SPE

SPE 22792

Sand Production Prediction Review:

Developing an Integrated Approach
CAM. Vaaken, D.R. Davies, C.J. Kanter, and A.P. Kooijman, Shell Research
SPE Members
a.v.
II
Copyright 1991, Society of Petroleum Engineers Inc.

This paper was prepared for presentation at the 66th Annual Technical Conference and Exhibition of the Society of Petroleum Engineers held in Dallas, TX, October 6-9, 1991.

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. Permission to copy is restricted to an abstract of not more than 300 words. Illustrations may not be copied. The abstract should contain conspicuous acknowledgment
of where and by whom the paper is presented. Write Publications Manager, SPE, P.O. Box 833836, Richardson, TX 75083-3836 U.S.A. Telex, 730989 SPEDAL.

in literature and developed in-house, are reviewed

ABSTRACT and evaluated. The field implementation of sand
prediction technology is also adressed in Section 3.
A reliable, but non-conservative, field The limitations of the sand prediction tools
validated prediction of sand production is essential and their field implementation are discussed in
to decide whether sand control measures need to be Section 4. The benefit of integrating information
installed during well completion. In this paper, obtained from field observations, laboratory
field measurements of sand production are classified experiments and theoretical modelling is
and quantified to obtain a better perspective of the demonstrated. Recommendations with respect to areas
downhole situation. Existing sand prediction requiring further development are formulated.
techniques are presented, critically evaluated and
their limitations discussed. The field application 2. SAND PRODUCTION
of the current generation of sand prediction
techniques is described. The importance of The classification of field measurements of
integrating field observations, laboratory sand production is considered an essential part of
experiments and theoretical modelling is sand prediction as it defines the situation
demonstrated. assessed. The term sand production envelops a wide
range of phenomena. A classification is developed,
1. INTRODUCTION based on field observations, to allow for a better
comparison and interpretation of sand production
When developing a sandstone oil or gas events. Subsequently, changes in the downhole
reservoir, a prediction of sand production is producing geometry are considered on the basis of
required to evaluate the necessity of sand control. the cumulative sand volumes produced.
Sand prediction technology also assists in selecting
the economically most attractive sand control 2.1 Types of sand production
technique, for example, by choosing between
selective perforating and gravel packing. The !E~~~!~~~_~~~~_EE~~~~~!~~
economic incentive for predicting sand production is Transient sand production refers to a sand
based upon the completion cost and the loss of well concentration declining with time under constant
productivity associated with sand control, well production conditions. This phenomenon is
especially in case of gravel packing. frequently observed during clean-up after
In Section 2, the sand production observed perforating or acidising, after bean-up [I] and
in-situ is classified and quantified to obtain a after water breakthrough. The sand concentration,
better description of the phenomena studied in sand the cumulative sand volume and the decline period
prediction. For example, the sand volumes produced vary considerably. Fig. 1 shows three field examples
in-situ are usually not considered in sand with a sand volume between 1 and 200 L and a decline
prediction studies and are frequently much larger period between 1 and 500 hrs.
than implicitly assumed.
Research into prediction of sand production has ~~~~!~~~~~_~~~~_EE~~~~~!~~
been ongoing for the last twenty years. In Section 3 In a great number of fields, continuous levels of
existing sand prediction techniques, both published sand production are observed [I]. The acceptable
sand concentration depends on operational
constraints with regard to erosion, separator
capacity, sand disposal, artificial lift, well
References and illustrations at end of paper
335
 SAND PRODUCTION PREDICTION REVIEW:
2 DEVELOPING AN INTEGRATED APPROACH SPE 22792
location ~tc. Typical tolerated sand cut levels are already during the clean-up phase. In unconsolidated
6-6006g~m (2.1-210 pptb) in oil producers and 16 and weakly consolidated formations the sand may
kg/lO m (1 lb/MMscf) in gas producers [2]. The slough towards the wellbore, arching or bridging
latter surface sand concentration is equiv~lent to a across the perforation openings instead of forming a
downhole sand conc~ntration of about 4 g/m cavity, see Fig. 3c. A more exotic sand production
(1.5 pptb) (3900 m reservoir gas equivalent to scenario, piping [3], is sketched in Fig. 3d.
6 3 The existence of large cavities has been
10 m surface gas). Much higher aJceptable sand cut
levels of the order of 28,000 g/m (10,000 pptb) confirmed in the field by changes in log
have also been reported [3,4]. density [6] and inflow performance. Effective
Part of the continuously produced sand settles wellbore radii up to several meters have been
inside the wellbore and increases the hold-up depth. inferred from a gradually increasing inflow
Depending on the lifting capacity of the fluid flow performance. Note, that the creation of cavities
and the sand concentration (part of) the increases the risk of casing failure due to
(perforated) producing interval may eventually be compaction [7]. This should be considered when
blocked. Normal production is (temporarily) restored deciding on the installation of measures to prevent
after wellbore clean-out. The volume of sand sand production.
settling in the ~ole depends on the well design but
can be several m • 3. SAND PREDICTION TECHNIQUES

