Modeling wax deposition with deposition-

release models

Kjetil Kandal Botne

Trondheim
December 2011
Preface

This project is carried out in TPG4510 Petroleum Production, Specialization Project. This
subject is compulsory in the 9th semester at NTNU for Petroleum Production. The subject
counts for 15 points, half of one normal semester. I would like to thank my supervisor,
Professor Jon Steinar Gudmundsson, for the topic and guidance through the semester.

II
Abstract

Deposition models are used to understand and predict deposition of solids. A major flow issue
in the oil industry is related to deposition of paraffin wax. In this project two wax deposition
models are evaluated. They are similar mathematical models called the exponential and
logarithmic model. The rate of deposition in each model is decided trough thickness of wax
layer and two correlations k1 and k2. Both models were matched against experimental data
from a single phase wax deposition experiment. The experiment consists of one series with
varying flow rate and one with varying temperature. The exponential model was a poor match
to these data and was not investigated further. The logarithmic model proved to be a good
way of modeling these experiments. Values of k1, controlling initial deposition rate, seem to
be a function of 1/Re for varying flow rates. When the temperature series is investigated k1
seem to be proportional to ΔT+. Values of k2, limiting deposition over time, seem to be
proportional to Re2 when varying flow rate series is evaluated. When the temperature series is
considered k2 seem to be proportional with 1/ ΔT+. Two new equations based on these
dependencies were made. Limited data made it hard to give dependencies on both Re and ΔT+
in the same equation.

III
Table of contents

Preface ....................................................................................................................................... II
Abstract .................................................................................................................................... III
Table of contents ...................................................................................................................... IV
1. Introduction ........................................................................................................................ 1
2. Paraffin wax and wax related problems ............................................................................. 2
3. Wax deposition experiments .............................................................................................. 3
3.1 Experimental data ........................................................................................................ 3
3.1.1 Experiments with varying flow rate ..................................................................... 3
3.1.2 Experiments with varying condensate temperature ............................................. 4
3.2 Additional data ............................................................................................................ 4
4. Wax deposition models ...................................................................................................... 5
4.1 Deposition-release model ............................................................................................ 5
4.1.1 Exponential model................................................................................................ 5
4.1.2 Logarithmic model ............................................................................................... 6
5. Matching the experimental data with a model ................................................................... 7
5.1 Evaluating k1................................................................................................................ 7
5.2 Exponential model and k2,exp ....................................................................................... 7
5.3 Logarithmic model and k2,log ........................................................................................ 8
6. Determine dependencies of physical parameters ............................................................... 9
6.1 Varying rate ................................................................................................................. 9
6.2 Varying temperature .................................................................................................... 9
7. Discussion ........................................................................................................................ 11
8. Conclusion ........................................................................................................................ 14
9. Nomenclature and Formulas ............................................................................................ 15
10. References ..................................................................................................................... 16
Tables ....................................................................................................................................... 17
Figures ...................................................................................................................................... 19
Appendix A: Digitizing graphical data .................................................................................... 29
Appendix B: Derivation of the exponential model: ................................................................. 30
Appendix C: Derivation of the logarithmic model: ................................................................. 31

IV
Appendix D: Estimating k1 and k2 in the models ..................................................................... 32
Estimating k1 ........................................................................................................................ 32
Estimating k2 ........................................................................................................................ 32
Exponential model............................................................................................................ 32
Logarithmic model ........................................................................................................... 33

V
1. Introduction

Wax deposition in pipelines is one of the biggest flow assurance challenges in the oil industry.
The need for understanding deposition becomes greater as hydrocarbons are being transported
over increasingly greater distances. Wax precipitates from oil when it is cooled and may
deposit on pipeline walls. Wall deposits can lead to severe problems and need to be removed
in an efficient way. It is difficult to perform accurate deposition measurements on real
pipelines. That is why laboratory experiments are performed, and models based on
experiments are developed. Such a model could be scaled up to field data and used to predict
deposition of solids.

