OTC-26508-MS

A New and Practical Model for Amount and Rate of Sand Production
Estimation
B. Wu, S. K. Choi, R. Denke, T. Barton, C. Viswanathan, and S. Lim, CSIRO; M. Zamberi, S. Shaffee,
N. Fadhlan, Z. Johar, M. B. Jadid, and B. B. Madon, PETRONAS

Copyright 2016, Offshore Technology Conference

This paper was prepared for presentation at the Offshore Technology Conference Asia held in Kuala Lumpur, Malaysia, 22–25 March 2016.

This paper was selected for presentation by an OTC program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Offshore Technology Conference and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Offshore Technology Conference, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the
written consent of the Offshore Technology Conference 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 OTC copyright.

Abstract
This paper presents a comprehensive study on sand production with the objective to develop a practical
sand production model to estimate amount and rate of sand produced. The study comprised three
components; laboratory experiments, numerical modeling and field case study for model validation. An
experiment program was carried out on seven weak sandstones with various strengths. The experiments
were conducted on relatively large samples under realistic effective reservoir stress and drawdown
conditions with oil and gas flow. The large sample size allowed a meaningful amount of sands to be
produced with minimum boundary effect. Real-time borehole monitoring using a borescope and X-ray CT
scanning on the tested samples enabled rock failure and sanding mechanisms to be investigated. The
laboratory experiments were simulated by numerical modeling using discrete element and finite difference
methods. A sand production quantification model was developed by correlating the critical plastic volume
obtained from the numerical modeling with the cumulative amount of sand produced from the sanding
experiments. The sand quantification model was applied to a field case study aiming for model validation.
The modeling result has been so far consistent with field observations.

Introduction
Sand production has been a major concern to the oil and gas industry for decades, and has become more
so nowadays as the industry is under increasing pressure for cost reduction. Downhole sand control
techniques are widely applied to keep the sands in the reservoir. However, these techniques increase
upfront CAPEX significantly to complete a well, and can make development of marginal fields uneco-
nomic. Furthermore they can increase operational complexity for well completion and hinder hydrocarbon
production. On the other hand, inadequate sand control may lead to unexpected large amount of sand
being produced. This increases difficulties in handling the sands topside, presents significant risk to well
and asset integrity, restricts production and can even lead to costly well or platform shut down.
Sand production is a consequence of fluid flow into a wellbore from a sanding-prone reservoir. The
process of sand production from reservoir to surface is very complex and may be divided into three stages,
i.e., loss of mechanical integrity due to degradation of the rocks surrounding a perforation or openhole,
separation of solid particles and clumps from the degraded sandstones due to hydrodynamic drag force of
2 OTC-26508-MS

fluid flow, and transportation of the sand particles to the surface by the produced fluids. The key process
leading to sand production in a weakly-consolidated and consolidated reservoir is the degradation of the
rock materials. This could be caused by a number of factors, such as drilling and perforation, cyclic effects
of well shut-in and open-up, reservoir pressure depletion, weakening of rock materials due to water-cut.
The drawdown induced fluid flow facilitates the separation of the solid particles and clumps from the
degraded zone in the reservoir formation.
To make an informed decision if and when to install sand control downhole or manage sand production
on surface, the ability to make a meaningful assessment of sand production risk and its severity prior to
completing the well becomes important. Over the decades, significant effort has been made to develop
robust and reliable sand production prediction models. As a result, there exists a large number of sand
production models ranging from empirical, analytical to numerical ones. The models become more
complex as more sand production mechanisms are captured. A comprehensive review on numerical sand
production models can be found in Rahmati et al. (2012).
Majority of the models have been developed for sand onset prediction, i.e., the first stage of the sand
production process or the condition for rock failure and degradation initiation. Several sand quantification
models have also been developed since 2000 that could be utilized to assess sand production severity, i.e.,
sanding amount and rate. Some of these models are formulated based on the sand production mechanisms
developed during the second stage of sand production (Papamichos et al., 2001; Chin and Ramos, 2002;
van den Hoek et al. 2003; Detouray et al., 2006; Detournay, 2009; Ray et al., 2014), whilst the other
models are formulated empirically based on laboratory or field sand production data (Willson et al., 2002;
Fuh and Morita, 2013; Fuh and Nozaki 2014). Despite the significant efforts spent in developing sand
production prediction models, quantification for sand amount and rate is still considered novel in the
industry (Addis et al., 2008). This is particularly so for sanding rate prediction; the predictive accuracy
of the currently available models could be in the range of 0.1 to 10 times of the actual sand production
rate (van den Hoek et al., 2005; Fuh and Nozaki, 2014).
This paper presents a comprehensive study on quantitative sand production prediction with an objective
to develop a practical sand quantification model. The model was developed by correlating the cumulative
amount of sand produced with the critical plastic volume evaluated by numerical modeling based on
observations from laboratory sand production experiments. The study comprised three major components;
laboratory sand production experiments, numerical modeling and a field case study. The following section
will cover the sand production experimental procedure and results, including characterization of the
sandstones used for the laboratory sand production experiments. This is followed by numerical modeling
on the sand production experiments, development of the sand production quantification model, its
calibration and validation based on the sand production experimental results. Finally, a field case study
aiming for model validation is presented to demonstrate how the developed model can be applied to field
studies.
Development of sand quantification model
It has been widely acknowledged that reservoir sandstones with an unconfined compressive strength
(UCS) falling into the so called ⬙grey area⬙ present great uncertainties in terms of making decision if and
when to install sand control downhole. This is where sand production prediction study is most valuable.
Although this ⬙grey area⬙ could be defined differently by different companies, it is commonly referred to
a rock strength ranging between ultra-weak sandstones and well consolidated sandstones. This area is
roughly defined as an unconfined compressive strength (UCS) of 3.5 MPa (500 psi) to 28 MPa (4000 psi)
in this paper. Outside the ⬙grey area⬙, the sand production prediction study may have little value since the
decision is usually quite straightforward. Sand control is considered to be a ⬙must⬙ for a reservoir
formation with UCS less than 3.5 MPa and may be unnecessary for a reservoir formation with an UCS
greater than 28 MPa unless in area with high tectonic stresses and expected high reservoir pressure
OTC-26508-MS 3

