations
The physical model that underlies the equations describing wellbore heat transmission consists of a straight, cased well that is
cemented to the formation and equipped with tubing for the transfer of fluids to the surface. Both casing and tubing have a constant
diameter. The diameter of the tubing is small with respect to the
length of the tubing. Initially, the tubing is filled with a fluid in
thermal equilibrium with the formation. At time zero, fluid starts
flowing from the bottom of the tubing to the top at a constant flow
rate. The fluid has a constant density and a constant heat capacity.
Heat conduction in the flow direction and frictional heating in the
tubing are negligibly small. The fluid that is initially present in the
tubing is the same as the fluid that enters the bottom of the tubing.
The temperature of the fluid at the bottom is equal to the formation
temperature at the bottom. Flow in the tubing is 1D (i.e., temperature and fluid velocity depend only on the distance along the
tubing). As the fluid moves up the tubing, it loses heat to the colder
formation. Heat losses to the formation take place through heat
conduction in the radial direction only. The effect of the tubing
wall, the annular space between casing and tubing, the casing wall,
and the cement zone on the heat transmission is included through
a single, steady-state heat-transfer coefficient. The formation surrounding the wellbore is of infinite extent. The initial temperature
of the formation increases linearly with depth, reflecting a constant
geothermal gradient.
The temperature distribution in the tubing for this physical
model is described by two heat balances: one for the tubing and
one for the formation surrounding the wellbore (see Appendix A).
In dimensionless form, the heat balance for the tubing reads:
TD TD
+
=
tD xD
tD

NRA

TD
quDNGztD ;UD d. . . . . . . . . . . . . . . . . . . (1)

For the associated initial and boundary condition, we have, respectively:


tD = 0: TD = 1 xD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
and
tD 0, xD = 0: TD = 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)
The dimensionless distance runs from zero at the bottom of the
tubing to unity at the wellhead. The dimensionless temperature
ranges from zero, the initial formation temperature at the wellhead,
to unity, the temperature at the bottom of the wellbore. The dimensionless time is based on the convection term; a dimensionless
time of unity corresponds to the residence time of a fluid particle
in the tubing when flowing from bottom to top.
The term at the right side of Eq. 1 represents heat losses of the
tubing fluid to the surrounding formation. It is essentially the
superposition of heat losses for a temperature in the tubing that is
varying with time using a unit heat-loss function. This unit heatloss function describes heat losses from the tubing to the formation
for a constant unit temperature drop between tubing and formation.
The unit heat-loss function can be derived from the solution of
the heat balance for the formation, which for the physical model
465

described reduces to the radial Fourier heat conduction equation.


In dimensionless form this equation reads (see Appendix A):

TfD
TfD
1
rD
=
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
rD rD
rD
tDFo
The initial condition is
tDFo = 0, rD 1:

TfD = 0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (5)

At the inner and outer boundary we have, respectively:


tDFo 0, rD = 1:

TfD
rD

= UD1 TfD . . . . . . . . . . . . . . . . . . . . (6)

and
tDFo 0, rD = :

Tf D = 0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7)

The dimensionless time in the Fourier equation is determined by


the thermal diffusivity of the formation and thus is different from
the dimensionless time in the tubing heat balance. To avoid any
confusion, we will refer to the dimensionless time in the Fourier
equation as Fourier time. The dimensionless radial distance is
measured from the center of the tubing and runs from unity to
infinity. Unity corresponds to the inner boundary of the formation,
the boundary between formation and cement zone. The dimensionless temperature in the Fourier equation is defined as a fraction of
the difference between the tubing temperature and the initial formation temperature. It ranges from zero, the initial formation temperature, to unity, the temperature in the tubing.
Eq. 6 is known as the radiation boundary condition4 and makes
allowance for the near-wellbore thermal resistances through the
dimensionless near-wellbore heat transfer coefficient UD. These
resistances are positioned between the inner boundary of the formation and the center of the tubing. For infinite UD, the temperature drop across the overall thermal resistance is zero and the
radiation boundary condition reduces to a constant temperature
boundary condition.
Solution of the Fourier equation subject to the initial and
boundary equations yields the temperature distribution in the formation as a function of time. Given this temperature distribution,
the unit heat-loss function then follows directly from
quDtDFo;UD =


TfD
rD

rD=1

. . . . . . . . . . . . . . . . . . . . . . . . . . . . (8)

= UD1 TfDrD = 1
The unit heat-loss function is a function of Fourier time and includes the dimensionless near-wellbore heat-transfer coefficient as
a parameter.
Inspection of the previous equations shows that the temperature
distribution in the tubing as a function of time is governed by three
dimensionless groups: NGz, NRa, and UD. The number NGz is
known as the Graetz number and arises from the two different time
scales that are at work in this problem: the Fourier time scale for
heat conduction in the formation and the time scale for convective
heat transport in the tubing. It is defined as the ratio of the dimensionless Fourier conduction time and the dimensionless time for
convection in the tubing:
NGz =

tDFo L
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
=
tD rcf2 u

The number NRa expresses the ratio of the heat lost by conduction
in the formation to the heat transported up the tubing by convection and is defined by
2f L
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10)
NRa =
cw
For small values of NRa (high production rates), the heat transport
is dominated by convection and little heat will be lost to the formation. By the same token, large values of NRa point to significant
cooling of the fluid in the tubing. In honor of Ramey, we will call
the number NRa the Ramey number.
466

