International Journal of Physical Sciences Vol. 7(26), pp.

4049-4060, 5 July, 2012

Available online at http://www.academicjournals.org/IJPS
DOI: 10.5897/IJPS12.042
ISSN 1992 - 1950 2012 Academic Journals

Review

Fuzzy logic in candidate-well selection for hydraulic

fracturing in oil and gas wells: A critical review
Mansoor Zoveidavianpoor1*, Ariffin Samsuri1 and Seyed Reza Shadizadeh2
1
Faculty of Petroleum and Renewable Energy Engineering, Universiti Teknologi Malaysia, Johor, Malaysia.
2
Abadan Institute of Petroleum Engineering, Petroleum University of Technology, Abadan, Iran.
Accepted 4 June, 2012

During the last two decades, Fuzzy Logic (FL) Systems have been increasingly applied to the research
area of petroleum engineering. Hydraulic Fracturing (HF) as an important discipline in this area and
stimulation of oil and gas wells is an important tool for natural gas and oil exploitation, whose success
relies on proper selection of target well and target formation. To date, however, no paper has attempted
to summarize and present a critique of the existing FL literature. This paper, therefore, aims to review
the FL literature that has been conducted in candidate-well selection for HF. Selecting a target
formation(s) among a vast numbers of zones/sub-layers within huge numbers of hydrocarbon-
producing wells in a reservoir is considered a difficult task, particularly if the selection goes through a
group of parameters having different domains, attributes and features. In fact, this process had been
recognized to be complex, nonlinear and adherent with uncertainty. It is proved that methods such as
FL could reduce the uncertainty thus permit superior selection of candidates well for HF treatment. The
comprehensive review provided in this paper offers new directions for FL and its application in HF
candidate-well selection.

Key words: Candidate selection, hydraulic fracturing, fuzzy logic, uncertainty.

INTRODUCTION

During the past decade, FL has shown itself to be an
appropriate alternative for engineering judgment to deal
with the uncertainties encountered in decision-making
processes. FL has been successfully applied in many
different areas of petroleum engineering and earth
sciences including petrophysics (Finol et al., 2001),
reservoir characterization (Quenes, 2000), drilling
(Garrouch and Lababidi, 2001), permeability and rock
type estimation (Kadkhodaie et al., 2006), rock
engineering systems (Khademi et al., 2010), petroleum
separation (Liao et al., 2008), stimulation (Xiong and
Holditch, 1999) and HF candidate-well selection (Xiong
*Corresponding author. E-mail: Mansoor353@yahoo.com.
and Holditch, 1995; Yang, 2009; Yin and Wu, 2009). FL
techniques, as dened in this paper, refer to fuzzy
concepts, which include fuzzy set and fuzzy logic.
To date, nevertheless, no paper has attempted to
summarize and present a critique of the existing FL
literature. The aim of this paper is to provide an overview
of the application of FL techniques in HF candidate-well
selection. In fact, FL established itself as a reliable
methodology for dealing with any sort of uncertainty,
especially in HF candidate-well selection, design and
execution. In this section, candidate-well selection, HF
and the needs for them will be explained briefly so as to
make it understandable to colleagues from a broad range
of scientific disciplines. Also, it will be discussed why FL
is of particular importance for helping extend the
productive life of the reservoirs and related assets in oil
4050 Int. J. Phys. Sci.
Figure 1. The HF triangle success.
and gas industry. Actually, this paper will review an
application that incorporates FL in HF candidate-well
selection.
Upstream oil and gas operations involve a range of
activities, including exploration, drilling and production of
crude oil and/or natural gas. Reservoir stimulation is one
of the main activities, whose main purpose is to enhance
the property value by the faster delivery of the petroleum
fluid and/or to increase ultimate economic recovery
(Economides and Nolte, 2000). Once a hydrocarbon-
bearing reservoir has faced difficulties to produce
economically, a vital question need answer by the
production team. How the recognized problem(s) could
effectively alleviated and/or eliminated? One may search
for a solution other than stimulation jobs, so the question
rose toward; what is done in other technical areas? Does
drilling a new well (infill drilling) to compensate for the
reduction of productivity, serve as the only reliable
answer? Vincent (2011) shows that HF of existing wells
had stronger economics than infill drilling. In addition, the
possibility of further discoveries of giant oil and gas
reservoirs in the world is low; new stimulation strategies
through utilization of available technologies, would lead to
maximizing recovery rates and efficiencies from the
existing fields. In fact, a stimulation technology such as
HF, which is the pumping of fluids at high rates and
pressures to break the rock to increase the output of a
well, considered as an ever-increasing focus area for
upstream industry. Hence, in order to be successfully
conducted; its directly depending on candidate-well
selection phase. Candidate-well selection is the process
of choosing wells that have potential for higher production
and better return of investment after stimulation job.
Actually, the effectiveness of the HF treatment is a
function of three critical parts, which tied together;
candidate-well selection, treatment design, and field
operation (Figure 1). In fact, they are the triangle success
factors that must be linked to each other. Applying the
best treatment design and field procedures to the wrong
candidate-well will results in a failure of the whole
operation. In other words, all of the three factors should
be performed well to guarantee the success of HF
treatment.
It has found that the methods, as applied in HF
candidate-well selection discipline, in the past two
decades, can be divided into two broad elds,
encompassing: (1) conventional techniques; and (2)
advanced approaches. The reason for their
nomenclatures had derived from the different approaches
that each of them is applying to choose a target well or
target formation. The former mainly deals with
engineering, geological, etc. aspects in decision making
process. More details about conventional candidate-well
selection for HF could be found by Zoveidavianpoor et al.
(2012). The later mainly fill the gap for classification and
manipulation of the parameters and mainly employs
Artificial Intelligence (AI) methods that offers the
opportunity to examine the data, their relationships, and
in other ways and to maximize the concealed potential.
In answering the question, whats the need for using FL
for HF candidate-well selection? One may observed the
independent nature of domains involved in HF candidate-
well selection procedure such as geological aspect,
reservoir and fluid characteristics, etc. and a relatively
sparse data set. Those considerations highlighted the
underlying problem of making predictions based on
existed approaches for which a large degree of
uncertainty exists. For such systems where only
ambiguous and imprecise information may be available,
FL provides a way to understand system behavior by
allowing us to interpolate approximately between
observed input and output situations.
Inspired by the need of AI methods, the previous
research conducted by the advanced methods to choose
candidate-well will be reviewed and the key points are
highlighted. As the main focus of this paper, a state of the
art FL approach will be explained afterwards. Next, as a
cutting-edge technology for managing uncertainty, it will
be illustrated that Type-2 Fuzzy Set and System (T2
FSS) could fill the gap between recent developments in
FL and HF candidate-well selection. So, this new trend
and option for the use of T2 FSS will be briefly explained
subsequently. In the discussion section, the main works
critically discussed and a conclusion sums up the main
results and points towards potential future developments.

