SPE 59121-Rock Mechanics Met Drilling

1

When Rock Mechanics met Drilling: effective implementation of real-time wellbore stability control

Walt Aldred Anadrill, Sugar LandIan Bradford Schlumberger Cambridge ResearchJohn Cook Schlumberger Cambridge ResearchJohn Fuller GeoQuest, Gatwick

Abstract

The aim of the Real-Time Wellbore Stability Control project was to develop and demonstrate a tool to give advice tooil well drilling staff in real-time (i.e., while the well is being drilled) of optimum mud weights, drilling practicesand well trajectories. The project was supported by the European Commission as a THERMIE project.

It is aimed at providing recommendations on drilling parameters to optimise the drilling process and reduce the riskof unscheduled events or lost rig time caused by wellbore instability. The technique uses surface and downholemeasurements, while drilling, to make regular updates to a model of the wellbore, and revise the drilling planaccordingly.

The real time wellbore stability process first generates a mechanical earth model using information from offset wellstogether with field and regional data. The proposed well trajectory for the new well is projected into the mechanicalearth model and the stability is examined along the proposed well for a given initial drilling plan. The productidentifies potential danger zones within a well plan.

During drilling, real time measurements including LWD and MWD, surface mechanical measurements, and fluidsand solids monitoring, are used to diagnose the state of the wellbore. Any significant hole instability is detected anda warning given to the driller. The existing state of the wellbore is compared to the model and any revision requiredto align the predicted with the detected state is made. This real time update of the mechanical model is then used topredict the future state of the wellbore, in front of and behind the bit for the given drilling plan. If the drilling plancan then be improved, a revision will be recommended; for instance a reduction in the rate of penetration, increasein mud weight, mud circulation and, if required, hole direction. The driller can also use the product to evaluate hisown recommendations for changes to the drilling plan and then decide on the best course of action. Real timewellbore stability also provides a record of wellbore stability information which can be input to the field descriptionfor use in future wells, enabling continuous improvement of the drilling process.

The process has been evaluated on the Valhall field in the Norwegian sector of the North Sea. Valhall developmentdrilling has been problematic in recent years. A plan to increase recoverable reserves from 600m bbls by a further50% through extended reach drilling to down-flank targets has been largely unsuccessful, with many of these wellsbeing suspended or abandoned in the overburden or penetrating shortened reservoir targets. Costs/losses areestimated to be in excess of $150m. The problems have been associated with wellbore instability in the very weakoverburden.

The team produced a geomechanical earth model for the Valhall field, working closely with the drilling engineers todevelop a well plan for a proposed ERD well. Implementation involved providing wellsite support to co-ordinatemonitoring and detection of wellbore instability from real time data through Anadrill PERFORM, and on-linesupport in the Amoco drilling office to interpret data, update the mechanical earth model and revise the well plan.Through this process the team proposed and implemented a strategy of drilling the well in controlled states offailure - not a conventional drilling approach. The well successfully reached its target ahead of schedule, a string ofintermediate casing was not required, mud losses (a previous problem contributing to instability and cost) wereminimal and the well was cased to below the unstable overburden intervals.

2

1. Introduction

Wellbore instability is a major problem during the drilling of many oil and gas wells. It is often quoted as costingthe industry between 0.5 and 1.0 billion dollars per year. It currently leads to major difficulties in such diverse areasas the North Sea, Argentina, Nigeria and the Tarim Basin. A spectacular, recent and well-documented example ofthe cost savings available from improved handling of wellbore instability is available for the Cusiana field operatedby BP and partners in Colombia. Wellbore instability was very severe there, leading to costs per well of the order oftens of millions of dollars; an integrated approach to the problem led to very large reductions in these costs (Last etal, 1995). A fundamental of the approach used was to accept that wellbore instability was inevitable, and to try tomanage it rather than to stop it altogether.

The main way that the drilling industry has attacked wellbore instability in the past has been either on an ad-hocbasis, collecting data and cores for specific problem formations, or through log interpretation methods such asSchlumberger’s IMPACT™ and its predecessors (these are well-planning tools using rock strength and stresspredictions based primarily on sonic logging). A number of new technical factors now mean that a differentapproach is possible:• the availability of caliper measurements while drilling (e.g., Rosthal et al, 1991). These will not be discussed indetail, but briefly it is now possible to gather information on hole enlargement while drilling, using LWD (logging-while-drilling) resistivity or ultrasonic caliper tools.• a better understanding of wellbore deformation and failure mechanisms, and their relation to stress state (e.g.,Papanastasiou, 1990; Bradford and Cook, 1994);• improved understanding of how drilling practices (e.g., frequency of wiper trips, swab and surge pressures)influence instability, and of how, in turn, instabilities of different kinds influence drilling.

These factors mean that it is now attractive to use real-time measurements and interpretation to manage wellboreinstability (real-time here means essentially during the drilling of the well; some real-time data arrives immediatelyas a formation is being drilled, other data can be delayed by up to a few hours). The Real-Time Wellbore StabilityControl (RTWBSC) project is aimed at constructing and testing a system to do this.

Wellbore instability can have either mechanical origins (for example, failure of the rock around the hole because ofhigh stresses, low rock strength, or inappropriate drilling practice) or chemical origins (damaging interactionsbetween the rock, generally shale, and the drilling fluid) or a combination of the two. Both types of problem havebeen studied extensively at Schlumberger Cambridge Research (SCR) and many other places. The integration ofour understandings of chemical and mechanical damage is still, however, problematical, and so the first attempt at areal-time product in this area was focussed on the mechanical issues.

The proposed RTWBSC product will be centred on a software tool, running in near-real-time as a well is beingdrilled, that will accept input data from a wide range of sources and use these to detect, diagnose and predictwellbore instability, and to refine a geomechanical model of the subsurface. The starting point will be ageomechanical model derived from offset well data, perhaps based on a seismic dataset. It is therefore intended as aproduct for GeoQuest, but the real-time operation will use data from LWD, mudlogging and rig sensors as available,and so will have a significant input from Anadrill.

The work is partly funded by the European Commission, under the THERMIE initiative (contract number OG-0199-95). The partners in the EU project are Amoco UK, GeoQuest Gatwick, and NITG-TNO, a Netherlands contractresearch organisation. SCR was the co-ordinator of the project and was responsible for most of the technical input;Amoco UK provided seismic, drilling and log data from a demonstration field in the North Sea, and the opportunityto test the product on a well as it was being drilled; GeoQuest was responsible for managing, manipulating andinterpreting the data; and NITG-TNO for carrying out finite element analysis of wellbore deformations.

3

2. Structure of real-time wellbore stability process

The ‘traditional’ approach to wellbore stability planning is as follows:• log data (usually sonic travel time) is collected from offset wells, and this is used to generate a rock strengthprofile and an earth stress profile for the interval of interest. Very often the wellbore stability code itself does theinterpretation of the sonic log data to give strength, etc.• a mud weight plan is calculated for the proposed well, taking into account the strength and stress from the above,and the trajectory of the well. This plan shows, as a function of depth, the maximum static mud weight (i.e., themaximum mud density) that can be used in the well without fracturing the formation, and the minimum mud weightthat can be used without inducing failure of the rock around the borehole. Much effort has been put into thecomplex mechanics of plasticity around boreholes, in order to refine our estimates of the lower mud weight.• the mud weight plan is sent to the drilling team, who are in the process of designing the well.• the well is drilled, with varying degrees of attention paid to the mud weight plan.• when the well is finished, an end-of-well report is written. Problems encountered during the drilling are listedhere, and the wellbore stability planning staff try to understand these and modify the plans for the next well.

