Drill String Vibration - Fial

DRILL STRING
VIBRATION
Techroleum Presentation
Definition of vibrancy
Wave propagation
Natural frequency
Resonance
Damping effect
 Viscous damping
 Coulomb damping
 Solid damping
Wave propagation
These are vibrations that pass through the system in the form of
waves.
So, the force that causes oscillation has influence firstly on the
contact point and then it spreads all over the system.
As a consequence, the force applied to the one end of the drilling
tube will have a time delay before reaching the other side of the
pipe.
If the wave propagation and particles shift in the same direction
then it is called longitudinal waves.
In other words, they are called compressional-tensional or axial
waves.
Natural frequency
Natural frequency is the preferred frequency on which a system
likes to vibrate.
The most important parameters to determine natural frequency
are its geometry and material properties.
When a force is applied to the spring, it travels in the direction of
the force that we applied. This happens till the moment when the
spring’s own force tries to return it to its initial position. This
process happens under natural frequency.
Resonance
If the new force is applied to the system at the moment when we
reach the original position, waves from two excitation sources will
mix and as result increase the amplitude of the already merged
wave.
This phenomenon is called resonance and it can have a severe
effect on drill string,
when we cannot allow the huge motions which may destroy the
string components.
Damping effect
Theoretically the force applied to the system will preserve all
energy.
This is a situation when the pendulum is in constant oscillation or
the spring remines in motion. In practice it was proved that this is
not possible.
As it said before, damping eliminates energy from our system and
stops the string motion bringing it to a stable position.
This is why resonance energy does not cause the failure of drill
string during vibration.
Dampening effect is rising as the object movement through the
viscous media is increased. Dispersion of energy which is the result
of movement material’s parts called Coulomb friction.
There are three kinds of damping:

Damping effect : Viscous Damping
Is related to the resistance of the body that travels with specific
velocity the fluid. It is the most preferred type of damping because
it is easy in calculations. Even if there is no viscous fluid it is
possible to calculate the damping ratio by taking the experimental
and theoretical data.
A Vicious one develops on the contact between the steel and the
mud. It is proportional to relative velocity at the end of the damping
tool.
It is proportional to relative velocity at the end of the damping tool.
Damping effect : Hysteretic Damping
Is the dissipation of energy that happens as a result of contact
between two dry surfaces.
Damping effect : Prevalent Damping
Is a result of dispersion of internal energy. Each solid body
attenuates all vibrations it encounters.
Mechanical Vibrations
Mechanical vibrations characterized as periodic exchange of
potential with kinetic
energy.
The mass and the stiffness are essential parameters of such a
mechanical system,
to which we relate our drilling string.
Components of mass connect two parameters: the force and
acceleration of the system.
Kinetic energy can be generated by the stiffness component's
movement.
Eventually, the displacement of energy will be handled by the
damping element.
Whenever it causes the system to damper, it will change potential
and kinetic energy as heat, that will be lost.
The 3 major types of mechanical motion are:
Free Motions
When a system is originally in an equilibrium position, it is
disturbed by a force that
moves it out of its equilibrium condition, causing free vibrations.
The system will vibrate until it returns to its original state of
equilibrium.
As the drill string cannot move in and become stacked in the bore
hole, jar firing’s tries to free the pipe and, in this situation, the full
drill string system starts to “free vibration”.
In case of no presence of external force the situation of string –
borehole wall interaction
can also be referred to as free motions.
Because of the damping effect, the energy diminishes throughout
time.
Free Motions
Figure 3.1 depicts an example of free vibration behavior.
Forced Motions
Forced vibration occurs when a continuous periodic excitation is
supplied to the
system instead of a single disturbance.
As first force applied to the system, the system will exhibit
transient state behavior before reaching steady state condition that
have same response as the disturbance function and the frequency
of vibration is equal to a forcing frequency.
It's vital to mention that after the repeating disturbance ends, the
system transforms into a free vibrating system that returns to its
original equilibrium point.
Forced Motions
Figure 3.2 shows how forced vibration demonstrated in a
magnitude to frequency domain
Self-excited Motions
This type of vibration depends only on the effect that it causes, not on the
vibrations
that it creates in the system.
This is the main difference between self-excited and forced vibrations, from other
aspects they are quite similar.
Also, it is important to mention that self-excited vibrations have a constant energy
source while in case of forced vibrations we have periodic excitation’s source of
energy.
As a result of friction between the walls of bore and string it is possible that string
will stop rotation because of enough friction.
The top drive will still continue to rotate because of the elastic qualities of the
drilling pipe, it will transfer energy to the string in which parts were in stationary
position in places of contact with borehole walls.
Self-excited Motions
In figure below demonstrates this vibration type.
Why Drill String Vibration is
Important ?
The presence of drill string vibrations is the main cause of loss of
performance
It provokes premature wear and tear of drilling equipment resulting
in fatigue and induced failures such as pipe wash-out and twist-off
It also cause significant wastage of drilling energy
In the oil industry, the improvement of drilling performance is a
matter of crucial economical interest
Drill string vibrations and shocks (V&S) can limit the optimization
of drilling performance
Axial, lateral, and torsional vibrations are generally quite complex
in nature. Phenomena such as bit bounce, stick-slip, forward and
backward whirl and parametric instabilities have been shown to
What is Drill String Vibration ?
Drill String Vibrations are due to the contact of the bit

with the formation and the contact of the string (drill
pipes, collars, stabilizers) with the well wall
 Also other causes like bent pipes and misalignment of
the drill string can cause or increase the severity of drill
string vibrations
Type of Vibration
These have led to the identification and

classification of vibrations into three
separate and distinctive categories
namely
 Torsional (stick-slip oscillations)
 Axial (bit-bouncing phenomenon)
 Lateral (whirl motion due to the out-
of-balance of the drill string)
Torsional Vibration
Downhole measurements show that applying a constant rotary
speed at the surface does not necessarily translate into a steady
rotational motion of the bit.
In fact, the downhole torsional speed typically exhibits large
amplitude fluctuations during a significant fraction of the drilling
time.
This self-excited rotational motion, also known as stick-slip, is
induced by the nonlinear relationship between the torque and the
angular velocity at the bit.
 The torsional flexibility of the drilling assembly exacerbates a non-
uniform oscillatory behavior causing rotational speeds as high as
ten times the nominal rotary table speed or a total standstill of the
bit.
 Torsional vibrations provoke fatigue to drill collar connections,
damages the drill bit and slow down the drilling operation thereby
prolonging the overall drilling process.
Axial Vibration
This vibration mode consists of irregular movements of the
drilling components along its longitudinal axis causing bit-
bounce and rough drilling behavior that destroys the drill bit,
damages the Bottom Hole Assembly (BHA) and increases total
drilling time.
Additionally, due to downhole coupling mechanisms, it also
excites lateral displacements of the string.
 The bit-bounce pattern may be detected at the surface, it is
likely to develop when drilling with a bit of roller-cone type, also
called tricone or rock bit, consisting of multiple lobes which
leads an erratic interaction of the bit with the bottom of the
well making the bit to loose contact with the rock formation.
Lateral vibration
One of the most destructive drill string oscillations is the whirling
phenomenon, since it may be unleashed with no indication at the
surface.
 Deep in the hole, the rotating BHA interacts with the borehole wall
generating shocks from lateral vibrations.
The collisions with the borehole wall will produce eccentric hole and the
shocks can damage components of the BHA.
 The lateral oscillations of the drill string cause severe damage to the
borehole wall and affect the overall drilling direction.
Drill collars whirling are simply the centrifugally induced bowing of the
drill collar resulting from rotation. If the center of gravity of the drill collar
is not initially located precisely on the centerline of the hole, then as the
collar rotates, a centrifugal force acts at the center of gravity causing the
collar to bend.
 Forward and backward whirling behaviors can further intensify due to
the combined effect of fluid damping, stabilizer clearance, and friction of
the drilling assembly against the borehole wall.
Parameters Affecting Drill String
Vibrations