depletion. Within the grey area, predicting the onset condition of sand production is insufficient to make
the decision, and the amount of sands likely to be produced is one of the critical considerations.
Consequently, this grey area of sandstone strength is the focus area in this study. The development of the
sand quantification model involved two aspects, i.e., laboratory sand production experiments and numer-
ical modeling on these experiments.
Laboratory sand production experiments and observation
Mechanical properties of the sandstones Sand production experiments were conducted on seven sand-
stones, including three synthetic sandstones and four natural sandstones. The strength range of the
sandstones was selected from the ⬙grey area⬙ prone to sand production. Natural sandstones with a strength
close to the lower bound of the ⬙grey area⬙ were particularly difficult to obtain with good homogeneity
and sufficient size required for the experiments. Hence, two weak synthetic sandstones were manufactured
by mixing silicate sands with desired particle size distribution with a binder and then compacted and
consolidated in a mould. Sodium silicate (SS) and poliset (PS) were used as the binding materials for the
two synthetic sandstones respectively. The porosity of these synthetic sandstones was approximately 38%
to 40%. The third synthetic sandstone was manufactured using the CIPS technology (Kucharski et al.,
1996). The technology can be utilized to manufacture synthetic sandstones with variety of mechanical and
physical properties. The synthetic sandstone manufactured for this study had a porosity of approximately
30%. The four natural sandstones were obtained from open cut quarries in Queensland, Australia and
USA.
The mechanical properties of the synthetic and natural sandstones were determined by conducting
unconfined compressive strength (UCS), triaxial and thick walled cylinder (TWC) tests. The key
parameters of the mechanical properties are summarized in Table 1. The sandstones were divided into
three categories based on their TWC strength in this paper, i.e., ultra-weak sandstones for TWC less than
20 MPa, weakly-consolidated sandstones for TWC between 20 and 40 MPa and consolidated sandstones
for TWC greater than 40 MPa.

Table 1—Summary of mechanical properties of the sandstones selected for sand production experiments
UCS C ␾ TWC
Sandstone E (GPa) v (MPa) (MPa) (deg) (MPa) Saturant Category

SS 0.55 – 1.35 0.1 2.98 1.08 19 7.02 Dry Ultra-weak sandstones

SS 0.16 – 0.56 0.11 1.69 0.55 15.8 5.56 Oil
PS 0.3 – 1.49 0.11 3.5 0.97 21.5 5.4 Dry
PS 0.2 0.12 3.5 – – 6.3 Oil
WG 1.2 – 2.8 0.14 – 0.35 8.05 2.5 26.3 22.2 Water Weakly-consolidated sandstones
WG 3.1 – 3.8 0.25 – 0.28 12.74 3.2 36.7 27.2 Dry
WW 0.89 – 2.82 0.21 – 0.29 6.72 2.11 25.9 22.5 Water
CIPS 1.0 – 5.1 0.13 4.5 1.56 36.1 22.5 Oil
CIPS 1.6 – 8.3 0.11 10.96 2.45 41.8 40 Dry Consolidated sandstones
SWN 1.5 – 7.7 0.12 – 0.44 13.11 3.24 37.5 53.29 Water
MW 8.1 – 10.3 0.29 – 0.34 25.43 6.02 39.3 66 Water

Sand production experimental apparatus and experiment setup The cylindrical sample size for the
sand production experiments was approximately 200 mm in diameter and 310 mm in length with an 18
mm diameter central hole. The large ratio of sample outer diameter to inner hole diameter was necessary
to allow meaningful amount of sands to be produced with minimum boundary effect. Loss of structural
integrity of the test sample can significantly affect the sand production amount and preferential cavity
evolution. Figure 1 shows a photo the experimental apparatus, comprising
4 OTC-26508-MS

Figure 1—Sand production experimental apparatus. The insert at top right is a schematic of a vertical section of the test cell and
sample, and the one at lower right is a photo of the sand and fluid capture system.

