Cement Plugs: Stability and Failure by Buoyancy-Driven Mechanism
Cement Plugs: Stability and Failure by Buoyancy-Driven Mechanism
J. P. Crawshaw, Schlumberger, and I. Frigaard, Schlumberger Dowell
Setting cement plugs off bottom is a well recognised problem
in the industry. Different causes of failure have been
identified over the years and many have been addressed by
improving practices used in the field. A remaining problem is
the instability of the lower interface between the cement and
the fluid below, driven by the greater density of the cement.
That is, once a cement plug has been injected into the wellbore
it must resist the tendency to fall through the less dense
drilling fluid below until the cement has set. The fluid
properties which ensure stability have never been satisfactorily
established and it is this gap which is addressed by the current
work.
Theoretical and experiment results are presented in this
paper which allow the estimation of the yield stresses required
of the fluids to stabilise a cement plug under any conditions of
hole size, inclination and density difference between the
cement and the fluid underneath. The experimental work
shows that the model is somewhat conservative and this is
discussed in terms of the assumptions made in each case.Introduction
Cement plugs can be placed either through tubing or by a
dump bailer. The dump bailer has limited applications and
only the placement of a balanced plug through tubing will be
discussed in this paper.
Plugging oilwells is a very common operation. Every well
must at some time be abandoned and this usually involves
setting cement plugs at several depths1. Renewed interest in
recovering more oil from well established fields has lead to an
increase in sidetracking from existing wells and this
commonly requires a cement plug to be set to kick off from.
Problems in drilling a new well are also sometimes addressed
by using cement plugs, for example to cure lost circulation.
However, plug cementing is an operation which has long2
been neglected as a minor task in which several attempts to set
a successful plug are the rule and not the exception. A widely
quoted figure3,4 for the industry average success rate in setting
cement plugs is 2.4 attempts per successful plug. A more
recent survey5,6 of plug failure mechanisms in the North Sea
concluded that, for off bottom plugs with no mechanical
barrier below the cement, only 30% were complete failures,
although 70% had the top of cement more than 30 meters
below planned depth. In general the failure rate can vary
considerably according to local conditions.
The reasons for the low success ratio have been
investigated by several authors7,3,4 and best practice guidelines
have been suggested to deal with many of the operational
problems. Some of the problems encountered can be related to
the circumstances in which the plugs are set, which are often
adverse in that the plug must be set quickly in response to an
unplanned event, and other problems are a consequence of the
fluid properties. The volume of cement which must be placed
is frequently small compared to the total volume of fluids in
the well, particularly in deep wells8, and this can lead to
contamination problems, both inside the tubing and in the
annulus. Additional contamination can also be induced when
the placement pipe is pulled out of the balanced plug.
Eliminating the causes of plug failure mentioned above by
application of best practices still leaves one mechanism for
which the physics has not been understood well enough to
derive quantitative, preventative measures: buoyancy driven
flow, in which the cement channels down through the less
dense fluid below. This is a well known problem investigated
in a largely empirical way in the literature. Only one paper9
addresses the modeling of this mode of plug failure and that
analysis was restricted to vertical wells and for a particular
symmetric mode of failure.
One solution would be to place a mechanical device below
the desired bottom of cement5. However, this incurs an
additional expense and it may not be needed in every
circumstance, as some off-bottom plugs succeed in the field
with no attempt to provide a base to support the lower
interface.
Other authors have carried out experiments with a viscous
pill7 or reactive viscous pill10 placed below the cement to give
additional stability. The most recent of these studies11
describes extensive experimentation in model wellbores at
several angles between vertical and horizontal. However, the
stability criterion developed in this work was flawed and
cannot be used to predict when a cement plug will remain
stationary after placement.
The aim of this paper is to develop an understanding of the
physics of the buoyancy driven failure mode, such that
quantitative predictions can be made of the rheology required
of both the cement and the fluid below to prevent fluid
movement after the pumping phase of the plug placement
process has ended. The theory will be briefly outlined in the
next section followed by a description of new experimental
results which will be compared to the theoretical predictions.