Nature of rock formation
Drilling Pipes
Bit type
Dimensions of borehole
BHA stabilizer
Consequences of Drill String
Failure
Mitigation of drill string vibrations
Vibration Models
Consequently, three types of modes of vibration are studied: axial,
torsional and lateral.
These models can be created by one another and also can be seen
in coupled versions.
Models are divided into two categories:
 Coupled and
 Uncoupled.
The main idea of using modeling is to predict the reaction of each

vibration mode as an
available limited exploitational conditions.
Because of the set of specific features of each vibration, engineers
can recognize the type of vibration mode.
 Torsional vibrations cause slip/stick which is caused by nonlinear friction
during interaction of bit and formations.

Vibration Models
Uncoupled type of models : Axial Models
The vibration along the drill string is referred to as
axial/longitudinal.
For years axial and torsional vibrations were obvious because they
have the ability to appear on the surface.
During axial vibration it is seen bouncing off equipment on the
surface.
The axial load on drill string is composed of two components:
 Static
 Dynamic
Static component has the upper limit of weight on the bit because
after that range buckling happens.
The dynamic one mostly related to bit/rock interaction and
compensated the WOB during penetration of well.
Vibration Models
In the case of the uncoupled axial model there was used differential
equations taking in account lateral bar vibration. (Kreisle and Vance, 1970))
In the primary study by using the mentioned equation it was possible to get
natural frequencies of axial oscillation of a given string. Experimental studies
helped to measure resulting force and displacement of the top of the drill
sting that was created by axial mode of vibration (Finnie and Bailey (1960)).
Study about discontinuous contact of drill bit teeth on axial oscillations
showed that axial vibrations could be seen on the surface (Paslay and Bogy
(1963)).
Common damped equation of axial vibration was linearized to study axial
oscillations of drill string. To get rid of undesired vibrations shock sub with low
stiffness was used.
During research it was also proved that BHA length affects the energy
transfer to the drill bit.
Vibration Models
As a result of research dynamic load of drill string can be less than drill string
weight in static conditions (Lubinski (1988)).
To reduce drill string and wellbore friction while tripping it was present the
idea of using
static model and rotation of drill string.
Future studies showed that quasi statistical analysis gives a more realistic
model than static one.
When there is a bit bounce the drill string will go up and go out the normal
penetration area and create vibrations.
Vibration Models
Uncoupled type of models : Lateral Model
Lateral motions are bending or reverse oscillations, related to transverse
rotation of
string.
There is not much information about vibrations occurring in the downhole
that is
the reason why it is harder to define lateral ones.
To study lateral vibration analytical and finite element models were used.
Due to the reverse movements of drill string to get mathematical models
mostly used beam elements.
For this purpose, Euler-Bernoulli beam theory is used (Baltus, C. (2007)).
The lateral vibration models are grouped into two groups: single- and three-
dimensional planes.
Lateral models assume that bending is not coupled with torsional and axial
oscillations.
Vibration Models
Uncoupled type of Models
Lateral Model : Whirl of BHA
BHA whirl is the situation when we

have bending of drill collars as a
reason of rotational force.
In the situation when the center of
mass is not located in the center
while the
centrifugal force is applied to the
center of mass it is possible to
observe curvature of collars.
Important aspect is distance
between stabilizers center to the
center of mass.
When we have significant curvature
Vibration Models
Uncoupled Type of Models
When there is narrowing of borehole it

happens that there is reduction of drill
collars collapse but increase of whirling.
When there is narrowing of borehole it
happens that there is reduction of drill
collars collapse but increase of whirling.
In case there is whirling in the direction of
drill string moves in borehole- forward
whirl.
When there is one cycle around the hole,
the collar has the same point of contact
with the wall.
When the slippage effect is minimal
Vibration Models
The pipe slides along the borehole wall in

the opposite direction of drill string
rotation in this style of whirl.
When there is no slippage effect, the term
"pure backward whirl" is used to
characterize backward whirl.
As a result of low displacement in the
back whirl makes no possibility to have
one contact point between
collars/borehole, nevertheless it is easy to
recognize it on surface friction.
Vibration Models
Lateral Model : Whirl of Bit
The pipe slides along the borehole wall in the opposite direction of drill
string rotation in this style of whirl.
Bit whirl is similar to BHA whirl in that the bit's initial center of rotation
is pushed away from the geometric center of the bore by an initial
eccentricity force.
An extra adhesion force is created when the bit gets in contact with the
bottom area.
The instantaneous center of spin will be at the contact point if there is
no slip in between the bit and the formation.
This is the same as an automobile tire, where the immediate center of
rotation is at the tire-road contact point.
It is preferable to prevent bit whirl entirely, as one of the whirl type's
Vibration Models
Brett et al. (1989) demonstrated that once the bit swirl started, both field and
laboratory observations revealed a regeneration propensity.
This is caused by a combination of two things.
 The first one is centrifugal force, which is particularly strong in swirl dynamics and is amplified
at high rotating speeds. The whirl's centrifugal force pushes the bit off center, increasing
abrasion
with the structure.
 The second aspect is that the bit teeth are constructed to have the center
of rotation at the center geometry of the well in order to reduce drill force unbalance. When
this rule is broken, the cutters are no longer lined out for full coverage, which increases the
drilling force imbalance.
Vibration Models
A whirling bit would dig an over gauged hole, and this will occur until the drill
collars' restoring force overcomes the whirling bit's regeneration forces (Brett
et al. 1989).
This results in the formation of sills in the well as well as cycling periods of
over gauge and true measurement drill.
Because of this tendency, caliper logs are an excellent diagnostic tool for
detecting spinning activity, since these cycles of true- and over-gauged drilling
possible to identify.
Vibration Models
Uncoupled Type of Models : Torsional
Model
The most used common model - torsional pendulum. This model
assumes that BHA is
rigid body and related to collars and rotary table, pipes assumed
without inertia. (Lin and
Wang, 1991; Jansen and Steen, 1995; Tucker and Wang, 1999).
Such phenomena as stick/slip and whirling were studied by a model
with two degrees of freedom.
Vibration Models
Torsional Model : Stick/Slip
The fundamental source of stick-slip vibrations has been the subject of several
theories.
The discrepancy in torque input to overcoming the friction force in the string
was initially considered to be the cause of stick-slip (Kyllingstad and Halsey
1987).
According to the Richard-Germay-Detournay (RGD) model, the coupling of axial
and torsional vibrations of the bit is the major cause of stick-slip vibrations.
Despite variations in root effect research of stick-slip vibrations, scholars had
acknowledged that stick-slip can be caused by bit or friction induced motions.
The bit rotation rate slows down during the "slip" stage of stick/slip until the bit
comes to a halt or is shifted outside the neutral point. Small intervals of
backward rotation can be noticed in the later situation.
Vibration Models
Figure 4.2 shows how the RPM approaches negative sign before a new
stick stage is launched, as shown by field measurements of stick-slip.
Vibration Models
The high bit’s rate detects the presence of downhole phenomena that
is ignored by observation of surface parameters.
In scenarios when angular movement starts to grow in direction top to
bottom in the direction of bit, drill string is simulated as a pendulum in
order to recognize and evaluate stick/slip.
When torsional resonance occurs, the drill collars vibrate at natural
frequencies that are substantially higher than the drill string’s overall
frequency.
The deflection during torsional resonance appears to be even less
significant, yet it varies across the drill string.
The BHA is basically free at the top because the drill pipe is just less
rigid than that of the collar.
Vibration Models
Coupled Vibration Models
 The main reasons of coupling different vibration modes are:
 Rock and bit interactions
 Drill string curvature
 Torque
 Modification from tension to compression of axial force lengthwise of drill string
Coupling various vibration modes helps to forecast the dynamic

