This document presents a new technique for predicting equivalent circulating density (ECD) values while drilling without using downhole tools. The technique uses artificial intelligence (AI) models to evaluate ECD based on surface drilling parameters. Two AI techniques were used: artificial neural networks (ANN) and adaptive neuro fuzzy inference systems (ANFIS). Both models achieved high accuracy in predicting ECD values compared to actual measurements, with errors less than 0.22%. The models provide a cost-effective and real-time method to evaluate ECD without needing expensive downhole sensors. Implementing these AI models could improve wellbore pressure control and drilling operations management.

1. SPE-192282-MS
New Technology to Evaluate Equivalent Circulating Density While Drilling
Using Artificial Intelligence
Mahmoud Elzenary, Salaheldin Elkatatny, Khaled Z. Abdelgawad, Abdulazeez Abdulraheem, Mohamed Mahmoud,
Dhafer Al-Shehri, King Fahd University of Petroleum & Minerals, All SPE Members
Copyright 2018, Society of Petroleum Engineers
This paper was prepared for presentation at the SPE Kingdom of Saudi Arabia Annual Technical Symposium and Exhibition held in Dammam, Saudi Arabia, 23–26
April 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
The equivalent circulation density (ECD) is a very important parameter in drilling high pressure high
temperature (HPHT) wells and deep water wells. In formations where the margins between the formation
pore pressure (PP) and the formation fracture pressure (FP) is narrow, ECD is a key parameter. In these
critical operations, the ECD is used to control the formation pressure and prevent kicks. The recent
approaches in oilfield depend mainly on using expensive downhole sensors for providing real time values
of ECD. These tools also have operational limits which may prevent using it in all downhole conditions.
The objective of this paper is to present a new technique for predicting ECD values while drilling
without the need for downhole tools. The technique uses surface drilling parameters in a model to evaluate
the ECD values precisely using artificial intelligence (AI) techniques. The model was developed using
two AI techniques; artificial neural network (ANN) and adaptive neuro fuzzy inference system (ANFIS).
Using this technique will save cost and time by eliminating needs for using expensive, complicated
downhole tools like measurement while drilling (MWDs) and pressure while drilling (PWD).
The results obtained showed that the accuracy of the developed model is very high with error factor
less than 0.22 % comparing the actual to predicted ECD values. Implementing this model in well design
will have great impact in accurately choosing correct mud weights to safely drill the well. It will also
minimize the drilling problems related to ECD such as losses or gains especially in critical area where
margins between pore and fracture pressure is very narrow.

2. SPE-192282-MS 2
Introduction
In most HPHT and deep water wells, the margins between the formation pore pressure (PP) and the
formation fracture pressure (FP) is narrow; and here appears the importance of evaluating correct ECD
values. Actually in these critical operations, the ECD is used to control the formation pressure and prevent
kicks. When the pumps are switched off, the loss of ECD may result in underbalanced conditions. In same
time, it is not possible to increase the mud weight due to fracture pressure limitations. Here we find also
the implementation of continuous circulating system (CCS) tools which allows for the connection to be
performed while pumps are on; in order to keep ECD values controlling the formation pressure (Barantho
et al. 1995; Ataga et al. 2012).
ECD can be defined as the sum of the mud hydrostatic pressure and the annulus pressure loss acting
on the formation (Haciislamoglu 1994). The following factors affect the annular pressure loss (APL):
1. Annular clearance
2. Mud weight
3. Mud rheology
4. Annular velocity (pump rates)
5. Cutting concentration in the annulus
6. Hole depth
It can be widely viewed as two main components are affecting the ECD; the cutting portion in the
annulus expressed as equivalent static density (ESD), and the mud related parameters (Zhang et al. 2013;
Hemphill et al. 2011). Equation 1 is relating some of these parameters for predicting the ECD values
Bybee et al. 2009).
𝐸𝐶𝐷 − 𝐸𝑆𝐷(1 − 𝐶𝑎) + (𝜌𝑠𝐶𝑎 +
∆𝑝
g. 10−3𝐻
)
𝑎
(1)
Where:
Δp: the pressure losses in the annular space, MPa.
ECD: the equivalent circulating density, g/cm3
.
ESD: the equivalent static density, g/cm3
.
𝐶𝑎: the solids concentration in the annular space, %.
H: the well depth along the vertical, m.
𝜌𝑠
: the cuttings (solids) density, g/cm3
.
ɑ: a constant taking into account the measurement units, equal to 8.345.
g: the gravitational acceleration, equal to 9.8 m/s2
.
Such numerical evaluations for predicting ECD values did not take into account many other factors