Finally the limitations of the approach and the implications for
field practice are discussed.Theory
It is difficult, if not impossible, to define the shape of the
mud/cement interface at the end of the plug placement process
and, as we shall see, the shape of the interface influences the
conditions for marginal stability. The approach taken in the
modeling, therefore, was to consider a slumping exchange
flow. This flow pattern is the most common way4,11 for the
plug to flow, when it begins to fail in an inclined wellbore
under the influence of gravity. This flow pattern was also
frequently observed in the experimental work described in the
next section and an example is shown in Fig. 1. The field
experience of a cement plug that drills soft but produces hard
cuttings is also evidence for the exchange flow, since the
uncontaminated cement fills only part of the wellbore.
The flow pattern can be divided into three regions where
mud and cement co-exist at a particular crossection of the
wellbore and these are shown schematically in Fig. 2. As the
more dense cement slumps down the low side of the wellbore,
a long axial exchange flow region is created where the motion
of both fluids is almost parallel with the axis of the wellbore.
We assume no fluids enter or leave the well during this stage
of the operation. The cement that moves down the well is
therefore matched by the volume of mud displaced up the
well. It is this phenomena which we refer to as an exchange
flow. The exchange flow region is connected to the
undisturbed cement above and mud below by two transitional
regions in which the flow is more complex
The mathematical analysis seeks to define the conditions
under which the fluids in the wellbore cannot move in an
exchange flow. This gives a conservative estimate of the
stability condition, as other interface shapes may be more
stable, (e.g. an interface that is close to horizantal), but is a
valid upper bound in that any interface which begins to move
under the influence of gravity will, eventually, form into an
exchange flow unless the well is exactly vertical.
The simplest rheological model which captures the general
features of the flow of cement and drilling fluids is that of a
Bingham fluid or plastic. A Bingham fluid has a yield stress
and therefore will not flow until a certain stress is exceeded.
Note that a yield stress is required to prevent movement in the
inherently unstable case of a more dense fluid above a less
dense one, a large viscosity would slow down the motion but
not prevent it.Description of Experiments
In contrast to its mathematical simplicity, it is very difficult
experimentally to produce a stationary exchange flow
geometry with fluids of a known yield stress. The most
reproducible interface shape (and the one chosen for most of
the experimental work) was a flat interface perpendicular to
the axis of the wellbore.
This was constructed by half filling, in the vertical
orientation, a long closed-ended pipe with a dense fluid and
then continuing to fill with a lighter fluid such that the
interface remained sharp and level. Samples of the two fluids
were taken at the time of filling the apparatus so that the yield
stress could be measured using a vane rheometer15. We note
in passing that it is less helpful to report the yield point, YP,
which is extrapolated from high shear rate rheology and may,
therefore, inaccurately represent the very low shear behavior
of the fluid.
To test the stability of the fluids, the pipe was rotated so
that the more dense fluid was above the less dense at a given
inclination to the vertical, q. At the same time the yield stress
of the samples was measured. This procedure was employed
for both 2 and 4 inch diameter pipes. At larger sizes, such as 8
inch, it became difficult to invert the pipe and the procedure
was modified. During the small diameter testing it became
apparent that the vertical orientation was more stable than an
inclined one for the flat interface configuration. For the
largest diameter, therefore, the pipe was half filled with the
less dense fluid first and then the more dense fluid was slowly
injected using a flow diverter tool positioned just above the
horizontal interface. Once the desired yield stresses were
obtained in the two fluids the pipe was moved to the required
inclination and the stability of the interface observed.
The fluids used in the experiments were: (i) Xanthan gum
solutions in water, which are viscous and Newtonian at very
low shear rates (less than 10-3 s-1) becoming shear thinning at
higher shear rates. This was used as the model fluid when no
yield stress was required. (ii) RDS grade Laponite suspensions
in water, which were used when a yield stress was required.
Laponite is a synthetic clay, similar to natural clays such as
hectorite or bentonite, but with a much smaller particle size.
The advantages of the Laponite for these experiments were its
good transparency and the ability to control the yield stress
over a wide range. The yield stress was manipulated by
changing the salt concentration of the 6 wt% Laponite
suspension and the time for which it was left static before an
experiment, as the yield stress continues to build over several
hours. This allowed small variations in the yield stress to be
made between experiments while using the same batch of
fluid. One of the fluids was weighted as required by the
addition of barite to generate the required density difference.
To obtain the marginally stable yield stress for a given Dr,
D and q a series of experiments was conducted with
progressively larger yield stresses until the interface remained
stable at the chosen inclination. Each series of experiments,
therefore, lead to a single point dividing stable from unstable
interfaces.