behavior in high accuracy.
These coupled models can be in two possible ways: linear or non-linear.
Extremely explored coupled models are:
 Axial and torsional
 Axial and lateral
 Torsional and bending
Vibration Models
Coupled Vibration Models : Axial–Lateral
Vibrations
 Lateral motions can cause defects and deformations of drill string.
 Axial extension becomes short because the full drill string faces
deformations and bending.
 The points where there are nodes connected and that present the
case of attenuation of lateral motions will have a reductional effect on
such behavior.
 Nevertheless, lateral vibrations are the most frequently occurring and
severe oscillation type which causes the vibration of a full drill string
and shows the relationship with the axial one.
 Numerical calculations are the main evidence of such process
occurrence.
Vibration Models
Vibrations
 In situations when string deviates, the shown
segment with have a form of arc
Vibration Models
Vibrations
 S and H can be calculated by assuming that the drill string deviation is

limited by the
wellbore and pipe size in the segment that is studied.
 As a result, the drill string length reduction can be determined per each
cycle every cycle can be calculated. For instance, for a string of length
12 ¼ with collars 8’’, H is equal to 0.1m. By knowing values of s that
equal to 10m and value of which is 0.08 respectively, putting these
values to the equation of Delta L was found that drill string length
reduction to be 3 ∗ 10*+m in each wave.
Vibration Models
The idea Vibrations
of this theory is that axial oscillations can be transferred to lateral one.
 According to the Dunayev study the first deviation which is quite small is the
initial string position.
 At time t1 it is given the maximum value of axial forces and it steeply decreases
as it reaches t2.
 The axial force changes sign, causing the diversion to reduce.
 The lateral displacement returns to neutral after the axial load has finished its
cycle.
 The lateral vibration has received an amount of energy expressed as extra
kinetic energy.
 Within the next hemi cycle of deflection, this energy leads to a significant
increase of lateral displacement.
 Each axial force cycle results in a continuously rise of lateral displacement
amplitude, which is known as parametric resonance.
 The term "parametric resonance" is not the same as "traditional resonance,"
which is widely practiced in drill string-dynamics simulations.
Vibration Models
 Vibrations
In situation of coupling stick and slip MSE measurement shows how much
energy is used during penetration.
 When there is a whirl, raising the WOB usually lowers the MSE. This is due
to the fact that boosting the WOB reduces whirling tendency. There is less
energy spent on cuttings and friction as a result of this process.
 Stabilizers and other full gauge components in the BHA serve as nodal
points, which means they can't move laterally in the bore.
 As a result, side forces are focused in these areas. When significant
amplitude lateral vibrations occur, the side forces become more powerful.
 Based on this side stresses, rounded shoulders can be observed on the
edges of the stabilizers.
 Because bit whirl creates over gauged holes, bit whirl and BHA swirl are
linked.
 These enormous side forces cause greater bore friction, which causes huge
amplitude torque swings.
Model Including Axial-Lateral
Vibration
In developing the axial-lateral vibrations of rotating drill string the total kinetic energy can be written
as :
 l is the total BHA length

 ρ is the density of the material
 A is the cross sectional area
 I is the cross sectional moment of inertia
 Ω is drilling angular velocity
 v, w and u are respectively deformations
in directions X, Y and Z (Fig. 1).
Vibration
The strain energy due to axial and transverse deformations can be expressed as:
 where E is Young’s modulus.

 In equation (2) the nonlinear axial strain produces the coupling between the axial and
transverse deflections.
Vibration
The contact force between borehole and the collar is defined using the Hertzian law as:
Where :
 K h is the Hertizan stiffness, a parameter which depends on the material properties and
the contact geometry
 D is the borehole clearance.
Vibration
The virtual work due to contact force is:
where zc is impact location.

 The velocity of the collar at the contact point given as:
Vibration
where rc is the position of the geometric center at the impact location and φ is given by:
A damping force resulting from hydrodynamic drag of mud is also acting on the collar. The virtual work due to this
force is considered as:
Where ρ f is the density of the drilling mud and CD is the hydrodynamic drag coefficient. For a rotating drill string ,
gyroscopic moments result from the change in the angular momentum in its bending motion.
Vibration
The virtual work of such moment can be written as:
where is the cross section mass moment of inertia
Finally, the axial WOB excitation, and bit formation interaction can be expressed
as:
 P0 , Pf, n, and Ω are the static component of the WOB, the amplitude of fluctuating
component, the bit constant, and the drilling angular velocity, respectively.
TheVibration
virtual work due to axial force is:
All component of potential and kinetic energy and virtual

work of forces are defined.
Vibration
The governing equations of system can be derived
by using Lagrange’s equation:
TheVibration
equations of motion for coupled axial-lateral
vibrations are obtained by using assumed mode method:
TheVibration
deformation u_st is due to the axial load given as:
Vibration
Vibration Models
Coupled Vibration Models : Axial–Torsional
 Vibrations
Drill strings that acquire a
particular length would
always be torsional flexible.
 The spin is the shear strain,
and the turning tension is the
shear stress inside the string.
 When solid rods are exposed
to torsional stress, they will
reduce axially (2013).
 Figure 4.7 depicts axial
shortness as a result of
twisting.
 The extent of one "fiber"
along the circular pipe is
depicted by the red line.
 Fiber extension remains
Vibration Models
Vibrations
Vibration Models
 The coordinates Vibrations
that characterize spiral:
 This formula shows that when twisting increases there is a

decrease of angle of slope.
 This way we calculate the total system’s twist, by taking in

