energies

Article
Determination of In-Situ Stress and Geomechanical
Properties from Borehole Deformation
Hong Xue Han and Shunde Yin *
Department of Petroleum Engineering, University of Wyoming, Laramie, WY 82071, USA; hhan3@uwyo.edu
* Correspondence: syin@uwyo.edu; Tel.: +1-307-7663511

Received: 12 December 2017; Accepted: 3 January 2018; Published: 5 January 2018

Abstract: This paper proposes a cost-effective technique to determine geomechanical properties

and in-situ stress from borehole deformation data. In this approach, an artificial neural network
(ANN) is applied to map the relationship among in-situ stress, borehole size, geomechanical
properties, and borehole displacements. The genetic algorithm (GA) searches for the set of unknown
stresses and geomechanical properties that matches the objective borehole deformation function.
Probabilistic recapitulation (PR) analysis is conducted after each ANN-GA modeling cycle and
will be repeated with a reduced number of unknowns for the next ANN-GA modeling cycle until
unequivocal results are achieved. The PR-GA-ANN method has been demonstrated by a field case
study to estimate borehole size, Young’s modulus, Poisson’s ratio, and the two horizontal stresses
using borehole deformation information reported from four-arm caliper log of a vertical borehole.
The methodology effectively solves the issue of the multiple solutions (various rock mechanical
properties and in-situ stresses combinations) for a certain borehole deformation. The case study also
indicated that the calculated horizontal stresses are in reasonable agreement with the filed hydraulic
fracture treatment observations and the reported regional stress study of the area.

Keywords: in-situ stress; geomechanical properties; caliper log; borehole deformation; borehole size;
bit size; probabilistic recapitulation

1. Introduction
In-situ stress and geomechanical properties play important roles in the drilling and completion
features of shale gas reservoirs, such as well design, well bore stability analysis, fracture design, as well
as reducing asymmetric fractures in a pad area to increase the recovery rate in shale gas extraction
with hydraulic fracturing. In-situ stress contrast between caprock and pay zone is also a controlling
factor for fracture containment, providing a scientific basis for assessment of the environmental
impact of hydraulic fracturing practice in shale gas extraction and other geo-energy development
activities. Therefore, effective determination of the parameters will be beneficial to these practices.
Geomechanical properties of rocks, such as Young’s modulus and Poisson’s ratio, are often statically
measured from core samples in lab tests. These properties can also be calculated dynamically from sonic
and density logs, which requires calibrations by static measurements. The magnitude of minimum
in-situ stress can be measured by leak off test (LOT) or hydraulic fracturing test [1,2]. Small scale
hydraulic fracturing test is called min-frac test. There are many types of mini-frac test, such as
diagnostic fracture injection test (DFIT) of Halliburton and modular formation dynamics tester (MDT)
of Schlumberger. These methods are most commonly used nowadays and considered most reliable.
For a vertical drilled borehole in a normal faulting stress regime [3], the measured minimum principal
stress is the minimum horizontal in-situ stress. The determination of maximum horizontal in-situ
stress magnitude is often achieved by calculation using Kirsch equation [4] from borehole breakouts,
minimum horizontal stress, and geomechanical properties such as cohesion, friction angle, unconfined

Energies 2018, 11, 131; doi:10.3390/en11010131 www.mdpi.com/journal/energies

Energies 2018, 11, 131 2 of 13

compressive strength (UCS), etc. [5,6]. Many attempts have been carried out on other means of in-situ
stress determination. Ervin and Bell used breakdown pressure or leak off pressure from formation leak
off test to calculate the maximum horizontal stress [7]. Cornet and Valette developed a method based
on normal stress measurements and fast flow rate reopening tests to calculate in-situ stress [8].
In actual practice, both laboratory geomechanics tests and field in-situ stresses measurements
require a substantial overhead cost and a prolonged waiting time, while the results may only be
available for limited formations in only few wells in an oilfield. Moreover, properties achieved
from lab are not in-situ, where mismatched confining stresses must be applied to mimic the in-situ
conditions underground. Therefore, many attempts have been made to develop techniques to
determine geomechanical properties or in-situ stress in an economical and prompt manner while
maintaining the measurement in-situ. Some researchers [9,10] tried to establish empirical relations
between rock mechanical properties and sonic and density logs or rock physical properties. It is
commonly accepted nowadays in the petroleum industry that mechanical properties and in-situ stresses
may be calculated from petrophysical logs with calibration and then by applied to geomechanical earth
modeling [11]. In 1980s, Aadnoy developed a method to determine the orientation and magnitudes
of horizontal in-situ stresses based on the formation breakdown pressure of a circular borehole of
arbitrary trajectory [12]. In drilling process, borehole deformation is closely related to in-situ stress
and rock mechanical properties. Therefore, it should be beneficial if such deformation data can
be used to determine in-situ stresses and/or rock mechanical properties. Laboratory tests have
been conducted to investigate borehole stability and deformation of drilling in granite under high
pressure and temperature [13]. The effect of the tensile and shear movement of rock on borehole
deformation was analyzed by numerical simulation [14]. Reservoir scale formation deformation or
surface uplift with regard to the in-situ stresses and mechanical properties variation can be modeled
by coupled geomechanical and reservoir simulation, the results can also be used to estimation some of
the geomechanical properties [15,16]. Artificial intelligence methods have been applied to map the
relationship between in-situ stress and the displacements of a borehole wall, and genetic algorithm has
been then applied to calculate the in-situ stress [17–19]. Recently, further study has been conducted to
determine in-situ stress from the measurement of width and depth of breakouts at borehole by using
finite element method [20].
The following challenges, however, remain in the abovementioned method for determination
of geomechanical properties and in-situ stress from borehole deformation: first, there is uncertainty
in uniqueness of the solution due to the unknown factors that affect borehole deformation; second,
the borehole size used in calculation cannot be simply assumed equal to the bit size. Addressing these
challenges will help facilitate the filed application of the methods. To that end, we developed a new
technique to estimate both in-situ stress and geomechanical properties from borehole deformation by
considering borehole size as an unknown.
In the following sections, the theoretical basis of the proposed approach is introduced; work
flow of probabilistic recapitulation-genetic algorithm-artificial neural network (PR-GA-ANN) method
is illustrated; and finally, a case study in Appalachian Basin is presented, which demonstrates the
feasibility of estimating geomechanical properties and in-situ stress from borehole deformation.