● A loading frame capable of applying a force up to 220 tonnes to a test sample,

● A sand production cell with a cell pressure capacity up to 70 MPa. The cell accommodated a
cylindrical sample with a diameter of 200 mm and length of 310 mm,
● A fluid injection system capable of applying an injection pressure up to 8 MPa. The fluid could
be water, oil or compressed air,
● A fluid and sand capture system. The fluid and sand produced from the sample was captured using
a plastic container located on a balance that was continuously measured and logged,
● A data logging and measurement system. The axial force, cell pressure, injection pressure at the
outer boundary of the sample were continuously measured and logged. The cell pressure and axial
stress were computer-controlled, and
● A borehole monitoring and measurement system, including a borehole gauge measuring borehole
displacement in two orthogonal directions and a borescope for visual observation on borehole
deformation and failure.
As shown in Figure 1, the experiment set up simulated sand production from a perforation. Multiple
layers of stainless steel mesh were wrapped around the external surfaces of the test sample to allow axial
as well as radial fluid flow. The central hole was oriented downward in vertical direction to enable the
sand amount to be quantified once the sand particles or clumps were detached from the test sample.
OTC-26508-MS 5

Experimental Procedure The procedure for a typical sand production experiment was as follows:
● Saturate test sample with water under a nominal cell pressure and axial stress. The injection
pressure for water was 0.7 MPa when the permeability of the test samples was high, or several
MPa when the permeability was low,
● Apply cell pressure and axial stress following a given stress path until failure initiation as indicated
by the large deformation detected by the borehole gauge and observed by the borescope,
● Inject test fluid at a constant injection pressure for a given period of time. If sand production was
observed, the fluid injection at the same injection pressure was repeated several times until the
sand production ceased or the amount of sand became sufficiently insignificant in subsequent fluid
injections. If no sand production was observed, the injection pressure was increased to a higher
pressure,
● Apply an increment in the cell pressure and axial stress along the given stress path, and
● Repeat above two steps until a sufficient amount of sand was produced or the pressure capacity
of the sand production cell was reached.
A regular borehole survey was conducted using the borescope during the experiment.
Experimental results and observations Typical sand production experimental results are presented in
Figure 2 to 7 for the three categories of sandstone. The amount of sand produced and drawdown pressure
were normalised with their maximum values respectively for each category of sandstone. For each
experiment, fluid injections were conducted at a number of cell pressures. The fluid was injected at several
injection pressures and was repeated for a number of times at each injection pressure. Each injection lasted
approximately 10 minutes or more. This injection period was based on the experimental observation on
sands produced and the injection was terminated when sand production essentially ceased or decreased to
an in-measurable level. The duration for each experiment or sample required 2-3 weeks. This experimental
procedure allowed various factors affecting sand production behavior to be examined separately, such as
the cell pressure (simulating depletion in reservoir pressure), injection pressure (simulating drawdown),
fluid type, post-failure restabilisation and the effect of axial stress. To facilitate comparison of sand
amounts produced under different drawdown pressures, the normalised sand amount was further divided
by normalised drawdown pressure, i.e., normalised sand amount by drawdown (DD).

Figure 2—Typical sand production experimental results for ultra-weak sandstones: normalised cumulative sand amount produced
versus cell pressure (A), and normalised sand amount by DD versus cell pressure (B).
6 OTC-26508-MS

Figure 3—Typical borehole images at failure initiation (A), collapse (B) and CT image on the tested sample (C) for ultra-weak sandstone.

Figure 4 —Typical sand production experimental results for weakly-consolidated sandstones: cumulative sand amount produced
versus cell pressure (A), and normalised sand amount by DD versus cell pressure (B).

Figure 5—Typical borehole images at failure initiation (A), collapse (B) and CT image on the tested sample (C) for weakly consolidated
sandstone.
OTC-26508-MS 7

Figure 6 —Typical sand production experimental results for consolidated sandstones: normalised cumulative sand amount produced
versus cell pressure (A), and normalised sand amount by DD versus cell pressure (B).

Figure 7—Typical borehole images at failure initiation (A), cavity enlargement (B) and CT image on the tested sample (C) for
consolidated sandstone.

