Kasdi Merbah University - Ouargla

Faculty Of Hydrocarbons, Renewable Energy, Science Of

The Earth and Universe
Department Of Drilling And Oil Mechanics

Master’s thesis in drilling

Presented by: supervised by:

Akik akrem Dr.Hachana oussama

Chelouche aissa

Abd esselam mohammed

Optimization of drilling parameters

using metaheuristic technique

Academic year:2017/2018
To our families, our mothers,
Our fathers, our brothers
And all the friends
Who stood with us
And supported us
 ACKNOWLEDGEMENTS

First and foremost, we thank Allah for endowing us with health, patience, and
knowledge to complete this work.
We would like to express our deep and sincere gratitude to our advisor Dr.Hachana
Ouassam. The doors to him were always open whenever we ran into a trouble or had
questions. It was a great privilege and honor to work and study under his guidance.
Finally, we are extremely grateful to our parents for their love, prayers, caring and
sacrifices for educating and preparing us for our future. Also we are very much thankful to
our sisters, brother, and lovers for their support and valuable prayers.
TABLE OF CONTENTS
DEDICATION
ACKNOWLEDGEMENTS
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
NOMENCLATURE
General introduction………………………………………………………………………..….1
Chapter I Drilling Parameters:Definition, Classification and Effect On Drilling
Performance
I. 1. Introduction…………………………………………………………………………….....3
I. 2. Drilling Principles…………………………………………………………………………3
I. 2. 1. Dependent Variables...........................................................................................4
I.2 .1. 1. Penetration Rate...................................................................................4
I. 2 .1. 2. Torque.................................................................................................5
I. 2 .1. 3. Flush Medium Pressure.......................................................................5
I. 2 .1. 4. Formation Pore Pressure………………………………….………….6
I. 2. 2. Independent Variables.......................................................................................6
I. 2. 2. 1. Weight On Bit....................................................................................6
I. 2. 2. 2. Revolution Per Minute.......................................................................7
I. 2. 2. 3. Drilling Fluids....................................................................................7
I. 2. 2. 4. Hydraulic Horse Power.....................................................................7
I. 2. 2. 5. Bit Type............................................................................................ 8
I.2 .2 .5 .1 .Roller-Cone Bits.................................................................8
I.2 .2 .5 .1 .Diamond Bits......................................................................9
I. 3.Conclusion………………………………………………………………………………..11
Chapter II ROP Modeling And Optimisation Techniques
II. 1. Introduction…………………………………………………………………………….13
II. 2. Drilling Optimization State Of The Art...........................................................................13
II. 3.Rate of Penetration Models ………………………………………………………….....14
II. 3. 1. Warren models……………………………………………………………….14
II. 3. 2. Bourgoyne and Youngs’ Model……………………………………………..15
II. 3. 2. 1. Effect of Formation Strength……………………………………..16
II. 3. 2. 2.Effect of Compaction……………………………………………...16
 II. 3. 2. 3. Effect of Differential Pressure……………………………………17
II. 3. 2. 4.Effect Of Bit Diameter And Bit Weight ,W/d……………………18
II. 3. 2. 5. Effect Of Rotary Speed,N………………………………………..19
II. 3. 2. 6. Effect of Tooth Wear, h…………………………………………...19
II. 3. 2. 7. Effect of Bit Hydraulic……………………………………………20
II. 4. Tooth wear modele..........................................................................................................21
II. 5. Bearing wear modele.......................................................................................................22
II. 6. Optimization Techniques.................................................................................................22
II. 6. 1. Multiple regression method.............................................................................23
II. 6. 1. 1. Multiple regression procedure..........................................................24
II. 6. 2. Metaheuristic optimization technique..............................................................25
II. 6. 2. 1. Particle swarm optimization.............................................................26
II. 6. 2. 2. Modified Particle Swarm Optimization...........................................28
II. 7. Drilling cost......................................................................................................................30
II. 8. Conlusion..........................................................................................................................32
Chapter III Result Analysis and Discussion
III. 1. Introduction…………………………………………………………………………....34
III. 2. Experimental Data……………………………………………………………………..34
III. 3. Result Analysis and Discussion………………………………………………………..35
III. 3. 1. Mutliple Regression…………………………………………………………35
III. 3. 2. Particle Swarm Optimization .........................................................................35
III. 3. 3. Modified Particle Swarm Optimization……………………………………..36
III. 4. Comparison between The Several Optimization Techniques …………………………38
III. 4. 1. A comparison of PSO and the M_PSO ……………………………………..38
III. 4. 2. A final comparison of the optimization techniques ………………………..39
III.5 Optimum bit weight and rotary speed..............................................................................40
III. 6. Conclusion .....................................................................................................................41
General conclusion....................................................................................................................42
REFERENCES.........................................................................................................................43
SUMMARY
LIST OF TABLES

Table. II. 1. Recommended tooth-wear parameters for roller cone bits...................................21

Table. II. 2. Recommended minimum data ranges for regression analysis..............................24

Table. III. 1 .field data taken in shale, offshore louisiana area................................................34

Table III.2. The results obtained from multiple regression method………………................35
Table III.3. The different PSO test results on three different population size (n)……………35
Table III.4 . Test results using modified particle swarm optimization………………………37
Table III.5 . Comparison between the best results of PSO with M_PSO……………………38
Table III.6 . References values we have been used for the relative error tests.........................39
Table III.7 . Relative error test results for PSO and M_PSO…………………………………39
Table III.8 . comparison of the several optimization technique results ……………………...39
Table III.9 . Required Data after bourgoyne study..................................................................40
LIST OF FIGURES
Figure I. 1. Drilling Variables associated With Rotary Drilling.................................................4
Figure I. 2. Roller-cone bit (inserts)............................................................................................9
Figure I. 3. PDC bit profiles......................................................................................................10

Figure. II. 1. Effect of normal compaction on penetration rate.................................................17

Figure. II. 2. - Effect Of Differential Bottom-Hole Pressure On Penetration Rate..................18
Figure. II. 3. Effect of bit weight on penetration rate..............................................................18
Figure. II. 4. Effect of rotary speed on penetration rate..........................................................19
Figure. II. 5. Effect Of Tooth Wear On Penetration Rate (Chipping-Type Tooth Wear).........20
Figure. II. 6. ROP as a function of bit Reynolds number.........................................................21
Figure. II. 7. Some of Modern or Nontraditional Optimization Techniques............................26
Figure. II. 8. Simple example of how particles move..............................................................28
Figure. II. 9. The flowchart of the standard PSO algorithm.....................................................29

Figure III. 1. Comparison of the three PSO results convergence…………………………….36

Figure III. 2. Comparison of the three M_PSO results convergence………………………....37
figure III. 3. Comparison of PSO and M_PSO results convergence for n = 100…………….38
NOMENCLATURE

a1 formation strength parameter -

a2 exponent of the normal compaction trend -
a3 under-compaction exponent -
a4 pressure differential exponent -
a5 bit weight exponent -
a6 rotary speed exponent -
a7 tooth wear exponent -
a8 hydraulic exponent -
c, c, and cc lithology coefficients -
B Bearing model -
Cb bit cost Dollars
Cf cost per drilled interval dollars/ft
Cr daily rig rate dollars/hour
D depth of borehole ft (m)
b bit diameter In
rate of penetration ft. /hr.
dn equivalent bit nozzle diameter In

dp exponent in dimensionless Pi term -

f1 formation strength function -
f2 formation normal compaction function -
f3 formation compaction function -
f4 pressure differential of hole bottom function -
f5 bit diameter and weight function -
f6 rotary speed function -
f7 tooth wear function -
f8 hydraulic function -
F distance drilled by bit [L], ft (m)
gp pore pressure gradient of the formation [M/L3], ppg
(sg)
h bit tooth dullness, fractional tooth height worn away -
H1, H2 , H3 constants for tooth geometry of bit types -
 f final bit tooth dullness -
i summation index for ith data point -
m modified jet impact force hp(N)
J summation index for frh drilling parameter -
J1 composite drilling parameter representing all but tooth -
wear
J2 tooth wear composite function used to calculate -
fractional tooth wear
N rotary speed rpm
n data point numbers used in regression analysis -
opt optimum rotary speed, rpm
Pe effective confining pressure psi
Pg represents the global previous best from the entire -
swarm
Pi represents the previous best of the current particle -
q volumetric flow rate [L3/T], gpm
(l/m)
R rate of penetration ft. /hr
r residual error in the drilling ROP equation -
r1 and r2 are random values -
S rock strength -
t time (usually bit rotating time) [T], hours
tb bit drilling time [T], hours
tc drill pipe connection time [T], hours
tt round trip time [T], hours
Vi Velocity of each Particle -
W weight on bit 1000 lbf (N)
w/db weight on bit per inch of bit diameter 1000
lbf/in(N/m)
(W/db)m bit weight per diameter where teeth fails 1000
instantaneously lbf/in(N/m)

opt
optimum bit weight per inch lb/in
b

(w/db)t threshold bit weight at which the bit starts to drill 1000 lbf/in
 (N/m)

X drilling rate of penetration independent parameter -

x2 normal compaction drilling parameter -
x3 under-compaction drilling parameter -
x4 pressure differential drilling parameter -
x5 bit weight drilling parameter -
x6 rotary speed drilling parameter -
x7 tooth wear drilling parameter -
x8 bit hydraulics drilling parameter -
xi Position of the Particle -

ϕ1, ϕ2 Cognitive and Social Components -

ω Weighted Inertia Component -
ρ density of the fluid Ppg
μ drilling fluid viscosity Cp
cone offset coefficient -
σ rock compressive strength Psi
ε rock ductility -
formation abrasiveness constant or life of teeth at Hours
standard
τB bearing constant Hours
ABBREVIATIONS

DMF drilling mud flow

ECD Equivalent Circulating Density
IADC international Association of Drilling contracter
IS International System
MR Multiple Regression
M_PSO modified version of Particle Swarm Optimiazation
NPT non productive time
O&G oil and gas
PSO Particle Swarm Optimiazation
PDC Polycrystalline Diamond Compact
PSO Particle Swarm Optimiazation
RPM rotation per minute
ROP rate of penetration
TDS Top Drive System
 General introduction