The existing sand prediction techniques are

£~~~~~E~E~!£_~~~~_EE~~~£~!~~
Catastrophic sand production refers to events where based on either field observations of sand
a high rate sand influx causes the well to suddenly production, laboratory sand production experiments
choke and/or die. Two catastrophic failure scenarios or theoretical modelling of sand production. The
can be imagined. The first one corresponds to slugs various approaches to sand prediction are presented
of sand creating sand bridges of moderate volume in and evaluated, divided in three corresponding
tubing or choke, e.g. during or after bean-up and categories. The theoretical analysis is complemented
shut-in operations. The second one refers to a by description of its field applications where
massive influx of sand, filling and obstructing the possible.
wellbore. The volume of sand associated with so
called massive sand failure depends on the well 3.1 Field observations of sand production
3
design and may amount from several to tens of m •
Note, that continuous and catastrophic sand Sand prediction techniques based on field
production may refer to the same event, catastrophic experience rely on establishing a correlation
sand production could also correspond to an between sand production well data and field and
excessive, continuous sand cut. operational parameters. Table 1 presents an
inventory of the parameters that may influence sand
2.2 Downhole geometry changes due to sand production production. Normally, only a small selection of
these parameters is used due to the practical
Sand production will lead to a change in the difficulties of monitoring and recording several
geometry of the producing sand face. ~ convenient years worth of data for all the wells involved in a
measure for the change of geometry is the cumulative study.
volume of sand production averaged over the
(perforated) interval length. Fig. 2 shows the Q~~_E~E~~~~~E
cumulative sand volume per unit interval length as a In its simplest form, the field data based sand
function of production period and sa~d concentration prediction tool uses only one parameter. For
for flow rates of 10, 100 and 1000 m /d produced example, a cut-off depth criterion for the
from a 10 m interval. Note, for example, that o~er a installation of sand control measures is used in
ten year period a moderate sand cut of 1-10 g/m several deltaic environments: sand control is not
adds ~p to a sand volume of 18-180 L/m (assuming installed below a certain depth. The critical depth
100 m /d flow rate). A temporary high sand is regionally dependent: refs. 8 and 9 mention
production level has little impact on the cumulative 12,000 and 7,000 ft respectively. Another cut-off
sand volume. criterion frequently applied specifies a
The volume of the intact perforations compressional sonic wave transit time (6t ) below
immediately after shooting is of the order of 1 L/m which sand control is not required; the l!mit 6t is
(13 shots/m, 20 cm long, 2 cm diameter). Initial again field or regionally dependent and may varyc
sand production may proceed through the gradual from 90 to 120 ~s/ft. Tixier et al.[8] have
enlargement of the initial perforations [5], as established a limit value for the sonic and density
sketched in Fig. 3a. After sufficient enlargement log derived parameter G/c (G is the dynamic shear
b
has occurred, separate perforation cavities will modulus, c the bulk compressibility): no sa~2ing
b
merge and cease to exist [5], see Fig. 3b. The problem is expected when G/c exceeds 0.8*10 psi 2 .
cumulative sand volume corresponding to the definite This limit value has been ap~lied successfully [10]
annihilation of perforations depends on the but appears to depend on the regional environment as
perforation policy and is estimated of the order of well [11]. The criteria specifying critical depth,
10-100 L/m. Such volumes are produced routinely, see 6t and G/c are related. For example, 6t decreases
b
Fig. 2. Any sand prediction needs to take this ascdepth increases; thus, the 6t criteri8n can be
change of cavity geometry during the life time of a translated into a12epth2criteriofi and visa versa.
field into account. Also, G/c =0.8*10 psi typically corresponds to
Note, that in weakly consolidated sandstone, 6t =115-120 ~s/ft [8]. The one-parameter approach is
perforations as such may never exist. In gravel c t'lcal, though conservative, and frequently used
prac
packing operations, the average volume of gravel due to its ease of use.
placed behind the casing can be of the order of
3
0.3 ft /ft, i.e. 28 L/m sand has been produced

