Professional Documents
Culture Documents
1 s2.0 S092041050300024X Main - 1
1 s2.0 S092041050300024X Main - 1
www.elsevier.com/locate/jpetscieng
Abstract
A comprehensive geomechanical approach to wellbore stability requires knowledge of rock strength, pore pressure and the
magnitude and orientation of the three principal stresses. These parameters are often uncertain, making confidence in
deterministic predictions of the risks associated with instabilities during drilling and production difficult to assess. This paper
demonstrates the use of Quantitative Risk Assessment (QRA) to formally account for the uncertainty in each input parameter to
assess the probability of achieving a desired degree of wellbore stability at a given mud weight. We also utilize QRA to assess
how the uncertainty in each parameter affects the mud weight calculated to maintain stability. In one case study, we illustrate
how this approach allows us to compute optimal mud weight windows and casing set points at a deep-water site. In another case
study, we demonstrate how to assess the feasibility of underbalanced drilling and open-hole completion of horizontal wells
utilizing a comprehensive stability analysis that includes application of QRA.
D 2003 Elsevier Science B.V. All rights reserved.
Keywords: Wellbore stability; Quantitative Risk Assessment; Underbalanced drilling
1. Introduction
Mechanical failure of a wellbore is a result of the
interplay between in situ stress, rock strength, and
engineering practice. While a number of techniques
have been developed to predict optimal operational
parameters such as mud weights or drilling trajectories, these techniques have been limited to deterministic analyses that are based on the assumption that in
situ conditions and rock properties are precisely
known. In reality, geomechanical parameters are never
known precisely, due to insufficient data and the need
* Corresponding author.
E-mail address: moos@geomi.com (D. Moos).
0920-4105/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0920-4105(03)00024-X
98
99
orientations of tensile wall fractures and the orientations and widths of breakouts (e.g., Moos and
Zoback, 1990; Peska and Zoback, 1995).
Fig. 1 shows a typical example of the application
of QRA, where the input parameter uncertainties are
given by probability distribution functions (curves
limiting the shaded areas in the figure) that can be
conveniently specified by means of the minimum,
the maximum, and the most likely values of each
Fig. 1. (a) Probability density functions (smooth, shaded curves) and the sampled values used in the QRA analysis (jagged lines) as defined by
the minimum, most likely, and maximum values of the stresses, the pore pressure, and the rock strength listed in Table 1. These quantify the
uncertainties in the input parameters needed to compute the mud weight limits necessary to avoid wellbore instabilities. (b) Resulting minimum
(quantified in terms of the likelihood of preventing breakouts wider than a defined collapse threshold) and maximum (to avoid lost circulation)
bounds on mud weights at this depth. The horizontal bar spans the range of mud weights that ensure a greater than 90% likelihood of avoiding
either outcomeresulting in a minimum mud weight of 12.4 ppg and a mud window of 0.75 ppg.
100
parameter (Table 1; Fig. 1a). The probability distribution functions shown here are either normal or
log-normal curves depending on whether the minimum and maximum values are symmetrical (e.g., Sv,
SHmax, Shmin, and Pp) or asymmetrical (e.g., Co)
with respect to the most likely value. In either case,
the functional form of the distribution is defined by
the assumption that 99% of the possible values lie
between the maximum and minimum input values.
Once the input uncertainties have been specified,
response surfaces for the wellbore collapse and the
lost circulation pressures can be defined. These
response surfaces are assumed to be quadratic polynomial functions of the individual input parameters.
Their unknown coefficients in the linear, quadratic
and interaction terms are determined by a linear
regression technique that is used to fit the surfaces
to theoretical values of the wellbore collapse and lost
circulation pressures. The theoretical values are calculated for multiple combinations of input values that
are selected according to the representative design
matrix based on the minimum, maximum and most
likely values. The calculations discussed in this paper
assume that the rock behaves elastically up to the
point of failure. The analysis can be generalized to
model poroelastic responses as well as chemical and
thermal interactions between the mud and the rock.