Experimental Results
Observations of failure modesIn most cases, when the pipe was inclined at more than a few
degrees away from vertical, the flow pattern during the
buoyancy driven failure was that of an exchange flow as
shown in Fig. 1. The vertical case was interesting in that the
flow pattern was much more complex and, as observed by
others16, followed a “roping” or helical pattern. However, this
is of little real world relevance as few wells are so close to
vertical.Comparison to Theoretical Results
The experimental values for the marginally stable yield
stresses in both fluids, as a function of angle and for all three
pipe sizes, are shown in Fig. 5. In this figure the view point
has been changed from that of Fig. 4 so as to reveal the
experimental points, which all lie below the theoretical
surface. From this viewpoint the experimental points are well
clustered and they follow the general shape of the theoretical
surface. This confirms that the initial selection of
dimensionless groups, equation (1), was appropriate and
should be regarded as a partial validation of the theory.
The degree of conservatism in the theoretical surface can
be seen in more detail in Fig. 6 in which the ratio of the
experimental yield stress modulus to the theoretical yield
stress modulus, y,exp y,u t /t , is plotted against the pipe
inclination. The experimental modulus varies between 70%
and 20% of the theoretical value with the better agreement
towards the horizontal orientation. To interpret these results
recall that the experiments were performed with a flat interface
perpendicular to the wellbore axis and that the theory was
developed for the axial exchange flow. Intuitively, the
interface shape which will be most unstable at any given pipe
inclination is that which lies parallel to the direction of gravity,
whereas the interface perpendicular to gravity will be most
stable. This trend is clearly seen in Fig. 6 and accounts for
some of the variation in conservatism. To test this idea two
additional experiments were carried out with the interface
initially at 45° to the pipe axis and these points are also shown
in Fig. 6. With the pipe horizontal, the change in the interface
had little effect on the marginally stable yield stress.
However, with the pipe vertical the experimental value was
much closer to the theoretical.
The influence of the distribution of yield stress between the
two fluids is examined in Fig. 7 where the theoretical and
experimental results are plotted for pipe inclinations of 10°
and 45°. This is equivalent to two slices through the marginal
stability surfaces shown in Fig. 5. The weak dependence of
the sum of the two marginally stable yield stresses on j can be
seen in the experimental data. As in the theoretical predictions
the effect is stronger at lower values of q.
Implications for Field PracticeThe comparisons outlined above give considerable
confidence that equation (6) is a valid representation of the
marginal stability surface. From the limited number of initial
interface shapes considered experimentally, it seems that the
theory over predicts the yield stresses by around 30%, a
reasonable engineering safety margin. However, in applying
the results to the field situation a number of additional
difficulties are encountered.
Firstly, the Bingham representation of the fluids in the
wellbore is a simplification and many real wellbore fluids have
a time and/or shear history dependant rheology. This together
with the pressure and temperature dependence of the rheology
make it difficult to estimate with precision the yield stress of
fluids thousands of meters below the surface.
Secondly, in all this work we have assumed that the
interface remains sharp and no significant mixing processes
take place. Further work is required to quantify the influence
of mixing during the pumping stage, particularly when an
open ended pipe is used and the cement emerges as a jet down
into the wellbore which must turn around before starting to fill
the annulus.
It is possible that mixing processes bring additional
stability to the field placement situation through two
mechanisms: a smearing out of the step in density locally
reduces the buoyancy driving force and incompatibility of the
mud and cement may result in a higher yield stress in the
mixed region than in either of the two fluids alone. This may
explain some of the successes in the field when stability
appears doubtful from a theoretical point of view.Conclusions
Setting cement plugs off bottom is a well recognised problem
in the industry. Many of the causes of failure can be addressed
by current best practice guidelines with the exception of the
swapping of the cement with the fluid below under the
influence of gravity when a significant density difference
exists. Buoyancy drives the interfaces between the cement
and the fluid below to deform as the cement slips down and
the mud up the wellbore in a stratified exchange flow.
Modeling and experimental work have established the
rheology required of the two fluids such that they will remain
stationary once the pumping phase of the plug cementing
operation is complete. This can be used to check the stability
of a given cement plug once it is pumped into the wellbore or
to design a fluid pill to correct an inherently unstable situation.Acknowledgements
We thank the management of Schlumberger for permission to
publish this paper.