The Real-Time Wellbore Stability Control project used a new approach:• all the available and appropriate information from the field in question is examined. Concrete data such asgeological structure, lithology, rock strength and earth stresses are incorporated into a model of the area, which wascalled the Mechanical Earth Model (MEM). This is consistent with Schlumberger’s overall approach to oilfielddata, which is to incorporate it all into a Shared Earth Model accessed by many applications. The MEM contains thebest available values of the data needed to estimate wellbore instability, i.e., strength values rather than sonic logvalues. Because of this, these data can come from a variety of sources, including lab testing for strength andfracturing or other field methods for stress estimation; input data are not restricted to those obtained from logs. TheMEM can be 3-dimensional, although this was unnecessary in this project.• Less precise or non-numeric data, such as drilling problems in a particular formation, or the presence of highbackground gas levels, are also incorporated into the MEM.• the trajectory of the well is projected through the MEM, and the relevant data extracted along it.• a relatively simple wellbore stability calculation is carried out using the trajectory and MEM data, to give the mudweight plan along the length of the well. Rather than making complex calculations using, for example, plasticitymodels, these calculations are made using an elastic model with a simple failure criterion. In general the input dataare not available for plasticity models, and even if they are it is usually computationally impractical to run aplasticity model over the entire length of a well.• this initial mud weight plan is discussed with the drilling team, to determine whether there will be any difficultiesin following it, and what might be done during drilling if it cannot be followed, or if it is simply wrong (as a resultof incomplete or incorrect input data).• during the drilling of the well, the wellbore stability team monitors the progress of drilling, with emphasis onpotential or actual wellbore instability problems, and on monitoring the data coming from the well. These data areinterpreted to give information about the state of the wellbore, and the earth model; the well plan is updated in realtime as the problems arise, and modifications are communicated to the drilling team.• changes to the trajectory, because of new drilling priorities or poor directional control, are immediatelyincorporated into the plan, and the mud weight limits recalculated.• the state of the wellbore, and the mud weight plan, are continuously presented to the driller in an easily-digestiblevisual format, rather than the depth-based well-log plots used in the past.

This process is summarised in Figure 1, and the heart of it, embodied to a large extent in the software package, isshown in Figure 2.

4

Figure 1: the design-execute-evaluate cycle for real-time wellbore stability control. The starting point is at the top,with initial data-gathering and construction of the first MEM being done in the planning phase. The remainder of

the cycle occurs as the well is being drilled.

NY

Real-timedata

Wellbore model

Predict existing wellbore state

Detect existing wellbore state

Diagnose existing wellbore state

Agree?

Compare

Refine drilling model

Recommend drilling model

Display

Operator’s recommendations

Predict future wellbore state

Drilling model

Minimumcost?

Predict future wellbore state

Y

N

Refine wellbore model

Fig. 2. The flow of information and decisions through the prototype system. Ellipses represent datainput, diamonds are decision or comparison points, and rectangles are processes. The starting points are

the two upper ellipses, and the finish point is the lower left-hand corner.

5

3. The Real-Time Wellbore Stability process

3.1 Data gathering

The original intention was that the RTWBSC process be tested on a well in the Everest Field of the UK sector of theNorth Sea. Data from a large number of wells was assembled, including log data, mechanical testing results,seismic surveys, and end-of-well reports. Most of the information was, as is usual in oilfield operations, from thereservoir sections (where wellbore instability problems are rare) but GeoQuest and Amoco UK were able to carryout petrophysical and geomechanical analysis of overburden data from offset well 22/23a-C3 operated by Amoex.The overburden sequence in this well is representative of the overburden encountered on the Everest structure andthe data were composited with Everest reservoir data. This provided a profile of the Everest stratigraphic sequenceto allow the building of the geomechanical field model.

At this point it emerged that development drilling on Everest was unlikely to continue within the timeframe of theproject, and so there would not be a well available there to test the RTWBSC process. An alternative was sought,and a proposal was made to run the process on a well in the Valhall Field operated by Amoco Norway. This wasaccepted by Amoco Norway, and agreement to this change was obtained from the EC in April 1998.

Appendix A describes the geology of the Valhall Field. Briefly, it is a very large chalk reservoir overlain by thickTertiary shale. Two central platforms have carried out extensive drilling to access the central reserves, but extendedreach drilling to reach the flanks has been much more difficult because of wellbore instability problems.

Partly because of the recent history of drilling problems, a good deal of data already existed for Valhall, and theprocess of assembling a Mechanical Earth Model proceeded quickly. The components of the model (and the sourcesof the data) are as follows:• rock strength and friction angle (from Amoco proprietary correlations to sonic log data; other non-proprietarycorrelations exist);• Poisson’s ratio (sonic data);• vertical stress (integration of density logs);• minimum horizontal stress (a biaxial strain model adjusted to fit leak-off and lost-circulation data);• maximum horizontal stress (from a model assuming the formation to be at failure);• pore pressure (Amoco’s model, from undercompaction and kick analyses);• formation tops (from seismic data and logging of offset wells);• drilling problems (from end-of-well reports and discussions with the drilling teams).

The numerical data in the MEM for the northern part of Valhall, where the field test well was located, is shown inFigure 3. Figure 4 shows the non-numerical data, and a 2-D representation of the trajectories planned for the testwell, A3C.

3.2 Wellbore mechanics

The stress state and deformation of a wellbore is influenced by the in-situ stress state in the formation, the rock’smechanical properties, and the angle of the wellbore axis relative to the directions of the principal stresses in theformation (i.e., in most cases its deviation from the vertical and its azimuth). For an isotropic elastic material, thestate of stress around a deviated borehole can be calculated analytically and rapidly; if this state of stress iscompared to a Mohr-Coulomb or other failure criterion an approximate condition for borehole failure can beobtained. This approach, however, does not take into account the effects of plasticity of the rock, which can reduceshear stresses around the hole and so lead to a more stable structure. For deviated wellbores with plastically-deforming rock, a finite element approach is needed, and NITG-TNO used DIANA, a commercial 3D finite elementcode, to study this geometry. Finite element plasticity computations tend to be time-consuming, and so it was notpracticable to investigate a large number of combinations of the input variables, which include:

6

• the values of the principal stresses• the orientation of the wellbore relative to the principal stresses• the strength, elastic properties and plastic properties of the rock• the criterion used to decide whether the wellbore has failed.Because of this, it was decided to run a single rock type, with a single stress state, and vary just the well orientation,the mud pressure, and the failure criterion. The intention was to find a calibration factor - a number by whichelastic, analytical solutions for wellbore failure should be multiplied to include the effects of plasticity.

Some results from the DIANA computations are shown in the remainder of this section. The details of these resultsare not important, but the conclusions drawn from them are. These are set out in section 3.3.

The model consists of three layers; the middle layer uses an elastic-ideally plastic Mohr-Coulomb material model,while the other layers behave elastically. This has to be done to reduce boundary effects at the top and bottom planesaround the borehole. The pressure in the hole is stepwise reduced from a starting value which is just above the valuecalculated by solving analytical equations defining the stress state when the material is at yield.

Figure 3: Mechanical Earth Model for the northern part of the Valhall Field. The vertical axis is true vertical depth,in metres. No data were available for the top 400 metres of the section.

7

Figure 4: True vertical depth (TVD) versus measured depth(MD) for the initial planned trajectory of A3C (in red)and the final plan (in blue), which on the scale of this figure was close to the actual drilled trajectory. See the textfor the reasons for the change in trajectory. The figure also shows the formation names in the Tertiary and chalksections, and on the right, the features that were expected to lead to wellbore instability and stuck pipe problems.