account the constant values of cross section zone, torque values
Vibration Models
 Vibrations
The extent of a red fiber in Figure 4.7 is constant, and it can be
determined for the spiral by considering the integral from every
progressive length starting from top till the point of full twist at
the bottom parts.
Vibration Models
Coupled Vibration Models : Bending-
 Torsional
Considering Vibrations
the bit and rock interaction, a set of various equations arrived.
 In order to solve these equations Newtonian method in polar coordinate
system used.
 By doing experimental search it was verified that stick/slip phenomena
vanish as whirl appears.
 Using the 4 DoF model which takes in consideration:
 lateral-torsional whirling oscillation and also wellbore/string friction it is possible to
predict the effect of position of drill string and friction coefficient.
 All these models can be coupled together and can evaluate the WOB and
TOB as an outcome of cutting rock. Mud characteristics can also influence
the stick/slip, bit bouncing and buckling of drill string.
 Obviously, a wide range of complex patterns were observed in order to
predict the dynamic response of the drill string.
 Depending on computation time and main aim of modeling the vibration
types should be properly chosen.
 As drill string is not just a construction that oscillates in air with only simple
Inclusion of Torsional Vibration
 Effects
Torsional vibrations in drill string play an important role in many issues of drilling such as rate of
penetration (ROP), over-torque drill pipe connection, twist-offs, premature bit wear and etc.
 These vibrations can be observed either as fluctuations in the current through the electric motor
that drives the rotary table. Commonly torsional vibrations in drill string are observed in three
parts, namely between motor and rotary surface, drill pipe and drill collar
 TheEffects
equations of motion for torsional vibration are obtained as:
 Effects
Main parameters in bit torque are weight on bit (WOB), type of bit and bit speed.
 where W is WOB, 0 , ff t are bit constants, and γ is the decay parameter
 Effects
The torque due to tangential contact force is expressed as:
Substituting Equations
Substituting equations (15)-(17) into equation (1)-(10) and applying Lagrange’s equation for
general coordinate, a set of non-linear ordinary differential equations for coupled axial-lateral
vibration is obtained.
In order to develop the governing equations we substitute equations (16-17) into equations (12-
14) to obtain the deformed shapes of the system. Next by forming the energy terms, equations
of motion are obtained using Lagrange’s equations, equation (11), In deriving the equations of
motion only one mode is considered in the assumed responses, i.e. the one specified in
equations (16) and (17). Choosing the generalized coordinate qi in equation (11) as α leads to:
Second governing equation is obtained by choosing qi in equation (11) as β :
Vibration
By employing equation (11) while γ is the generalized coordinate one obtains the
following equation:
Substituting equations (20) and (21) into equations (18) and (19) one obtains two
following equations which are coupled with equations (22-24) and together form the
set of equations governing the dynamic behavior of drill string :
Numerical Simulations
The problem is solved using Runge-Kutta method in MATLAB (command
line ODE45).
The parameters used in the simulation are shown in table (1), which
represent a typical case in an oil well drilling operation.
Numerical Simulations
The problem is solved using Runge-Kutta method in MATLAB (command line
ODE45).
The parameters used in the simulation are shown in table (1), which represent a
typical case in an oil well drilling operation.
1 radians per second (rad/s) is equal to 9.5492968 rotations per minute (rpm).
Methods to Understand and Mitigate Stick-Slip
Torsional Vibrations Introduction
 It will be shown how stick-slip vibration distributions can be used to evaluate drill string and
parameter redesign options to mitigate stick-slip on the next well.
 It will also be shown how stick-slip vibration data can be used to estimate the time interval
during which the bit is stuck when operating at or beyond full stick-slip.
 The torsional vibration example to be discussed in this paper is based on the distribution of
stick-slip vibration data when downhole data is available, or alternatively using surface torque
swing data.
 The operator's frequency-domain dynamic model relates torque swing at the surface to the
variation in bit rotary speed, as a function of the drill string design and surface RPM.
 Linear relationships between this distribution function and several variables are investigated,
including the critical torque swing at full stick-slip, rotary speed (RPM), and downhole torque.
Torsional Vibrations Background
 It will be shown how stick-slip vibration distributions can be used to evaluate drill string and parameter
redesign options to mitigate stick-slip on the next well.
 It will also be shown how stick-slip vibration data can be used to estimate the time interval during which the
bit is stuck when operating at or beyond full stick-slip.
 The torsional vibration example to be discussed in this paper is based on the distribution of stick-slip
vibration data when downhole data is available, or alternatively using surface torque swing data.
 The operator's frequency-domain dynamic model relates torque swing at the surface to the variation in bit
rotary speed, as a function of the drill string design and surface RPM.
 Linear relationships between this distribution function and several variables are investigated, including the
critical torque swing at full stick-slip, rotary speed (RPM), and downhole torque.
 Torsional vibrations, also known as "stick-slip", occur as the rotational speed of the bit and drill string vary
due to stiffness inertia, and torsional friction dynamic interactions.
 In some wells, these torsional/rotational fluctuations are not severe and may not reach full stick-slip" much
of the time.
 Full stick-slip occurs when the bit comes to a full stop in the torsional vibration cycle.
Torsional Vibrations Background
 When drill bit torque loads are high in competent rocks with
aggressive bits, these vibrations can be severe.
 The bit may not only come to a full stop, but it may be "stuck",
or not rotating, for a significant portion of the stick-slip cycle.
 In such cases, the penetration rates may be slow, and the
interval length may be short as a result of bit and/or drilling
tool damage.
Torsional Vibrations
Torsional Severity Estimate (TSE)
 In the case of stick-slip vibrations, the Torsional Severity Estimate, or TSE, is a measure of
proximity to full stick-slip (Ertas et al., 2014; Bailey et al., 2017).
 A TSE value of 0.0 represents no torsional vibrations whatsoever, where the entire drill
string is turning at a constant angular velocity.
 A TSE of 1.0 is the case of full stick-slip in which the rotary speed at the surface is constant
but the bit speed varies sinusoidally between zero and twice the surface rotary speed.
 Values of TSE greater than 1.0 indicate time intervals in which the bit comes to a complete
stop and then accelerates to more than twice the surface RPM.
 In the most severe cases, negative rotation of the bit may occur.
 Service providers have similar indices of stick slip vibration data based primarily on
downhole tool measurements.
 The sample rate from downhole tools is typically on the order of one value every several
minutes, much slower than that of TSE calculated from surface torque data.
Stick-Slip Distribution Modification
 A histogram of TSE values indicates how severe the vibrations

have been over an interval.
 Small excursions of data at full stick-slip may not warrant much
concern, whereas significant duration in excess of full stick slip
calls for mitigation efforts as drilling performance is likely
affected by the vibration.
 As presented in detail in the Appendix, the histogram may be
used as a tool in redesign. The distribution of stick-slip may be
modified by changing certain parameters.
 From the Appendix, the following relation is one instance of this method that describes a
modified TSE value calculated from an initial TSE value and the changes in drill string stiffness
property (ΔTQSSRef), rotary speed (RPM), and downhole torque (DTOR).
 Note that a set of TSE values may be combined into a distribution of values, and a scaling
function applied to the TSE values will result in a rescaling of the initial distribution, resulting
in a contraction or expansion along the x-axis.
 This mathematical relation, and others provided in the Appendix, are similar in concept to the
Combined Gas Law that combines Boyle’s Law, Charles’s Law, and Gay-Lussac’s Law
 The quantification of stick-slip vibration mitigation using stick-slip

distribution modification may best be explained through a simple
example.
 The torque swing at full stick-slip increases linearly with rotary
speed (Bailey et al., 2016a), such that the system may come out of
stick-slip if the rotary speed is increased a sufficient amount, with
all other parameters equal.
 Consider a drill string with Reference Specific Torque Swing
(ΔTQSSRef) of 100 ft-lbs/RPM.
 This is primarily a function of the drill pipe torsional stiffness and
length. If operated at 100 RPM, full stick-slip would be expected at
 The TSE values are, to first order, linearly related to the following three factors:
 Drill string stiffness: the effective torsional stiffness of a string may be