General introduction
In today's drilling industry, all considerations are involved to reduce drilling operation
expenditure even when the returns on capital in the oil and gas (O&G) sector was 100$
USD/bbl. In several cases, Drilling parameters play a large role in helping drillers achieve a
good rate of penetration (ROP), superior drilling performance and long bit life. They are basic
recommendations that help the driller to avoid damaging bits and other drilling equipments,
Also this means a reduction in non productive time (NPT) and a minimum drilling cost. In
consequence a drilling parameters optimization is a key point to make a drilling operation
economically satisfied.For these problem, The objective for our study is to focus on the
Optimization of the Drilling Parameters, To achieve this goal we will begin our study by a
dominant and widely utilized method for drilling rate prediction that called Bourgoyne and
Young’s Model. It demonstrates a relation between (ROP) and the parameters affecting on it,
There are eight (8) variables influencing the drilling rate and they depend on ground
formation type and must be determined based on the data gathered in advance. Bourgoyne
and Young have suggested the multiple regression (MR) analysis method in order to define
these constants.Then our study aims to propose one of the best metaheuristic optimization
techniques to improve and compare the quality of solution founded by the multiple regression
method . in other words, we used a particle swarm optimization (PSO) algorithm. it’s a
biologically inspired computational search and optimization method developed in 1995 by
Eberhart and Kennedy based on the social behaviors of birds flocking or fish schooling.
Compared with other optimization algorithms, the PSO is more objective and easily to
perform well, That’s why we enhance our study by a some modification in the Particle swam
optimization algothim (M_PSO) .
At last, This technique of optimization can be implemented by any programming language
and we have chosen MATLAB to solve the optimization model of drilling parameters which
is based on the rate of penetration. The simulation results will prove the efficiency of the
metaheuristic technique we have used (PSO) specially with the modified one (M_PSO). So
more faster drilling rate would result, and the objective of a least possible cost and in the
shortest time in compliance with safe operation will achieve in the drilling operation .
In the first chapter we talk about Drilling Parameters:Definition, Classification and Effect On
Drilling Performance.
In the next chapter we discussed about Rate Of Penetration Modelling And Optimization
Techniques
In the last chapter we make Result Analysis & Discussion
1
Chapter I: Drilling Parameters

Chapter I
Drilling Parameters

2
 Chapter I: Drilling Parameters

I. 1. Introduction
Oil and gas companies have played a major role in the energy sector, and constantly try to
develop technology to maximize their overall revenue. However, as the wells continue to get
drilled farther, the drilling wells cost continue to rise. Many researchers have worked for
optimizing constant operational parameters. However, these parameters lead to wasted time
and money for the operators if they are not well estimated. This is because they constantly
change throughout the drilling process. Therefore, it is required to know widely the behavior
and the influence of the several parameters, usually known as drilling variables, on the system
drilling quality, among these variables we notice: weight on bit (WOB), rotation of the bit
(rotation per minute (RPM)) and drilling mud flow (DMF). Especially, determining the
optimal rate of penetration (ROP) has been always one of the main concepts of drilling
engineering.
In this chapter, we will introduce, classify and define the drilling variables, and we will
make a study about bit types and their characteristics, also the different formation we
frequently used in.
I. 2. Drilling Principles
The wide variations in drilling conditions encountered under field conditions make it
difficult to develop general rules of operation for maximum drilling efficiency. Field
experience usually provides the basis for operations in a particular area, but testing often is
too costly and experience too late. Consequently, a method for determining optimum drilling
techniques and parameters for any particular drilling condition, with a minimum of
engineering effort and drilling experience is greatly needed [1]. The drilling parameters, or
variables, associated with rotary drilling have been analyzed and divided in two groups as
independent and dependent parameters as shown in Figure I.1. The independent variables are
those which can be directly controlled by the drilling rig operator and dependent variables are
those which represent the response of the drilling system to the drilling operation. There are,
of course, many factors other than those discussed here that effect drilling efficiency and
footage cost. These include such factors as formation hardness, abrasiveness of formation and
well depth. As these items cannot be conveniently controlled, their influence on costs must
simply be accepted [2].

3
 Chapter I: Drilling Parameters

Drilling variables

Weight on bit
Independants varaibles

Dependants varaibles
Penetration rate

Rotational speed
torque

Drilling fluids

Flush medium
pressure
Hydraulic horsepower

Formation pore
pressure
Bit type
Deeb

Drilling performance

Figure I.1. Drilling variables associated with rotary drilling.

I. 2. 1. Dependent Variables
The dependent variables associated with rotary drilling represent the response of the
drilling system to the imposed conditions and are the penetration rate of the bit, the torque and
the flush medium pressure and formation pore pressure [2].
I. 2 .1. 1. Penetration Rate
The rate of penetration (ROP), also known as drill rate, is the speed at which a drill
bit breaks the rock under it to deepen the borehole. It is normally measured in feet per minute
or meters per hour, but sometimes it is expressed in minutes per foot. This parameter is the
most important parameter, since all the calculations in this study are based on estimations of
ROP in the drilling industry [3].

4
 Chapter I: Drilling Parameters

The factors which effect on rate of penetration are listed under two general classifications
such as controllable and environmental. Controllable factors are the factors which can be
instantly changed such as weight on bit, bit rotary speed, hydraulics. Environmental factors on
the other hand are not controllable such as formation properties and drilling fluids
requirements. The reason that drilling fluid is considered to be an environmental factor is due
to the fact that a certain amount of density is required in order to obtain certain objectives
such as having enough overpressure to avoid flow of formation fluids. Another important
factor is the effect of the overall hydraulics to the whole drilling operation which is under the
effect of many factors such as lithology, type of the bit, downhole pressure and temperature
conditions, drilling parameters and mainly the rheological properties of the drilling fluid. It
has been observed that the drilling rate of penetration generally increases with decreased
Equivalent Circulating Density (ECD). Another important term controlling the rate of
penetration is the cuttings transport. It was concluded that average annular fluid velocity is the
dominating parameter on cuttings transport, the more the flow rate is high the less cuttings
bed is developed [3].
I. 2 .1. 2. Torque
Torque is a rotational force and it can be described as the ability to overcome resistance to
rotation. Its magnitude is measured by multiplying the perpendicular component of the force
applied by the distance between the axis of rotation and the point where the force is applied.
In drilling applications this distance would of course be the drill pipe radius. It is measured by
means of Top Drive System (TDS) systems. Previously the readings for this parameter were
relative. This parameter is going to be significantly important for inclined and highly deviated
wells, which is also related with the wellbore cleaning issues [2].
I. 2 .1. 3. Flush Medium Pressure
Drilling fluids in the wellbore can be in either a static or dynamic state. The static system
occurs when the fluid stands idle in the well. The dynamic state occurs when the fluid is in
motion, resulting from pumping or pipe movement. The static pressure of a column of fluid
pressure is known as "hydrostatic pressure" which is an essential feature in maintaining
control of well and preventing kicks or blowouts. The hydrostatic pressure of a fluid column
is a function of the mud weight or density and the true vertical well depth.
The ROP obtained while a well is drilled generally shows a steady decline as well depth
increases. The causes of the reduction in ROP with depth can be divided into two categories
[2] :

5
 Chapter I: Drilling Parameters

1) Processes that affects the unbroken rock;

2) Processes that act on the rock once it is broken into chips.
I. 2 .1. 4. Formation Pore Pressure
Formation pore pressure can be major factor affecting drilling operations especially in deep
wells. An operator planning a well needs some knowledge of overburden and formation fluid
pressure in order to select the necessary hydrostatic or drilling fluid pressure. If this pressure
is not properly evaluated, it can cause drilling problems such as lost circulation, blowouts or
kicks, stuck pipes, hole instability and excessive costs.
The Formation fluid or pore pressures are usually categorized as normal, subnormal and
abnormal or over pressured. When formation pore pressure is approximately equal to
hydrostatic pressure of drilling fluid for a given vertical depth, formation pressure is described
to be normal. When the formation is opened to the atmosphere during drilling, a column of
drilling fluid from the ground surface down to the formation depth (hydrostatic pressure)
would balance the formation pressure. If the formation pressure is less than that of the
hydrostatic pressure, then it is called subnormal formation pressure. Formations with pressure
higher than hydrostatic are encountered at various depth in many areas. These formations are
referred to as being abnormally pressured or over pressured. Generally, abnormal pore
pressures are associated with fluids trapped within the pore spaces of rocks by low
permeability barriers such as salt domes, folds or faults. Numerous authors have demonstrated
the severe reduction in ROP with different rotary bits as the borehole pressure increases [2].
I. 2. 2. Independent Variables
The independent variables are the drilling fluids, weight on bit, the bit rotational speed, bit
type and the hydraulics horse power.
I. 2. 2. 1. Weight On Bit
It represents the amount of weight applied onto the bit, that is then transferred to the
formation which in turn is the energy created together with string speed that advances
drillstring.
This amount of downward force exerted on the drill bit provided by thick-walled tubular
pieces in the drilling assembly that are known as drill collars.
It is an essential part of drilling optimization to ensure that the well deepens as drilling
moves forward. Finding the right amount of WOB per application is crucial to drilling
operations. If the WOB is greater than the optimum value, the drill bit has a higher chance of
wear or damage and there is even a chance for the drill string to buckle. [2].

