You are on page 1of 13

This article was downloaded by: [Aston University]

On: 25 January 2014, At: 19:27


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Petroleum Science and Technology


Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/lpet20

A New Method for Determining


Bourgoyne and Young Penetration Rate
Model Constants
a a a
H. Rahimzadeh , M. Mostofi & A. Hashemi
a
Petroleum Engineering Department , Petroleum University of
Technology , Tehran, Iran
Published online: 22 Mar 2011.

To cite this article: H. Rahimzadeh , M. Mostofi & A. Hashemi (2011) A New Method for Determining
Bourgoyne and Young Penetration Rate Model Constants, Petroleum Science and Technology, 29:9,
886-897, DOI: 10.1080/10916460903452009

To link to this article: http://dx.doi.org/10.1080/10916460903452009

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Petroleum Science and Technology, 29:886–897, 2011
Copyright © Taylor & Francis Group, LLC
ISSN: 1091-6466 print/1532-2459 online
DOI: 10.1080/10916460903452009

A New Method for Determining Bourgoyne and


Young Penetration Rate Model Constants

H. RAHIMZADEH,1 M. MOSTOFI,1 AND A. HASHEMI1


1
Petroleum Engineering Department, Petroleum University of Technology,
Tehran, Iran

Abstract The objective of any drilling operation is to drill a well with the least
Downloaded by [Aston University] at 19:27 25 January 2014

possible cost and in the shortest time in compliance with safe operation. This objective
is usually achieved by applying the drilling parameters in such a way that the
fastest drilling rate would result. Several drilling penetration rate models have been
proposed for estimation of penetration rate based on different drilling parameters.
One of the most frequently used models for estimation of drilling penetration rate is
the Bourgoyne and Young model. This model relates the penetration rate to several
drilling parameters. There are eight unknown constants in this model. Bourgoyne
and Young have proposed multiple regression analysis for obtaining these constants.
Because the constant values obtained by multiple regression analysis are sometimes
meaningless and are not in the recommended ranges, other methods for determining
these coefficients are suggested. Some authors have used least squares data fitting
methods and trust region approaches to obtain these coefficients. In this article,
a new method is used for this purpose. A set of possible answers is chosen from
the recommended bounds. Then the best fitted answers are selected and compared
with a new set of possible answers. This trial is repeated until the desired answers
are obtained. The results show the proficiency of the new methodology to determine
constant coefficients of the Bourgoyne and Young penetration rate model.

Keywords Bourgoyne and Young penetration rate model, drilling optimization,


penetration rate, progressive stochastic method

Introduction
One of the main issues for any drilling engineer is to drill a well in the shortest time and
at the lowest cost. The concept of time taken for any drilling operation can be stated in
terms of the drilling rate of penetration. Therefore, in order to drill a well in the shortest
time, a good estimation of the penetration rate and knowledge of how different parameters
affect it are required. To achieve this objective, a drilling model should be used. A drilling
model is the relation between the rate of penetration and different parameters that have
significant effects on it. These relationships are usually stated in the form of an equation.
Penetration rate is affected by many parameters such as bit hydraulics, weight on bit,
rotary speed, bit type, mud properties, formation characteristics, etc. (Akgun, 2007). There
is no exact mathematical relation between drilling rate and different drilling variables
because not only do a large number of uncertain drilling variables influence the drilling

Address correspondence to Hedayat Rahimzadeh, Petroleum Engineering Department, Pe-


troleum University of Technology, Khosro Jonoobi Ave., Sattarkhan St., Tehran, Iran. E-mail:
hedayat.rahimzadeh@gmail.com

886
Bourgoyne and Young Model Constants 887

rate, but their relationship is nonlinear and complex (Ricardo et al., 2007). However,
during recent years, several drilling models have been proposed for the rate of penetration,
among which the Bourgoyne and Young model (Bourgoyne and Young, 1974) is the most
well known. This model consists of eight parameters and eight constant coefficients.
The constant coefficients should be determined in order to have a good estimation
of the penetration rate. Because the accuracy of this estimation depends on how the
constant coefficients are computed, different methods have been used to achieve more
accurate results. The coefficients are determined using previous drilling experience. In
order to determine the unknown constants, Bourgoyne and Young suggested the multiple
regression technique (Bourgoyne and Young, 1974). This technique is limited to the
number of data points. In addition, in some cases, multiple regression analysis gives
physically meaningless result.
Some authors have examined the use of mathematical methods to determine the
Bourgoyne and Young constant coefficients. For example, nonlinear least squares data
fitting with a trust region method have been applied to determine these constants (Bahari
Downloaded by [Aston University] at 19:27 25 January 2014

