You are on page 1of 10

q

SPE
SPE 18183 A Computational Model of Well Cornpletmn Design
by S. Dunn-Norman* and J.M. Peden,f Heriot-wati W and D.R. Bedford) Ferranti Computer Systems Md. *SPEMsmbafs
.
*WI 1W9, SOClOty d Potrohum Engirwora This peper woo propemd tor proaonklion at tho S3rd Annual Technkal Oonkrence and Exhibitionof the Society of PetrolbumEngineers held in Houston,lx, CMber 2-5, 19ss. TM peper was edwtul for &enntatkn by q n SPE ProgramOornmittw foliowlngrwtew ot informMon amtdnad In m abstract $ubmittodby the author(s),OontonmO!Ma pepor, as preoontad,havs not boonr@dewc5W the SooIetyO!Potrohum Enginooraq nd q ro $ubjwt to owrootkrt by tho Papua author(o). The matuiat, S9 praented, don not nemaarlly rotted arty posllkn Gtthe Soolely at Potrotwm EncIIwor& ItsottkerB, or mern5ere., tooopyb ~@~*m~_@MmW *W E_l Wm-dtiqd W*m E_. P~ MtratWmmay not boopkd.lhaebmraot clWJ!dcuntainoo@OWu9~ot rWMctadman aWaotofnot rnomthanwwWda Rkhudmn, TX ~ Telex, 7MSSS SPEOAL. whWeanllbywhorn the~b~. Wrlta PWoaUam Merug8r,SPE, P.O.,SO!I-,

soOkItuor?strkEndneam

ABSTRACT

oxtensi ve array of dealgn $m%metera.

In addition,

deaignera The design of well ootttplet ions 1s a Qomplex process


whioh requires the engineer to seleot, rationalize and integrate a large num>e~ of design elements. In the literature, it has been recognized that a syetems epproaoh to well cotnplet ion design must be adopted if optimum and coet-effioient Ooatpletione are to be generated. Historically, diagratmnatioal suoh as logic diagrams, have been representations, usedto model and implement the systems approach for

must

optimize

designai

for

ourrent

requirersenta, proJeoted requiretaenta of fut we well oper,: ions (e.g. stimulation, re-entry), future and producing conditions (e.g. produotian souring), potent ial produot ion problems (e.g. sanding, aoale,

or oorroaion).
oomplex prohlam,

ASJwould be expeoted
many resultant

completion

are sub-optimal solutions.

in auoh a designs

well completion design, This paper discusses, as an alternative, a computational model of well completion design industry developed through an sponsored projeot, collaborative researoh The model and i~teraotion formalize9 the relationships between conventional computations for completion design, suoh as tubing hydraulic or stress analysis, and design knowledge gained through design and operating experience. This model provides a comprehensive representatioof the systems approaoh to well completion design, and serves as the basis for a computerized implementation. The implementation integrates both conventional and expert systems programming techniques. Program development and the advantages of using a oomputer design program to provide a rational and consistently optimun design capability are disouesed, INTRODUCTION Well completion deeign refers to the selection of all downhole equipment, fluids, and procedures neoessary either to establish, or to improve, the production or injeotion of wellbore fluids. In designing oomple;ions, engineers must oonsider an Referenoea and illustrations at end of paper

In an attempt to rationalize the prooess of well completion design, Patton and Abbott proposed the systems approaoh to wel: completion design() In this approaoh, they suggest an integrated design prooec!ure in which eaoh oomponent of the design is With respect to its attributes, configured et:vironment, goals, standards, aocess to resouroes, This approaoh was presented as a &nd constraints. possible basis for oonstruoting a c!ompletior? design oomputer program. The systems approach is significant in that it sets a framework for describing relationships between varioua wellbore components and operations, their effeot on well productivity, and any limitations However, simply imposed on future operations. identifying these relationships is not sufficient Rather, the for oormtruoting a design program. design knowledge must be expressed in a struoture whiah explicitly details the reasoning and processes Further, this used to reaoh design oonolusions, struoture must provide a means of constructing completion design solutions. Logio diagrams have been used in an attempt to These struotur,) completion dvsign knowledge(a). purely impart a teohniquee diagrammatical procedural, or sequential, view of the completion In addition, the diagrammatical design prooess, approaoh reqiilres that the design knowledge be structured in a binary, query format, that 1s, a question whioh oan only be answered by yes or no. 27