In this project I investigate two versions of the deposition-release model. The exponential
model and the logarithmic model are evaluated against experimental data presented by
Rosvold (2008). The first task is to find out which model that best suits the experimental data.
The second part is to relate model constants to physical parameters that are important
mechanisms in deposition. In this project the two mechanisms investigated is varying flow
rate and varying flow temperature.
2. Paraffin wax and wax related problems

Deposition of paraffin wax is a major issue in the oil industry. Wax precipitates from both
crude oil and condensate when temperature falls below a certain value. As oil temperature
decreases further, more wax will precipitate. If wax precipitates it may deposit in pipes and
equipment causing flow problems. Wax deposition is mainly a problem in pipelines,
production equipment and in wells. Deposited wax in pipelines may cause increased pressure
drop, decreased production and clogged pipes. It may also damage production equipment or
make it less efficient.

Wax molecules are mostly long chain n-alkanes, and weight% of 1-15 is considered typical in
both crude oil and condensates (Aske 2011). These n-alkanes normally have a carbon number
between C20 and C40 (Gudmundsson 2010). When oil is cooled below a certain temperature,
wax will start to precipitate. This temperature is called the wax appearance temperature
(WAT), and is normally found around 30 - 40°C (Gudmundsson 2010). The term cloud point
temperature is another term used to describe the WAT. Below the cloud point there is another
temperature called the pour point temperature. When the pour point is reached the paraffin
wax will become a soft solid (a gel). This solid loses the ability to flow and may deposit.

Normal ways of preventing wax deposition in pipelines includes heating or insulation, pigging
and chemical injection. Active heating is expensive and limited by distance, especially for
subsea pipelines. Insulation is common on long land pipelines (Gudmundsson 2010). Pigging
is also common and removes deposited wax mechanically by scraping it of the wall. In a
startup phase pigging is usually performed when wax thickness reaches 2-3 mm (Labes-
Carrier 2002). This criterion is set to avoid incidents with a stuck pig. Chemical additives may
prevent agglomeration of wax molecules and prevent wax from depositing on the wall
(Gudmundsson 2010).

2
3. Wax deposition experiments

The experimental data used in this report was primarily taken from Karianne Rosvold’s
master thesis from 2008. The experiments were performed in a test rig at Statoil’s multiphase
flow loop laboratory in Porsgrunn. The test fluid was a waxy gas condensate and it was
operated at atmospheric pressure (Rosvold 2008). Detailed information about the rig can be
found in Table 1. The same experimental data are also presented in a paper by Rainer
Hoffmann and Lene Amundsen (2010).

3.1 Experimental data

The experimental data available to me was the graphs from Rosvold’s thesis. These data was
presented as plots with measured wax thickness (mm) vs. time (h). In order to utilize the
experimental data and make new plots, the plots were digitized. This procedure is explained
in Appendix A. Some additional data were taken from Hoffman and Amundsen (2009). This
data was used in order to perform calculations.

Experiments were performed by flowing condensate trough a horizontal steel pipe with length
5.55 m. The fluid is assumed to be single phased. The outside of the pipe was cooled by water
at 10°C in all experiments. The wax thickness in the pipe was estimated using measurement
of pressure drop. Two different flow series experiments were performed. One series varied
flow rate and kept condensate temperature constant. The other one did the opposite and varied
condensate temperature while keeping flow rate constant. The total of eight experiments were
given by Rosvold (2008).

3.1.1 Experiments with varying flow rate

Five runs with flow rate ranging from 5 m3/h to 25 m3/h are given by Rosvold (2008). The
condensate temperature Tcond = 20°C in all five runs. Experiments were run between 65 and
160 hours. The digitized version of these experiments can be seen in Figure 1. Specific values
from this flow series are found in Table 2, and constant values in Table 3. The fluid velocity,
u, in Table 2 ranges from 0.66 m/s to 3.31 m/s. A typical fluid velocity for oil would be from
2 m/s to 4 m/s (Gudmundsson 2009). Reynolds number ranges from 13836 to 69179, all in the
turbulent flow area.

3
3.1.2 Experiments with varying condensate temperature

Four experiments with condensate temperature between 15°C and 40°C are given. Flow rate
is kept constant at 21 m3/h. These experiments were run between 100 and 335 h. The
experiment with Tcond = 20°C and Q = 21 m3/h is part of both series. The digitized
experiments can be seen in Figure 2. Specific values from these runs are given in Table 4, and
constant values in Table 5. The oil density is given as a constant for temperature because it is
only given at 20°C (Hoffmann and Amundsen 2010).

