Ongoing projects showcase how AI can be leveraged within drilling
By Linda Hsieh, Editor & Publisher
Amid the drilling industry’s digital transformation, staying competitive is requiring higher and higher levels of investment into AI technologies. In this issue of DC, we delve into this AI world and showcase several examples of recent and ongoing work to leverage the technology for better performance and better safety.
Patterson-UTI, for example, is using Hidden Markov Models (HMMs) to detect drillstring breakover. HMMs are probably already a common feature in your life, if you ever use the Siri digital assistant on your iPhone or other Apple devices. It is a statistical model that is, according to ChatGPT, foundational in many fields for tasks involving sequential data analysis and pattern recognition.
By deploying this technology, Patterson-UTI was able to create a model that can detect breakover with approximately 98% accuracy – and to do so within 1-3 seconds. Not only can this help to optimize connection times, but it also holds potential to improve other areas like torque and drag monitoring, or automatic weight on bit. Click here to read our report.
And while the industry is still working to adopt AI and machine learning, a newer type of AI – generative AI – is now starting to find its way into the drilling arena. Amazon Web Services (AWS) launched its Bedrock platform last year. It is already being used in industries like pharmaceutical, automotive and manufacturing to streamline access to historical and real-time data and could find its way into oil and gas in the near future.
Through Bedrock, organizations can access foundational models to build custom gen AI apps using company-specific data. This saves organizations from the costly and time-consuming effort of training their own models, since AWS has pre-trained them already.
In 2023, Spain-based operator CEPSA started using Bedrock in conjunction with another AI-based tool, Smart Safety Assistant, to improve its safety reporting. By combining historical data on safety incidents and real-time data, plus other contextual information like weather or near-misses involving similar pieces of equipment, a much more granular report can be generated to assist safety personnel in doing their jobs. Click here to read the article.
Another type of AI: agentic AI
Yet another type of AI is now emerging, as well – agentic AI. This field of AI represents another significant advancement as it moves beyond simple rule-based systems to more autonomous and decision-capable entities that can operate in complex and dynamic environments (again, according to Chat GPT).
DC recently spoke with eDrilling, which has just kicked off an ambitious R&D project that aims to leverage agentic AI to build what it’s calling an AI drilling agent. The goal, ultimately, is to enable fully autonomous drilling.
Agentic AI is new but evolving rapidly, according to eDrilling. The company plans to combine agentic AI with other field-proven AI- and physics-based solutions, including dynamic drilling simulators and digital twins, to create the AI drilling agent. The agent will be capable of adapting independently to dynamic drilling conditions even while it manages more repetitive drilling tasks. In effect, it will supply the “brain” to support truly autonomous drilling.
Importantly, the agent will be cognitive and have the ability to consider the most effective strategy for the entire well, not just the next stand.
While development of the AI drilling agent is in the early stages, eDrilling says it aims to have a full version for the market in approximately 2.5 years and will have a minimum marketable product for trial before then. Read the article here, and explore other AI-related articles, including one about Repsol’s InWell.ai tool (available here), in the rest of this edition. DC