The dimensionless near-wellbore heat transfer coefficient UD is


defined by
UD =

Urcf
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11)
f

It represents the ratio of the conductive heat resistance of the


formation to the near-wellbore thermal resistance. Very large values of UD mean little thermal resistance near the borehole and thus
heat losses that are dominated by heat conduction in the formation.
Likewise, small values indicate heat losses dominated by nearwellbore thermal resistance. The reciprocal of UD may be called
the thermal skin factor, analogous to the familiar skin factor of
oil and gas wells, which quantifies near-wellbore resistance to
fluid flow.
Ramey Solution
Rameys solution for the temperature distribution is based on the
following two simplifying assumptions in the tubing heat balance
Eq. 1: (i) the accumulation term is negligibly small, and (ii) the
convoluted heat-loss term can be replaced by the product of the
unit heat-loss function and temperature. The first assumption definitely does not hold during the early stages when the hot fluid
entering the tubing at the bottom is displacing the initial tubing
fluid. In this period, the temperature profile in the tubing is
coupled to the displacement front, the movement of which is controlled by the accumulation term. However, as the change in tubing
temperature with time becomes increasingly smaller, the accumulation term may become negligibly small at some point during the
later stages of the process. The second assumption means that the
effect of temperature history on heat losses can be ignored. This is
only true for a temperature-step function or for a time-independent
heat-loss function. These conditions may be approached also after
some time when both heat losses and temperature become slowly
varying functions of time.
Introducing the Ramey assumptions into the tubing heat balance Eq. 1, we obtain what we may call the Ramey heat transmission equation:
TD
= NRaTD 1 + xDquDtDFo. . . . . . . . . . . . . . . . . . . . . . (12)
xD
Eq. 12 describes the temperature profile in the tubing at time tDFo.
In this equation, the unit heat-loss function appears as a (timedependent) constant. Direct integration of Eq. 12 subject to the
boundary condition (Eq. 3) yields
TD = 1 xD +

1 expNRaquDxD
. . . . . . . . . . . . . . . . . . . . . . (13)
NRaquD

From Eq. 13 we see that the temperature profile depends on a


single parameter NRaquD, the product of the Ramey number and
the unit heat-loss function. The dimensionless number UD is implicitly included in the unit heat-loss function. The temperature
profile depends on time through the time dependence of the unit
heat-loss function. The Graetz number does not occur in Eq. 13
because the assumption of a negligible accumulation term makes
the convection time scale immaterial.
As we shall see in the next section, the temperature profile (Eq.
13) for small values of NRaquD is of special interest. Series expansion of the exponent yields
TD = 1

NRaquD 2 NRaquD2 3
xD +
xD . . . . . . . . . . . . . . . . . . . . (14)
2!
3!

A second-order approximation of Eq. 14 predicts the temperatures


within 1% for values of NRaquD smaller than 0.24. For these small
values, the temperature profile takes a quadratic shape and all the
temperatures are in excess of 89% of the inlet temperature.
Application of the Ramey solution requires as input the Ramey
number and the unit heat-loss function for a given value of the
thermal skin. The Ramey number follows directly from the formation heat conductivity, tubing length, fluid properties, and well
rate for the case in question. Evaluation of the unit heat-loss funcDecember 2004 SPE Journal

Fig. 2Effect of NRa on Ramey wellhead temperature.


Fig. 1Ramey temperature profile at successive times.

tion calls for the solution of the radial Fourier equation, subject to
the radiation boundary condition. We have solved this equation by
means of Laplace transformation in combination with Stehfests
numerical Laplace inverter.5 (See Appendix B.) Rameys approximation of the unit heat-loss function will be discussed in a
later section.
Fig. 1 illustrates the general features of the Ramey solution.
Here we have plotted the temperature distribution in the tubing for
NRa1 and 1/UD0 (zero thermal skin) at three successive Fourier times: 1, 10, and 100. The profile is concave on the left,
reflecting the increase in heat losses at shallower depths. Because
of continuous heating of the formation by the hot flowing fluid in
the tubing, heat losses become progressively smaller with time.
Therefore, the temperature profile moves to the right, closer to the
inlet temperature of unity.
Fig. 2 shows the effect of the Ramey number on the time
profile of the wellhead temperature. It displays the profile for
NRa0.25, 1, and 4 and 1/UD0. As expected, the larger the
value of NRa, the stronger are the heat losses and thus the lower is
the wellhead temperature. We further see that the wellhead temperature increases gradually, a consequence of the decreasing heat
losses. After an infinite time, the wellhead temperature in all cases
must be equal to the inlet temperature of unity.
Fig. 3 illustrates the effect of skin factor on wellhead temperature. Here we have plotted wellhead temperature as a function of
time for NRa1 and for 1/UD0, 1, and 4. From Fig. 3 we see that
the larger the skin factor, the lower the heat losses are and thus the
closer the wellhead temperature is to the inlet temperature. In
addition, we observe that the time profile of the wellhead temperature is flatter for larger skins. This is a direct consequence of the
flatter time profile of the unit heat-loss function for larger skins. In
the extreme case of heat losses that are completely dominated by

Fig. 3Effect of skin (1/UD) on Ramey wellhead temperature.


December 2004 SPE Journal

near-wellbore heat resistance, the unit heat-loss function is constant and so is the wellhead temperature.
Rigorous Solution
We have rigorously solved the tubing heat balance (Eq. 1) and the
associated Fourier equation (Eq. 4) by using Laplace transformation. (See Appendix C.) In Laplace space, both equations can be
solved analytically and can be combined into a single analytical
equation for the Laplace-transformed temperature as a function of
distance. We have inverted this analytical solution to time by
means of the numerical inversion algorithm of Stehfest.5 Before
discussing the general character of the rigorous solution, we first
present its behavior at early and late times, and for large thermal
skins. Under these conditions, the Laplace-transformed temperature can be approximated to a form that can be analytically converted to real time. These solutions provide helpful insight into the
temperature distribution under limiting conditions. They also serve
as benchmarks for the Stehfest numerical-inversion algorithm.
The early-time solution for the temperature ahead of the fluid
front (xD>tD) is given by (see Appendix C)
TD = 1 xD + tD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (15)
Early means a dimensionless time much smaller than unity and
thus long before the fluid front has reached the wellhead. Eq. 15
reflects the displacement of the initial fluid with the linear temperature profile from the tubing; the profile ahead of the front
retains its linear shape but moves up linearly proportional to time.
The late-time approximation of the temperature profile reads
(see Appendix C):
TD = 1

NRaquD 2
xD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (16)
2!

