SPE-181001-MS

An Integrated System for Drilling Real Time Data Analytics

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

Ahmed M. Alsalama, Joseph P. Canlas, Salem H. Gharbi, Saudi Aramco

Copyright 2016, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Intelligent Energy International Conference and Exhibition held in Aberdeen, United Kingdom, 6-8 September 2016.

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
Drilling operations is one of the potential areas where data mining and advanced analytics technology
can be used to improve the company's bottom line. With real time drilling technology, huge amounts of
data can be generated and captured. Exploiting this data and transform it into knowledge will help oil and
gas companies to increase revenue, lower costs, and reduce risks. Building a data mining and analytics
engine that uses drilling real time data is a difficult challenge. Data transmission, retrieval, storage, data
accuracy, format of the data and the timeliness of the analysis are major considerations for developing a data
mining and analytics solution. Most of the current analytical engines require more than one tool to have a
complete system. Several tools will be needed for data transmission, retrieval, storage, and data formatting.
In addition, separate tools for building data models and visualization are required. Therefore, adopting an
integrated system that combines all required tools will significantly help an organization to address the
above challenges in a timely manner.
In this paper, we propose an architecture of an integrated system that combines all required tools of
building a data mining and analytics engine for drilling real time data. Our architecture uses a single system
that retrieve and preprocess data, apply data analytics models, extract the knowledge and visualize it, and
store the results for further analysis. The advantage of such a system is the speed of the delivery of the
analysis, which is crucial in drilling operations. The proof of concept experiments show that our proposed
architecture can address and overcome the challenges of building a data mining and analytics engine for
drilling real time data.

Introduction
Today the term data analytics draws a lot of attention, but behind the hype there's a simple story. For decades,
companies have been making business decisions based on data collected and stored in their databases.
Beyond that critical data, however, is a potential treasure trove of non-traditional, less structured data:
emails, maps, logs, sensor data, etc. that can be mined for useful information. Decreases in the cost of both
storage and computing power have made it feasible to collect this data - which would have been thrown
away only a few years ago. As a result, more and more companies are looking to include non-traditional yet
potentially very valuable data with their traditional enterprise data in their business intelligence analysis.
Oil and Gas companies including Saudi Aramco generate massive drilling real time data. It is advantageous
2 SPE-181001-MS

for businesses to be able to exploit this data and transform raw data into knowledge that can be used to help
increase revenue, cut costs and reduce risks.
In the last few years, drilling operation had become more complex. Drilling has been conducted more in
challenging environment such as deep water drilling, extended reach shale formation, etc. which requires
more sophisticated drilling technologies. To make the situation even more challenging, there is a big number
of drilling engineers and rig foremen are retiring each year. So while the operation become more difficult,
the professionals in the industry become less experienced. This led to the demand to have a drilling-task-

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

force, which consist of subject matter experts in drilling operations, whom can provide the support to these
operations, especially the critical ones. And since most of the drilling operations are in remote area, the
drilling industry introduced the drilling real time operation centers (RTOC), where sensors transmit the
drilling operation activities from rig sites to these centers, allowing those task force to provide the needed
interaction with drilling engineers and rig foremen remotely. Unfortunately, due to the human capabilities
and volume of data, the RTOC cannot handle all the complex wells. The drilling engineers and rig foremen
cannot always monitor, follow up and act on early warning signs of potential drilling problems which can
be indicated by the real time drilling data. With the sharp increase in the computational capabilities, this
critical gap can be addressed by the development of advanced computer models that can process the drilling
data in real-time and provide recommendations to prevent drilling troubles and improve efficiency.
To derive real business value from drilling real time data analytics, we need the right tools to capture
and organize a wide variety of data types from different sources, and to be able to easily analyze it within
the context of all our enterprise data. In this paper, we propose an architecture of an integrated system that
combines all required tools of building a data mining and analytics engine for drilling real time data to
predict early signs of drilling failures. In addition, the paper highlights a decoupled system developed early
in Saudi Aramco that share the same purpose of the proposed system. Also, we compare the architecture
and experiments results of the proposed system with the decoupled system.

Decoupled Architecture
Methodology
The objective is to develop an artificial intelligence (AI) engine which will monitor and analyze real-time
drilling data. The model will calculate the expected values, such as stand pipe pressure, on particular depths
and compare these values with actual drilling data. Any slight deviation in the pressures which can be
interpreted as early signs of troubles is highlighted by the model. One of the models developed is the Stand
Pipe Pressure (SPP) model and is used to identify and predict potential hole-leaning, bit nozzle plug, or
drilling string leak and twisting issues. The same methodology was applied to develop the Hook Load (HKL)
model (Al Gharbi, Al sanie, Al Zayer 2015) to identify potential drilling troubles.

