You are on page 1of 10

A METHOD FOR BIT SELECTION BY MODELLING ROP AND

BIT-LIFE

H.XU T.TOCHIKAWA T.HATAKEYAMA

this article begins on the next page F


THE PETROLEUM SOCIETY PAPER 97-78 A Method for Bit Selection by Modelling ROP and Bit-life H. Xu, T Tochikawa, T. Hatakeyama Japan National Oil Corporation
This paper is to be presented at the 48th Annual Technical Meeting of The Petroleum Society in Calgary, Alberta, Canada, June 8 - 11, 1997. Discussion of this paper is invited and may be presented at the meeting if filed in writing with the technical program chairman prior to the conclusion of the meeting. This paper and any discussion filed will be considered for publication in CIM journals. Publication rights are reserved. This is a pre-print and is subject to correction.
THE PETROLEUM SOCIETY PAPER 97-78

,
.~

A Method for Bit Selection by


Modelling ROP and Bit-life

H. Xu, T. Tochikawa, T. Hatakeyama


Japan National Oil Corporation

This paper is to be presented at the 48th Annual Technical Meeting of The Petroleum Society in Calgary, Alberta, Canada, June 8 - 11,
1997. Discussion of this paper is invited and may be presented at the meeting if filed in writing with the technical program chairman prior
to the conclusion of the meeting. This paper and any discussion filed will be considered for publication in CIM journals. Publication rights
are reserved. This is a pre-print and is subject to correction.
and tl>, Eq.! can be re-written as:
CPUD=..E.r-+ Cb + Cret z
ABSTRACT ROP ROPetb ROPetb Eq.!.1

This paperpresents a practical methodfor bit selection According to Eq. 1.1 the following observations can be
through modeling ROP and Bit-lift from the conventional made:
and routinely available bit record and logging data. This • To any bit run, CPUD is effectively determined by
method employs simple statistical analysis to quantify the ROP and 4" because bit cost (Cb) and rig rate (Cr) are
effects of individual parameters on bit performance and constant and trip time (t l ) vary little; to any well, the sum
then identify the most significant parameters from the of trip times depend on the number of bit runs and the
uncountable variables associated with drilling processes. number of bit runs are determined by both 4, and ROP. In
The parameters incorporated in the model can be short, improving bit performance is the direct and efficient
demonstrated to be statistically significant, physically approach to drilling cost-reduction.
interpretable and dynamically controllable. Based on the • To individual bit run, changes in ROP have greater
model ofROP and Bit-life, the value of cost-per-footage impacts on CPUD. It can be seen from Eq.l.l, ROP
can be normalized and therefore bit selection can be appears in all the 3 terms while 4, appears in 2 terms, C, in
conducted on a scientific basis. In addition, the method 2 terms and Cb in 1 term. Furthermore' partial derivative of
for developing model to optimize drilling parameters for Eq.l.l can explain the impacts of individual components
~ndividual bit runs and for dynamic drilling conditions is on CPUD analytically. However, it should be noted that
zllustrated. Also the application of the model in ROP both ROP and ~ may be themselves correlated and, to a
prediction, in diagnosing drilling problem is discussed well, the sum of trip time (tl) depend on the number of bit
briefly. This method unlocks a whole plethora ofmethods runs and the number of bit runs are determined by tb and
to model bit performance and has the potential to be ROP as well.
automatedfor routine application. Apparently Eq.l.l has been straightforward, but to
apply it practically is difficult For instance, there are two
bits, bit No.16 and bit No.74, with the same name and
same size while the value ofCPUD for No.16 is 61unit and
1. INTRODUCTION for No.74 is 189unti (Table la). It will be debatable to say
that the performance of bit No.16 is better than that of bit
When drilling a well, bit related costs account for a No.74, since both ROP and ~ may themselves correlated
colossal value. It follows that even a small percentage of and affected by a number of other variables, particularly
savings in it can yield significant cost-reduction. WOB, RPM, Mud-density, Flowrate and formation
Quantitatively the formula below has long been recognized propertyl,2,3 (Table Ib). It follows that, to evaluate bit
as a tool to measure bit performance: performance and then select the most suitable bit for
CPUD= Cb+CreCtb+tz) subsequent drilling practice, it is essential to normalize the
liD Eq.! value of CPUD. And to normalize the value of CPUD, it is
where CPUD stands for Cost-Per-Unit-Depth, C for the essential to model ROP and bit-life, because modeling
cost of bit, C r for the operating cost of the rig per ~it time; ROP and bit-life can quantify the impacts of individual
4, for th: sum of the actual drilling time; 1t for tripping and
parameters and then identify the most significant variables
connection time; AD for the given interval of depth. in particular drilling conditions 4,5.6,7. As a result Eq.l.l is
Because.6.D is the product ofROP (Rate Of Penetration) re-written as Eq.1.2, where subscript "m" means