336
SPE 22792 C.A.M. VEEKEN, A.P. KOOIJMAN, C.J. KENTER AND D.R. DAVIES 3
!~~_~~E~~~!~E~ about 5-10 bpd and relatively independent of sand
The above one parameter models do not explicitely mixture, cavity size, boundary stress and pore
include the depletion of the reservoir pressure pressure. In contrast, sand production and cavity
(AP ) and the drawdown pressure (AP ). These enlargement for friable-consolidated sandstone is
d dd
para~eters are considered in the two parameter governed to a large degree by the boundary stress.
petrophysical tool illustrated in Fig. 4. In Fig. 4 A simplified model test using a thick-walled
the total drawdown pressure (AP d=AP +AP ) is cylinder sample has been developed for field
d dd
plotted versus the sonic transi~ timeefor sand and application based on sand production tests carried
no-sand producing wells located in the same oil out on hollow cylinder samples [22,23).
field. A risk region with a slope of -0.74 .
MPa/(~s/ft) (-108 psi/(~s/ft» and a width of 10.8
!~!~~:~~!!~~_=X!!~~~E_~~~E~~~~
MPa (1560 psi), see Fig. 4, has been established on The initial failure of a perforation can be related
the basis of data from several fields. Sand-free to the initial failure of a hollow cylinder core
production can b~ realistically expected to the left sample [22,23,28). The maximum near-wellbore
of the risk region, while it is essentially vertical effective stress (0 ) sustained by a
impossible to produce wells to the right of the risk horizontal perforation is eq~aI to the initial
region. Fig. 4 indicates that increasing the total failure pressure of a representative thick-walled
drawdown may trigger sand production. The position cylinder (Otwc,i)'
of the risk region is field dependent; sand
production tests or routine monitoring can be used 0=0 . (la)
v,w twc,l
to determine its position.
Another two-parameter tool is described by where initial failure corresponds to visual damage
Stein et al.[12,13) relating a cut-off drawdown of the inner wall. The standard dimensions of the
pressure to the (sonic and density log derived) TWC (thick-walled cylinder) sample are 25 and 8.5 mm
dynamic shear modulus (G). inner and outer diameter and 50 mm length; the test
configuration is shown in Fig. 6. The near-wellbore
~~!!!:~~E~~~!~E vertical effective stress is rather arbitrary
The width of the risk region in Fig. 4 can be defined as
attributed to the influence of other parameters.
Multi-parameter correlations can improve the o (2)
v,w
resolution between sand and no-sand producers.
Fig. 5 illustrates the use of the multiple where 0 is the far-field vertical effective stress.
discriminant analysis technique for the data set of A greatVmany TWC collapse tests carried out on
Fig. 4. Sand production is correlated with a wide friable-consolidated sandstone have established that
range of parameters including depth, sonic transit the collapse pressure of the TWC (Ot ) is 0-30%
time, production rate, drawdown pressure, higher than the initial failure pres~~re: on average
productivity index, shaliness, water cut and gas 0twc i~0.86*Otwc· Eq. ~la) may now be expressed in
cut. The sand and no-sand producing wells are well term~ of the more readlly observable TWC collapse
separated. The parameter influencing sand production pressure (see Fig. 6),
most in case of Fig. 5 is water cut: sand and no-
sand producers are characterised by an average water 0v,w = 0.86*Otwc· (lb)
cut of 19% and 2% respectively. The discriminant
function describing the influence of the various The representativeness of the TWC collapse test for
factors is regionally dependent. In a similar initial perforation failure has been investigated
analysis, Ghalambor et al.(2) used multiple linear- both experimentally and numerically. For example,
regression to correlate the critical drawdown the effect of the different stress regime (isotropic
pressure observed in water-producing gas wells with in the lab, anisotropic in-situ) and of the limited
seven parameters. ratio between outer and inner diameter of the TWC
The multi-parameter techniques are not commonly sample have been investigated over a realistic range
used because of the extensive data requirements. of conditions. The influence of these parameters
lies within the uncertainty range quoted above
3.2 Laboratory sand production experiments (±15%).
Eq. (1) describes initial perforation failure;
Laboratory sand production experiments are but not the subsequent enlargement and (post-
carried out to observe and simulate sand production failure) stabilisation. Also, it should be noted
in a controlled environment. It helps develop that eq. (1) is based on intact rock testing.
insights into the sand production mechanisms and in Perforating introduces a zone of damaged, weakened
the influence of the various field and operational rock around the perforation which volume can easily
parameters on sand production. Theoretical sand exceed 10 L per perforated meter (27). The
prediction models can be validated against the prediction based on eq. (lb) is compared to field
laboratory observations. Moreover, the laboratory observations of sand production events (transient,
sand production experiments can be used as a field continuous and catastrophic) in Fig. 7. As can be
sand prediction tool after translation of the test seen from Fig. 7, eq. (lb) has been found to be
results to the field situation. conservative and may be used with confidence.
Laboratory sand production tests have been
carried out using both unconsolidated sand [10,14- 3.3 Theoretical modelling
21) and friable-consolidated sandstone [22,27). In
the tests on unconsolidated material, sand The majority of existing sand prediction tools
production is dominated by the flow rate and are based on theoretical modelling of perforation
capillary forces. Sand production creates a cavity and cavity stability. The theoretical sand
which gradually enlarges with increasing flow rate prediction tools require a mathematical formulation
until at a critical flow rate it collapses. The flow of the sand failure mechanism. The mechanisms
rate corresponding to cavity failure [15-19) is currently held responsible for sand production are
337
 SAND PRODUCTION PREDICTION REVIEW:
4 DEVELOPING AN INTEGRATED APPROACH SPE 22792
sketched in Fig. B: compressive Tensile failure
failure [5,6,11,23,2B-34], tensile failure
---------------
A reasonable consensus exists on how to model
[5,20,24,33-36] and erosion [10,34]. tensile failure. The stability criterion can be
Compressive failure refers to an excessive, expressed in terms of the normalised drawdown
near cavity wall, (compressive) tangential stress pressure gradient (g ) at the cavity wall [5,33],
pn
(0 ) which causes shear failure of the formation
ma~erial (Fig. B). This condition can be triggered (3 )
by both far field stresses (depletion) and drawdown
pressure. Tensile failure refers to a tensile radial where rand R are the radius of investigation and
stress (0 ) exceeding the tensile failure envelope the cavity radius respectively. Note, that g
and is trIggered exclusively by drawdown pressure depends on the near-wellbore permeability (k~naS
(Fig. B). Laboratory sand production experiments illustrated in Fig. 9: g is higher in case of
support the existence of both types of failure - impairment [5,20,33,35] ~8ue to e.g. perforating,
tensile failure predominates in unconsolidated fluid invasion, fines movement) and lower in case of
sands, while compressive failure predominates in stimulation (due to e.g. acidising or material
consolidated sandstone). dilation [24,34,36]).
Erosion occurs when the drag forces exerted on A critical value of g was first derived by
a particle at the sand face exceed its apparent Bratli et al.[20,35]. He a~~umed an ideally-plastic
cohesion. It is a special form of tensile failure. frictional material, but the conclusion also applies
to other material models. For a cylindrical
£~~~!~~~l~~_!~l!~!~ geometry, stress equilibrium dictates
Several types of compressive failure models have
been published. An elastic-brittle failure model
[6,11,29,31,32] is easy to implement, but has the
disadvantage that it does not offer a very realistic where Sr and S9 are the radial and tangential total
description of friable and loose materials. An stress respect1vely. At r=R, using Sr=or+P=P and
elastic-plastic material model [5,23,30,33,34] Se=oe+P' one derives
involves more computational effort and, in return,
enables a more realistic description of the material (5)
behaviour.
The modelling result is extremely sensitive to A tensile radial stress develops if aor/a(r/R) is
the choice of yield envelope and failure criterion. negative or
One may choose between e.g. the Drucker-Prager
[5,23,29,32,33] and the Mohr-Coulomb [6,11,30,31] (6a)
yield envelopes and between failure criteria based
on maximum plastic strain [5,23,33], maximum plastic irrespective of material model, far field effective
zone size [30] or maximum stress [6,11,29,31,32] stresses and other parameters. For a spherical
etc. The use of different material models may lead cavity one derives
to completely different results [37,3B], despite
being based on the same set of triaxial test data. gpn ~ 20 e (r=R). (6b)
The material model therefore needs to be
validated against lab and field sand production Since °(r=R)=O, the maximum 0e(r=R) is limited by
data. This is not normally done. The conservative the unc6nfined compressive strength 0ucs Hence,
TWC empirical approach has been used as a benchmark 0eSoucs and
to compare the various compressive failure models
against. Most stability calculations are (cylinder) (7a)
conservative with respect to the empirical tool and and °ucs
do not offer an advantage compared to the TWC 2*0 (sphere) (7b)
ucs
approach.
A favourable comparison between field data and set an upper limit to g • A similar result has been
elastic-brittle sand prediction is reported in ref. reported previously [20~24,35]. Tensile failure
39. This is probably a coincidence due to an provides a reasonable description of laboratory sand
erroneous definition of effective stress and the production tests carried out using unconsolidated
choice-of the most stable cavity geometry (vertical sand [20,24] (cohesion corresponding to capillary
cylinder) • forces) and weakly consolidated material [24] (which
A selective perforating policy was implemented experienced dilation).
based on a finite element prediction model [40]. The Eq. (7) can be translated into a drawdown
prediction incorporates a realistic simulation of pressure criterion if sufficient information is
the completion and a sophisticated material model. available. Assuming uniform permeability and steady
It is judged to be more accurate than analytical state flow conditions:
tools. Still, some degree of arbitrariness remains
e.g. the choice of the Drucker-Prager yield envelope g log(R /R) (Ba)
and the maximum plastic strain criterion [5]. pn e
and
Theoretical modelling of compressive failure is gpn = 20ucs (8b)
useful in qualitative terms, e.g. for developing an
optimum perforation policy (density, phasing, for a cylindrical and spherical geometry
size), selective perforation of stronger zones respectively; R is the drainage radius. In ref. 1
[6,40] and formulation of guidelines for maximum eq. (8) has bee~ used to interpret sand production
flow rate, maximum drawdown pressure, bean-up and test data obtained in a weakly consolidated
shut-in. Morita et al.[5] have demonstrated this formation.
application.

