Ahmed Gowida, Salaheldin Elkatatny, Abdulazeez Abdulraheem, and Dhafer Al Shehri, King Fahd University of
Petroleum & Minerals
This paper was prepared for presentation at the International Petroleum Technology Conference held in Dhahran, Saudi Arabia, 13 – 15 January 2020.
This paper was selected for presentation by an IPTC Programme Committee following review of information contained in an abstract submitted by the author(s).
Contents of the paper, as presented, have not been reviewed by the International Petroleum Technology Conference and are subject to correction by the author(s). The
material, as presented, does not necessarily reflect any position of the International Petroleum Technology Conference, its officers, or members. Papers presented at
IPTC are subject to publication review by Sponsor Society Committees of IPTC. Electronic reproduction, distribution, or storage of any part of this paper for commercial
purposes without the written consent of the International Petroleum Technology Conference 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 where and by whom the paper was presented.
Write Librarian, IPTC, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax +1-972-952-9435.
Abstract
Synthetic well-log generation using artificial intelligence tools is presented as a robust solution when the
logging data are not available or partially lost. Formation bulk density (RHOB) logging data greatly assist
in identifying downhole formations. It is measured in the field using density log tool either while drilling
by logging while drilling technique (LWD) or mostly by wireline logging after the formations are drilled
because of the operational limitations during the drilling process.
Therefore the objective of this study is to develop a predictive tool for estimating RHOB while drilling
using artificial neural networks (ANN) and Adaptive network-based fuzzy interference systems (ANFIS).
The proposed models used the drilling mechanical parameters as feeding inputs and the conventional RHOB
log-data as an output. These drilling mechanical parameters including the rate of penetration (ROP), weight
on bit (WOB), torque (T), stand-pipe pressure (SPP) and rotating speed (RPM), are usually measured while
drilling and their responses vary with different formations.
A dataset of 2400 actual data points obtained from horizontal well in the Middle East is used for building
the proposed models. The obtained dataset is divided into 70/30 ratios for training and testing the model
respectively. The optimized ANN-based model outperformed the ANFIS-based model with correlation
coefficient (R) of 0.95 and average absolute percentage error (AAPE) of 0.72 % between the predicted and
the measured RHOB compared to R of 0.93 and AAPE of 0.81 % for the ANFIS-based model. These results
demonstrated the reliability of the developed ANN model to predict the RHOB while drilling based on
the drilling mechanical parameters. Afterwards, the ANN-based model is validated using unseen data from
another well within the same field. The validation process yielded AAPE of 0.5 % between the predicted and
the actual RHOB values which confirmed the robustness of the developed model as an effective predictive
tool.
Introduction
Formation density is considered one of the main factors for identifying the nature of the subterranean
formations (Ellis, 2003). It is categorized as one of the porosity logs, which gives an indication on the
2 IPTC-19787-MS
electron density of the drilled formation (Rolon et al. 2009). It can be used to provide valuable information
for the geologists and geoscientists such as : (1) identifying the drilled formations, (2) detecting the evaporite
mineral existing within the formation (Rolon et al. 2009), (3) detecting high-pressurized formations, (4)
detecting the fluid content of the drilled formation (Wraight et al., 1989), (5) investigating the invasion
zones for the drilled formations (Bassiouni, 1994), (6) assisting in developing geo-mechanical models which
provides information on the mechanical properties of the formations (Gatens 1990) and (7) evaluating the
porosity of the reservoirs (Reichel et al, 2012).
Density logging tool was firstly presented in the petroleum industry 1960s. It can by simply described as
a nuclear tool which uses a radioactive source for emitting gamma rays through the formations with medium
energy, and a detector to collect the reflected gamma rays from the formations (Wahl et al., 1964). The
radioactive source and the detector are lowered combined through the drilled well opposite to the selected
formations. Afterwards, the gamma rays are emitted from the radioactive source thereafter interact with the
electron of the formations resulted in scattering rays. These scattered rays will then be collected using the
detector which is placed at a certain distance from the source within the logging tool. The number of the
gamma rays collected by the detector gives an indication on the density of the electrons within the formation
and accordingly the formation bulk density can be calculated (Alger 1963).
