Russian Scientists Use AI to Analyze Rock Samples Faster and More Accurately

Researchers from Rosneft’s Tyumen institute have built an AI-powered tool to speed up the interpretation of geological core data.

Russian scientists have developed a system that automates the analysis of lithological descriptions of rock core samples. The tool, called LitoText, is based on large language models (LLMs) and was created by experts at the Tyumen Scientific Research Institute under Rosneft.

Using artificial intelligence, LitoText scans geological reports and identifies 16 key physical parameters of the rock samples. These include characteristics such as color, type, texture, fracturing, porosity, and more.

Lithology specialists, who study sedimentary rocks, are helping train the system. They assist the AI in learning domain-specific terminology and identifying geological traits. As a result, the system has increased the speed of analysis by five to seven times while significantly reducing human error. It can also spot inconsistencies in research reports.

All findings are compiled into a centralized information database. As of now, the system has processed digital data from core samples totaling over 52 kilometers in depth.