After the response surfaces have been determined,
Monte Carlo simulations can be efficiently performed
to establish uncertainties in the wellbore collapse and
the lost circulation pressures. Ten thousand random
values of each input parameter are generated (jagged
curves in Fig. 1a) using a mean and standard deviation
identical to that calculated for the appropriate probability density function (shaded). The 10,000 random
numbers then enter the polynomial response surface
Table 1
Minimum, most likely, and maximum values of the input data used
in the QRA analysis shown in Fig. 1
Parameter
Minimum
value
Most likely
value
Maximum
value
Sv, ppg
SHmax, ppg
Shmin, ppg
Pp, ppg
UCS, psi
13.22
16.24
12.84
10.25
1000
13.92
17.10
13.52
10.79
1400
14.62
17.96
14.20
11.33
1470
101
Fig. 2. (a) Example showing a deterministic analysis of the mud weights required to prevent drilling problems (mud windows for each casing
interval are indicated by shaded rectangles), along with recommended casing set points to maintain an adequate mud window between the
collapse mud weight and the mud weight above which lost circulation may occur. (b) The cumulative likelihood as a function of mud weight to
avoid lost circulation and collapse while drilling the 6th casing interval. The shape of the upper bound is defined by uncertainties in the fracture
gradient at the previous casing shoe. The shape of the lower bound is defined by the collapse gradient at the bottom of the interval.
102
Fig. 3. Gamma ray, lithology (chert in darker, sandstones in lighter, and shales in intermediate grey), compressional-wave sonic transit time
(inverse velocity) and predicted rock strengths calculated from sonic and density data using the relationship from Moos et al. (2001).
103
Fig. 4. (a) Histogram of predicted strength values for the reservoir interval proposed for underbalanced drilling and open-hole completion. (b)
Log-normal probability distribution function for rock strength Co consistent with the variation shown in the histogram in (a).
Fig. 5. The shaded regions show Gaussian probability density functions that provide a measure of the uncertainties in the values (in ppg) of the
parameters comprising the geomechanical model used to predict the likelihood of success shown in Fig. 6 (the rock strength distribution is
shown in Fig. 4). The jagged lines show the values of these parameters actually used in the Monte Carlo simulations.
104
Fig. 6. Cumulative probability functions for wellbore collapse for the reservoir section of the well proposed for underbalanced drilling and
openhole completion, using a critical breakout width of (a) 30j and (b) 60j. The solid line corresponding to a mud weight of 8.2 ppg indicates
the pore pressure in the reservoir. Also shown is a vertical line that indicates a mud weight equivalent to a 1 ppg underbalance.
balanced drilling is possible (although a 1 ppg underbalance is likely to be risky), but without further
analysis open-hole completion is not recommended.
Fig. 7 shows the QRA response surfaces that reveal
the sensitivity of the predicted mud weight to the
uncertainty of each input parameter. The response
surfaces are quite flat for Sv, SHmax, and Shmin
magnitudes and for the pore pressure, suggesting that
these parameters are known with sufficient precision
not to require additional analysis. However, the predictions are extremely sensitive to the compressive
rock strength (Co). For the weakest rocks likely to be
encountered (i.e., Co f 3500 psi) a mud weight of 10
ppg is required. The strongest rock, in contrast, will be
stable even if the fluid pressure in the well is much
lower than the pore pressure in the reservoir. Thus the
critical parameter necessary to refine the predictions
of the lower bound safe wellbore fluid pressure is the
rock strength.
Fortunately, it is possible to acquire the data
necessary to measure the rock strength prior to completing the well using LWD and/or wireline log
measurements. Because the uncertainties in the predictions are essentially independent of uncertainties in
any parameters other than the rock strength, these
measurements provide the information necessary to
refine the results of QRA (which indicate that safe
completion requires sand control measures) after drilling the well but before running casing or screens.
105
Fig. 7. Response surfaces for individual geomechanical parameters that illustrate the sensitivity of the mud weight predictionsexpressed in
ppgassociated with each parameters uncertainty, as shown in Figs. 4 and 5.
106
Fig. 8. (a) Proposed casing design for a vertical well in deep water, showing mud weight windows (shaded rectangles) obtained using Pp + 0.5
ppg, and FG 0.5 ppg. (b) Revised mud windows for the original program, which include the wellbore collapse pressure (heavy dark line). (c)
An alternative casing program that honors the collapse and fracture gradient constraints and provides at least a 0.5 ppg mud weight window
throughout.
107
Table 2
Minimum, most likely, and maximum values of the input data used in the QRA analysis shown in Fig. 9
Parameter
Sv, ppg
SHmax, ppg
Shmin, ppg
Pp, ppg
UCS, psi
4668 ft
5894 ft
Minimum
value
Most likely
value
Maximum
value
Minimum
value
Most likely
value
Maximum
value
10.28
11.73
9.28
8.32
N/A
10.82
12.35
9.77
8.76
N/A
11.36
12.97
10.26
9.20
N/A
11.72
14.90
10.60
8.98
1660
12.34
15.70
11.16
9.45
1750
12.96
16.46
11.72
9.92
1840
Fig. 9. Analysis of the probability of avoiding drilling problems, as a function of the mud weight used, while drilling the second casing interval
shown in Fig. 8c based on the uncertainties in the stresses, pore pressure and rock strength shown in Table 2. The left-hand curve shows the
cumulative probability of avoiding collapse at the bottom of the interval. The right-hand curve shows the cumulative probability of avoiding lost
circulation at the previous casing shoe. The horizontal bar shows the range of mud weights between 9.3 and 9.8 ppg necessary to maintain at
least a 50% probability of avoiding both events.