The general observations on these experimental results are summarized below:

● For weakly-consolidated and consolidated sandstones, the mechanical failure must be induced on
the borehole surface prior to sand production. The cell pressure was the predominant factor to
cause borehole failure. Drawdown induced fluid flow had a significant effect on sand production
(Figures 4 and 6). For the ultra-weak sandstone, sand production could take place prior to borehole
failure visually observed (Figure 2),
● Type of flowing fluids had a significant effect on sand production initiation and sand amount
produced. In general, the samples tested with gas flow could sustain a higher cell pressure prior
to experiencing borehole failure and produced less sands, in comparison with the samples tested
with oil flow (Figures 2 and 4),
● The effect of the axial stress (the stress acting parallel to the borehole) had a considerable effect
on sand production by comparing experimental results obtained under a hydrostatic stress (i.e., cell
pressure and axial stress were equal) and with those under a non-hydrostatic stress (i.e., the axial
stress was 50% of the cell pressure). The axial stress appears to have a strengthening effect on the
borehole (Figures 2 and 4),
● For a given stress and drawdown condition, sand production usually took place at the early stage
of fluid flow, and would decrease rapidly and even cease. This observation was particularly
evident for the weakly-consolidated and consolidated sandstones (Figures 4 and 6). For the
8 OTC-26508-MS

ultra-weak sandstones, the sand production might not lead to complete ceasing, but would decrease
and approach a background sand production rate,
● For weakly-consolidated and consolidated sandstones, the predominant mechanism for borehole
failure was shear failure around the borehole. This was followed by erosion of the loose materials
due to flowing fluid in the region of rock degradation (failure zone). The loose sand particles and
small pieces of sandstone from the shear bands would be mobilised and produced (Figure 7). The
large clumps of material would fill up the borehole (Figures 3 and 5), and
● Sand production rate was observed to be highest during the collapsing of the borehole, and the rate
would decrease rapidly once the borehole had fully collapsed. The sand production mechanism
from a collapsed borehole would be erosion of individual particles along the surface of the shear
fractures and in the shear bands.

Numerical modeling on experimental results

Two approaches were adopted to model the sand production experiments, namely discrete element method
(DEM) and continuum mechanics based finite difference method (FDM). The DEM was used to model
and understand the rock failure and sand production mechanisms observed in the sand production
experiments. However, the DEM approach is computationally intensive, which places a limitation on the
model domain size. The FDM was adopted to develop field scale sand production model based on the sand
production experimental results and the fundamental understanding obtained from the DEM modeling.
DEM Modeling The DEM was developed by Cundall (1971). The method can be used to simulate the
disintegration of granular media, such as sandstones, subjected to loading. In the DEM, the granular media
is discretised into particles. The translational and rotational motions of each particles are governed by
Newton’s second law of motion. The contact force between particles can be calculated by employing the
Hookean or Hertian contact law in conjunction with Coulomb’s friction law. A state of equilibrium is
reached whenever the internal forces are equal to the external forces. The main advantage of the DEM
applied to sand production modeling is that it allows the discontinuous nature of rock failure and the
mechanisms at the grains scale to be explicitly modeled.
Prior to modeling sand production experiments, 2D DEM models were constructed for and calibrated
based on UCS and TWC tests. In these modeling studies, the properties of all the grains and cement
between the grains were assumed to be uniform. Grains were deformable and elastic and failure occurred
in the cement. Grain crushing was therefore not modeled. When the sample was loaded within the elastic
range, deformation occurred mainly in the grains. At failure, large relative displacements occurred
between some of the particles and shear and/or tensile fractures might form in the cement. As all the void
space was assumed to be filled up with cement material, the strength properties assigned to the cement
material and the deformation properties assigned to the grains were based on those obtained from the UCS
and triaxial compressive strength tests.
By adopting such an approach, the model input material properties was calibrated with some small
adjustment such that both the laboratory measured UCS and the strain at failure could be matched by the
DEM results almost exactly. It was interesting to note that the mode of failure predicted by the model
agreed reasonably well with what were observed in the UCS tests. Furthermore, by using the same model
input material properties calibrated against the UCS tests, the DEM model was able to predict collapse
strength of TWC tests on the same sandstone as well as the failure mode. Figure 8 shows examples of
comparison on the failure modes observed in experiments and in the DEM model.
OTC-26508-MS 9

Figure 8 —Modeling of weakly-consolidated sandstone and comparison with the failure mode of the tested samples. A and B are photo
from UCS test and numerically modeled shear failure band respectively. C and D are the photo from TWC test and numerically modeled
shear fracture respectively.

A similar approach as for modeling TWC test was used to model the sand production experiments. The
main difference between the models for TWC and sand production experiment was that fluid pressure was
applied at the outer boundary of the model. The sand particles were assumed to be impermeable and fluid
flow occurred only in the void space occupied by the cement between the sand particles.
In the 2D DEM model for sand production experiments, small discrete elements were used around the
borehole wall where failure was expected to occur. Further away from the borehole, bigger elements were
used in order to reduce computing time. Figure 9 shows the DEM mesh, the stress release areas around
the borehole obtained from the DEM modeling and the photo of the borehole in the sand production test
sample.

Figure 9 —DEM mesh for sand production experiment model (A), DEM modeling result showing the areas around the borehole with
stress release in particles (B), and photo of borehole collapsing observed in sand production experiment (C).

