Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

SPE-193779-MS

Application of Data Analytics to Improve Drilling Performance and Manage


Drill Stem Vibrations

Mohammed Al Dushaishi, Texas A&M International University; Svein Hellvik, National Oilwell Varco; Ahmad
Aladasani, Consultant; Mortadha Alsaba, Australian College of Kuwait; Qutaiba Okasha, Kuwait Oil Company

Copyright 2018, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE International Heavy Oil Conference and Exhibition held in Kuwait City, Kuwait, 10-12 December 2018.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Data mining and Artificial Intelligence (AI) methodologies are underdeveloped in the oil and gas industry,
despite the need to improve drilling performance and remain globally competitive in all capital-intensive
projects.
Drilling companies allocate significant resources to improve well planning, drilling schedules and rig
management. Well planning comprises of two main elements; drilling performance and the reduction of
drill stem vibrations. Therefore, modeling methodologies such as drill string statics, dynamic tools and rate
of penetration modeling are applied to determine the optimum bottom hole assembly (BHA) components
and drill bit design. However, more attention is required on drill stem fatigue, non-productive time (NPT)
and their impacts on drilling operations.
In this paper, Data Analytics (DA) is applied to drilling logs taken from three wells that recorded
vibration readings from different geological stratification. In turn, the work in this paper establishes
a relationship between drill stem vibrations and various measurement and logging data while drilling.
Statistical regression and multivariate analysis were used to examine correlations of drilling parameters,
including BHA assembly, to vibration data. Therefore, the results include a composite vibration model that
describes the drilling stem vibration behavior as a function of drilling parameters, and geological formations.
Results of the vibration models built in this study indicate that the drill stem lateral vibration behaves
parabolically as a function of the drill pipe length, length of drill collar, gamma ray (GR) response, and
weight on bit (WOB). The analysis of drill stem vibration effect on the mechanical specific energy (MSE)
was inconclusive for depths below 1350 meters. However, for depths above 1350 meters a strong correlation
was observed to ROP.

Introduction
Severe drill stem vibration causes downhole equipment failure and accelerates downhole equipment wear.
The drill stem encounters axial, torsional, and lateral vibration while drilling, as shown in Figure 1, however,
one vibration mode will be more dominate than the others (Dunayevsky et al. 1993). With each vibration
2 SPE-193779-MS

mode, destructive mechanic could occur in severe cases, for instance, bit bounce during axial vibration,
stick-slip during torsional vibration, and forward/backward whirl during lateral vibrations.

Figure 1—Drilling Vibration Modes

To understand drill stem vibrations, theoretical modeling and downhole vibration measurements are
used. Numerous drill stem vibrations models were proposed in the literature to address different effects
(Paidoussis et al. 2008; Ritto et al. 2009; Al Dushaishi 2015; Al Dushaishi et al. 2016; Alsaffar et al.
2018; Dong and Chen 2018). Drill stem vibrations are measured in both real time (Arevalo and Fernandes,
2011) and in memory mode for post-well analysis (Wright et al. 2014, Al Dushaishi et al. 2015). Drill
stem vibration measurements are important to understand factors affecting downhole dynamics in order to
mitigate damages to the BHA (Akinniranye et al. 2009).
Drill stem vibrations are affected by many factors including the energy input parameters such as rotational
speed and WOB, BHA and bit designs, and geological formations/structures. The energy parameters are
also significant in measuring drilling performance. Drilling performance is often measured using the MSE
developed by Teale in 1965. Since drill stem vibration are affected by the applied WOB and RPM, which
are also a function of the MSE, an investigation of the direct relationship between drill stem vibration and
its effect on MSE is not well established.
Both DA and AI have been used for geological identification and well steering functions (Bao et al.
2012). However, there is limited application of DA and AI for drill stem vibration, and drilling performance,
analysis. Adhikari et al. 1989 and Scoble et al. 1989, respectively, performed an analysis of the induced
vibrations and drilling performance for blast surface hole, in mining application. Adhikari et al. 1989
reported a poor relationship between rock properties and induced blast vibrations, and suggested including
structural discontinuities to better correlate rock properties and induced blasting vibrations. Another study
used real time vibration data, in geosteering application, to correlate vibration with geological events
(Sitinjak et al. 2017). This study revealed a strong correlation between the measured near bit vibrations
and geology, with axial vibrations being more dominant than lateral vibrations during lithology change.
However, their conclusion was not supported with statistical evidence and was only based on qualitative
analysis.
SPE-193779-MS 3