108
required likelihood of success, it is possible to establish both the minimum mud weight necessary for safe
drilling and the mud window, honoring the data
uncertainties. In this case the horizontal bar indicates
that a mud weight of 9.3 ppg provides a 0.5 ppg mud
window with a 50% combined likelihood of avoiding
drilling problems.
A number of variables contribute to the uncertainty in the above analysis. In order to investigate
which of these is most important in defining the
lower bound of the mud window for this casing
interval, Fig. 10 presents a sensitivity plot of the
required mud weight to prevent collapse as a function of each input parameter. As can be seen, the
uncertainty in the vertical stress has no influence on
the results. Variation in the vertical stress between
11.72 and 12.96 ppg equivalent results in no change
in the 9.28 ppg mud weight required to stabilize the
well. The pore pressure has the single largest uncertainty and contributes the most to the uncertainty in
the required mud weight ( F 0.27 ppg). The uncertainty in UCS contributes a relatively small amount,
whereas the magnitudes of SHmax and Shmin each
contribute approximately F 0.15 ppg. Thus, a considerable additional measurement and analysis effort
is required to reduce the uncertainty in the collapse
pressure. On the other hand, the lost circulation
pressure in this interval is controlled only by fairly
5. Conclusions
The above discussion illustrates a new approach to
wellbore stability utilizing Quantitative Risk Assessment to predict the required mud weights to avoid
stability problems and the uncertainty in those predictions. This approach differs from previous methods
in that it allows determination of the uncertainties in
both the collapse and the lost circulation pressures,
and explicitly analyzes these in terms of all three of
the in situ stresses, the pore pressure, and the rock
Fig. 10. Sensitivity plots showing the effects of uncertainties in the input stresses, the pore pressure, and the rock strength on the minimum mud
weight (vertical axis, in ppg) required to avoid collapse over the second casing interval shown in Fig. 8c. With the exception of the vertical
stress, large uncertainties in the required mud weight are associated with all of these parameters.
109
References
Horsrud, P., 2001. Estimating mechanical properties of shale from
empirical correlations, SPE 56017. SPE Drilling and Completion, 68 73.
Huffman, A.R., 2001. The future of pore pressure prediction using
geophysical methods, OTC 13041. Presented at 2001 Offshore
Technology Conference, Houston, TX, 30 April 3 May.
Ito, T., Zoback, M.D., Peska, P., 2001. Utilization of mud weights
in excess of the least principal stress to stabilize wellbores:
theory and practical examples. SPE Drilling and Completion
16, 221 229.
McLellan, P.J., Hawkes, C.D., 1998. Application of probabilistic
techniques for assessing sand production and wellbore instability risks. SPE/ISRM, 47334.
Moos, D., Zoback, M.D., 1990. Utilization of observations of well
bore failure to constrain the orientation and magnitude of crustal
stresses: application to continental, deep sea drilling project and
ocean drilling program boreholes. Journal of Geophysical Research 95, 9305 9325.
Moos, D., Zoback, M.D., Bailey, L., 2001. Feasibility study of the
stability of openhole multilaterals, Cook Inlet, Alaska, SPE
73192. SPE Drilling and Completion, September, 140 146.
Ottesen, S., Zheng, R.H., McCann, R.C., 1999. Wellbore stability
assessment using quantitative risk analysis, SPE/IADC 52864.
Presented at the SPE/IADC Drilling Conference, Amsterdam,
The Netherlands, 9 11 March.
Peska, P., Zoback, M.D., 1995. Compressive and tensile failure
of inclined well bores and determination of in situ stress and
rock strength. Journal of Geophysical Research 100 (7),
12791 12811.
Raaen, A.M., Brudy, M., 2001. Pump-in/flowback tests reduce the
estimate of horizontal in-situ stress significantly, SPE 71367.
Presented at 2001 SPE Annual Technical Conference and Exhibition, New Orleans, LA, 30 September 3 October.
van Oort, E., Nicholson, J., DAgostino, J., 2001. Integrated borehole stability studies: key to drilling at the technical limit and
trouble cost reduction, SPE/IADC 67763. Presented at the SPE/
IADC Drilling Conference, Amsterdam, The Netherlands, 17
February 1 March.