Yellow means medium severity, red means very severe.

The parameters used in the model are as follows:

Young’s Modulus E = 3700 MPa Vertical effective stress = 37.32 MPaPoisson's ratio = 0.35 Max. horizontal effective stress = 24.88 MPaCohesion = 11 MPa (central layer) Min. horizontal effective stress = 12.44 MPaFriction angle = 13 degreesDilatancy angle = 13 degrees

• Case 1: azimuth of the maximum horizontal stress is 0°

In this case the maximum horizontal stress lies in the same vertical plane as the borehole axis. Results for thedifferent inclinations are presented in Figure 5. Note that 0° is here the vertical borehole. The failure criterion to beused here is a critical value of the maximum plastic strain achieved anywhere around the hole, and so the resultsplotted are maximum plastic strain versus mud pressure, for different orientations.

8

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0 500 1000 1500 2000 2500 3000

Mud density kg/m^3

Max

. prin

cipa

l pla

stic

str

ain

30

60

90

0

45

Figure 5: Maximum principal plastic strain versus mud density for different borehole inclinations. Azimuth ofmaximum horizontal stress = 0°, i.e., maximum horizontal stress lies in same vertical plane as axis of borehole.

• Case 2: azimuth of the maximum horizontal stress is 90°

In this case the minimum horizontal stress lies in the same vertical plane as the borehole axis. The results are shownin Figure 6.

0.00E+00

5.00E-03

1.00E-02

1.50E-02

2.00E-02

2.50E-02

3.00E-02

3.50E-02

4.00E-02

4.50E-02

0 500 1000 1500 2000 2500 3000

Mud density (kg/m^3)

Max

. prin

cipa

l pla

stic

str

ain

90

60

45

30

0

Figure 6: Maximum principal plastic strain versus mud density for different borehole inclinations. Azimuth ofmaximum horizontal stress = 90°, i.e., minimum horizontal stress lies in same vertical plane as axis of borehole.

9

Table 1 shows how wellbore orientation affects the mud pressure required to reach a failure criterion of 1% plasticstrain.

angle of inclination mud density (kg/m^3)az=0°

mud density (kg/m^3)az=90°

0° 630 65530° 1175 28545° 1530 65060° 1860 113090° 2090 1465

Table 1: mud densities for the two cases for 1% maximum plastic strain

The plateaux in Figure 6 indicate directly that the use of a maximum strain criterion is not straightforward and evennot applicable for this stress situation. The evolution of the plastic strain differs much from case 1. Initially the strainevolves in the same area as case 1, then it starts to distribute itself around the borehole when the pressure is lowered.

To illustrate this difference in evolution for the area of plastic strain, two snapshots from the two cases arepresented. Both snapshots show a normal cross section of a 45 degree inclined borehole. Looking from left to rightand from top to bottom, the pressure in the borehole has a value of 25 MPa, 17.5 MPa, 7.5 MPa and 0 MParespectively. The first snapshot (Figure 7) is from the borehole using the stress situation of case 1. The nextsnapshot (Figure 8) shows the borehole using the stresses of case 2. Here the maximum horizontal stress is normal tothe borehole axis.

Figure 7: Area of plastic strain at a mud pressure of 25 MPa, 17.5 MPa, 7.5 MPa and 0 MPa respectively, lookingfrom left to right and from top to bottom (case 1).

From these result, it is easy to see the different shape of the area where plastic strain occurs. The first plastic strainappears in the area as in case 1, but when the mud pressure is lowered, the area of plastic strain spreads around thehole. This leads to the inapplicability of the maximum plastic strain as a failure criterion. Following muchdiscussion of this topic, several possible criteria were formulated:• Maximum plastic strain limit (used here)• Maximum displacement limit

10

• Maximum angle of plastic strain around the boreholeThe problem is that all these criteria will give totally different results. It is therefore likely that different criteria mustbe used for different stress situations, and this begins to make this kind of detailed modelling impossible to apply inthe borehole environment, where conditions vary widely over the length of the hole.

Figure 8: Area of plastic strain at a mud pressure of 25 MPa, 17.5 MPa, 7.5 MPa and 0 MPa respectively, lookingfrom left to right and from top to bottom (case 2).

One of the aims of this part of the project was to generate a calibration factor, by which the critical mud weightderived from an elastic analysis should be multiplied in order to mimic the effects of plasticity. The calibrationfactor is the ratio between the values for the critical mud density derived from the numerical calculations and theanalytical equations. This ratio can be defined using different methods; two of them are presented here. The easiestway to define the calibration factor is to find the critical mud density for a point on the borehole where yielding firstoccurs (Fjaer et al., 1992). Then, this mud density will be compared with the mud density derived from thenumerical model in the same point when the plastic strain is 1%. The ratio between the mud density values definesthe calibration factor, hence

Calibration factor =mud density at 1% plastic strain (numerical) at point A

mud density when only point A is at yield (analytical)

This factor is always greater than 1. Another method to calculate the calibration factor is to use the same muddensity as found in the previous method from the numerical model and to compare it with the mud density when50% of the circumference of the hole is at yield (using the analytical expressions). Then,

Calibration factor =mud density at 1% plastic strain (numerical) at point A

mud density when 50% of circumference is at yield (analytical)

Point A is the point where maximum plastic strain occurs. Figure 9 shows both calibration factors as a function ofthe inclination. Curves are given for the two different azimuths of the maximum horizontal stress.

11

a)

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

0 10 20 30 40 50 60 70 80 90

Inclination (degrees)

Calibrationfactor

Az. sH = 0

Az. sH = 90

b)

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

0 10 20 30 40 50 60 70 80 90Inclination (degrees)

Calibrationfactor

Az. sH = 0

Az. sH = 90

Figure 9: Calibration factor v. inclination. a) method 1 b) method 2 (see text for explanation)

The calibration factor is clearly a function of azimuth, inclination and material parameters. Using other azimuths, a3D surface can be created for one material. Ideally, this has to be compared with other materials and relations haveto be sought. For the different lithologies these surfaces might be constructed and implemented in the algorithms ofthe wellbore stability prediction package. Another approach would be the comparison of the plastic strain criterionused in the numerical calculation with real breakouts. Then the breakout determines the level of plastic strain andhence the calibration factor.

12

This part of the project was very revealing, demonstrating the complexities of the plastic behaviour of a borehole.The aim of having a single value, or even a simple function, to represent the calibration factor appears to be over-ambitious and simplistic. It is also clear that the choice of failure criterion has a strong influence on the predictedmud pressure required for stability of the hole. There is no agreed failure criterion for a borehole, that can beexpressed in terms of stresses or strains; even if there was, it might well be dependent on orientation, azimuth,material type, etc. The idea of having complex plasticity models for hole stability within the algorithm wastherefore abandoned.

3.3 Rock failure and wellbore failure

This forced an important change in the way that borehole failure was perceived within the project, and this changehas been incorporated into the way that we think of wellbore stability in general. The new approach is as follows: itdoes not matter how badly the rock around the hole has deformed, or even whether the rock has failed - theimportant consideration is whether the borehole itself has failed as a piece of engineering. In drilling terms, theprimary function of the borehole as an engineering structure is to allow the bit and drillstring to be inserted andremoved, and to allow the casing to be installed. It has failed if it does not fulfil these functions; it has not failed if itdoes. (There are other functions of a wellbore such as being a way to make measurements with wireline loggingtools, where hole enlargement in itself might constitute failure, but access for drillstring and casing is undoubtedlythe most important). Figure 10 illustrates the difference between rock and borehole failure.