PREVIOUS RESEARCH
As the volume of data increases, human cognition is no
longer capable of deciphering important information from
it by conventional techniques. Mohaghegh (2005)
demonstrated that data mining and machine learning
techniques must be used in order to deduce information
and knowledge from the raw data that resides in the
databases. Although the required parameters for
selecting a well for HF are relatively identical,
classification and manipulation of the structure data is
different. This fact comes from different characteristics for
each hydrocarbon reservoir. Thus, there is a need to
provide a methodology that allows selecting the desired
well/layer with minimum time and costs and provide a
framework to overcome the difficulties in conventional
techniques. As indicated by Moore and Ramakrishnan
(2006) first-pass evaluation using Moving Domain
Analysis (MDA) technique by utilization of existing data
can help prioritize wells for in-depth analysis. Using ANN
methods, Shelley (1999) explains that operators can
screen candidate-well before performing more detailed
engineering analysis. The number of hours required to
complete the ANN procedures are minimal compared
with hours required for conventional geologic and
engineering evaluation. Reeves et al. (1999) developed a
multi-process candidate selection methodology to identify
viable stimulation candidates. It consists of production
comparisons, engineering based performance
assessments, and Pattern Recognition technology. Malik
et al. (2006) presented new methodology, which uses
different trained ANN that is flexible to design GA based
on different parameters. To develop a methodology for
candidates in a short time with limited budget, Flores
(2009) took advantages from case histories from prior
stimulation treatments; they then evaluated to find trends
and success indicators by designing a stimulation
workflow. Using this improved methodology, success was
increased to 75%, and costs were reduced by 40%.
Emrich et al. (2001) and East and Craig (2006) showed
that diagnostic algorithms improved candidate selection
and contributed to improved performance.
There are a number of different emerging AI
approaches that have been applied in this way in recent
years (Mohaghegh et al., 2000) to the manipulation of the
typical fracturing data sets. This approaches are quite
varied in their requirements and their upside potential
(Oberwinkler et al., 2004; Conway et al., 1985) and
include approaches such as FL, ANN, GA, self-
questioning matrices, pattern recognition, self-organizing
maps, etc. AI has two important components, that is, FL-
based models that embed existing structural human
knowledge such as experience, expertise, and heuristics,
Zoveidavianpoor et al. 4051
rules of thumb into efficient framework such as IF-THEN
rules. Experimental data learning methods such as ANN
and Support Vector Machines are those that learn from
experimental data (examples, samples, measurements,
records, patterns, observations).
Although the work reported by Oberwinkler and
Economides (2003) was performed in order to primarily
identify HF candidate potential, it is an excellent
demonstration of the measure of application that can be
achieved with the AI approaches when quality data is
available. Grieser et al. (2008) describe a case whereby
a conventional approach, including cross-plots, bar
charts, cumulative frequency plots, etc., did not
determine any definitive trends or correlations. They then
proceeded to perform a more detailed data-mining
exercise and applied the Self-Organizing Map approach.
The results were clear, with definitive trends, which could
then be applied to improve the future programs. McVey et
al. (1996) developed an ANN, which trained with existing
data to identify influential parameters in HF; it was found
that the flow parameters have more influence than other
parameters. Reeves et al. (1999, 2000) concluded that
virtual intelligence methods were relatively better
compared to other approaches. The advantage of various
virtual intelligence techniques is that it can be designed to
mimic the thinking process of a completion engineer
entrusted with the job of selecting refracturing
candidates, but the downside is data requirement and
expertise (Reeves et al. 1999). Other studies such as
Xiong and Holditch (1995); Yang (2009); Yin and Wu (2009),
demonstrated that fuzzy logic is a promising tool that can be
used for stimulation candidate-well selection. The first
research used a fuzzy logic evaluation method to select and
rank candidate wells for different stimulation treatments such
as HF, acid fracturing, and matrix acidizing. They called their
method "fuzzy expert system", which selects a candidate
well based on nine fuzzy variables. The second study
determined and analyzed the factors influencing the effect of
oil well fracturing and called his model a "fuzzy mathematics
model" to choose target well and formation for HF from
twelve factors (fuzzy variables). In the third study, Yin and
Wu (2009) were analyzed quantitatively the main factors of
