SPE 84893
A New Method to Characterize the Size and Shape Dynamics of Asphaltene
Deposition Process in CO2 Miscible Flooding
Douglas Fisher, SPE, Alberta Research Council; Nintoku Yazawa, SPE, Japan National Oil Corporation; Hemanta
Sarma, SPE, The University of Adelaide; Marcel Girard, Alberta Research Council; Alex Turta, SPE, Alberta Research
Council; Haibo Huang, Alberta Research Council
Copyright 2003, Society of Petroleum Engineers Inc.
This paper was prepared for presentation at the SPE International Improved Oil Recovery
Conference in Asia Pacific held in Kuala Lumpur, Malaysia, 20–21 October 2003.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300
words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836 U.S.A., fax 01-972-952-9435.
Abstract
The use of miscible solvents in Improved Oil Recovery (IOR)
processes requires knowledge of how the solvent will behave
over all mixing ratios of oil and solvent. In particular it is
important to know: (i) at what solvent concentration the
asphaltenes start to precipitate, and (ii) what conditions will
cause the particles to flocculate and eventually deposit in the
reservoir pore network. In order to study this process we have
developed a new method of describing the process of
asphaltene precipitation/flocculation/deposition under typical
reservoir conditions.
This method uses image analysis to convert the information
obtained from a typical sequence of images obtained from a
dynamic high pressure mixing system acquired using a Micro-
Visual cell and special optics. These images are analyzed and
a single number per image is computed. This number is
defined as the “Particle Growth Factor” (PGF), which is a
compound number that includes information about the size of
the particles, the number of particles and their shape.
This new approach shows how PGF increases with respect to
solvent concentration. If we use a suitable non-linear model to
fit PGF with respect to concentration, we can use the
characteristics of the function to predict the onset of
precipitation, the point of maximum flocculation, and when
deposition is at its maximum. We can then produce a least
squares correlation, which relates the concentration of
asphaltene precipitation onset to both pressure and
temperature, using a 3D polynomial surface.
In the field, this information can then be used to predict
production/injection performance and different operational
problems related to asphaltene deposition in CO2 miscible
injection and also CO2 sequestration process in a depleted oil
reservoir.
Introduction
Understanding the phenomena related to asphaltene
precipitation, flocculation and deposition is of significant
interest in the application of miscible CO2 flooding. Our
previous studies have shown that with proper equipment it is
possible to study asphaltene/solvent behavior under reservoir
conditions, for a range of crude oils, without resorting to
dilution of the oil with an aromatic solvent such as toluene1.
Although the concept of spectrophotometric methods as a tool
to study precipitation has been used widely2, the use of the
time to frequency conversion in “time of flight” measurements
to obtain a much richer range of information from has not
been reported in such experiments. Time to frequency
conversion gives detailed information about the changes in
particle size, as the solvent/oil ratio changes. This class of size
information is very sensitive to changes in the distributions
found in the flow stream, and as such can be of great use in
studying the flocculation process as a time dependant variable.
A somewhat less common method is to use a Micro-Visual
cell and acquire images as the solvent/oil ratio changes2. These
images contain information about the total light transmitted
(just as the spectrophotometer does) and the fraction of the
cell area occupied by the solid particles. If the image is
sampled properly, we can convert the spatial information
contained in the image, with respect to the solid particles, into
frequency domain information in the same fashion as the
spectrophotometric time/frequency transformation. The net
result of this transformation is a 2D frequency domain image,
which gives detailed information about the changes in particle
size in space and time.
This paper describes the application of frequency domain
imaging to study the asphaltene precipitation / flocculation /
deposition process for a Japan National Oil Corporation
2 SPE 84893
(JNOC) CO2 miscible flood. The objectives of this
project were:
To determine the onset of asphaltene precipitation,
flocculation and deposition outside porous media
(bulk flow)
To assess CO2 miscibility in the region around 20
MPa (2900 psi)
To investigate the possible appearance of a second
liquid phase
To develop a method which can determine the
maximum amount of information from the dynamic
mixing of solvent and oil, and do this in such a way
as to be fast and efficient
Experimental
In this work, we investigate the changes in asphaltene particle
size as the solvent/oil ratio changed during bulk oil-CO2
mixture flow.
The JNOC oil/solvent system was first tested with pentane as
the solvent, for the first set of validation experiments, since
pentane, is a well-known solvent for the precipitation of
asphaltenes. These measurements helped provide a baseline
case to better understand the behavior of the JNOC oil
with CO2.