The main observations from the DEM modeling are:

● The use of UCS value calculated from the cohesion and angle of internal friction derived from a
series of UCS and triaxial tests was more appropriate in DEM modeling than the UCS value
directly measured from UCS tests, and would improve the DEM model predictions,
10 OTC-26508-MS

● The model input mechanical properties that would give a good match between the model predicted
and laboratory measured TWC, would also give a good prediction of the condition for sanding
onset in sand production experiments. As the samples in the TWC and sand production experi-
ments were subjected to more similar stress conditions, TWC would be a better test than UCS for
sand onset prediction,
● The mode of failure of the different types of sandstones was very important for sand amount
prediction as different amount of sands might be produced from the same volume of failed rock.
Individual sand grains and small clumps could be easily mobilized and produced by the hydro-
dynamic force of fluid flow whilst large clumps might remain in place and would not be produced
unless subjected to further degradation, and
● The mode of failure was not only dependent on the strength of the material, but also on the relative
contribution of cohesion and internal friction to the shear strength. The formation of larger clumps
was expected for sandstones which were more plastic. For sandstones with similar UCS but more
brittle, smaller clumps were expected to form.
FLAC Modeling and Sand Quantification Model Development Since discrete element methods are
computationally intensive, an alternative is to use the equivalent continuum approach which averages the
behaviour across many particles in a small volume in the constitutive model. The commercial software
FLAC (2008) was used to model the sand production experiments. Sandstones were treated as a strain
softening elasto-plastic material with a Mohr-Coulomb yield function and non-associated flow rule. Based
on the triaxial test results on sandstones, the strain softening was assumed to be cohesion reduction with
plastic shear strain whilst the internal friction angle remained constant.
Following a similar procedure adopted for the DEM model calibration, an axisymmetric FLAC model
was firstly constructed for calibration based on TWC tests. The deformation and strength properties for
the sandstones were based on those from the uniaxial and triaxial compression tests. The steel end platens
and the interfaces between the sample and platens were included in the FLAC model explicitly. The cell
pressure and axial stress were increased equally in small steps until the whole sample had yielded (plastic
failure had occurred in the whole sample) when the cell pressure was assumed to be the TWC strength
for the sandstone. Different degrees of cohesion softening as a function of shear plastic strain was applied
to the different sandstones based on the stress versus axial strain curves obtained in the uniaxial and
triaxial tests. With some small adjustment to the model input properties, an almost exact match between
model predicted and laboratory measured TWC could be obtained for all the different types of sandstone
tested.
Once the model was calibrated, the same constitutive model, deformation and strength properties used
in the TWC model were used as input for the FLAC model for sand production experiments. In this model,
a 2D plane strain condition was assumed representing a typical cross-section of the test sample.
The FLAC model computed effective plastic strain around the borehole for a given stress condition and
fluid injection pressure (drawdown). This was continued by changing the cell pressure and axial stress
following the stress path applied in the sand production experiments, and applying fluid pressure at the
outer boundary corresponding to the conditions at the different stages of the sand production experiments.
The effective plastic strain on the borehole wall was defined as the critical effective plastic strain at the
sand production onset observed experimentally. The effective plastic strain in the entire model was
monitored and the area within which an effective plastic strain was greater than the critical effective
plastic strain at sand onset was defined as the critical plastic area. The sandstone in this area would have
undergone sufficient mechanical degradation and potentially be producible, subjected to the hydrody-
namic drag force of the fluid flow. Figure 10 presents a FLAC modeling example for the effective plastic
strain contour plot showing shear bands in near borehole region, which were almost identical to the shear
failure pattern observed from the CT scan image on a tested sample on ultra-weak sandstone. This
OTC-26508-MS 11

example demonstrated that by incorporating the strain softening behaviour, the FLAC model had captured
the fundamental sand production mechanism from this sandstone material.

Figure 10 —Comparison between the effective plastic strain contour plot in the near borehole region from FLAC modeling (A) and a
typical CT scan image of the tested sample of sand production experiment on ultra-weak sandstone.

It was assumed that the amount of sand produced could be related to the critical plastic volume, i.e.,
Equation 1

where MN and VpN is the amount of sand production and the critical plastic volume divided by the
borehole volume respectively. The critical plastic volume was obtained by assuming that the effective
plastic strain distribution was uniform along the borehole length. Figure 11 shows two examples of the
correlations between the normalised amount of sand produced and the normalised critical plastic volume
for the weakly-consolidated and consolidated sandstones. Equation 1 may not fit every data point obtained
experimentally for all the categories of the sandstone, it certainly captured the trend of the data. From
Figure 11, it may be concluded that the critical plastic area or volume is a good indicator of the sand
amount that will be produced.