338
SPE 22792 C.A.M. VEEKEN, A.P. KOOIJMAN, C.J. KENTER AND D.R. DAVIES 5
In Figs. lOa and lOb, the sand concentration experience: in loosely consolidated formations sand
measured during sand production tests is plotted production from open holes tends to be less than
versus the drawdown pressure for two tests carried from perforated completions [4,41]1 in line with the
out in a loosely consolidated and a consolidated fact that the fluid velocity at the open hole
formation respectively. The sand concentration shows surface is three orders of magnitude smaller than
a sharp increase when the drawdown pressure (or flow the velocity at the (intact) perforation
rate) exceeds a certain treshold. The drawdown surface [10]. Erosion is related to tensile failure,
pressure criterion of eq. (8) is compared to sand but needs to be considered as a separate mechanism
production field data in Fig. 11 where the treshold due to its particulate nature.
drawdown pressure is plotted against the unconfined
compressive strength. Fig. 11 shows that 3.4 Field implementation
AP dd =0.5*U provides a conservative prediction of
field sandu~~oduction. The decision on sand control depends on sand
g is high during the transient flow stage prediction but also on other factors [42], such as
follow~Rg bean-up. The maximum tensile radial stress the effect of sand control on well deliverability,
caused by a bean-up increment APdd,b is well integrity and the ability of the surface
facilities to handle sudden, massive sand influxes.
U
r -APdd,b (9 ) Only the sand prediction aspects are adressed in
this paper.
leading to the following criterion for tensile The field implementation of sand prediction
failure during bean-up: tools is characterised by a high degree of
uncertainty. This is partly associated with the
APdd,b = u t (10) inadequacy of current sand production models, as
described in the previous sections. For example,
where U is the tensile strength (positive by only a few of the parameters listed in Table 1 are
convention). The bean-up criterion tends to be generally considered in sand prediction tools. The
conservative as in practice g is reduced by fluid other factor impairing the quality of sand
compressibility and wellbore ~¥orage effects. Note, prediction is the uncertainty of the input data.
that eq. (10) needs to be refined for practical The input parameter affecting the accuracy of
applications where the bean-up rate is of sand prediction most is the formation strength (e.g.
importance. Controlled bean-up has been observed to the unconfined compressive strength or the TWC
reduce (transient) sand production in the field. collapse pressure). The uncertainty is greatly
A different mechanism leading to tensile reduced if core material is available. Even then,
failure is triggered by shut-in. The plastically incomplete core recovery and a practical limitation
deformed material near the cavity wall may develop to the number of strength tests performed introduces
tensile damage if stress unloading during shut-in is uncertainty as to whether (all) weak zones have been
excessive [5,33]. The damaged material will be (more characterised. Also, the selection of the formation
easily) produced during a subsequent bean-up. The strength to be used in actual sand prediction is not
amount of sand produced as a result of a production unambiguous 1 selecting the minimum strength measured
cycle will depend on the magnitude of the pressure may for example be overly conservative. In practice,
cycle and the strength of the material [5,33]. The the characteristic strength concept used in other
critical pressure cycle magnitude APdd,c is related engineering disciplines is considered adequate1 the
to u ucs ' characteristic strength is defined as the value
exceeded by, for example, 90% of the strength data.
AP = L*u (11) If core material is not available one has to
dd,c ucs
rely on correlations with log derived parameters to
According to ref. 5, L_1-3. Combining the elastic assess the formation strength [6,11,32,40].
expression for unloading at shut-in and the Mohr- Correlations are useful, for example to extrapolate
Coulomb yield envelope one derives L-0.7, see Fig. to other wells of the same field [40] or to indicate
12. Field observations suggest that pressure cycling weak zones [6]. Fig. 14 is a cross plot of u
2 t
may lead to transient sand production. versus Pbv (v =l/~t ), where P is the matrl~c
c c c b .
density and v refers to a gas saturated rock. Flg.
The combined use of compressive and tensile 14 shows meas8red data points for a particular field
failure in the shape of a stability diagram has been together with the average linear trend line. In most
introduced by Morita et al.[5,33]. In Fig. 13 the applications the average trend line is used to infer
tensile failure curve has been adapted according to the formation strength from the log derived
eq. (8). Tensile failure is triggered by an parameter.
excessive drawdown pressure gradient. This results The data scatter in Fig. 14 is caused by the
in perforation or cavity enlargement, thus reducing fact that u is influenced by other parameters
g to within acceptable limits. Compressive failure beside ~t ~~7g. shaliness, grain Sha~e). The
r~~ults from an excessive drawdown pressure AP and uncertain£y in u inferred from Pbv is of the
dd
may lead to catastrophic sand production. The order of 50-l00%~w80nsequently, the a~erage trend
position of the compressive failure envelope depends line should be used with caut~on. For example,
on the cavity geometry and the far field stresses. consider a formation with Pbv ranging from 20 to 30
GPa. According to the trend l!ne, this corresponds
Erosion to u ranging from 27 to 55 MFa. However, in
The term-erosion implies a gradual production of realir~ u varies from -15 to -70 MFa, see Fig.
individual sand grains from the cavity surface. 14. A staS!Iity prediction based on the log derived
According to Durrett et al.[lO] erosion will take 27 MFa minimum formation strength would not be
place if the drag force exerted on a surface representative of the weaker intervals.
particle exceeds the (apparent) cohesion between
surface particles. Hence, fluid velocity becomes an
important parameter. This is confirmed by field
339
 SAND PRODUCTION PREDICTION REVIEW:
6 DEVELOPING AN INTEGRATED APPROACH SPE 22792
4. DISCUSSION 4.3 Laboratory-sand production experiments