Equipment
The schematic of the apparatus is shown in Figure 1. The
valves, bypass loops, transfer vessels and heating systems,
needed for operational systems are not shown.
The Micro-Visual cell and the spectrophotometric cuvette
have one major feature in common; the windows of both cells
can be placed in close proximity of each other. The minimum
inter-window distance for both of these optical cells is in the
order of 100 to 200 microns. Small path length optical systems
are required when the system is to be used with medium to
heavy crude oils, since these oils are very dark and transmit
very little light.
A second important point in the operation of these cells is the
need for high intensity light sources. The design of the light
source for the spectrophotometer is a case in point. Here we
have used what is termed a non-imaging concentrator to direct
light from a “300 watt” quartz halogen reflector bulb directly
into the window aperture. This method supplies light at
intensities much greater than systems based on lenses.
It was necessary to record the transmitted light from the
spectrophotometer in digital form. This was done using a
computer and an analog to digital converter. The
spectrophotometer data acquisition system is 16 bits, in order
to achieve the required resolution in the frequency domain.
The power line frequency (50/60 Hz) represents one of several
interferences to this class of measurement and must therefore
be kept to a minimum. To achieve this we used a high voltage
DC power supply to run the light source. In order to facilitate
the time of flight studies of particle size, the computer must
digitize the photometric information at sample rates in the
order of 256 Hz, or higher.
The Micro-Visual cell is monitored using a video camera and
a 10bit video digitizer. This process requires a special
software, which allows the user to acquire images at variable
rates, and store the image files on the computer hard drive in a
format that allows the time at which the image was captured to
be stored in the image header. In order to achieve this goal we
developed a special image file format that stores the capture
time and other information in each image file.
A typical experiment will acquire images every 16 seconds
over the range of solvent to oil ratios needed for the
particular study.
Experimental Procedure
The solvent/oil ratio is controlled by using two pumps which
co-inject both oil and solvent into a static mixer, at system
pressure. The total flow rate is kept to 1000 micro litres a
minute, with flow ratio changes in steps of 50 to 100 µl/min
for approximately 7 min per step. This results in an almost
continuous concentration change at the detector with respect
to time. The CO2-JNOC oil experiments this value was 50
µl/min steps in 5.0 min increments.
Numerical Processing – “Time of Flight”
Figure 2 shows the essential elements of the “Time of Flight”
particle size characterization method, where the windows
represent the high-pressure spectrophotometric cuvette.
The basic principle involved is that as a particle flows through
the light path it changes the amount of light measured at the
detector in two ways. First, the size of the particle decreases
the amount of transmitted light; second, the time that a particle
takes to transit the window affects the size of the depression in
light transmittance.
If we couple these two effects and monitor the output signal
with respect to time (at a high sample rate), the resulting data
set can be converted from the time domain into the frequency
domain by using the Fast Fourier Transform (FFT) or related
methods. The transformed data can be displayed as a 2D
image (Figure 3) where the frequency is on one axis and
solvent volume fraction on the other (time and volume fraction
are directly related to each other), with the power density as
the intensity of the pixels in the image.
At this point we need to expand on the use of the terms
frequency and power. For the spectrophotometer data we have
a data stream, which is measured with respect to time,
therefore the frequency is in reciprocal seconds, or Hertz. The
power in this case is the distribution of transmitted light
SPE 84893
energy at a given frequency, since we are measuring this
variation through a transducer this connection is uncalibrated
and expressed in relative terms, note: it is not normalized.
With calibration using standard particles at a constant flow
rate, such an image gives detailed information about the
particle size in the flow stream in real time. If the flow rate is
not constant then calibration is not possible, but the
measurement still retains useful information about the
dynamic changes in the distribution of solid particles,
assuming that the changes in flow rate are not too large.
Numerical Processing – Micro-Visual Cell
The diagram shown in Figure 4 represents the Micro-Visual
cell in the alignment used in this work. Figure 5 shows a
typical image acquired when a large number of asphaltene
particles are present.
The first step in manipulating the image data is to determine
the range of intensity in an image using a histogram of the
intensity of every pixel in the image. Since the range of
intensities that can be represented by the video system is
expressed in terms of the number of bits the video digitizer
uses. We have a range of values from 0 to 255 for 8 bits, and 0