and Baradaran Seyed, 2007). Although this method has overcome the problem of giving
meaningless results by restricting answers with lower and upper bounds, it does not result
in a sufficiently accurate estimation of penetration rate.
In this research, first a set of possible answers is chosen from the recommended
bounds. This is done by defining an auxiliary variable that can translate the random
numbers taken from the interval of [0 1] to the constant parameter values. Then, best
fitted results are selected to be transmitted to the next set of answers. In each iteration,
the constants are chosen from a more restricted interval. The iteration is continued until
the desired results are obtained. In addition to reaching meaningful results, this method
leads to a more accurate model for penetration rate.

Bourgoyne and Young Drilling Rate Model


Bourgoyne et al. (2003) proposed the following equation to model the drilling rate of
penetration when using roller cone bits:

ROP D f1  f2  f3  f4  f5  f6  f7  f8 (1)

f1 , f2 , f3 etc., represent the functional relations between penetration rate and different
drilling variables. The functional relations chosen usually are based on trends observed
either in the laboratory or during field studies (Bourgoyne et al., 2003). These relations
are stated as follows:

f1 D e 2:303a1 D K (2)

f 2 D e 2:303a2.10000 D/
(3)
0:69 .g
f3 D e 2:303a3D p 9/
(4)

f4 D e 2:303a4D.gp c / (5)
  3
W W
2
6 db db 7
f5 D 6   t7 (6)
4 W 5
4
db t
888 H. Rahimzadeh et al.
 a6
N
f6 D (7)
60

f7 D e a7 h (8)

Fj a8
 
f8 D (9)
1000
The function f1 represents the effect of formation strength, bit type, mud type,
and solid content, which are not included in the drilling model. This term is usually
called formation drillability. The functions f2 and f3 model the effect of compaction
on penetration rate. The function f4 takes into account the effect of overbalance on
penetration rate. The functions f5 and f6 model the effect of bit weight and rotary speed
on penetration rate, respectively. The function f7 implies the effect of bit tooth wear on
penetration rate. The function f8 models the effect of bit hydraulics on penetration rate.
Downloaded by [Aston University] at 19:27 25 January 2014

Jet impact force was chosen as the hydraulic parameter of interest, with a normalized
value of 1.0 for f8 at 1,000 lbf.
As can be seen from Eqs. (2) to (9), the Bourgoyne and Young model consists of eight
constant parameters (a1 to a8 ) in addition to the variable parameters. These constants can
be determined for each formation using the drilling data taken from the previous drilling
records of the field. Bourgoyne et al. (2003) suggested individual bounds for each of
the constants a1 to a8 , based on the reported ranges for the coefficients from various
formations in different areas to achieve meaningful results. These bounds are shown in
Table 1.

SP Field
SP field is located on the Qatar-Fars Arch, one of the major structural elements of the
Central Persian Gulf Area. The Persian Gulf is part of what is referred to in the plate
tectonic literature as the Arabian Plate or the Middle East Sedimentary Basin, which
is approximately 3,000 km in length and 2,000 km in width. The Kangan and Dalan
formations are the gas-bearing horizons of this field. Figure 1 illustrates the stratigraphy

Table 1
Recommended bounds for coefficients
to a1 to a8

Lower Upper
Coefficient bound bound

a1 0.5 1.9
a2 0.000001 0.0005
a3 0.000001 0.0009
a4 0.000001 0.0001
a5 0.5 2
a6 0.4 1
a7 0.3 1.5
a8 0.3 0.6
Data from Bourgoyne et al. (2003).
Bourgoyne and Young Model Constants 889
Downloaded by [Aston University] at 19:27 25 January 2014

Figure 1. Stratigraphy column and geological description of the formations for a typical well in
South Pars gas field.

column and geological description of each formation for a typical well in this field.
Because the most time-consuming and problematic section of the drilling in this field is
the 12 1/4-in. hole, three formations in this section were chosen for this study. These
formations are Hith, Surmeh, and Upper Dashtak. The Bourgoyne and Young model
coefficients and by which the model itself have been obtained for the three mentioned
formations. Drilling data from two wells in the SP field (namely wells A and B) were
used for this purpose. Well A was used to determine the constant coefficients and well
B was used to verify the validity of the results obtained in well A.
890 H. Rahimzadeh et al.