2. Mathematical Model for Borehole Displacement

When a borehole is drilled, there will be several borehole shape scenarios (depending on the
mud pressure, geomechanical properties, and the in-situ stresses), such as breakouts, drilling-induced
fractures, elliptical boreholes, circular holes, or the combination of two or more of these events. If the
borehole rock has not yielded under two deferent stresses that are orthogonal to the borehole cross
section, an elliptical borehole will be formed, as shown in Figure 1, with a vertical borehole considered
as an example.
Energies 2018, 11, 131 3 of 13
Energies 2018, 11, 131 3 of 12

Figure 1.
Figure 1. Schematics
Schematics of
of borehole
borehole deformation.
deformation.

According to
According to the
the elasticity
elasticity theory,
theory, elastic
elastic solids
solids subjected
subjected to
to loadings
loadings will
will deform,
deform, and
and these
these
deformations can be quantified by knowing the displacements of material in the body [21]. When aa
deformations can be quantified by knowing the displacements of material in the body [21]. When
circular hole
circular hole is
is drilled,
drilled, the
the displacements
displacements around
aroundborehole
boreholecan
canbe
bewritten
writtenas:
as:
h1  σ   i
σHh + σ h  ×  r + R2   σ − σ    R 4 4 4 × R 2 2
µrr =μrr−=E1− σH+ θ()2θ

×  r 
+ R2

+ +  σHH−σh h ××  rr −− R3 3 ++ 4× R  × cos(2
× cos  )
h E 2  2    r r
  22    r r r ri  
R2 2 R44
+ Eυ υσH− σh
× r − + σH −σh
2 σ − σ h   r R   σ2H − σ h   × r − r3 
R × cos ( 2θ ) (1)
+  H  ×  r −  +   ×  r − 3  × cos(2θ )  (1)
+4 × υEE ×  R2 2× σH −σh r×cos  (2θ2) + υ × r ×rσv+ (1+υ) ×
2

r 2 E E
 Pm × r
υ R  σ −σh 
2
+4 × × ×  H
2
υ
cos(2θ ) + × r × σ v +
(1 + υ ) × P × r
 ×the m
E r radius;
where: r is original borehole  2 R 
is distanceE from E
borehole center; µrr is the displacement of
the borehole at R distance from the borehole center; E is Young’s modulus; υ is Poisson’s ratio; σv is
where: r is original borehole radius; R is the distance from borehole center; μrr is the displacement of
vertical stress; σH is maximum horizontal stress; σh is minimum horizontal stress; Pm is borehole mud
the borehole at R distance from the borehole center; is Young’s modulus; υ is Poisson’s ratio;
pressure; θ is the angle from maximum horizontal stress direction.
is vertical stress; is maximum horizontal stress; is minimum horizontal stress; Pm is borehole
When R = r, the displacement at the wellbore wall can be calculated by:
mud pressure; θ is the angle from maximum horizontal stress direction.
When R = r, theµrr displacement
= − E1 [(σH + σat the wellbore wall can be calculated by:
h ) × r + ( σH − σh ) × 2r × cos(2θ )]
(2)
+2 × υE × 1r × (σH − σh ) × cos(2θ ) + Eυ × r × σv + (1+E υ) × Pm × r
2

(σ H + σ h ) × r + (σ H − σ h ) × 2r × cos(2θ ) 
μrr = −
E
point A in Figure(2)1,
On the borehole wall2 along the maximum horizontal stress1direction,
υ
× r × (σ Hby:
υ
− σ h ) × cos(2θ ) + × r × σ v +
( +υ )
the displacement can+be
2 ×determined × Pm × r
E E E
rh 2 i
On the borehole µwall
rrA =along υ (
theσH −
maximum
σ h ) + + ( 1 +
horizontal
υσv υ ) P −
stress
m ( 3σ −
direction,
H σ h )point (3)
A in Figure 1, the
E
displacement can be determined by:
On the borehole wall along the minimum horizontal stress direction, point B in Figure 1,
r 2
the displacement can beμdetermined
rrA =
υ (by:
σ H − σ h ) + υσ v + (1 + υ ) Pm − (3σ H − σ h )  (3)
E
rh 2 i
rrB =along
On the borehole µwall 2υthe − σH ) + υσ
(σhminimum horizontal ) Pm − direction,
v + (1 + υstress (3σh − σH )point B in Figure 1, the
(4)
E
displacement can be determined by:
The longer axis (C13) and shorter axis (C24) of the four-arm caliper measurements, which are
corresponding to the deformation
r
μ rrB = 2υof2 (the
E
[ σ H ) + υσwall
σ h −borehole + υpoint
v + (1at ) Pm − B(3σ
and ]
σ H ) A respectively, can(4)
h − point be
determined by the Equations (5) and (6):
The longer axis (C13) and shorter axis (C24) of the four-arm caliper measurements, which are
2r h 2 i
corresponding
C13to=the
2(µdeformation
rrB + r ) = of2υthe
(σborehole
h − σH ) +wall
υσv +at (point
1 + υ)BPmand − (point
3σh − AσHrespectively,
) + 2r can (5)
be
E
determined by the Equations (5) and (6):
2r h
[ ]
i
C24 = 2(µrrA + r ) = 2r 2υ2 (2σH − σh ) + υσv + (1 + υ) Pm − (3σH − σh ) + 2r
C13 = 2( μrrB + r ) = E 2υ (σ h − σ H ) + υσ v + (1 + υ ) Pm − (3σ h − σ H ) + 2r
(6)
(5)
E
Energies 2018, 11, 131 4 of 13

Therefore, the relationship among the caliper measurements (longer axis C13, shorter axis C24),
geomechanical properties, and in-situ stress can be described as follows:

2r h 2 i
C13 = 2υ (σh − σH ) + υσv + (1 + υ) Pm − (3σh − σH ) + 2r (7)
E
2r h 2 i
C24 = 2υ (σH − σh ) + υσv + (1 + υ) Pm − (3σH − σh ) + 2r (8)
E

3. FR-GA-ANN Method
Theoretically, in a vertical well, if Young’s modulus, Poisson’s ratio, vertical stress, original
borehole size, borehole pressure, and two horizontal stresses are known, Equations (7) and (8) can
be applied to calculate the longer and shorter diameters of the elliptical borehole. In actual cases
of a drilled borehole, the longer and shorter diameters can be measured by four-arm-caliper tools.
Vertical stress can be calculated form overburden weight using density logs. Borehole pressure can be
estimated form mud weight. Young’s modulus, Poisson’s ratio, and two horizontal stresses are usually
unknowns. Although bit size is known in any drilling process, the borehole size is not supposed to be
equal to it, due to the factors such as lithology, erosion, damage and whirling of the bit. Therefore, for a
drilled hole at certain depth, with known vertical stress and mud weight, there will be five unknowns
(Young’s modulus, Poisson’s ratio, two horizontal stresses, and borehole size) and two knowns (longer
and shorter diameters measured form calipers logs) in Equations (7) and (8).
In order to solve the unknowns with the limited known parameters, we developed a method that
combines artificial neural network (ANN), genetic algorithm (GA), and probabilistic analysis.
The ANN model was first developed by McCullock and Pitts [22]. Since then, ANN models have
evolved and have been considered a modeling tool in finding patterns based on the characterized
relationship between the inputs and outputs. Once the relationship between inputs (e.g., geomechanical
parameters in this paper) and outputs (e.g., borehole deformations in this paper) from training samples
has been established, the ANN model forwards the relationships to GA for further estimation of in-situ
stress and geomechanical properties based on the desired wellbore displacements.
GA was introduced by Holland [23] as an abstraction of biological evolution. The method is
used to search and optimize parameters based on the model of natural selection and natural genetics.
Initially, the GA starts with values to the solution of a problem, known as the population. Inside the
population, there are several potential solutions called chromosomes that can change to converge
on an ideal solution. As these chromosomes evolve over time through several successful iterations,
known as generations, stronger chromosomes will be generated, which are evaluated by the objective
function (fitness). The chromosomes that have a stronger fitness are more likely to be selected in the
evolution process [24,25]. Once certain chromosomes have been selected, they become parents and are
combined together with other parents to produce new chromosomes for the next successive generation
through the genetic process.
However, because of the large number of unknown parameters, more than the number of
equations, there could be multiple combinations of in-situ stress, geomechanical parameters and
borehole size that all lead to the same borehole deformation. To solve this dilemma, we adopted the
probabilistic analysis for a number of realizations of the ANN-GA model. After each probabilistic
analysis, the obviously higher frequency value for certain parameters will be identified. Then the
specific values for these parameters will be chosen to be locked, and thus the total number of unknown
parameters will be reduced for the next-run of ANN-GA model. The process is repeated until satisfied
results are achieved with sufficient clarity. The final results will be further validated by finite element
method to verify the relationship revealed by the ANN-GA model. We name this repeated probability
analysis process as probabilistic recapitulation (PR). The work flow for the PR-ANN-GA method is
shown in Figure 2.
 Energies 2018, 11, 131 5 of 12

repeated
Energies 2018, probability
11, 131 analysis process as probabilistic recapitulation (PR). The work flow for the PR-
5 of 13
ANN-GA method is shown in Figure 2.

Figure 2. Workflow of the probabilistic recapitulation-genetic algorithm-artificial neural network (PR-

Figure 2. Workflow of the probabilistic recapitulation-genetic algorithm-artificial neural network
GA-ANN) method.
(PR-GA-ANN) method.

4. Case Study
4. Case Study
In this section, the proposed PR-GA-ANN method is demonstrated by a field study in West
In this southern
Virginia, section, the proposed PR-GA-ANN
Appalachian methoddata
Basin, USA. Drilling is demonstrated
and boreholeby a field study
geometry in WestofVirginia,
information MIP
southern Appalachian Basin, USA. Drilling data and borehole geometry information
3H vertical drilling borehole were used for the determination of geomechanical properties of MIP 3H vertical
and
drilling borehole
horizontal werestresses.
in-situ used for the determination of geomechanical properties and horizontal in-situ stresses.
AtAt depth
depth ofof7538.5
7538.5feet
feet(true
(truevertical
vertical depth
depth at 7536
7536 feet)
feet)ininLower
LowerMarcellus
Marcellusformation
formation of of
thethe MIP
MIP
3H3Hwell, drilling bit size is 8.75 inches, bore
well, drilling bit size is 8.75 inches, bore hole mud pressure while drilling is estimated to
mud pressure while drilling is estimated to be 35 MPa,be 35 MPa,
thethe
vertical stress
vertical stresscalculated
calculatedfrom
fromdensity
density log
log is 54 MPa,
MPa, the
themeasured
measuredlonger
longerborehole
borehole diameter
diameter C13C13
is is
8.91
8.91 inches,
inches, themeasured
the measuredshorter
shorter borehole
borehole diameter
diameterC24C24isis8.88 inches.
8.88 The
inches. data
The is listed
data in Table
is listed 1. 1.
in Table

Table1.1. Available
Table Available known parameters.
parameters.

Parameters Pm SigV Bit Size C13 C24

Parameters P SigV Bit Size C13 C24
Units m Mpa Mpa inch inch inch
Units Values Mpa 35 Mpa 54 inch
8.75 8.91inch8.88 inch
Values 35 54 8.75 8.91 8.88
The input training and testing data for ANN are calculated based on the parameters listed in
Table 2. Totally 1024 combinations were generated. Thus 1024 sets of longer and shorter diameters,
C13The
andinput
C24,training and testing
were calculated data for
through ANN are
analytical calculatedusing
calculation based on the parameters
Equations (7) and (8).listed in Table 2.
Totally 1024 combinations were generated. Thus 1024 sets of longer and shorter diameters, C13 and C24,
were calculated through
Tableanalytical calculation
2. Parameters using
used for the Equations
generation (7) and
of training (8).
and testing data.