. 2
Thi8 Btruoture MODl?L O F WE!LL aMPLS!WCaW A C@lPl?XATI@UiL ..-.- C --. ----- - nmSTCN ----=. h an awkward model of the completion . The design prooe8s uttl.z~s
-r=

C!nu*a*a*
.SJawa

fotward not

driven

design prooese. An alternative approaoh ror representing the preoess of completion design is a computational model which provides a dynamic, non-procedural representation. A Oomputatlonal model combinee both empirical design knowledge with nunerioal, engineering computations. The following paper revfevs the procese of well completion deeign, presents a computational model of that prooess, and disousses the implementation of thxs model as a oomputer based design program. WELL COMPLETION DESIGN, PROCESS The wel,l completion design process refers to the funotions which must be performed in specifying all cbwnhole equipment, fluids, and procedures required to cmplete a well. The design process, covers a wide range of produot!on engineering topios. The principal techniza; topics are shown in Figure 1. It is important tc note that there is actually no single prooess of designing well completions. The
specific tasks, or functions performed, .Iepend on the objective of tfle design. For example, tasks involved in formul+b:.ng a preliminary design for

reasoning,

and ia non-procedural, muut be Constructed,

. Design aclutiona enumerated.

. The design prcblem oan be reduced into sub-problems or eub-oomponenta, such as sizing, packer selection, etc.

tubing

, fiesign criteria or requirements are frequently ae the design process updated or deleted proceeds. . The teohnical domain of well completion design sub-tasks such as interpretation, includes planning or forecasting. Several of these characteristics are significant elements of the computational model, and therefore warrant further discussion. design if a forward, or completion First, This implies that deeign data-driven problem().
solutions muet be constricted unumor&ted, Enumeration means that rather than it is feasible

to list
these

all potential

eolutlona,

and then struoture

expendable exploratory or delineation wells may vary ooneiderably from the tasks performed in detailed design of development WellS. Consequently, any computational model of well completion deeign must or objeotive of the design have a defii~ed level, process to simulate. The computational model presented herein simulates the detailed level of design normally required for or tasks, infield development wells. All funotions, required to select downhole equipment or fluids, are modelled, Examples of these design functions selecting paakers and inolude tubing, sizing movement compensation devices, and specifying a downhole safety valve for the well. In addition to ascertaining the level of the design proc~ss to be modelled, a startir: point of the desisn process must be defined. Frequently, in the design of development well completions, it is assumed that the well has already been drilled to total depth and that production casing is set, It may be argued that this assumption constrains the design of the completion string, and that to design a truly optimum completion requires the assumption that any hole, or easing size, is available, In some oases this argument is valid. However, this was deemed to be an unrealistic approaoh for the purposes of the computational model, CHARACTERISTICS OF WELL COMPLETION DESIGN In constructing a model or simulation of a problem, it is fundamental to identify the primary task and its associated oharaoteristios. Accordingly, well completion design was identified as a !Iderign taskil. In design tasks, objects are configured under goals, or satisfy constraints in order tq aohieve requirements. Figure 2 depicts this oonce~t. The primary characteristics of well design may be described as follows: completion

in suoh a way that Find the oorreot answer.

they may be searched to It aan be argued that

enumeration 1s possible if the design problem dcaain is sufficiently small(). Clearly, this argument is not valid for well completion design. Well oompletlon design ia also a non-procedural process. It is extremely diffioult to identify an ordered set of Stepsw, which defines a general Further, method of designing wellbore completions. the starting point of the design process is oontext dependent, For example, if the well is subsea, the engineer may wish to begin with selection of the tubing sizing, rather than tree, subsea Consequently, diagrammatical, or solely procedural approaches are awkward in representing the entire It should be well completion design process. recognized, however, that thefle are portions of the design process whioh are procedural, and these necessitate conventional modelling techniques. problem into The ability to decompose the provides a sub-components sub-problems, or manageable approach to constructing solutions when However, with the technical domain is extensive. this approaah, design oonflicts may occur between items which have been configured independently. In this case some filtering teohnique must be applied to eliminate conflicting oonfigurations$ Another important poinb is that completion design is a dynamio prooess, As the design develops, the design criteria or requirements are frequently altered. This is a complicated characteristic to an Kindott mechanism model, as it neoessit,ltes capable of removing any existing components whioh criteria or with new design the oonfliot
requirements.

