Keywords truth for reservoir characterization challenges. We found



Rock Physics, Porosity, Segmentation, Optimization

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!

order now




Core samples
have always been a luxury for measuring reservoir properties. However, in most
cases the cores become non-usable after a single experiment. Methods such as
Digital Rock Physics (DRP) based on image processing is offering an alternative
to model these reservoir properties with better control on subjective biases of
the experimentation and is non-destructive in nature. DRP involves imaging the
formation and simulating the field performance to account for various
non-homogeneity in the reservoir formation. Over some time now, it has become a
popular method but in case of complex reservoirs such as carbonates and
unconventional resources, it is still at the feasibility stage only. The
reasons are plenty ranging from availability of calibration libraries and
transition space error and its quantification. In this paper we used DRP to
obtain porosity in carbonate samples at various scales and compared the results
obtained using established laboratory methods which at the moment serves as
ground truth for reservoir characterization challenges. We found that DRP
results mostly align with the results obtained using methods like QEMSCAN. The
analysis mostly points to the resolution limits input to the respective




reservoirs account for most of the world’s oil and gas reserves and thus are
likely to dominate the hydrocarbon production through the next century (Akbar
et al., 1995). The extremity of carbonate reservoirs is such that these can be extremely
large while having pores which are microscopic in nature. This makes the
characterization of these heterogeneous reservoirs very complex and important
(Akbar et al., 2000, 2001).


In DRP, 3D pore volume is a favourite
output involving FIBSEM for deconstruction and COMSOL for meshing and
simulation (Brown, 2011).  As a result of flow
simulation, absolute permeability is another attribute that can be successfully
obtained from this exercise. But, obtaining a 3D scanned volume of the sample
is a costly method. So, 2D scanned images can be used to obtain a 3D volume of
the sample by conditional reconstruction process (Karimpouli & Tahmasebi, 2016) or empirical
relations (Karimpouli,
Khoshlesan, Saenger, & Koochi, 2017).
This reconstructed 3D volume is then used to determine the permeability of the
sample. Dvorkin (2009), attempt the DRP processes on a heterogeneous Berea
sandstone sample and investigate the variation on properties on various scales
by subdividing the digital volume into eight small cubes. A similar work with
overlapping volume in carbonates was attempted by Saenger et al. (2016). Grader et al.
(2010) compared the material property in carbonates in different facie types,
such as granular and vuggy DRP techniques have also been extended for elastic
property estimation (Zhang, Saxena, Barthelemy,
Marsh, Mavko, & Mukerji, 2011). A Fontainebleau sandstone
sample was used wherein the rock
compressibility was replaced by the elastic parameter of the pore filling
material as suggested by Ciz and Shapiro (2007). DRP workflow on drill cuttings
was dealt in detail by Dvorkin et al. (2003). The work
considered three shale cuttings photographed using thin sections and Scanning
Electron Microscope (SEM) to resolve the pore space. Kalam (2012) compared the Special
Core Analysis (SCAL) tests with the DRP measurements for the case of complex
carbonates obtained from Middle East reservoirs.


As evident with literature review, DRP
has been successfully attempted for in various forms for physical property
estimation but specific work addressing the optimization of algorithm to
address scaling issue is still missing. The objective of this work was to highlight
the issues of scales and resolution across homogenous to heterogeneous
carbonate facies. We used micro CT images of the two carbonate samples that had
same mineralogy but were distinctly different in fabric morphology. The
porosities were determined and compared with the respective values from
conventional methods and limitations were deliberated upon.




In order to
characterize a reservoir and thereby determine the petrophysical parameters,
various conventional methods are used. Conventionally there are 2 ways of
measurement, i.e. laboratory method based on Archimedes Principle and steady
state CMS-300 method (Core Measurement System) based on Boyle’s law. In both
these methods the porosity of the whole core sample, which was of the size
2-4in, was determined. But with the advent of technology we have been able to
look at the rock at a much higher scale of resolution which helps us to better
understand the pore network. This is specially required in the cases where rock
structure is heterogeneous, such as carbonates.