3.2 Additional data

Very little data about the condensate was given by Rosvold (2008). The only fluid data given
was the wax appearance temperature of 45°C. A WAT of about 30°C was reported by
Hoffman and Amundsen (2009). This value was determined through several tests that all
came up with a WAT of about 30°C. The fluid used in the experiment was a North Sea
condensate with wax content of about 4.5 % (Hoffman and Amundsen 2010). A WAT of
26°C at atmospheric conditions was given in another study of a North Sea gas condensate by
Labes-Carrier et al. (2002). The wax content of this condensate was similar at 4.4 %. Using
this information it was decided to base calculations on a WAT of 30°C.

Fluid density (at 20°C), viscosity and wall temperatures during some experiments were given
by Hoffman and Amundsen (2010). Viscosities are given in a figure plotting viscosity against
temperature. The viscosities are taken from this plot as accurate as possible. Wall
temperatures are given by Hoffmann and Amundsen (2010) for all runs in the varying
temperature series. For the varying flow rate series wall temperature was only given for two
out of five runs. In order to obtain a value a linear relationship between wall temperature and
flow rate was assumed. Estimated values for the three other runs were obtained using this
assumption. They may not be completely accurate but they are given in Table 2. Reynolds
number given in Table 4 may also be inaccurate since oil density at 20°C is used in all
calculations.

4
4. Wax deposition models

Deposition models can be used to predict how wax builds up with time. A figure plotting
deposition thickness as a function of time is an effective way of showing deposit buildup.
Figure 3 show some of the different trends that are possible outcomes of a deposition
experiment. These trends are linear buildup trend, an exponential (asymptotic) trend and a
logarithmic trend. Of course it may be other possibilities or mixtures of the three trends given.
In Figure 4 we can see how the rate of deposition is influenced by deposition thickness. A
linear trend might cause a plugged pipe. In the other cases deposition may stop when a certain
wax thickness is reached. Many models exist and some are even a part of simulation software.
The focus here is on the deposition release model.

4.1 Deposition-release model

When experimental results are obtained we try to match them with a model. If data can be
matched with a model, the model can be developed in order to predict wax buildup. This
project focuses on two simple mathematical models and tries to match them with
experimental data. These models are the exponential model and the logarithmic model. Both
of these models are based on the deposition release principle. One part of the model concerns
the buildup of deposition with time. The other part of the model concerns deposits breaking
down. When these two mechanisms are equally big the deposition will stop. This is seen in
Figure 4 when rate of deposition goes towards zero. The deposition release model is typical
for deposition caused by changes in pressure or temperature. Like wax deposition which
occurs when oil is cooled.

4.1.1 Exponential model

This model is based on the exponential line in Figure 3. It has been showed to work well for
different types of deposition situations by Gudmundsson (1981). Rate of wax deposition is
described by the equation

Both k1 and k2 are constants or functions, x is wax thickness [mm] and t is time [hour]. When
integrated the deposition thickness as a function of time is described by

See Appendix B for the full derivation. The initial rate of deposition is

5
 .

When time is increased the deposition thickness will reach its maximum asymptotic level at

4.1.2 Logarithmic model

In some cases the exponential deposition-release model does not match the data. A similar
model with logarithmic increase has been proposed by Gudmundsson (2010). This model
resembles the logarithmic line in Figure 3. Rate of wax deposition is described by

Similar to the exponential model both k1 and k2 are constants or functions, x is wax thickness
[mm] and t is time [hour]. If the equation is integrated and solved for deposition thickness we
get
.

See the full derivation in Appendix C. As for the exponential model the initial rate of
deposition is

( )

When using this model the deposition thickness will not stop increasing, but the rate of
deposition will decline. In practical cases an asymptotic level is likely to be reached, if not
this deposition trend leads to a clogged pipe. An example is k2 = 500 and x = 2mm: 500-2 =
4E-6 ≈ 0 => dx/dt = 0. The bigger the value for k2, the faster k2-x goes towards zero.

6
5. Matching the experimental data with a model

Both the exponential and the logarithmic deposition model were matched to the experimental
data. In order to do this, values for k1 and k2 had to be identified. The detailed procedure for
estimating k1 and k2 is explained in Appendix D. Parameter k1, which is initial wax buildup
rate, is the same parameter for both models. Parameter k2 has a different value for each of the
two models, these are called k2,exp and k2,log. For the best possible match both models were
adjusted by trial and error. The eight experiments and the best fit with each model are given in
Figure 5-12. The values for k1 and k2 from each model is given in Table 6 and 7, for all eight
experiments. The values given in the tables are the values which gave the best fit for each
model.