The computational model provides a struoture whioh However, to embodies these problem oharaateriatics. attain an exa?t simulation of the completion design

28

m% 18183

S. DUNN-NORMN~ J.M. I

EMMD D.R. SSDFCW

prooeaa would require a oomplete understanding of the human reasoning procesaea. FOr the Purposes Of

developing the computational assumed that it is sufficient problem ohvacteristi~s. tOMFUTATIONAL MODEL ~ Definition

to

model, it has been model the principal

the Computational results in order to determine an optimuv size. Where it le. not possible to immediately tietermine the optimum value, all alternatives are held for later evaluation?. When an attributes value has been determined, it beoomes a constraint on subsequent elements of the design problem. As mare components are Speoified, and inc~easing numbers of constraints are added to the design problem, a non-option design state may be reached. This leaves the designer with a dilemma of finding which oonflioting design parameter to alter. The designers of 12ECsexpert SYStem WON, recognized this phenomena, and simplified the!.r stculation of the design prooess to eliminate the prcbiem(c). Mechanical acinpletion components, or completion reservoirs, anti fluids, objects such as wells, possess inherenz rt .atianahips. For example, in Figure 3 a subsurface safety valve is a r)mponent. ofw a completion string, and fluids are .untained in!! a reservoir. These aomponent relationships have been formalized in the aomp~tational model, and point<ers in the design serve se intelligent
reasoning, Several hlerarohiaal, or alsesification s$ructuree The of eb.jeots also exist incompletion design. general elaes of paakera has eub-oleese$ oalled dual paakers and single paokera. These sub-ulaeses share many of the same attributes and values se the parent It is therefore convenient tv have class, psaker. inheritance etruotures in the computational model which allow properties to be passi)d from one objeat to another. Object structures, relations, and inheritance are used to represent knowledge about techniques components or objects of the completion system and Knowledge regarding their inherent relationships. the specification and seleotion of these objects is formalized in design reaeoning.

rhe computational model ie a formal representation Df the well completion degign problem. The model ?onsjsts of two principal elements: empirical design ?easonlng and engineering computations. rhe empirical design reasoning is simply what 1s wellbore completions. It known about designing enocmpasses a description of the objeots and components Whiah comprise the completion aytem, selecting those relevant about reasoning components, and knowledge of how field operations may affeot the d~si gn, The or aonstrain computational model utilizes symbolic representations to describe the components and system, and the completion objects of the completion
design reasoning: Engineering amputations refer to 3alculatione such ae tubing hydraulics, stress and expeoted movefaent. These aaloulations are a fundamental element of the well ,oompletion design proaesa. Calculated values suoh se total string expansion, fluid velooity, and the foruee developed on the packer, are required to obviate impossible design choices.

The engineering Computations must be invoked at appropriate pants in the design proaess. Further, the engineering computations and empirical design reasoning must be well co-ordinated, so that information may pass from one element to another at various points in the reasoning prooess. Completion Components The components used in completion design can be ola&sed acaording to their type or fUflCtiOfl. Figure 3 shows some of the major components, or objeots, of a wellbore oompletim system. Eaoh oomponent, or object in the completion, possesses attributes. These attributes normally have assigned ialues. For example, a paoker has the attribute IInumber of borts!! and this attribute oan normally have a value of 1, 2 or 3. Further, most obdects possess many attributes, and bhese may be organized in an objeat-attribute-value struoture for olarity(s)o In the computational model, most completion components are desoribed using this eitructuro. Figure 4 shows objeot-attribute-value structures for two completion components. The value, or values of an objeot~a attributes may be determiner, through empirioal design reasoning, direct numerical oomputatlon, or a uomblnation of the two methods. For example, optimizing tubing size means determining a value for the tubing attributed ltODilarid lIDI1, This is aooomplished by first computing tubing hydraulics for a range of possibilities, and then applying design reasoning to