In the previous pages the types of sand The TWC approach assesses initial failure. The
production observed in the field and available sand presence of the outer boundary causes the sample to
prediction models have been described and critically collapse [37] and prevents the study of e.g. hole
evaluated. They will now be placed in context and enlargement. The size of reservoir core samples is
recommendations made for improving the sand generally limited to 4 in. diameter. This limits
prediction tools. laboratory sand production testing to e.g. single
perforations [27] or cavities whose enlargement is
4.1 Sand production limited. In case of unconsolidated and loosely
consolidated materials the TWC collapse pressure is
Up to date, sand prediction studies have less meaningful as sample failure is then governed
predominantly adressed the more readily observable by the pressure necessary to extend the plastic zone
transient and catastrophic types of sand production to the outside of the sample [20]. Thus, the
without specifying sand concentration or volume. In influence of the boundary stress on sand production
practice, continuous sand production and the volumes from a weakly consolidated core sample may be
of sand associated are of equal or even more exaggerated.
importance. In many cases operational guidelines In the absence of detailed field information
specify a maximum tolerable (continuous) sand concerning the effect of sand production on the
concentration. Moreover, the downhole geometry downhole geometry, large scale testing is necessary
changes as large cumulative sand volumes are to facilitate a realistic simulation of in-situ sand
produced. Periodic measurement of sand production production [44]. A laboratory test of a completion
levels is essential to monitor this. In addition, a including casing, cement and perforations situated
cavity log is required to assess whether the sand in a large sample would allow the investigation of
production has been uniform or whether it was perforation enlargement and coalescence, and of the
concentrated from certain weaker zones. Cavity influence of perforation policy and borehole
logging also serves to validate model predictions orientation on sand production. Such equipment is
[6] • available for industry use [45]. By comparing large
Sand production field testing is one means of scale and small scale sand production tests,
determining the maximum flow rate that does not correction factors necessary to translate the test
cause a sand-problem (Fig. 10). It should be results on small scale core samples [26] to the
recognised that the sand concentration following field situation may be established.
clean-up and bean-up generally decreases with time
(Fig. 1). Therefore, sand production field tests 4.4 Theoretical modelling of sand production
should be of sufficient duration to be able to
establish the continuous level of sand production Morita et al.[5,33] demonstrated that the
(Fig. 1). They need to be repeated at regular influence of various field and operational
intervals as the reservoir depletes and the well's parameters on transient and catastrophic sand
production characteristics change (e.g. water cut production can be understood qualitatively using
develops). current rock mechanical modelling techniques.
Continuous sand production was not explicitely
4.2 Correlating sand production field data considered, but could have been included. For
example, by tensile failure maintaining an
Linear regression techniques using data from equilibrium between damage deposition and damage
different wells may obscure the actual influence of removal or by the gradual production of the weakened
field and operational parameters. In Fig. 15a sand zone around the perforation [27]. Compressive
concentration is plotted against drawdown pressure; failure of perforations or cavities leading to a
the drawdown pressure does not notably influence the more stable geometry [5] may offer an alternative
sand cut and would not appear as significant in a explanation. In addition, perforations would tend to
correlation exercise. In Fig. 15b changes in sand fail sequentially depending on the local formation
cut are plotted against changes in drawdown pressure strength.
for individual wells in the same field. A definite However, these explanations are considered
influence of drawdown pressure can now be seen. somewhat restrictive. An erosion type process could
The more similar the characteristics of the offer a more general explanation of continuous sand
various wells, the greater the expected success of production and would introduce a credible influence
correlation techniques. The division in water free of flow velocity [10]. Little study of sand
and water producing wells in ref. 2 illustrates this production by erosion has been reported.
point. The on/off influence of water cut would have To improve the rock mechanical sand prediction
dominated the multi-variable linear regression, thus models, validation with respect to lab or field sand
making it less sensitive to the other factors. production data is essential. Advanced numerical and
Records of sand production spanning a longer material modelling will be required to further study
period are most valuable for assessing the influence the sand production mechanisms e.g. to realistically
of depletion and water production [2]. Variations simulate cavity enlargement, the influence of
associated with differences in formation strength, material dilation, and the interaction between
inflow performance, perforation policy etc. are thus compressive and tensile failure.
excluded. In Fig. 16, sand cut, water cut and gross
production rate are plotted against time [43]. The 4.5 Field implementation
onset of sand production with water breakthrough is
clearly established. In this case the flow rate was The uncertainty in input data affects the
beaned back to restrict the sand production rate. results of sand prediction. For example, the
uncertainty in the formation strength inferred from
a log derived parameter influences the success of
selective perforating. This is illustrated by Fig.
340
SPE 22792 C.A.M. VEEKEN, A.P. KOOIJMAN, C.J. KENTER AND D.R. DAVIES 7
14: using the linear trend line a critical formati9n - Correlations of sand production field data with
strength 0t corresponds to a single value 0f Pbv • field and operational parameters increase in
2
However, ex~Iuding the zones with lower Pbv does c accuracy with the number of parameters measured.
not exclude all zones with lower ° in vi~w of the However, the amount of data required and field