Determining Bourgoyne and Young Coefficients Using a


Progressive Stochastic Method
Bourgoyne and Young applied multiple regression techniques to determine the constants
a1 to a8 . Because no bound can be defined for the results using this method, it may
lead to meaningless results. For example, in some cases the bit weight constant would
result in a negative value, which means by increasing weight on the bit, a reduction in
penetration rate would be experienced.
During recent years, attempts to determine Bourgoyne and Young equation constants
showed significant improvement in the accuracy and validity of the results of this model.
For instance, the trust region approach was used in 2007 by Bahari and Baradaran Seyed.
In this method the recommended bounds were used to avoid obtaining the meaningless
results. However, using the method described here will lead to more accurate penetration
rate prediction in comparison with the trust region approach.
In this method, a1 through a8 are chosen randomly so that they can be bounded by
Downloaded by [Aston University] at 19:27 25 January 2014

lower and upper values. ai is limited by the lower bound bi and upper bound ci :

bi  ai  ci

Figure 2. Comparing actual and estimated penetration rate with three methods for the Hith
formation.
Bourgoyne and Young Model Constants 891

Subtracting bi from all sides of the inequalities results in:

0  ai bi  c i bi

Dividing by ci bi yields:
ai bi
0 1
ci bi
ai bi
The term ci bi
, which is always between 0 and 1, is called ki :

ai bi
ki D (10)
ci bi
Because ki can be any random number between 0 and 1, it can be simply produced by
computer any time it is required.
Downloaded by [Aston University] at 19:27 25 January 2014

In order to compute ai values, Eq. (10) should be rephrased as:

ai D ki .ci bi / C bi (11)

A group of a1 to a8 sets is computed using this method and then the penetration
rate is estimated based on the Bourgoyne and Young model using each set of a1 to a8 .
The coefficients that result in the best estimation of penetration rate are taken as one of
the answers for the next step. Based on this best-fitted set of coefficients, the lower and

Figure 3. Comparing actual and estimated penetration rate with three methods for the Surmeh
formation.
892 H. Rahimzadeh et al.

upper bounds are modified and more restricted. Therefore, new bounds are defined for
the next step.
The same procedure as above is followed for the next step. The only difference is
that the results are compared with the best answers from the previous step. This will be
continued until the desired constant coefficients that lead to the minimum possible error
in the estimation of penetration rate result.

Results and Discussion


The constants a1 to a8 were computed for three formations of the SP field using multiple
regression techniques, the trust region approach, and a new method called the progressive
stochastic method. Table 2 shows the constants obtained using each method. As can be
seen from this table, application of multiple regression analysis may result in negative or
zero values, which are physically meaningless. In contrast, the constant coefficients com-
Downloaded by [Aston University] at 19:27 25 January 2014

puted by the two other methods are always meaningful and in the recommended ranges.
Figures 2 to 4 illustrate the predicted and actual penetration rates versus total depth
for the Hith, Surmeh, and Upper Dashtak formations of well A, respectively. There are

Figure 4. Comparing actual and estimated penetration rate with three methods for the Upper
Dashtak formation.
Downloaded by [Aston University] at 19:27 25 January 2014

Table 2
Computed coefficients for three formations of South Pars field using three methods (well A)

Formation Method a1 a2 a3 a4 a5 a6 a7 a8

Hith Regression 25.384 0.0364 0.6727 0.047 0.010 1.833 140.0 0.0001
Trust region 1.226 0.00023 0.00007 0.000001 0.513 0.486 0.831 0.4009
Progressive stochastic 1.176 0.0003 0.00017 0.000039 0.997 0.918 0.953 0.4024

893
Surmeh Regression 4.317 0.0006 0.8743 0.0069 0.388 1.571 4.298 0.7837
Trust region 1.689 0.0004 0.000003 0.0001 0.521 0.954 0.862 0.5951
Progressive stochastic 1.731 0.00029 0.0009 0.000001 0.500 0.999 0.392 0.5999
Upper Dashtak Regression 0.911 0.0021 0.0233 0.0008 0.620 1.454 5.357 12.01
Trust region 1.893 0.00048 0.0002 0.000002 0.559 0.976 0.514 0.3725
Progressive stochastic 1.900 0.0005 0.0009 0.000001 0.624 1.000 0.792 0.3000
894 H. Rahimzadeh et al.
Downloaded by [Aston University] at 19:27 25 January 2014

Figure 5. Applying the model obtained in well A to well B (Hith formation).

three different predictions for penetration rate in the figures, which are based on the
three mentioned methods for determination of a1 to a8 constants. As these figures show,
the application of progressive stochastic method leads to the most accurate prediction of
penetration rate in comparison with the other methods.
In Table 3, the squared error for each method’s prediction is tabulated for the three
formations of SP field. As the table depicts, minimum errors will be produced by using
the progressive stochastic method, which shows that it is the most reliable method for
the determination of a1 to a8 .