On the left is a borehole where the mud weight has been too low, or the earth stresses have been too high, or therock has been too weak. The rock has failed around the hole, become separated from the borehole wall, and theresulting cavings have been swept up the hole by the mud flow to the surface. In terms of rock mechanics orplasticity modeling, this hole has failed drastically; the rock has undergone enormous strains and displacements, hasbeen heavily fractured, and is far beyond the limits of continuum analysis. In terms of drilling, it is a good hole, inthat drillstring and casing can be passed through it without any problems.

On the right is exactly the same borehole, having been subject to exactly the same stresses and mud pressures. Theonly difference is that the broken material has not been swept away. This situation would be a major drillingproblem, and probably a terminal one; if the bottomhole assembly was below this section, it would be difficult to getit out, and if it was above, it would be difficult to re-enter the hole, perhaps leading to a sidetracking tendency.

There is no difference between the two sections in terms of rock mechanics or the traditional approach to wellboreinstability. The only change is that for the left-hand hole, the driller has been informed and diligent about hole-cleaning procedures, and maintained enough flow rate and viscosity in the drilling mud to remove the broken rock.The scientific boundary conditions are identical, but the driller has made the difference between a failed boreholeand a perfectly viable one.

This example demonstrates how important drilling practice, as well as rock mechanics, is in wellbore stabilitycontrol. Other situations can be described that tell the same message, and this re-appraisal of the important factorsin wellbore stability strongly influenced the direction of the project.

13

Figure 10: two sections though a hypothetical borehole, illustrating the nature of borehole failure. See text forexplanation.

3.4 Design of the RTWBSC software

The software was required to fulfil the functions outlined in Figure 2, and to run on a laptop so that it could be moreeffectively deployed at a client drilling office or rigsite, rather than in a data interpretation centre. The decision wasmade to write the code using Matlab, a commercial scientific computation and visualisation code from TheMathWorks Inc.

The code was planned to do the following:• use the MEM and trajectory to calculate the critical mud weights for shear failure of the rock around the wellbore,and for lost circulation• use MEM and trajectory to predict the condition of the wellbore under a given mud weight• use data acquired in real time to diagnose the condition of the wellbore• refine inputs to the predictive model in order to reconcile its predictions with the real wellbore state• refine the drilling model (mud weights, etc.) to minimise the cost or trouble of the drilling process, in the light ofthe refined predictive model• make recommendations for drilling practices• display the results in an easy-assimilated form.

Each of these will be described below.

3.4.1 Critical mud weight window calculationOnce the MEM and trajectory are loaded, the data can be used to calculate the mud weight below which plasticdeformation of the rock around the wellbore begins, and also the mud weight above which tensile fracturing occurs.This was done with an elastic model with a Mohr-Coulomb yield criterion. The mud weight window is calculated at0.25 m intervals over the entire trajectory, the yield function is calculated at 36 points around the wellbore wall ateach level, and a search routine is needed to find the minimum mud pressure that suppresses yield. Since there areabout 25,000 0.25 m intervals in a typical trajectory, this is a lengthy calculation, and required about 20 minutes forthe entire trajectory. Since model updating was not expected to occur rapidly, this was felt to be satisfactory. Thelost circulation calculation was much more rapid, since the lost circulation pressure was taken to be equal to theminimum principal stress. Figure 11 shows a typical mud weight window for well A3C from Valhall.

3.4.2 Wellbore condition under a given mud weight.This means applying a predetermined borehole pressure to each interval in the trajectory, and calculating the yieldfunction under those conditions. Since no search is involve, it is more rapid; calculating the wellbore state for the

14

2.5 metre length used in the main display takes around 1 second, and calculating the risk display for the wholetrajectory (see below) takes about 1 minute.

Figure 11: a mud weight window plot from Valhall well A3C. The vertical axis is measured depth, i.e., distancealong the well. The horizontal axis is the mud weight or density, which leads to the required borehole pressure at agiven depth. The green curve is the estimated pore pressure; the borehole pressure must not drop below this, for

safety reasons. The blue curve is the least principal stress in the earth; the borehole pressure must not exceed this, orlost circulation becomes likely. The red curve is the borehole pressure below which borehole failure becomes likely.

Figure 12 shows an example of the display of the predicted wellbore state along a 2.5 metre length. The wellbore iscorrectly oriented with respect to geographic directions and the vertical, shown by the red axes. The green axisshows the direction of one of the principal stresses. The display is rotatable using the mouse, so an all-round viewcan be obtained. The conditions for the prediction can be changed using the buttons around the display, and thedisplay updates immediately one of these is pressed. The blue parts of the displayed wellbore are where the rockstress state is predicted not to have exceeded the failure criterion. The red parts are where it has; further more theradial enlargement of the wellbore in the red zones is an indication of the extent to which the failure criterion hasbeen exceeded. This can be expressed as follows:

If• radius of display of undamaged wellbore is R• radius of display of damaged wellbore is r

15

• Mohr-Coulomb yield function is F (a linear function of maximum and minimum stresses, and unconfined rockstrength)• unconfined compressive rock strength is U

r = R*max(1, F/U + 1).Whenever r is greater than R, the surface is coloured red.

This is not a rigorous approach to calculating the size of the damaged borehole, or even the size of the predictedplastic zone. It is intended to give a quick and qualitative view of the severity of the predicted damage, and itslocation around the borehole (for comparison, for example, with imaging logs or with cavings morphology).Furthermore, by clicking the ‘drill’ button, the simulation can be advanced though the formation along thetrajectory, extracting the correct earth data and calculating the wellbore state at each step in measured depth, andgiving a view of ‘virtual drilling’. The main benefit of this is to focus the attention of the watchers on upcomingproblems.

Figure 12: detailed view of the predicted state of the wellbore under a given mud weight. See text for description.

Figure 13 shows a view of the entire well (once again in a rotatable 3D view) where the trajectory is colour-codedaccording to the local risk of wellbore instability. This is a view of the data that was consistently requested duringthe development phase, whenever the project was shown to drilling staff. The state of the wellbore is calculated inthe same way as described above, then the enlarged radius (i.e., the value of ‘r’) is integrated around thecircumference to give a damage value, and the particular line segment is coloured from green (for an undamagedsegment) to red (for a segment with damage beyond a threshold value). Carrying out this procedure for the entire

16

well requires a great many calculations, and to do them at the resolution of Figure 12 was far too slow. Thecalculations are therefore accelerated by carrying them out at every twentieth measured depth interval (i.e., every 5metres), and by taking 12 points around the wellbore rather than the 36 points used in Figure 12. This takes about 1minute on a 150 MHz laptop. Clicking on a point on the trajectory in this window generates the prediction in Figure12 for that point, so it is easy to form a view of the wellbore instability trouble spots using this system.

Figure 13: view of the entire well, colour-coded for risk of wellbore instability. See text for description.

3.4.3 Diagnosis of wellbore instability.The intention here was to construct an automated system to establish the real state of the wellbore from real-timedata, which would then be compared to the predictions of the model. This proved to be an over-ambitious goal.

The decision was taken to use a fuzzy logic approach, since this can deal with imprecise data. The data that can beused to arrive at a diagnosis include:• Downhole weight on bit • Downhole torque• Rate of penetration • Hookload• Surface torque • Cuttings/cavings rate• Cuttings/cavings morphology • Lithology• Mud losses or gains • Mud chemistry• Caliper • Gamma ray• Compressional slowness • Shear slowness

A deterministic approach to diagnosis would require inputs such as:• the downhole torque is 22% higher than the expected value• cavings generation rate is 14 kg/hour

17

• lithology is a fractured shaleand would also need an underlying mechanics model, which would be extremely complex.