to 1023 for 10 bits. This is a rather narrow range compared to
normal data acquisition systems for voltages and temperatures,
which generally use 12 bits or 16 bits. Although 8 bits is not
as good as 10 bits we can extract useful information from both
systems. It is only when we deal with heavy oils that the 10 bit
system is a requirement (heavy oil systems in the order of
50,000 cp have been measured). This means that the total
range of intensity for the fluids must fall in this range for
100% oil (close to zero transmittance) to 100% solvent (100%
transmittance). Figure 6 shows a set of images optimized to
cover the whole range while not clipping the minimum or
maximum intensity for any pixel in the area of interest (i.e. the
window area).
In order to study the changes in images over the whole
concentration range we frequently must subtract the first
image from all of the other images in the sequence. This
results in images with negative pixel values in some of the
steps. The need for negative numbers means that we cannot
use an unsigned integer representation during processing. The
solution to this is to use floating point numbers at this point,
and finally convert the results to 8 bit images for presentation.
Other calculations described as described in the following
require the use of floating point numbers as well.
A typical image sequence will be further manipulated to
stretch the dynamic range of the image so that the particles
that appear black (low light intensity) will have the maximum
contrast between them and the surrounding solvent/oil
mixture. This is achieved by computing the histogram for each
image, and then converting it to its cumulative form by
integration. This cumulative curve is then normalized, and the
intensities associated with the 0.1% and 99.9% points are
found. With the knowledge of these intensities all pixels above
3
or below these thresholds are set to these values and the
resulting image is renormalized. This gives a much more
dynamic contrast range for the early precipitation part of
the sequence.
The Micro-Visual cell also contains information relating to the
size of the particles, but since these particles move from image
to image and do not in general produce pixels of zero
intensity, the traditional method of binary segmentation is
inappropriate for these images. The question is what then can
we do to extract information about the changes in size the
particles are undergoing from image to image.
The solution is to use a method that can be best referred to as
“Space to Frequency Domain Transformation”. This method
involves the concept that in a 1D3 case the variation in the
light intensity across the cell can represent the variation in
width of objects encountered by the 1D sample line (Figure 7).
We will use the 1D case for simplicity for most of the
examples, whereas the bulk of the reported measurements
were computed the 2D FFT method.
When we discuss the use of frequency domain analysis for
spatially encoded data, such as images, we need to reiterate
the definition of frequency, and power. Frequency in this
context is in reciprocal centimeters (or any other distance
unit). The power term is again the variation of light
transmittance for each defined spatial frequency.
This representation when applied to image data, gives a
distribution of the spatial frequencies found in the image, and
the magnitude of power for the variation of transmittance.
Since the raw image data cannot be defined by any analytical
function, we must use a discrete numerical method such as the
Fast Fourier Transform (FFT) to convert the spatial
information into the frequency domain. In Figure 8 we have
the 1D data from the previous image Figure 7. It is obvious
that there is noticeable variation in the signal at both the low
and high frequency part of the power spectrum (Figure 9).
We know that in this case the low frequency variation is due
to inhomogeneties in the light source. We can remove this
influence by subtracting the locally weighted average value
(One of the locally weighted least squares methods) from each
point in the set. Once this has been done, we can then compute
the power spectrum using a suitable method such as the FFT.
The FFT method requires that the data set be a power of 2 in
size (256, 512 etc). Since we cannot always ensure this
condition for 1D samples (i.e. NTSC images of 640x480
aspect ratio), we chose the Maximum Entropy Method to
compute the power spectra for these samples (1D only). The
result is shown in Figure 9.
Figures 10 through 12 show a sample taken where some
asphaltene particles are present. Here one can see that there
are small but noticeable drops in the raw 1D signal along the
line, but that they are not very large in magnitude. The
resulting power spectrum in Figure 12 shows noticeable
4 SPE 84893
changes in the distribution of frequencies as compared to
Figure 9.
We repeat the process in Figures 13 through 15 but here the
asphaltene particles are both large and very dark. This
produces major drops in signal strength along the sample line
in Figure 14 and major amounts of power distributed in the
low frequency zone of the spectrum (Figure 15).
The final set of data - Figures 16 through 17 - represent the
cell when it has been cleaned using toluene. Here the power
spectrum in Figure 17 has a different character than that found