Formation porosity can be derived from these logging data once the rock matrix density and the saturating
fluid density are known (Spross 1993; Darwin 2003), using the following formula:
(1)
where, Ø is the formation porosity, ρm is the rock matrix density, ρb is the formation bulk density, and ρf
is the fluid density.
Formation bulk density (RHOB) can assist in optimizing the drilling operation while drilling by
optimizing the bit selection which is dependent on the nature of the formation being drilled. Moreover, it
helps avoid many interrupting problems such as loss of circulation, kicks and wellbore instability, by proper
detecting of the downhole formations while drilling (Falconer et al. 1988).
There are two possibilities to measure RHOB in site; either (1) using logging while drilling (LWD) tools
while drilling, or (2) using wireline logging after the hole has been drilled (Wraight et al. 1989). However,
LWD measurements have many challenges due to the harsh environment in the hole wile drilling which
makes them need several corrections and increase their operational cost (Jackson 1994) therefore these data
are not always available during the drilling operations and it is preferable to run the logging tools after
drilling the hole to avoid many logging difficulties while drilling. As a result, RHOB measurements may
not be available during the drilling operation and identifying the drilled formations while drilling would
be confusing due to lack of data. There is another way for identifying the drilled formation, rather than
using logging data, using the analysis of the collected cuttings however this method has a lag time so that
it cannot provide real-time information on the drilled formations (Xu 2009). Synthetic well log generation
using different methods have been introduced as a n alternative and robust solution for obtaining log data
while drilling and even for those sites where the well logs data are not available or partially absent (Zhang
et al., 2018; Gowida et al. 2019).
The nature of the subterranean formations significantly affects the drillability of these formations. Each
formation has its own characteristics which affect its resistance for drilling. WOB, T, and RPM are adjusted
depending on the nature of the formation like soft and hard formations (Bourgoyne et al., 1986). Also, the
nature of the cuttings for each formation is a key parameter for adjusting SPP to control hole cleaning.
All the aforementioned parameters in addition to the formation type play a main role in controlling ROP
(Head, 1951; Mensa-Wilmot et al., 1999). Therefore, these drilling parameters are somehow related to the
nature of the drilled formations and in turn their densities. The usual availability of the drilling mechanical
parameters while drilling rises the idea of using them as inputs to estimate RHOB. Therefore, the objective
IPTC-19787-MS 3
of this study is to develop new approach by building new models to predict RHOB while drilling using
artificial intelligence tools (ANN, ANFIS) in conjunction with drilling mechanical parameters (ROP, WOB,
T, SPP, and RPM) and conventional well-logging data in order to generate synthetic formation bulk density
log data with economical and low cost.
The ranges of the selected data are: ROP ranges from 5.81 to 65.9 ft/hr., WOB ranges from 4.6 to 35.3
kIb, RPM ranges from 58.5 to 135.9 rpm, T ranges from 1.03 to 8.02 kIb-ft, SPP ranges from 2393.7 to
3483.9 psi, GPM ranges from 195.1 to 305.23 gpm and RHOB ranges from 2.43 to 2.91 g/cm3.
Data processing
The accuracy of the prediction process relies on the quality of the data used for training the model, so it is
very important to filter and analyze the data before building the models (Hemphill et al., 2007). Therefore,
the obtained data have been filtered from any non-reasonable values like negative values and 999 values. In
addition, MATLAB codes have been used to remove outlies using different statistical methods. Moreover,
RHOB log data have been analyzed and filtered by tracking the value of the correction data. Reliable RHOB
data should have correction value varies from – 0.25 to 0.25 (Ellis and Singer 2007) and accordingly the
data points with correction value beyond this range are removed.
6 IPTC-19787-MS
The optimization process showed that the optimized network parameters for the developed ANN model
can summarized as follows:
A typical architecture of the developed network is depicted in Fig. 3. The results obtained using the
optimized parameters yielded great agreement with coefficient of determination (R2) of 0.93 and 0.91
between the actual and predicted RHOB for training and testing respectively as shown in Figs. 4a and
4b which figured out that plotted points significantly coincide with the fourty-five line indicating that the
predictions of the ANN model are greatly close to the measured RHOB values.