influencing the fracturing effects, gives the relation between
fracturing effects and the parameters, and divides the grade
intervals of each influencing factor quantitatively, then
establishes fuzzy judging mathematical model of choosing
fracturing well. It is believed that advanced methods such
as FL could be better handle uncertainty (Zadeh, 2006)
existed in candidate-well selection.
Table 1 illustrated the application of AI methods by
different authors in HF candidate-well selection. Some
key issues could be concluded from the literature of
advanced candidate-well for HF:
4052 Int. J. Phys. Sci.
Table 1. The application of AI methods in HF candidate-well selection.
Investigators
Moore and Ramakrishnan (2006)
Shelley (1999); McVey et al. (1996)
Reeves et al. (1999)
Malik et al. (2006)
East and Craig (2006); Emrich et al. (2001)
Economides and Martin (2007); Mohaghegh et al. (2000);
Oberwinkler and Economides (2003); Reeves et al. (2000); Grieser
et al. (2004)
Xiong and Holditch (1995); Yin and Wu (2009); Yang (2009)
1. ANN and FL are the two dominant AI methods that had
been applied successfully in this filed.
2. The nature of candidate-well selection is non-linear
because it consists of several input parameters, which
are qualitatively and quantitatively different with each
other.
3. FL enables a convenient framework to combine and
analyze these data even independently of their source or
characteristics.
4. The major disadvantage of ANN method is that it
needs large input/output observation for training or
learning procedures and does not account for
uncertainty. Two former key issues had been the impetus
behind deep investigation on three main works that
utilized FL-based method in HF candidate-well selection.
In this way, this paper will look at a cutting-edge
procedure to fill the research gap between recent
developments in FL to decrease the uncertainty existed
in HF candidate-well selection.
FUZZY LOGIC
FL is an extension of conventional Boolean logic
developed to handle the concept of partial truth truth
values between completely true and completely false.
It was introduced by Dr. Lotfi Zadeh of UC/Berkeley in the
1960s as a means to model uncertainty (Zadeh, 1965).
Zadeh says that rather than regarding fuzzy theory as a
single theory, we should regard the process of
fuzzification as a methodology to generalize any specific
theory from a crisp (discrete) to a continuous (Fuzzy)
form. According to Zadeh (2006), uncertainty is an
unavoidable attribute of information. By using the rules of
probability, scientists were capable of dealing with such
uncertainties in information. With fuzzy set theory coming
into existence, it is performed better by FL. Zadeh (2008),
Methods Used
Moving Domain Analysis and Artificial Neural Network
Artificial Neural Network
Pattern Recognition
Artificial Neural Network and Genetic Algorithm
Diagnostic Algorithms
Fuzzy Logic, Artificial Neural Network, Genetic Algorithm,
Self-Questioning Matrices, Pattern Recognition, and Self-
Organizing Maps
Fuzzy Logic
in answering the question "is there a need for FL?",
believes that, today, close to four decades after its
conception, fuzzy logic is a precise logic of imprecision
and approximate reasoning, which shows itself to be
more effective than an attempt at the
formalization/mechanization of human reasoning
capabilities. Fuzzy sets theory, as a soft computing
technique, has established itself as a new methodology
for dealing with any sort of ambiguity and uncertainty.
Soft computing, as introduced by Zadeh (1992), includes
approaches to human reasoning, which try to make use
of the human tolerance for incompleteness, uncertainty,
imprecision, vagueness, and fuzziness in decision-
making problems (Jang et al., 1997).
FL can be used in order to reduce the uncertainty in the
data collected and can model vagueness in data and/or
relationship in a formal way. This technique is able to
manipulate fuzzy qualitative data in terms of linguistic
variables. In FL, unlike standard conditional logic, the
truth of any statement is a matter of degree. For example,
for the rule if (skin is high) then (candidate is very good),
both variables, high and very good, map to range of
values. Fuzzy systems rely on membership functions to
explain to the model how to calculate the correct value
between zero and one. The degree to which any fuzzy
statement is true is denoted by a value between zero and
one. In fuzzy inference system, there are 5 steps such as
fuzzy inputs, combination of inputs with AND (OR)
method, implication, aggregation of all outputs and
defuzzification (Tinkir, 2011). A typical fuzzy system
consists of a rule base, membership functions and an
inference procedure, as presented in Figure 2. The
membership function A(x) describes the membership of
the elements x of the base set X in the fuzzy set A,
whereby for A(x) a large class of functions can be taken.
Reasonable functions are often piecewise linear
functions, such as triangular or trapezoidal functions.
 Zoveidavianpoor et al. 4053