Variable Item 2r υ E SigH Sigh

Table 2. Parameters used for the generation of training and testing data.
unit inch Gpa Mpa Mpa
start value 8.925 0.16 5 40 25
Variable Itemincrement2r 0.0875 υ 0.06 5E 5 SigH 5 Sigh
unit end value inch 9.1875 0.34 20
Gpa 55 Mpa 40 Mpa
start value 8.925 0.16 5 40 25
Once the ANNincrement
is trained, the 0.0875 0.06 modeling
inverse analysis 5 is used 5to characterize
5 the relationship
endoutput.
between input and value Then9.1875 0.34
fitness (objective 20
function), the55borehole 40
longer and shorter
diameters in this case, is established and GA is used as an optimization tool to search chromosomes
(solutions)
Once the from
ANN ais wide
trained,range of inputs
the inverse that modeling
analysis meet the is established objective function.
used to characterize The
the relationship
chromosomes
between thatoutput.
input and have a Then
stronger fitness
fitness are more
(objective likely to the
function), be selected
boreholeaslonger
the results. Each modeling
and shorter diameters
in this case, is established and GA is used as an optimization tool to search chromosomes (solutions)
from a wide range of inputs that meet the established objective function. The chromosomes that have a
stronger fitness are more likely to be selected as the results. Each modeling cycle runs 100 realizations for
Energies 2018, 11, 131 6 of 13
Energies
Energies2018,
2018,11,
11,131
131 66ofof12
12

cycle
cycleruns
probabilistic 100
100realizations
runsanalysis. The firstfor
realizations 100
for probabilistic
GA-ANN model
probabilistic analysis. The
The first
realizations
analysis. 100
firstfor GA-ANN
100five unknown
GA-ANN model realizations
input
model parameters
realizations forare
for
five
shown unknown
in Figure input
3. It parameters
can be seen are
that shown
except in
theFigure 3.
borehole It can
size be
to seen
bit that
size except
ratio, the
five unknown input parameters are shown in Figure 3. It can be seen that except the borehole size tothe borehole
rest four size to
parameters,
thebit
bit size
sizeratio,
Young’s the
therest
modulus,
ratio, four
fourparameters,
Poisson’s
rest the
theYoung’s
ratio, maximum
parameters, modulus,
horizontal
Young’s modulus, Poisson’s
stresses, ratio,
ratio,maximum
minimum
Poisson’s horizontalhorizontal
maximum stresses are
horizontal
stresses,
evenly
stresses, minimum
scattered without
minimum horizontal stresses
any obvious
horizontal are
high are
stresses evenly
frequency scattered
evenly values.
scattered without
In the any
histogram
without obvious high
high frequency
of the borehole
any obvious size to bit
frequency
values.
size ratio In the
(Figure histogram
4), an of
obvious the borehole
high size
frequency to bit size
indicates a ratio
value (Figure
about 4),
1.018.an obvious
The high
corresponding
values. In the histogram of the borehole size to bit size ratio (Figure 4), an obvious high frequency frequency
borehole
indicates
sizeindicates a value
is 8.9075ainches. about 1.018. The corresponding borehole size is 8.9075
value about 1.018. The corresponding borehole size is 8.9075 inches. inches.

Figure
Figure 3. 3.
Figure 3.Results
Results ofof
Results hole/bit
hole/bit
of size
hole/bitsize ratio,
sizeratio, Poisson’s
Poisson’sratio,
ratio,Poisson’s ratio,Young’s
ratio, Young’smodulus,
Young’s modulus,and
modulus, andhorizontal
and stresses
horizontal
horizontal of
stresses
stresses ofof
100
100100ANN-GA
ANN-GA
ANN-GA realizations.
realizations.
realizations.

Figure
Figure4.4.Histogram
Histogramof
ofhole/bit
hole/bitratio.
ratio.
Figure 4. Histogram of hole/bit ratio.
Therefore,
Therefore, aahole
hole size
sizeof 8.9075 inches was then
thenapplied
appliedas asaaaknown factor for the re-calculation
Therefore, a hole size ofof8.9075
8.9075inches
incheswas
wasthen applied as known
knownfactor
factorfor
forthe
there-calculation
re-calculation
of
of the
the training
training data
data with
with the
the ranges
ranges of
of Young’s
Young’s modulus,
modulus, Poisson’s
Poisson’s ratio,
ratio, maximum
maximum horizontal
horizontal
of the training data with the ranges of Young’s modulus, Poisson’s ratio, maximum horizontal stresses
stresses
stresses and
and minimum
minimum horizontal
horizontal stresses.
stresses. The
The results
resultsof
of another
another 100
100realizations
realizations of
ofGA-ANN
GA-ANN model
model
and minimum
with horizontal stresses. The results of another 100 realizations of GA-ANN model with
with four
four unknown
unknown parameters
parameters isis shown
shown inin Figure
Figure 5.5. When
When compared
compared with with the
the five
five parameters
parameters
four unknown
results parameters is shown in Figure 5. When compared with the five parameters results in
resultsininFigure
Figure3,3,the
thevalue
valuerange
rangeof ofeach
eachparameters
parametersbecome
becomeaabit bitnarrower;
narrower;however,
however,the thevalues
values
Figure 3, the value range of each parameters become a bit narrower; however, the values are still quite
 Energies 2018, 11, 131 7 of 12
Energies 2018, 11, 131 7 of 13
Energies 2018, 11, 131 7 of 12
are still quite scattered without showing any constrained value. The histogram of these four
are still quite
parameters scattered
are shown without
in Figure 6. Weshowing
chose anyhighest
the constrained value.
frequency The histogram of these atfour
scattered without showing any constrained value. The histogram ofvalue
these in theparameters
four Poisson’s ratio
are shown 0.285in
and the highest frequency value of minimum horizontal stress at 34.2 MPa. Thereafter, the0.285
parameters are shown in Figure 6. We chose the highest frequency value in the Poisson’s ratio at input
Figure 6. Wehighest
chose the
and the parameters
highestvalue
frequency
frequency value inhorizontal
of minimum
the Poisson’s ratio 34.2
at 0.285
stress atmaximum
and the highest
MPa.horizontal
frequency
Thereafter,stresses.
the input
unknown can be reduced to three: Young’s modulus, The
value of minimum
unknown horizontal
parameters can be stress
reduced at to
34.2 MPa.
three: Thereafter,
Young’s modulus,themaximum
input unknown parameters
horizontal stresses. can be
The
results of 100 new GA-ANN modeling realizations for the reduced two unknown parameters are
reduced
resultstoofthree: Young’s modulus, maximum horizontal stresses. The results of 100 new GA-ANN
100 new
shown in Figure 7. GA-ANN modeling realizations for the reduced two unknown parameters are
modeling
shown in Figure 7. for the reduced two unknown parameters are shown in Figure 7.
realizations