6
 Chapter I: Drilling Parameters

I. 2. 2. 2. Revolution Per Minute

The definition of RPM is a measure of frequency of rotations performed by an equipment
in one minute. It is a technical term which is associated with any equipment that conducts its
operations by performing rotations over a fixed axis. It is an International System (IS) unit of
rotations and is abbreviated by many other common terms such as rpm/RPM (or rotations per
minute) or rev/minute.
Some of the examples of equipment used in drilling sector that consists of revolutions per
minutes include: top drive, drilling mud motor, compressor reciprocating pumps and motors
downhole, motor internal combustion engine [2].
I. 2. 2. 3. Drilling Fluids
The bottom hole must be always cleaned, so we have to remove the cuttings from the
borehole. This one obtained by using drilling fluids with sufficient flow flushing medium that
can be air, water, oil, oil/water emulsion, mud or foam. Drilling rate is proved to be faster and
bit life longer with air as compared to water or mud. Drilling was originally performed with
air or water as a drilling medium used to cool the bit and flush away the drill cuttings. As
these two media were usually, easily available, cheap and satisfactory for the shallow
boreholes and hard formations being drilled at that time. Through the years many additional
requirements have been placed on the drilling fluid. To satisfy these demands, as boreholes
began to be drilled deeper, and especially with the rapid development of oil well drilling in
soft and often caving sedimentary formation, the composition has been modified greatly from
the air or water that was originally used. A drilling fluid called mud was developed,
consisting of water and bentonite clay to overcome problems such as borehole instability,
Mud has a number of properties such as its caking ability, its higher density, viscosity and its
thixotropic properties, which make it particularly suitable for drilling deep and soft
formations that would otherwise prove difficult to drill. However , The selection of the type
of drilling fluid is largely determined by the expected hole conditions. The adjustment of
drilling fluid properties is intimately related to the well depth, casing program and the drilling
equipment [2].
I. 2. 2. 4. Hydraulic Horse Power
Hydraulics has long been recognized as one of the most important considerations in the
design of drilling programs. Improved bottom hole cleaning afforded by jet rock bits and high
levels of bit hydraulic horsepower permit the use of the most effective combination of weight
and rotary speed and minimizes the risk of bit fouling. These benefits became apparent during

7
 Chapter I: Drilling Parameters

the early days of jet bit drilling as contractors began to search for ways to maximize the
effectiveness of their hydraulic systems. The results are extended bit life and faster
penetration rates. An increasing number of commercial bits are becoming available with
interchangeable nozzles, providing the flexibility of rig-site hydraulics optimization. With
these interchangeable nozzles, the hydraulic power of the drilling fluid that is dissipated
across the bit face can be adjusted to match that portion of the rig's hydraulic power that is
available for the bit after other system losses have been considered. The degree to which
drilling rate was affected by bit hydraulic horsepower depends on the rock/drilling-fluid
combination [2].
I. 2. 2. 5. Bit Type
The drill bit is the main tool of the drilling process, positioned at the end of the drill string.
Its rotation cuts and the weight on bit indents, resulting in penetration of the formation.
Drilling fluid circulates through the bit to decrease bit wear by cooling, and to help the
penetration rate by removing cuttings. The aim of every drilling engineer when selecting a
drilling bit is to achieve the highest rate of penetration with the least possible bit wear. and
because formation properties and bit type are the largest factors that affect penetration rate,
the correct bit type is a major importance in achieving high rates of penetration.
There is a great selection of bits available where rotary drilling has two main groups of
bits in which we find numerous varieties of bit designs. These are roller-cone bits and fixed-
cutter or diamond bits.
I. 2. 2. 5.1 Roller-Cone Bits
Roller-cone bits can be categorized by insert or milled tooth. Insert bits have a cutting
structure consisting of a sequence of inserts pressed into the cone. Milled tooth bits have a
cutting structure of teeth milled out of the cone. Tooth design and bearing types vary greatly
for roller-cone bits, making them applicable for several formation types. Milled tooth bits are
usually used in soft formations. Insert bits are appropriate for a wider variety of formations,
including hard formations.
Three cones and legs of similar size, connected to a pin, normally make up roller-cone bits.
The cones are mounted on each of their bearings, and able to rotate with respect to the bit
body. Connection to the drill string is provided by the pin section. Drilling fluid is pumped
down the drill string and through the nozzles of the bit. Openings by the legs provide fluid
circulation, and give the possibility to achieve high pressure jetting through the nozzles of the
bit. A representation of a typical roller-cone bit is provided below in Figure I.2.

8
 Chapter I: Drilling Parameters

Figure I.2. Roller-cone bit (inserts).[38]

Roller-cone bits are made of steel, which requires sufficient hardenability, yield strength,
heat treatment, machinability, and impact resistance. Design of the bit has generally four
focus areas: geometry and type of cutting structure, hydraulic requirements, material
selection, and mechanical operating requirements. The bit design is chosen based on how it
will operate and in what conditions it will operate in. Operating factors influencing the bit
design are primarily weight on bit, rotary speed and hydraulics. Operating conditions such as
formation, depth, drilling fluid, and hole deviation are also important parts considered when
designing a bit. The geometry and type of cutting structure is the significant design area of the
bit for providing an efficient penetration. Wear-resistance is also important during the
selection of geometry and type of cutting structure. Cutter shape and grade is normally
differentiated by its placement on the cone for insert teeth. There is a number of available
geometries, sizes and grades for cutters to be optimized depending on the cutters location and
conditions [4].
I. 2. 2. 5.2 Diamond Bits
Diamond bits can be regarded as fixed-cutter bits, as the bits have no separately moving
parts. Diamond is the hardest readily available material, thus using it as material provides
superior hardness. Both rotating as one piece and using diamond material gives a long bit life.
The diamond bits are mainly used in soft to moderate formation. In hard formations, the bit
has limitations regardless of recent developments [5]. Limitations such as low ROP and high

9
 Chapter I: Drilling Parameters

wear is also a result for deep continental gas developments [6]. Two categories of diamond
bits are currently on the marked: Polycrystalline Diamond Compact Bits and Natural
Diamond Bits. The Polycrystalline Diamond Compact (PDC) Bit is the most common
diamond bit, relatively equal in popularity as the roller-cone bit. PDC bits uses inexpensive,
fabricated diamonds. Their long bit life and capability of maintaining a high ROP has resulted
in wide popularity. Fixed-cutters induce a shearing action more effective than the crushing of
the inserts or teeth on the cones of the roller-cone bit [7, 10]. A PDC bit is designed based on
four considerations: materials, formation properties, hydraulic conditions, and mechanical
parameters. There are four different types of blade profiles for a PDC bit:
1. Flat profile – for hard and non-abrasive formations ;
2. Short parabolic – for hard and medium abrasive formations ;
3. Medium parabolic – for medium/hard and abrasive formation ;
4. Long parabolic – for soft and abrasive formations.

Figure I.3. PDC bit profiles.[39]

Figure I.3.shows various PDC bit profiles, broken into five zones: cone, nose, taper,
shoulder, gauge (from center). The profile or shape of the bit is dependent on cutter
placements, cutter geometry, cutter density, hydraulics, well geometry, and formation. All
elements need to be considered to design a bit capable of high ROP and low bit wear. The
shape will have a direct influence on steerability, stability, ROP, durability, fluid circulation,
and cutter density [4].
There are many proposed methods for bit selection and often more than one is used before
reaching a decision. Bit selection methods include [2] :

10
 Chapter I: Drilling Parameters

1) Cost analysis;
2) Offset well bit record analysis;
3) Offset well log analysis;
4) IADC bit coding;
5) Manufacturer's product guides;
6) Geophysical data analysis;
7) General geological considerations.
I.3. Conclusion
In this chapter, we have mentioned several parameters (variables) of drilling, many
references have classified them into two mean categories, dependent and independent.
Dependent variables such as ROP, Torque, flush medium pressure, formation pore pressure,
and independent variables : RPM, WOB, drilling fluids, hydraulic horse power and bit types.
These parameters are keys of successfully drilling operation because of their positive
influence on time, cost and security.
In the next chapter we will focus on ROP modeling and the description of the several
optimization techniques used in this work.

11
Chapter II: Rate Of Penetration Modelling And Optimization Techniques

Chapter II
Rate Of Penetration Modelling And Optimization
Techniques

12
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

II. 1. Introduction
In this next study we will make a state of the art for different models used in drilling
parameters optimization and discus about some models of Warren and specially the model of
Bourgoyne & Young in order to obtain optimal value of drilling parameters during drill
operation, this could be by one of these important optimization techniques like a Multiple
Regression (MR), Particle Swarm Optimization (PSO) and a modified version of Particle
Swarm Optimization (M_PSO). So we identify the principle process of each one.
II. 2. Drilling Optimization State Of The Art
In the beginning of the 1900. During conception period the rotary drilling principle
developed by the introduction of rotary bits, casing installation and cementing techniques, and
developments in drilling fluids. after 20 years During the development period, more powerful
rigs, better bits, improved cementing and drilling fluid treatment techniques were introduced
which took place following Spindletop.
In 1950s the scientific period took place with expansion in drilling research and most
important of all optimized drilling, better understanding of the hydraulic principles,
significant improvements in bit technology, improved drilling fluid technology. After 1970s
rigs with full automation systems, closed-loop computer systems, with ability to control the
drilling variables started to operate in oil and gas fields.
One of the first attempts for the drilling optimization purpose was presented in the study of
Graham and Muench in 1959 [11]. They derive empirical mathematical expressions for bit
life expectancy and for ROP as a function of depth, RPM, and WOB by evaluated the WOB
and RPM combinations.
Maurer in 1962 [12] derived ROP equation that based perfect cleaning condition where all
of debris is considered to be removed between tooth impact and that equation is for roller-
cone type of bits considering the rock cratering mechanisms.
Galle and Woods in 1963 [13] presented procedures for determining the best combination
of constant WOB and RPM ; the best constant weight for any given rotary speed and the best
constant rotary speed for any given weight. For each of these procedures ,by considering a
combination of bit teeth and bearings life, and drilling rate limits economical bit life ,they
presented eight cases. They established empirical equations on drilling rate for the effects of
weight on bit, rotary speed, and cutting structure dullness, rate of tooth wear and bearing life.
Eckel in 1968 [14] performed microbit studies expressing the drilling rate exponentially as
a function of pseudo bottom hole or near bit-nozzle Reynolds number. The relation introduced

13
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

was reported to be independent of bit weight and speed .and differential borehole pressure,
formation.
One of the most important drilling optimization studies performed was in 1974 by
Bourgoyne and Young [15]. In order to obtain the optimized drilling parameters They
proposed the use of a linear drilling ROP model and performed multiple regression. They
have used minimum cost formula, showing that maximum ROP may coincide with minimum
cost approach [3].
Warren in 1987 [4] when using roller cone bits that includes the effect of both the initial
chip formation and cuttings removal process he defined a new model to explain rate of
penetration. And he developed an initial basic model that will be refined by addition of a
more varied set of test conditions every time that new data are added
Miska in 1988 [16] presented three governing differential equation: rate of penetration, rate of
teeth wear, and rate of bearings wear.
Maidla and Ohara in 1991 [17] they compared a tested drilling model on offshore drilling data
with the Bourgoyne and Young’s model. Rommetveit and al. in 2004 [18] developed a new
innovative drilling automation and monitoring system. The project was named as drilltronics.
in order to optimize the drilling process all available surface and subsurface drilling data was
utilized, one of the introduced modules was “bit load optimization module” which modulated
rotary speed and WOB and observed the how respective changes effected the ROP [19].
II. 3. Rate Of Penetration Modeling
In order to optimize a system we must have a model. It has been found that drilling rate of
penetration could be modelled in real time environment as function of independent drilling
variables, the ability to the drilling ROP with respect to depth characteristically with certain
parameters for specific formation on real time basis could bring new insights to the nature of
drilling operation. Therefore, many researchers have developed models that try to capture the
physics of the drilling process for all types of bits. Below, we will describe the several models
developed by Warren, and then we will focus on the model developed by Bourgoyne and
Young. ROP model, bearing wear model and tooth wear model.
II. 3. 1. Warren models
Warren in 1981 [20], developed an ROP model to relate weight on bit (WOB), revolutions
per minute of the bit (RPM), bit diameter, rock strength, and bit type to rate of penetration.
b
(II.1)
b