5.1 Evaluating k1

Figure 5-12 makes it easy to evaluate the two models against the experimental data. From the
figures we observe that the initial deposition rates for both models match the experiments
well. That means that estimates of k1 are good in most cases. The different k1 values matched
to variable rate experiments are found in Table 6. There seems to be no real consistency in
this value related to the increase in condensate rate. The k1 values estimated from the variable
temperature experiments are given in Table 7. One k1 value in this Table, Tcond = 20°C, is
much higher than the others. If we exclude this value the other three k1’s decreases with
increasing condensate temperature.

5.2 Exponential model and k2,exp

From Figure 5-12 we observe that the exponential model does not match most experiments.
The only experiment with a good match to the exponential model is Tcond 40 = °C and Q = 21
m3/h. All other experiment does not reach an asymptotic value. At least not in the timeframe
used in these experiments. Another issue is that the asymptotic value of the exponential model
is reached to fast. A lower value of k1 and adjusted k2,exp makes it possible to increase the time
it takes to reach this asymptotic value. But then we are faced with the problem that the initial
deposition rate is predicted as too low. Because of the poor match between experiments and
this model it is hard to determine the best k2,exp values. This poor match also makes it hard to
comment on trends in k2,exp values given in Table 6 and 7.

7
5.3 Logarithmic model and k2,log

When evaluating Figure 5-12 we observe that the logarithmic model gives a good match in
most cases. The only bad match is the Tcond = 40 °C and Q = 21 m3/h, which was the best
match to the exponential model. Because this model seems to be more compatible to the data
it is easier to comment on the k2,log values. In Table 6 we can see that k2,log increases with
increasing condensate rate. When comparing k2,log values for the varying temperature in Table
7 one value stands out. The k2,log for Tcond = 20°C is higher than the rest, the same trend as for
the k1 value. If this value is excluded an increasing k2,log with increasing temperature is
observed.

8
6. Determine dependencies of physical parameters

Changes in k1 and k2 suggests that changes in both flow rate and temperature influence the
wax deposition. The same trend is observed in Figure 1 and 2. After evaluating models and
determining k1 and k2 values in each experiment the focus is now on determining physical
parameters. These physical parameters are based on flow rate changes or temperature
changes. The focus will now be on the logarithmic model. The exponential model was a poor
fit to the experiment and will not be investigated further.

6.1 Varying rate

When condensate flow rate is increased values for both k1 and k2,log change. Parameter k2,log
increases with increasing flow rate as seen in Table 6. It is hard to say anything about k1
because it shows no clear trend. Reynolds number (Re) is a function of fluid velocity and
increase with increasing velocity. In Figure 13 k1 is plotted against 1/Re. In this figure we
observe a possible linear relationship between rates 10 m3/h, 15 m3/h and 25 m3/h. That could
indicate that k1 is a linear function of 1/Re. That means that in this case k1 = k3 /Re + k4,
where k3 and k4 are new constants. Another linear trend is observed when k2,log is plotted
against Re2 as seen in Figure 14. This trend indicates that k2,log is proportional to Re2. That
means that k2 = k5 * Re2. The points furthest away from this linear trend have the rate 21 m3/h
and 25 m3/h. If the correlations for k1 and k2 are used the result becomes

⁄
( ⁄ )

Where constants in this case are k3 = 1.75*104, k4 = -0.186 and k5 = 4*10-7. This equation is
only dependent on Re and wax thickness, the temperature is not included. In Figure 17 k1 is
plotted against ΔT+/Re. The same linear trend seen in Figure 13 is observed. That could give a
possible correlation of k1 with both ΔT+ and Re.