Figure 11—Examples of correlations between the normalised amount of sand produced and the normalised critical plastic volume for
weakly-consolidated (A) and consolidated (B) sandstones.
12 OTC-26508-MS

For weak sandstone reservoirs, sand production is mainly induced by shear failure around a borehole
(Chin and Ramos, 2002). Sand production in field is a complex and dynamic process influenced by a
number of mechanical, physical and fluid properties. With the objective of developing a practical sand
quantification model that retains primary sand production mechanisms, a number of assumptions had been
made in deriving the model in Equation 1. These include:
● For a given stress, drawdown and fluid flow condition, the amount of sand producible is finite. The
long term fluid flow induced-erosional background sand production is not included in the model,
● The amount of sand produced is directly related to the critical plastic volume within which the
condition of the sandstone satisfied the sand onset condition. Any increase in sand production is
due to the increase in the critical plastic volume, and
● The model does not take into consideration of the cavity evolution type that may form around the
borehole, such as wormhole, slit type of cavity. Cavity evolution due to sand production is a very
complex process under field condition. Although different types of cavity has been observed in
laboratory under simplistic condition, it is difficult to understand and quantify how the cavity
evolution will be affected by stress and fluid type as well as the geometry of completions and fluid
flow.
Preliminary field application
Data preparation
FLAC models for openhole and cased and perforated completions were constructed to estimate the likely
amount of sand to be produced in field for a given depletion in reservoir pressure and drawdown pressure.
The correlation of the amount of sand produced with the critical plastic volume was incorporated into the
FLAC model. The outer boundary of the model was located at 40 times of the borehole radius. A
workflow for evaluating amount of sand produced is shown in Figure 12.

Figure 12—Workflow for evaluating volumetric sand production in field.

The study field is located in offshore shallow water. The objective of the field study was to assess the
applicability of the sand quantification model for sand production evaluation from an existing production
well and a new well to be drilled in the field. A MEM (Mechanical Earth Model) containing the data of
in-situ stresses, pore pressure and mechanical properties was constructed for the reservoir section in the
existing production well, as shown in Figure 13. No core testing data were available from this well
although the relevant wireline logs were available to construct the MEM. Hence, the strength model was
calibrated in other wells nearby with core testing data and the calibrated strength model was then applied
to the existing production well. The TWC strength was derived from UCS using a proprietary correlation
OTC-26508-MS 13

that is valid for sandstones only. The stress model was determined from the density logs, LOT/FIT, MDT
formation pressure measurements. The MEM was validated by conducting a wellbore stability analysis
using the actual mud weights applied during drilling of the existing production well and comparing the
wellbore surface stress condition (ratio of stress to formation strength) with the caliber logs. As shown in
Figure 13, the stress condition on the wellbore wall was generally consistent with borehole enlargement,
which may be considered as stress-induced.

Figure 13—The Mechanical Earth Model (MEM) for the existing production well.

The new well is to be drilled in the same reservoir and is located approximately 2.5 km away from the
existing production well. The MEM, except the in-situ stresses, was then mapped from the existing
production well onto the new well trajectory based on the formation tops in both wells. A new MEM
model was established for the new well by re-calculating the in-situ stresses.
Sand production evaluation
Sand production evaluation was conducted for both the existing production well and the new well. The
existing production well is cased and perforated completion with no downhole sand control, while the new
well is planned for openhole completion with standalone sand screen. The CDP (critical drawdown
pressure) at the original reservoir pressure is presented in Figure 14 for both wells. Figure 14 shows that
the minimum CDP in the perforated section in the existing production well was approximately 7 MPa.
Since the drawdown pressure applied in this well was well below 7 MPa, no sand production should be
expected which is consistent with the field observation so far. For the new well, no positive drawdown
can be applied without inducing sand production. This means sand production is likely from the beginning
of production. In evaluating sand onset condition for the new well, the scale effect in hole size has been
taken into account (Kooijman and Elzen, 1991).
14 OTC-26508-MS

Figure 14 —CDP profiles for the existing production well and the new well at the original reservoir pressure and the depth interval and
CDP bins.

To make the FLAC modeling along the well trajectory within the reservoir section feasible, the log
derived properties were averaged as discussed below. The entire reservoir was firstly divided into depth
intervals and then into CDP bins within each depth interval as shown in Figure 14. For the existing
production well, the entire perforated section was defined as one depth interval whilst for the new well,
the reservoir was divided into two depth intervals for the two sandstone sections. Within each CDP bin,
the mechanical properties, and in-situ stresses and pore pressure were upscaled by averaging, so that one
FLAC modeling was required for each CDP bin for a given depletion and drawdown condition to evaluate
sand production amount. For the cased and perforated completions, the perforation geometry, density and
phasing were taken into consideration in the FLAC modeling.
The FLAC modeling was conducted for two drawdown pressures at several reservoir pressures. The
sand amount produced for the existing production well was minimal, only several kilograms with a
drawdown of 3.5 MPa (500 psi) after a reduction in reservoir pressure by 12 MPa. This modeling result
appears to be consistent with field observations so far, and will continue to be validated by the sand
monitoring data from that well.
OTC-26508-MS 15

For the new well, which is planned for openhole completion, the amount of sand produced is over 500
kg by end of planned depletion in reservoir pressure by 14 MPa with a drawdown pressure of 3.5 MPa
(500 psi). Assuming that the reservoir pressure depletes linearly with time, the modeling results for the
new well is summarized in Figure 15. It shows that the drawdown magnitude has a quite significant effect
on the final amount of sand produced.