IPTC-19787-MS 7
Table. 4—Optimizing the cluster radius for Genfis_2 fuzzy logic type.
Comparing the results obtained from the two developed models showed that ANN outperformed ANFIS
in predicting RHOB. It is found that R between the predicted and the actual RHOB values has values of 0.96
using ANN compared to 0.95 using ANFIS for training process while R has values of 0.95 using ANN and
0.93 using ANFIS for the testing process. In addition, ANN-based model yielded a lower AAPE between the
actual and predicted RHOB values than that obtained from the ANFIS-based model. The ANN-based model
resulted in a lower AAPE of 0.67 % and 0.72 % for training and testing processes respectively compared to
the ANFIS-based model which yielded AAPE of 0.71% and 0.81%. Figs. 6a and 6b comparison between
the actual and predicted RHOB values using ANN-based model and the ANFIS-based model. Accordingly,
the ANN-based model is selected to be used for the validation process.
Figure 6a—Comaprison between the predicted and actual RHOB for the testing data using the ANN-based model.
10 IPTC-19787-MS
Figure 6b—Comaprison between the predicted and actual RHOB for the testing data using the ANFIS-based model.
Model Validation
The developed ANN model has been validated using field data for another well within the area under study.
The dataset used for the validation process has not been used during building the model. The validation
data involved 900 data points including the input parameters (ROP, WOB, T, SPP, and RPM) and the
corresponding RHOB well-log data. The prediction results from the developed ANN model using these
data, showed a significant agreement with the actual RHOB values with AAPE of 0.5 % as shown in Fig. 7.
Conclusions
In this study, two models were developed using ANN and ANFIS for predicting RHOB based on drilling
mechanical parameters measurements ROP, WOB, RPM, T, and SPP. Actual measurements (2400 field data
points for a horizontal well) are used to build the models. The findings of this work can be summarized
as follows:
1. The ANN-based model outperformed the ANFIS-based model in the accuracy of the RHOB predicted
values with AAPE of 0.72% between the predicted and actual RHOB compared to 0.81 % for the
ANFIS-based model.
2. The optimized ANN model can predict RHOB with high accuracy indicated by R of 0.95 and AAPE
of 0.72 % between the predicted and the measured RHOB values.
IPTC-19787-MS 11
3. The validation process for the ANN-based model using field data from another well confirmed its
outstanding prediction performance indicated by AAPE of 0.5 % between the predicted and actual
RHOB.
4. The developed ANN-based model can be used with high reliability to predict RHOB values with high
accuracy especially in wells where the well-logging data are not available or partially absent.
Nomenclature
ANN Artificial Neural Network
ANFIS Adaptive network-based fuzzy interference systems
AAPE Average absolute percentage error
GPM Pumping rate is gallon per minute
LWD Logging while drilling
RHOB Bulk density, gm/cm3
ROP Rate of penetration, ft/hr.
RPM Revolution per minute
SPP Stand-Pipe pressure, psi
T Torque, kIb-ft
WOB Weight on bit, kIb
Tansig Hyperbolic tangent sigmoid transfer function
Hardlim Hard-limit transfer function
Logsig Log-sigmoid transfer function
Pure-linear Linear transfer function
Elliotsig Elliot symmetric sigmoid transfer function
Tribas Triangular basis transfer function
Satlin Saturating linear transfer function
Radbas Radial basis transfer function
Trainlm Levenberg-Marquardt backpropagation
Trainr Random order incremental training with learning functions
Traingda Gradient descent with adaptive learning rule backpropagation
Trainc Cyclical order incremental update
Trainbr Bayesian Regularization backpropagation
Traingd Gradient descent backpropagation
References
1. Alger, R.P. and Raymer, L.L. Jr, 1963. Formation density log applications in liquid-filled holes.
Journal of Petroleum Technology, 15(03), pp.321–332.
