Data is something that this industry has an abundance of and all the data must be cleaned, processed, and interpreted, over and over until the data becomes understood and actionable to make exploration decisions on. It’s the foundation of exploration and production, and its quality and coverage are paramount to operational success. As the world’s largest provider of subsurface data, TGS is committed to providing operators with the highest quality and the latest information, techniques, technology, and applications. The quality, consistency, coverage, and volume of data that TGS possesses is what allows for a robust analytics solution.

Over the past few years TGS has worked to standardize the vast TGS data library and put it into a Data Lake hosted on the cloud. This data has been cleaned, standardized, normalized, tagged with metadata and is ready to visualize, query, and analyze. With the introduction of artificial intelligence and machine learning (AI/ML) in the industry, the importance of quality, clean, searchable and analytics-ready data sets is more important than ever. The quality and completeness of the data enable creation of AI models with high accuracy and generalization capability.

Deep learning
Deep learning based image segmentation using conventional neural networks (CNN)

“One of the things we see at Google Cloud is that it's not the company with the best algorithm that's going to win — it’s the company with the most data”

-Darryl Willis, Google Cloud, VP oil, gas, and energy sector

TGS’ vast library of global subsurface data services the entire upstream life cycle from exploration through appraisal and development including seismic, magnetic and gravity data, multibeam, coring information, digital well logs, production and completion data. There’s a lot that goes into training a model for good outputs from AI and ML, because they are only as good as the data that gets put in. “Good” data is a perfect storm of breadth – an expansive high quality library – concentrated, comprehensive, and rich data. This is why TGS has stepped into AI and ML solutions.

Combining the world’s largest digital library of well logs with machine learning, the ARLAS log prediction algorithms calculate missing log curves and fill in gaps to give complete coverage from top to bottom, standardized, and ready for optimal interpretation. The volume, extensive coverage, and high quality of data that TGS has is what differentiates the success of ARLAS from other products on the market. TGS’ prediction algorithms can turn every LAS into a complete Quad-Combo suite, using as little as one Gamma Ray, Bulk Density, Neutron Porosity, Sonic Pwave, or Deep Resistivity curve.

sonic lines
Sonic cross section lines in the Permian. Left, standard LAS cross section, Right TGS ARLAS cross section

“The ultimate goal with our current analytics ready LAS product is to get the most complete dataset available so that the operators can make better decisions in the subsurface; drill less wells, drill more productive wells, drill wells faster. All of those things go into why we chose to go down that path.”

-Rob Gibson, TGS, Director of strategy and sales, Data and Analytics, from Beneath the Subsurface, a TGS original podcast.

One of the most difficult steps when building AI solutions is finding the right problem to solve. TGS first stepped into AI and ML with a community of developers on Kaggle. Kaggle is an international crowdsourcing platform that hosts competitions where developers can take on challenges to build algorithms for cash prizes. TGS created a $100,000 challenge for this community to crack – to train an algorithm to accurately recognize salt bodies in preliminary seismic images, a process that has been done for facial recognition on Facebook or object discrimination in self-driving cars.

“3000 teams comprised of almost 4,000 data scientists across the world worked on this problem for three months and gave us more than 80,000 different solutions. We would have never gotten anything like this working day and night with the whole of TGS working on this problem.”

-Arvind Sharma, TGS, VP Data and Analytics – from Beneath the Subsurface, a TGS original podcast.

The winning algorithm was matured with TGS’ full library of seismic data, and has since undergone regularization, perturbation, and further training to evolve into SaltNet, an ensemble model that accurately identifies salt bodies in preliminary seismic imaging processes.

Salt net
Gulf of Mexico seismic data. Left, expert picked salt bodies, Right, SaltNet picked salt bodies

ARLAS and Saltnet can be found on TGS’ AI platform tgs.ai where there are live demos of each available Geoscience AI product, showcasing the enhancements that AI and ML have to offer this industry leveraging TGS’ vast data library. This platform also offers cloud computing to meet high-performance compute needs on demand, and data management solutions for easy data access.

Visit TGS.ai to dig deeper and listen to the TGS original podcast.