for 100% oil in Figure 9. The explanation for this observation
is that we are observing solids left on the windows, which is
only visible when there is highly transparent fluid. We must
remember that we have been subtracting the local average
from each point, so the effect of the large signal strength in the
clear cell has been removed. Thus, we see only the variation
due to the solids left behind, which cannot be detected when
the cell is full of oil.
In Figure 18 we have taken the power spectra from all the
images, and created a 2D image where the x-axis represents
time/image number (or concentration for that matter) and the
y-axis represents frequency in reciprocal cm. The z-axis or
intensity represents the power, or in effect, the amount of size
information at each image/time step and at each size expressed
in reciprocal cm. In addition images from the difference series,
which are representative of each type of process, have been
added to the frequency image, in order to help emphasize the
different characteristics the particles exhibit as the solvent/oil
ratio changes.
Since these example images are made up of 1D samples from
a given image, it is important to know that the statistical
nature of such a sample represents a reasonable representation
of the whole. To solve this problem we either used all of the
horizontal lines, or a previously mentioned use the 2D FFT. If
we use the 2D FFT, we must insure that the dimensions of the
image are a power of 2 in each direction. We can do this in
two different ways. The first method is to sample a square
section of the image, which is the largest square, which can be
extracted from a circular window, or to use 2D linear
interpolation and convert the rectangular aspect ratio (i.e.
640x480) to a 512x512 image. Once this has been done the 2D
FFT can be computed. Once we have the sequence of images
converted to power spectra images, we then compute the total
power integral found over the frequency range of choice (this
range depends on the range of sizes one wishes to examine).
The images contain additional information about the
distribution of the solids versus the liquid part of the image.
This information can be extracted from the histogram
calculated from the difference image series. The process used
to extract this information is called “Histogram
Deconvolution”. This method uses non-linear Least Squares
to fit a set of Gaussian curves to the histogram. The
parameters from this set of equations can be used to extract
information about the separate parts of the image. In Figures
19 through 21 we show three separate histograms and the
image they represent. The two Gaussian curves are plotted
under each histogram. The parameter of interest here is the
mode or the position of the peak maxima. If we plot the mode
value on the “x” axis for all images, we get two curves, one
representing the modal intensity for the liquids, and the other
for the solids.
These curves give different information about the process. In
Figure 22, these curves are plotted as an image where the x-
axis represents the solvent concentration, the y-axis represents
the histogram intensity and the z-axis represents
the frequency.
The liquid curve is most closely related to the total average
light transmitted from the cell but it has had the effects of the
solids removed. This means it is more related to dilution with
the obvious caveat that as the asphaltenes are removed the
amount of light transmitted goes up faster than by Beer’s law
dilution alone.
The solids curve tells two stories; first, when the precipitation
phase is just starting, the particles are small and do not
produce something big enough to fill the space between the
two glass windows (ca. 200 µm). As such, the intensity
associated with these particles is higher than when the
particles are larger or are adhering to the glass (as in
deposition). Therefore the solids modal value must decrease as
the asphaltenes become bigger and darker. This method allows
us to identify these changes.
The second liquid phase detection
The presence of the second dense liquid phase can be seen as
the appearance of a significant number of pixels at maximum
intensity; these pixels represent the solvent bubbles that
transmit light efficiently. We have done experiments on a
wide range of oil samples, with solvents such as C3, C4, CO2
and pentane. In all of the cases where a second liquid phase
appears when the solvent is no longer totally miscible in the
oil the asphaltene precipitation onset has occurred at a solvent
concentration lower than the appearance of the second liquid
phase. In general we have observed that it is the change in
properties of the oil created by the loss of the asphaltenes that
changes the solvent oil system from completely miscible to
one in which the solvent concentration in the oil has reached
saturation. An additional observation is that the solvent can
under these conditions extract light ends from the oil and
therefore causing the viscosity of the oil to
increase substantially.
In the reservoir case it is important to note that this type of
complex phase behavior is hard to depict accurately using
phase behavior simulators alone. The existence of these
transitions and the conditions at which they occur in
quantative terms can reduce the uncertainty of the design and
operation of miscible systems in the field.
SPE 84893 5