14
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

The negative of this model is not taking into account hydraulic effects, and assumed perfect
cleaning
So Warren later by taking into account the hydraulics added to this model, a new imperfect
cleaning model incorporating a new term into the ROP equation [21]. This term is a function
of the diameter of the bit, density of the fluid, drilling fluid viscosity, and modified jet impact
force.
b b
(II.2)
b m

Another time this model was again further developed to take into account roller cone offset
and formation ductility which added an additional term to the ROP model
[22], consisting of the cone offset coefficient, rock compressive strength, and rock ductility.
b b ϕσ b
(II.3)
b m ε

A few years later, Warren’s model was modified by adding another term for the chip hold
down effect [23]. In the end ,the model now takes into account the position that the fluid is
with respect to the mud overbalance [24]. This new term is a function of effective confining
pressure, and lithology coefficients [25].

c b b
c c Pe-120 (II.4)
b m

II. 3. 2. Bourgoyne and Young’s Model

The model proposed by Bourgoyne and Young has been adopted for this study in order to
derive equations to perform the ROP estimation using the available input data. This model is
considered as one of the complete mathematical drilling models in use of the industry for
roller-cone type of bits that’s why we have chosen it in our study.
The drilling model selected for predicting the rate of penetration, ROP, by considering the
effect of the various drilling parameters is described as: ROP = (f1)(f2)(f3)….(fn).
Where f1, f2, ..., fn represents the functional relations between penetration rate and various
drilling variables. Each of these functions contains constants which are shown as a1 through
an. Determination of these constants is accomplished by using a multiple regression analysis
and genetic algorithm of collected drilling data [26].
f1 function is defined as the formation strength and it should have the same unit as rate of
penetration, which is also known as drillability of the formation.x1 is the dummy variable
which is equal to 1 for every observation of rate of penetration. The effect of formation
compaction on rate of penetration is represented with two functions. The primary effect is

15
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

normal compaction, f2, it is given with an exponentially decreasing response with increasing
depth. In other means this function assumes increasing rock strength with depth due to the
normal compaction. The secondary effect of normal compaction is represented by f3. This
function considers the effect of under-compaction in abnormally pressured formation. Within
over-pressured formations rate of penetration is going to end up with increased magnitudes.
There is an exponential increase in penetration with increased pore pressure gradients f4
represents the function for pressure differential of bottom hole. The less the pressure
differential at the hole bottom, the less the penetration rate to be observed. True vertical depth
corresponding magnitudes have been used in calculation of the latter three functions. f5
represents the function for bit diameter and weight applied onto the bit. Rate of penetration
general equation is directly linked with the weight applied over the hole diameter. This
function is normalized for 4000lb per bit diameter. (w/d) t is the threshold bit weight, which is
known as the force at which rock fracturing begins. f6 represents the function for rotary speed.
Likewise the direct defined relation of bit weight on penetration rate, the rotary speed is also
set to have a direct effect [19].
Note : fn=EXP(anxn)
The drilling model selected for predicting the effect of the various drilling parameters, x j
on penetration rate, dD/dt, is given by :
1 =2
a ) (II.5)
Where Exp (z) is used to indicate the exponential function .
The modeling of drilling behavior in a given formation type is accomplished by selecting the
constants al through a8 in Eq. II.5, since it is linear.
II. 3. 2. 1. Effect of Formation Strength
The constant a1 primarily represents the effect of formation strength on penetration rate.
It is inversely proportional to the natural logarithm of the square of the drillability strength
parameter discussed by Maurer [12]. It also includes the effect on penetration rate of drilling
parameters that have not yet been mathematically modeled; for example, the effect of drilled
solids.
II. 3. 2. 2. Effect of Compaction
The terms a2x2 and a3x3 model the effect of compaction on penetration rate. x2 is defined
by :
x2 = 10000.0 - D· (II.6)

16
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

And thus assumes an exponential decrease in penetration rate with depth in a normally
compacted formation. The exponential nature of the normal compaction trend is indicated by
the published micro bit and field data of Murray [27], and also by the field data of Combs
[28] (see Figure. II. 1).

Figure. II. 1. Effect of normal compaction on penetration rate.[15]

Where, x3 is defined by :
0.69
3= gp-9.0 (II.7)
And thus assumes an exponential increase in penetration rate with pore pressure gradient.
The exponential nature of the effect of under compaction on penetration rate is suggested by
compaction theory, but has not yet been verified experimentally. Note that the effect of
a2 2 a3 3
compaction on penetration rate, , has been normalized to equal 1.0 for a normally
compacted formation at 10000ft.
II. 3. 2. 3. Effect of Differential Pressure
The term a4x4 models the effect of pressure differential across the hole bottom on
penetration rate. x4 is defined by :
4= p ρc ) (II.8)
And thus assumes an exponential decrease in penetration rate with excess bottom-hole
pressure. Field data presented by Vidrine and Benit [29] and by Combs [28], and laboratory
data presented by Cunningham and Eenink [30] and by Garnier and van Lingen [31] all
indicate an exponential relation between penetration rate and excess bottom-hole pressure up

17
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

to about1000psi (see Figure. II. 2.). Vidrine and Benit also noted an apparent relation between
the effect of differential pressure on penetration rate and bit weight. However, no consistent
correlation could be obtained from the available data, so no bit weight term was included in
Eq. II.8.

Figure. II. 2. Effect of differential bottom-hole pressure on ROP.[15]

II. 3. 2. 4.Effect Of Bit Diameter And Bit Weight ,W/d
The term a5x5 models the effect of bit weight and bit diameter on penetration rate. x5 is
defined by
t
5 ) (II.9)
t

Figure. II. 3. Effect of bit weight on ROP.[40]

18
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

a5
and thus assumes that penetration rate is directly proportional to .Note that the term
ea5x5 is normalized to equal 1.0 for 4000 lb per inch of bit diameter. The threshold bit weight,
(W/d)t, must be estimated with drill-off tests.
II. 3. 2. 5. Effect Of Rotary Speed
The term a6x6 represents the effect of rotary speed on penetration rate. x6 is defined by :

6 (II.10)
a6
and thus assumes that penetration rate is directly proportional to .Note that the term e a6x6
is normalized to equal 1.0 for 100 rpm. Reported values of the rotary speed exponent range
from 0.4 for very hard formations to 0.9 for very soft formations.

Figure. II. 4. Effect of rotary speed on ROP.[40]

II. 3. 2. 6. Effect of Tooth Wear, h

The term a7x7 models the effect of tooth wear on penetration rate. x7 is defined by :
7 (II.11)
Where h is the fractional tooth height that has been worn away. Previous authors have used
more complex expressions to model tooth wear .However those expressions were not ideally
suited for the multiple regression analysis procedure used to evaluate the constant a7 from
field data. Figure. II. 5.shows a typical comparison of the previously published relations and

19
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

a
7 7
. The value of a7 depends primarily on the bit type and, to a lesser extent, the formation
type. When carbide insert bits are used, penetration rate does not vary significantly with tooth
wear.
Thus the tooth wear exponent, a7, is assumed to be zero, and the remaining exponents, a1
a
7 7
through a6 and a8, are regressed. Note that is 1 when either h or a7 is zero.

Figure. II. 5. effect of tooth wear on ROP (chipping-type tooth wear).[15]

II. 3. 2. 7. Effect of Bit Hydraulic
The term a8x8 models the effect of bit hydraulics on penetration rate. Where, x8 is defined
by :

8 (II.12)
n

And is based on micro bit experiments performed by Eckel [14]. Which found that

penetration rate was proportional to a Reynolds number group . Since µ the apparent
n

viscosity at 10000 sec-1, is not routinely measured and recorded it must be estimated using the
relation [12].
p τy
(II.13)

20
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

Figure. II. 6. ROP as a function of bit Reynolds number.[40]

II. 4. Tooth Wear Model
Instantaneous tooth wear could be calculated by means of finding the abrasiveness constant
for a known bit record in the subject formation. Formation abrasiveness constant is a
parameter when reached the bit in use will become inefficient to drill ahead. The
instantaneous tooth wear equation is given in terms of the relation in Eq. II.14. It has been
defined by the combination of tooth geometry, bit weight and rotary speed.
2
m-4
b
(II.14)
m 2
b b

Where, is formation abrasiveness constant, hours, hf fractional tooth wear, H1 and H2 are
tooth geometry constants. The recommended tooth-wear parameter constants for roller cone
cutter bits are as given in Table. II. 1. These parameters should be based on general field
experience and drilling practices observed in field applications. A particular study could be
conducted to update these parameters [12].
Table. II. 1. Recommended tooth-wear parameters for roller cone bits.
Bit Class H1 H2 H3 (W/D)max
1 – 1 to 1 – 2 1.90 7.0 1.00 7.0
1 – 3 to 1 – 4 1.84 6.0 0.80 8.0
2 – 1 to 2 – 2 1.80 5.0 0.60 8.5
2–3 1.76 4.0 0.48 9.0
3–1 1.70 3.0 0.36 10.0
3–2 1.65 2.0 0.26 10.0
3–3 1.60 2.0 0.20 10.0
4–1 1.50 2.0 0.18 10.0

21
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

Note that the tooth wear formula given above is going to be normalized at 60 rpm of bit
rotation speed and a constant bit weight of 4,000 lbf/in. The normalization magnitudes are
selected accordingly for the specific conditions in the scope of this study.
In order to be able to calculate the formation abrasiveness constant a tooth wear parameter
is required to be introduced, which is basically the reciprocal of the some of the given terms
in the composite tooth wear Eq. II.10. The tooth wear parameter is symbolized as J2, Eq.II.15.
m
2= (II.15)
b b
2
m
b