Figure 15—Summary of the modeling results on the amount of sand produced for the new well.

Nominal sanding rate evaluation

Currently, the FLAC model incorporating the sand quantification model can provide an estimate on
cumulative amount of sand produced for a given reservoir pressure and drawdown pressure. Sand
production concentration is commonly reported as pptb (pound per thousand barrel) for oil wells, or
lbm/MMscf (pound per million standard cubic feet) for gas wells. The sand production process in field
may be separated into two scenarios, i.e., sand production induced by drawdown pressure while the
reservoir pressure is constant and by reservoir pressure depletion while the drawdown pressure is
maintained constant. Reservoir pressure depletion induced sand production rate with a constant drawdown
can be evaluated by coupling the FLAC model with a reservoir simulator which can provide input data
on reservoir pressure, drawdown pressure and oil or gas production rate to the FLAC model. For example,
a number of FLAC modeling can be conducted along the reservoir pressure depletion plan for a given
drawdown pressure, as shown in Figure 15. A nominal sand production rate can then be evaluated based
on oil or gas production rate, cumulative amount of sand produced for a given time period corresponding
to the reservoir pressure depletion.
In the early time of field production or the reservoir pressure being maintained constant, sand
production can be induced by the drawdown pressure due to rock failure around the borehole or
perforation. In this scenario, sand production can eventually diminish or stop when the reservoir pressure
surrounding the borehole has reached a pseudo steady state in the near borehole region and all the
materials producible in the degraded zone has been mobilized and eroded. Sand production rate can be
estimated from the amount of sand produced from the FLAC modeling and the time required to reach the
pseudo steady state and oil production rate. The time required to reach the pseudo steady state after
application of a drawdown pressure in the borehole would ideally be evaluated by a reservoir simulator.
For an idealized reservoir condition, the time required to reach the pseudo steady state can also be
estimated by exploring the transient solution to the diffusion equation for single phase fluid radial flow,
i.e.,
16 OTC-26508-MS

Equation 2

where p is fluid pressure, r and t are radial and time variables respectively and c is generalized
consolidation coefficient relating to formation permeability and fluid viscosity (Detournay and Cheng,
1988). Assuming a fluid viscosity of 4 cP and reservoir radius of 300 m, the time required to reach a
pseudo steady state for a 7 MPa (or 1000 psi) drawdown would be in the order of days for a permeability
of 100 md, as shown in Figure 16. However, it will be approximately 3 weeks to reach the same pseudo
steady state for a permeability of 10 md (Wu, 2015).

Figure 16 —Variation of reservoir pressure with time in the near borehole region and comparison with steady state reservoir pressure

Summary and conclusions

A comprehensive study on quantitative sand production prediction has been carried out, comprising
laboratory experiments, numerical modeling and field case study for model validation. The major
observations and results are summarized below:
● It was observed from the experiments that rock failure was by shearing. The failed materials were
in large clumps for low strength and ductile sandstones, and in individual particles and smaller
clumps for relatively competent but brittle materials,
● For weakly-consolidated and consolidated sandstones, sanding rate decreased rapidly with time
and would cease once the failed materials transportable by fluid flow had been produced under
given drawdown and stress condition,
● The type of fluids had a significant effect on sand production initiation and amount of sand
produced. The samples tested with gas flow could sustain a higher cell pressure (or depletion in
field) prior to sand production and produced less sand,
● The stress along the borehole (or perforation) had a strengthening effect, i.e., it could delay sand
production onset given other conditions being the same,
● Failure patterns observed on the borehole wall in realtime and CT scan images on tested samples
were almost identical to the simulated failure modes using discrete element and continuum
numerical modeling. This gives greater confidence of the predicted plastic strains, therefore
estimates on the amount of sand produced,
● A practical sand quantification model were developed which related the amount of sand produced
with critical plastic volume, and
OTC-26508-MS 17

● A preliminary field case study was carried out using the sand quantification model. The model
prediction has so far been consistent with field sand production observation. More field observa-
tion and measurements are required for the amount of sand production.
However, it should be noted that sand production is a very complex problem. It involves not only
geomechanical aspect, but also other aspects, such as fluid flow and sand/solid transportation. The
experiments conducted in this study was under an isotropic stress condition on cylindrical samples, i.e.,
the two principal stresses perpendicular to the borehole were equal. This may not be the case in general
for a perforation or an openhole in reservoir. Furthermore, the borehole size used in the experiments was
18 mm in diameter, simulating likely sand production from a perforation. The effects of stress condition
and borehole size will likely have an impact on the sand quantification model. These will be assessed in
future by conducting sand production experiments in true three dimensional stress condition and using
samples with large hole sizes.