Results of Tests with CO2 – JNOC Crude Oil System clear CO2 phase was first detected) boundary conditions were
determined using the image data, where it was possible to
We now describe the results obtained with CO2 as a solvent, at identify the point at which two phases were present to the
the following pressure and temperature conditions (Table 1). nearest image number (i.e. time).

Table 1: Pressure and temperature conditions for the CO2- Table 2: Onset of points of asphaltene precipitation and two-
JNOC crude oil system tested phase formation, expressed as CO2 flow fractions
40°C 60°C 80°C 98°C
TWO-PHASE
14.5 MPa TEMPERATURE
0
C
PRESSURE
MPA
CO2
COMPRESSIBILITY
CO2
DENSITY
ONSET OF
PRECIPITATION
ONSET
CO2 FLOW
GM/CC CO2 FLOW FRACTION
15.8 MPa 15.8 MPa FRACTION

17.2 MPa 17.2 MPa 17.2 MPa 17.2 MPa 98.0 17.2 0.5954 0.4076 0.324 0.375
18.6 MPa 18.6 MPa
98.0 18.6 0.5838 0.4512 0.270 0.395
20.0 MPa
98.0 20.0 0.5783 0.4949 0.274 0.443
The temperature of 98.3 0C represents the reservoir
temperature, while the pressure of 20 MPa (2900 psi) is the 80.0 15.8 0.5138 0.4614 0.232 0.395
minimum miscibility pressure (MMP) for CO2 with the live oil 80.0 17.2 0.5020 0.5176 0.291 0.444
at reservoir temperature. The pressure and temperature
80.0* 17.2 0.5020 0.5176 0.343 0.480
conditions in the Table 1 were chosen such that they represent
the prevailing conditions at the CO2 displacement front, 80.0 18.6 0.4992 0.5561 0.276 0.476
around the production well and in the borehole, over a few
hundred meters, when the oil-solvent mixture flow upward 60.0 17.2 0.4084 0.6780 0.295 0.488
vertically in the production tubing. In the production tubing, 60.0 14.5 0.3990 0.5864 0.254 0.450
during the natural flow, both temperature and pressure
decrease. An additional test at 40 0C was carried out to 60.0 15.8 0.4000 0.6305 0.280 0.473
consolidate our conclusions. It is generally accepted that when
the temperature decreases, there is a higher possibility of the 40.0 17.2 0.3575 0.8100 0.266 0.494
appearance of a second liquid phase. We made an attempt to
check this, as well. * Repeatability test