calculated by a model and represented by the "Reference Specific Torque Swing
at Full Stick-Slip per RPM (ΔTQSSRef)";
 Downhole torque: the torque required by the bit and string, which may in some
cases be effectively modeled by the bit friction factor, weight on bit, and hole
diameter, or in other cases may be the surface or downhole torque value; and,
 Rotary speed.
 When operated at 200 RPM, as seen in Fig. 1b, the TSE for
the same torque swing decreases to 0.5, and the bit will not
be in full stick-slip.
 The peak-to-peak amplitude of the bit speed variation is
proportional to the amplitude of the surface torque variation,
so the same bit speed peak-to-peak variation occurs as that
seen with 100 RPM at surface.
 However, the bit never stops in this case (does not reach
zero RPM) because of the higher rotary speed.
 The effects on the stick-slip TSE distribution for changes in
 Using Eq. 1 provided above and discussed in more detail in

the Appendix
 The rotary speed was doubled; therefore, the TSE was

reduced by a factor of two, with identical drill string and
torque factors.
 This is seen by comparing the TSE distributions in Fig. 1a and
Parameter Redesign to Mitigate Stick-Slip
 Drilling data from two wells is used to illustrate this
Vibrations
methodology in a real-world application.
 Figs. 2a and 2b provide raw drilling data and calculated
values related to torsional vibrations seen in two drill wells,
henceforth referred to as Well 1 and Well 2.
 The torsional vibrations were severe in Well 1 and
significantly mitigated in Well 2, and Well 2 drilled this
challenging interval much faster than Well 1.
Vibrations
 Starting at left, the first parameter is the surface torque TQ. The next parameter is ΔTQ or
Vibrations
the "torque swing" measured from one cycle to the next. This torque swing is directly
related to the stick-slip peak-to-peak values at the bit as seen in the example of Fig. 1. In
Fig. 2, it is evident that the torque swing is more severe for Well 1 (Fig. 2a) than for Well 2
(Fig. 2b).
 The next parameter to the right is the surface rotary speed "RPM". Neither of these
assemblies had a downhole motor. Well 1 showed minimal stall, whereas Well 2 had some
stalling events as it was operated at higher WOB. In some intervals, a top-drive soft drive
control algorithm was tested, however this was off most of the time because it increased the
stall tendency and stick-slip was not the limiter.
 Following the RPM track is ΔTQS or the "specific torque swing per RPM". This is the
measured torque swing divided by the average rotary speed. The next column is the TSE
value for bit stick-slip based on the surface torque swing method. The following column is
the TSE value based on measured downhole bit RPM data, in this case from the MWD
supplier. Overall, the TSE trends are relatively constant in depth, in spite of the testing of
the soft top drive controller in Well 2. Again, stick-slip was not the limiter in Well 2 that it
was in Well 1. The last column shows the measured downhole torque value DTOR, also from
Parameter Redesign
 These results warrant some discussion:
to Mitigate Stick-Slip
Vibrations
 For both actual drill Wells 1 and 2, TSE estimates from both surface torque (TSETQ) and
downhole MWD measurements (TSEBRPM) show high stick-slip in Well 1 and much lower
stick-slip in Well 2, although the shapes of the TSETQ and TSEBRPM distributions are
different.
 The P(TSE>1) indicator based on the cumulative distribution track closely,
demonstrating good agreement between both methods.
 Transformation of the Well 1 distributions for the modified operating conditions yield
distributions that are indeed close to those for Well 2, with good agreement among the
cumulative distributions and the P(TSE>1) criteria, despite differences in operating
conditions previously noted.
 These results were obtained using Eq. 1 with average values for torque and RPM.
The method could be refined by including distributions of these values instead of
average values, which would essentially provide a point-wise evaluation of the
data. The point-wise scaling method would be more complicated but is certainly
feasible, entirely within the current framework of this Parampara distribution
Stick-Slip Vibration "Stuck Time“ Estimation
 In severe stick-slip cases, it is known that significant "stuck
time" can occur. With increasing severity, the sine wave
amplitude increases beyond the full stick-slip swing from
zero to two times the surface rotary speed. The bit slows as
it approaches a rotational speed of zero and can come to a
full stop. This corresponds to the maximum bit speed
exceeding two times the surface rotary speed, and in some
cases, it can be much more. Indeed, the bit may even spin
backwards.
 When the bit comes to a full stop, the waveform becomes a
half-rectified sine wave as seen in Fig. 9. The bit may remain
Stick-Slip Vibration
 Observation of the "Stuck Time“ Estimation
stick-slip period and the TSE can be used to estimate the total
stuck time by summing the two contributions. TSE may be obtained from surface
torque or downhole data.
 In this relation, the actual period and TSE are obtained from drilling data, and the
theoretical period is typically obtained from a dynamic model of the drilling
assembly.
 Eq. 2 can be used to estimate the stuck time. This may be a useful surveillance
diagnostic because in most wells we do not have the detailed data recorded in
Fig. 10, and in any event this data was obtained in recorded mode and not
available at the surface while drilling.
 This concept may "Stuck Time“ Estimation
best be explained with an example. Fig. 10 illustrates a dataset
in which the bit was stuck for 4.7 seconds and slipping for 3.0 seconds. The
measured period was 7.7 seconds, but modeling of the drill string at a depth of
3100 meters determined a theoretical free oscillation period of 5.4 seconds.
 The surface rotary "Stuck Time“ Estimation
speed was 120 RPM and the maximum RPM (blue line) was
490 RPM, and therefore the TSE = (490-120)/120 = 3.1. The actual period that
was measured is 7.7 seconds, and the duration of the cycle extension was 2.3
seconds. From Eq. 2, the total calculated stuck time is 4.4 seconds, determined in
the following way:
 The difference between calculated and measured stuck time is therefore about
0.3 seconds, or about 6% error. However, the stuck time is only 1 second less
than the theoretical free period without sticking and almost 50% greater than the
slip time. Torque dissipates and accumulates during the stuck time, setting up the
conditions for the next severe torsional vibration wave.
Identifying the Optimum Zone for Reducing Drill
String Vibrations
Introduction
 The parameters collected during drilling are weight on bit (WOB), rotation rate
(RPM), pump parameters
(SPM), depth, inclination, azimuth and rate of penetration (ROP).
String Vibrations
Vibration Issues Facing the Driller during
 Two kinds of vibration are of significant
Drilling Operations
concern:
 First is Stick-Slip.
 In this case, the bit periodically stops rotating in a
torque up moment then spins freely, this goes on
through a non-uniform rotation of the drill string.
 During stick slip, the downhole RPM is can be 3x to 15x
average surface RPM.
 The consequences of Stick-slip are PDC bit damage,
lower ROP, connection over-torque, back-off and drill
string twist-offs.
String Vibrations
Vibration Issues Facing the Driller during
 Two kinds of vibration are of significant concern:
Drilling