3. SPE-192282-MS 3
affecting ECD while drilling, as implementing these factors in the equations will increase the error factors
while prediction (Caicedo et al. 2010; Costa et al. 2008).
Now, the oil industry is depending on using downhole tools for predicting the values for ECD and
monitoring the changes in these values and the related impact on well control issues (Erge et al. 2016;
Rommetveit et al. 2010). The main tools used now are Measurement while drilling (MWD) and Pressure
While Drilling (PWD) tools. These tools contain pressure sensors that can measure the total pressure
affecting on the bottom hole of the well, regardless of the factors controlling the ECD (Ettehadi et al.
2013; Dokhani et al. 2016).
It can give an accurate reading for ESD representing and giving a value for the cutting concentration
in the mud, and also readings for the ECD for the total pressure acting on the well during circulation. By
comparing the ESD with ECD we can get a clear view of what are the reasons for ECD change (Vajargah
et al. 2016; Osisanya et al. 2005; Lin et al. 2016).
Drilling service companies provide these tools with expensive daily rates plus the operating limits for
these tools (pressure – temperature – tool failures. Etc.). Using this model will eliminate the needs for
these tools, saving rig time and cost.
Artificial Intelligence
Artificial intelligence (AI) is the brain of the oil and gas drilling industry it became one of the most
important supports to that industry development. Artificial neural networks and fuzzy logic are common
AI techniques being applied today in oil and gas reservoir simulation, production and drilling
optimization, drilling automation and process control, and data mining. Recently, artificial neural
networks (ANN) have gained widespread popularity in many engineering fields, especially the petroleum
one as these kinds of networks have outstanding ability to solve complex and non-linear problems
(Wikipedia; Naganawa et al. 2014; Razi et al. 2013).
Artificial Neural Network (ANN)
An artificial neural network (ANN) reflects a system that is based on operations of biological neural
networks and hence can be defined as an emulation of biological neural systems. ANN's are at the forefront
of computational systems designed to produce, or at least mimic, intelligent behavior.
Figure 1. illustrates such a situation. There, the network is adjusted, based on a comparison of the
output and the target, until the network output matches the target. Typically, many such input/target pairs
are needed to train a network (Wikipidia; Naganawa et al. 2014; Razi et al. 2013).

4. SPE-192282-MS 4
Fig. 1— ANN system
Neural networks function is (cloning or reproducing) the underlying processing mechanisms that
increases the cleverness as a growing property of complex, adaptive systems and this type of system is
different from the classical artificial intelligence (AI), that aims immediately match rational and logical
reasoning. Neural network systems have been (successfully) raised and diffused to go through pattern
recognition, capacity planning, business intelligence, robotics, or intuitive problem related aspects (DHT
Technologies; Hemphill et al. 2007).
In computer science (CS), neural network had earned much steam in the last few years on data mining,
data analytics and forecasting. Data mining explains the process of finding out new patterns that has a
wide data sets and it apply this by an enormous set of methods that results out of statistics, artificial
intelligence, or database management. Data analytics is used to experiment the raw data with the aim to
analyse conclusions about that information which is different from the data mining. Data analytics is
utilized in plenty industries to permit companies to not only take successful decisions in their business or
science, but also to prove and refute the present patterns or theories (Andagoya et al; DHT Technologies).
The actual data mining function inverts an automatic (or semi-automatic) analysis of big quantities of
data and the purpose is to evolve previously unknown patterns such as groups of data records (cluster
analysis), uncommon records (anomaly detection), or dependencies (association rule mining). So, data
mining commonly concentartes on sorting through big data sets (by using sophisticated software
applications) to find undiscovered patterns and to set hidden relationships (aka extract value). On the other
hand, data analytics focuses on inference. Inference is the process of deriving a conclusion that is just
based on what the researcher have already know (DHT Technologies; Omosebiet al. 2012).
To summarize, a neural network can be known or defined as a highly parallel system eligible of solving
paradigms that linear computing can not tackle (DHT Technologies).