Figure 2. A typical fuzzy logic system.

Table 2. The standard operation in fuzzy logic system.

Algebraic Equation Based on:

Fuzzy operator name Symbol Description
operator name Algebraic Fuzzy
applied to two fuzzy sets A and B with the
Intersection AND AB (x)= {A(x), B(x)}, xX AB (x)= min{A(x), B(x)}, xX
membership functions A(x)and B(x)
applied to two fuzzy sets A and B with the
Union OR AB (x)= {A(x), B(x)}, xX AB (x)= max{A(x), B(x)}, xX
membership functions A(x)and B(x)
Complement NOT
applied to fuzzy sets A with the A ( x ) 1 A ( x ), x X A ( x) 1 A ( x), x X
membership functions A(x)

Table 2 shows the standard operation in FL
system. Whilst the operations according to
Intersection and Union are based on min/max
operations, the complement is an algebraic
(arithmetical) one. Union and intersection can also
be defined in an algebraic manner but giving
different results as indicated in Table 2.
Selection of operators solely depends on the
problems of applications. The shaded region in
Figure 3 shows the membership function of the
fuzzy set AB, which results from the intersection
of fuzzy sets A and B. Here A and B are the
membership functions for the fuzzy sets A and B,
respectively. For example, A could be a
membership function for the fuzzy set of skin
value or water saturation. Triangular or
trapezoidal functions usually considered as linear
functions. As shown in Figure 3, the two fuzzy
sets have presented as triangular membership
functions. Normally, a membership function
selected by trial and error to be representing the
base set. There are five basic membership
functions. These functions along with their
descriptions are represented in Table 3.
4054 Int. J. Phys. Sci.

Figure 3. Membership function for fuzzy set AB realized by function min (A,
B).

Table 3. The major function used in fuzzy logic system.

Function Name Parametric Representation Graphical Representation

(a) Triangular MF

0, xa
( x a)
1

, x ( a, b) 0.8

(b a )

Membership Grades
Triangular Function ( x, a, b, c) 0.6

(c x ) , x (b, c)
(c b ) 0.4

0, xc 0.2

0
0 20 40 60 80 100

0, xa
( x a ) /(b a), x (a, b)

A
Trapezoidal function
1, x (b, c)
(d x) / d c), x (c, d )

Normalized Gaussian function

Two Sigmoidal functions

 Zoveidavianpoor et al. 4055

Table 3. Contd.

Generalized Bell function

Table 4. Reasons for the utilization of T2 FSS.

Why We Required T2 FSS?

Against undesirably affect from uncertainty conditions of parameters of the rule base and membership sets
No previous work had used T2 FSS
On account to the fact that T2 FL estimation provides the possibility to indicate the uncertainties in which is crucial in maintaining
high quality results
To test the proposed model to make T2 FL systems reasonable and reliable
To present a new application for using T2 FL intervals
To understand the performance of the T2 FSS
To deal with the potential uncertainty of experts knowledge and experience
The human experts have uncertainty in membership function specification
T2 FL system permitted them to model rule uncertainties
To handle the large amounts of uncertainties present in real world environments
For demonstration of application of T2 FL systems to predict a critical parameter in a system
The experts will have different opinions, expectations the interpretations for the requirements specification as well as for the
applicants matching and ranking

