AI machine learning

AI machine learning

New technologies based on Artificial Intelligence are now making significant inroads into the E&P Sector, especially within the real-time, data-heavy domains of operational drilling and reservoir surveillance. AI/ML algorithms are also playing an ever more prominent role in the predictive maintenance of facilities, in reservoir modelling and production forecasting.

With access to modern web-based tools and continual increases in processing speed, combined with the versatility of cloud computing, the oil industry’s adoption of Artificial Intelligence looks set to vastly improve existing efficiencies.

During a recent project, Flare Solutions outlined one of the key differences between data driven and knowledge driven approaches to developing AI/ML algorithms, the concepts of which are also explained in an interesting article posted on MC.AI.
Data driven Machine Learning

An international energy company was conducting its own research and development into AI/ML tools to enhance data extraction and verification.  However, the tool-set required significant input for specific, narrow use-cases and was missing opportunities for data acquisition and verification due to a lack of consistent terminology within data sets.

Flare’s Sirus toolkit was deployed to allow content at scale (>200M+ files) to be selected as candidates for input to specific AI use-cases.

Flare’s Taxonomies with their standard reference terms and synonyms, for keywords, asset names and deliverables also delivered a step-change in understanding the terminology.  This meant the AI system started with a University level of industry knowledge rather than a primary school one.  Leveraging this knowledge reduced the need for complex deep learning.