Acknowledgements
The financial support and permission to publish the paper from both PETRONAS and CSIRO manage-
ment are gratefully acknowledged. In addition, a number of people have contributed to this work. In
particular, the effort by Mr M. Camilleri, Mr. K. Quinlan and Mr. D. Barker of CSIRO to provide support
on machining, electronics, software and equipment computer control is gratefully acknowledged.

References
Addis, M. A., Gunningham, M. C., Brassart, Ph., Webers, J., Subhi, H. and Hother, J. A. (2008). Sand quantification: the
impact on sandface completion selection and design, facilities design and risk evaluation. SPE 116713, SPE Ann.
Tech. Conf. and Exhib. Denver, Colorada, USA.
Chin, L. Y., and Ramos, G. G. (2002). Predicting volumetric sand production in weak reservoirs in Proceedings of the
SPE/ISRM Rock Mechanics Conference, pp. 161–170, Irving, Tex, USA, Oct.
Cundall, P. A. (1971). A computer model for simulating progressive large scale movement in blocky rock systems, in
Proceedings of the Symposium International Society of Rock Mechanics, Nancy, vol 1, paper II–8.
Detournay, E. and Cheng, A. H. D. (1988). Poroelastic response of a borehole in a non-hydrostatic stress field. Int. J. Rock
Mech. And Min. Sci. & Geomech. Astr. Volume 25, Issue 3, Page 171–182
Detournay, C. Tan, C. and Wu, B. (2006). Modeling the mechanism and rate of sand production using FLAC, in
Proceedings of the of 4th International FLAC Symposium on Numerical Modeling in Geomechanics, pp. 8 –10.
Detournay, C. (2009). Numerical modeling of the slit model of cavity evolution associated with sand production. SPE
Journal Dec. 2009.
FLAC (Fast Lagrangian Analysis of Continua in 2 Dimensions), Version 6.0 (2008). Minneapolis, Minnesota: Itasca
Consulting Group (ICG).
Fuh, G. F. and Nozaki, M. (2014). Completion design using sand management approach based on sanding prediction
analysis for HPHT gas wells. SPE 170954. SPE Ann. Tech. Conf. and Exhib. Amsterdam, the Netherlands, 27-29 Oct.
Fuh, G. F. and Morita, N. (2013). Sand production prediction analysis of heterogeneous reservoirs for sand control and
optimal well completion design. IPTC 16940, Int. Petr. Tech. Conf. Beijing, China.
Kooijman, A. P., van den Elzen, M. G. A. (1991). Hollow cylinder collapse: measurement of deformation and failure in
an X-ray CT scanner and observation of size effect. Proc of the 32nd U.S. Symp. on Rock Mechanics. Norman, July
10-12.
Kucharski, E. S., Price, G. P., Li, H., and Joer, H. A. (1996) Laboratory Evaluation of CIPS Cemented Calcareous and
Silica Sands, 7th Australia New Zealand Conference on Geomechanics, Adelaide, South Australia, July 1-5, 1996,
Conference Proceedings 102.
Papamichos, E., Vardoulakis, I., Tronvoll, J. and Skjrstein, A. (2001). Volumetric sand production model and experiment.
International Journal for Numerical and Analytical Methods in Geomechanics, vol. 25, no. 8, pp. 789 –808.
Rahmati, H., Jafarpour, M. Azadbakht, S., Nouri, A., Vaziri, H., Chan, D. and Xiao, Y. (2012). Review of sand production
prediction models. Journal of Petroleum Engineering, volume 2013.
Ray, P., Rijken, M., Cameron, J., Jones, C. and EI-Fayoumi, A. (2014). Estimating sand production volume in oil and gas
reservoir. SPE 170814, SPE Ann. Tech. Conf. and Exhib. Amsterdam, The Netherlands, 27-29 Oct.
18 OTC-26508-MS

van den Hoek, P. J. and Geilikman, M. B. (2003). Prediction of sand production rate in oil and gas reservoirs. SPE 84496.
SPE Ann. Tech. Conf. and Exhib., Denver, Co.
van den Hoek, P.J. and Geilikman, M.B. (2005). Prediction of sand production in oil and gas reservoirs: field validation
and practical use. SPE 95715, SPE Ann. Tech. Conf. and Exhib., Dallas, Texas, USA, 9-12 October.
Willson, S. M., Moschovidis, Z. A., Cameron, J. R. and Palmer, I. D. (2002) New model for predicting the rate of sand
production, in SPE/ISRM Rock Mechanics Conference, pp. 152–160, Irving, Tex, USA.
Wu, B. S. (2015). Transient solution to the diffusivity equation for liquid flow. Personal communication.