If both sides of Eq. II.14. is written in a terms of J2, the following relation is achieved when
integrated with Eq. II.15:
b f
2 2 (II.16)
When equation II.16 is integrated, the following relation yields, Eq. II.17.
f
b= 2 f 2 (II.17)

The formation abrasiveness constant could then be written as given in Eq. II.18.
b
(II.18)
f
2 f 2

Once formation abrasiveness constant is known, a tb, time of bit rotation as a function of
predefined constants and tooth wear as a fraction, could be calculated, solving Eq. II.17.
An arbitrary hf value could first be selected and until a tooth wear fraction is iterated, the
selection of hf should be determined, provided that the bit rotating time that is back calculated
equal to the actual bit rotation time that is available in the database [30].
II. 5. Bearing wear model
Bearing wear was estimated by using the following equation :

(II.19)
B

Where the constant b depends upon bearing type and mud type and the bearing constant B is
calculated from a dull bit grading.
II. 6. Optimization Techniques
The word optimum, meaning “best”, is synonymous with “most” or “ma imum” in one
case and with “least” or “minimum” in another. The term, optimize, means to achieve the
optimum, and optimization refers to the act of optimizing. Thus, optimization theory
encompasses the quantitative study of optimal and methods for finding them. There is no such

22
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

thing as a “true” optimum drilling program, invariably compromises must be made because of
limitations beyond our control that result in something less than optimum.
In general terms, an optimization problem consists in selecting from among a set of
feasible alternatives, one which is optimal according to a given criterion. The optimization
term in this thesis are considered as the drilling procedure, which the best constant weight and
rotary speed together with another controllable drilling parameters yield the penetration rate
with the minimum drilling cost.
II. 6. 1. Multiple Regression Method
Eq. II.6 through Eq. II.12 define the general functional relations between penetration rate
and the other drilling variables, but the constants a2 through a8 must be determined before
these equations can be applied. The constants a2 through a8 are determined through a multiple
regression analysis of detailed drilling data taken over short depth intervals. The idea of using
a regression analysis of past drilling data to evaluate constants in a drilling rate equation is not
new. For example, it was proposed by Graham and Muench in 1959 [11] in one of the first
papers on drilling optimization. This approach was used by Combs in his work on the
detection of pore pressure from drilling data. However, much of the past work in this area has
been hampered by the difficulty in obtaining large volumes of accurate field data and because
the effect of many of the drilling parameters discussed above were ignored. Recent
developments in on site well monitoring have made it possible to routinely regress the more
complex drilling equation Eq. II.5.
A derivation of the multiple regression-analysis procedure is presented in detail in the
section (II. 6. 1. 1).
Theoretically, only eight data points are required to solve for the eight unknowns a1
through a8. However, in practice this is true only if Eq. II.5 models the rotary drilling process
with 100-percent accuracy. Needless to say, it never happens. When only a few data points
are used in the analysis of field data, even negative values are sometimes calculated for one or
more of the regression constants. A sensitivity study of the multiple regression-analysis
procedure indicated that the number of data points required to give meaningful results
depends not only on the accuracy of Eq. II.5 but also on the range of values of the drilling
parameters x2 through x8. Table. II. 2. summarizes the recommended minimum ranges for
each of the drilling parameters and the recommended minimum number of data points to be
used in the analysis. When any of the drilling parameters, x j, have been held essentially
constant through the interval analyzed, a value for the corresponding regression constant, a j ,

23
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

should be estimated from past studies and the regression analysis should be carried out for the
remaining regression constants. As the number of drilling parameters included in the analysis
are decreased, the minimum number of data points required to calculate the remaining
regression constants is also decreased (see Table. II. 2). In many applications, data from more
than one well had to be combined in order to calculate all eight regression constants.
Table. II. 2. Recommended minimum data ranges for regression analysis

Parameter Minimum range Number of Parameters Minimum Number of Points

x2 2.000 8 30
x3 15.000 7 25
x4 15.000 6 20
x5 0.40 5 15
x6 0.50 4 10
x7 0.20 3 7

x8 0.50 2 4

For example, the first of the eight equations defined in the section (II. 6. 1. 1) is given by
Taking the logarithm of both sides of (Eq. II.5) yields [15].

na1 + a2 Σx2 + a3 Σx 3 + a4 Σx4+ a5 Σx5 + a 6 Σx6 + a7 Σx7 + a8 Σx8 = Σ In (II.20)

II. 6. 1. 1. Multiple Regression Procedure

The equation of the proposed model is :

1 (II.21)

Taking the logarithm of both sides of the above equation yields :

1 (II.22)

If the residual error of the ith data point, ri, is defined by :

i= 1 (II.23)

In order to minimize the square of the residuals i , the constants from a1 to a8 should
be determined properly by taking derivative from the square of the residuals i .
i i
i i (II.24)

For j = 1, 2, 3, ........,8. The constants a1 through a8 can be obtained by simultaneously solving

the system of equations obtained by expanding i for j = 1, 2, 3,............, 8. The

expansion of i yields:

24
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

1 2 2 3 3 8 8

1 2 2 2 3 2 3 8 2 8 2

1 3 2 3 2 3 3 8 3 8 3

. . . . .
. . . . .
. . . . .

1 8 2 8 2 3 8 3 8 8 8

After substituting the appropriate functions into Eq. II.6 and Eq. II.12 and by using
multiple regression-analysis, in order to calculate the constants a1 through a8, the following

linear equation system can be obtained by matrix :

i2 i3 i8
2 a1
i2 i2 i2 i3 i2 i8 a2 i2
2 a3
i3 i3 i2 i3 i3 i8 i3

2 a8
i8 i8 i2 i8 i3 i8
i8

II. 6. 2. Metaheuristic Optimization Technique

For every optimization problem there exists a set of possible solutions which is called the
solution space. A solution can be seen as an input for a known model. The model also called
objective function that calculates an output for a given input. The output is considered the
quality of a solution. A globally optimal solution of an optimization problem is found, if there
exists no other solution which evaluates to a better quality. The best quality can be either the
highest or the lowest, depending if it is a maximization or a minimization problem. The
difficulty of a problem depends, among other factors, on its complexity. Linear optimization
problems are problems where the objective function can be described as a linear function.
These problems are solvable in polynomial time. There are algorithms such as the Simplex
method developed by George Danzig which can solve linear problems efficiently. Non linear
or discrete optimization problems are much harder to solve efficiently, except for some
special cases. In the general case, no algorithm is known until today which can solve such
problems exactly in polynomial time with a deterministic turing machine. To search the
solution space of combinatorial problems for an optimal solution, a backtracking algorithm

25
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

might be used which just enumerates all possible solutions. The best way to tackle large
solution spaces of non linear optimization problems is to start with a constructed or randomly
created solution and iteratively improve it. To improve the solution, problem independent
heuristics also called metaheuristics can be applied. Such metaheuristics are usually based on
the quality of solutions, which allows the abstraction of the algorithm from the underlying
problem. Of course, the evaluations of a solution as well as manipulating a solution are still
problem-specific operations, but the algorithmic steps can be abstracted from the problem.
Metaheuristic optimization techniques typically consist of some stochastic steps and
consequently their results underlie a stochastic distribution. It is not guaranteed that a globally
optimal solution is found. shows the classification of optimization techniques in a wide
context, yet the following sections will describe only metaheuristic methods in more detail,
since they are most relevant for this thesis. Metaheuristic algorithms can be categorized into
trajectory based and population based metaheuristics [32].

Particle
Swarm
Optimization

Metaheuristic
Genetic Ant colony
Algerithms
Optimization Optimization
Technique

Neural
Network and
Fuzzy
Optimization

Figure. II. 7. Some of modern or nontraditional optimization techniques.

II. 6. 2. 1. Particle Swarm Optimization
Particle swarm optimization is a powerful and widely used optimization technique that
covers a wide range of research areas. PSO is one of the most popular nature inspired
metaheuristic optimization algorithm developed by Kennedy and Eberhart in 1995, and it was
modelled to mimic how certain groups of animals move in the natural world : such as a school
of fish, flock of birds, etc. For this algorithm, a group of animals is referred to as a "swarm"
and each animal inside the group is considered a "particle". This algorithm uses a combination
of information from the group as a whole and the information from each individual particle to

26
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

search the space for the optimal solution. For each individual particle, the PSO algorithm uses
the current “velocity” of each particle, along with the information from the best values found
from both the individual particle and the best global from the swarm, to move the particle
around the space.
PSO starts initially by randomly selecting values for all dimensions corresponding to each
particle inside a swarm. The swarm is evaluated and the new velocity of each particle and
position are updated. The velocity and position equations [33] are shown below, where :
- Vi represents the velocity of a particle;
- Pi represents the previous best of the current particle;
- Xi represents the current position of the particle;
- Pg represents the global previous best from the entire swarm;
Each one of these variables is a vector of d in length, representing the number of
dimensions in the problem. The other variables are :
- ϕ1 and ϕ2 which are considered acceleration constants, respectively, 1 and 2.
- ω which is a weighted inertia constant [0.4 to 1.4].
- r1 and r2, are random values that are taken from the uniform distribution [0, 1].

(II.25)

+ (II.26)
The velocity equation, Eq. II.25, above is comprised of three components, social, cognitive
and momentum [34]. The social component, ϕ2, forces the particles towards the global best
solution found; the cognitive component, ϕ1, forces the particles back towards the previous
best solution found by each particle; and the momentum component, ω, forces the particle to
continue on the current trajectory. Al three components help the particle swarm optimization
technique traverse the exploration/exploitation dilemma that surrounds all optimization
problems.
In Our study The PSO algorithm uses the ROP model by having the particles search
the solution space and converge on the optimal WOB, RPM, bit selection, and pull depth. The
inputs for this algorithm include: rock strength, WOB and RPM operational ranges, and
available bit selections [35].

27
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

Figure. II. 8. Simple example of how particles move.

