Society of Petroleum Engineers

SPE 28852

A Study of the Physical Mechanisms of Sanding and Application

to Sand Production Prediction
J.M. Cook, * I.O.R. Bradford, and R.A. Plumb, * Schlumberger Cambridge Research
'SPE Members

Copyright 1994, Society of Petroleum Engineers, Inc.

This paper was prepared for presentation at the European Petroleum Conference held in London, U.K., 25-27 October 1994.

This paper was selected for presentation by an SPE Program Committee following review of information contained in an abstract submitted by the 8uthor(s), Contents of the paper,
as presented. have not been reviewed by the Society 01 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 Librarian, SPE, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A. Telex 163245 SPEUT.

ABSTRACf predicting sanding potential on a foot-by-foot basis,

in particular to refine decisions on gravel-packing
The complications involved in understanding and and selective perforating.
predicting sand production are discussed. The need
to predict sanding potential on a foot-by-foot basis Many factors have been proposed as being
means that a simplifying approach is required, and important in sand production: for example, stress
this paper sets out experimental and initial modelling state, perforation strategy and type, production rate
progress in this direction. and history, rock strength and plasticity, nature of
the produced fluid (e.g., water cut), and perforation
We have induced sand production in simple, interaction. With so many variables it is difficult to
reproducible and easily monitored systems, in order decide on the basis of field experience which are the
to be able to vary one parameter at a time and most important for a given field or fonnation. The
examine its effect. This has meant the use of difficulties are compounded by the sparse nature of
synthetic rocks, moulded rather than perforated field sand production data.
holes, and two simple geometries.
To improve this situation, the work described here
The effects of stress, flow rates, fluid viscosity and aims not to simulate downhole conditions but to
rock strength have been examined so far. For the simplify them, reducing the number of variables and
high penneability rocks used in the experiments, the increasing the number of measurements to get a
stresses in the rock produced by radial flow into the clearer picture of what the important factors are and
perforation tunnel have little influence on sanding; how they are inter-related. The variables chosen for
flow along the axis of the tunnel is found to be the initial study have been stress and flow; other
controlling flow. This allows the sanding problem aspects, such as sensitivity to water cut, will be
to be split into the prediction of damage around the examined during future work in the light of the
tunnel induced by far-field stresses, and the transport insight obtained here.
of damaged material by axial flow.
To achieve this simplification, a synthetic rock has
INTRODUCfION been used, to obtain better reproducibility than is
usual with natural rocks, and also to be able to tailor
Sand production is a major problem in many fields variables such as strength and penneability.
throughout the world, and has been the subject of Moulded perforation tunnels have been used rather
research for many years. A number of sand than real tunnels generated by perforating with
production models and criteria exist 1-8, ranging shaped charges, since the changes produced in the
from simple depth criteria to complex finite element rock by a real perforating charge are still not clear.
codes, and there have been recent advances9,10 in
experimental simulation of the sanding process. All the experiments and discussion here concern the
There remains, however, a demand for a method of early stages of sand production by flows of light,