6.2 Varying temperature

When temperature of condensate increases values k1 and k2,log change. Previously it was
pointed out from Table 7 that k1 seems to decrease and k2,log increase with increasing
temperature. ΔT+ is a dimensionless temperature driving force parameter and a function of
wall, fluid and cloud point temperature. The equation is given in Nomenclature and Variables

9
and was given by Gudmundsson (2010). If k1 is plotted against ΔT+ a linear relationship
between three out of four points can be seen in Figure 15. This linear relationship implies that
k1 is proportional to ΔT+. That means that k1 = k3* ΔT+. The one not in line with the others is
Tcond = 20°C. If constant k2,log is plotted against 1/ ΔT+ a linear relationship between three
points can be seen in Figure 16. This relationship implies that k2,log might be proportional to 1/
ΔT+. That means k2 = k4 / ΔT+. The same point, Tcond = 20°C, does not appear on this line. If
the new correlations of k1 and k2 are used in the model the result becomes

⁄
⁄

Where constants in this case have the values k3 = 5.72*10-2 and k4 = 1.31*102. In this
equation the only parameters are the temperature parameter and wax thickness. The ideal case
would be if the data made it possible to make an equation of both Re and ΔT+. The linear
relationship between k1 and ΔT+/Re was worse than with k1 and ΔT+. That means that the
correlation found for the flow rate series was not a good match. The correlations presented
here in the equations were the best correlations with the data available.

10
7. Discussion

A claim stated by Gudmundsson (2010) suggests that “The initial rate of deposition and the
asymptotic deposition both decrease with increased flow rate”. Figure 1 shows the digitized
version of the experiments with varying rate presented by Rosvold (2008). The initial rate of
deposition seen in Figure 1 is not easy to evaluate. But it seems like three series Q = 10, 15
and 25 m3/h show the expected trend. The other two, Q = 5 and 21 m3/h, deviates from this
trend. In the model k1 shows the initial deposition rate. Values of k1 in Table 6 indicate the
same as Figure 1. That Q = 5 and 21 m3/h deviates from the trend that would be expected.
Three out of five series confirms the claim that the initial rate of deposition decrease with
increasing flow rate. The asymptotic deposition can also be evaluated from Figure 1. The
asymptotic level seems to decrease with increasing flow rate. One exception is Q = 21 m3/h
which have a lower asymptotic deposition than expected. Values of k2,log in table 6 increase
with increasing flow rate. An increased k2,log gives a decreased asymptotic level. Both Figure
1 and k2,log values seem to confirm that asymptotic level decrease with increasing flow rate.
Experiment Q = 5 m3/h had a big spread in data and initial deposition rate was hard to digitize
correctly. This is shown in Figure 19. That might be a reason for why it does not behave as
expected.

Another claim by Gudmundsson (2010) states that “The initial rate of deposition and the
asymptotic deposition both increase with increased difference between solution cloud point
and wall temperature”. When condensate temperature is decreased the difference between
solution cloud point (WAT) and wall temperature is increased. That means that with lower
condensate temperature a higher rate of deposition and asymptotic deposition is expected.
From the experiments in Figure 2 it is clear that three out of four experiments behave as
expected. The run with Tcond = 20°C and Q = 21 m3/h shows a deviation. The same trend is
observed for k1 and k2,log values in Table 7. The k1 value decrease with increasing flow rate
and k2,log values increase with increasing flow rate. This expected trend is seen in the same
three runs as in Figure 1. That means that the claim is confirmed for three runs in both Figure
2 and values from Table 7. The Tcond = 20°C and Q = 21 m3/h does not show the expected
behavior. This run is part of both the rate and the temperature series. It does not really behave
as expected in either of the series, with the biggest deviation the temperature series. One
possible explanation for this is that the wax thickness is estimated trough measurement of
pressure loss. This may cause inaccuracies, especially since wax roughness is not known.

11
When observing Figure 5 - 12 it is clear that the logarithmic model matches the experiments
far better than the exponential model. The exponential model was a poor fit to seven out of
eight experiments. The constant k2,exp for the exponential model was therefore not evaluated
against physical parameters. The logarithmic model was a good fit to the experiments and
constant k2,log was evaluated against physical parameters. Because of limited data available
the constants were only evaluated against simple rate dependent and simple temperature
dependent parameters.

