Russian Researchers Develop Neural Network to Prevent Downhole Pump Failures
The system can warn of a malfunction 10 to 15 days before it occurs.
Researchers at the Astrakhan State Technical University's Advanced Engineering School "Higher School of Oil" (Vysshaya Shkola Nefti), in collaboration with the ITMO University Advanced Engineering School and Tatneft-Dobycha, have developed a new digital platform called ARGUS-USHGN. It is designed for the timely prediction of failures in sucker-rod pumps.
The system is powered by artificial intelligence, which analyzes telemetry data transmitted to the well management system to identify potential signs of failure. In some cases, the platform was able to warn of a possible breakdown 10 to 15 days in advance. This allows scheduled maintenance to be carried out in time and prevents unplanned production shutdowns. Notably, the system does not require additional equipment for deployment — it works with existing infrastructure.
Sucker-rod pumps are used in more than 40,000 wells, accounting for approximately two-thirds of the country's active well stock.