- '-,
"modified". measured accurately; also it is a disadvantage to
incorporate the variables which are not significant in the
CPUD =-.s:....-+ Cb + Crett conditions in which the model was derived and to be
m ROPm ROPm etb
m
ROPm etb
m
Eq.l.2
applied.
Overall, coefficienty is introduced to temper the value In view of the inadequacy of the conventional method
of the .mo~ified Cost-per-footage (CPUDm), where the in ROP modeling, a new method has been develope<f,s·6.7.
determmation ofy depends on engineering judgment. For.illustration, an exercise with the following steps is
earned out:
C UP'Dm _( Cr
- --+
cb + Crett )·rE 13 (1) Prepare the data. 46 records of bit with same bit
T ROPm ROPm etbm ROPm etbm q.. type (Insert-Rock-Bit), same size (12-114"), used to drill
Conventionally, there exist two typical methods for the same formation namely Hiuchiyama which is a typical
e~aluating bit performance. One method is to compare the formation in Japan, drilled in same mud-type (Water-
bIt performance by reviewing the conventional bit record based). Because each records is averaged from a number
which condenses all the variation of the operating of meter-by-meter measured data points, the effects of a
conditions of individual bit run into a single line. few outliers is minor4,s. In this formation, two PDC bits
However, the brevity of the data may result in the loss of and one diamond bit were used. However, the records of
valuable information and in the ignorance to the cause and those three bits were excluded because, according to the
e~fect of poor or good performance, consequently, it is
logging reports, "diamond bits and turbine however did not
~fficult to mak~ practical recommendations for improving
improve the rate of penetration. Also in both the runs the
s bits did not last in the hole half the time of what the
bIt performance m subsequent bit runs . At a more detailed
level than bit records, integrated plots of mud and electric medium and hard insert bits lasted, and, when pulled out of
logging data are now becoming more common for hole bit wear was found to be around 60%."
performance diagnosis in well sections of interests. These (2) Review the background of individual data. Read
plots facilitate a visual impression ofthe interaction of bit relevant documents, such as daily reports and logging
performance and different variables on a foot-by-foot or reports, to ascertain the background to the bit run. This
meter-by-meter b~is. For example, overlaying ROP, may be useful for both explanation and application of the
WOB and formatIon logging data on a single plot may results or may highlight the data which may be
reveal how WOB affects ROP for a range of lithologies. inappropriate for detailed analysis. It was noted that all the
Although such plots are of undoubted value in drilling 46 data were from normal drilling conditions and no lose-
practice, they are not intended to quantify the relationship bit cone(s) or other failure were reported; also the
be~een ROP aD:d its influential variables from the large
personnel involved in drilling that section were inquired,
amount of loggmg data. Interpretation of such plots however non drilling problems could necessarily be
!he~e~ore tends ~ be subjective and depends largely on
attributed to the suitability of individual bits. In addition,
mdIvIdual expertIse. The use ofthe conventional method the relation between torque and ROP or 4. was investigated
should be complemented by using the method described in but no significant phenomenon was found.
this paper as well as other works by Xu et ar,s·6.7. (3) Perform Regression Analysis. Multiple linear
regression analysis benefits from being simple to apply and
straightforward to interpret. A visual inspection of the
residuals (the difference between actilal and predicted ROP
2. METHOD FOR MODELING ROP
values) after a linear regression analysis indicates the
suitability of the model and may suggest how the model
Literally modeling ROP is not new 1•2,3. Although the may b~ improved by including non-linear variables such as
previous methods in literature described drilling WOB for example. According the general assumption
mechanism and indicated the variables which should be about drilling practice 1•2 ,3, ROP is on the side of the
considered when developing a model, it is difficult to apply dependent variable, while WOB, RPM, Mud-density,
the model in literature in drilling practice. For instance, the Flowrate ,which are all dynamically controllable are on the
field applicability of models developed from laboratory side of the independent variables. Also Ro~k-density,
controlled experiments may be limited, because a model is measured every 10-meters based on the cutting samples, is
~imp~~ a mathematical representation of a physical process, on the side of the independent variables standing for the
Its utihty must largely depend on the similarity between the formation prope~. As a result, a model incorporating
conditions in which it was derived and those in which it is RPM, Mud-densIty and Rock-density is derived (Eq.2.1,
to be applied. However, real drilling conditions almost Fig.l).
always differ radically from those experienced in the
laboratory. Also, models in literatilre usually prescribe the ROP=32.27+O.024.RPM-6.0S.MD-9.12.RD Eq.2.1
general form of the relationship between ROP and the
independent variables, which is unrealistic given the wide From Fig.l with statistical analysis, the followin a
potential for variation in the bit/rock interaction with observations can be made: .,
different lithologies. Also it is difficult to obtain the • 70% of the ROP variations can be explained by the
accurate coefficients and constants incorporated in the three independent parameters namely RPM, Mud-density
models in literatilre2,3, for example, bit constants, rock and Rock-density (Eq.2.1), which indicates that little could
strength, pressure differential of formation etc.. In fact, it be improved by borrowing models including numerous
may damage the accuracy and reliability of a model to variables or by retaining insignificant variables. Also it is
incorporate the coefficients/constants which can not be evident that little can be improved by including variables in

2
I

I--------

You might also like