The fuzzy logic approach can deal with inputs such as:• the downhole torque is rather high• cavings generation seems normal• I don’t know what the lithology is.It can be based on experience and on more qualitative models for instability and drillstring behaviour. The fuzzylogic is not more accurate than the deterministic approach, but is more tolerant and robust.

An algorithm was initially written using Java, and implemented via a World-Wide Web interface; after feedback onthis it was written into Matlab. Figure 14 shows the input interface, where qualitative information - for example,‘the pump pressure has decreased a lot’ - could be fed to the algorithm.

Figure 14: window for input of data to diagnosis system. The sliders are moved to high or low values according tothe judgement of the operator, rather than on a numerical basis. The 'present?' check boxes are for future use, wherenot all measurements might be available; the state of the check boxes is be used to construct a new diagnosis model

18

The output of the algorithm is a vector of 13 numbers, corresponding to 13 sources of wellbore instability problems;the magnitude of each component of the vector represents the severity of the mechanism. The 13 mechanisms are asfollows:

1. Breakouts2. Sloughing3. Natural fractures4. Drilling-induced fractures5. Fault crossing6. Fault activation

7. Undergauge hole8. Interbedded formations9. Overpressured formation10. Unconsolidated formation11. Mobile formation12. Highly permeable formation13. Chemically-active formation

A ‘dry run’ of the field test was held, to help to spot problems with the system before its use during drilling, and thisshowed that the diagnosis algorithm was not effective given the data coming in. There are good reasons forexpecting this (with the benefit of hindsight); essentially the fuzzy logic approach requires all the data channels usedin building the diagnostic model to be given a value, in order to make a valid diagnosis. It is not the case that allthese channels are always available during drilling, and a more suitable approach would allow for missing data.Building a new system, based on a different architecture (such as, for example, Bayesian belief networks), wasbeyond the resources and timescale of the project, and so one of the conclusions of the dry run was that we shouldrely on manual (i.e., human) diagnosis of instability problems during the field test rather than on the algorithm.

These difficulties have taught us that the diagnosis of wellbore instability in real-time is anextremely difficultproblem, compounded by complex but relatively uncertain physics and mechanics, the fairly limited range ofmeasurements that can be made at present, and the very wide variety of configurations of geology and drillingequipment that must be accommodated.

3.4.4 RefinementThe aim here was to update the Mechanical Earth Model to make its predictions more consistent with reality. Forexample, if breakouts are predicted, but none are seen, our initial interpretation was that either the real rock strengthis higher than the values in the MEM, or that the real stress state is lower. The choice of which parameter to updatewas to be made on the basis of confidence in the data; for example, if rock strength values were well-constrained bymeasurement, the stress state would be changed. In the case of the test on the Valhall well, we had confidence in therock strength values, both from previous laboratory work, and from the ISONIC tool data, which agreed well withpredictions. The ‘Refinement’ module was designed to accommodate these possibilities, and Figure 15 shows theinterface that was developed to allow this.

In the light of the discussion in section 3.3 on rock failure versus borehole failure, however, it is clear that there areother, non-model-related reasons why reality and predictions might not agree. For example, our evidence forbreakouts is only from observation of the cavings produced. When the model predicts breakouts, and none are seen,this may be because of poor hole cleaning rather than an incorrect rock mechanics model. When breakout are notpredicted, but they are seen, this may be because of cleaning of an old cavings bed by improved hole cleaningprocedures, or by intermittent breakouts generated by swab pressures during a trip. Very direct evidence is neededin order to refine the model; for breakouts, a caliper-while-drilling is required, to show whether or not the rock hasbroken at a particular spot. This technology is not yet available. The aspects of the model that can beunambiguously refined are the minimum principal stress and the pore pressure, on the basis of leakoff and kick datarespectively. The stress was refined during the field test (using the Matlab command line rather than the interface inFigure 15), but fortunately no kicks occurred and the pore pressure did not need to be refined.

3.4.5 Recommendations.The intention of this module was to examine the current state of the wellbore, and make recommendations fordrilling practice to minimise future problems.

The code takes the output vector from the diagnosis system and suggests actions to deal with the worst of thecomponents. It includes a calculation of trip speed, based on a hydraulic model of the wellbore and drillstring, tohelp avoid the generation of excessive swab or surge pressures, which can fail an already-damaged wellbore (forexample in fractured shale), and an integration of wellbore instability risk over the open hole section in order to

19

reach an optimum mud weight. An example of the output of the recommendations module is shown in Figure 16.The recommendations module worked well during the dry run, but was only used once or twice during the field test.

Figure 15: interface to the ‘Refinement’ module.

Figure 16: an output from the recommendations module, based on the recognition of naturally-fractured formationsas the dominant instability mechanism. The recommended actions are derived from drilling industry handbooks of

wellbore stability and stuck pipe control, together with a mathematical model for swab and surge pressuresdeveloped previously at SCR.

20

4 The Valhall field test

Amoco Norge agreed to run the field test on well A3C, in the Valhall Field. This would be an extended reach wellrunning north-west from the platform. An old well would be re-entered and plugged back, a window cut in thecasing string above the plug, and the new well drilled from there. The first sections will be 20 inch and 16 inch, andthe real-time wellbore stability monitoring would be conducted during the drilling of the 11 3/4 and 9 5/8 inchsections through the Tertiary.

Characteristics of the sections are as follows:- 13 1/4" inch section1527-4584 m MD (measured depth, along the hole), i.e., 1318-2275 m TVD (true vertical depth)Wellbore instability is particularly likely to occur in the zones 1600-1800 m and 2000-2200 m TVD, and is bothinclination- and time- dependent.Hole cleaning and lost returns issues.Problems most likely with POOH (pulling out of hole).

- 12 1/4" inch section4584-5530 m MD (2275-2570 m TVD )Middle Eocene, Balder and Lista are the more unstable formations, with Lista requiring high mud weight and beingparticularly unstable.Hole cleaning and lost returns issues.Lost returns when drilling into the Tor formation, immediately below the Tertiary.

In order to prepare for the field test, a dry run was held in Cambridge at the beginning of July 1998, with membersof GeoQuest and Amoco, to simulate the operations that the system and the team might be expected to carry out.The well chosen for this exercise was F-13, also in Valhall but drilled in a south-easterly direction from a differentplatform. This was in progress at the time of the dry run, and was expected to have similar problems to thoseanticipated in A3C. The dry run consisted of running the Wellbore Instability Evaluation software (WISE), andexamining the daily drilling reports, to see what information was important, and how to make decisions regardingthe drilling of the well.