473
low viscosity oils from cased and perforated
completions. Sanding in gas and heavy oil wells,
and in open-hole completions, are important
phenomena but we do not address them here, nor the
possible development of cavities and sand arches
behind perforations after extensive sand production.
EXPERIMENTAL DETAILS
Equipment
Two types of sanding experiment were run. The
initial tests, designed for rapid turnover to give a
survey of the phenomenon, used a near-uniaxial
strain geometry, with a simulated perforation tunnel
perpendicular to the maximum principal stress
(Figure 1). Fluid was pumped through four ports in
the vessel wall, directly into the synthetic rock, and
flowed out through the perforation tunnel, which
was under atmospheric pressure. In some of these
tests an endoscope was placed at the exit of the
perforation, allowing observation of the sanding
process as it occurred.
Subsequent tests used a cylindrically-symmetric
geometry (Figure 2). The perforation tunnel ran
along the axis of the cylinder of rock, and liquid
could be flowed independently and simultaneously
along the axis of the tunnel and radially through the
rock. The sample could be subjected to independent
radial and axial stresses, although in all of the tests
here the stress state was isotropic.
Many methods were tried for continuous monitoring
of the quantity of sand produced, but none had
sufficient range and sensitivity combined. Sand
production was therefore initially assessed semi-
quantitatively by visual observation, placing the
instantaneous rate of production into ten classes
from zero to catastrophic. Later experiments used a
batch filtering and weighing procedure to obtain
fully quantitative data.
Materials
A large number of experiments was planned,
requiring reproducible and very weak rock, so it was
decided to use a synthetic material rather than take
on the problems of acquiring a natural one.
The material needed to be very weak (around 1-3
MPa unconfined compressive strength - UeS), very
permeable (1 Darcy or higher), stable in the presence
of water, and capable of being produced in large
sizes. After some testing, the final choice was a 200
micron sand cemented with very small volume
fractions of a low-viscosity epoxy. Most of the work
reported here used samples with 0.75% by volume
of epoxy, having a ues of about 2.5 MPa. Porosity
and permeability were 38-40 % and 50 Darcies. The
latter is of course very high, and not representative
even of sanding reservoirs; as stated above,
however, these experiments are not intended as
simulations and the high permeability can be taken
into account in models of the sanding process.
474
Permeability could be reduced by changing the
sorting of the sand.
The sand/epoxy mix was cast either into a
cylindrical mould or directly into the pressure vessel
(both internally coated with low friction material),
depending on the geometry of the experiment. The
loose mixture was tamped into the mould in layers,
using a fixed number of blows per layer, but was not
compacted beyond this. For the cylindrically-
symmetric experiments, a bead pack was then cast
around the curved surface of the sand cylinder, to
provide a path for the fluid flowing radially through
the rock. This bead pack was extremely permeable
and very weak, and is assumed not to have a
significant influence on either the flow or the stress
state within the sand cylinder. The simulated
perforation tunnel (passing all the way through the
sample for the equipment in Figure 2) was produced
by moulding the sand around a 10 mm diameter
metal rod with a hemispherical end.
The material was characterized using conventional
triaxial testing of individually moulded 50 mm
diameter cores, to determine parameters used in
elastoplastic models of the sanding process. The
epoxy is water sensitive and so wet samples of the
rock were weaker than dry ones; this effect,
however, stabilized very quickly after wetting, and
all tests, sanding and characterization, were carried
out in a wet state.
Unconfined strength of nominally-identical samples
varied by approximately ±20%, and the porosity by
±5% (Le., approx. ±2 p.u.). This was judged to be
acceptable, as was the homogeneity of the samples.
EXPERIMENTAL RESULTS
Experiments in general involved changing either
stress at constant flow rate, or flow rate at constant
stress, and monitoring the resulting sand production.
Figure 3 shows a map of the severity of sanding
during a number of tests in the uniaxial strain
geometry. The horizontal axis is flow rate through
the rock into the perforation, and the vertical axis is
the stress applied perpendicular to the perforation
axis. Each horizontal or vertical stripe represents
one experiment. A white area of the map means
that no experiment has been carried out under those
conditions; otherwise denser shading of the map
corresponds to more severe sanding. It is clear that
sand production rate increases with both stress and
flow rate, and that there are threshold values for
each below which negligible sanding occurs. Figure
3 shows the behaviour of a rock made with 0.75%
epoxy, having a ues of approx. 2.5 MPa, with
water as the produced fluid. Other experiments
show, as expected, that weaker rocks produce
equivalent amounts of sand at lower stresses and
flow rates. Increased fluid viscosity has a similar
effect.