Figure
Figure 5. Results
5. Results of of Poisson’sratio,
Poisson’s ratio,Young’s
Young’s modulus,
modulus, Maximum
Maximumhorizontal
horizontalstresses, and
stresses, horizontal
and horizontal
Figure 5. Results of Poisson’s ratio, Young’s modulus, Maximum horizontal stresses, and horizontal
stresses of 100 ANN-GA realizations.
stresses of 100 ANN-GA realizations.
stresses of 100 ANN-GA realizations.

Figure 6. Histogram of Poisson’s ratio, Young’s modulus, minimum horizontal stress and maximum
horizontal stress of 100 realizations of four parameters GA-ANN.
Figure6.6.Histogram
Figure Histogram ofof Poisson’s
Poisson’s ratio,
ratio, Young’s
Young’s modulus,
modulus, minimum
minimum horizontal
horizontalstress
stressand
andmaximum
maximum
horizontal stress of 100 realizations of four parameters GA-ANN.
horizontal stress of 100 realizations of four parameters GA-ANN.
Energies 2018, 11, 131 8 of 13
Energies 2018,
Energies 11,11,
2018, 131131 8 of812of 12

Figure Figure 7. Results

7. Results of Young’s
of Young’s modulusand
modulus andmaximum
maximum horizontal
horizontalstress of of
stress 100100realizations of GA-ANN
realizations of GA-ANN
model
Figure 7.with
model with fixed. fixed.
Results of Young’s modulus and maximum horizontal stress of 100 realizations of GA-ANN
model with fixed.
It can It can obviously
obviously be be seenfrom
seen fromFigure
Figure 77 that
that thethe calculated
calculatedresults
results of ofYoung’s
Young’s modulus
modulus and and
maximum
maximum It can horizontal
obviously
horizontal stresses
be seen
stresses are
arefrom quite consistent
quiteFigure in
7 thatinthe
consistent multiple realizations.
calculated
multiple The
results ofThe
realizations. average
Young’s
averagevaluevaluethe
for
modulus andthe
for
Yong’s modulus is 17.1 GPa; the average value for the maximum horizontal stress is 49.8 MPa.
Yong’s modulus is 17.1 GPa; the average value for the maximum horizontal stress is 49.8 MPa. the
maximum horizontal stresses are quite consistent in multiple realizations. The average value for
Yong’s Therefore,
modulus in the case
is 17.1 study of MIP 3H well, which was drilledhorizontal
in the lowerstress Marcellus
is 49.8formation
Therefore, in the caseGPa;
studytheofaverage
MIP 3Hvaluewell,for the maximum
which was drilled in the lower Marcellus MPa. formation
with 8.75
Therefore,inches in bit
the in the
case Appalachian
study of MIP Basin,
3H well,if borehole
which mud
was pressure
drilled in theis 35 MPa,
lower vertical
Marcellus stress is
formation
with 548.75
MPa,inches bit in the borehole
the estimated Appalachian Basin,be
size should if 8.9075
borehole mudIt pressure
inches. was determined is 35 MPa, vertical stress is
by PR-GA-ANN
54 with
MPa, 8.75
methodtheinches
estimated
that
bit inborehole
the maximum
the Appalachian
size should
horizontal
Basin, if8.9075
in-situ be
borehole mud pressure
stress is 49 inches.
MPa, theItminimum
is 35 MPa, vertical
was determined horizontal by
stress is
PR-GA-ANN
in-situ stress
54 MPa, the estimated borehole size should be 8.9075 inches. It was determined by PR-GA-ANN
method thatMPa,
is 34.2 the themaximum horizontal
Yong’s modulus in-situ
is 17.1 GPa,stress
and the is 49 MPa, the
Poisson’s minimum
ratio is 0.285. horizontal in-situ stress is
method that the maximum horizontal in-situ stress is 49 MPa, the minimum horizontal in-situ stress
34.2 MPa,Inthe Yong’s
order modulus
to verify eachisANN17.1 GPa,
model, andthethe Poisson’s ratio
mean-square erroris(MSE)
0.285.is used to evaluate the
is 34.2 MPa, the Yong’s modulus is 17.1 GPa, and the Poisson’s ratio is 0.285.
convergence
In order to of ANN.each
verify FigureANN 8 shows the MSE
model, the variations
mean-square for training,
error (MSE)validation, and testing
is used with the
to evaluate
In order
iterations. The tobest
verify each ANN model,isthe
validation mean-square error
shows(MSE) is used to evaluate the
convergence of ANN. Figure 8 performance
shows the MSE 2.59variations
× 10−3. Figure
for9 training, the validation,
linear regression
andbetween
testing with
convergence
the modeled of ANN.
output Figure 8 shows
and the corresponding the MSE variations
target. The for training,
regression
−3 . Figurevalues validation,
are shown and
on testing
the top with
of
iterations.
iterations.
TheThe
best validation
best validation
performance
performance
is 2.59 × Figure
10 9the
shows the linear regression
pictures of training, testing, validation, andisall2.59 × 10−3.respectively.
samples 9 showsAll of linear
the regression
regression valuesbetween
are
between the modeled
thegreater
modeled output theand the corresponding target. The regression values areonshown onofthe
thanoutput and meant
0.99, which corresponding target. The
very good performance. regression values are shown the top
toppictures
of pictures of training,
of training, testing,
testing, validation,
validation, and
and all all samples
samples respectively.
respectively. All ofAll
theof the regression
regression valuesvalues
are
aregreater
greaterthan
than0.99,
0.99,which
which meant very good performance.
meant very good performance.