scatter of data around the trend li~~. specific character increases accordingly.
Geostatistical techniques can be used to account for - A thick-walled cylinder approach for predicting
the uncertainty in the formation strength and in initial perforation failure is presented and
other input parameters so as to obtain a realistic validated with field data.
assessment of sand production. Sand prediction - Current theoretical sand prediction tools are
studies should therefore aim to develop a sand mainly of qualitative use due to model and data
failure risk scenario rather than predicting single uncertainty. This is compounded by the absence of
dates or production conditions for sand failure. lab and field validation.
Such a scenario would give a prediction as to - Field application of sand prediction techniques is
whether, in a particular field, all wells will start influenced by input data uncertainty. The quality
producing sand at the same moment or whether they of, for example, a sophisticated theoretical model
would fail gradually over a period of time (Fig. can never exceed the quality of the input
17) • formation strength data. This uncertainty needs to
be recognised in the final sand prediction.
5. AN INTEGRATED APPROACH - Integration of sand production field data,
laboratory testing and theoretical modelling will
This paper has reviewed and evaluated existing be necessary to alleviate the limitations of
sand prediction techniques. The techniques based on existing prediction techniques.
laboratory experiments and theoretical modelling
remain of qualitative use unless validated and/or ACKNOWLEDGEMENTS
calibrated by field sand production data. Similarly,
the extrapolation of laboratory sand production The authors gratefully acknowledge the
tests on core samples to field conditions requires permission of Shell Internationale Research
theoretical modelling. Also, in the absence of Maatschappij to publish this paper. We want to thank
relevant field experience, laboratory sand the staff of the Koninklijke/Shell Exploration and
production tests are capable of investigating the Production Laboratory who have contributed through
influence of e.g. water breakthrough and progressive the years to sand prediction research, with special
depletion. Laboratory strength data form the thanks to D. Antheunis, P.B. Vriezen, K.W. Mess,
necessary input to theoretical modelling. J. Geertsma. A. van Asperen, R. Fernandez Luque,
Clearly, field observations, laboratory A.J.T. Grimberg and J.V. Walters.
experiments and theoretical modelling are highly
complementary and should be combined as far as REFERENCES
possible in order to ensure a realistic and
consistent approach to sand prediction. The current 1. Risnes, R., Bratli, R.K. and Horsrud, P.: "Sand
generation of prediction techniques is considered Arching - a Case Study," paper EUR 310 presented
useful, provided their limitations are realised. The at the European Petroleum Conference, Oct.25-28,
representativeness and accuracy of the final 1982.
prediction will vary widely depending on the quality 2. Ghalambor, A., Hayatdavoudi, A., Alcocer, C.F.
of the available input data. and Koliba, R.J.: "Predicting Sand Production in
Balance should be maintained between the three U.S. Gulf Coast Gas Wells Producing Free Water,"
sources of information. Sophisticated theoretical JPT (Dec. 1989) 1336-1343.
modelling is incapable of making great progress in 3. Elkins, L.F., Morton, D. and Blackwell, W.A.:
the absence of sand production field data (or "Experimental Fireflood in a Very Viscous Oil-
representative laboratory sand production data) and Unconsolidated Sand Reservoir, S.E. Pauls Valley
formation strength data. In most field cases the Field, Oklahoma," paper SPE 4086 presented at
sophistication of e.g. current numerical modelling the 47th Annual Fall Meeting, San Antonio, Oct.
is not matched by equally high quality input data. 8-11, 1972.
Theoretical (numerical) modelling should aim to 4. Phillips, F.L. and Whitt, S.R.: "Success of
establish design rules with moderate requirements Openhole Completions in the Northeast Butterfly
concerning input data [46]. Examples of such design Field, Southern Oklahoma," paper SPE 11555
rules are the TWC approach and the drawdown pressure presented at the 1983 Production Operation
criterion of eqs. (1) and (8). Symposium, Oklahoma City, Feb.27-March 1.
5. Morita, N., Whitfill, D.L., Massie, I. and
CONCLUSIONS Knudsen, T.W.: "Realistic Sand-Production
Prediction: Numerical Approach," SPE Prod. Eng.
- Existing sand prediction techniques, both (Feb. 1989) 15-24.
previously published and developed in-house, have 6. Edwards, D.P., Sharma, Y. and Charron, A.:
been critically reviewed and evaluated. "Zones of Sand Production Identified by Log-
- Sand production is classified as transient, Derived Mechanical Properties: a Case Study,"
continuous or catastrophic. Proceedings SPWLA 8th European Formation
- Typical levels of continuous sand production Evaluation Symposium, London, March 1983, Paper
significantly change the downhole producing S.
geometry, corresponding a.o. to perforation 7. Suman, G.O.: "Casing Buckling in producing
enlargement and subsequent coalescence, also in Intervals," Petroleum Engineer (April 1974) 36-
rather competent formations. 42.
- Existing sand prediction tools are based on one or 8. Tixier, M.P., Loveless, G.W. and Anderson, R.A.:
more of the following: field observations, "Estimation of Formation Strength from the
laboratory testing, theoretical modelling. Mechanical-Properties Log," JPT (March 1975)
283-293.
341
 SAND PRODUCTION PREDICTION REVIEW:
8 DEVELOPING AN INTEGRATED APPROACH SPE 22792
9. Lantz, J.R. and Ali N.: "Development of a Mature Annual Technical Conference and Exhibition, San
Giant Offshore Oil Field," JPT (Apr il 1991) 392- Antonio, Oct. 8-11, 1989.
454. 27. Halleck, P.M.: private communication 1991.
10. Durrett, J.L., Golbin, W.T., Murray, J.W. and 28. Geertsma, J.: "Some Rock-Mechanical Aspects of
Tighe, R.E.: "Seeking a Solution to Sand Oil and Gas Well Complet ions," SPE J. (Dec.
Control," JPT (Dec. 1977) 1664-1672. 1985) 848-856.
11. Coates, D.R. and Denoo, S.A.: "Mechanical 29. Nordgren, R.P.: "Strength of Well Completions,"
Properties Program Using Borehole Analysis and Proc. 18th US Symp. on Rock Mechanics, Keystone,
Mohr's Circle," Proceedings SPWLA 22nd Annual Co (June 1977) 4A3-1,4A3-9.
Logging Symposium, June 23-26, 1981, Paper DD. 30. Peden, J.M. and Yassin, A.A.M.: "The
12. Stein, N. and Hilchie, D.W.: "Estimating the Determination of Optimum Completion and