Investigation of the tunnel during sand production,
using an endoscope, revealed the following points:
• solids are produced in two fonns; as individual
grains and as flakes.
• the majority of sand is produced from the sides of
the tunnel lying at 90° to the maximum compressive
stress. This is where the highest stress
concentration is expected to be.
• both grains and flakes separate from the wall over
a period of time rather than suddenly. It was not
possible to determine whether this is because of
intrinsic time-dependence of the material, or
because of flow effects.
• if the fluid velocity in the perforation drops below
about 0.2 m.s- l , solids are no longer transported
along the tunnel. A few grains and flakes continue
to separate from the wall of the tunnel, but fall to the
bottom of the tunnel and fonn a bed there. If the
flow rate is increased after a low-flow period, this
bed is swept out of the perforation, resulting in a
sudden transient increase in sand production rate.
This effect may be partially responsible for the
transient sanding seen in wells after increases in
production rate6 .
• at high stresses the tunnel may collapse, with
slabs and plates of rock moving into the tunnel and
fonning a loose pack. This is accompanied by a
substantial increase in sand production rate.
Although the uniaxial strain geometry allows rapid
testing with anisotropic stresses, experiments with
cylindrical geometry are easier to characterize and to
model, so further tests were carried out with this
geometry to resolve some of the issues that arose
from the early tests. Fluid flows will be defined as
follows:
• radial flow is from the outer curved surface of the
sample to the perforation tunnel;
• axial flow is flow directly into one end of the
tunnel;
• total flow is the sum of the two, i.e., the flow from
the outlet of the tunnel.
Figure 4 shows sand production rate in a cylindrical
sample, as a function of external radial stress and
axial flow rate along the perforation tunnel. There
was no radial flow in the rock during this test. The
two columns under each stresslflow condition give
the rates for 0 - 30 seconds into the flow interval,
and for 60 - 90 seconds, and clearly show that
sanding rate declines with time after a change in
conditions. These data show that, for these
conditions, the known stress and flow rate
dependencies of sand production can be produced
without involving the stresses caused by fluid flow
within the rock. The reasons for this will be
examined in the Discussion section. This figure also
shows that, in this material (whose UCS is approx.
2.5 MPa), sanding begins at a stress of 6 MPa and
becomes suddenly severe between 9 and 10 MPa; a
separate test carried out with no flow of any kind
showed that collapse of the hole (observed by eye)
475
begins at about 6.5 MPa. Tests on weaker material,
with a UCS of approx. 1 MPa, showed the onset of
sanding at 2 MPa radial stress and severe sanding at
5MPa.
Figure 5 shows two examples of sand production
rate versus time, together with flow rates through the
rock into the perforation (radial flow) and directly
into one end of the perforation (axial flow). Radial
stress was constant with time. The sanding rate is
measured during 4 consecutive 30 second intervals
for each flow condition, and again usually declines
with time after a change. Note that, with some
exceptions, radial and axial flow rates have similar
effects on sand production rates. This is shown
more clearly in Figure 6, which shows mean sand
production rates (averaged over each two minute
flow interval) versus axial and radial flow rates for
the two experiments in Figure 5. Multiple linear
regression of the mean sanding rate against axial and
radial flow rates, for these experiments and others,
shows that the apparent effect of radial flow rate on
sanding is similar to or less than that of axial flow
rate. We will now show that this means that the
direct effect of radial flow is negligible, and that its
apparent influence is by contributing to total flow.
DISCUSSION
Influence offlow rate
Flow from the rock into the perforation is nonnally
thought to influence sand production by changing
the radial effective stress near the rock surface
(which is generated by the stress concentration
around the tunnel). These changes can encourage
shear or tensile failure at the rock surface. The
experiments above show that another flow (along the
perforation) is also important, and this will now be
examined in more detail.
The two physical flow variables in the experiment,
which are expected to influence sand production, are
the radial flow rate (through the rock into the
perforation, changing the state of stress around it)
and the total flow along the axis (made up of the
sum of the radial flow and the experimentally-
imposed axial flow rate). It is not possible to
detennine the effect of radial flow directly. We can
do this indirectly, however, from the equal apparent
importance of radial and axial flows. Total flow is
the sum of these, and the following argument shows
that it is the only important variable. If the sanding
rate is S, and it depends linearly on total and radial
flow rates Ft and Fr (above threshold values) then
(1)
where Ct, Cr, al and a2 are constants. Rearranging