Solution Architecture
The architecture of this solution start by collecting the drilling real time data which are transmitted from
the rig site sensors in Wellsite Information Transfer Standard Markup Language (WITSML) format and
stored in a centralized database. A visualization system is then used to fetch this data and display it to the
drilling engineers and Real Time Operation Centers (RTOC). It is worth mentioning that these drilling data
are massive and are transmitted every second. The artificial intelligence engines are developed as separated
entity using MATLAB technology. These engines will receive the data, process it and generate the calculated
results. Due to the real-time nature of drilling operations, the model must process the data fully and produce
the results in timely manner. A custom written centralizer software was written to work as bridge and
controller between the real-time database and the artificial intelligent engines. This centralizer software
will pull specific parameters of the real-time data from database, reformats the WITSML to MATLAB
compatible format, then does preliminary quality checks on the data. It would then push these data to the
SPE-181001-MS 3

intended engine such as the SPP model. Once the engine produces the calculated results, the centralizer
software will transform the data back to WITSML, and push it to be stored in database.
From the users point view, the application is seamless. They will visualize the raw real-time data,
associated with it the calculated ones so they can compare actual vs. calculated values. If there values start
to deviate then this can be interpreted as early sign of drilling troubles and had to be acted upon.

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

Figure 1—Decoupled solution architecture

Tools
Oracle Database was used as the database to store the drilling real-time data. MATLAB was used to develop
the artificial intelligent engines. Java was used to develop the centralizer software, which process the data
between the real-time database and the artificial intelligent engines.

Results
This architecture was successfully deployed and was able to process drilling real time data and predict
potential drilling troubles in a timely manner. It managed to generate warning signs and alerts up to more
than an hour in advance. This provided the drilling engineers and operation crew with much bigger window
to mitigate the problem and resolve it before the issue worsen such as in a stuck-pipe condition. In addition,
this architecture is capable of integrating other artificial intelligent engines even black-box ones.

Limitation
Although, this architecture was successfully deployed, it has some limitations. Since this architecture is
comprised of three different decoupled components (database, artificial intelligent engines, and centralizer),
there is inherent lag in the processing time. This lag is due to the need to transform and transmit
massive amounts of drilling real-time between these components in different formats. Moreover, these three
components were developed in three different technologies and were deployed separately. This introduced
support and compatibility issues. The whole setup needs continuous and detailed analysis before updating
any component. In case of failures, it took the support team quite a time to identify the failing component
4 SPE-181001-MS

and fix it. Due to the criticality of the drilling operation, any delay can mean failure to identify a potential
drilling trouble and prevent it. Since this environment will evolve to host more models, which will make
the situation more complex, there was a need to revisit the architecture.

Integrated/In-Database Architecture
Methodology

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

We have replicated two drilling predictive models (Stand-Pipe Pressure and Drag on Bit) that were
developed using the Decoupled Architecture as highlighted in the previous section using an integrated
architecture. The objective of these models is to predict early signs of drilling failures such as stuck pipe, and
the drag on the drill bit. The drilling data used in the models were collected in real-time using sensors which
measure and log data while drilling. These measures were used by the developed models to generate the
predicted drilling parameters which will be used to identify drilling failures. Using the same data allowed
us to analyze, validate and compare the results with the previously developed models.

Solution Architecture
The system highlighted in the Decoupled System section proposes a solution for drilling real time data
analytics which requires several tools to be integrated in order to have a complete system. To prepare the
data, a number of tools are needed to retrieve, clean, merge and transform the raw data. This preprocessing
step can be tedious and time-consuming particularly when using different tools. The transformed data is then
used by the data analytics software to generate models used to solve particular drilling real time problems.
Finally, the results are presented using the visualization tool.
Developing a data analytics system requires several steps as showing in Figure 2, and the solution
proposed in the Decoupled System section requires independent tools for each step which has several
limitations as highlighted in the previous section.

Figure 2—Data analytics process

SPE-181001-MS 5

The proposed architecture in this section is an integrated system developed that combines all required
tools for building a data mining and analytics engine for drilling real time data. An example of this system
is one developed by a vendor (Oracle 2012). The Integrated / In-Database Architecture (Oracle 2012) uses a
single system that retrieves and preprocesses the data as well as applies the data analytics models, extracts the
knowledge and visualizes it. The proposed architecture embeds a wide-range of data manipulation, mining
and analytics functions inside the database where the data is stored. Coupled with the power of popular
data retrieval and statistical computing languages such as SQL and R Language, the proposed architecture

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

eliminates data movement and duplication, maintains security and minimizes latency time from raw data
to valuable information.

Figure 3—Integrated/in-database solution architecture

Tools
PL/SQL (Oracle 2014) was used for the data preparation. Oracle Advanced Analytics (Oracle 2012) was
used to create the two models. Specifically, Support Vector Machine (Campbell, Ying 2011) – Non Linear
Regression (Ritz, Streibig 2008) were adopted as data mining algorithms. The visualization was also done
using Oracle Advanced Analytics (Oracle 2012).