Eq. 16 is identical to the second-order approximation of the


Ramey solution (Eq. 14), which is valid for dimensionless temperatures above 0.9. This proves that at any rate the Ramey solution is an excellent approximation of the rigorous solution at
high temperatures.
In the extreme case that the thermal skin dominates the heat
transfer to the formation, the rigorous solution takes a particularly
simple form. (See Appendix D.) In that case, the unit heat-loss
function is equal to UD and thus is independent of time. As a
consequence, the profile reaches a steady state as soon as the fluid
front has reached the wellhead. This steady-state profile is the
same as the Ramey profile for the appropriate value of the constant
parameter NRaquD. Before this point, the profile consists of the
Ramey profile up to the front and a linear profile ahead of the
front. As the front moves to the wellhead the linear part moves up
and vanishes when the front reaches the wellhead. The plot of
wellhead temperature vs. time consists of two straight lines: one
line through the origin up to a Fourier time equal to NGz and the
steady-state wellhead temperature of TwhD(NRaquD), and the other
line parallel to the time axis representing the steady-state temperature. The approximate Ramey solution for this special case consists
467

Fig. 4Wellhead temperature vs. time for constant unit heatloss function.

Fig. 5Wellhead temperature vs. time for NRa=1, NGz=1, and


1/UD =0.

of a constant temperature all the way from time zero. Fig. 4 depicts
the analytical and the approximate Ramey solution for NGz0.5
and NRaquD0.8.
Fig. 5 illustrates the general characteristics of the rigorous solution vis--vis the Ramey solution. Here we have plotted the
rigorously calculated wellhead temperature vs. Fourier time for
NRa1, 1/UD0, and NGz1, along with the Ramey solution
(Eq. 13) and the early- and late-time approximations of the rigorous solution (Eqs. 15 and 16, respectively). The rigorous solution
consists of three distinct time regimes: an early transient period, a
late-time period, and an intermediate period in between. In the
early transient period, the Ramey method underestimates heat
losses and thus overestimates the wellhead temperature. This we
have already observed in Fig. 4. The arrival of the front at the
wellhead (tDFoNGz1) is not attended by a clear kink in the
wellhead temperature as we saw in Fig. 4; however, it remains a
characteristic point. The late-time regime is the period with temperatures in excess of 0.9, in which there is always excellent agreement between the Ramey and the rigorous solution. In this example, the late-time regime begins at a Fourier time of approximately 1,500. The point at which the Ramey and the rigorous
solution converge marks the beginning of the intermediate-time
regime. This occurs, for the conditions of this case, at approximately a Fourier time of 10, one log cycle after the production of
one tubing volume.
Fig. 6 shows the effect of the Graetz number on wellhead
temperature. It depicts the wellhead temperature as a function of
time for NRa1, 1/UD0, and for NGz0.1, 1, 10 along with the
approximate Ramey solution, which is the same for all values of
NGz. As we can see, the length of the initial transient period de-

pends on NGz; the larger the NGz, the longer it takes for the two
solutions to converge and thus the longer is the early transient
period. This is in line with the observation in Fig. 5 that the
intermediate regime must begin after the arrival of the fluid front
at the wellhead, which occurs at tDFoNGz.
Fig. 7 shows how Ramey number affects wellhead temperature.
Here we have plotted wellhead temperature as a function of time
for NGz1, 1/UD0, and for NRa0.25, 1, 2, and 4, again compared with the approximate Ramey solution for each value of NRa.
Smaller Ramey numbers mean lower heat losses to the formation
and thus higher wellhead temperatures. We further observe that the
smaller the Ramey number, the shorter the early transient period.
In the case of the smallest Ramey number of 0.25, the intermediate-time regime is all but absent and the late-time regime directly
succeeds the early transient period.
Fig. 8 illustrates the effect of thermal skin factor on wellhead
temperature. It shows the time profile of the wellhead temperature
for NGz1, NRa1, and for 1/UD0, 1, and 4, along with the
approximate Ramey wellhead temperature for each value of 1/UD.
The trend is as expected: larger skins mean lower heat losses and
thus higher wellhead temperatures. The larger the thermal skin, the
shorter is the early transient period. For the largest skin of 4, the
early transient period is just above the minimum value of a Fourier
time of unity, and the intermediate-time regime is absent again.
This agrees with the behavior of the wellhead temperature for very
large skins as depicted in Fig. 4, where the early transient period
ends at tDFoNGz1.
In summary, comparison of the Ramey solution with the rigorous solution demonstrates that the Ramey solution is an excellent approximation, except at early times. In the early period, the
two Ramey assumptions are clearly not satisfied and cause significant errors in the calculated temperatures.

Fig. 6Effect of NGz on wellhead temperature.

Fig. 7Effect of NRa on wellhead temperature.

468

December 2004 SPE Journal

Fig. 8Effect of skin (1/UD) on wellhead temperature.