to express in tenns of the experimental variables,
i.e., axial and radial flow rates, gives
From experiment we find that al "" (a1 + a2), so a2,
the influence of radial flow alone, is negligible. A
similar conclusion was reached by Tronvoll et al 9
from modelling of experiments on a weak: sandstone.
This indicates that, in these experiments, the radial
flow in itself (more specifically, the radial effective
stress change introduced by radial flow) is of minor
importance in sand production. The real
significance of the flow through the rock is that it
feeds the flow along the axis of the perforation.
This view is supported by the behaviour shown in
Figure 4, where sanding is seen as a result of axial
flow alone; radial flow is evidently unnecessary.
Vriezen et al 11 also used an independent axial flow
(of gas rather than liquid) in a simulated perforation
~unnel. They found that axial flow velocity
mfluenced ~he rate of sand production, but not the
fina1 collapse stress. They did not vary radial flow
rates appreciably.
The unimportance of radial flow is a strong
conclusion, and its applicability to other sand
production conditions needs careful examination
(especially the case of sand production from rocks
with lower permeability than the synthetic material
used here). A starting point is to consider how the
zone of damaged material around the tunnel affects
sanding behaviour. Virtually all produced sand
must come from this zone. If sand is not removed
from here, it does not matter whether reservoir
depletion, for ex.ample, or changes in drawdown,
lead to changes m the stress state around the
perforation tunnel, because the sand stays where it
IS.
The forces acting to remove material from this zone
are those from gravity and flow. Gravity alone can
only remove sand from a near-vertical perforation
tunnel above the well bore, so flow forces are
dominant. The relative importance of radial flow
into th~ tunnel, and axial flow along it, can be
determmed by examining the flow velocities.
Taking a highly productive, small perforation tunnel
(25 barrels per day, 12 mm diameter, 200 mm long)
and assuming uniform flow into the tunnel along its
length, the fluid radial velocity at the rock surface
can be calculated as 6.5 mm sec-I. The axial
velocity of the fluid in the tunnel rises from near
zero at the tip to 430 mm sec- 1 at the mouth. The
corresponding values for the highest flow rate used
in the experiments are 21 mm sec- 1 and 1700
!llm sec-I ..In general t~e maximum axial velocity
m a tunnel IS 50 - 100 times the radial velocity.
The radial velocities are less than or similar to the
settling velocities of sand grains in water or light
oil 12, so the forces exerted on grains must be less
than or ~imilar to t~e buoyant weight of the grains.
Any reSIdual coheSIOn of the damaged material, even
476
very small values, will prevent its removal by the
radial flow.
The axial flow velocities, on the other hand, are in
general larger than those need for transport of sand
grains along the tunnel. Transport of solid grains by
fluids has been extensively studied (e.g., 12-14),
although there are no simple models for the
phenomenon. There are, however, a number of
semi-empirical methods for estimating critical
velocities for suspension or transport. For water or
light oil under downhole temperatures, the critical
velocity for transport of sand-sized particles in a
typical horizontal perforation tunnel is between 0.1
and 0.2 m.s- 1 (compare this to the 0.2 m.s- 1 seen in
these experiments, using an endoscope, as the flow
velocity below which sand was not moved).
Axial flow with this velocity will remove loose sand
from the tunnel, but will not remove material with
any residual cohesion from the walls. Higher
velocities will do this, however, reducing support
around the perforation and potentially allowing
further growth of the elastoplastic zone, and
continued sanding. There is little understanding at
present of the erosion of cohesive materials 12; there
have been many tests on clay soils but little
agreement, and no way to extend the results to sand-
sized particles. All we can say is that the velocities
needed for removal of cohesive material will be
greater than those for loose grains. Figure 4
indicates that, for the synthetic rock damaged by
stress, there is a threshold velocity of -0.4 m.s- 1,
above which sanding becomes severe, and this may
correspond to the velocity for removal of material
with residual cohesion.
Durrett et al 15 considered sand transport in this way
and concluded that most formation grains would be
transported by typical production velocities in a 1/2
inch perforation. This is by no means the case for all
production velocities, and bigger perforations and
removal of cohesive material will require higher
values. Th~s .means tha.t in many cases fluid velocity
can be a drivmg factor m whether or not a well is
seen to produce sand.
Influence of stress
The first signs of sanding begin, in the material we
are using, at a radial stress of 6 MPa. This is
slightly lower than the stress at which the first signs
of damage are visible in a test with no flow at all.
This strengthens the view that the far-field stress
alone produces damage in the rock around the
perforation, and the axial flow causes sanding by