II. 6. 2. 2. Modified Particle Swarm Optimization
In standard PSO, because the particle has the ability to know the best position of the group
particles have been searched, we need one particle to find the global best position rather than
all particles to find it, and other particles should search more domains to make sure the best
position is global best position not the local one.
The modification in PSO consists of three categories: extension of field searching space,
adjustment the parameters, and hybrid with another techniques. The procedure of modified
PSO is as following:
1) Initialize the position and velocity of each particle;
2) Calculate the fitness of each particle;
3) Concern the particle with the biggest fitness value, reinitialize its position; and evaluate the
particle with the smallest fitness value whether its new position is acceptable, if the answer is
yes, update its position, otherwise, a new position is assigned to the particle randomly in its
neighborhood with radius r; then renew the position and velocity of other particles according to
Eq II.25.and II.26;
4) For each particle, compare its current fitness value with the fitness of its pbest, if the current
value is better, then update pbest and its fitness value;
5) Determine the best particle of group with the best fitness value, if the current fitness value is
better than the fitness value of gbest, then update the gbest and its fitness value with the
position;
6) Check the finalizing criterion, if it has been satisfied, quit the iteration; and return to step 3.
[36].

28
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

Randomly initialize
the swarm particles

Calculate the fitness function

for each particle

YES Is the current fitness value NO

better than the individual best
position ?

Assign current fitness value Keep the previous individual

as the individual best position best position unchanged

Assign the best position among

all particles as the global best
position

Using individual and global best

positions, Update the position
and velocity of each swarm
particle

NO Is the stopping criterion YES End

reached ?

Figure. II. 9. Flowchart of the standard PSO algorithm.

29
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

II. 7. Drilling cost

Drilling cost per foot equation is as defined in Eq II.18. It has been defined to be a function
of daily rig rate, bit cost, and timing required in the course of the bit runs. This equation is
known to be the mostly applied drilling cost formula in the literature.
b r t c b
f= (II.27)

Where, Cf is the cost per drilled interval, Cr is the daily rig rate, Cb is the bit cost, tt round trip
time, tc connection time, hr, and tb bit drilling time. ΔF is the footage drilled with the bit in the
use, ft. Bit drilling time, tb, (with respective to tooth wear), Eq II.16.and drilling interval, ΔF,
(with respect to general drilling functions and tooth wear), the equation:
a 7h
= 1 2τ 2h

a2
2=
2
When inserted into equation : , after modifying the same, the cost per foot equation
could be redefined.
r b f
f= t c 2τ 2h (II.28)
1 2τ 2 r

Calculus states that any differentiable equation when differentiated to the first order it would
have been maximized. This statement could be written as below in order to optimize the
drilling cost with respect to WOB.
f)
f)= (II.29)
b

The second derivative of the Eq II.29.would define whether the response of the function is
a relative minimum or a maximum [37], such as :
f)>0, then f) is a relative minimum
f) 0, then f) is a relative maximum
When Eq II.28 is re-arranged, one gets the following :
b f
r t c 2
f= (II.30)
r

2 1 2 1

The detailed drilling cost equation could be written in a form including the equivalent
forms of composite drilling, J1, and tooth wear, J2 parameters respectively from the equations:

a5 b

a1 a2 2 a3 3 a4 4 a6 6 a8 8
2 b (II.31)
and Eq II.15 :
b f
r t c 2
f= (II.32)
r

30
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

Where ;

5
1 2 2 3 3 4 4 t 6 6 8 8
b (II.33)

t (II.34)
b b

2
(II.35)

m
b b
(II.36)
m
b

t
b b
5
1 2 2 3 3 4 4 t 6 6 8 8
b (II.37)

Differentiating Eq II. 32. to the first order and equalizing to zero with respect to WOB
independent parameter would result in drilling cost optimization for WOB parameter
f
f) r 5 5
f)= t (II.38)
b
b m
b b

Where;
2 (II.39)
b
t c (II.40)
r

1 2 2 3 3 4 4 6 6 8 8
2 (II.41)

t
b b
5
b (II.42)
1 2 2 3 3 4 4 6 6 8 8
(II.43)
Rearranging Eq II.38 with a simplified evaluation :

2
W
5
W
b
t
f
W
5
W
(II.44)
b
t m t
b b b b
b b
2 2

Using the distribute law of mathematics, equation above can be re-written,

b
t c t t
f f
5 2τ
b b b
5- (II.45)
r
2
b m
b b
2

Simplifying equation above results in having the relation below,

31
 Chapter II: Rate Of Penetration Modelling And Optimization Techniques

t
b f
5 2τ
b b
t c 5- 2 (II.46)
r m
b b

Eq II.30 should also be differentiated as a function of rotary speed, N.

f)
f)= (II.47)

The solution of the respective derivative for Eq II.47. is as give,

b 1 f
t c 2τ 2 (II.48)
r 6

The optimum equation for the weight for each diameter of bit size is as given
below, Eq II.49 .
5 1 ma 6 t
opt= (II.49)
b b

b 5 1 6

In a similar manner the optimum bit speed can be expressed in the following form Eq. II.50
after being obtained using the Eq II.21.
ma opt
opt=60 (II.50)
b b

b ma
b

II. 8. Conclusion
In this chapter we talked about ROP modelling and their different estimation during the
years, so we defined the importance of them, and we have detailed in the model of Bourgoyne
and Young which we is the selected one in our study. Some several techniques of
optimization (MR, PSO and MPSO) are explained with their different procedure and
characteristics have been proposed to optimize the different drilling parameters.
In the next chapter we will calculate the model constant and the optimal drilling parameters
with the aforementioned optimization techniques by using MATLAB programming language.

32
Chapter III: Results Analysis & Discussion

Chapter III

Results Analysis & Discussion

33
 Chapter III: Results Analysis & Discussion

III. 1. Introduction
In this final chapter we will describe our results after using the concept of the following
methods : Multiple Regression, Particle Swarm Optimization and Modified Particle Swarm
Optimization in order to obtain the best optimal solution for the objective function (ROP
value). In the metaheuristic techniques we will make a test for three different swarm
populations with a fixed number of iterations (300). Then we will compare the techniques
results by using different approximation errors (relative error, absolute error and the Root
Mean Square Error).
III. 2. Experimental Data
The field data are taken from an offshore Louisiana well that are shown in Table. III. 1
[15].
Table. III. 1. Field data taken in shale, offshore Louisiana area.

Driling Rotary Reynols Pore

Data Depth Bit Bit weight Tooth ECD
rate speed number Gradient
entry (ft) Number (1000lb/in Wear (lb/gal)
(ft/hr) (rpm) Function (lp/gal)
1 9515 7 23 2.58 113 0.77 0.964 9.5 9.0
2 9830 8 22 1.15 126 0.38 0.964 9.5 9.0
3 10130 9 14 0.81 129 0.74 0.827 9.6 9.0
4 10250 11 10 0.95 87 0.15 0.976 9.7 9.0
5 10390 12 16 1.02 78 0.24 0.984 9.7 9.0
6 10500 19 1.69 81 0.61 0.984 9.7 9.1
7 10575 13 1.56 81 0.73 0.984 9.7 9.2
8 10840 13 16.6 1.63 67 0.38 0.932 9.8 9.3
9 10960 15.9 1.83 65 0.57 0.878 9.8 9.4
10 11060 15.7 2.03 69 0.72 0.878 9.8 9.5
11 11475 15 14 1.69 77 0.20 0.887 10.3 9.5
12 11775 18 13.5 2.31 58 0.12 0.852 11.8 10.1
13 11940 21 6.2 2.26 67 0.2 0.976 15.3 12.4
14 12070 22 9.6 2.07 84 0.06 0.993 15.7 13.0
15 12315 15.5 3.11 69 0.40 1.185 16.3 14.4
16 12900 23 31.4 2.82 85 0.42 1.150 16.7 15.9
17 12975 24 42.7 3.48 77 0.17 1.221 16.7 16.1
18 13055 38.6 3.29 75 0.29 1.161 16.8 16.2
19 13250 43.4 2.82 76 0.43 1.161 16.8 16.2
20 13795 25 12.5 1.60 81 0.56 0.272 16.8 16.2

34
 Chapter III: Results Analysis & Discussion

21 14010 26 21.1 1.04 75 0.46 0.201 16.8 16.2

22 14455 28 19 1.76 64 0.16 0.748 16.9 16.2
23 14695 18.7 2.00 76 0.27 0.819 17.1 16.2
24 14905 29 20.2 2.35 75 0.33 0.419 17.2 16.4
25 15350 30 27.1 2.12 85 0.31 1.290 17.0 16.5
26 15740 14.8 2.35 78 0.81 0.802 17.3 16.5
27 16155 32 12.6 2.47 80 0.12 0.670 17.9 16.5
28 16325 14.9 3.76 81 0.50 0.532 17.5 16.6
29 17060 34 13.8 3.76 65 0.91 0.748 17.6 16.6
30 20265 40 9 3.41 60 0.01 0.512 17.7 16.6
Note that the primary drilling variables required for the MR, PSO and MPSO are
depth, penetration rate, bit weight per inch of bit diameter, rotary speed, fractional tooth wear,
Reynolds number parameter, mud density, and pore pressure gradient. To calculate the best
values of the model constants a1 through a8 using the data shown, the parameters x2 through
x8 must be calculated using Eq. II.6 through Eq. II.12 for each data entry.
III. 3. Result Analysis And Discussion
III. 3. 1. Multiple Regression
By using the equation of the proposed model of Bourgoyne and Young’s and the
procedures of MR that have been indicate in the previous chapter . the solutions are shown in
the table below.
Table III. 2. Results obtained from Multiple Regression method.
A1 A2 A3 A4 A5 A6 A7 A8
3.90557 1.96 2.0035 4.2839 0.40740 0.45315 0.48380 0.06024

III. 3. 2. Particle Swarm Optimization

Based on the general working principal of PSO and its standard settings where the personal
acceleration constant : = 2 and the weighted inertia constant Ω = 0.9, the test results
are shown in the table and figure bellow for three different swarm population (n).
Table III. 3. Different PSO test results on three different population size (n).
PSO A1 A2 A3 A4 A5 A6 A7 A8 It. n OF

n=30 3.51587 1.4083 1.7067 4.3373 0.31975 0.38019 0.43195 0.28276 35 0.06468

n=50 3.17740 1.3824 1.8056 4.2326 0.23963 0.17381 0.33643 0.42340 30 0.06324

n=100 3.24331 1.3582 1.7483 4.2081 0.25953 0.22309 0.34038 0.39374 24 0.06250

35
 Chapter III: Results Analysis & Discussion

Figure III.1 : Comparison of the three PSO results convergence.