Completion Design Reasonin~ A substantial part of the oorpletion design praoess is empiriaal. Engineers learn, through experience, Muuh of this what works in field operatlona. knowledge is based on aseaaiating some end result or For consequence with one or more conditions, example, if two tubing strings are in tihe well, it is not possible to rotate to unseat a packer. Thus, if a dual completion is being planned, the paoker This type retrieving method must be straight pull. of completion design knowledge aan be oodified using design rules. In design formt rules, knowledge is > > I Figure 5 shaws an example tubing deeign rule whioh generates possible,tubing grades given CO~ and H*S, partia~ pressure dkta. Design reasoning is represented symbolically in rule presented in the

If < conditions < aatians then

20

4
format. physioal

A COMP~ATXOIW, -L This means that


oonatraints

Q? WSLL CQMS~ON DESIGN


used to prooeas. provide a etruoture

,,
o? the

SPS 18183
completion,

de8igr

onoepte, quch se
may be desoribed

or oonditionc,

tith a syntax similar to English language. the idea that 3 paoker unseats with a 8traight i.e. no rotation, oan be ooded as: (retrieving-method Paoker etraight-pull)

Thus, pull,

The atruoture developed may be deaoribed as having two leve18. The first, or outw level, defines requirements of the design problem. The inner level, ie where design reasoning oreates and mahtains completion components whioh may satl?y raquiremente of the problem, la This atruoture shown in Figure 6,

Design rules operate eolely on the ourrent sLate of design knowledge. In order for a rule to exeoute, the conditions o? the design rule must be validated by faots. The straight-pull paoker oited above is Bn example of a faot capable of validating some design rUld. ,, Design rules alter the state of reasoning, either by adding, deleting or modifying ourrent racta abOUt Eaah time the state of the completion deeign. reaa(aning ohanges, new fatts end design knowladge oauee other design rulee to ooctinue the design It ehould be ~oted, however, reasoning process. that rules are not t.esigned to exeoute in a pr~-d~!ter-irled Lype order, aa in procedural pr.~grsms. A desi8n faut, which csuses one rula to be invoked, may aotua?iy be deleted by the aotiun o! rule has an enother design rule Def~e the first opportunity to e&cmte. Accordingly, this manner of
design reasoning h both completion ooding non-prooedur:l and non-determinWio, and obviatea the need to explicitly determine the Problem solution paths. design completion knowledge aymbolioally in rule format provides a powerful meohanism for modelling the design reasoning used in speeirying wellbore completions. However, the oomputatlonal model must also have a logioal

I*funotional design The requirements, or are requiramentalt as they oalled in the i?omputational ~del, derine the utilities whioh a include the completion nuat provide. Suoh utilities ability t.o produce fluids via tubi!jg, isolate the annulus, bullhead to kill the well, uirculate chemiaal treatments, or provide a meana or lifting the well if it dies. Many functional requirements appear to be related to geographical praoticea, Ae functional requirements are asserted in the computational model, design reasoning begins to speciry equipment, fluids or ite~s capable or those satisfying requirement. Tree-1ike struoLures are developed which deaoribe the valid combi~ations or ,attributes for each completion component. Theue structures result from the design model, An example of reaaonlng in the computational thie is shown in Figure 7, whioh depiota alternative
possible attributes. for tubing. Engineering parameters are an important oriteria for de?)ign alternatives, as the oomponenta apeciried must meet the physical oonatrainte of the Consequently, environment or completion ayatem. engineering, calculations are invoked at appropriate points in the deeign prooeas to return valuee auoh

screening

Repreaenttng

structure or rramework for the design rules, design solutions are oonatructed. Computational t40del Strgoture The struoture of the the manner in whioh organized, In order is first necessary hypothetical worlds.