5. SPE-192282-MS 5
Adaptive Neuro Fuzzy Inference System (ANFIS)
ANFIS is a kind of artificial neural network that is based on Takagi–Sugeno fuzzy inference system
(Figure 2). The technique was developed in the early 1990s. Since it integrates both neural networks and
fuzzy logic principles, it has potential to capture the benefits of both in a single framework. Its inference
system corresponds to a set of fuzzy IF–THEN rules that have learning capabilityto approximate nonlinear
functions. Hence, ANFIS is considered to be a universal estimator.
Fig.2—Adaptive neuro fuzzy inference system
New Technology
The objectives of this paper are to evaluate accurate ECD values while drilling and Provide an AI model
which can be used to evaluate real ECD values while drilling by using surface (caliber) data, thus there
will be no need for down hole measurements.
Methodology
To build an AI model using ANN and ANFIS techniques, inputs and output had to be defined. In this
study I used the following surface parameters as inputs for an AI model:
a) Rate of penetration (ROP), m/hr.
b) Mud weight going into the hole (MWI), ppg.
c) Drill pipe pressure (DPP), psi.
The reason these parameters were called (surface), is they can be measured from surface without down
hole measurements. By looking to these three input parameters, it will be noticed that all other drilling

6. SPE-192282-MS 6
parameters affecting ECD values are related to one or more of these three parameters.
More than 3000 data point of previous mentioned parameters were collected from drilling 8.5” section in
an Egyptian well.
Data Filtration and Analyzing
In order to have accurate prediction from the use of AI techniques, data will need to be filtered and
analyzed before being used in the AI model. Filtration process starts with removing all random values that
cannot represent the measurements such as negative values, 999 values and null ones (Andagoya et al.
2015; Hemphill et al. 2007; Omosebiet al. 2012). Second process of filtration was the use of histogram
plot to remove the data out layers. To use histogram over the three input parameters will not be effective
as it might remove useful data, so using histogram over one parameter will filter the data in more efficient
way. From engineering point of view, ROP values will be the suitable parameter to be used in histogram
plotting filtration process. The initial histogram plot for ROP values had shown out layer over 30 m/hr
value as shown in Figure 3.
Fig.3—ROP histogram before filtering
After removing ROP out layer values greater than 30 m/hr, 2376 data points remained and the histogram
plot changed to Figure 4.

7. SPE-192282-MS 7
Fig. 4— ROP histogram after filtration
To ensure how effective is the filtration process on the quality of the data to be used AI model, data
analyzing using statistics was done as shown in Table 1. These results for statistical analysis for the inputs
values showing good quality for the data arrangement and variation which means that we can use it for
accurate prediction for ECD values using AI techniques.
Table 1— Statistics Analysis of the used data set
ROP MWI DPP
Max. 29.62 12 6920.04
Min. 0.26 10.5 4945.85
Mean 14.19 11.09 5873.93
Mode 22.6 10.5 5634
Range 29.36 1.5 1974.19
Skewness 0.21 0.43 0.33
Coefficient of
variation
0.47 0.06 0.08

8. SPE-192282-MS 8
First AI Model – ANN:
Multiple layered neural network model was created with three layers. The 2376 data points representing
ROP, MWI & DPP as inputs were randomly used as following:
• 70% for training the network.
• 15% for testing the results.
• 15% to validate the network results.
Figure 5 shows the training and testing results of predicting ECD values, while Figure 6 is showing the
predicted ECD profile of booth ANN training and testing.
Fig. 5— Predicted ECD for booth ANN training and testing compared with Field measurement

9. SPE-192282-MS 9
Fig. 6— ECD profile of booth ANN model training and testing compared with ECD from field measurement
Significant results were obtained for real predicted ECD values (Table 2), this was shown from using
correlation coefficient and average absolute error percent (AAPE) (Janjua et al. 2011; Diaz wt al. 2004)
terms to check prediction accuracy as following:
Table 2—ANN model training and testing Correlation Coefficient and AAPE
Parameter Training Testing
Correlation Coefficient 0.9971 0.9982
Average absolute
percentage error (%)
0.2252 0.2237

10. SPE-192282-MS 10
Second AI Model – ANFIS:
An ANFIS model was created using 5 membership functions, using gaussmf as input membership
function, and linear as output membership function with random selection of training and testing data
points. Figure 7 shows the training and testing results which have a good match with the measure ECD
values. Also the ECD-depth profile showed the high accuracy of the training and testing results (Figure
8) with high correlation coefficient and AAPE as shown in Table 3.
Fig. 7—Predicted ECD from ANFI Model (training and testing) compared with field measurement.