Maximum Production Rate Possible from Friable Production Conditions for Sand-Free Oil
Sandstones Without Using Sand Control," JPT Production," paper SPE 15406 presented at the
(Sept. 1972) 1157-1160. 61st Annual Technical Conference and Exhibition,
13. Stein, N., Odeh, A.S. and Jones, L.G.: New Orleans, Oct.5-8, 1986.
"Estimating the Maximum Sand-Free Production 31. Stein, N.: "Calculate Drawdown that will cause
Rates from Friable Sands for Different Well Sand Production," World Oil (April 1988) 48-49.
Completion Geometries," JPT (Oct. 1974) 1156- 32. Edwards, D., Joranson, H. and Spurlin, J.:
1158. "Field normalization of formation mechanical
14. Hall, C.D. Jr. and Harrisberger, W.H.: properties for use in sand control management,"
"Stability of Sand Arches: a Key to Sand Proceedings SPWLA 29th Annual Logging Symposium,
Control," JPT (July 1970) 821-829. June 5-8, 1988, Paper Y.
15. Tippie, D.B. and Kohlhaas, C.A.: "Effect of Flow 33. Morita, N., Whitfill, D.L., Fedde, O.P. and
Rate on Stability of Unconsolidated Producing Lovik, T.H.: "Parametric Study of Sand-
Sands," paper SPE 4533 presented at the 48th Production Prediction: Analytical Approach," SPE
Annual Fall Meeting, Las Vegas, Sept.30-0ct.3, Prod. Eng. (Feb. 1989) 25-33.
1973. 34. Dusseault, M.B. and Santarelli, F.J.: "A
16. Cleary, M.P., Melvan, J.J. and Kohlhaas, C.A.: Conceptual Model for Massive Solids Production
"The Effect of Confining Stress and Fluid in Poorly-Consolidated Sandstones," Proceedings
Properties on Arch Stability in Unconsolidated ISRM-SPE Int. Symp. Rock at Great Depth, Pau,
Sands," paper SPE 8426 presented at the 54th Aug. 28-31, 1989, Vol.2, 789-797.
Annual Fall Technical Conference and Exhibition, 35. Risnes, R., Bratli, R.K. and Horsrud, P.: "Sand
Las Vegas, Sept.23-26, 1979. stresses around a wellbore," SPE Journal (Dec.
17. Wood, D.C.: "The Effect of Sand Size on Arch 1982) 883-898.
Stability in Unconsolidated Sands," Thesis T- 36. Vaziri, H.H.: "Theoretical Analysis of Stress,
2165, Colorado School of Mines (1979). Pressure, and Formation Damage during
18. Lasaki, G.O.: "Arch Stability and Failure Production," J. Can. Petro Techn. 27 (1988) 111-
Behaviour in Unconsolidated Natural Sands," 117.
Thesis T-2299, Colorado School of Mines (1980). 37. Veeken, C.A.M., Walters, J.V., Kenter, C.J. and
19. Winslow, P.M.: "The Effect of Fluid Pore Davies, D.R.: "Use of Plasticity Models for
Pressure on the Formation and Stability of Sand Predicting Borehole Stability," Proceedings
Arches," Thesis T-3100, Colorado School of Mines ISRM-SPE Int. Symp. Rock at Great Depth, Pau,
(1985). Aug. 28-31, 1989, Vol.2, 835-844.
20. Bratli, R.K. and Risnes, R.: "Stability and 38. McLean, M.R. and Addis M.A.: "Wellbore Stability
Failure of Sand Arches," SPE Journal (April and Analysis: a Review of Current Methods of
1981) 236-248. Analysis and their Field Application," paper
21. Selby, R.J. and Farouq Ali, S.M.: "Mechanics of IADC/SPE 19942 presented at the IADC/SPE
Sand Production and the Flow of Fines in Porous Drilling Conf., Houston, Feb. 27-March 2, 1990.
Media," J. Can. Petro Techn. 27 (1988) 55-63. 39. Barrow, D.C. and Lasseigne, C.A.: "A Field
22. Vriezen, P.B., Spijker, A. and VIis, A.C. van Evaluation of the Mohr's Circle Technique for
der: "Erosion of Perforation Tunnels in Gas Predicting Sand Strength," paper SPE 13087
Wells," paper SPE 5661 presented at the 50th presented at the 59th Annual Technical
Annual Fall Meeting, Dallas, Sept.28-0ct.l, Conference and Exhibition, Houston, Sept. 16-19,
1975. 1984.
23. Antheunis, D., Vriezen, P.B., Schipper, B.A. and 40. Massie, I., Nygaard, o. and Morita, N.:
VIis, A.C. van der: "Perforation Collapse: "Gu11faks Subsea Wells: an Operator's
Failure of Perforated Friable Sandstones," paper Implementation of a New Sand Production
SPE 5750 presented at the European Spring Prediction Model," paper SPE 16893 presented at
Meeting, Amsterdam, April 8-9, 1976. the 62nd Annual Technical Conference and
24. Perkins, T.K. and Weingarten, J.S.: "Stability Exhibition, Dallas, Sept. 27-30, 1987.
and Failure of Spherical Cavities in 41. Barril, R.J. and Gay, L.G.: "Controlling Sand
Unconsolidated Sand and Weakly Consolidated Production in High Rate Gas Wells," World Oil
Rock," paper SPE 18244 presented at the 63rd (Sept. 1983) 52-56.
Annual Technical Conference and Exhibition, 42. Weissenburger, K.W., Morita, N., Martin, A.J.
Houston, Oct.2-5, 1988. and Whitfill, D.L.: "The Engineering Approach to
25. Holt, R.M., Fjaer, E. and Raaen, A.M.: "New Sand Production Prediction," paper SPE 16892
Elements of Sand Failure Prediction," presented at the 62nd Annual Technical
Proceedings SPWLA 11th European Formation Conference and Exhibition, Dallas, Sept. 27-30,
Evaluation Symposium, Oslo, Sept. 1988. 1987.
26. Halleck, P.M. and ·Damasena, E.: "Sand Production 43. Labrecque, B.: private communication (1991).
Tests under Simulated Down Hole Pressure 44. Kooijman, A.P., E1zen, M.G.A. van der and
Conditions using Shaped-Charge Perforated Well Veeken, C.A.M.: "Hollow Cylinder Collapse:
Cores," paper SPE 19748 presented at the 64th Measurement of Deformation and Failure in an x-
342
SPE 22792 C.A.M. VEEKEN, A.P. KOOIJMAN, C.J. KENTER AND D.R. DAVIES 9
Ray CT-Scanner, Observation of Size Effect,"
Proc. of 32nd u.S. Symposium on Rock Mechanics, COMPLETION
Norman, July 10-12, 1991. - Wellbore orientation, wellbore diameter
45. Behrmann, L.A. and Elbel, J.L.: "Effect of Completion type (open hole/perforated)
Perforations on Fracture Initiation," paper SPE Perforation policy (height, size, density,
20661 presented at the 65th Annual Technical phasing, under/overbalance)
Conference and Exhibition, New Orleans, Sept. Sand control (screen, gravel pack, chemical
23-26, 1990. consolidation)
46. Starfield, A.M. and Cundall, P.A.: "Towards a Completion fluids, stimulation (acid volume, acid
Methodology for Rock Mechanics Modelling," Int. type)
J. Rock Mech. Min. Sci. & Geomech. Abstr. 25 Size of tubulars
(1988) 99-106.
PRODUCTION
TABLE 1: PARAMETERS INFLUENCING SAND PRODUCTION - Flow rate
Drawdown pres~ure
FORMATION Flow velocity
Rock Damage (skin)
Strength Bean-up/shut-in policy
Vertical and horizontal in-situ stresses (change Artificial lift technique
during depletion) Depletion
Depth (influences strength, stresses and Water/gas coning
pressures) Cumulative sand volume
Reservoir
Far field po~e pressure (changes during depletion) * influenced by completion and production
- Permeability * philosophy.
- Fluid composition (gas, oil, water)
- Drainage radius
- Reservoir thickness
- Heterogeneity