Theoretical dynamic stability in conjunction with real time vibration data, using machine learning
approach, was used to select optimum combination of RPM and WOB in order to avoid severe vibrations
(Christian, 2017). Predetermined dynamic stability boundaries, based on WOB and RPM are established as
a base and real time machine learning algorithm is used to identify the optimum drilling zone within this
window (Christian, 2017). Multivariate and regression analysis were used to correlate drilling performance
with geological, geo-mechanical and drilling parameters. It was found that the unconfined compressive
strength (UCS), internal friction angle, and flow rate are the most contributing parameters affecting drilling
performance when considering the geo-mechanical properties (Behboud et al. 2016). Considering only the
mechanical properties, regression analysis of Davarpanah et al. (2016) concluded that TRQ, ROP, WOB,
flow rate, RPM, differential pressure, and measured depth are the most influential independent parameters
on drilling performance. The use of data analytics, to understand drill stem vibrations behavior, is still
under-developed. For instance, drill stem vibration data was used only to qualitatively describe the vibration
behavior with respect to geological event (Sitinjak et al. 2017) without making full use of the vibrations data.
Quantitative analysis addressing the effect of drilling and geological data on drilling performance i.e.
MSE, were addressed from mechanical, geological, and geo-mechanical prospective (Behboud et al. 2016;
Davarpanah et al. 2016), however drill stem vibration was not considered. Christian (2017) used machine
learning algorithm for selecting optimum drilling zone based on WOB and RPM, however, the data used
in his analysis did not include vibration data, and were based on WOB and RPM correlation with MSE.
Thus, there is a need to understand and correlate measured drill stem vibrations with both mechanical and
geological factors. Due to the complexity of drill stem vibration, the use of DA and AI for data driven
modeling is one of the key answers for drilling performance and managing drill stem vibrations.
The objectives of this paper are to correlate various drilling data and geological data to drill stem
vibration and drilling performance. Using downhole and surface data, the influence to drilling and geological
parameters on drill stem vibration is investigated via data analytics. The goal of this work is to address the
following gaps in literature summarized in Table 1.

Table 1—Summary of Literature Review

Well Data and Descriptive Analytics


Surface and downhole drilling and vibration data of vertical wells, previously published in Al Dushaishi et
al. 2015, were used in this study. The data includes length of drill pipe (Ldp), length of drill collar (Ldc),
hook load (HL), WOB, surface and downhole RPM, torque (TRQ), flow rate (FR), standpipe pressure (SPP),
4 SPE-193779-MS

rate of penetration (ROP), and GR. The first step in the analysis consisted of locating the parameters with
the most significant influence on drill stem lateral and torsional vibrations. Initially, each section of the
three wells was analyzed separately, where each well consisted of three sections as shown in Table 2. The
analysis of each individual section was then used to build the vibrations behaviors of the entire well.

Table 2—Summary of Wells of Study

The lithology effect was considered by calculating the UCS, where the UCS is obtained using the acoustic
velocities as (Hareland and Nygaard 2007):

(1)

where k1, k2, and k3 are constants that depends on lithology, Δtc is compressive wave travel time in µsec/
ft, and UCS have a unit measure of MPa.
Using the energy drilling input parameters, MSE was calculated following Teale (1965), and Dupriest
and Koederitz (2005) as:

(2)

where Db is the bit size in meters, N is rotational speed, T is the resultant torque (KN.m) and MSE have
a unit of Pa.
Stick-slip severity is determined based on the measured downhole RPM and the applied surface RPM
using (McCarthy et al. 2011):