The statement discussed above is that rate of deposition, k1, decreases with increasing flow
rate. This is only clear for 3 out of 5 runs in the flow rate series. It was shown in Figure 13
that k1 for these three runs form a linear relationship with 1/Re. The same linear trend was
also observed for other rate dependent parameters (eg. u, u*). The wall temperature is also
dependent on flow rate, although the difference between runs is small. Since only two wall
temperatures are given for the flow rate series the estimated ones may be slightly inaccurate.
The temperature driving force have been introduced along with Re in Figure 17. Plotting k1
against ΔT+/Re does not change the linear relationship between the three points. When it
comes to the k2,log values all flow runs show the same trend, decreased asymptotic deposition
with increasing flow rate. A relationship that suggest that k2,log is proportional to Re^2 is
observed in Figure 14. The run furthest from this trend is the run mentioned before, Tcond
20°C and Q 21 m3/h. Using the relationship between k1 and k2,log with Re we get equation 6.1.
This equation considers Re and three new constants given as k3, k4 and k5. The k2,log values in
this series show a better consistency than k1. That might indicate that k2,log is more stable and
easier to estimate than k1. Or it could mean that initial deposition is harder to calculate
correctly using the pressure loss method.

The other statement discussed above states that k1, initial rate of deposition, should decrease
with increasing temperature. The k2,log should increase with increasing temperature, indicating
a decrease in asymptotic deposition level. As mentioned both of these are true for three out of
four runs, all except Tcond = 20°C. As seen in Figure 15 an almost linear relationship between
these three is observed when k1 is plotted against ΔT+. This indicates that k1 is proportional to
ΔT+. Plotting k1 against ΔT+/Re gave a good trend for the rate series seen in Figure 17. If k1
for temperature series are plotted against ΔT+/Re a linear relationship is harder to see. But
change in Re with temperature is only based on change in viscosity and not in density. That
may cause Re to be inaccurate. When considering k2,log the best linear relationship between

12
these three points appear when k2,log is plotted against 1/ΔT+. This indicates that k2,log is
proportional to 1/ΔT+. The same relationship appears if k2,log is plotted against Re/ΔT+.
Considering temperature dependence a new equation is formed, equation 6.2. This equation
introduces two new constants k3 and k4. As stated before the Tcond = 20°C does not match the
other runs. Considering the other three runs it seems like k1 α ΔT+ while k2,log α 1/ΔT+.
Limited data and possible inaccuracies made it hard to evaluate dependence on both
temperature and Re at the same time. Some possible correlations with both was discovered
but was not consistent for both series. That is why the equations 6.1 and 6.2 only use either Re
or ΔT+, not both.

13
8. Conclusion

- The logarithmic deposition release model does a much better job modeling these
experimental data than the exponential model. This is observed in Figure 5-12.
- For the varying rate series only three out of five runs show the expected trend in initial
deposition rate. Values for k1 on these three runs seems to be a linear function of 1/Re.
The linear function is equally good when compared to ΔT+/Re. The asymptotic
deposition level is more consistent, and values for k2,log appears to be proportional
with Re2. The dependencies of Re are used to form equation 6.1.
- When varying temperature three out of four series show the expected trend related to
initial deposition and asymptotic level. If focusing on these three the initial rate k1
seems proportional to ΔT+. The asymptotic level controlled by k2,log seems to be
proportional with 1/ ΔT+. This help form equation 6.2.

14
9. Nomenclature and Formulas

Q [m3/h] – Flow rate

ID [m] – Inner diameter of flow pipe

A [m2] – Inner area of flow pipe,

u [m/s] – Flow velocity,

ρ [kg/m3] – Density of gas condensate

μ [Pas] – Viscosity if gas condensate

Re – Reynolds number, dimensionless,

Tc [°C] – Cloud point temperature, also known as WAT

Tw [°C] – Wall temperature

Tcond [°C] – Condensate temperature

ΔT+ – Dimensionless deposition temperature driving force (Gudmundsson 2010),

15
10. References

1. Aske, N., Statoil ASA (2011): Wax Control, Guest Lecture in TPG 4135 Processing of
Petroleum, 7 March,
[Link]

2. Gudmundsson, J.S. (1981): Particulate Fouling, Fouling of Heat Transfer Surfaces,

Somerscales, E.F.C. and Knudsen J.G. (eds.), Hemisphere Publishing Corporation,
Washington, 357-387

3. Gudmundsson, J.S. (2010): Flow Assurance, Solids in Oil and Gas Production (First
Draft), Department of Petroleum Engineering and Applied Geophysics, Norwegian
University of Science and Technology, Trondheim.