Figure 8. Variation of the mean square errors.

Figure8.8.Variation
Figure Variationof
of the
the mean
mean square
square errors.
errors.
Energies 2018, 11, 131 9 of 13
Energies 2018, 11, 131 9 of 12

Figure
Figure 9.
9. Performances
Performances of
of linear
linear regressions.
regressions.

5. Discussion
5. Discussion

5.1.
5.1. Comparison
Comparison with
with Field
Field Data
Data
Table
Table 33showsshows the the comparison
comparison of the of the PR-GA-ANN
PR-GA-ANN based inverse based inverse
analysis analysis
results results of
of geomechanical
geomechanical properties in-situ
properties and horizontal and horizontal in-situ
stresses with thestresses with the field
field observation observation
or reports of the or reports
area for theofwell
the
area for the well MIP 3H. The basin wide stress study indicated σh/σν value of up to 0.7 in the
MIP 3H. The basin wide stress study indicated σh /σν value of up to 0.7 in the corresponding depth
corresponding depth[26].
at this well location at this
With well
thelocation
vertical [26].
stressWith
of 54the MPa vertical
from Table stress1,ofthe54upper
MPa frombound Table 1, the
minimum
upper bound minimum horizontal stress will be 37.8 MPa. The estimate
horizontal stress will be 37.8 MPa. The estimate by PR-GA-ANN in terms of minimum horizontal by PR-GA-ANN in terms of
minimum horizontal in-situ stress
in-situ stress is consistent with this study. is consistent with this study.
Another sourcefor
Another source forverification
verification of the
of the minimum
minimum horizontal
horizontal in-situin-situ
stress bystress by PR-GA-ANN
PR-GA-ANN method
method is from hydraulic fracture treatment records. The first stage
is from hydraulic fracture treatment records. The first stage of the multi-stage hydraulic fractureof the multi-stage hydraulic
fracture
treatment treatment in this formation
in this formation of the 3Hof the
well3Hreported
well reported a pressure
a pressure of 34.8of MPa
34.8 MPa
that that
is 15ismin
15 min
afterafter
the
the instantly-shut-in-pressure (ISIP). The pressure is considered close to the
instantly-shut-in-pressure (ISIP). The pressure is considered close to the fracture closure pressure, fracture closure pressure,
which
which isis representative
representativeof ofthe
thesmallest
smallestprinciple
principle in-situ
in-situ stress,
stress, i.e.,i.e.,
thethe minimum
minimum horizontal
horizontal stress
stress in
in this
this
case.case.
In thisIn study,
this study, the estimated
the estimated minimum minimum
horizontal horizontal
stress ofstress
34.2 MPa of 34.2
is in MPa is in reasonable
reasonable agreement
agreement with the field
with the field observations. observations.
Evans
Evans alsoalso stated
stated inin the
the Appalachian
Appalachian StressStress Study
Study Report
Report thatthat the magnitude of
the magnitude of σσH varies from
H varies from
high
high ininthe
thenorthern
northernpart partofof
thethe
basin to low
basin values
to low in the
values in south; the stress
the south; state in
the stress the in
state Devonian shale,
the Devonian
of which Middlesex formation is a part, is either strike slip or normal fault
shale, of which Middlesex formation is a part, is either strike slip or normal fault regime due to the regime due to the pinch-
out of the of
pinch-out underlying salt [26].
the underlying The location
salt [26]. of the
The location studied
of the well
studied wellis around
is around the
thepinch-out
pinch-outareaarea[27].
[27].
Therefore, unit value of σ H/σν stress ratio should be an upper limit. That means the maximum
Therefore, unit value of σH /σν stress ratio should be an upper limit. That means the maximum
horizontal
horizontal stress
stress should
should be be smaller
smaller than
than 5454 MPa,
MPa, thethe vertical
vertical stress.
stress. ThisThis also
also validates
validates the
the maximum
maximum
horizontal
horizontal in-situ
in-situ stress
stress estimated
estimated by by the
the proposed
proposed PR-GA-ANN
PR-GA-ANN method. method.

Table 3. Comparison of PR-GA-ANN back calculated rock mechanical properties and in-situ
horizontal stresses with field observation in Lower Marcellus formation.

PR-GA-ANN Method Field Observation

E (Gpa) ν σH (MPa) σh (MPa) E (GPa) ν σH (MPa) σh (MPa)
17.1 0.285 49.8 34.2 To be tested To be tested <54 34.8 (15 min after ISIP)
Energies 2018, 11, 131 10 of 13

Table 3. Comparison of PR-GA-ANN back calculated rock mechanical properties and in-situ horizontal
stresses with field observation in Lower Marcellus formation.