11. SPE-192282-MS 11
Fig. 8— Fig. 6— ECD profile of booth ANFIS model training and testing compared with ECD from field
measurement
Table 3—ANFIS model training and testing Correlation Coefficient and AAPE
Parameter Training Testing
Correlation
Coefficient
0.9982 0.9982
Average absolute
percentage error (%)
0.2258 0.2262
Summary
This study has been focusing on the importance of ECD evaluation and even prediction before start drilling
operations. Good estimation of ECD values provides a powerful tool for choosing accurate mud weights
for optimum well control operations. The recent approaches in oilfield depend mainly on using expensive
downhole tools for providing real time values of ECD. These tools, in addition to being expensive and
time consuming, it’s also has operation limits which may prevent using it in all downhole conditions. The
study was able to build two AI models (ANN & ANFIS), for predicting ECD values using surface
parameters without need for downhole tools. The results of these two models were highly accurate with
error factor of 0.22 %. Using these models and to be integrated with new automated rig cyber systems will
give immediate results for actual ECD values affecting the bottom hole, helping decision makers for better
control on the well.

12. SPE-192282-MS 12
Acknowledgements
The authors thank the management of KFUPM for their permission to present this paper

13. SPE-192282-MS 13
References
Andagoya Carrillo, K. I., Avellán, F. J., and Camacho, G. 2015. ECD and Downhole Pressure
Monitoring While Drilling at Ecuador Operations. Presented at the SPE Latin American and
Caribbean Petroleum Engineering Conference, Quito, Ecuador, 18-20 November. SPE-177062-
MS. https://doi.org/10.2118/177062-MS
Ataga, E., Ogbonna, J., & Boniface, O. 2012. Accurate Estimation of Equivalent Circulating Density
During High Pressure High Temperature (HPHT) Drilling Operations. Presented at Nigeria
Annual International Conference and Exhibition, Lagos, Nigeria, 6-8 August. SPE-162972-MS.
https://doi.org/10.2118/162972-MS
Baranthol, C., Alfenore, J., Cotterill, M. D., and Poux-Guillaume, G. 1995. Determination of
Hydrostatic Pressure and Dynamic ECD by Computer Models and Field Measurements on the
Directional HPHT Well 22130C-13. Presented at the SPE/IADC Drilling Conference, Amsterdam,
Netherlands, 28 February-2 March. SPE-29430-MS. https://doi.org/10.2118/29430-MS
Bybee, K. 2009. Equivalent-Circulating-Density Fluctuation in Extended-Reach Drilling. 2009.
Journal of Petroleum Technology, 61(2): 64 – 67. SPE-0209-0064-JPT.
https://doi.org/10.2118/0209-0064-JPT
Caicedo, H. U., Pribadi, M. A., Bahuguna, S., Wijnands, F. M., and Setiawan, N. B. 2010.
Geomechanics, ECD Management and RSS to Manage Drilling Challenges in a Mature Field.
Presented at the SPE Oil and Gas India Conference and Exhibition, Mumbai, India, 20-22 January.
SPE-129158-MS. https://doi.org/10.2118/129158-MS
Costa, S. S., Stuckenbruck, S., Fontoura, S. A. B., and Martins, A. L. 2008. Simulation of Transient
Cuttings Transportation and ECD in Wellbore Drilling. Presented at the Europec/EAGE
Conference and Exhibition, Rome, Italy, 9-12 June 2008. SPE-113893-MS.
https://doi.org/10.2118/113893-MS
DHT Technologies. An Introduction to Artificial Neural Networks (ANN) – Methods, Abstraction,
and Usage. Internal company report, DHT Technologies, Dripping Springs, Texas
Diaz, H., Miska, S., Takach, N., & Yu, M. 2004. Modeling of ECD in Casing Drilling Operations and
Comparison with Experimental and Field Data. Presented at the IADC/SPE Drilling Conference,
Dallas, Texas, 2-4 March. SPE-87149-MS. https://doi.org/10.2118/87149-MS
Dokhani, V., Ma, Y., and Yu, M. 2016. Determination of equivalent circulating density of drilling
fluids in deepwater drilling. Journal of Natural Gas Science and Engineering. 34:1096-1105.
https://doi.org/10.1016/j.jngse.2016.08.009
Erge, O., Vajargah, A. K., Ozbayoglu, M. E., and van Oort, E. 2016. Improved ECD Prediction and
Management in Horizontal and Extended Reach Wells with Eccentric Drillstrings. Presented at the
IADC/SPE Drilling Conference and Exhibition, Fort Worth, Texas, USA, 1-3 March. SPE-
178785-MS. https://doi.org/10.2118/178785-MS
Ettehadi Osgouei, R., Yoong, W. L. S., & Ozbayoglu, E. M. 2013. Calculations of Equivalent
Circulating Density in Underbalanced Drilling Operations. Presented at the International
Petroleum Technology Conference, Beijing, China, 26-28 March. IPTC-16601-MS.
https://doi.org/10.2523/IPTC-16601-MS
External website – https://www.wikipedia.org
Haciislamoglu, M. 1994. Practical Pressure Loss Predictions in Realistic Annular Geometries.