(3)

where ΔRPMDH is the difference between the maximum and minim downhole RPM measured in one period,
and RPMs is the measured applied surface RPM.
Multi-regression analysis was used to construct linear models that predict drill stem lateral vibrations,
stick-slip tendency, and MSE using different measurement while drilling (MWD) and logging while drilling
(LWD) data. The standard least square method was used to construct models, where the goal is to find
factors that maximize or minimize the response. The standard least square model provides a continuous
response fit to a linear model with multiple factors. When a model is created, the model is evaluated based
on the summary of fit, which includes estimation of variation in the response rather than to random error
(Jmp and Proust, 2012). The summary of fit is indicated using R2, which measures the proportion of the
variation accounted for by fitting means to each factor level. The remaining variation is attributed to random
SPE-193779-MS 5

error, including measurement range and accuracy. An R2 value of 1 indicates that the means account for all
the variation with no error. A low R2 value suggests that there may be variables not included in the model
that account for the unexplained variation.
Several models were constructed to account for all the variation in the model and decrease errors. In this
study, the models that showed the least error were constructed using response surface method. The response
surface method is a collection of mathematical expressions used for modeling and analysis of problems,
where a response is influenced by multiple parameters. The response surface method can be either first
order linear model with intersection, first order linear model without intersection, or second order quadratic
models, which are respectively, expressed as (Alexander, 2000):
(4)

(5)

(6)
where Y is the response to be predicted, which is dependent on the xij representing the independent variables.
When the model is created, the parameters that does not contribute to the overall model response are removed
and then the model is recalibrated.

Empirical Models
A regression analysis using the standard least-square method was performed to investigate the relationship
between different MWD/LWD parameters and drill stem vibrations, mainly, the lateral vibrations and the
stick-slip tendency. Then the same methodology was used to investigate the relationship between MSE and
drill stem vibrations with the use of drilling data. The regression analysis was built for each individual
section, and then used to build a full regression analysis for the entire well. Different statistical models were
tested using all MWD and LWD data. The model that showed the best fit was the model that was constructed
based on surface response effect.
For the lateral vibration regressions analysis, the factors that showed the highest contribution to the model
fit are listed in Table 3, which includes the model fit summary for each well.

Table 3—Lateral Vibration Model Parameters and Summary of Fit for Each Well

The best model summary was obtained with the data collected from Well A with summary of fit of
R2=0.77. For Well B and Well C, the sonic travel data was lacking, and thus UCS data was not included
in the analysis for those wells. The effect of each parameter on lateral vibration is visualized in factor
prediction profile plot as shown in Figure 2 for well A. The lateral vibration prediction profile shows the
prediction of each parameter in black line, where the lateral vibration value in the y-axis corresponds to the
vertical dotted redline location (i.e. current value) of each factor, while values in brackets corresponds to
the 95% confident interval of the mean response of the lateral vibration. The blue lines with shaded areas
show the 95% confident interval of each factor. The surface RPM and the length of drill pipe effects on
6 SPE-193779-MS

lateral vibration showed similar profile for all models, where increasing RPM causes an increase in lateral
vibration.

Figure 2—Factors Profile With Respect to Lateral Vibrations for the Models Obtained From Well A

The model showed that drill stem lateral vibration behaves parabolically with the Ldp, Ldc, GR, and
WOB, where lateral vibration tends to increase at both low and high values of each parameter. The
aforementioned parabolical behavior is caused by operator intervention of the RPM, as the RPM is properly
adjusted by the operator to match the WOB to mitigated stick-slip and stabilize the drilling parameters.
Therefore, Figure 3 indicates that manual over/under shooting required RPM has the highest influence on
stick-slip occurrence. The stick-slip model was based on the calculated stick-slip severity. Following the
lateral vibration model procedure, Figure 3 shows the stick-slip model profile.