PR-GA-ANN Method Field Observation

E (Gpa) ν σH (MPa) σh (MPa) E (GPa) ν σH (MPa) σh (MPa)
Energies
17.1 2018, 11, 131
0.285 49.8 34.2 To be tested To be tested <54 10 of 12
34.8 (15 min after ISIP)

5.2. Validation by Finite Element Method

5.2. Validation by Finite Element Method
For a vertical borehole discussed in this paper, the deformation can also be forward calculated
For a vertical borehole discussed in this paper, the deformation can also be forward calculated by
by finite element modeling [28] if the rock mechanical properties and in-situ stresses values are
finite element modeling [28] if the rock mechanical properties and in-situ stresses values are available.
available. In order to further validate the calculation, we used the calculated geomechanical
In order to further validate the calculation, we used the calculated geomechanical properties and the
properties and the horizontal in-situ stresses to forwardly calculate the borehole deformation using
horizontal in-situ stresses to forwardly calculate the borehole deformation using finite element method.
finite element method. Figure 10 shows the mesh of the finite element model. The radius of the
Figure 10 shows the mesh of the finite element model. The radius of the borehole used the half of the
borehole used the half of the calculated original boreholes diameter, which is 4.46 inches. Parameters
calculated original boreholes diameter, which is 4.46 inches. Parameters used for the finite element
used for the finite element model, which are determined from the proposed PR-GA-ANN method,
model, which are determined from the proposed PR-GA-ANN method, are listed in Table 4.
are listed in Table 4.
Table 4. In-situ stress, rock mechanical properties, and hole size.
Table 4. In-situ stress, rock mechanical properties, and hole size.
ParameterParameter Value Value
Young’s
Young’s Modulus, Modulus,
E (GPa) (GPa) 17.1 17.1
PoissonPoisson
Ratio, ν Ratio, ν 0.285 0.285
RadiusRadius
of borehole, r (in)
of borehole, (in) 4.454 4.454
Maximum principal
Maximum stress, σstress,
principal H (MPa) (MPa) 49.8 49.8
Minimum principal
Minimum stress, σstress,
principal h (MPa) (MPa) 34.2 34.2
VerticalVertical
stress, σstress,
v (MPa) (MPa) 54 54
Mud pressure, Pm (MPa) 35
Mud pressure, (MPa) 35

Figure 10. Mesh of the finite element model.

Figure 10. Mesh of the finite element model.

Using the
Using theestimated
estimated 8.9075
8.9075 inches
inches as original
as the the original borehole
borehole diameter,diameter, the calculated
the calculated shorter
shorter borehole
borehole diameter
diameter (C24) is inches
(C24) is 8.880069 8.880069
forinches for the
the shorter shorter
axis of theaxis of theborehole;
elliptical ellipticalwhile
borehole; while the
the calculated
calculated longer diameter (C13) is 8.910139 inches. The ratio of the calculated shorter
longer diameter (C13) is 8.910139 inches. The ratio of the calculated shorter diameter and longer diameter and
longer diameter is 0.996625. The finite element modeling results are listed in
diameter is 0.996625. The finite element modeling results are listed in the Table 5. When comparedthe Table 5. When
compared
with with the shorter
the measured measured andshorter
longerand longer reported
diameters diametersfrom
reported from thecaliper
the four-arm four-arm caliper
data, which data,
are
which are 0.88 inches and 8.91 inches respectively, the deference is minor and the
0.88 inches and 8.91 inches respectively, the deference is minor and the errors are very small (around errors are very
small (around 10−4 to 10−5). The relationship that the PR-GA-ANN method revealed is in consistent
with the calculations from the finite element method.

Table 5. Calculated deformation from the estimated parameters using FEM.

Measured Finite Element Method Errors

C24 C13 C24 C13 C24 C13
Energies 2018, 11, 131 11 of 13

10−4 to 10−5 ). The relationship that the PR-GA-ANN method revealed is in consistent with the
calculations from the finite element method.

Table 5. Calculated deformation from the estimated parameters using FEM.

Measured Finite Element Method Errors

C24 C13 C24 C13 C24 C13
C24/C13 C24/C13 C24/C13
(inches) (inches) (inches) (inches) (inches) (inches)
8.88 8.91 0.996633 8.880069 8.910139 0.996625 0.00007 0.0001 −0.00001

6. Conclusions
It is possible to estimate geomechanical properties and horizontal in-situ stresses from borehole
deformation data. Uncertainties of results can be reduced by combining probabilistic analysis with
GA-ANN method. Since there is no direct stress measurement and core mechanical test data for this
well to calibrate the calculated results, future filed measurement or lab test work might be needed to
further justify the methodology.
The correct estimate of borehole size is an important factor influencing the magnitude of borehole
displacement and must be addressed different from bit size. This paper treats the original borehole
size as an unknown input and estimate its value by inverse calculation from deformation data.
However, the relationship between the bit size and the borehole size is still unknown. Future lab
and/or field tests will be required to further validated the relationship.

Acknowledgments: The support of the United States Department of Energy (DE-FE0026825, UCFER-University
Coalition for Fossil Energy Research) is greatly acknowledged. Authors are grateful to the US Department of
Energy and West Virginia University for supplying the field data. Special thanks are given to Bezalel Haimson at
University of Wisconsin who has provided constructive suggestions.
Author Contributions: Hong Xue Han and Shunde Yin conceived and designed the numerical experiments;
Hong Xue Han performed the experiments; Hong Xue Han and Shunde Yin analyzed the data and wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest.

Nomenclature
µrr Displacement of the borehole at R distance from the borehole center
µrrA Displacement at the borehole wall along the maximum horizontal stress direction
µrrB Displacement at the borehole wall along the minimum horizontal stress direction
C13 Longer axis of the elliptical borehole
C24 Shorter axis of the elliptical borehole
E Young’s Modulus
ν Poisson’s Ratio
Pm Borehole pressure
σv Vertical stress
σH Maximum horizontal stress
σh Minimum horizontal stress
R Distance from borehole center
r Original borehole radius
θ Angle from maximum horizontal stress direction

References
1. Haimson, B.C. A Simple Method for Estimating In Situ Stresses as Great Depth; Field Testing and Instrumentation
of Rock, STP32150S; Clark, G., Ed.; ASTM International: West Conshohocken, PA, USA, 1974; pp. 156–182.
2. Haimson, B.C.; Fairhurst, C. Initiation and Extension of Hydraulic Fractures in Rock. Soc. Pet. Eng. J. 1967,
7, 310–318. [CrossRef]
3. Anderson, E.M. The Dynamics of Faulting. Trans. Edinb. Geol. Soc. 1905, 83, 387–402. [CrossRef]
Energies 2018, 11, 131 12 of 13