so that

Computational model rerers to the design reasoning has been to disousa this struoture, it to understand the oonoept of

as fluid velooity, pressure drop, paoker toroe, tubing expanaion or oontraotion, and expeoted well performance. The numerical values returned are ueed for physioal tests, to prevent generation of In Figure 7, no tubing impossible doeign choioes. weights are shown for 3.5 inch) J55 %rade tubing, Thie demonstrates that a numerioal test, suoh as burst, oollapse or tensile load, has prevented tubing weights of insuftioient ntrength from being generated for this size and grade of tubing. At some point in the design process, a sufficient number of functional requirements and corresponding completion aomponenta will have been generated to deqign allow the mtidel to begin constructing completion oompone~ts The valid solutions. epeoified at this Staga will have been seleoted independently, without eignifiaant considerations of the other equipment to be inoluded in the design. Thus, some meana of oombining the oomponenta must exist in order to oonstruot final design solutions. :onatruoting Daaign Solutions

Hypothetical worlds are states where alternative, possible values or the same parameter are he~d. For example, in the initial stages or the design process an engineer may wish to oonsider both permanent and In thie ease two retrievable type paokers, One world hypothetical wcrlds would be areated. would hold the assumption that the paoker should be retrievable, while the other world would indioate that the paoker should be permanent. Hypothetical worlds may be oreated, deleted, or manipulated in a number or ways. Consider the If, at sane point in paoker example cited above. the desi8n reasoning, other design parameters are asserted whioh oonrliot with the permanent paoker, the world oontalntng the permanent packer ehould be Henoe, hypothetical deleted and never recraated. to explore design worlds provide an abi ity possibilftiea tentatively / );
Jn the

. At various points in the design prooess, completion components or items must be oombined t~ verify In the oompatibilitiea or perform oaloulations. computational model, this ie aooomplished by merging the hypothetical worlde whioh oontain theee items. Figure 8 depiots merging within the computational , model,
When a merge ooours, the design att:!h:tes, or ?%ota

computational

model, hypothetical

worlds are

al

.-

*
16183 ----8. DIIMa-UORmaN. .

BBWtMD . - . --. -. . J.M. PBDENANDD.R. promieing deei~, mskea preliminary optimization.

K .

relevant to eaoh independently c!salgned item, are cwubined in a new state, Thu&, if 3.5 inoh, N80 tubing ie combined with an Otis HC paoke~, the attributes of both items will exiet in the Oombined Btate. If, however., any of the faota to be joined

choices for nerging, the model determinations concerning deeign

are incoinpatible, the merge will be prevented from ooourring. This is accomplished by explicitly identifying and ooding design contradiotiona or inoompatibflitiea within the model. The points in the design prooesa where mergea should ooour appear to be either a funotton of the nature of the problem, or a prerequisite for numerioal computations. This point is illustrated with two examples. reason fop merging components, It is intuitive that all components whioh have a profile must be {?enaidered Jointly to ensure that they are.compatibly sized. Consequently, all profile devices are merged in the computational model.
Profile sizing is an example of the firet

With this approaoh, the computational model provides the neoea,aary atruoture for oonstruotlng multiple valid design solutions. The optimum solution is then identified within this set of possible designs. Optimizing Completion Designs The optimum completion deeign refers to a dezign which either maxtmtzea or minimizee oertain parameters, while achieving all specified goals and
objectives of the design. The parameters optimized are normally measures suoh as oost, reliability,

safety, availability, or ease of workover. These measures may be specified by the designer at the outset of the design pruoess. There are at least two met~ods of optimizing designs. The first, and probably most theoretically aorrect method, is to develop and apply a funotion to evaluate each measure at every state, or merge, of the design process. The optimum solution ie thue
the path whioh maximlzea , or measure(s). This forward, aearoh generate a single deeign solution. minimizes the approach would