14. SPE-192282-MS 14
Presented at the SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana, 25-
28 September. SPE-28304-MS. https://doi.org/10.2118/28304-MS
Hemphill, T., and Ravi, K. 2011. Improved Prediction of ECD with Drill Pipe Rotation. International
Petroleum Technology Conference. Presented at the International Petroleum Technology
Conference, Bangkok, Thailand, 15-17 November. IPTC-15424-MS.
https://doi.org/10.2523/IPTC-15424-MS
Hemphill, T., Bern, P. A., Rojas, J., & Ravi, K. 2007. Field Validation of Drillpipe Rotation Effects
on Equivalent Circulating Density. Presented at the SPE Annual Technical Conference and
Exhibition, Anaheim, California, U.S.A, 11-14 November. SPE-110470-MS.
https://doi.org/10.2118/110470-MS
Janjua, I., Faizal, S. A., & Dabyah, A. A. 2011. ECD Management in a Middle East well with the aid
of Formation Pressure While Drilling. Presented at the SPE Saudi Arabia section Young
Professionals Technical Symposium, Dhahran, Saudi Arabia, 14-16 March. SPE-155414-MS.
https://doi.org/10.2118/155414-MS
Lin, T., Chenxing Wei, C., Zhang, Q., and Sun, T. 2016. Calculation of Equivalent Cirulating Density
and Solids Concentration in The Annular When Reaming The hole in Deepwater Drilling.
Chemistry and Technology of Fuels and Oils. 52(1):70-75.
Naganawa, S., Sato, R., & Ishikawa, M. 2014. Cuttings Transport Simulation Combined with Large-
Scale Flow Loop Experiment and LWD Data Enables Appropriate ECD Management and Hole
Cleaning Evaluation in Extended-Reach Drilling. Presented at the Abu Dhabi International
Petroleum Exhibition and Conference, Abu Dhabi, UAE, 10-13 November. SPE-171740-MS.
https://doi.org/10.2118/171740-MS
Omosebi, A. O., Osisanya, S. O., Chukwu, G. A., & Egbon, F. 2012. Annular Pressure Prediction
During Well Control. Presented at the Nigeria Annual International Conference and Exhibition,
Lagos, Nigeria, 6-8 August. SPE-163015-MS. https://doi.org/10.2118/163015-MS
Osisanya, S. O., & Harris, O. O. 2005. Evaluation of Equivalent Circulating Density of Drilling Fluids
Under High Pressure/High Temperature Conditions. Presented at the SPE Annual Technical
Conference and Exhibition, Dallas, Texas, 9-12 October. SPE-97018-MS.
https://doi.org/10.2118/97018-MS
Razi, M., Arz, A., Naderi, A. 2013. Annular Pressure Loss while Drilling Prediction with Artificial
Neural Network Modeling, European Journal of Scientific Research. 95(2):272-288.
Rommetveit, R., Odegard, S. I., Nordstrand, C., Bjorkevoll, K. S., Cerasi, P. R., Helset, H. M.,
Havardstein, S. T. 2010. Drilling a Challenging HP/HT Well Utilizing an Advanced ECD
Management System with Decision Support and Real-Time Simulations. Presented at the
IADC/SPE Drilling Conference and Exhibition, New Orleans, Louisiana, USA, 2-4 February.
SPE-128648-MS. https://doi.org/10.2118/128648-MS
Vajargah, A. K., Sullivan, G., & Oort, E. van. 2016. Automated Fluid Rheology and ECD
Management. Presented ath the SPE Deepwater Drilling and Completions Conference, Galveston,
Texas, USA, 14-15 September. SPE-180331-MS. https://doi.org/10.2118/180331-MS
Zhang, H., Sun, T., Gao, D., Tang, H. 2013. A New Method for Calculating the Equivalent Circulating
Density of Drilling Fluid in Deepwater Drilling for Oil and Gas. Chemistry and Technology of
Fuels and Oils. 49(5):430-438.