removing and transporting the damaged material.
This indicates that the problem of sanding
potentially can be split into the simpler ones of hole
yield/stability and sand transport. This conclusion
was also reached by Vriezen et alII for gas wells,
and by Tronvoll et al 9.
Much effort has been devoted in recent years to
models of wellbore deformation, including non-
linear elasticity, plasticity, anisotropy and
localization, and considerable progress has been
made in predicting post-yield behaviour. Models
that incorporate softening, however, are still in their
infancy, and so it is not possible at present to predict
the reduction in cohesion, as a result of yielding, of
the material around the wellbore or perforation
tunnel. Knowledge of this reduced cohesion is likely
to be necessary to allow prediction of the conditions
under which a given axial flow will initiate erosion
and sanding.
Instructive predictions can nevertheless be made,
with a suitable choice of constitutive model to
represent the rock. A model incorporating strain-
hardening plasticity 16, with input parameters
derived from triaxial testing of the synthetic rock,
has been used to predict the onset of the behaviour
seen in the experiments. For a plastic strain value at
the wellbore ofO.OI, the confining pressure is
predicted to be 3 MPa. This is considerably less
conservative than the 1 MPa predicted on an
elastic/brittle basis, but still falls short of the 6 MPa
observed in the experiments. This is presumably
because the model does not incorporate the
softening of the rock following loss of cohesion.
Development of models that can do this is in
progress.
Prediction of sanding
Until such models are established, predictions of
sanding stresses can be made empirically, when core
is available, using the thick-walled cylinder collapse
pressure as proposed by Vriezen et al ll and Veeken
et a16. The decoupling of sand production into a
stress problem and a fluid flow problem means that
this approach is valid in providing a necessary but
not sufficient condition for sand production. The
thick-walled cylinder test may predict failure of the
perforation wall under the stress conditions imposed
by the drawdown, but fragments released from the
wall will not be seen as produced sand unless the
flow velocity in the perforation is high enough to
transport them, and continued sanding will not occur
unless the velocity is high enough to scour loose
material from the wall. Since we do not have a
direct measurement of fluid velocity in the
perforation, it must be calculated from the flow rate
of the interval and the number and size of the
flowing perforations. Good cleanup of as many
perforations as possible, and a knowledge of how
many are actually flowing, are obviously important
in controlling the flow velocities in individual
perforations.
477
CONCLUSIONS
Experiments on weak, very permeable synthetic
sandstone have shown that radial flow of liquid into
a perforation tunnel has little direct effect on sand
production. Flow along the axis of the tunnel
influences sand production strongly, and sanding
can be observed with no radial flow at all.
This is because the much greater velocity of the
axial flow is better at removing damaged material
from the sand face, and this removal is necessary in
order to maintain the evolution with time of the
plastic zone around the tunnel, and so maintain sand
production.
This means that for many situations the problem of
sanding can be split into a stress problem,
determining when a specified state of low cohesion
is produced at the sand face, and a flow problem,
determining how high a fluid velocity can be
tolerated in the tunnel without scouring damaged
material and stimulating further damage. Larger
perforation tunnels and a greater number of active
perforations will clearly help to increase the
maximum sand-free production.
REFERENCES.
1. Tixier, M.P., Lovelss, G.W. and Anderson, R.A.
"Estimation of formation strength from the
Mechanical-Properties log". 1. Pet. Tech., March
1975,283-293. SPE paper no. 4532.
2. Risnes, R., Bratli, R.K. and Horsrud, P. "Sand
stresses around a wellbore". Soc. Pet. Eng. J.,
Dec. 1982, 883-898. SPE paper no. 9650.
3. Morita, N., Whitfill, D.L., Massie, 1. and
Knudsen, T.W. "Realistic sand prediction
prediction: numerical approach". Presented at
62nd. Ann. Tech. Conf. of the SPE, Dallas, TX,
Sept. 1987. SPE paper no. 16989.
4. Morita, N., Whitfill, D.L., Fedde, 0.P. and
L~vik, T.H. "Parametric study of sand
prediction: analytical approach". Presented at
62nd. Ann. Tech. Conf. of the SPE, Dallas, TX,
Sept. 1987. SPE paper no. 16990.
5. Vaziri, H.H. and Byrne, P.M. "Analysis of
stress, flow and stability around deep wells".
Geotechnique, 40:63-77, 1990.
6. Veeken, C.A.M., Davies, D.R., Kenter, C.J. and
Kooijman, A.P. "Sand production prediction
review: developing an integrated approach".
Presented at 66th. Ann. Tech. Conf. of the SPE,
Dallas, TX, Oct. 1991. SPE paper no. 22792.
7. Weingarten, J.S. and Perkins, T.K. "Prediction
of sand production in gas wells: methods and
Gulf of Mexico case studies ". Presented at 67th.
 Ann. Tech. Conf. of the SPE, Washington, DC, Axial load
Oct. 1992. SPE paper no. 24797. Perfomtion
Pressure ............ ... ...
vessel tunnel
8. Kessler, N., Wang, Y. and Santarelli, FJ. "A