The seoond reason for merging items is to provide neoeseary inputs for numerioal computations. For example, in order to oalaulate tubing movement and. the foroee developed on the paoker, . verioue
attribute

of the tubing, packer, downhole fluids, and tubing movement devioee muet be known. If we
that many deeign possibilities exist for each

assume
of

these items, there will be many oombinationa whioh oocur by mar#ing these poaaibilities. E&oh merge should invoke the numerioal ooisputation to oheok if either the seal length or paoker strength M adequate for the oonditione and equipment oonaidered. Figure 9 chows a conceptual view of how Iscrglng is with an engineering oo-ordinated

The problem with devel<,i~


Is that
must be

the 6pe0ifiC
quantifiable

forward type fumtiona measurea (e.g. 00et, SSfetY)


a manner which slows this

in

typO of evaluation. For example, if the measure is equipment reliabili*J, there would need to be some eesossment of the reliability af an item in an ieolated oiroumatanoe, and then in t?aebination with

computation. Within the f.computational model, merging provides a


means of aompatibilities completion verifying oc+nponent and preventing impossible design

a s+quence of other items. It is unlikely that completion reliability data, or other measures, will be available in this form. ;he second method of finding an optimum design is to generate the most promising mechanical designs. At this point, all the oomplotion oanponents for each of these designs will be known. The designs could then be ranked aoccrding to a eingle measure suoh as oost, or a pve-determined combination of measures. implementing problem in design The primary extreme difficulty in the optimization is quantifying measures suoh as equipment reliability and safety. In principle, a oost measure should be quantified. However, further more easily investigation of completion design optimization techniques is required. RESULT? In the systemu approach to well Completion design, the oonoept of defining a total Completion system is This syetem inoluden completion presented, components and attributes, reeouroes, environment, sub-systems, goals, and objectives. constraints, These briteria are noted to possess interaotiono with one anothe~ in the design prooess. Further, an integrated problem eolving method ia auggeated in whioh an engineer oonsifiers all aepeots of these oriteria in making design otmioes.

generated, combinations being from a technique importantly, merging provides aonstruoting design solutions.

More for

t3iven the strtoture of the computational model, it ehould be evident that design solutions oocur when all functional requirements have been eatisfied by one or more completion components, and when these components have been oombined suooessfully, Theoretically, all valid design solutions oould be cionatruoted by merging all design possibilities, one oholoe for each component, in an unrestricted manner. However, there are praotioal limitations to design solutions. this approaoh of constructing Consequently, an intelligent oontrol strategy must be applied to merging in the model. The control strategy in the computational model limits the number of design possibilities merged at any point in the deeign prooess, while retaining 611 other valid component design possibilities, A aeleotion crib,eria ia applied to eaoh completion oomponent to determine whioh design possibilities is often a will be merged. The seleotion oriteria suoh aa largest through bore meohanioal basis, internal diameter, but may also inolude an arbitrary By seleoting the most ordering of selections,

a.

6 -,

A COMPUTATIONAL MODEL OF WELL CWI&TfON DSSI~

SPE 18183

The oompu:atlonal model of well ocmpletion design *epreaente a formal atruoture of the eystems approaoh to well completion design. Speoifica:ly, the computational model provides:

Figure 11.
The~e are many advantages would apply alternatives, alternatives. the ?aoility a rational while uhich could be gained from implementing the computational model aa a well completion design program. First, a design prcigram and consistently all other W#timum valid

,
q

formal

description

and representation
attributes.

of

completion constraints,

components and th~ir goals,

desi~n philosophy. rapidly generate

a representation

Second, a design program would the most promit3ing design


maintaining

for design parameteiw sach aa resources, and conditions.

and . a struoture for describing the relationships interactions between completion components and design parameters.
of the design reasoning and . formalization knowledge whioh engineers employ in selecting downhole equipment and fluids.
. a two level structure of the completion design process, .in which design requirements are with valid identified and then satisfied mechanical designs.

In addition, a program aould provide to record design success and f6ilures, and this knowledge could be maintained for tutoring Finally, depending on the Oomputing purposes. environment used, the program could provide portable completions knowledge.