4. Gudmundsson, J.S. (2009): Kompendium TPG 4135, Prosessering av Petroleum,

Grunnleggende enhetsoperasjoner i produksjon av olje og gass, Department of
Petroleum Engineering and Applied Geophysics, Norwegian University of Science
and Technology, Trondheim.

5. Hoffman, R. and Amundsen, L. (2010): Single-Phase Wax Deposition Experiments,

Energy & Fuels, Vol. 24, 1069-1080.

6. Labes-Carrier, C., Rønningsen, H.P., Kolnes, J., Leporcher, E. (2002): Wax

Deposition in North Sea Gas Condensate and Oil Systems: Comparison Between
Operational Experience and Model Prediction, SPE Annual Technical Conference and
Exhibition, 29 September-27 October, San Antonio, Texas, SPE 77573, 12pp

7. Rosvold, K. (2008): Wax Deposition Models, [Link]. Thesis, Department of Petroleum

Engineering and Applied Geophysics, Norwegian University of Science and
Technology, Trondheim, 104 pp.

16
Tables

Table 1: Porsgrunn wax rig data. Ref Rosvold (2008)

Parameter Value
Oil Pipe, ID 52,58 mm
Oil Pipe, OD 60,56 mm
Water Pipe, ID 131,33 mm
Water Pipe, OD 0,1397 m
Tank Volume 300 l
Epoxy coating pipe diameter 51,7 mm
Differential pressure length 5,55 m
Water jacket length 5,31 m
Fluid A waxy gas- condensate

Table 2: Specific values for varying rate experiments.

Rate of condensate [m3/h], Q 5 10 15 21 25
Fluid velocity [m/s], u 0,66 1,32 1,98 2,78 3,31
Reynolds Number, Re 13836 27672 41508 58111 69179
Wall temperature[°C], Tw 11,9 12,7 13,5 14,4 15,0
Temperature driving force, ΔT+ 0,91 0,87 0,83 0,78 0,75

Table 3: Constant values for varying rate experiments. Ref. Hoffmann and Amundsen (2010).
Constant rate parameters Value
Condensate temperature, Tcond 20 °C
Oil density, ρ 809 kg/m3
Viscosity, μ 2,00E-03 Pas
WAT (Cloud point temperature),Tc 30 °C

Table 4: Specific values for varying temperature experiments. Wall temperature and viscosity
from Hoffman and Amundsen (2010).
Condensate temperature [°C], Tcond 15 20 30 40
Wall temperature[°C], Tw 12,1 14,4 19,4 24,7
Temperature driving force, ΔT+ 1,19 0,78 0,35 0,13
Viscosity [Pa s], μ 3,00E-03 2,00E-03 1,00E-03 9,00E-04
Reynolds Number, Re 38740 58111 116221 129135

Table 5: Constant values for varying temperature experiments. Ref. Hoffmann and Amundsen
(2010).
Constant temperatur parameters Value
Rate of condensate, Q 21 m3/h
Oil density, ρ 809 kg/m3
WAT (Cloud point temperature),Tc 30 °C

17
 Table 6: Model constants for varying rate experiments.
Rate of condensate [m3/h], Q 5 10 15 21 25
k1 0,21 0,45 0,23 0,29 0,07
k2 exponential model 0,22 0,50 0,34 0,48 0,18
k2 logarithmic model 40 210 580 1050 2000

Table 7: Model constants for varying temperature experimenst.

Condensate temperature [°C], Tcond 15 20 30 40
k1 0,07 0,29 0,03 0,01
k2 exponential model 0,10 0,48 0,06 0,03
k2 logarithmic model 80 1050 360 1000

18
Figures

Figure 1: Wax thickness vs. time for digitized flow rate experiments.

Figure 2: Wax thickness vs. time for digitized temperature experiments.

19
Figure 3: Possible deposition buildup trends.

Figure 4: Possible rate of deposition trends.

20
Figure 5: Matching models to experimental data.

Figure 6: Matching models to experimental data.

21
Figure 7: Matching models to experimental data.

Figure 8: Matching models to experimental data.

22
Figure 9: Matching models to experimental data.

Figure 10: Matching models to experimental data.

23
Figure 11: Matching models to experimental data.

Figure 12: Matching models to experimental data.