As mentioned above, the dry run showed that the diagnosis algorithm was not effective given the data coming in. Italso helped us to develop a drilling strategy for A3C, which will be discussed in the context of Figure 4. Drilling inthe Valhall Tertiary normally started with a mud weight of about 14.3 pounds per gallon (ppg), i.e., a mud density of1715 kg.m-3; as drilling proceeded and cavings (caused by shear failure of the wellbore) were observed, this densitywould be increased steadily, often exceeding 16 ppg. This caused problems in the lower sections, as it producedwellbore pressures above the fracture gradient, mud was lost, and large amounts of blocky cavings were producedfrom naturally-fractured zones, resulting in pack-offs. The strategy proposed as a result of the dry run was thatdrilling should begin at 14.2 ppg, a little lower than usual, and that this value should not be increased unlessabsolutely necessary, and primarily in response to gas, positive flow checks or other signs of overpressure. Ifcavings were produced by shear failure, they would be removed by good hole cleaning practices rather thansuppressed by higher mud weight. The equivalent circulating density (ECD) would at all costs be kept below thevalue corresponding to the minimum horizontal stress (15.3 ppg in the problem zones), since exceeding this limitwould invade the fractured zones known to cause problems in Valhall, and loosen them. (ECD is the mud weightthat would generate the downhole pressure observed while pumping; it is generally greater than the mud weightmeasured at surface because of frictional pressure drop in the annulus). ECD would normally be allowed to riseabove this, with consequent loss of mud to the fractures in the formation, an expensive problem but not one regardedas terminal; the new drilling strategy emphasised the connection between mud losses and activation of the fracturezones, which could be terminal.

This strategy made the explicit assumption that cavings produced by shear because of low mud weight, would occurin quantities controllable by hole cleaning, but that cavings produced by mud invasion and loosening of fracturezones would be uncontrollable. It was clearly important to know whether mud invasion was occurring, and so afurther part of the strategy was to monitor mud volume for losses, and also monitor the cavings at surface to identifytheir source. This would be done through their shapes; shear-induced cavings are angular, those from fracture zones

21

are blocky and parallel-sided (for an example, see Figure 17). If in spite of the low mud weight, blocky cavingswere seen at surface, this would mean that the fracture zones were being invaded. This would require addition oflost circulation material to the mud in order to block the fractures.

The implementation of the real-time wellbore stability system was planned to be as follows:- on the rig, Anadrill engineers monitor the downhole tools and measurements; these include drilling motor,orientation system, telemetry, Compensated Dual Resistivity (CDR), torque and weight on bit (IWOB), AnnularPressure while Drilling (APWD) and Sonic while Drilling (ISONIC). Data from these would periodically betransmitted back to shore, to the Amoco office in Stavanger, to an IDEAL (Integrated Drilling, Evaluation andLogging) workstation there.- also on the rig, the cuttings and cavings coming up the well would be monitored, for quantity, shape and geologicalprovenance, as a primary input to the diagnosis of the state of the wellbore. Some of the cavings would bephotographed with a digital camera, and the images emailed to shore.- in the office, the data would be received, and transferred as text files to the RTWBS computer. Scientists fromSCR, GeoQuest, Amoco and NITG-TNO were to be on duty there on a 24 hour rota, to display and analyse the data,run the RTWBS software, incorporate changes and implications into the earth and trajectory models, and makediagnoses and recommendations. These would then be communicated to Amoco staff, who would decide whetheror not to take action.

To help with the planning process, and the actual execution of the field test, a website was set up. This method ofinformation gathering and exchange has been found very useful for other wellbore stability projects, where largeamounts of diverse information needs to be easily available. The web site was made accessible to Schlumberger,Amoco and TNO staff. Figure 18 shows the front page of the website, with a link to the daily drilling reports(detailed activity logs for the rig for the previous 24 hours) from well F13, which was used for the dry run. Theability to call up the drilling reports, and the other types of information, rapidly and systematically from anynetworked computer, was invaluable for monitoring progress on the well.

Figure 17: an example image of a caving from F13, revealing the presence of natural fractures.

22

Figure 18: the front page of the Valhall field test website.

The date on which drilling of A3C was scheduled to begin changed constantly as operations on the platformproceeded. Since the wellbore stability team intended to have a presence (at least two people, from GeoQuest,Anadrill, SCR, TNO and Amoco UK) in the Amoco office in Stavanger for the duration of the test interval, thisrequired a great deal of juggling of schedules. A scheme was arrived at eventually which would allow two people tobe there, but the idea of working shifts had to be abandoned, since it would lead to too much disruption. The teammembers would both be present during the day, albeit with times staggered to cover early morning to late night, andwould be available during the night by mobile phone. In the event, the phones were not needed.

Eventually the drilling of the field test section began on Sunday 20th September. Final changes in thecasing and well trajectory plan were made by Amoco Norge, following discussions with the wellbore stability team.The change meant removing the 11 3/4" inch liner planned from the base of the Nordland to about 2275 m TVD,and making the trajectory shallower over this interval (see Figure 4). This was intended to avoid drilling thenaturally-fractured zone below 2000 m at a shallow angle, as this had been found to cause problems in the past.

23

A network was established between the Anadrill data-logging computer on the rig, and an NT workstation on shore.It was intended that data would be passed directly to shore, as soon as it was acquired, but it was soon found that thiswas vulnerable to any network disruption, with a knock-on effect that when the network connection to shore stoppedworking, the data acquisition often also stopped. This was unacceptable from the point of view of Anadrill'sprovision of a commercial service (since the MWD data are required immediately when directionally drilling, and sodata acquisition downtime means a stop to drilling). A variety of solutions to this were tried, and data was sent inbatch mode by FTP while this was happening. Finally, a second network was established on the rig, isolating thedata acquisition from the rig-to-shore communications, and this proved stable enough for continuous real-timetransmission and display. The data was also stored on the rig and transmitted periodically in batch mode. ExtraAnadrill engineers were brought in to assist with reformatting these data, to produce time- and depth-based plots orhard copy on shore.

The cavings data collected were primarily:- the rate of cavings production at the shale shakers (the coarse solids separators on any rig) measured every 30minutes by the filling time of a bucket. This may seem crude, but is versatile (in terms of the number of differentmodels of rig to which it can be applied) and reliable; more sophisticated solids measuring devices have been triedon a number of rigs, but very few have been satisfactory;- the dominant shape of the cavings. Initially the intention was to classify the cavings into three types: angularcavings originating from breakouts, blocky cavings from naturally-fractured zones, and elongate or splintery cavingsfrom zones of elevated pore pressure. It was very soon clear that many cavings did not fit into this convenientclassification, and were simply uninformative pieces of broken rock. However, some of the cavings did indicate thatthey came from breakouts, and some indicated that they came from overpressure zones; only two cavings were seenduring the entire drilling program seen that came unambiguously from fracture zones, which points to thecorrectness of the selected drilling strategy.- the geological age of the cavings. This would give an idea of where they were coming from in the interval. Thisrequired micropalaeontological analysis, which was not immediate; when the results arrived they indicated that allthe cavings were coming from the upper section of the open hole, where the exposure time to the mud was longest.

On shore in the Amoco drilling team office, the real-time drilling data were displayed, the WISE software was run,and the wellbore stability team attended the morning drilling meetings and advised on stability issues. The drillingdisplay proved very popular, and gave the onshore drilling and wellbore stability staff very close contact with thedrilling operations. One of the tasks that the wellbore stability team took up was to monitor with particular care therate of penetration and the ECD. The latter is a critical parameter; if it were to drift upwards beyond 15.3 ppg, thisran the risk of invading the fracture zones and causing permanent damage to the wellbore, and if it jumped suddenlyit could indicate the onset of blockages in the annulus, and packoff (when cuttings and cavings collect around thetop of the bottomhole assembly, preventing fluid flow and sticking the drillstring in the hole). Rate of penetration isimportant in controlling ECD; if too much rock is drilled too quickly, the cuttings suspended in the mud in theannulus increase its density and hence the ECD. While it is easy to state that this might lead to problems, one of thetraditional aims of the drilling crew on a rig is to drill as fast as possible, assuming that this will help them to reachthe bottom of the well quickly (i.e., cheaply). In many areas of the world this is true, but in areas such as the NorthSea a longer-term view must be taken, and high instantaneous drilling rates often lead to problems which cost moreto solve than is saved in drilling time.

