"Drilling File Classifier" AI Agent

"Drilling File Classifier" AI Agent

as opposed to HI* agent on the picture above??

*(human intelligence)


Welcome to our 2nd edition of Upstrima AI Agents Newsletter

We will showcase our File Classifier Agent and why this is a foundation for any further AI Development in Oil & Gas Industry.

But first - penguins ??

Why penguins??? ??

Penguins might not be the smartest animal on Earth, but they're pretty incredible in social skills, working together to defend, hunt, navigate and even coordinate "rafts" of multiple birds to come ashore smoothly. Some penguins can filter out extraneous noise and focus on what they need to hear, a phenomenon known as the "cocktail party effect".?

What am I getting at ? One pinguin has a role and can be good at something, but lots of penguins working together can execute quite a complex task.

File Classifier is a "penguin" that plays a fundamental role in the entire chain of events for automating O&G workflows.

Challenge

We need a smart "penguin" to navigate through massive volumes of unstructured, scattered historical wells data often stored in a variety of formats and plagued by duplicates and data anomalies. Not only does this affect productivity of day-to-day tasks but it also represents a key barrier to further extraction from the big data. A conventional approach requires manual cleaning, structuring, classification and verification to prepare appropriate datasets for analysis. Despite the tremendous amount of time spent on data processing, skipping this step would lead to erroneous analysis outcome negatively impacting the decision-making process.

Solution

While reading the input data, File Classifier AI Agent compares the input files against common oil and gas data file types to split the files into categories. Subsequently, the AI agent renames the files, creates an appropriate folder hierarchy and allocates the files to optimize data search and extraction. Concurrently, it employs indexing, namespace management, and metadata extraction to segment and integrate the data into a high-efficiency vector database. This process enhances data organization and retrieval capabilities, establishing a robust foundation for subsequent analytical queries and data engineering.

Case Study

Our File Classifier Agent was presented with an unstructured data set consisting of randomly named files and different formats (e.g. Daily Drilling reports, Daily Geological reports and Daily mud reports). To confuse the Agent even further we named some files the same way (Fig 1).

Our little penguin identified the well name, type of report and date of report creation and renamed the files accordingly (Fig 2), followed by organizing them in the respective folders (Fig 3). Processing the entire dataset of a real field took mere seconds to complete with an accuracy of 91% (there were some files that our File Classifier had no clue what they were, but you can manually label those to increase accuracy).


Fig 1 Unstructured input data
Fig 2. Files automatically renamed according to their types, well names and dates

.

Fig 3. Files automatically arranged into folders per their types


Conclusion

  • Is this File Classifier AI Agent a ground breaking technology that will re-invent Oil and Gas ? Of course not.
  • Is it helpful for drilling engineers on a daily basis? For sure!
  • Is it going to be useful for other AI Agents that deal with data search, retrieval, mining and analytics ? Hell yes!!!

The key is not treat a DDR (daily drilling report) as just another pdf, like all the generic chatbots do, but as a DDR, known to the AI System.

Obviously this goes beyond ddr, dgr and dmr. Upstrima can successfully recognize LAS files, surveys, EOWRs, manuals, operational guidelines, HSE policies, etc...

And the best part: users don't have to pre-process the data in any way. Just dump all your data into Upstrima and File Classifier Agent will not only extract data but also organize all these files neatly for you.         
that famous phrase that we were all so happy repeating during the last decades: "Garbage In / Garbage Out" will soon become obsolete

Read in our next editions how we are planning to achieve that using Data QA/QC and other AI Agents working separately and together, just like a colony of penguins.


Disclaimer: no penguins were hurt in the making of the Upstrima AI Agents.


Follow as at Gain.Energy

Sign up for Beta Testing here https://www.gain.energy/


Neil Armstrong

Engineering Director at Merlin ERD Ltd

2 周

Is the file classifier penguin analogy all yours or AI inspired or generated? Either way, it's a good one!

要查看或添加评论,请登录

Gain.Energy的更多文章