Limitations
As with any integrated system, there is always a limitation that some of the components are not always
the best of the breed, e.g. visualization. Typically, integrated software has longer product cycles and makes
the wait for new features longer. On the other hand, integrating this type of architecture with a third party
product is not an easy task.

Experiments and Results

This section presents experimental studies conducted using the both architectures explained in the previous
sections: the Decoupled Architecture and Integrated Architecture. The Stand Pipe Pressure (SPP) model was
used as an analytical model. The datasets containing real time drilling data from two different oil wells were
used for the SPP models in both architectures. This allowed their accuracy and performance be measured
and compared relatively to each other.
6 SPE-181001-MS

Stand Pipe Pressure (SPP) Model

As mentioned in an earlier section, the objective of this model is to predict and identifies early signs of
drilling troubles such as stuck pipes or drilling string failures. Specifically, the model compares the stand
pipe pressure with the mud pump flowing pressure; if they don't match the same pattern, the model will
fire alarms as showing in Figure 4.

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

Figure 4—Stand pipe pressure (SPP) model

Accuracy
Real time drilling data from Well A was used for this study because there was an actual SPP alert that
happened and acted on by the Drilling crew. The SPP models in both architectures must be able to detect
this beforehand.
Using this data, the SPP model developed using the Decoupled Architecture was able to identify two (2)
events that could be the start of this drilling trouble. These were a 135 psi pressure decrease at 02:06 PM
and a 680 psi pressure decrease at 25 minutes later as shown in Figure 5.
SPE-181001-MS 7

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

Figure 5—Events detected by the decoupled architecture

The same SPP model running in the Integrated/In-Database Architecture was able to replicate the same
results and identified the same two (2) events as shown in Figure 6. The figure above shows that results
produced by the Integrated/In-Database Architecture is accurate as the results produced by the Decoupled
Architecture.
8 SPE-181001-MS

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

Figure 6—Results of applying the SPP model developed by both architectures for well A

The events identified by the model before the actual alert was detected at 3:14 PM by RTOC as shown in
Figure 7. In a real drilling situation, this would have given the Drilling crew more than an hour to investigate
and take corrective actions to prevent a possible drilling trouble.
SPE-181001-MS 9

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

Figure 7—Alert sent by RTOC

Figure 8 shows the results of applying the SPP model developed by the two architectures for the same
dataset for another well (Well B).
10 SPE-181001-MS

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

Figure 8—Results of applying the SPP model developed by both architectures for well B

Performance
The previous section shows that the both implemented architectures have the same level of accuracy. But,
the question is: how about the performance? As mentioned, the Decoupled Architecture has three different
components: database, artificial intelligent engines, and centralizer. Implementing these three components
with different tools and technology introduce a performance issue. On the other hand, the Integrated/In
Database Architecture overcomes the performance issue as it combines all required tools for building a data
mining and analytics engine.
In these studies, the dataset for each oil well contained 1.2 M records and using the Integrated/In-Database
Architecture shows significant advantage in performance over the Decoupled Architecture.
The result as the following:
SPE-181001-MS 11

Table 1—Performance result

Downloaded from http://onepetro.org/SPEIE/proceedings-pdf/16IE/All-16IE/SPE-181001-MS/1412556/spe-181001-ms.pdf/1 by Saudi Aramco, Baraket Mehri on 21 February 2022

Figure 9—Performance result in seconds

Conclusion
In conclusion, the Integrated/In-Database Architecture has been proven to produce comparable results as
the traditional Decoupled Architecture. The main benefit of this architecture is the significant improvement
in the processing time because of easier and faster data preparation. This advantage is critical especially for
time sensitive applications such as real time drilling. There are some drawbacks of this architecture such
as not being able to use the best-of-breed component, etc. However, the significant performance of this
architecture outweighs these disadvantages.

References
Al Gharbi, S. H., Al sanie Faisal S., & Al Zayer, M. R. (2015, September 15). Automated Real time Data Cleansing
and Summarization; Case Study on Drilling Hook load Real time Data. Society of Petroleum Engineers. doi:
10.2118/176755-MS
Oracle. 2012. Oracle Data Mining 11g Release Competing on In-Database Analytics. Oracle, February 2012, http://
www.oracle.com/technetwork/database/options/advanced-analytics/odm/twp-data-mining-11gr2-160025.pdf (acces-
sed November 2014)
Oracle. 2014. Database PL/SQL Language Reference. December 2014 https://docs.oracle.com/cd/E11882_01/
appdev.112/e25519.pdf (accessed November 2014)
Ritz, C., Streibig, J. C. 2008. Nonlinear Regression with R, first edition. New York: Springer-Verlag
Campbell, C., Ying, Y. 2011. Learning with Support Vector Machines, first edition. : Morgan & Claypool