Table 3
Squared errors for different method in the determination of
Bourgoyne and Young coefficients (well A)

Regression, Trust Progressive


Formation % region, % stochastic, %

Hith 21.0 7.9 3.5


Surmeh 10.1 9.1 6.0
Upper Dashtak 11.7 10.1 8.0
Bourgoyne and Young Model Constants 895
Downloaded by [Aston University] at 19:27 25 January 2014

Figure 6. Applying the model obtained in well A to well B (Surmeh formation).

The constant coefficients computed for the formations of well A were used to estimate
penetration rate in the same formations of well B. The results were confirmed for all
cases except for the application of multiple regression techniques in the Hith and Surmeh
formations. Using this method for the two mentioned formations led to zero or very high
values of the penetration rate. Therefore, the resulting coefficients obtained by multiple
regression cannot be applied in the same formations of the other wells.
Figures 5 to 7 validate the models worked out by the trust region and progressive
stochastic methods. These figures and computed errors for each method (Table 4) in

Table 4
Computed errors for applying results from well A into well B

Progressive
Regression, Trust stochastic,
Formation % region, % %

Hith — 9.4 4.6


Surmeh — 14.2 4.4
Upper Dashtak 13.8 8.4 6.3
896 H. Rahimzadeh et al.
Downloaded by [Aston University] at 19:27 25 January 2014

Figure 7. Applying the model obtained in well A to well B (Upper Dashtak formation).

prediction of penetration rate show that the progressive stochastic method leads to the
most accurate model.

Conclusion
The inability of multiple regression techniques to lead to meaningful and reliable constant
coefficients for the Bourgoyne and Young drilling rate model requires the use of other
methods. Although the trust region approach does not have this deficiency, application
of this method does not lead to sufficiently accurate results. Application of progressive
stochastic methods, in addition to meeting the meaningful and in the recommended
ranges results requirements, predicts the penetration rate more accurately compared with
the trust region approach. The main feature that causes the stochastic progressive method
to provide such an accurate drilling rate model is that the method is progressive. This
means that in each step, a shorter interval is defined as the new bounds for the coefficients,
causing the answers to converge to the desired values faster.

References
Akgun, F. (2007). Drilling rate at the technical limit. Int. J. Petrol. Sci. Tech. 1:99–118.
Bourgoyne and Young Model Constants 897

Bahari, A., and Baradaran Seyed, A. (2007). Trust-region approach to find constants of Bourgoyne
and young penetration rate model in khangiran Iranian gas field. Paper No. SPE 107520.
Proceedings of SPE Latin American and Caraibbean Petroleum Engineering Conference,
Buenos Aires, Argentina, April 15–18, pp. 1–60.
Bourgoyn, A. T., Millhem, K. K., Chenevert, M. E., and Young, F. S. (2003). Applied Drilling
Engineering, 9th ed. Richardson, TX: Society of Petroleum Engineers.
Bourgoyne, A. T., and Young, F. S. (1974). A multiple regression approach to optimal drilling
and abnormal pressure detection. Paper No. SPE 4238. SPEAIME Conference on Drilling and
Rock Mechanics, Austin, TX, January 22–23, 1973.
Ricardo, J., Mendes, P., Foseca, T. C., and Serapaio, A. B. S. (2007). Applying a genetic neuro-
model reference adaptive controller in drilling optimization. World Oil Mag. 228:29–38.

Nomenclature
a1 to a8 , ai Bourgoyne and Young model constant coefficients
bi lower bounds for Bourgoyne and Young model coefficients
Downloaded by [Aston University] at 19:27 25 January 2014

ci upper bounds for Bourgoyne and Young model coefficients


D true vertical depth (ft)
db bit diameter (in.)
Fj jet impact force (lbf)
f1 to f8 Bourgoyne and Young model fractional functions
gp pore pressure gradient (lbm/gal)
h fractional bit tooth wear
K formation drillability (ft/sec)
ki auxiliary variables which translate random numbers to constants
N rotary speed (rpm)
W
  weight on bit (1,000 lbf)
W
db
threshold bit weight per inch of bit diameter (1,000 lbf/in.)
t

Greek Letters
c equivalent circulating density (lbm/gal)

You might also like