The JNOC crude oil used in tests has an asphaltene content of
3.0 wt %. In all tests, dead oil of a viscosity of 11.8 mPas at
25 0C and 1000 psi was used. The viscosity decreases to 9.23
mPa.s at 40 0C and 3000 psi, and at 1.89 mPa.s at 100 0C and
3000 psi. The CO2 MMP is 2900 psi (20 MPa) at 98.3 0C for
the live crude oil having a bubble point pressure of 1000 psi
and a solution gas-oil ratio of 300 scf / std. bbl.
Both the data obtained from the spectrophotometer and the
Micro-Visual cell was used to determine the concentration of
CO2 at which the precipitation of asphaltenes starts.
The onset of precipitation was determined by integrating the
power spectrum with respect to frequency for each point on
the time axis. Once we obtained the integral for each image,
the portion representing the range from onset to deposition
was fitted using a non-linear least squares model, based on the
Gaussian integral (a algebraic sum of two or more scaled error
functions). After normalizing the fitted curve, it was possible
to find the point on the time axis that corresponds to the region
on the asphaltene image represented by the 5% value above
the baseline.
In Table 2, we show the flowing fraction of CO2
corresponding to the onset points for asphaltene precipitation
as well as for the formation of a second phase (CO2 rich
phase). It should be noted here that the flow rate was changed
in steps every 5 minutes, which results in an error in the
concentration of ±2.5%. The two phase (the point where a
The flowing fraction data presented in Table 2 is the volume
fraction of CO2 for a total flow rat e of 1.0 ml/min. This value
can be converted to grams of CO2, by using the density of CO2
at each pressure and temperature, over the
experimental region.
The results shown in Table 2 indicate that at lower
temperatures, for the same pressure, a lower volumetric
fraction of CO2 is needed to reach the onset of asphaltene
precipitation. At the same time, at lower temperatures, for the
same pressure, a higher volumetric fraction of CO2 is needed
to reach the two-phase formation onset point (the appearance
of CO2 rich phase). Moreover, the fact that the appearance of
the second phase occurred at 98 0C and 20 MPa, as well,
indicates that at these conditions first contact miscibility does
not exist, when the volumetric CO2 fractions are higher than
44%. This statement should also be considered in the context
of a “vaporizing” mechanism for the dynamic miscible
(multiple contact) displacement with CO2 at this high
temperature (98 0C).
The multiple contact miscibility condition is more complex
than first contact miscibility and requires intensive mass
transfer of light hydrocarbons from the oil to the CO2.
Additional studies will be necessary in order to better
understand this process.
Our experimental data supports the observation that
flocculation is a continuous process and therefore it is difficult
6 SPE 84893

