Analytics and Automation to Improve Drilling Performance
Pradeep Ashok is a PhD in Mechanical Engineering and a Senior Research Scientist in the Rig Automation and Performance Improvement in Drilling (RAPID) Group at The University of Texas (UT) at Austin. He manages the data analytics program in RAPID that works towards developing data analytics skills in the future workforce. He actively volunteers his time for the industry and is a committee member in the SPE ATCE and the SPE/IADC Drilling conferences. He is an active participant in the Open Subsurface Data Universe (OSDU) effort that is currently ongoing. He is also the CTO of Intellicess, an oil and gas services software company he co-founded in 2010.
Abstract: Does Oil and Gas Have an AI Adoption Problem?
AI is sometimes dismissed as not usable for an actual drilling operation and is often treated with suspicion and distrust. In this presentation, we will explore if that is indeed the case, and to what degree that might be true. We will also explore what can be done to make AI broadly acceptable to the oil and gas community at large. We will explore concepts of explainability and interpretability and discuss why these are almost as important as accuracy in the development of AI systems. Finally, we will touch on a few key data analytics challenges in drilling that have the potential to substantially decrease well construction costs – both for Oil and Gas as well as Geothermal.
Paul Pastusek is a Drilling Mechanics Advisor at ExxonMobil. His expertise is in: automation, drill string dynamics, steerable systems, borehole quality, bit applications, cutting mechanics, rig instrumentation and control systems, and failure analysis. He received the 2020 SPE International Drilling Engineering Award and the 2017 GCS Regional Drilling Engineering Award. Paul has a BSME from Texas A&M University and an MBA from the University of Houston. He has 43 years’ experience redesigning drilling processes and tools to the economic limit. He is a Registered Professional Engineer, holds 42 US patents, and has delivered 41 papers and presentations on drilling technology. He is currently leading two industry efforts: upgrading the IADC Code System and creating an Open Source Drilling Community.
Abstract: Data Analytics & The Scientific Method
One of the first issues that comes up in Data Analytics is having a good ecosystem for data exploration. Most of the time is often spent getting access to clean reliable data. This investment is essential to developing long term robust capabilities.
As this is being done we need to teach Drilling/Production Operations to the analytics team before starting the analysis. This is the only way for the analyst to know what the sensors are measuring, what is data, what is noise, what time frames are important, and what intervals may be of interest. Many times we apply too many Data Scientist and not enough Engineers to a given problem. The raw data often needs engineering manipulation and derivative data calculations prior to the analytics efforts.
Statistical models can be used for prediction, but we need to create physics based models to understand the process. This is more important than prediction alone and is essential for continuous improvement.
The output needs to be an integrated application for the end user, not just a dashboard. The tools and resulting databases must be open source and extendable, not close proprietary solutions. The solutions must be useful in future decades, not just this year…archive quality.
For engineers, solutions should include analytics for design and testing as well as observational analytics. We must have analytics support for active testing of independent variables as part of the scientific method. It is essential to distinguish between correlation vs cause and effect.