Early Transient Period


To apply the Ramey method, it is essential to know the duration of
the early transient period. The absolute minimum duration of this
period is tDFoNGz. This is the time required to displace the initial
fluid from the tubing. In all cases where the Ramey method predicts high wellhead temperatures at tDFoNGz (i.e., within 10% of
the inlet temperature), the early transient period closely approaches
this minimum. As we have seen above, in these cases there is
strong agreement (within 1%) between the rigorous solution and
the Ramey solution; therefore, the early time period must be close
to the absolute minimum. In practice, this situation may occur in
prolific, high-capacity oil and gas wells, where the product of the
Ramey number and unit heat-loss function is often less than 0.24.
Another example is provided by injection wells in thermal recovery projects that are specially designed to minimize heat losses.
Regardless of the temperature level, the minimum duration may
also be approached if the thermal skin factors are large (say, more
than 5), and the product of the Ramey number and the unit heatloss function is larger than 0.24. In these situations, heat losses are
considerable and most of the temperature drop occurs across the
near-wellbore thermal-resistance zones. As a consequence the unit
heat-loss function is insensitive to time and so is the wellhead
temperature. (See Fig. 4 for the limiting case of a constant unitheat-loss function.) In practice, these conditions may occur in oil
and gas wells that have been drilled through formations with a
large thermal conductivity (e.g., salt).
To estimate the duration of the initial transient period for arbitrary values of NGz, NRa, and UD, we have developed an empirical, graphical correlation based on the observation that, for a given
thermal skin, the initial transient period is shorter as the wellhead
temperature is closer to the inlet temperature of unity. (See Fig. 9.)

The vertical axis represents, on a log scale, the ratio of the length
of the initial period tDFoi to NGz, the arrival time of the fluid front.
The horizontal axis shows the wellhead temperature calculated by
the Ramey formula (Eq. 13) at a Fourier time of NGz. We have
determined the length of the initial period for three thermal skin
factors 1/UD0, 1, and 4, and for 0.1<NGz<10 and 0.25<NRa<4.
As a convergence criterion, we have used a difference between the
Ramey solution and the rigorous solution of 1%.
As we can see, the numerical data points in Fig. 9 cluster into
three groups, depending on the thermal skin. All clusters converge
to the ratio of unity at wellhead temperatures close to the inlet
temperature of unity. At these high temperatures, the duration of
the early transient period must be close to NGz. We further observe
that the larger the thermal skin, the shorter the initial transient
period. This stands to reason because in the limit of a very large
skin, the heat-transfer coefficient becomes independent of time,
and as a consequence, the transient period is equal to a Fourier
time of NGz. (See also Fig. 4).
Unit Heat-Loss Function
Application of the Ramey equation for the tubing temperatures
requires as input the unit heat-loss function at a certain time for a
given thermal skin factor. In the previous example calculations, we
have determined this function by rigorously solving the radial
Fourier heat-conduction equation. As a shortcut, Ramey1 proposed
to estimate the heat-loss function (or overall heat-transfer coefficient) by simply adding up the transient heat resistance for conduction in the formation for zero thermal skin and the total thermal
skin. In our dimensionless notation, the unit heat-loss function is
the reciprocal of the total dimensionless heat resistance. In formula
form, Rameys approximation thus reads
1
1
1
= 0
+
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . (17)
quDtDFo quD
tDFo UD
where the superscript zero denotes zero thermal skin. The total
thermal skin 1/UD is equal to the summation of the reciprocal of
the dimensionless heat-transfer coefficients of the individual resistance zones7:
1
=
UD

U
j=1

, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18)

jD

where n denotes the number of near-wellbore resistance zones.


When the resistance zone is made up of the solid wall of a circular
cylinder (e.g., tubing wall) of inner radius ri and outer radius ro,
the dimensionless thermal resistance is given by6
1 f lnro ri
=
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (19)
UjD
j
For the dimensionless thermal resistance at the inner wall of the
tubing or the annular space between tubing and casing, characterized by a steady-state heat-transfer coefficient hj, we can write6
1
f
=
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (20)
UjD rjhj

Fig. 9Correlation for duration of early transient period.


December 2004 SPE Journal

Strictly speaking, Eq. 17 holds true only for a constant heat-flux


boundary condition. In that case, the temperature drop across the
near-wellbore heat resistance is constant and may be added directly to the variable transient temperature drop. To check the
Ramey approximation (Eq. 17) for the radiation boundary conduction, we have compared the rigorously calculated unit heat-loss
function with the approximation based on Eq. 17 for 1/UD1 and
4. Fig. 10 shows the results. If Eq. 17 were perfectly true, the
rigorously calculated reciprocal heat-loss functions would run perfectly parallel to the zero-skin reciprocal heat-loss function with a
separation equal to the skin factor. Inspection of Fig. 10 shows that
the reciprocal heat-loss function for the nonzero skins indeed runs
nearly parallel to the one for the zero skin. The approximation is
better at larger Fourier times. For the skin of 1 and 4, the accuracy
of the calculated heat-loss rate is within 1% for Fourier times
larger than 200 and 400, respectively. Therefore, for large Fourier
469

tDFo 105 :

1
0
quD
tDFo

= 0.5ln

4tDFo
. . . . . . . . . . . . . . . . . . . . . . (24)

Eq. 22 follows from the early-time approximation of the rigorous


solution and includes a correction factor (0.79526) in the last term
to exactly match q0uD at tDFo1. The accuracy of Eq. 22 is within
0.1% of the rigorous solution. Eq. 23 is based on a best polynomial
fit of the rigorous solution in the pertaining time interval and is
accurate within 1%. Eq. 24 is the well-known long-time solution
presented by Carlslaw and Jaeger.4 Its accuracy is 1.2% at
tDFo105, and less than 0.8% for tDFo>106.
Example Calculation