There are two types of Fuzzy Sets and Systems
(FSSs); Type-1 (T1) and Type-2 (T2). In T1 FSS, the
membership grade is a number in the range [0, 1]. T2
FSS is an answer to the question of what to do when
there is uncertainty about the value of the membership
function itself. In T2 FS, the membership grade is itself a
fuzzy set in [0, 1]. As a result, a T2 FSS is governed by
fuzzy membership functions that are three-dimensional
and include a mark of uncertainty. T2 itself has two
categories named (1) General T2 FSS, and (2) Interval
T2 FSS. Because of using sophisticated mathematical
concepts, the application of the first category is not
popular and hence little publications are available. On the
other hand, utilization of the later is widespread among
literature. For an Interval T2 FS that third-dimension
value is the same (for example, 1) everywhere, which
means that no new information is contained in the third
dimension of an Interval T2 FS. Therefore, for such a set,
the third dimension is ignored, and only the Footprint of
Uncertainty (FOU) is used to describe it. It is for this
reason that an Interval T2 FS is sometimes called a first-
order uncertainty fuzzy set model, whereas a general T2
FS is sometimes referred to as a second-order
uncertainty fuzzy set model.
IMPLICATIONS FOR THE FUTURE RESEARCH
DIRECTIONS
Gas Research Institute (GRI) believes that candidate-well
selection phase is where the greatest industry benefit
resides (Ely et al., 2000). Selecting a target formation(s)
among a vast numbers of zones/sub-layers within huge
numbers of hydrocarbon producing wells in a reservoir, is
considered a difficult task, particularly if the selection
goes through a group of parameters having different
domains, attributes and features. The trend of candidate-
well selection process for HF treatment had recognized to
be complex, nonlinear, un-equilibrium, and adherent with
ambiguity and uncertainty. Therefore, utilization of
4056 Int. J. Phys. Sci.
modern and right problem-solving tool such as T2 FSS to
handle such uncertainties should be considered a great
concern to the petroleum industry. Indeed, new
requirements force the previous methods to advance and
novel techniques expected to meet the requirements and
remove the existing weakness. In highlighting this need,
the question has been answered about why T2 FSS
should be used. By reviewing its application in the
industry, the second question raised; what are the
industry motivations to use this method? Despite of the
application in several areas of science and engineering,
yet, they application in petroleum industry have not been
fully developed and a few papers available (Olatunji et
al., 2011a, b). A review of industrial applications of T2
FSS in solving complicated problems (Dereli et al., 2011)
indicates that there have been different reasons for
authors to use type-2 FSS applications. Those reasons
are illustrated in Table 4.
DISCUSSION
The nature of candidate-well selection is non-linear
because it consists of several input parameters, which
are qualitatively and quantitatively different. This is the
privilege of fuzzy set theory that enables a convenient
framework to combine and analyze these data even
independently of their source or characteristics. Two of
the major concepts associated with FL, Fuzzy
membership functions and linguistic variables, can be
especially adapted to suit applications to solve the
complex nature of HF candidate-well selection. In fact,
the former enable one to perform quantitative calculations
in fuzzy decision making, while the concept of linguistic
variables serves the purpose of providing a means of
approximate characterization of phenomena that are too
complex to be amendable to description in conventional
quantitative terms.
In addition to the four major key points of utilization of
AI methods that highlighted the advantages of FL-based
approaches in Previous Research Section, we chose
particularly three main cutting-edge works because they
utilized FL method in HF candidate-well selection, that is,
Xiong and Holditch (1995), Yang (2009) and Yin and Wu
(2009). Table 5 lists the applied steps in their FL system.
There are some problems associated with the current
fuzzy systems applied in the field of HF candidate-well
selection. The first problem is the structure identification
of the relation R that links the inputs to the output. In
fuzzy system modeling, the structure represented by
means of fuzzy is, if-then rules. None of mentioned
literatures had shown or even discussed the rule base
section of their created FL system. This important step
was correlated to weighting factor, which is different in all
proposed methods. As can be seen from Table 6, there
are noticeable differences between their allocated
weighing factors, which could finally affect the candidate-
well selection process result. Secondly, this
heterogeneity in the relative importance of the fuzzy
variable, which makes the weighting factor, is not the only
noticeable problem; the number of the variables is
different in each work; 9, 12 and 7 for Xiong, Yang, and
Yin and Wu, respectively. This may be justifying, because
the structure of the fuzzy, ifthen rules, is determined
priori subjectively from other sources such as experts
knowledge. However, these rules varied among the
experts and even for the same expert at different times.
Third, by applying inaccurate rules and incorrect
functions, the achievement will be possible imprecise
conclusions. For example, ignoring the effects of fluidity
in the system (Yin and Wu, 2009), low weight and priority
for skin factor (Yang, 2009), and participation a variable
such as wellbore condition in the system, which could be
affect the system results in a mature filed by decreasing
the number of selected parameters (Xiong and Holditch,
1995). Table 5 shows the fuzzy system application in the
mentioned studies. The major steps toward establishing the
fuzzy system in each of the study are presented. Finally, as
it is obvious in Table 5, the basic steps of the proposed
works are all similar and in contrast with the current
development in fuzzy logic system (Jang et al., 1997), a
significant new trend should be used to managing
uncertainty in which the current FL systems (T1 FSS) in
literature of HF candidate-well selection could not handle.
Consider a fuzzy set (very low, low, medium, high, and
very high) for a fuzzy variable such as porosity. As can
be seen from Table 6, different boundaries were
associated with it. It is worthy to note that these
differences are logical, because the nature and
characteristics of reservoir are dissimilar from region to
region. Therefore, its clear that there are uncertainties
and we need a model to capture them.
T1 FL System was previously developed and
implemented to overcome the difficulties that were
presented in conventional methods in HF candidate-well
selection (Xiong and Holditch, 1995; Yang, 2009; Yin and
Wu, 2009). However, T1 FSS is not capable of fully
capturing the linguistic uncertainties in the terms used
and inconsistency of the expert's decision-making.
Actually, uncertainty involved in any problem-solving
situation is a result of some information deficiency. There
exist many different causes of uncertainty such as (a)
lack of information, (b) abundance of information, (c)
conflicting evidence, (d) ambiguity, (e) measurement and
(f) belief (Zimmermann, 2000). In general, these various
information deficiencies may result in different types of
 Zoveidavianpoor et al. 4057

Table 5. Fuzzy system application in the literature regarding candidate-well selection.