Operations
The second vibration type is drill string whirling.
 The bulk of drill string whirling happens in the bottomhole
assembly (BHA).
 During whirling, parts of the BHA face lateral displacements which
generate bending stresses and lateral shocks when the BHA
contact the borehole wall (JPT Staff 1998). The act of moving
around the wellbore and not its centerline is the whirling
phenomenon.
 Three types of whirling can occur; forward whirling is a scenario
where the drill string is rotating around the wellbore in the same
direction with its rotation around its own centerline; backward
whirling is a situation where the drill string is rotating around the
wellbore in a direction opposite the direction of its rotation around
String Vibrations
How Machine Learning is utilized for
 WOB and RPM causing whirling and
Vibration Problems
stick slip can be predetermined if
the total drilling conditions are
known (Wu et al 2010).
 A boundary condition for stable
drilling can be obtained in a plot
with WOB on the Y axis and RPM on
the X axis (Figure 2).
 This means if the driller keeps his
drilling parameters to keep the bit
in the optimum zone, then drilling
will be stable depending on the bit
and mechanical properties of the
rock.
 The boundaries of the optimum
zone help determine the best
combination of WOB and RPM for
String Vibrations
How Machine
 Thus, the Learning is utilized for
need for machine learning. In order to identify the optimum zone
Vibration Problems
effectively, a process, as shown in Figure 3, is adopted to ensure all drilling
parameters have an impact on the optimum zone.
 The process uses available real-time data from the drilling rig, performs a
variable transformation and reduction, and then utilizes machine learning
algorithms to identify the optimum zones.
 The zones are then transformed back into the original variables and
displayed to the driller. Principal Component Analysis (PCA) is used to
achieve the variable transformation and reduction.
 K-means clustering and decision tree classification are then used to identify
the optimum zones; after which the boundaries are represented based on
WOB and RPM.
String Vibrations
Compressing Data
 Thus, the need for machine Using PCA
learning. In order to identify the optimum zone
effectively, a process, as shown in Figure 3, is adopted to ensure all drilling
parameters have an impact on the optimum zone.
 The process uses available real-time data from the drilling rig, performs a
variable transformation and reduction, and then utilizes machine learning
algorithms to identify the optimum zones.
 The zones are then transformed back into the original variables and
displayed to the driller. Principal Component Analysis (PCA) is used to
achieve the variable transformation and reduction.
 K-means clustering and decision tree classification are then used to identify
the optimum zones; after which the boundaries are represented based on
WOB and RPM.
String Vibrations
Compressing Data Using
 PCA can be used for identifying PCA
hidden patterns in high dimension data; it is
commonly applied in image compression.
 PCA is used to find differences and similarities in data.
 Part of the PCA advantage is dimension reduction without losing vital
information.
 PCA is an orthogonal transformation of a correlated data matrix of n objects
by p variables into uncorrelated axes called principal component that are
linear combinations of the original p variables.
String Vibrations
Compressing Data
 Consider the data to be Using PCA
matrix M where the transpose of the matrix is MT.
 A scalar λ is called an eigenvalue of the n × p matrix M. If Mx = λx , then x
is called an eigenvector corresponding to the eigenvalue λ.
 The PCA procedure involves obtaining the eigenvectors for either MMT or
MTM.
 Applying this to the initial data, the principal eigenvector axis is the one
with its points well spread which is where the data has its variance
maximized.
 These points lie along this axis with very little deviation away from the axis.
 The second axis is the eigenvector with the second highest eigenvalue in
which the variance of distances from the first axis is greatest, and so on
(Leskovec et al 2014).
String Vibrations
Compressing
 In other words, PCA is Data Using PCA
a technique which generates new features which are
linear combination of the original feature.
 For instance, given a dataset in a "d dimensional space" is transformed to a
"z dimensional subspace" where z < d.
 The new dimensions in z are called principal components and they are
directed in the path of maximum variance without repeating variance in
preceding components.
 This means the first component has the highest variance while each
component that follows it have a corresponding smaller variance value
(Vasan et al 2016). The Principal Components can be represented as the
following
String Vibrations
Case Study
 Surface and downhole data from a well in continental US was used in this
study to evaluate the potential of machine learning to identifying optimum
zones while drilling. Six drilling parameters were of concerning to ensure
optimum drilling operations.
 These parameters formed the original data input into the machine learning
process. Table 1 shows the significance of each of these parameters to the
drilling process.
 The data for this well form a 1000 by 6 matrix which will now follow the
process shown in Figure 3 in order to inform the driller what the optimum
zone for drilling is per formation drilled.
String Vibrations
Case Study
String Vibrations
Results
 The PCA results show that 99.9535 percent of the input variables was
explained by the first principal component as shown in Figure 4 hence
achieving dimensionality reduction.
String Vibrations
Results
 The plot shown in Figure 5 displays the principal components subspace
alongside the coordinates of the original data points. It can be observed
that most of the points lie along the MSE coordinates showing that MSE
contains an aggregate of the information contained in the other five
variables.
String Vibrations
Results
 The result of K-means done after PCA shows that for this data, 2 clusters are
the most stable indicating stick slip and whirling present as drilling took
place as shown in Figure 6. The number of stable clusters was picked based
on the peak mean silhouette value which is 2 for this data. This value was
confirmed when K-means clustering was also carried out on the original data
set. The peak mean silhouette value was also 2 as shown in Figure 6. The
major benefit of running the K-means after PCA is done is to reduce the
noise in the data that often disturb the validity of the results.
String Vibrations
Results
 Figure 7 compares clusters formed by applying K-means on just the original
data and K-means on PCA produced data. It is clearly seen that the
distinction of the two clusters is higher after the PCA has been done. This
gives the driller a distinct idea of where the optimum zone is.
String Vibrations
Results
 For more distinct clarity on where the optimum zone is, a decision tree is
designed.
 The classes are chosen based on conventional drilling solutions for stick slip
and whirling as described in figure 8.
 In order to form the basis for decision, the process was reversed to give the
worst-case scenario where stick slip and whirling will definitely occur.
 The results in certain boundaries as shown in Figure 9.
 First, the truncated mean for WOB and RPM in the input data is obtained
and then four equation can be derived.
 The stick slip line is the third reverse of 5 percent decrease in WOB, the low
ROP line is the third reverse of 10 percent increase in RPM, the backward
whirling line is the third reverse of 10 percent decrease in RPM while the
forward whilring line is the third reverse of 10 percent increase in WOB.
 Based on these conditions, new data can be classified into 5 classes. Class 1
is the "potential stick clip zone," class 2 is the "potential optimum zone,"
class 3 is the "potential backward whirling zone," class 4 is the "potential
String Vibrations
Results
String Vibrations
Conclusions
 The process described in this paper practiced the use of supervised
(decision tree classification) and unsupervised (K-means clustering)
machine learning to help the driller visualize what appropriate value of WOB
and RPM should be used when change is needed.
 Based on literature, an optimum zone is formed which shows the driller how
to avoid stick slip and whirling problems that occur during drill string
vibrations.
 The boundaries of the optimum zone help determine the best combination
of WOB and RPM for optimum ROP.
 Running PCA on the data before clustering gives the driller a more distinct
view of where the optimum zone is.
 The zone for each data point can be deterred using decision tree
classification.
Mechanical Specific Energy (MSE)
What is MSE ?
The mechanical specific energy or MSE, is