Fig. 10Comparison of approximate and rigorous total heat


resistance (1/quD).

times, Rameys method for estimating the overall heat transfer


coefficient is a perfectly valid approximation.
The largest difference between the approximate and rigorous
overall heat-transfer coefficient occurs at small values of the Fourier time near 1: 8.3 and 4.4% for the skin of 1 and 4, respectively. For values of NGz larger than unity, this time range falls
definitely within the early transient time and thus outside the range
of applicability of the Ramey solution. For values of NGz smaller
than unity, the time range with the largest deviations falls within
the applicability range of the Ramey solution. Then, the error in
the approximate heat-loss rate may cause a significant error in the
predicted tubing temperatures. Whether the method can still be
used depends on the error in the temperature that is acceptable for
the case in question. To estimate the error in wellhead temperature
resulting from a given error in the overall heat transfer coefficient,
we can use the following relation based on the approximate Ramey
solution (Eq. 13):
TwhD =

dTwhD
quD =
dquD

1 TwhDquDNRa + 1

quD
. . . . . . . . . . . . . . . . . . . . . . . . . . . (21)
quD

As an example, for a skin of unity and a Fourier time of unity, the


rigorous and the approximate heat-loss rate are 0.532 and 0.488,
respectively. The relative error in the approximate heat-loss rate is
thus a little more than 8%. The wellhead temperature according to
Ramey for a Ramey number of unity is 0.776. The relative error in
the wellhead temperature as given by Eq. 21 is then 2%. This
example illustrates that the Ramey solution (Eq. 13) is rather forgiving for errors in the unit heat-loss function.
To use Eq. 17, all that is required is an evaluation of the
heat-loss function for zero skin at a particular time. This function
is well documented in the literature and available in graphical
form; see, for example, the log-log graph presented by Earlougher.7 To quickly and accurately reproduce this function, we
present the following mathematical approximations for early, medium, and late Fourier times, respectively:
0
tDFo 1.0: quD
tDFo =

0.79526
4

tDFo

To illustrate the results of this work, we present an example of the


calculation of wellhead temperature of a flowing oil well after a
certain production time. We have taken this example from the SPE
Monograph Multiphase Flow in Wells,8 where it is described as
Example 2.4. Table 1 lists the relevant input data.
To find out the time regime for the production time of 2 weeks,
we first evaluate the dimensionless numbers and the dimensionless
Fourier time. For the Ramey number, Graetz number, and thermal
skin factor we calculate 2.115, 0.402, and 1.328, respectively. The
skin factor (1/UD) is equal to the near-wellbore heat resistance
times the thermal conductivity of the formation (0.94871.4). Two
weeks of production correspond to a dimensionless Fourier time of
120.96. The absolute minimum Fourier time for application of the
Ramey solution is 0.402, equal to the Graetz number. The Ramey
wellhead temperature at this minimum time is approximately 0.65.
From the correlation in Fig. 9, we then estimate for the length of
the early transient period approximately 4*0.4021.6. The Fourier time of 120.96 is well beyond the early transient period, so we
can safely apply Rameys method. In addition, we expect that at
this Fourier time the approximate calculation of the unit heat-loss
function by using Eq. 17 is fairly accurate. (See Fig. 10.)
Table 2 shows the dimensionless input and final results for the
analytical Ramey solution with approximate (second column) and
rigorous (third column) unit heat-loss function, the rigorous solution as discussed in this work (fourth column), and the outcome of
exercise 2.4 reported in Ref. 8 (final column).
As expected, the Ramey solution with both the rigorous and the
approximate unit heat-loss function comes very close to the rigorous solution. The wellhead temperatures of Column 2 and Column 5 are both based on the Ramey solution with approximate unit
heat-loss function and, therefore, ought to be the same. The difference is due to a small error in the calculation of the unit heatloss function quD in Ref. 8 [1/f(t) in the nomenclature of Ref. 8].
There, the authors approximated the heat-loss function by using
the long-time approximation (Eq. 24) which gives rise to an error
of 3.1%. This causes an error in the (real) wellhead temperature of

+ 0.5

tDFo
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (22)

1 tDFo 105 :

1
0
quD
tDFo

= 0.0012ln tDFo3

+0.0249ln tDFo2 + 0.3083ln tDFo + 1.0504 ,


. . . . . . . . . . . . . . . . . . . . . . . . . . (23)
470

December 2004 SPE Journal

T1
u
U
Ui
w
x

almost 0.4%, in agreement with Eq. 21. For the Fourier time of this
example, Eq. 23 is more accurate than Eq. 24.
Conclusions
1. Inspection analysis of the Ramey wellbore heat transmission
equations shows that temperature distribution in a wellbore is
governed by three dimensionless numbers: Graetz number,
Ramey number, and thermal skin factor.
2. Rameys approximate analytical solution for the temperature
distribution depends only on Ramey number and thermal skin
factor.
3. Comparison of Rameys analytical solution with a rigorous solution shows that the Ramey solution is an excellent approximation except for an early transient period. In this early period,
the Ramey solution significantly overestimates the temperatures.
4. The duration of the early transient period can be estimated from
a simple graphical correlation based on the calculated Ramey
wellhead temperature at the moment the tubing fluid has been
completely displaced.
5. Rameys method for estimating the overall heat transfer coefficient holds good only for large values of the Fourier time.
Nomenclature
a constant
b constant
c heat capacity, fluid
cf heat capacity, formation
h heat-transfer coefficient
j index of near-wellbore heat resistance zone
L tubing length
n number of near-wellbore heat resistance zones
NGz Graetz number
NRa Ramey number
q heat loss rate per unit of tubing length
qu unit heat-loss function
r radial distance from center of tubing
rcf radius of formation inner boundary (at interface of
cement zone and formation)
ri inner radius, circular cylinder
ro outer radius, circular cylinder
rt inner radius, tubing
s Laplace variable
t time
tD dimensionless time for convection
tDFo dimensionless time for conduction (Fourier time)
tDFoi duration of initial transient period (Fourier time)
T temperature in tubing
Tbot temperature at bottom of tubing
Tf temperature in formation
Ttop initial temperature at top of tubing
Twh wellhead temperature
T0 initial temperature of formation
December 2004 SPE Journal