Figure 3—Comprehensive Stick-Slip Model Factors Profile

In regards to the coefficient of determination of the stick-slip model, the model showed a lower coefficient
of determination than expected (R2=0.69). This indicates a low frequency of occurrence and a dominating
influence of RPM on stick-slip behavior, and more parameters such bit specification and BHA designs are
required to contrast and evaluate stick-slip control measures.
The effect of drill stem vibration on MSE was addressed by building a regression model that includes
drill stem vibration factors. The regression analysis, considering the entire data, showed no correlation (i.e.
low coefficient of determination) with MSE and drilling parameters. Thus, following the MSE equation (Eq
2), the regression analysis was performed based on drill bit size. Figure 4 and Figure 5 show the model
factor profile for the 12¼" and the 8½" bits, respectively. The coefficient of determination for the 12¼"
model was R2=0.68, while for the 8½" model the coefficient of determination was R2=0.6. While both
models have moderately low R2 coefficients, model factors behavior differently as expected for each model.
The 12¼" model with R2= 0.68, showed that lateral vibrations, stick-slip tendency, Ldp, GR, RPM, and
flow rate have minor influence on MSE (Figure 4) The 8½" model however, showed that lateral vibrations,
stick-slip tendency, and GR have high influence on MSE (Figure 5). It is expected and confirmed that
significant energy is transmitted/stored in the drill string when well depth exceed 1350 meters. The 12 ¼"
model indicates that insignificant MSE is present in the drill string. As such, all the drilling profiles indicate
SPE-193779-MS 7

flat response to MSE except for WOB. Because energy is transmitted from the drill stem to the bit, where
the amount of MSE is throttled. Future work should focus on level of energy transmission between various
bit size, drill pipe length, and gamma log readings.
In Figure 5 it is noticed once more, how operator intervention is reflected parabolically, as the ROP is
undershoot or overshoot by the operator. Maximum instability is caused by overshooting the ROP especially
at high RPM (RPM >110), poor confidence levels are indicated for all drilling parameters especially the
lateral vibration and the stick-slip. The aforementioned is a dangerous condition that often arises when
drilling operation are expedited due to the financial incentives of completing the well. The lateral vibration
magnitude moves inversely to the stick-slip, which is indicative of a change in vibration type.

Figure 4—MSE 12¼" Model Factor Profile

Figure 5—MSE 8½" Model Factor Profile

Discussion
It is worthwhile mentioning that all measured drilling and geological parameters were initially included
in the models. However, parameters that showed no influence were removed from the models, for
8 SPE-193779-MS

simplification, without compromising the model accuracy. The model accuracy is acceptable to describe
drilling lateral vibration, shown in Figure 2. The developed work in this paper shows:
a. The model indicates that drill stem lateral vibration behaves parabolically (explaining operator error)
as a function of Ldp, Ldc, GR and WOB. This behavior agrees with field observations of lateral
vibrations measurement, where lateral vibration increase at low WOB, extended reach wells, and low
and high GR values (Al Dushaishi et al. 2015). Similar behavior can also be seen in the comprehensive
stick-slip model, where stick-slip tendency increases at low and high RPM and decreases with the
increase of rock strength i.e. GR. The modeling in this paper, suggests that operator improper WOB
adjustment dominantly impacts lateral vibration, whereas operator improper RPM adjustment impacts
dominantly stick-slip.
b. The MSE model's accuracy improved with increased depth. This is representative of MSE stored in
the drill stem. The 12¼" model showed that both lateral vibrations and stick-slip tendency have no
effect on MSE, while the 8 ½" model showed that lateral vibrations and stick-slip tendency increases
MSE i.e. reduces drilling efficiency.
c. Future work should focus on the level of energy transmission between various bit size, drill pipe
length, and GR readings. Therefore, more data is required to address the bit specifications, BHA
designs, and well trajectory for deviated wells. In the current status of the analysis, the models can
be used as a base for artificial intelligent models and drilling automation to optimize drilling based
on data driven analysis.
The following is a summary of risk drilling envelopes that develop as drilling operation progresses. The
WOB and RPM increase in par with increasing depths and/or formation strength, as such MSE build ups.
The pre-requisite for lateral and stick-slip is established with MSE. The first risk is initiated by operator
improper adjustment of WOB which generates lateral vibration, which is mitigated by the weight of the
drill collar (i.e. increasing the BHA inertia). As the length of drill pipe exceeds 1350 meters, the second
risk is initiated by the operator improper adjustment of RPM which results in stick-slip. The sensitivity of
RPM to stick-slip is greater than the sensitivity of WOB to lateral vibration (shown in figure 2 and figure
3). Stick-slip has no mitigative measure such as stiffening the drill pipe with the drill collar, as such the risk
magnitude is higher. If we consider risk management principles, RPM should have alarms set in place when
drilling depths exceed 1350 meters or formations that have high strength.
A drilling risk matrix is developed based on statistical analysis (Table 4). The risk matrix comprises of the
key parameters that have a significant impact on the performance of all drilling program, indicating critical
ranges that risk regimes for operators to observe and consider during the development of their well plans.
SPE-193779-MS 9