Digital Rock Physics is a non-invasive and non destructive method
wherein the core sample is at first scanned and then a digital image of the
rock (separating pore space from the mineral matrix) is constructed through a
process called segmentation. Different procedures are applied on the segmented
image/volume to visualize   its   internal structure (Andrae, et al., 2003)
and simulate the physical properties thereof. However, the approach of “image
and compute”, relies largely on the chosen magnification of image acquisition.
Image processing is then performed to correct the image for noise if any by
applying filters and then segmentation is done for segregating different
volumes of energy intensity, identifying them a possible mineral and pore
volumes. In the end, physical processes are simulated computationally to
calculate the effective porosity, permeability, elastic properties etc.
(Andrae, et al., 2003).


Segmentation is
a technique which delineates the pores from the solid matrix. In order to
perform this step a histogram is generated to view the pixel distribution in
the image and decide the threshold for segmenting the pores and grains, on the
basis of their corresponding intensity values. Each image element
carries an 8-bit signal (256 grey values) corresponding to the X-ray attenuation
experienced within its volume. A pixel completely filled by empty pore space data
approaches black (0), while a pixel completely filled by high density mineral approaches
white (255). Pixels completely filled by minerals with lower densities, or by
various proportions of minerals and/or pore space cover the entire grey scale
range from black to white (Kalam, 2012).Once the pores are identified the total
porosity of the sample can be determined.


In this paper, in the first step we
have worked on the SEM images of different resolutions, such as1mm, 1µm, 5µm,
20µm, 100µm and 500µm. Also, we have used the sliced images obtained from Micro
CT scan of the same Carbonate sample. Then we have used MATLAB code to perform
segmentation on these carbonate data and then determined the porosity of each
SEM image, the bulk porosity of the sliced images, i.e. combining all the
slices to give a single porosity and also the porosity from the sections
combining 50 slices each.


In the second step, the porosities obtained
in the previous step were then compared with the conventional methods. In doing
so, we determined the factors or issues that could be responsible for or reason
behind the difference in the porosity values.


MATLAB Algorithm


1. Convert the RGB
image file into a binary (B&W) image.

2. Define the
location of the first image file and the number of image slices you want to

3. Set the
threshold manually from the histogram showing the pixel distribution in the
image. (Make sure that the threshold is adjusted properly so that it detects
all the pores present in the images)

4. Create an
iterative loop such that the MATLAB reads all the image slices.

5. Add another
iterative loop to calculate the porosity. The porosity detected within each
image is given as:


Porosity = (Area of pores)* 100 ? (total
area of the image)

 Result and Discussion


We had
the porosity values of the entire core obtained from the conventional methods.

1.Laboratory measurements gave the porosity
value of 18%

2.CMS measurement gave the porosity value of

3.QEMSCAN measurement gave the porosity value
of 6.27%.The Segmentation
process performed on these SEM images (shown in Figure 1) and slicedMicro CT
scanned images (shown in Figure 2) of the carbonate sample using the
MATLABalgorithm resulted in the porosity valuesIn this
paper, it was found that the bulk porosity found from the sliced Micro CT
images resulted in the porosity of 6.30 %, the porosity obtained from sections
containing 50 Micro CT slices each resulted in an average porosity of 6.1%
whereas individual SEM images of different resolutions resulted in an average
porosity of 7.8 %. So, the calculated porosity lies in the range 6%-8%.This calculated porosity is
comparable to the QEMSCAN porosity because the Micro CT images are of the order
of 25 microns whereas QEMSCAN are of the order of 2-10 microns. Since, the
scales of these two measurements are almost similar, they resulted in
comparable porosities.


On the contrary, the lab and CMS
resulted in porosity higher than the calculated porosity i.e. in the case of
calculated porosity even the pores are being considered as grains. This could
be due to many reasons:


The Lab measurements consider the
entire core and are carried out at a lower resolution and thus have larger
field of view. Whereas, in DRP the measurements are done at higher resolution
and it looks at the core at micro or nano scale. So, in DRP a lot of space is
not sampled which the Lab measurements are able to do.



This paper applies the digital
image processing step of Digital Rock Physics method, i.e. segmentation
process. In this paper we have addressed the scaling issues that one encounters
when comparing the Digital Rock Physics method with the Conventional methods. The
scaling issues mentioned in this paper are subjective and can be overcome by
applying upscaling process, using optimization algorithm.