24
Figure 13: Constant k1 vs. 1/Re for varying flow rate series.

Figure 14: k2,log vs. Re2 for varying flow rate.

25
Figure 15: k1 vs. ΔT+ for varying temperature series.

Figure 16: k2,log vs. 1/ΔT+ for varying temperature series.

26
Figure 17: k1 vs. ΔT+/Re for varying flow rate series.

Figure 18: Successful digitizing of experiment (yellow and black line).

27
Figure 19: Experiment hard to digitize with one line (yellow and black line).

28
Appendix A: Digitizing graphical data
As mentioned before the experimental data was presented as graphs in Rosvold’s (2008) MSc
thesis. For me to utilize these data properly I needed to digitize them. To do this I used a
software called Didger (ver. 2). Here is a short presentation on how this was done.

1. The thesis exists as a PDF file and plots from this file was saved as a picture file. The
picture was then loaded into the Didger software.
2. First four points on each graph was manually selected as accurate as possible (using
zoom). The coordinates for the selected point was typed in. For simplicity I used the
four points (0, 0), (max x, 0), (0, max y) and (max x, max y). The value of max x and
max y varied from graph to graph.
3. Using zoom the entire graph was identified as accurate as possible. Each graph was
identified with 80 – 140 points. Any graphs with a big spread were digitized as good
as possible by staying in the middle.
4. And output file was saved as a data file and opened in Microsoft Excel. This data file
contained the x and y coordinates to all digitized points on a graph. The data was
plotted using Excel and the plot was compared to the original from the thesis.

In total 8 plots were digitized. A very good match with 7 graphs was accomplished. One of
the graphs (Q = 5 m3/h) had a big spread. The only thing I could do was to make a sort of an
average line. A graph digitized accurately can be seen in Figure 18 and the difficult case can
be seen in Figure 19.

In order to use this digitized data I had no other option than to trust the digitizing software.
But the visual comparison between my plots and plots from the thesis was good.

29
Appendix B: Derivation of the exponential model:

Rewrite

Integrate

∫ ∫

Fill in

[ ]

Rewrite

Rewrite

Rewrite

Use exponential function

Rewrite

Rewrite

Rewrite

[ ]
Rewrite

[ ]
Rewrite

[ ]
The solution becomes

Considering both x and dx/dt the following is obtained

30
Appendix C: Derivation of the logarithmic model:

( )

Rewrite

Integrate

∫ ∫

Fill in

[ ]

Rewrite

Rewrite

Rewrite

Use natural logarithm (ln)

Rewrite

The solution becomes

Initial rate is given by

31
Appendix D: Estimating k1 and k2 in the models
This is a brief introduction to how I estimated the values for parameters k1 and k2 in each
model. Both models have one parameter in common. The k1 is the same for both models as
shown in Appendix B and C.

Estimating k1
Initial wax buildup is determined by k1. It can be identified by drawing a tangent line to the
start of a graph. I used Microsoft Excel and plotted the first 4-8 points as wax thickness vs.
time. Then I used the trend line function through the origin and got an equation for the line.
This equation was a linear function like y = ax. Then a [mm/h] is the k1 value we are looking
for. To ensure a good match the tangent was plotted against the entire graph. All tangents
found had a good match to the experimental data (digitized data).

Estimating k2

Exponential model
Values for k2 in each model was found using tables from digitizing and values for k1. The
exponential function derived in Appendix B was used to estimate k2.

This function or equation is not solvable for k2 in a regular way. I used the solver function in
Excel to help me with this. Paramter k2 was set equal to 1 and wax thickness (xcalc) was
calculated using time (t) and already estimated values of k1. The solver function was then
used in order to find the k2 that would give xcalc = xreal.

In order to match k2 with all points in the experimental table this method was applied to all
points. A macro was used so that Excel would perform this procedure automatically on all
points. This produced a different value for k2 in each point. A small range of values was
selected from this table and used to make a wax thickness plot. This plot was then compared
to the experimental plot. The k2 value with the best fit to the experimental plot was selected as
the best match.

32
Logarithmic model
The logarithmic k2 was estimated using the same procedure as for the exponential model. The
only thing that was different in this case was the equation for wax thickness as a function of
time.

Other than that the procedure is exactly the same. A simple try and fail method from the start
would probably produce the same result. But in order to save time I choose the procedure
described here.

33