to determine a threshold for flocculation. A threshold for Conclusions

deposition can be determined from this data set with
additional computation.
In Table 3 the final concentrations of CO2, for both the onset
of precipitation and the boundary of the two-phase region, in
moles/L are shown. The boundary of the two-phase region can
be regarded as the bubble point pressure (saturation pressure)
envelope in a Pressure-Temperature diagram; which is the
phase equilibrium diagram of the CO2-crude oil in a
P-T diagram.
1. The models used here are based on the traditional
chemical description of how solids precipitate from
solutions, and then flocculate into agglomerates large
enough to settle out of the liquid and deposit on the
rock surface. This new method allows the
determination of the onset of asphaltene precipitation
to be determined quantitatively, while semi-
quantitative information on flocculation rate and
deposition conditions cal are defined.

A correlation surface for each parameter presented in Table 3 2. The apparatus developed in this study also permits
can be constructed using a linear least squares polynomial the determination of the two-phase envelope in a
surface. Both sets of data were successfully fitted with Pressure-Temperature system.
pressure and temperature as the independent variables, using a
polynomial of order 2, with one cross term of order 1. The 3. A new method – “The Determination of the Onset of
resulting surface for the onset of precipitation is shown in Asphaltene Precipitation by Frequency Domain
Figure 23, as a contour graph showing the magnitude of Analysis”– was developed based on a very detailed
the concentrations. computational approach to the analysis of
experimental data. This method allows changes in
The two-phase envelope and saturation pressure envelope are pseudo-continuous solvent/oil ratios, and it can
shown in Figure 24. Both data sets have a region where, due to generate one set of data in approximately one to two
the lack of experimental data points, the correlation surface hours, discounting the time required to do the
cannot be relied on to extrapolate correctly. This region has analysis (which can be done off-line). This
been set to black in the images and in the contours a lower methodology allows the study of multiple “pressure
limit of 2.0 Moles/L has been used to define the lower limit. and temperature” steps in a relatively short
The correlation coefficient for the onset of precipitation is time frame.
0.928 and for the two-phase system is 0.99. Such a high
correlation coefficient for both systems helps us overcome the 4. The experimental data contains additional
2.5% error for individual measurements. information about the rate of flocculation and the
onset of deposition. Further calculations based on the
In Figures 25 and 26 the correlation surfaces have been model used to describe the relationship between PGF
converted into false color maps of the onset points with and solvent / oil fraction, are required to achieve
respect to temperature and pressure, where the color of the this end.
map represents the amount of CO2 required to produce
each effect. Acknowledgement
We thank JNOC for the permission to present this paper.
Table 3: Onset points of asphaltene precipitation and two- We thank Kevin Ripsler for his work in operating
phase formation, expressed as CO2 Moles/Litre the equipment.
We would like to thank Steve Sim, Leslie Randall and Blaine
ONSET OF TWO PHASE Hawkins for their comments and review of this paper.
TEMPERATURE PRESSURE
0 PRECIPITATION ENVELOPE
C MPA
MOLE/L MOLE/L
References
98.0 17.2 3.00 3.47
1
98.0 18.6 2.77 4.05 A. Turta, D. Fisher, A.K. Singal, J. Najman, “Variation of Oil-
98.0 20.0 3.08 4.98 solvent Mixture Viscosity in Relation to the Onset of Asphaltene
Flocculation and Deposition”, Special Edition Journal of
80.0 15.8 2.43 4.14 Canadian Petroleum Technology 38, No. 13 Paper: 97-81, 1999
80.0 17.2 3.42 5.22
2
80.0* 17.2 4.03 5.65 Jamaluddin, A.K.M, Creek, J., Kabir, C.S., McFadden, J.D.,
D’Cruz, D., Mankalathil, J., Joshi, N., Ross, B., “Laboratory
80.0 18.6 3.49 6.02
Techniques to Measure Thermodynamic Asphaltene
Instability”, JCPT, July 2002, Vol. 41.
60.0 17.2 4.55 7.52
60.0 14.5 3.39 6.00 3
When we refer to 1D and 2D we use the convention that 1D
60.0 15.8 4.01 6.78 represents a row or column from the image matrix, or
transmittance with respect to time, whereas 2D refers to an
40.0 17.2 4.90 9.09 image matrix.
* Repeatability test
SPE 84893 7

Light Source
Asphaltenes
Pump 1 Micro Visual Cell
Mixer

Detector

Pump 2

Spectrophotometer BPR

Capillary Windows
Viscometer

Flow Direction Computer

Data Acquisition

Figure 1: Schematic of Asphaltene Apparatus Figure 2: Schematic of Time of Flight

Spectrophotometer
20.7 MPa 60 °C
Digital Image High Pressure Housing
Concentration Step Changes Spacer
Acquisition
High

Light Source
Frequency as 1/d

Power

Second Dense Liquid Phase

Flocculation

Precipitation

Video
Low

Camera
Solvent Ratio
Glass Windows
Figure 3: Time to Frequency Image
Figure 4: Schematic of Micro-Visual Cell