2. Alkinani, H. H., Al-Hameedi, A. T. T., Dunn-Norman, S., Flori, R. E., Alsaba, M. T., & Amer,
A. S. (2019, March 15). Applications of Artificial Neural Networks in the Petroleum Industry: A
Review. Society of Petroleum Engineers. doi:10.2118/195072-MS
3. Angelini E, Ludovici A (2009) CDS Evaluation model with neural networks. J Serv Sci Manag
02:15–28. https://doi.org/10.4236/jssm.2009.21003
4. Bassiouni, Z., 1994. Theory, measurement, and interpretation of well logs. In: SPE Textbook
Series, vol. 4, 372 pp.
5. Bourgoyne, A. T. Jr, Millheim, K. K., Chenevert, M. E., & Young, F. S. Jr 1986. Applied drilling
engineering. Volume 2.
6. Ellis, D., 2003. Formation porosity estimation from density logs. Petrophysics, 44(05).
12 IPTC-19787-MS
7. Ellis, D.V., and Singer, J.M., 2007, Well Logging for Earth Scientists, 2nd edition, Springer.
ISBN: 978-1-4020-3738-2,DOI: 10.1007/978-1-4020-4602-5.
8. Falconer, I.G., Burgess, T.M., and Sheppard, M.C., 1988, Separating Bit and Lithology
Effects from Drilling Mechanics Data, Paper SPE-17191 presented at the SPE/IADC Drilling
Conference, Dallas, Texas, USA, 28 February–2 March. DOI: 10.2118/17191-MS.
9. Gatens J. M., Harrison, C. W., Lancaster, D. E., & Guidry, F. K. 1990. In-Situ Stress Tests and
Acoustic Logs Determine Mechanical Properties and Stress Profiles in the Devonian Shales. SPE
Form Eval 5 (03). SPE-18523-PA. http://doi:10.2118/18523-PA
10. González, J.W., Valdez, R., Torres, J., and Medina, F., 2018, Identification of Zones of Abnormal
Pressures and Determination of the Mechanical Properties of the Rock Through Pseudo-Sonic
and Pseudo-Density Logs in Conventional and Unconventional Reservoirs, Paper SPE-191866
presented at the SPE Argentina Exploration and Production of Unconventional Resources
Symposium, Neuquen, Argentina, 14–16 August. DOI: 10.2118/191866-MS
11. Head, A. L. 1951. A drillability classification of geological formation. In 3rd World Petroleum
Congress. World Petroleum Congress.
12. Hemphill, T., Bern, P., Rojas, J., and Ravi, K. 2007, Field Validatio of Drillpipe Rotation Effects
on Equivalent Circulatin Density, Paper SPE-110470 presented at the SPE Annua Technical
Conference and Exhibition, Anaheim, California, USA, 11–14 November. DOI: 10.2118/110470-
MS.
13. Hinton GE, Osindero S, Teh Y-W. 2006. A fast learning algorithm for deep belief nets. Neural
Comput 18:1527–1554. https://doi.org/10.1162/neco.2006.18.7.1527.
14. Jackson, C.E., and Heysse, D.R., 1994, Improving Formation Evaluation by Resolving
Differences Between LWD and Wireline Log Data, Paper SPE-28428 presented at SPE Annual
Technical Conference and Exhibition, New Orleans, Louisiana, USA, 25–28 September. DOI:
10.2118/28428-MS.
15. Jang J-SR (1993) ANFIS: adaptive-network-based fuzzy inference system. IEEE Trans Syst Man
Cybern 23:665–685. https://doi.org/10.1109/21.256541
16. Jang J-SR (1996) Input selection for ANFIS learning. In: Proceedings of IEEE 5th international
fuzzy systems. IEEE, pp 1493–1499
17. Jang J-SR, Sun Chuen-Tsai (1995) Neuro-fuzzy modeling and control. Proc IEEE 83:378–406.
https://doi.org/10.1109/5.364486
18. Lippman RP. 1987. An introduction to computing with neural nets. IEEE ASSP Magazine 4:
pages 4–22. https://doi.org/10.1109/massp.1987.1165576.
19. Mensa-Wilmot, G., Calhoun, B., & Perrin, V. P. 1999. Formation drillability-definition,
quantification and contributions to bit performance evaluation. In SPE/IADC Middle East
Drilling Technology Conference. Society of Petroleum Engineers.
20. Nakamoto P. Neural networks and deep learning: deep learning explained to your granny a visual
introduction for beginners who want to make their own deep learning neural network (machine
learning), 2017, CreateSpace Independent Publishing Platform.