The wellbore stability team acted as advisors and consultants on matters relating to the state of the wellbore, anddrilling strategy, partly from their experience and the knowledge available through them from the rest ofSchlumberger, and partly through use of software tools. This role, and the focus on problem avoidance that itbrought to the drilling office, was very well received by Amoco; the team participated in all the morning drillingmeetings, when the shore-based drilling staff discuss the previous 24 hours progress and problems with the drillingengineer on the rig. Another key task that was welcomed was the teaching of a class on wellbore stability in generaland cavings monitoring in particular. This lasted about two hours, and was given three times per week to all drillingand mud-logging crew going offshore to drill A3C. The response to this from the crew was so positive that the classwas requested and given to the crew of the other platform operating in Valhall as well. It was felt that the classesfocussed the attention of the crew on the avoidance of instability problems (rather than the traditional reactiveapproach once the problem has occurred), and also allowed them to meet and question the scientists who would behaving an influence on their drilling practices. Drilling staff are generally highly independent and sceptical, and itwas important for them to have the opportunity to gain confidence in the scientific members of the team.

24

The WISE software, developed as part of this project, was run on a Pentium laptop next to the real-time drillingdisplay. Its main function was recalculation of the safe mud weight windows following modifications to theMechanical Earth Model, as detailed below. The wellbore stability team was, by this time, so familiar with thesituation on A3C that the diagnosis and recommendation aspect of the code was hardly needed; this will not be thecase in a commercial implementation, where the engineers will not have the time to focus so completely on a singlewell.

The team monitored the drilling data regularly, updated the Mechanical Earth Model, and made many comments andrecommendations, usually to avoid raising the mud weight, to keep the rotation speed of the drill string at 130 rpm(to assist hole cleaning), and to reduce the rate of penetration (to minimise cuttings loading in the annulus). Somemore specific examples were as follows:- a leak-off test was performed (as is usual) after drilling out of the 13 3/8 inch casing shoe. This indicated a likelylower bound for the minimum horizontal stress of 15.0 ppg, higher than the pre-drill estimate, and so the MechanicalEarth Model was revised to take this into account, and the predicted mud weights to avoid shear collapse of thewellbore were recalculated. Figures 19 and 20 show the MEM's before and after revision (Figure 21 also includes arevision after a mud loss at 4120 m MD).

Figure 19: initial (pre-drill) Mechanical Earth Model. The friction angle curve has been omitted, because it obscuresmuch of the detail of the stress levels.

- background gas levels were high at around 2100-2200 m MD. Normal response to this would be a substantialincrease in the mud weight, which would have led to the destabilisation of the critical fractured zone between 4100-4300 m MD. On the recommendation of the wellbore stability team the mud weight increased was restricted (to14.6 ppg), and the rate of penetration was lowered to below 30m/hr. The reduced ROP decreased the rate at whichgas was released into the annulus due to rock being crushed, and these actions reduced background gas levels from20% (gas peaks of 35% were observed) to less than 5%, while avoiding problems further down the well. Figure 21

25

shows some of the data collected from this period, from the electronic data (ECD, ROP and depth) and themudlogging reports (gas level).

Figure 20: modified Mechanical Earth Model, where the horizontal stresses have been changed in response to aleak-off test at about 1370 m TVD, and subsequently (much later during drilling) to mud losses at 2130 m TVD

(4120 m MD). The friction angle curve has been omitted, because it obscures much of the detail of the stress levels.

- the rock strength data in the Mechanical Earth Model were verified using new sonic data from the ISONIClogging-while-drilling tool. This was done in order to be have confidence in the offset data used to construct themodel; this had been derived from a variety of sources, mainly through older sonic data and a proprietary AmocoNorge correlation to rock strength. The new strength values were in fact very close on average to the old ones(although because of the normal variability of geological conditions there were many differences in detail), andcould be spliced into the earth model without difficulty; they did not therefore lead to any significant changes in thedrilling strategy.

The results of the cavings monitoring were sent to the wellbore stability team regularly by the Anadrill staff on therig. The main aim of the monitoring was to watch for tabular cavings caused by fracture activation, but because ofthe success of the mud weight strategy used, very few of these were seen. Systematic observations of the cavingsdid, however, help to clarify our picture of the state of the hole as drilling proceeded; for example, a decrease incavings rate generally pointed to a drop in hole cleaning efficiency rather than a lack of hole instability. If theimportance of hole cleaning procedures was re-emphasised, and the rotation speed of the drillstring increased to itsrecommended value, a flush of cavings would often be produced, giving confidence that the hole was now in acleaner state.

26

Figure 21: the influence of rate of penetration (m/hour) and ECD (pounds/gallon: controlled by static mud density atthe surface) on the background gas levels (parts/million) monitored at the drill floor. The vertical axis is time, andthe right-hand chart shows measured depth in metres. In the upper part of the central plot, the rate of penetration

(blue; units ) shows a erratic increase as the drilling team aim to cut down on rig time. As they do this, thebackground gas levels (pink stars) increase to worrying levels, and the drilling team suggested a large increase inmud weight. This would cause problems lower down the well, and so a compromise was reached in consultation

with the drilling team of a small increase in mud weight (0.4 ppg in two stages, giving two small steps on the ECD),and a decrease in rate of penetration (which is controlled by the driller using the weight on the bit). The gas levels

drop rapidly as a result.

Figure 22 shows cavings data as a function of time for the whole of the drilled interval. Because of the nature ofdrilling operations, the cavings record is incomplete at many points (e.g., when the drillpipe is being tripped into thehole, no cavings are brought to surface). The overall rate of cavings production is relatively low (the numbers onthe vertical axis of the cavings rate chart, second from the top, are in arbitrary units). The most interesting feature ofthe data is the onset of surges of cavings after 320 hours. One tentative explanation of this is linked to the welldeviation; for the earlier sections of the well, the hole angle was greater than 65 degrees from vertical, but where thesurges occurred, it was below this value. It is believed that there are three regimes of cuttings movement in adeviated well: if the deviation is less than about 60 degrees, the cuttings and cavings fall smoothly down the welland are cleaned out smoothly by the mud flow; if the deviation is greater than about 65 degrees, they form stablecuttings beds and are again removed smoothly. Between these two limits however, unstable beds of cavings areformed on the low side of the hole; any disturbance in the wellbore tends to cause avalanches, which sweep cavings

27

intermittently into the mud flow, and so produce cavings surges at surface. The period in Figure 22 between 330and 420 hours shows surges, and the deviation of the well at this point is in the avalanche range. It may also be thatthe smoother response in the earlier section of the well is because of the beneficial effect of the rotating drillingsystem that was used there; this system was not, however, used in the short section at 200 - 220 hours.

There were other aspects of the drilling operations in which the wellbore stability team, their technical background,and the WISE software, played a role during the drilling of A3C. These are too lengthy to cover in detail, but theyincluded:- wellbore monitoring supported elimination of unnecessary wiper trips- refined hole cleaning/tripping plan to account for hard limestone bands- developed mud weight plan to avoid wellbore damage during/after trip- location of strong zones for bit positioning during rotation off-bottom- updating predictions following unplanned changes in trajectory

Figure 22: Cavings data from the test interval of well A3C, versus time in hours. The top chart shows the depths inmetres of the bit (blue) and the bottom of the hole (green). The bit depth is often less than the hole depth, for

example, when the bit is being removed from the hole. The second chart is rate of cavings collection at the surface(arbitrary units; the number is proportional to the reciprocal of the length of time taken to fill a bucket). The thirdchart is the deviation of the well from vertical, in degrees. The fourth chart shows the rate of flow of the drilling

fluid, in gallons per minute.