commonly defined as the amount of energy
required to destroy a unit volume of rock.
A measure of drilling efficiency.
It is expressed in terms of lbs/in2 (psi) or N/m2
(more commonly referred to as a Pascal, Pa).
Why MSE Is Important ?
MSE has become a common way to analyze drilling efficiency post-run and,
in some cases, make corrections in real time to improve rate of penetration
(ROP).
A typical use is to compare the MSE to the rock strength to see whether the
right amount of energy is utilized at the bit and not wasted or dispersed
somewhere else.
However, MSE alone cannot tell if drilling inefficiency is due to a change in
the rock hardness, or due to vibrations, or bit wear or bit balling.
For optimal drilling efficiency, the objective is to minimize the MSE and to
maximize the rate of penetration (ROP).
As MSE is mainly affected by the level of downhole torque (TOB), the effect
of WOB is often neglected and is not taken into account in standard MSE
analysis.
MSE measures the connection between input energy and ROP as a ratio.
Calculate MSE
Where:
 MSE = Mechanical Specific Energy, psi
 AB = Bit Area, in2
 RPM = Bit rotation speed, rpm
 TOB = Torque on bit, ft-lbs
 ROP = Rate of penetration, ft/hr
It is important to mention that:
 The torque term used in the above equation should be the torque on the bit in order to
estimate the amount of downhole energy required to cut the rock. If surface torque is used
instead, MSE will be highly overestimated, especially in deviated wells, because of the
overall friction along the drill string.
 Additionally, RPM should always be the bit RPM, therefore when using a mud motor the
revolutions per gallon (rev/gal) and flow rate should be used to calculate the bit RPM and
then added to the string RPM for use in MSE calculations.
Calculate TOB
 Where:
 ΔP = Differential Pressure (psi)
 Tmax = Mud Motor max-rated torque (ft-lbs)
 Pmax = Mud Motor max-rated ΔP (psi)
Some MWD tools are now equipped with downhole TOB sensors and when telemetry rates allow, data can be
sent to surface at a frequency which allows good TOB data to be used for MSE analysis.
In the absence of measurement sensors close to the bit or along the BHA, mud motor differential pressure can
also provide valuable information about the torque delivered at the bit and be used as a TOB measurement.
However, this depends entirely on the diligence of the driller to zero out the differential pressure at every
change in pump rate.
This method also requires an accurate motor performance curve to describe the relationship between torque
and differential pressure which is assumed here to be linear.
ROP Prediction and Optimization in Real-Time Using
Supervised Machine Learning
ROP optimization has long been a quest in reducing the cost of
drilling wells.
This has become more acute in the long lateral sections of
unconventional reservoir wells.
 The drilling industry is moving towards the automated optimization
of key controllable parameters such as differential pressure, weight-
on-bit, rotary speed, and flow rate.
There is the often discussed “sweet spot” where ROP is maximized
and lateral, axial, and torsional vibrations are minimized.
Drilling efficiency parameters such as Mechanical Specific Energy
(MSE), and depth of cut must be kept in check to protect PDC bit life.
Case Study
Roughly #1
4,000 horizontal wells have been drilled in
about half a dozen reservoirs in Texas and New
Mexico in an area collectively called the Permian
Basin.
These wells are remarkably similar in trajectory
design as compared to other areas.
Lateral sections are getting longer as time
progresses with 10,000 feet currently being a good
average and some operators pushing to 13, 000
feet.
This makes the Permian an ideal area to test new
drilling analysis techniques. The trajectory design
of these wells (shown in Figure 1) has four sections:
1) Surface vertical
2) “S” shaped trajectory nudge (in 60% of wells)
3) Build to horizontal – with build rates between 6 and 16
deg/100ft
4) 8,000 to 12,000 ft lateral in a Permian shale group
Case Study #1Data Preparations
Drilling
The drilling data gathered from these wells was a 16 parameter
subset of the standard WITSML data
collected by the electronic data recorder (EDR) providers with a
timestamp.
 These include six direct measurements:
 STQ – surface rotary torque,
 SRPM – surface rotations per minute
 SPPA –standpipe pressure (usually averaged)
 HKLD – hook load
 Mud flow rate in FLOWI
 Measured depth MD
Drilling
timestamp.
 five computational parameters:
 ROP – rate of penetration
 SWOB – surface (calculated) weight-on-bit
 DPRES – differential surface pressure
 MSE – mechanical specific energy
 HSI – bit hydraulic horsepower per square inch
Drilling
timestamp.
 five parameters:
 MW – Mud density or weight
 PV – Mud plastic viscosity
 YP – Mud yield point
 BS – Bit size
 TFA – Bit total flow area
Case Study #1
Determining a correlation or linear relationship
among the parameters and ROP (the bottom row)
is very difficult.
The Spearman calculations result in a coefficient
that can be considered sensitive to linear
correlation.
This is shown in Table 1. This will be the starting
point for testing machine learning models.
Since we have both input features and
corresponding output value available as training
examples, we used only supervised learning
algorithms for model training.
Case Study #1
Case Study #1
Additionally, the drilling parameters or predictive features were evaluated to
determine if they were normally distributed.
Usually the goal in machine learning and artificial intelligence work is to
strive to organize data in terms of linearity.
 Processing data to obtain a linear relationship between cause and effect
helps to reduce variance and computational complexities in further work.
It helps in handling outliers and data that carries too much analysis weight
called leverage points or influencing points. If “X” (drilling parameters) is
normally distributed and “Y” (ROP) is also normally distributed, one is more
likely to fit a straight line with most of the data centered in the middle of the
line rather than at the end points, aka outliers, leverage, and influencing
points.
Case Study #1
Testing Machine Learning Models
 The relationship between ROP and the
different predictive features is complex.
 Eight machine learning algorithms were
selected for testing.
 45 of the 50 wells were picked at random for
this work.
 The remaining five were reserved for a blind
Case Study #1
Testing Machine Learning Models
 Three linear models were chosen:

 1) Multivariate linear regression
 2) Least absolute shrinkage selector operator (LASSO) regression
 3) Ridge regression.
 Three non-linear models were tested;
 1) Random forest/decision trees
 2) Artificial neural networks (ANN) / deep learning
 3) Recurrent neural networks (RNN)
 The remaining algorithms are:
 Principal Component Analysis which is a decomposition technique
 Multivariate adaptive spline regression or spline regression for short
Case Study
 Mean #1 percentage error (MAPE) was used to judge the effectiveness of
absolute
the eight models.
 It is given by the following formula:
 Where “n” is the number of fitted data pairs, “At” is the actual value of ROP and
“Ft” is the forecast value.
Case
GammaStudy #1
ray measurements are basically the only petrophysical data gathered in
unconventional wells.
No correlation was seen between lagged gamma ray and ROP.
This was not surprising in total shale environments.
The next approach was to add engineering features that used lagged operational
drilling data which could act as
a proxy for formation information.
Several combinations were tested: Lagged MSE, SRPM, DPRES and STQ were found to
significantly improve the predictions.
This reduced MAPE by approximately 20% for our winning model