21. Niculescu S. Artificial neural networks and genetic algorithms in QSAR. J Mol Struct 2003,
622:71–83.
22. Rable, B. The future is here: 3 ways AI roots itself in O&G in the surge Magazine., 2017. http://
thesurge.com/stories/future-artificial-intelligence-roots-oil-gas-industry.
23. Rao S, Ramamurti V. A hybrid technique to enhance the performance of recurrent neural
networks for time series prediction. In: IEEE international conference on neural networks, 1993,
IEEE, pp 52–57.
IPTC-19787-MS 13
24. Razi M, Arz A, Naderi A 2013. Annular pressure loss while drilling prediction with artificial
neural network modeling. Eur J Sci Res 2013, 95, pp. 272–288.
25. Reichel, N., Evans, M., Allioli, F., Mauborgne, M.-L., Nicoletti, L., Haranger, F., Rabrei, R.
2012. Neutron-Gamma Density (Ngd): Principles, Field Test Results and Log Quality Control of
a Radioisotope-Free Bulk Density Measurement. SPWLA-2012-082 presented at SPWLA 53rd
Annual Logging Symposium, 16-20 June, Cartagena, Colombia.
26. Rolon, L., Mohaghegh, S.D., Ameri, S., Gaskari, R. and McDaniel, B., 2009. Using artificial
neural networks to generate synthetic well logs. Journal of Natural Gas Science and Engineering,
1(4-5), pp.118–133.
27. Tahmasebi P, Hezarkhani A (2012) A hybrid neural networks fuzzy logic-genetic algorithm for
grade estimation. Comput Geosci 42:18–27. https://doi.org/10.1016/j.cageo.2012.02.004
28. Wahl, J.S., Tittman, J. and Johnstone, C.W., 1964. The dual spacing formation density log.
Journal of Petroleum Technology, 16(12), pp.1–411.
29. Walia N, Singh H, Sharma A (2015) ANFIS: adaptive neuro fuzzy inference system—a survey.
Int J Comput Appl 123:32–38.https://doi.org/10.5120/ijca2015905635
30. Wraight, P.D., Evans, M., Marienbach, E., Rhein-Knudsen, E. and Best, D., 1989, January.
Combination formation density and neutron porosity measurements while drilling. In SPWLA
30th Annual Logging Symposium. Society of Petrophysicists and Well-Log Analysts.
31. Wraight, P.D., Evans, M., Marienbach, E., Rhein-Knudsen, E., and Best, D., 1989, Combination
Formation Density and Neutron Porosity Measurements While Drilling, Paper B, Transactions,
SPWLA 30th Annual Logging Symposium, Denver, Colorado, USA, 11–14 June.
32. Xu, H., Li, X., Li, Y., Lei, Z., and Sui, S., 2009, A Calculation of Cuttings Lag Time for Foam
Drilling, Petroleum Exploration and Development [Shiyou Kantan Yu Kaifa], 36(4), 503–507.
33. Yagiz S, Sezer EA, Gokceoglu C (2012) Artificial neural networks and nonlinear regression
techniques to assess the influence of slake durability cycles on the prediction of uniaxial
compressive strength and modulus of elasticity for carbonate rocks. Int J Numer Anal Methods
Geomech 36:1636–1650. https://doi.org/10.1002/nag.1066
34. Yılmaz, I. and Yuksek, A.G. An example of artificial neural network (ANN) application for
indirect estimation of rock parameters. Rock Mechanics and Rock Engineering 2008, 41(5),
pp.781–795.
35. Zhang, D., Chen, Y. and Meng, J., 2018. Synthetic well logs generation via Recurrent Neural
Networks. Petroleum Exploration and Development, 45(4), pp.598–607.
36. Gowida, A., Elkatatny, S. and Abdulraheem, A., 2019. Application of Artificial Neural Network
To Predict Formation Bulk Density While Drilling. Petrophysics, 60(05), pp.660–674.
Muito mais do que documentos
Descubra tudo o que o Scribd tem a oferecer, incluindo livros e audiolivros de grandes editoras.
Cancele quando quiser.