temperature at inner radius of tubing


linear flow velocity in tubing
near-wellbore heat transfer coefficient
near-wellbore heat transfer coefficient of zone j
mass flow rate in tubing
distance from bottom of tubing
exponential of Euler constant (1.7810724)
heat diffusivity formation
heat-conduction coefficient
density tubing, fluid
density, formation

Subscripts
D dimensionless quantity
F formation
J index of near-wellbore heat resistance zone
Superscript
* Laplace transform
0 zero thermal skin
References
1. Ramey Jr., H.J.: Wellbore Heat Transmission, JPT (April 1962) 427;
Trans., AIME, 225.
2. Willhite, G.P.: Overall Heat Transfer Coefficients in Steam and Hot
Water Injection Wells, JPT (May 1967) 607.
3. Wu, Y.-S. and Pruess, K.: An Analytical Solution for Wellbore Heat
Transmission in Layered Formations, SPERE (November 1990) 531.
4. Carlslaw, H.S. and Jaeger, J.C.: Conduction of Heat in Solids, second
edition, Oxford U. Press (1959) 184.
5. Stehfest, H.: Numerical Inversion of Laplace Transforms, Communications of the ACM (January 1970) 13, No. 1, 47.
6. Beek, W.J. and Muttzall, K.M.K.: Transport Phenomena, John Wiley
& Sons Ltd., Chichester, U.K. (1975) 176.
7. Earlougher Jr., R.C.: Advances in Well Test Analysis, SPE Monograph
Series, Vol. 5, Dallas, Texas (1977) 38.
8. Brill, J.P. and Mukherjee, H.: Multiphase Flow in Wells, SPE Monograph Series, Vol. 17, Richardson, Texas (1999).
9. Lauwerier, H.A.: The Transport of Heat in an Oil Layer Caused by the
Injection of Hot Fluid, Appl. Sci. Res. (1955) 5, Section A, 145.
10. Marx, J.W. and Langenheim, R.H.: Reservoir Heating by Hot Fluid
Injection, Trans., AIME (1959) 216, 312.

Appendix AHeat-Balance Equations


For the physical model described in the text, the tubing heat balance is given by

qT,x,t
T
T
+ cu
=
. . . . . . . . . . . . . . . . . . . . . . . . . . (A1)
t
x
rt2

Here, q is the heat-loss rate from the tubing to the formation per
unit length of tubing. In general, q depends on temperature, position, and time, and is determined by heat conduction in the formation and by near-wellbore heat resistance. The latter includes
heat transfer at the tubing wall, heat transfer by conduction through
the tubing wall, heat transfer in the annular space between tubing
and casing, heat transfer by conduction through the casing wall,
and finally, heat transfer through the cement zone.2
In the case of a constant and uniform temperature in the tubing,
the heat-loss rate is described by a unit heat-loss function, which
is defined as the heat loss per unit length and per unit drop in
temperature. This unit function follows from a solution of the
formation heat balance, which for the physical model adopted
reduces to the radial Fourier heat-conduction equation. Let us, for
the time being, assume that the temperature distribution for a unit
change in the temperature of the tubing at time zero is known. The
unit heat-loss function is then given by
qu =

q
2rcf f dT
=
T1 T0
T1 T0 dr

, . . . . . . . . . . . . . . . . . . . . (A-2)
r=rcf

471

where T0 is the faraway temperature in the formation and T1 the


(constant) temperature in the tubing. When the temperature in the
tubing is not constant but varies with time, as in the present problem, the heat-loss rate per unit length follows from Duhamels
convolution theorem:
t

q=

T
q t d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-3)
u

Insertion of Eq. A-3 into Eq. A-1 gives

T
T
+ cu
=
t
x

rt2

Likewise, we can write for the Fourier heat conduction (Eq. A-7),
the initial condition (Eq. A-8), and the boundary conditions (Eqs.
A-9 and A-10), respectively:

TfD
TfD
1
r
=
, . . . . . . . . . . . . . . . . . . . . . . . . . . (A-15)
rD rD D rD
tDFo
tDFo = 0, rD 1:
tDFo 0, rD = 1:

t = 0 : T = Tbot Tbot Ttop

. . . . . . . . . . . . . . . . . . . . . . . (A-4)

x
. . . . . . . . . . . . . . . . . . . . . . . . (A-5)
L

quDtDFo;UD =

t 0, x = 0 : T = Tbot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-6)
The unit function in Eq. A-4 can be derived from the solution of
the radial Fourier heat-conduction equation for the formation surrounding the wellbore:

1
Tf
1 Tf
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-7)
r
=
r r
r
t
where f /(f cf).
Initially, the temperature is constant throughout the formation,
thus:
t = 0 : Tf = T0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-8)
For the inner boundary condition at the interface between cement
zone and formation, we have
Tf
= UT1 Tf, . . . . . . . . . . . . . . . . . (A-9)
r

where U is the heat-transfer coefficient for heat flow across the


near-wellbore heat resistances.
At the outer boundary at infinity, we have
Tf = T0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-10)