Case Study #1
Results
 The performance of the eight machine learning models on the 45 test wells is detailed in Table 2.
 Decision Trees and Random Forest models performed well on training data but poorly on blind data due to
overfitting.
 Deep learning and artificial neural networks models gave acceptable accuracy but were rejected because
their black box nature resulted in poor interpretability.
 Shrinkage methods such as LASSO and Ridge Regression gave clear interpretability but reduced accuracy
because of the assumed linear relationship.
 The winning model was multivariate adaptive spline regression because it is computationally inexpensive
and fast to keep up with real time drilling, provides the best accuracy, and offers clear interpretability.
Case Study #1
Results
 The next step was to predict ROP in the remaining five wells in a blind test.
 Table 3. shows those results.
 While there was a wide range in ROP in the wells and in the average ROP for a well,
the error was fairly
consistent at about 13%. This suggests stability in the predictive model.
Case Study #1
Results
 Figures 4&5 show the predicted versus actual ROP for sections of two of the blind test wells.
Case Study #1
Optimizing ROP
 Once a reliable and consistent machine learning model is

established, the next step is to use it to predict
better values for the forecast feature.
 In this case ROP optimization occurs when features can be
tested to yield a higher forecast number for ROP.
 The model develops a weighting function for the input
features.
Case Study #1
Optimizing ROP
 Once a reliable and consistent machine learning model is established, the next step is to use it
to predict better values for the forecast feature.
 In this case ROP optimization occurs when features can be tested to yield a higher forecast
number for ROP.
 The model develops a weighting function for the input features.
 Table 4. Lists the general weights for the features.
 DPRES was identified as the most important feature by a significant margin.
 Differential pressure caused by the energy required for a PDM to rotate and cut rock is an
indirect measurement of downhole drilling torque at the bit.
Case Study #1
Optimizing ROP
 Figure 6. shows improvements in ROP forecasts as a function of DPRES.
 Since there is a weighted relationship among the input features, optimization works
through all the parameters to determine the best ROP.
Case Study #1
Optimizing ROP
 There is usually a “sweet spot” where any further testing of the features does not
improve or actually reduces
ROP.
 It is dimensionally very difficult to plot this process. Figure 7. shows a flow chart for
this.
Case Study
Internal #1
Live Test
 The 50 well population led to the development of this spline regression machine
learning model for first
predicting and then recommending optimization for ROP.
 This was, by definition, a post-mortem data analysis process.
 Qualifications and normalizing techniques must mesh with that data stream,
process lagged features, predict ROP, and make recommendations for optimization
and not fall behind in a real world drilling environment.
Case Study
Internal #1
Live Test
 Figures 8 and 9 show two snapshots from real-time screens.
Case Study
Internal #1
Live Test
 Figures 8 and 9 show two snapshots from real-time screens.
Case Study #1
Conclusions
 This work evaluated eight machine learning techniques and found that a spline
regression model could maintain a consistent error in predicting a MAPE of 13% in
the fairly uniform rotary drilling of unconventional lateral sections of Permian Basin
wells.
 The novelty of this work is that the model does not require retraining on every new
formation.
 Differential pressure caused by downhole motor action was found to be the most
significant input feature which was not surprising in the work since it is a proxy for
downhole torque, though this differs from other work in this field.
References
An Overview on the Modeling of Oilwell Drilling Vibrations
 B. Saldivar ∗ I. Boussaada ∗ H. Mounier ∗ S. Mondi´e ∗∗ S. I. Niculescu ∗
 Proceedings of the 19th World Congress The International Federation of Automatic Control Cape Town, South Africa. August 24-29,
2014
DRILL STRING VIBRATION MODELING INCLUDING COUPLING EFFECTS

 H. Ahmadian, S. Nazari and H. Jalali
 IUST International Journal of Engineering Science, Vol. 18, No.3-4, 2007, Page 59-66 Industrial & Mechanical Engineering Special
Issue
DRILLING VIBRATION MONITORING & CONTROL SYSTEM

 Martin E. Cobern, Ph.D., Director of New Product Development & Mark E. Wassell†, Manager of Analysis APS Technology, Inc. 800
Corporate Row Cromwell, CT 06416, USA
Study of Critical Speed for a Flexible Drill String System Based on Fluid–
Structure Interaction
 Applied Mechanics and Materials Online: 2011-09-27
 ISSN: 1662-7482, Vols. 105-107, pp 545-552
 doi:10.4028/www.scientific.net/AMM.105-107.545
 © 2012 Trans Tech Publications, Switzerland
DRILL STRING VIBRATION: A PROXY FOR IDENTIFYING LITHOLOGIC

References
Use of Mechanical Specific Energy Calculation in Real-Time to Better Detect
Vibrations and Bit Wear While Drilling
 S. Menand and K. Mills, DrillScan US Inc.
 Copyright 2017, AADE-17-NTCE-033
Real Time Optimization of Drilling Operations and Performance Based on

Drilling Specific Energy Determined from Actual Field Data
 Lazăr D1, Halafawi M2*, Avram L2, Sorin R1 and Sorin G3
 Petroleum & Petrochemical Engineering Journal : ISSN: 2578-4846
Cloud Based ROP Prediction and Optimization in Real-Time Using Supervised

Machine Learning
 Kriti Singh*, Sai Sharan Yalamarty, Mohammadreza Kamyab, and Curtis Cheatham, Corva AI.
 Copyright 2019, Unconventional Resources Technology Conference (URTeC) DOI 10.15530/urtec-2019-343
From Science to Practice: Improving ROP by Utilizing a Cloud-Based

Machine-Learning Solution in Real-Time Drilling Operations
 Kriti Singh, Sai Yalamarty, Curtis Cheatham, Khoa Tran, and Greg McDonald, Corva AI
 Copyright 2021, SPE/IADC International Drilling Conference and Exhibition

References
Drilling Vibration Modes and Penetration Rate Modeling using Arti¦cial
Neural Network and Multiple Linear Regression Analysis in Khoman
Formation at the Egyptian Western Desert
 Sherif A. Ezz, Mohamed S. Farahat, Said Kamel, Ahmed Z. Nouh
 DOI: https://doi.org/10.21203/rs.3.rs-337955/v1

Drill String Vibration - Fial

Uploaded by

Copyright:

Available Formats

Drill String Vibration - Fial

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Drill String Vibration - Fial

Uploaded by

Copyright:

Available Formats

DRILL STRING

Drill String Vibrations are due to the contact of the bit

These have led to the identification and

The main idea of using modeling is to predict the reaction of each

BHA whirl is the situation when we

When there is narrowing of borehole it

The pipe slides along the borehole wall in

Coupling various vibration modes helps to forecast the dynamic

 S and H can be calculated by assuming that the drill string deviation is

 l is the total BHA length

 where E is Young’s modulus.

where zc is impact location.

where is the cross section mass moment of inertia

All component of potential and kinetic energy and virtual

 This formula shows that when twisting increases there is a

 This way we calculate the total system’s twist, by taking in

 A histogram of TSE values indicates how severe the vibrations

 The quantification of stick-slip vibration mitigation using stick-slip

 Drill string stiffness: the effective torsional stiffness of a string may be

 Using Eq. 1 provided above and discussed in more detail in

 The rotary speed was doubled; therefore, the TSE was

The mechanical specific energy or MSE, is

 Three linear models were chosen:

This reduced MAPE by approximately 20% for our winning model

 Once a reliable and consistent machine learning model is

DRILL STRING VIBRATION MODELING INCLUDING COUPLING EFFECTS

DRILLING VIBRATION MONITORING & CONTROL SYSTEM

 ISSN: 1662-7482, Vols. 105-107, pp 545-552

 © 2012 Trans Tech Publications, Switzerland

DRILL STRING VIBRATION: A PROXY FOR IDENTIFYING LITHOLOGIC

 Copyright 2017, AADE-17-NTCE-033

Real Time Optimization of Drilling Operations and Performance Based on

 Petroleum & Petrochemical Engineering Journal : ISSN: 2578-4846

Cloud Based ROP Prediction and Optimization in Real-Time Using Supervised

 Copyright 2019, Unconventional Resources Technology Conference (URTeC) DOI 10.15530/urtec-2019-343

From Science to Practice: Improving ROP by Utilizing a Cloud-Based

 Copyright 2021, SPE/IADC International Drilling Conference and Exhibition

You might also like