28

In summary, the reservoir was penetrated ahead of plan, with much lower mud loss to the formation than usual(around 10% of the typical value for a well such as this on Valhall; 130 bbls as opposed to 1000-2000 bbls), andnegligible activation of fracture zones. Some damage was done to the well on the trip out of the hole, and the wipertrip aimed at cleaning up this damage unfortunately initiated a sidetrack, from around 3600 m MD. This sidetrackalso penetrated the reservoir, using the same wellbore stability/hole cleaning strategy as for the first hole (althoughthe wellbore stability team were no longer in Amoco's office by then). The casing string then re-entered the originaltrack, which had been open through the fractured zone for several weeks, and was landed below the fractured zone.It could not, however, could not reach the bottom of the hole. The well was suspended at this point for some time,owing to other operational matters, and the reservoir section is not yet complete. The well as a whole thereforecannot yet be said to be a success; nonetheless, the facts that the reservoir was penetrated ahead of schedule, and thecasing could still be installed in the troublesome fracture zone after several weeks of open-hole exposure to thedrilling fluid, are evidence of the success of the original wellbore stability strategy, and the real-time approach.

5. Results, comments and conclusions

The use of the real-time wellbore stability process in a field demonstration was successful. Ways of implementingreal-time wellbore stability as a commercial service are being examined by GeoQuest, and the software is underdevelopment for commercial use in another part of the company.

The capabilities of the real-time wellbore stability process can be developed as follows:-• the current automated diagnosis algorithm does not work effectively and should be replaced. The new systemshould, in particular, be able to analyse situations where there is missing data. This is a difficult issue and work isbeing undertaken at SCR to address it.• a relatively simple model is currently used to predict classical rock mechanics instabilities, namely breakouts andmud losses. This was adequate for Valhall, but in general calibration or adjustment factors will be needed, and afacility to use these factors will make the approach more generally applicable. The addition of more sophisticatedtheories to calculate the factors themselves is not required, as they can be generated using tools such asSchlumberger's ROXANTM or commercial finite element codes such as DIANATM or ABAQUSTM.• a wellbore stability tool is not a drilling simulator. It may, however, be desirable to add components more oftenfound in such simulators, for example a hole cleaning model, to the wellbore stability control process in order toclarify and quantify the interaction of drilling practices and borehole instability.• the recommendations module is currently rather simple, and based on look-up tables. The approach used to linkcause and effect in a new diagnosis tool may also be a more appropriate technique to relate problems to remedialactions. Recommendations also have to be presented in a form more suitable to drilling engineers. For example,"increase flow to 900 GPM and increase RPM to 130", rather than “increase hole cleaning”.

The implementation of the process in the A3C field test took account of experience gained, principally by theIntegrated Project Management staff of Schlumberger, from wellbore stability projects in fields such as Cusiana,ETAP and Hibernia. This implementation model will, therefore, probably be close to future forms. There is,however, scope for enhancement as follows:-• the software should have an interface to the rig-site acquisition system for better information flow.• data from all sources (mud logging, LWD, …) should be sent continuously to one computer. There should be areal-time data display in a public area in the office, rather than solely in the wellbore stability team office.• the daily deliverables should include written reporting of wellbore issues. Other deliverables, such as logsgenerated at the end of bit runs, should be specified with the client at the beginning of any project.• rig crews and relevant office-based personnel should be trained such that they (a) appreciate general wellborestability issues, (b) possess a basic understanding of the interpretation of surface, fluids and particularly solidsinformation, and (c) are aware of the consequences that changes in drilling parameters may have on the integrity ofthe wellbore.

29

6. References

Bradford, I.D.R., and J.M. Cook: 'A semi-analytical elastoplastic model for wellbore stability with application tosanding'. Eurock'94: joint SPE/ISRM conference on Rock Mechanics in Petroleum Engineering, Delft, Aug. 1994.SPE no. 28070.

Fjaer E., Holt R.M., Horsrud P., Raaen A.M., Risnes R. (1992) Petroleum-related Rock Mechanics. Developmentsin Petroleum Science 33, Elsevier, pp. 338.

Last, N. Plumb, R.A, R. Harkness, R., Charlez, P., Alsen, J. and McLean, M. An Integrated Approach ToEvaluating And Managing Wellbore Instability In The Cusiana Field, Colombia, South America. Ann. SPE Tech.Conf. (Dallas, 22-25th Oct. 1995) pp.147-160. SPE 30464.

Papanastasiou, P.C. Numerical analysis of localisation phenomena in deep boreholes. PhD thesis, University ofMinnesota, 1990.

Rosthal, R.A., Best, D.L. and Clark, B. Borehole Caliper While Drilling From a 2-MHz Propagation Tool andBorehole Effects Correction. Proc. 66th Ann. Tech. Conf. S.P.E. (Dallas, 6-9th Oct. 1991) pp. 491-502. SPE no.22707.

Appendix: A geological overview of the Valhall FieldThe Valhall Field, operated by Amoco Norway Oil Company, is located in offshore blocks 2/8 and 2/11 in theCentral Graben area of the southern part of the Norwegian North Sea (see Figure A1). The field was discovered in1975, when the discovery well 2/8-6 encountered over 100m of hydrocarbon-bearing section in Late Cretaceouschalk formations. Following successful delineation, and approval from the Norwegian authorities, Valhall Fieldbegan production in 1982 from the highly porous Tor and Hod chalk formations of Upper Cretaceous age. Currentlyestimated recoverable reserves are 650 MM barrels although several projects are ongoing to further increaserecoverable reserves to 1000 MM barrels.

The Upper Cretaceous was a period of abnormally high sea level over much of Western Europe and resulted in thedeposition of calcareous chalks. These chalks form the producing reservoir in Valhall Field. Chalk deposition atValhall ended at the beginning of the Paleocene following Laramide movements, which resulted in an influx ofclastic material in response to uplift of the sediment source area. Directly overlying the Upper Cretaceous is theRogaland Group, a predominantly claystone sequence, subdivided into the Lista, Sele and Balder Formations.Locally developed within the Lista are very thin sandstones, which are the lateral equivalents of the Andrew andForties formations developed elsewhere in the Central North Sea. The Balder Formation also characteristically hastuffaceous siltstone interbeds.

The overlying Eocene to Recent saw continued subsidence of the Central North Sea area, which resulted in thedeposition of a thick, up to 2500m in thickness, monotonous sequence of predominantly claystone and siltstonesediments in a marine environment.

The Valhall Field was ideally suited to demonstrate the real-time detection and control of wellbore instability for thefollowing reasons:

1) It can readily be demonstrated that wellbore instability is a major concern at Valhall, impacting the safety andefficiency of drilling operations, particularly in highly-deviated and extended-reach wells.

2) Sufficient data are already available to develop a geomechanical model for Valhall Field. These data includemudlogs, wireline logs, LWD/MWD data, core data, stress measurements, pore pressure measurements andstructural data.

3) Two extended-reach wells would be drilled in Valhall Field during 1998. Both wells would be available todemonstrate the capability of the real-time product.

30

4) Extended-reach drilling in Valhall Field is expected to continue for at least a further five years. The successfuldemonstration of the real-time product will have a direct impact on the safety and efficiency of these drillingoperations.

Figure A1: location of Valhall Field (from Munns: Marine and Petroleum Geol., Feb. 1985)