100
3000lF.:=?E:=::sr====;----,----,------,

~ 300 JL..:===c.::....:",-,,"--p...L...,
L--r--t----r 8O
i 100

c: 10
'5
.g
()
g 10
] 30 §
8
3 0
04 06 08 10 12 14 16 18 20 22 24 02 04 06 08
Time (hrs)
0.1
30 r -----,=c=rr:;:::;::::;;:::::::::;;;::::::;,
b. Light oil well
Sand volume - 351 0.01 0.1 10
B Interval = 1-10 m Time (yellrs)
! 20

~ Fig. 2 Cumulative sand volume produced

"g (in Um @ flowrate in m3/d; 10m interval)
10
~

0
0 2468363840
Time since startup (days)
500
c. Heavy oil well
.;-400 Sand volume - 200 I
E Interval = 20 m
S300
~
"C 200
lij
Ul 100 c. Sloughing sand d. Pipe

0
0 40 80 120 160 200 240 280 'x
/
,)..--
\ l
Time since startup (days) --f- J'" ( r
I
I
j/ -+-/ '

Fig. 1 Transient sand production

\ - ~
;< ---- /
'..... ..../

Fig. 3 Sand production - downhole situation

343
 SPE 2 2792
30 --or----------=--:-----;-::-:-:~
Single completions
• sand problems I
o no sand problems 2
I
I
•
25 I
I
00 I
o I
20 I
.gc I:> 0 --u
o "'_ :.

l ~
0
00 000
0
1
I •
•
~ 15 l;j 0 -1- __o ~ __ ...Q... __ L~, __••-------.
~ 00 :G 0 :.- • • •
<l ~ 8 I •••
10 Il!
Eo 0 01
iii • • ••
o ~ -1 I

5 Io I
safe
region o o
I
I
I
I
Single-zone producers:
o I
• sand problem
• -2
1> o no sand problem
o --J-------,------,----.::~---I T I
70 80 90 100 -3 -2 ~ 0 \ ~ ~ 4
L\lc (jl.SIft) Discriminant function
Fig. 4 Total drawdown versus transit time for
intervals with and without sand problems Fig. 5 Plot showing result of multiple-discriminant
analysis

80

!-P
JI
• • •
Hollow cylinder sample
Sleeve
60
•
I Immm- End cap <is
0-
-- (Jtwc ~40
:J
-t
20

0
0 5 10 15 20 25 30 35 40 45 50
O'twc(MPa)

Fig. 7 Near-wellbore vertical effective stress

Fig. 6 Thick-walled cylinder collapse
versus TWC collapse pressure (field data)

Compressive failure

~. Po
Impaired zone Stimulated zone

r
0'
Far field stress +
drawdown induced ll.Pdd

Erosion Pw

Drawdown induced
EFlow induced
2
rlRi
3 4

Fig. 9 Dependance of normalised drawdown pressure

gradient on near-cavity permeability
2 3 4
r/Ri
2 3 4
r/Ri

Fig. 8 Sand production mechanisms

344
 SPE 2 2792

12

10 • •
8
til
Q. •
~ 6
"C
"C
Q.
<:l 4

0
•
0 2 3 4 5 6 7 8 9 10
O'ucs (MPa)

Fig. 11 Drawdown pressure versus unconfined

compressive strength (field data)

9pn

Compressive
failure

Mohr-Coulomb

Tensile failure

O'r 0'9 (= O'ucs>

It I

- 1.5' APdd
Fig. 13 Sand stability diagram
(adapted from Ref. 5)
Fig. 12 Tensile damage due to shut-in

345
 SPE 2 :J 79 2

20
a) 0 0 0

15
:a
is.
.e: ..
a
"0
10 -
00
0

Iii 0

••
0

- Linear trend line CJ) 0 0

140 • e• •
• 5 0
0
.- . ..
0

• •
120 •• •••~I • ••
•• : :... -,(.r. -.-::.'. ,.. ~
o -+-~..!i.!.I,..a!..!lL...a....!..l,pe..!."":"-r-I-!..-'::I--r-r----I
100 o 100 200 300 400 500 600
lil Drawdown pressure (psi)
a..
~ 80
!
t:) 30
60 b)
20
40 ~
g
'S 10
20 0
"0
C
as 0
0 f/)

0 5 15 25 35 45 55 .5
~ -10
Pb'1c (GPa)
Iii
.l::
0 -20
2 •
Fig. 14 TWC collapse pressure versus Pbvc
-200 -100 0 100 200 300
Change in drawdown pressure (psi)

Fig. 15 Effect of drawdown pressure on sand

production (field data)

6"" 60 ~------W-a-te-r-o-n-se-t----------'
.€ ~ sand cut I ~
E.! 40 • Water cut •
// '"
a • Gross rate ~ t ///
"0 20

~ oL-_......----..;n:.~L-:~~-~
.. ~
t :
; I
!
I
/

'S !I
~ t-. :
~? 40
:::.13
S'"
-.
• •
I

e ~ 20
:ll~
e
C) o-l---r--,..-...--..---;:....--,--.----.-r----,--i A, 8, orC?
Jun Sap Dec Apr Jul Okt Jan May Aug Nov Mar Jun
Time - - -

Fig. 16 Record of gross rate, water cut and Fig. 17 Sand failure scenarios
sand concentration

346