Authors Xiong and Holditch (1995)
Fuzzy Variables (9)
Permeability/Viscosity Ratio [0.25]
Skin factor [0.2]
Net pay thickness [0.1]
Water saturation [0.1]
Formation pressure gradient [0.1]
Fuzzy
Wellbore condition [0.1]
Variables
Drainage area [0.05]
Formation depth [0.05]
Porosity [0.05]
Yang (2009) Yin and Wu (2009)
Influential Factors (12) Factor Volumes (7)
Formation pressure [0.13] Skin factor [0.3]
Permeability [0.13] Permeability [0.15]
Recovery percent of reserves [0.13] Fluidity [0.15]
Water saturation [0.13] Water cut [0.1]
Reservoir thickness [0.12] Reservoir pressure [0.1]
Skin factor [0.1] Fractured width [0.1]
Effective porosity [0.07] Fracturing effects [0.1]
Production rate of the well [0.05]
Recovery percent of reserves [0.05]
Flow coefficient [0.03]
Vertical heterogeneous coefficient [0.03]
Producing pressure drop [0.02]

Levels Grades Comment Values

Excellent Candidate [1] Optimum [0.64] First Selection
Levels Good Candidate [0.7] Better [0.25] Second Selection
Possible Candidate [0.5] Selection [0.1] Unsuitable Selection
Not a Candidate [-1] Non-Selection [0.01] Not Available

Constants m & n Upper limit (dup) & Lower limit (ddown) Index Parameters a & b
d up d down d iju d ijd
m a aij
Constants 2 2
d up d down (d iju d ijd ) 2
n b bij
2 ln 2 4 ln 2

Gaussian function Gaussian function Gaussian function

Membership x m 2 ( x a ) 2 ( d i aij ) 2
( ) [ ( ) ] [( )]
Function
F ( x) e n rij e b
uij (d i ) e
bij

Relationship Matrix Fuzzy Analogues Matrix Fuzzy Relationship Matrix

F11 F12 F13 F14 r11, r12, ..., r1N r11 r12 ... r1m
r , r ,..., r
r r ... r2 m
R ... ... ... ... R1 21 22 2N
Matrix name
R uij ( d i ) 21 22
F F F F r31, r32 ,..., r3 N ... ... ... ...
91 92 93 94 rn1 rn 2 ... rnm
r41, r42 ,..., r4 N

Min Max Max

Operator
9
B
b j min 1, I i Fij b j AK bki V AR V1 ,V2 ,V3 ,...Vm
T

i 1 K 1
4058 Int. J. Phys. Sci.
Table 6. Allocated weighting factor in the literature.
Fuzzy Variables Xiong and Holditch, 1995
Permeability (mD) >1
Skin < 5.5
Water Saturation (%) > 37.5
Porosity (%) 6 - 26
Net Pay Thickness (ft) > 100
Drainage Area (acre) > 110
Formation pressure (psi) 2000-3000
Formation Depth (ft) < 10000
uncertainty (Klir and Wierman, 1999). IT2 FS can handle
such uncertainties in a better way because they provide
more parameters and more degrees of freedom
(Hameed, 2009). IT2 FL system offer better capabilities
to handle linguistic uncertainties by modeling the
uncertainties using T2 Membership Functions and
provide diagnosticians with decision-making flexibilities.
The development of a new methodology for HF
candidate-well selection via utilization of T2 FL system
forms a proposed new research trend in this paper, which
will provide the capability to handle linguistic uncertainties
better than previous studies. Finally, because there is
uncertainty regarding the value of the membership
function, and variables associated with each set in HF
candidate-well selection, more sophisticated types of
fuzzy sets, T2 FSS is recommended at the end of this
paper.
CONCLUSIONS
The process of selection wells and layers for HF is the
process through which a service oil company or an
operator determines the best fit from a dozen of well and
formation candidates whose impact on the operation
performance can be very strong. As a consequence, the
role of selecting target wells and target layer within this
industry is critical. Typically a company will utilize
different types of data and information in different
domains and attributes having different format and
configuration such as petrophysical well logs, drilling,
production, and reservoir files, rock mechanical data, and
etc. to screen and ultimately select a candidate(s).
Considering the complex, nonlinear, and imprecise
nature of data in this process, decision makers are
beginning to add other methods to the selection process
in order to handle uncertainty and further assist in
determining the best fit for a treatment. It becomes
obvious that the ability to manage uncertainty turns out to
Yang, 2009 Yin and Wu, 2009
0-0.3 0 - 30
5-25 4-5
0 - 40 75 - 80
0 -10 N/A
33-164 10 -16
N/A N/A
1450 - 1741 1450 - 2030
N/A N/A
be a critical issue for selecting HF candidate selection. In
this paper, the independent nature of domains involved in
HF candidate-well selection procedure highlighted the
underlying problem of making predictions based on
existed approaches for which a large degree of
uncertainty exists. However, it is impossible to develop a
precise mathematical model that can predict the right
candidate. For such systems where only ambiguous and
imprecise information may be available, FL provides a
way to understand system behavior by allowing us to
interpolate approximately between observed input and
output situations.
The weaknesses of the other AI methods (large
input/output observation and no uncertainty capturing) in
contrast with the advantage of FL that enables a
convenient framework to combine and analyze these
data even independently of their source or
characteristics, had been the impetus behind deep
investigation on the main works that utilized FL-based
method in HF candidate-well selection. In this way, this
paper will look at a cutting-edge procedure to fill the
research gap between recent developments in FL to
decrease the uncertainty existed in HF candidate-well
selection. Certainly, a reliable methodology is needed to
handle uncertainty. After conducting a comprehensive
literature review on the applications of AI and specially FL
technique in HF candidate-well selection research, a
research area has been identied for further study. It is
recommended that T2 FSS can be incessantly adopted in
the research area of candidate-well selection because
they can assist in developing models to make decisions
which are often viewed as complicated, and uncertain.
ACKNOWLEDGEMENT
The authors of this paper would like to express their
gratitude to Universiti Teknologi Malaysia due to the
administrative and financial supports during this study.