The advantages of a computerized design program both extend and enhance tne manner in Ehioh wellbore designed. completions are currently This iS trUe regdrding the point of generating particularly promising design alternatives multiple, while maintaining other valid solutions. Engineers currently must limit number of design the alternatives considered at any point in the design
prooeas. Consequently, the ability to maintain valid alternatives s@nifiaantly enhances outrent process of well completion design. all the

. a mechanism for oonstruating mechqnieal deeigne.

multiple,

valid

. a methodology for optimizing completion designs equipment based on measures such as oost, reliability, or safety. Figure
the

10 providee

computational

well

completion

a comparison between aspects of model, and the systems appraaah to design.

Implementation of the Computational Computer Design Program

Model

as: a

Finally, a oomputer design prosrs!n will lik?ly be a approaoh to completion design. more efficient Preliminary designo may be generated and verified in a muoh shorter p~riod of time than with current techniques. In eddttlon, useful Saailities auoh as completions equipmelt database queries, graphios funotions, and completions reccmd keeping would alleviate tedious aspects of the design prooese. CONFUSIONS

Testing the validity oF the computational model requires an aotual implementatim, or.~imulation, whose results may be oo,apared with solutions to aotual completion design problemg. Accordingly, the computational model is being implemented as a computerized design program. input design and The program acoepts data requirements ,of the probl+m, and incrementally develops multiple, valid oomplebion designs which for A provision aatlefy those requirements. ilireutly over-riding the design or speoifioation uf oertain completion components exists in the program. Thus, the program functions as a decision suppt.% system oapable of aseisting engineers in formulating completion designs. The implementation of the computational model as a oomputer program requires both conv~ntional and expert syetems programming techniques. Empirical design knowledge is represented symbolically using an expert system development environment. Numerioal, engineering programs are aoded in conventional languages suoh as Fortran, and are oo-ordfnated to operate in Conjunction with the expert system environment. The implementation is being Oonduoted on a large utillzi.ng bitmap graphios. worketatior, This development environment allows the program to produoe aompletlon design sohematios as shown in
.,

The follOWing Oonolusions are development of a computational


completion designt

drawn from ?ne model in well

10 The systems approaoh to well completion design provides a framework for describing well relations. completion components and their knowledge muet be given a However, this struoture in order to conetruot final completion designs. 2, Conventional, procedural techniques provide an awkward representation of the completion design prooess, It is possible to model the prooess of design in a non-procedural ~ completion non-determinisbio manner, This requires the Use and expert systems of both aonventlonal programming teohniquee. The Computational model of well completion design aan be implemented as a aomputer design way in whioh to enohanoe the program, completions ard ourrently designed.

3.

4,

ACKNONMDOEMENTS
The work presented herein is the result of an researoh collaborative industrially eponsored

. . ea

projeot.
~ppreoiatlon

The

authors would like to expraae to Baker Oil Toola, Ltd., Britoil P1o. ,

3;

Wmoo, Chevron Petroleum Ltd., thw UK,Department of Znergy, Elf Aquitaine, Norak Hydra a.,s., North Sea sun Oil Coispany, ,Oooidental Petroleum Ltd., Otisi Preaaure Control, Pall Well Technology, Saga Petroleum as., Total Oil Marine Ltd., and Tristar )ilfield Servioea for their uontributiona to thts awk and permission to publish the re$mlts.

Bromaton; L., Farrell, R., Kant, E.; and Martin, N.: PSWPemmingExpert Systems in 0PS5, Addison-Wesley, Reading (1985) 7.

4;

Reiohgelt, H;, and Van Harmelin, F.: Criteria Choosing Representation Languagea and for Control Regimes for Expert SyatemsW, The Knowledge Engineering Review (Deoember 1986)
Vl, N4, 10.

5. 1. Patton, L.D., and Abbott, A.: Well Completions and Workovera: Tne Systems Approaohw, Energy Publications; Dallas, Texas (1985) PP 1-7. Peden, J.M., and Leadbett.er, A;: Rationality in Completion Design and Equipment Seleotion in the North SC:. paper SPE 15887 presented at the 1986 SPE European Petroleum Conference, London, 20-22 Ootober.

and King, D.: tArtifioial Harmon, P., Intelligence in Business: Expert SystemsIt, John Riley add Sons, NewYork (1985) 38-39. !~il Rule-based MoDermott, -1.! Computer Systema*t, Artificial v19, 39-88. Configure of Intelligmce,

6.