Table 4—Drilling Risk Matrix

Researcher of drilling performance should comprehend and capture acting forces in each drilling regime
in order to create more representative models that are likely to generate meaningful results.

Conclusions
In this paper, descriptive data analytics was applied to drilling, and vibration data, which was collected from
three drilled wells, with the objective of studying the impacts of different drilling and geological factors on
drill stem vibration. A relationship between drill stem vibrations and drilling/geological parameters were
established using regression analysis. The vibration model created in this works indicates that drill stem
lateral vibration behaves parabolically as a function of the drill pipe length, length of drill collar, gamma
ray, and weight on bit. Lateral vibration increases at low WOB, extended reach wells, and low and high
GR readings. The stick-slip tendency increases at low and high RPM, i.e. parabolic profile, and decreases
with the increase of GR readings.
The significance of the work in this paper is based on the following salient conclusions:

• Geological characteristics influence lateral vibration. Gamma ray data correlations are comparable
to the WOB.
• The relationship between lateral vibration and WOB is inversely proportional.

• Stick-slip frequency increased with well depth in par with the model accuracy for 8 ½" drill bit
section compared to 12¼" drill bit section.
• Mechanical specific energy is influenced by ROP and is a prerequisite for lateral vibration and
stick-slip tendency
• Mechanical stored energy builds up as the ROP increases. Stick-slip likelihood increases in par
with RPM, followed by depth. Lateral vibration ramps up with WOB and length of drill collar. A
neural network can be created for MSE combined with stick-slip and lateral vibration.

Acknowledgment
The author would like to acknowledge the travel support from a University Research Grant awarded by
Texas A&M International University.

Abbreviation
AI Artificial Intelligence
10 SPE-193779-MS

DA Data Analytics
BHA Bottom Hole Assembly
FR Flow rate
GR Gamma Ray
HL Hook Load
Ldc Length of Drill Collar
Ldp Length of Drill Pipe
MSE Mechanical Specific Energy
R2 Coefficient of Determination
ROP Rate of Penetration
RPM Rotation per Minute
SPP Stand Pipe Pressure
TRQ Torque
WOB Weight on Bit
UCS Unconfined Compressive Strength