To make the previous set of equations dimensionless, we introduce


the following dimensionless variables:
TD =
tD =

T Ttop
Tf T0
x
r
,T =
,x = ,r = ,
Tbot Ttop f D T1 T0 D L D rcf

u
1
t, tDFo = 2 t, and quDtDFo =
q .
L
2f u
rcf
. . . . . . . . . . . . . . . . . . . . . . . .(A-11)

The tubing heat balance (Eq. A-4) and the associated initial (Eq.
A-5) and boundary (Eq. A-6) conditions then become, respectively:

tD = 0 :

TD = 1 xD , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-13)

tD 0, xD = 0:
472

TfD
rD

rD=1

NRa =
NGz =

2L

curt2

2L
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-20)
cw

tDFo L
=
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-21)
tD urcf2

and
UD =

Urcf
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-22)
f

In petroleum engineering, related heat-transfer problems, viz. 1D


convective heat flow with heat losses perpendicular to the flow
direction, have been examined by Lauwerier9 and Marx and Langenheim.10
Appendix BSolution of the Radial
Fourier Equation
The radial Fourier equation can be conveniently solved by Laplace
transformation. The subsidiary equations of the Fourier equation
(Eq. A-15) and the associated boundary conditions (Eqs. A-17 and
A-18) are, respectively,

1 d
dT*D
rD
= sT*D , . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-1)
rD drD
drD
rD = 1 :

1
dT*D
= UD T*D , . . . . . . . . . . . . . . . . . . . . . . . (B-2)
drD
s

rD = : T*D = 0 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-3)
where s is the Laplace variable and the asterisk denotes Laplace
transform.
Eq. B-1 is known as the modified Bessel equation of order zero.
Its general solution reads:
T*D = aI0rDs + bK0rDs , . . . . . . . . . . . . . . . . . . . . . . (B-4)
where I0 and K0 are the modified zero-order Bessel functions of
the first and second kind, respectively, and a and b are constants
that follow from the boundary conditions. The boundary condition
in Eq. B-3 requires that a must be zero. The constant b follows
from the radiation boundary condition (Eq. B-2) and is given by
UD
sK1ss + UDK0s

. . . . . . . . . . . . . . . . . . . (B-5)

The Laplace transform of the unit heat-loss function then becomes

TD()
quDNGztD d , . . . . . . . . . . . . . . . . . . (A-12)

NRa

The dimensionless numbers NRa and NGz in Eq. A-12 and UD in


Eq. A-17 are defined by, respectively,

b=

TD TD
+
=
tD xD
tD

= UD1 Tf D, . . . . . . . . . . . . . . . (A-17)

UD1 Tf DrD = 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-19)

The boundary condition at the entrance of the tubing reads:

t 0, r = :

rD

The dimensionless unit heat-loss function follows from:

Initially, the vertical temperature in the tubing is equal to the linear


geothermal temperature distribution in the surrounding formation.
Therefore,

TfD

tDFo 0, rD = : Tf D = 0 . . . . . . . . . . . . . . . . . . . . . . . . . (A-18)

T
q t d
u

t 0, r = rcf :

TfD = 0 , . . . . . . . . . . . . . . . . . . . . . . . . . (A-16)

TD = 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-14)

q*uD =


T*D
rD

rD=1

UDK1s

sK1ss + UDK0s

. . . . . . . . . . . . . . . . . . . . . (B-6)

December 2004 SPE Journal

Back transformation of Eq. B-6 with the help of the Stehfest inverter yields the unit heat loss as a function of dimensionless
Fourier time. We have used an implementation of the Stehfest
algorithm with 12 Stehfest terms. The algorithm very closely reproduces the analytical early- and long-time approximations of the
constant terminal temperature solution (zero thermal skin).
Appendix CRigorous Solution of the Tubing
and Formation Heat-Balance Equations
We have solved the tubing and formation heat-balance equations
by means of Laplace transformation. The subsidiary equations of
the tubing heat-balance equation (Eq. A-12) and its boundary condition (Eq. A-14) are
dT*D
sT*D 1 xD +
=
dxD

quDc
* =

UD
1
. . . . . . . . . . . . . . . . . . . . . . . (C-9)
NGz s NGz + UDs NGz

From Eq. C-9, we see that the Laplace transform of the unit heatloss function becomes negligibly small for large s. The Laplace
transformed temperature (Eq. C-6) then simplifies to
T*D =

1
2

1 exp sxD +

Back transformation of Eq. C-10 gives for the temperature ahead


of the fluid front (tD<xD)
TD = tD + 1 xD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (C-11)
For large times, s is small and the Laplace transform of the unit
heat-loss function is large. Therefore, Eq. C-6 becomes

NRasT*D 1 xDquDc
* , . . . . . . . . . . . . . . . . . . . . . . . (C-1)
1
xD = 0 : T*D = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (C-2)
s

T*DxD =

1
2

s NRaquDc
*

* sxD +
1 exp NRaquDc

where
* = LquDNGztD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (C-3)
quDc
Rearranging Eq. C-1 leads to the ordinary differential equation
dT*D
= NRaquDc
* + 1sT*D 1 xD . . . . . . . . . . . . . . . . . (C-4)
dxD
Integration of Eq. C-4 subject to the boundary condition (Eq. C-2)
yields
lnsNRaquDc
* + 1sT*D 1 + xD + 1
= sNRaquDc
* + 1xD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (C-5)

1
s NRaquDc
* + 1

* + 1sxD +
1 exp NRaquDc

1 xD
. . . . . . . . . . . . . . . . . (C-6)
s

To complete the solution in Laplace space, we have to convert the


Laplace transform of the unit heat-loss function Eq. B-6 into the
Laplace transform of the unit heat-loss function for the convection
time scale as used in Eq. C-6. This can be done by using the
relationship
1
LFat = LFt ss a . . . . . . . . . . . . . . . . . . . . . . . . . . (C-7)
a

Thus,
quDc
* =

T*DxD =

K1s NGzs NGz + UDK0s NGz

Neglecting the third and higher-order terms, we obtain


1 xD xD 1
+ NRaquDc
* xD2
s
s 2!