REFERENCES
Conway MW, McMechan DE, McGowen JM, Brown D, Chisholm PT,
Venditto JJ (1985). Expanding recoverable reserves through
refracturing. Paper SPE 14376 presented at the SPE annual
technical conference and exhibition. 22-25 September. Las Vegas,
Nevada, USA. p. 8.
Dereli T, Baykasoglu A, Altun K, Durmusoglu A, Trksen B (2011).
Industrial applications of type-2 fuzzy sets and systems: A concise
review. Comput. Ind. 62(2):125-137
East L, Craig D (2006). Refracs reverse production dips: single- and
multiple-zone wells see output gains after refracturing. E&P Mag.
79(11):45-47.
Economides MJ, Nolte KG (2000). Reservoir stimulation. 3rd Edition,
John Wiley & Sons Ltd, London p. 856.
Ely JW, Rothenberg M, Krupa A, McDougal F, Conway M, Reeves S
(2000). Restimulation program finds success in enhancing
recoverable reserves. Paper SPE 63241 presented at the SPE
Annual Technical Conference and Exhibition, 1-4 October. Dallas,
USA. p. 12.
Emrich E, Shaw D, Reasoner S, Ponto D (2001). Codell restimulation
evolve to 200% rate of return. Paper SPE 67211 presented at the
SPE production and operations symposium. 24-27 March. Oklahoma
City, Oklahoma, USA. p. 8.
Finol J, Guo YK, Jing DY (2001). A rule based fuzzy model for the
prediction of petrophysical rock parameters. J. Pet. Sci. Eng. 29(2):
97-113.
Flores DQ (2009). Finding additional resources in a mature field:
refracturing engineering workflow for a successful campaign in
Vicksburg basin, south Texas. Abstract SPE 125260 presented at the
SPE Tight Gas Completions Conference. 16-17 June. San Antonio,
Texas, USA.
Garrouch AA, Lababidi HMS (2001). Development of an expert system
for underbalanced drilling using fuzzy logic. J. Pet. Sci. Eng. 31(1):
23-39.
Grieser B, Shelley B, Johnson BJ, Fielder EO, Heinze JR, Werline JR
(2008). Data analysis of Barnett Shale completions. SPE J. 13(3):
366-374.
Hameed IA (2009). Simplified architecture of a type-2 fuzzy controller
using four embedded type-1 fuzzy controllers and its application to a
greenhouse climate control system. Proceedings of the Institution of
Mechanical Engineers, Part I: J. Syst. Control Eng. 223(5): 619-631
Jang JSR, Sun CT, Mizutani E (1997). Neuro-fuzzy and soft computing.
A computational approach to learning and machine intelligence.
Prentice-Hall: Englewood Cliffs, New Jersey, p. 614.
Kadkhodaie IA, Rezaee MR, Moallemi SA (2006). A fuzzy logic
approach for estimation of permeability and rock type from
conventional well log data: an example from the Kangan reservoir in
the Iran Offshore Gas Field. J. Geophys. Eng., 3(4):356-369.
Khademi HJ, Shahriar K, Rezai B, Bejari H (2010). Application of fuzzy
set theory to rock engineering classification systems: an illustration of
the rock mass excavability index, Rock Mech Rock Eng. 43(3):335-
350.
Klir GJ, Wierman MJ (1999). Uncertainty-based information: elements
of generalized information theory, Physica-Verlag Heidelberg, New
York p. 185.
Liao RF, Chan CW, Hromek J, Huang GH, He L (2008). Fuzzy logic
control for a petroleum separation process. ENG APPL ARTIF
INTEL. 21(6):835-845.
Malik K, Mohaghegh SD, Gaskari R (2006). An intelligent portfolio-
management approach to gas storage field deliverability maintenance
and enhancement: part one- database development and model
building. Paper SPE 104571 presented at the SPE eastern regional
meeting. 11-13 October. Canton, Ohio, USA. p. 6.
McVey DS, Mohaghegh S, Aminian K, Ameri S (1996). Identification of
parameters influencing the response of gas storage wells to hydraulic
fracturing with the aid of a neural network. SPE Comput Appl. 8(2):
Zoveidavianpoor et al. 4059
54-58.
Mohaghegh SD (2005). Recent developments in application of artificial
intelligence in petroleum engineering. J. Pet. Technol. 57(4):86-91.
Mohaghegh SD, Reeves S, Hill D (2000). Development of an intelligent