4. Kirsch, E.G. Die Theorie der Elastizität und Die Bedürfnisse der Festigkeitslehre; Zeitschrift des Vereines deutscher
Ingenieure; Springer: New York, NY, USA, 1898; Volume 42, pp. 797–807.
5. Zoback, M.D.; Moos, D.; Mastin, L.; Anderson, R.N. Wellbore Breakouts and In-situ Stress. J. Geophys. Res.
1985, 90, 5523–5530. [CrossRef]
6. Peska, P.; Zoback, M.D. Compressive and Tensile Failure of Inclined Wellbores and Determination of In-situ
Stress and Rock Strength. J. Geophys. Res. 1995, 100, 12791–12811. [CrossRef]
7. Ervine, W.B.; Bell, J.S. Subsurface In Situ Stress Magnitudes from Oil-well Drilling Records: An Example
from the Venture Area, offshore eastern Canada. Can. J. Earth Sci. 1987, 24, 1748–1759. [CrossRef]
8. Cornet, F.H.; Valette, B. In Situ Stress Determination from Hydraulic Injection Test Data. J. Geophys. Res.
1984, 89, 11527–11537. [CrossRef]
9. Chang, C.; Zoback, M.D.; Khaksar, A. Empirical Relations between Rock Strength and Physical Properties in
Sedimentary Rocks. J. Pet. Sci. Eng. 2006, 51, 223–237. [CrossRef]
10. Abdulraheem, A.; Ahmed, M.; Vantala, A.; Parvez, T. Prediction of Rock Mechanical Parameters for
Hydrocarbon Reservoirs Using Different Artificial Intelligence Techniques. Presented at the 2009 SPE
Saudi Arabia Section Technical Symposium and Exhibition, Alkhobar, Saudi Arabia, 9–11 May 2009.
Paper SPE126094.
11. Najibi, A.R.; Ghafoori, M.; Lashkaripour, G.R.; Asef, M.R. Reservoir Geomechanical Modeling: In-Situ Stress,
Pore Pressure, and Mud Design. J. Pet. Sci. Eng. 2017, 151, 31–39. [CrossRef]
12. Aadnoy, B.S. Inversion Technique to Determine the In-Situ Stress Field from Fracturing Data. J. Pet. Sci. Eng.
1990, 4, 127–141. [CrossRef]
13. Zhao, Y.; Feng, Z.; Xi, B.; Wan, Z.; Yang, D.; Liang, W. Deformation and Instability Failure of Borehole at High
Temperature and High Pressure in Hot Dry Rock Exploitation. Renew. Energy 2015, 77, 159–165. [CrossRef]
14. Liu, J.; Sun, H.; Hu, Q. Surface Borehole Synthesis Tension Deformation Fracture Time-Space Rule. Int. J.
Min. Sci. Technol. 2012, 22, 465–470. [CrossRef]
15. Lonardelli, J.N.; Silva, R.O.; Falcao, F.O.L.; Santos, M.A.C.; Abreu, C.E.B.S. Evaluation of Oil Production
Related Effects through Geomechanical Modeling: A Case Study from Marimbá Field, Campos Basin, Brazil.
J. Pet. Sci. Eng. 2017, 158, 186–201. [CrossRef]
16. Newell, P.; Yoon, H.; Martinez, M.J.; Bishop, J.E.; Bryant, S.L. Investigation of the Influence of Geomechanical
and Hydrogeological Properties on Surface Uplift at In Salah. J. Pet. Sci. Eng. 2017, 155, 34–45. [CrossRef]
17. Zhang, S.; Yin, S. Determination of Horizontal In-Situ Stresses and Natural Fracture Properties from Wellbore
Deformation. Int. J. Oil Gas Coal Technol. 2014, 7, 1–28. [CrossRef]
18. Zhang, S.; Yin, S. Determination of Earth Stresses Using Inverse Analysis Based on Coupled Numerical
Modeling and Soft Computing. Int. J. Comput. Appl. Technol. 2015, 52, 18–28. [CrossRef]
19. Zhang, S.; Yin, S. Determination of In Situ Stresses and Elastic Parameters from Hydraulic Fracturing Tests
by Geomechanics Modeling and Soft Computing. J. Pet. Sci. Eng. 2014, 124, 484–492. [CrossRef]
20. Zhang, H.; Yin, S.; Aadnoy, B.S. Poroelastic Modeling of Borehole Breakouts for In-Situ Stress Determination
by Finite Element Method. J. Pet. Sci. Eng. 2017, in press. [CrossRef]
21. Sadd, M.H. Elasticity Theory, Applications, and Numerics, 3rd ed.; Academic Press: Cambridge, MA, USA,
2014; ISBN 978-0-12-408136-9.
22. McCulloch, W.S.; Pitts, W.H. A Logical Calculus of the Ideas Immanent in Nervous Activity. Bull. Math.
Biophys. 1943, 5, 115–133. [CrossRef]
23. Holland, J.H. Adaptation in Natural and Artificial Systems, 1st ed.; University of Michigan Press: Ann Arbor,
MI, USA, 1975.
24. Goldberg, D.E. Genetic Algorithms in Search, Optimization, and Machine Learning; Addison-Wesley: Reading,
MA, USA, 1989.
25. Holland, J.H. Adaptation in Natural and Artificial Systems, 2nd ed.; MIT Press: Cambridge, MA, USA, 1992.
26. Evans, K.F. Appalachian Stress Study. J. Geophys. Res. 1989, 94, 17619–17645. [CrossRef]
Energies 2018, 11, 131 13 of 13

27. Pierce, W.G.; Rich, E.I. Summary of Rock Salt Deposits in the United States as Possible Storage Sites for Radioactive
Waste Materials; US Government Printing Office: Washington, DC, USA, 1962.
28. Yin, S.; Towler, B.F.; Dusseault, M.B.; Rothenburg, L. Fully coupled THMC modeling of wellbore stability
with thermal and solute convection considered. Transp. Porous Media 2010, 84, 773–798. [CrossRef]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).