Early
Oil

Late Toluene

Figure 5:Typical Micro-Visual Cell Image Figure 6: Set of Micro-Visual Images in which the
intensity range has been optimized
 8 SPE 84893

0.32
Position of Sample Line 0.30
0.28
100% Oil 0.26
0.24

Intensity
0.22
0.20
0.18
0.16
0.14
0.12
0.10
100 200 300 400 500 600

Horizontal Position (Pixels)

Figure 9: 1D Power Spectrum

Figure 7: 1D Sample location

0.0020
Position of Sample Line

0.0015 Some Particles

Power

0.0010

0.0005

0.0000
0.0 0.1 0.2 0.3 0.4 0.5

Frequency (1/pixels)

Figure 9: 1D Power Spectrum

Figure 10: Image with some particles
Figure 9: 1D Power Spectrum
Figure 10: Image with some particles
0.0020

0.40

0.35 0.0015
Intensity

0.30
Power

0.0010
0.25

0.20
0.0005

100 200 300 400 500 600

0.0000
Horizontal Position (Pixels)
0.0 0.1 0.2 0.3 0.4 0.5

Frequency (1/pixels)
Figure 11: 1D Sample line plot
Figure 12: 1D Power Spectrum
SPE 84893 9

Photometric Transmittance at a Given Vertical Point

0.7

Position of Sample Line

0.6
Large Particles
0.5

Intensity
0.4

0.3

0.2

0.1
100 200 300 400 500 600

Horizontal Position (Pixels)

Figure 13: 1D Sample Line Figure 14: 1D Sample Line Plot

1D Power Spectrum of a Sample Line

0.0020 Photometric Transmittance at a Given Vertical Point

0.0015
1.0
Power

0.9
0.0010
Intensity

0.8
0.0005

0.7

0.0000
0.0 0.1 0.2 0.3 0.4 0.5 0.6

Frequency (1/pixels) 100 200 300 400 500 600

Horizontal Position (Pixels)

Figure 15: 1D Power Spectrum Figure 16: 1D Toluene Plot

1D Power Spectrum of a Sample Line

0.0020

0.0015
Power

0.0010

0.0005

0.0000
0.0 0.1 0.2 0.3 0.4 0.5

Frequency (1/pixels)

Figure 17: 1D Toluene Power Spectrum

10 SPE 84893

Frequency Æ

Time →
Figure 18: 1D to 2D image from 1D power spectra samples

Image Number 1 Image Number 50

24000
44000

Liquid
18000
33000 Liquid
Frequency

Frequency

12000
22000

Solids Solids

11000 6000

0 0
80 85 90 95 100 105 60 70 80 90 100 110 120

Pixel Intensity (0-255) Pixel Intensity (0-255)

Figure 19: Image Histogram 100% Oil Figure 20: Image Histogram, some solids
SPE 84893 11

20.7 MPa 60 °C
Im age Num ber 100
12000

High
9000 Liquid

Pixel Intensity

Histogram Frequency
Frequency

6000

Solids
3000

0
50 70 90 110 130 150 170

Low
Pixel Intensity (0-255)
Solvent Ratio

Figure 21: Image Histogram, some solids Figure 22: Histogram Image

Figure 23: Precipitation Onset Envelope Figure 24: Second Phase Onset
(Contours in Moles CO2) Envelope(Contours in Moles CO2)

Onset of Asphaltene Precipitation Onset of Second Dense Liquid Phase

18.6 20.0 20.7 22.1 22.7 24.1 24.8 27.6
18.6 20.0 20.7 22.1 22.7 24.1 24.8 27.6

High
High

Pressure MPa
Pressure MPa

Moles CO2
Moles CO2

17.2
17.2

15.8
15.8

14.5

Low
14.5

Low

40°C 60°C 80°C 98°C

40°C 60°C 80°C 98°C

Temperature °C
Temperature °C

Figure 25: Precipitation Onset Envelope Figure 26: Second Phase Onset Envelope