system approach for restimulation candidate selection. Paper SPE
59767 presented at the SPE/CERI gas technology symposium, 3-5
April. Calgary, Canada. p. 11.
Moore LP, Ramakrishnan H (2006). Restimulation: candidate selection
methodologies and treatment optimization. Paper SPE 102681
presented at the SPE annual technical conference and exhibition. 24-
27 September. San Antonio, Texas, USA. p. 9.
Oberwinkler C, Economides M (2003). The definite identification of
candidate wells for refracturing. Paper SPE 84211 presented at the
SPE annual technical conference and exhibition. 5-8 October 2003.
Denver, Colorado, USA. p. 17.
Oberwinkler C, Rythammer G, Zabgl G, Economides MJ (2004). New
tools for fracture design optimization. Paper SPE 86467 presented at
the SPE international symposium and exhibition on formation
damage control. 18-20 February. Lafayette, Louisiana, USA. p. 13.
Olatunji SO, Selamat A, Abdul Raheem AA (2011a). Modeling the
permeability of carbonate reservoir using type-2 fuzzy logic systems.
Comput. Ind. 62(2):147-163.
Olatunji SO, Selamat A, Abdul Raheem AA (2011b). Predicting
correlations properties of crude oil systems using type-2 fuzzy logic
systems. Expert Syst. Appl. 38(9):10911-10922.
Quenes A (2000). Practical application of fuzzy logic and neural
networks to fractured reservoir characterization. COMPUT GEOSCI.
26(8):953-962.
Reeves SR, Bastian PA, Spivey JP, Flumerfelt RW, Mohaghegh SD,
Koperna GJ (2000). Benchmarking of restimulation candidate
selection techniques in layered, tight gas sand formations using
reservoir stimulation. Paper SPE 63096 presented at the SPE Annual
Technical Conference and Exhibition. 1-4 October. Dallas, Texas,
USA. p. 16.
Reeves SR, Hill DG, Tiner RL, Bastian PA, Conway MW, Mohaghegh
SD (1999). Restimulation of tight gas sand wells in the Rocky
Mountain region. Paper SPE 55627 presented at the SPE Rocky
Mountain regional meeting. 15-18 May. Gillette, Wyoming, USA. p. 9.
Shelley RF (1999). Artificial neural networks identify restimulation
candidates in the Red Oak field. Paper SPE 52190 presented at the
SPE mid-continent operations symposium. 28-31 March. Oklahoma
City, Oklahoma, USA. p. 7.
Tinkir M (2011). A new approach for interval type-2 by using adaptive
network based fuzzy inference system. Int. J. Phys. Sci. 6(19):4502-
4518.
Vincent MC (2011). Restimulation of Unconventional Reservoirs: When
Are Refracs Beneficial? J. Can. Pet. Technol. 50(5):36-59.
Xiong H, Holditch SA (1995). Using a fuzzy expert system to choose
target well and formations for stimulation. In Braunschweig et al (eds)
Artificial intelligence in the petroleum industry: symbolic and
computational applications. Editions Technip, Paris, pp. 361-379.
Yang E (2009). Selection of target wells and layers for fracturing with
fuzzy mathematics method. paper presented at the sixth international
conference on fuzzy systems and knowledge discovery, pp. 366-369.
Yin D, Wu T (2009). Optimizing well for fracturing by fuzzy analysis
method of applying computer. paper presented at the 1st IEEE
international conference on information science and engineering. pp.
286-290.
Zadeh LA (1965). Fuzzy sets. Inform Control 8:338-353.
Zadeh LA (1992). Fuzzy logic, neural networks, and soft computing.
One-page course announcement of CS 294-4. The University of
California at Berkeley.
Zadeh LA (2006). Generalized theory of uncertainty (GTU) principal
concepts and ideas. Comput Stat Data Anal. 51:15-46.
Zadeh LA (2008). Is there a need for fuzzy logic? Inf. Sci. 178:2751-
2779.
Zimmermann HJ (2000). An application-oriented view of modeling
4060 Int. J. Phys. Sci.

uncertainty. Fuzzy Sets Syst. 122:190-198.

Zoveidavianpoor M, Samsuri A, Shadizadeh SR (2012). A review on
conventional candidate-well selection for hydraulic fracturing in oil
and gas wells. Int. J. Eng. Technol. 2(1):51-60.