/~
simplified pseudo-3D model to evaluate sand
production risk in deviated cased holes".
Presented at 68th. Ann. Tech. Conf. of the SPE,
Houston, TX, Oct. 1993. SPE paper no. 26541.

9. Tronvoll, J., Morita, N. and Santarelli, F.J.

"Perforation cavity stability: comprehensive
laboratory experiments and numerical analysis".
Presented at 67th. Ann. Tech. Conf. of the SPE,
Washington, DC, Oct. 1992. SPE paper no.
24799.

10. Kooijman, A.P., Halleck, P.M., de Bree, P. and

Kenter, C. "Large scale laboratory sand
production test". Presented at 67th. Ann. Tech. monitor pressure
Conf. of the SPE, Washington, DC, Oct. 1992.
SPE paper no. 24798. Upstream
pressure Pump and flow
control valves Sand
11. Vriezen, P.B., Spijker, A. and van der Vlis, trap
A.C. "Erosion of perforation tunnels in gas
wells". Presented at 50th. Ann. Fall Meeting of Figure 1: initial configuration of sanding
the SPE, Dallas, TX, Sept. 1975. SPE paper no. experiments.
5661.

12. Raudkivi, A.J. Loose Boundary Hydraulics,

Pergamon, Oxford, 1976. Pressure

13. Carson, M.A. The Mechanics of Erosion,

Pion, London, 1971.

14. Govier, G.W. and Aziz, K. The Flow of

Complex Mixtures in Pipes, Robert Krieger,
Malabar, 1982.

15. Durrett, lL., Golbin, W.T., Murray, J.W. and

Tighe, R.E. "Seeking a solution to sand control",
Rock Beal
J. Pet. Tech., Dec. 1977, 1664-1672. SPE no. t Radial sample pack
6210. flow
Sand
16. Bradford, I.D.R. and Cook, J.M. "A semi- trap
analytic elastoplastic model for well bore stability
with applications to sanding". Presnted at ,...--~
EUROCK 94, Delft, Aug. 1994. SPE no. 28070.

Upstream
Pump and flow
pressure
control valves

Axial
flow

Figure 2: configuration of cylindrically symmetric

sanding experiments with independent radial and
axial flows, and independent radial and axial
stresses.

478
 A 4 60

Q) 3.5
50 .J:J
.='5E 3
40
a.
.e-
(;; 2.5 Q)

~
Q)
2 30 ~
'=- Cl
1.5
20 :6c:
<tI
10 (J)
~ 0.5
o
i:L 0
o 500 1000 1500 2000 2500 3000

Time/seconds
B 4 30

Q) 3.5
'5 25 .J:J
c:
'E
3
20
a.a.
en 2.5 Qj
~ 2 15 ~
1.5 Cl
10 c:

5
"c:
(J)
<tI
~ 0.5
o
i:L 0
o 500 1000 1500 2000 2500

Time/seconds

- Axial flow rate - Radial flow rate - - - - Sanding rate

Figure 5: sanding rate (in pounds per thousand barrels) and axial and radial flow rates versus time for
:onstant stress tests in cylindrical geometry. Test A was carried out under 6.5 MPa, test B under 7.2 MPa.

flo ... rate/litres.min- 1- . , 140

2.0
Stress~
IIPa

~
Fluid Far-field radial stress
3.0 velocity IMP.
/m.= ·1
Figure 4: Sand production rate (in pounds per
thousand barrels) versus radial stress and fluid
velocity along the perforation. For each velocity
".0 value, darker columns refer to first 30 seconds under
those conditions, and lighter columns refer to 60-90
second period. Sample Ues is approx. 2.5 MPa.

:>.0
I
Figure 3: rate of sanding from samples in equipment
shown in Fig. I. Samples have ues of about 2.5
MPa. See text for details.

479
 A-----=~l:.

Axial
flow rate
llitres.min -1
Radial flow rate
Jlitres.min -1
B

Axial
flow rate
Jlitres.min ~ 1
Radial flow rate
Jlitres.min -1
Figure 6: I1leao Sanding rare over. 21Ilinure interval
(in POunds per thousand banels) versus ""i.1 and
radial flow rates for the tests shoWn in Figur., 5. A
White square I1leaos that no data Were taken under
the corresponding set of Conditions.

480