Where, It.n is the iteration number abbreviation and OF is for the objective function. From
the figure III.1 and the table III.1 above, and after several tests we observed that the best value
of the ob ective function in PSO is obtained in the third test with n=100 . As well it’s the
fastest test because the best result of the objective function ( OF = 0.06250 ) is obtained after
just 24 iterations.
III. 3. 3. Modified Particle Swarm Optimization
A Several PSO standard settings are modified on the test of the objective function during
the course of our study and we evaluated their performances each time. the main modification
of MPSO that we used where in the constants personal acceleration coefficient( and )
into this equation :
(III.1)

With α = 1 , = 1.5 , = 2 and b = . Therefore = and = .

also we modified in the weighted inertia constant from Ω = 0.9 in PSO into Ω = 0.4 in
MPSO . The performance of the resulting algorithm in three population size is shown in the
next table and figure.

36
 Chapter III: Results Analysis & Discussion

Table III. 4. Test results using Modified Particle Swarm Optimization.

MPSO A1 A2 A3 A4 A5 A6 A7 A8 It. n OF
n=30 3.27143 1.3715 1.7455 4.2565 0.27357 0.22273 0.34158 0.39087 54 0.06259

n=50 3.34587 1.3787 1.7241 4.2744 0.28732 0.26191 0.36737 0.35411 45 0.06177

n=100 3.34602 1.3788 1.7241 4.2745 0.28735 0.26195 0.36742 0.35404 61 0.06176

Figure III.2 : Comparison of the three M_PSO results convergence

From this results, we found that Modified PSO achieves better values of the objective
function the more we increased the population size from n=30 through n=100. So we identify
that this parameter (n) always affect on the convergence rate and the optimal solution in this
optimization technique too, in addition to that we noted that the simulation time in MATLAB
increased (the simulation time took about 23.5 seconds when we used MPSO with n = 100).

37
 Chapter III: Results Analysis & Discussion

III. 4. Comparison Between The Several Optimization Techniques

III. 4. 1. Comparison Between PSO and MPSO
In the next tables and figure we will pick and compare the best previous results for PSO
and MPSO with each other.
Table III. 5. Comparison between the best results of PSO with MPSO.
For
A1 A2 A3 A4 A5 A6 A7 A8 It. n OF
n=100

PSO 3.24331 1.3582 1.7483 4.2081 0.25953 0.22309 0.34038 0.39374 24 0.06250

MPSO 3.34602 1.3788 1.7241 4.2745 0.28735 0.26195 0.36742 0.35404 61 0.06176

Figure III.3 : Comparison of PSO and MPSO results convergence for n = 100.
The table III.3 and figure III.3 above have been compared with all the previous results,
And after we checked The objective function results many times, the value OF=0.06176 of
the Modified PSO proved and confirmed its best performance specially for n=100. Therefore
this best value needs more iterations number (61). And for more accuracy and precision of the
comparison we will use two Relative Error ( tests for both of this optimization techniques,
where :

(III.2)

While, is the reference value, is the other calculated value, is a relative error.

38
 Chapter III: Results Analysis & Discussion

Table III. 6. References values we have been used for the relative error tests.
References
A1 A2 A3 A4 A5 A6 A7 A8
values

PSO 3.24331 1.3582 1.7483 4.2081 0.25953 0.22309 0.34038 0.39374

MPSO 3.34602 1.3788 1.7241 4.2745 0.28735 0.26195 0.36742 0.35404

Table III.7. Relative error test results for PSO and MPSO.

Relative error εa1 εa2 εa3 εa4 εa5 εa6 εa7 εa8
tests

5.69 % 0.0898 % 3.63 % 3.34 % 20.45% 36.65 % 15.23 % 16.86 %

Test1
PSO
6.91 % 3.97 % 6.18 % 4.45 % 20.56 % 59.20 % 24.68 % 22.25 %
Test2

0.0016 % 0.0016 % 0.0028 % 2.16e-04 % 0.0039 % 0.0091% 0.0019 % 0.0083 %

Test1
MPSO
0.0011 % 0.0031 % 0.0039 % 9.72e-04 % 0.0036 % 0.0070% 0.0013 % 0.0065 %
Test2

rom table III.5, it’s easy to find out that the percentage error of MPSO is so far smaller
than the one of PSO in all tests , and that’s an indication of how good the constants A1
through A8 are when we used MPSO. Otherwise , this means that the particles of the MPSO
have searched for more optimal solution than the PSO. That’s why the MPSO convergence is
better and takes more iterations number.
III. 4. 2. Final Comparison Of The Optimization Techniques
In this section , we compared the test results of PSO with MPSO and MR based on the
Root Mean Square Error (RMSE) and the Absolute Error, as presented in Table III.8.
Table III. 8. comparison of the several optimization technique results.
RMSE RMSE
Technique of optimization Absolute Error
Ln(ROP) ROP

n=30 0.16735 3.11840 0.04425

PSO n=50 0.16655 3.08693 0.04492
n=100 0.16631 3.11550 0.04416
n=30 0.16625 3.06860 0.04411
MPSO n=50 0.16619 3.09460 0.04384
n=100 0.16619 3.09457 0.04384

MR 0.21 3.5213 0.1471

39
 Chapter III: Results Analysis & Discussion

From the table III.8 it can be seen that both of the proposed optimization technique MPSO
and PSO give us best optimal solution than the Multiple Regression, they can achieve
excellent convergence on this optimization problems. In addition to that the simulation results
have shown that the MPSO is a better algorithm to solve complex optimization problems and
the best result is improved on a high number of population (n=100). Which indicates the
better searching performance and the more excellent convergence ability.
III.5 Optimum Bit Weight And Rotary Speed
Table III. 9. Required data according to Bourgoyne study.
Trip time, hours 6.0

Bit class 1-3

Bit weight, 1,000 lb/in. 4.1
Rotary speed, rpm 60
Tooth wear T-6
(W/d)t, 1,000 lb/in. 0.5
And From the best results of M_PSO that obtained in the previous study ,we took
1=3.34602, 2= 1.3788 , 3 = 1.7241 , 4 = 4.2745 , 5 = 0.28735, 6 = 0.26195,
7 = 0.36742, 8 = 0.35404.
Solution :
1. Calculation of formation abrasiveness constant by using Eq. II.10
2
m-4
3
b
b (III.3)
m 2
b b

From Table. II. 1, H1 = 1.84, H2 = 6, H3 = 0.8, (Wld)m = 8.0.

-4
= 15.75

2. Calculation the optimum bit weight by using Eq II.29 :

)
opt= = 5.18 lb/in
b

3. Calculation of the expected bit life by using Eq.II 13:

b= = 16.1h

4. Calculation the optimum rotary speed by using Eq. II.30

opt=60 = 41.25 rpm

40
 Chapter III: Results Analysis & Discussion

III. 6. Conclusion
This chapter focuses on the optimization of the drilling parameters results and the
capability of the modified particle swarm optimization to find the best optimal solutions when
its parameters are optimized specifically for this problem. We have shown in more than 20
tests that the more we increase the population size the more we obtain better values of the
objective function in both PSO and MPSO, but it takes more time of convergence.
We have shown that the MPSO has a better performance in terms of stability and speed of
convergence to find the global optimum.

41
 General conclusion

General conclusion
One of the most important perspectives from both technical and economic sides
for the petroleum industry during the drilling operations is the decreased ROP. So the deeper
we understand the processes, properties, and driving mechanisms of the ROP, also the
different classification of the drilling parameters and their effects the more we will capable of
increase The ROP. For this reason we have used a viable approach to understand and
calculate the above-mentioned details, It’s the Bourgoyne and Young ROP model.
In this study and after we have determined and calculated the constant values A1
through A8 of the proposed model by the multiple regression method, another metaheuristic
technique of optimization called particle swarm optimization (PSO) we have utilized to reach
more optimal values of the objective function and more reliable constants. The results we
have achieved were better than the multiple regression method, but we have faced a problem
that the PSO doesn’t always converge to the best values and optimal solution. That’s why we
proposed a Modified particle swarm optimization (M_PSO), The modification were generally
in the constants personal acceleration coefficient of the PSO.
Finally, through the results we have achieved in several tests and the different
comparison of the MR,PSO, and M_PSO using some of the approximation errors, we validate
that the best convergence, optimization performance and the optimal values are all obtained
by the M_PSO. This leads results lead us to calculate and Optimum bit weight and rotary
speed for the drilling operation.
In the future work, we will plan to implement our results on a real time drilling
operation in order to increase the ROP and the well productivity also to reduce the drilling
cost. In addition this study leads the readers to search and look for a better modification of the
PSO approach.

42
 References

References

[1] Shah, S. M. A., The optimization of PDC bit performance in coal measures rocks.
Unpublished Ph.D. thesis. University of Nottingham. 1992.
[2] Fasheloum Mohammed, Investigation of drilling parameters indicators. PhD thesis,
University of Nottingham, 1997
[3] Eren, T. Real-time-optimization of drilling parameters during drilling operations,
PhD Thesis, METU, Ankara. (2010).
[4] Morten A.H.ROP Modelling and Analysis,Master thesis,Stavanger,2015

[5] Brett , J.F., Warren , T.M. and Behr , S.M., Bit Whirl: A New Theory of PDC Bit
Failure; paper SPE 19571 presented at the SPE Annual Technical Conference and
Exhibition, San Antonio, TX, 8-11 October 1989.
[6] Hareland, G, et.al, “Cutting Efficiency of a Single P C Cutter on ard Rock” (2007)
[7] Wise J.L., Grossman, J.W., Wright, E.K., Gronewald, P.J., Bertagnolli, K. And
Cooley, C.H., Latest Results Of Parameter Studies On Pdc Drag Cutters For Hard
Rock Drilling; Grc Transactions, Vol. 29, 2005.
[8] Swenson , D.V., Wesenberg , D.L. and Jones , A.K., Analytical and Experimental
Investigations of Rock Cutting Using Polycrystalline Diamond Compact Drag
Cutters; paper 110. SPE 10150 presented at the SPE Annual Technical Conference
and Exhibition, San Antonio, TX, 4-7 October 1981.