References
Al Dushaishi, M. F., Nygaard, R., Hoel, E., Hellvik, S., and Andersen, M. 2015. Post Well Vibration Analysis in the North
Sea: A tool to Understand Drilling Performance. Proceedings of the ASME 2015 34th International Conference on
Ocean, Offshore and Arctic Engineering. Newfoundland, Canada, May 31-June 5. OMAE2015-42227. doi: 10.1115/
OMAE2015-42227.
Al Dushaishi, M., Nygaard, R., and Stutts, D. 2016. Effect of drilling fluid hydraulics on drill stem vibrations.
Journal of Natural Gas Science and Engineering, Volume 35, Part A, Pages 1059-1069. https://doi.org/10.1016/
j.jngse.2016.09.041.
Al Dushaishi, Mohammed Fayez, "Numerical and field data analysis of drill stem vibration" (2015).
Doctoral Dissertations, Missouri University of Science and Technology. http://scholarsmine.mst.edu/
doctoral_dissertations/2435.
Alexander, M.T., 2000. Response surface optimization using JMP® software. http://www2.sas.com/proceedings/sugi22/
STATS/PAPER265.PDF.
Alsaffar, Y., Sassi, S. and Baz, A. 2018. Band Gap Characteristics of Nonrotating Passive Periodic Drill String. Journal
of Vibration and Acoustics, 140(2), p.021004. doi: 10.1115/1.4037851.
Arevalo, Y. I., and Fernandes, A. J. 2011. Quantification of Drillstring Integrity Failure Risk Using Real-Time Vibration
Measurements. Presented at the SPE Asia Pacific Oil and Gas Conference and Exhibition, 20-22 September, Jakarta,
Indonesia. https://doi.org/10.2118/147747-MS.
Bao, F., He, X. and Zhao, F., 2010. Applying data mining to the geosciences data. In Computer, Mechatronics,
Control and Electronic Engineering (CMCE), 2010 International Conference on (Vol. 5, pp. 290-293). IEEE. https://
doi.org/10.1109/CMCE.2010.5609971.
Dong, G. and Chen, P. 2018. The vibration characteristics of drillstring with positive displacement motor in compound
drilling Part1: Dynamical modelling and monitoring validation. International Journal of Hydrogen Energy, 43(5),
pp.2890-2902. https://doi.org/10.1016/j.ijhydene.2017.12.161.
Dunayevsky, V., Abbassian, F., and Judzis, A. 1993. Dynamic stability of drillstrings under fluctuating weight on bit. SPE
Drilling and Completion, 8(2), 84-92. https://doi.org/10.2118/14329-PA.
Dupriest, F. E., and Koederitz, W. L. 2005. Maximizing Drill Rates with Real-Time Surveillance of Mechanical
Specific Energy. SPE/IADC Drilling Conference, Amsterdam, Netherlands, 23-25 February. SPE-92194-MS. http://
dx.doi.org/10.2118/92194-MS.
Hareland, G., and Nygaard, R. 2007. Applications of Rock Strength in Drilling Evaluation. SPE Latin American
and Caribbean Petroleum Engineering Conference, Buenos Aires, Argentina, 15-18 April. SPE-106573-MS. http://
dx.doi.org/10.2118/106573-MS.
Jmp, A. and Proust, M., 2012. Modeling and multivariate methods. JMP Inc., Cary.
McCarthy, J., Forster, I., Burnett, T., and Kabbara, A. 2011. Careful Planning and Application of an Asymmetric Vibration
Damping Tool Dramatically Improves Underrating While Drilling Performance in Deepwater Drilling. Presented at
Offshore Technology Conference, 4-6 October, Rio de Janeiro, Brazil. OTC-22439. https://doi.org/10.4043/22439-
MS.
SPE-193779-MS 11

Paidoussis, M. P., Luu, T. P., and Prabhakar, S. 2008. Dynamics of a long tubular cantilever conveying fluid downwards,
which then flows upwards around the cantilever as a confined annular flow. Journal of Fluids and Structures, 24(1),
111-128. https://doi.org/10.1016/j.jfluidstructs.2007.07.004.
Ritto, T., Soize, C., and Sampaio, R. 2009. Nonlinear dynamics of a drill-string with uncertain model of the
bit-rock interaction. International Journal of Non-Linear Mechanics, 44(8), 865-876. https://doi.org/10.1016/
j.ijnonlinmec.2009.06.003.
Teale, R. 1965. The Concept of Specific Energy in Rock Drilling. International Journal of Rock Mechanics and Mining
Science, 2 (1), 711–725. doi: https://doi.org/10.1016/0148-9062(65)90022-7.
Wright, J., Borri, L., Simeone, D., Longo, J., Weber, M., and Barnes, B. 2014. Nikaitchuq Project Vibration Mitigation
Strategy. Presented at the SPE Western North American and Rocky Mountain Joint Meeting, 17-18 April, Denver,
Colorado. https://doi.org/10.2118/169559-MS.

You might also like