1 1
= NRaquDc
* xD2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (C-14)
s 2!
Back transformation yields
TDxD,tDFo = 1

1
N q t x2 . . . . . . . . . . . . . . . . . . . (C-15)
2! Ra uD DFo D

Appendix DSolution of the Heat Balance for


Constant Heat-Loss Function
In the case of a constant unit heat-loss function (Eq. A-12) reduces to

where TD(0) is the initial temperature profile in the tubing given by


. . . . . . . . . . . . . (C-8)

Eqs. C-6 and C-8 constitute the solution of the temperature distribution in Laplace space. Back transformation by the Stehfest inverter yields the temperature distribution in real space. We have
used the same implementation of the Stehfest algorithm as mentioned in Appendix B. We have checked this algorithm against the
analytical solution for the special case of a constant unit heat-loss
function. The algorithm does fairly well in calculating the linearly
increasing wellhead temperature during the transient period and
very well in the period thereafter when the wellhead temperature
remains constant. The difference between the constant temperature
and the calculated temperature is less than 1% for Fourier times
larger than 1.3 times the Graetz number.
For small times and thus large s, the ratio K1/K0 approaches
unity and the Laplace transformed unit function then becomes
December 2004 SPE Journal

1
N q * sx 2 + ..
2! Ra uD D

TD TD
+
NRaquDTD TD0 , . . . . . . . . . . . . . . . . . . . (D-1)
tD xD

1
1
q*
=
NGz uD ss NGz NGz s NGz
UDK1s NGz

1
1 xD
+ 2
s
s NRaquD
*

. . . . . . . . . . . . . . . . . . . . . . . .(C-13)

T*DxD =

1 xD
. . . . . . . . . . . . . . . . . . . . . (C-12)
s

For small s, the product of s and the Laplace transform of the unit
heat-loss function is also small. Expansion of the exponent in Eq.
C-12 gives

1 1 NRaquD
* sxD +

Solving Eq. C-5 for the Laplace transformed temperature, we obtain


T*D =

1 xD
. . . . . . . . . . . . . . . . . . . . (C-10)
s

TD0 = 1 xD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (D-2)
Eq. D-1 can be conveniently solved by Laplace transformation.
The subsidiary equation of Eq. D-1 is
sT*D TD0 +

dT*D
=
dxD

1
NRaqDT*D + NRaqDTD0 . . . . . . . . . . . . . . . . . . . . . . . (D-3)
s
The boundary condition (Eq. A-14) in Laplace space becomes
1
xD = 0 : T*D = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (D-4)
s
Rearranging Eq. D-3 leads to the ordinary differential equation
473

dT*D
= NRaqD + sT*D +
dxD

TD0 1 + NRaqD

TD = 1 xD +

1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (D-5)
s

Integration of Eq. D-5 subject to the boundary condition (Eq. D-4)


yields
ln NRaquD + ssT*D + 1 xD =

1 exp NRaquDtD
, . . . . . . . . . . . . . . . . . . . (D-9)
NRaquD

which represents a linear temperature profile parallel to the initial


profile. For tD1, the temperature profile (Eq. D-8) becomes
TD = 1 xD +

1 exp NRaquDxD
. . . . . . . . . . . . . . . . . . . (D-10)
NRaquD

The profile (Eq. D-10) is the same as the Ramey profile (Eq. 13).

NRaq*
uD + s xD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (D-6)
Solving Eq. D-6 for the Laplace transformed temperature, we obtain
1
T*D =
sNRaquD + s

1 exp NRaquD + sxD +

1 xD
. . . . . . . . . . . . . . . . . . (D-7)
s

Back transformation of Eq. D-7 gives


1 exp NRaquDtD
TD = 1 xD +
HtD xD
NRaquD
1 expNRaquDtD xD
exp NRaquDx
NRaquD
. . . . . . . . . . . . . . . . . . . . . . . . .(D-8)
where H denotes the Heaviside function.
The displacement front between the original fluid in the tubing
at time zero and the fluid that has entered the tubing during a time
tD is positioned at xD=tD. Ahead of the displacement front
(xD>tD), Eq. D-8 reduces to

474

SI Metric Conversion Factors


bbl 1.589 873
E01
Btu 1.055 056
E+00
ft 3.048*
E-01
F (F-32)/1.8

in. 2.54*
E+00
lbm 4.535 924
E01

m3
kJ
m
C
cm
kg

*Conversion factors are exact.

Jacques Hagoort has been a reservoir engineering consultant


with Hagoort & Assocs. B.V. since 1983. e-mail: jacques@
haa.nl. Previously, he worked for Shell as a reservoir engineer in
The Netherlands, the U.S., and Canada. From 1987 to 2003,
Hagoort was also a part-time professor of reservoir engineering
at the Delft U. of Technology. Hagoort holds MS and PhD degrees from the Delft U. of Technology. A Distinguished Member
of SPE, Hagoort served as a 199091 SPE Distinguished Lecturer
and is a recipient of the 1975 Rossiter W. Raymond Award, the
1998 Lester C. Uren Award, and the 2001 Cedric K. Ferguson
Certificate.

December 2004 SPE Journal

S-ar putea să vă placă și