[9] Warren , T.M. and Armagost , W.K., Laboratory Drilling Performance of PDC Bits;
SPE Drilling Engineering, Vol. 3, No. 2, pp. 125-135, June 1988.
[10] ] Zeuch , D.H. and Finger , J.T., Rock Breakage Mechanisms with a PDC Cutter;
paper SPE 14219 presented at the SPE Annual Technical Conference and Exhibition,
Las Vegas, NV, 22-26 September 1985.
[11] Graham .W. and Muench N. ., “Analytical etermination of Optimum Bit Weight
and Rotary Speed Combinations,” SPE 1349-G, Fall Meeting of the Society of
Petroleum Engineers, Dallas, October 1959.
[12] Maurer, W. C.: "The 'Perfect Cleaning' Theory of Rotary Drilling," ]. Pet. Tech.
(Nov. 1962) 12701274; Trans., AIME, Vol. 225.
[13] Galle E.M and Woods A.B., “Best Constant Weight and Rotary Speed for Rotary
Rock Bits,” rill. And Prod. Prac., API 1963, pp 48-73

43
 References

[14] Eckel, J. J.: "Microbit Studie s of the Effect of Fluid Properties and Hydraulics on
Drilling Rate," ]. Pet. Tech. (April 1967) 541-546; Trans., AIME, Vol. 240.
[15] Bourgoyne Jr., A. T. and Young Jr., F. S. (1974). A multiple regression approach to
optimal drilling and abnormal pressure detection. Society of Petroleum Engineers
Journal, August, 371-384. doi:10.2118/4238-PA.
[16] Stefan Miska., "Developments in Petroleum Engineering", Gulf Publishing
Company, (Collected works of Arthur Lubinski), 1988,Vol-2, pp 266-275
[17] Maidla E.E., and S. Ohara, “ ield Verification of rilling Models and Computerized
Selection of Drill Bit, WOB, and rillstring Rotation,” SPE rilling Engineering,
SPE Paper 19130, September 1991, pp 189-195.
[18] Rommetveit R., Bjorkevoll K.S., Halsey G.W., Larsen H.F., Merlo A., Nossaman
.N., Sweep M.N., Knut M.S., and Inge S., “ rilltronics: An Integrated System for
Real-Time Optimization of the rilling Process,” IA C/SPE 87124, IA C/SPE
Drilling Conference, Dallas, Texas, 2-4 March 2004.
[19] Eren, T. and Ozbayoglu, M. E.. Real time optimization of drilling parameters during
drilling operations, SPE Oil and Gas India Conference and Exhibition, Mumbai,
India, 20-22 January 2010. doi:10.2118/129126-MS.
[20] Warren, T. M. (1981). Drilling model for soft-formation bits. Journal of Petroleum
Technology, 33(06), 963-970.
[21] Warren, T. M. (1987). Penetration rate performance of roller cone bits. SPE Drilling
Engineering, 2(01), 9-18.
[22] Winters, W. J., Warren, T. M., & Onyia, E. C. (1987, January). Roller bit model with
rock ductility and cone offset. In SPE Annual Technical Conference and Exhibition.
Society of Petroleum Engineers.
[23] Hareland, G., & Hoberock, L. L Use of drilling parameters to predict in-situ stress
bounds. In SPE/IADC Drilling Conference. Society of Petroleum Engineers. . (1993,
January).
[24] Charlez, P. A.. The Concept of Mud Weight Window Applied to Complex Drilling.
Society of Petroleum Engineers. doi:10.2118/56758-MS(1999, January 1)
[25] Ryan Self, Use Of Particle Swarm Optimization Algorithm To Reduce Drilling Costs
By Finding Optimal Operational Parameters, Master Of Science, Oklahoma State
University, July, 2016.

[26] O. A. Falode and C. J. Agbarakwe,Optimisation of Drilling Parameters for

44
 References

Directional and Horizontal Wells Using Genetic Algorithm, Article, University of

Ibadan, Published 1 July 2016.
[27] Murray, A. S., and Cunningham, R. A.: "The Effect of Mud Column Pressure on
Drilling Rates," Trans., AIME (1955) Vol. 204, 196-204.
[28] Combs, G. D.: "Prediction of Pore Pressure From Penetration Rate," Proc., Second
Symposium on Abnormal Subsurface Pore Pressure, Baton Rouge, La. (J an. 1970).
[29] Vidrine, D. J., and Denit, E. J.: "Field Verification of the Effect of Differential
Pressure on Drilling Rate," ]. Pet. Tech. (July 1968) 676-682.
[30] Cunningham, A. J., and Eenink, J. G.: "Laboratory Study of Effect of Overburden,
Formation and Mud Column Pressure on Drilling Rate of Permeable Formations,"
Trans., AIME (1959) Vol. 216, 9-17.
[31] Garnier, A. J., and van Lingen, N. H.: "Phenomena Affecting Drilling Rates at
Depth," Trans., AIME (1959) Vol. 216, 232-239.
[32] Christoph Neumüller, Parameter Meta-Optimization of Metaheuristic Optimization
Algorithms, Master thesis, Hagenberg,2011
[33] tashnezhad, A., Wood, D. A., Fereidounpour, A., Designing and optimizing deviated
wellbore trajectories using novel particle swarm algorithms. Journal of Natural Gas
Science and Engineering. 21. 2014.
[34] Blum, C. and Li, X. Swarm intelligence in optimization. Springer., 2008
[35] Re eb, ., Guessoum, Z., & M’ allah, The e ploration-exploitation dilemma for
adaptive agents. In Proceedings of the Fifth European Workshop on Adaptive Agents
and Multi-Agent Systems. March 2005.
[36] Ai-Qin Mu, De-Xin Cao , Xiao-Hua Wang, A Modified Particle Swarm
Optimization Algorithm, Articl, College of Science, China University of Mining &
Technology, XuZhou, China, Received 17 August 2009; revised 28 August 2009;
accepted 30 August 2009.
[37] Wrede R., and Spiegel M.R., “Theory and Problems of Advanced Calculus,”
McGraw-Hill Companies Inc, 2002, pp 73

45
 References

[38] Singh et al., “Cutting Structure for Roller Cone Bits”, United States Patent, United
States Patent and Trademark Office, Patent #: US006374930, Sheet 2 of 8,
http://www.uspto.gov/, (23. April, 2002)

[39] “P C bit profile”, PetroWiki by SPE International, http://petrowiki.org/, Created

Aug 29, 2012, Retrieved April 3, 2015, from PetroWiki by SPE International,
http://petrowiki.org/,
[40] Bourgoyne Jr., A. T., Millheim, K. K., Chenevert, M. E., and Young Jr., F. S.
Applied Drilling Engineering. Society of Petroleum Engineers, Richardson, TX.
(1991).

46
 Summary

SUMMARY
More than ever in the drilling industry, all considerations are involved to reduce
drilling operation e penditure. That’s why The objective for our study is to focus on the
Optimization of the Drilling Parameters, To achieve this goal we began our study by a
dominant and widely utilized method for drilling rate prediction that called Bourgoyne and
Young’s Model. And their suggested analysis method the multiple regression (MR) to define
these constants.
Then our study aims to propose one of the best metaheuristic optimization techniques
(PSO) and the modified one (M_PSO) to improve and compare the quality of solution founded
by the multiple regression method. Through the final the results we have achieved in several
tests and the different comparison of the MR,PSO, and M_PSO using some of the
approximation errors, we validate that the best convergence, optimization performance and the
optimal values are all obtained by the M_PSO. These results lead us to calculate and Optimum
bit weight and rotary speed for the drilling operation

Keywords:Drilling expenditure Optimization drilling rate multiple regression

metaheuristic errors

‫ كل التفكير و الدراسات باتت للحد من نفقات عمليات‬,‫أكثر من أي وقت مضى في مجال التنقيب عن البترول‬
‫لتحقيق هذا الهدف بدأنا بالطريقة‬,‫ لذلك كان الهدف في دراستنا هو التركيز على تحسين خصائص الحفر الى االمثل‬.‫الحفر‬
‫باستعمال طريقة التحليل‬. ‫( للباحثين بوركوين و يونغ‬ROP)‫السائدة والتي تستخدم على نطاق واسع لتنبؤ بمعدل الحفر‬
‫ ثم تهدف دراستنا إلى اقتراح واحدة من أفضل الخوارزميات المصممة إليجاد حلول عاليه‬.‫( الذي اعتمداه‬MR) ‫المتعدد‬
‫( وهذا لتحسين ومقارنة‬M_PSO) ‫( والنسخة المعدلة و المهجنة منها‬PSO) ‫الجوده المسماة باستمثال عناصر السرب‬
‫ ومن خالل النتائج النهائية التي حققناها في العديد من االختبارات‬.(MR) ‫جودة الحل الناتج عن طريقة التحليل المتعدد‬
‫ أكدنا أن أفضل‬.‫ باستخدام بعض االخطاء المطلقة و النسبية و غيرها‬M_PSO ‫ و‬PSO ‫ و‬MR ‫والمقارنة المختلفة بين‬
‫ هذه النتائج تقودنا إلى حساب الوزن على اداة الحفر‬.M_PSO ‫تقارب وأداء والقيم المثلى يتم الحصول عليها جميعًا من‬
.‫األمثل والسرعة الدورانية المناسبة لعملية الحفر‬

‫االخطاء المطلقة‬ ‫الخوارزميات‬ ‫التحليل المتعدد‬ ‫ التنقيب نفقات تحسين معدل الحفر‬:‫الكلمات المفتاحية‬

Plus que jamais dans l'industrie du forage, toutes les considérations sont impliquées
pour réduire les dépenses d'opération de forage. C'est pourquoi l'objectif de notre étude est de
concentrer sur l'optimisation des paramètres de forage, nous avons commencé notre étude par
une méthode dominante et largement utilisée pour la prédiction du tau de forage, c’est le
model de Bourgoyne and Young's. Tell que leurs méthode d'analyse suggérée est la régression
multiple (MR) pour définir ces constantes.
Ensuite, notre étude vise à proposer l'un des meilleures techniques d'optimisation
métaheuristique appelée l’optimisation par essaims particulaires (PSO) et sa version modifiée
(M_PSO) pour améliorer et comparer la qualité des solutions fondées par la méthode de
régression multiple. Grâce aux résultats finaux obtenus dans les plusieurs tests et la
comparaison entre le MR, PSO et M_PSO et après l’utilisation des erreurs d'appro imation,
Nous avons confirmé et valider que la meilleure convergence, performance et les valeurs
optimales sont toutes obtenues à partir de M_PSO. Ces résultats nous conduisent à calculer le
poids optimal sur l'outil de forage et aussi la meilleure vitesse de rotation adaptée au fora

Les mots clé : forage dépenses optimisation taux de forage régression multiple
métaheuristique erreurs