2.

7.

Hayes-Roth~ F., Waterman, D.A., and Lenat, D.B.: lBuilding Expert Systemsw, AddisonWesley, Reading (1983) 85.

00

..

: : EE2:
.
-_-s9 : :

p--

.cn

9=-s===

:NUU95

COUPLSNG-OD COUPLINGID COIJE%INGWEIGHT c&y&13E

3.0s3 4.7 IJMT N30 2-3/3 Eu?33RD 2=3/8WE 3RD MATERIAL 3.0 VALUES

G=

LOWEIt~
MATERIAL LENGTH SETTING DEF?II

302S NIAXWFACTU== = D)m_rIPICATION

prmimmmTNG TCP llETECD %5 Iii GUN-SIZE 51BHYPmknmlI SHOT-TYPE SHOT-SIZE :* sHoi-DENsmz SHOT-P.EASfNG120DEG GUN-LENGTH 5,10 G7JNs-PEE-Itm 2 DET-NDZITiOD HYDRAULIC lrUMBFX-RuNs 1 2-10-33 DATE PERPD-ZONES (3500-3505J (3525-sss5)
m. ~
stmumafOr Qmaxr@NkmaamYxa=

w.
1) ATUMNGSllUNG=HIS
AND 2)

xHEBmluMmM~Is-mD.JmF

AND

smBPAmTAIl~mm>w

@ [~)
(EQUIP)

.*.-

COMPONENT ATTRIBUTES AT IIWIEIt LEVEL (EQI1lP)

. , ~

# --

SYMF2MS ELEMZNlYJ C4wmnu


Altrlbtrta lhwitmmmb

OF COMPLETION DINX(lN APPROAOH M dwrholt qulpmm$

COMPUTATIONAL hlODEi
rlulds and ohlti

Owtrlotlanofcmn pwwnb, Ewh UUi%ut4 m8y hwo on. C?mom valueI Iltattmd, ?ixtd urd wntmlldrh slwchofexumal Ibms aK.(tlMUwmmpltUon dw&r# I%@ wwironmmt ftirt m ha mnitrdnth mwurm, w dmply Am % cbndltlonoa Rtitrlctioncon the dmlun. Umalb rtiora to VIIYIIC41 titwla @com nant must exwtd. Camtrnlnto arc dqicri d by dodrn ruhh

COrwtralntt

lrI

mndlum

Objcctk

Ooali of tht dttlgn. Intlulw optlmlutlon $f ll181tU~W iwh M M4h mllahUUY, laftly, hem} wdlablo for Uw dodgtr. M erdh complollorh P td!! L$Jwh wwh u reMrvolrq mdng, m . oorthole. A dngh wmri+nentofan enllm comldmlca

nowuws

IWAIllyabm 8ub9y8twn

TAEKR

Oodgll A#r&
Catml

BR8 E
FIII. 9.-MWSIIVJ coorrllnMVl wllh a mrmwlcd ooir$rultllon.

Chwua

the modd Is Irianded ~ dmulsh tlw woc. of wdl com~ltttwr dwlgn, lie main @k It thw dwlum Bubtalkfi tuch M d! mie, monuminu, wrd tontrnl ore mm T cm (o IIW detlkm mcdol.

COMPLETION DESIGN REPRESRNTATION

RtOW

Chmta

Dynlmle, nmt.pro+tdurnl mcde.1Includlng emplriral and numerical dwlem knowlwJ@ ,. and 9yN0MWProNh 10Wdlcomplatlon dtdom

flu, 10-8umm@ry at tho oomputatlmlal mad

llvuwt

,Imu.1um.1

Im.nlwlm.l

r
I
1

;l... .

Irwtlllbuwlml

AMIH.lllR1,I

WWAIIB.J*M.1 amwkwtr.a

Rwlllw.wll.l

You might also like