TotalEnergies predicts pump failure with 95% accuracy

Within the oil and gas industry, electrical submersible pumps (ESPs) are used to lift moderate to large volumes of fluid out of wells. Similar to other rotating machinery like compressors, turbines or different pumps, ESPs are operated under very specific conditions and certain parameters need regular monitoring to ensure the pumps stay in optimal condition. Vibration analysis, in particular, is used to determine if rotating equipment is working as it should.

Performing maintenance on ESPs can be very costly, as it usually requires stopping production, but pump downtime is also expensive when you add up the cost of lost output and equipment repair. In addition, ESPs themselves are big-ticket items, so regular maintenance is critical to make sure they continue to work correctly and last as long as possible.

The challenge

Many oil and gas companies take a preventive approach to ESP maintenance, replacing parts and scheduling maintenance at predetermined intervals to avoid potential breakdowns, but these schedules are based on statistical averages that don’t necessarily reflect the specific operating conditions or usage history of the pump in question. And since purchasing replacement parts and stopping production both come with a hefty price tag, this “whether it needs it or not” approach is not necessarily the most cost-effective strategy when it comes to pump maintenance.

TotalEnergies knew if they could find a way to identify and follow variables that reliably indicated a specific pump was likely to fail, they could save money by taking a predictive rather than preventive approach to maintenance, undertaking the cost of replacing parts and scheduling repairs only when necessary.

Fieldbox proposition

Fieldbox gathered contextual data from electrical submersible pumps located in four of TotalEnergie’s wells, as well as 100 sensors located in wellbores and around wellheads. It then deployed its “ESP Failure Prediction” solution for TotalEnergie’s complex onshore and offshore assets, using the history of multivariate data streams to detect the appearance of weak signals early on.

Fieldbox used machine learning algorithms to identify the warning signs of pump failure early enough to replace a part or perform maintenance before a breakdown happened, avoiding unnecessary downtime and reducing or eliminating the need for a full repair altogether.

An operational dashboard with a real-time overview of these key variables was created and is now available to operations staff 24/7, allowing them to anticipate trends and drill down into data to investigate operational issues in detail.

Results and benefits

• Save Money and Improve Employee Morale and Safety
  Improve operational efficiency and avoid subjecting your team to unnecessary stressful or emergency situations that could also pose a safety risk.
• Order Replacement Parts Ahead of Time
  Optimize your purchase price and avoid getting caught off guard by ordering replacement pieces before you urgently need them.
• Reduce the Need for Repairs
  Minimize unnecessary wear and tear on machinery so that equipment needs fixing less frequently.
• Optimize Equipment Efficiency
  Schedule maintenance for low-demand periods when equipment is less needed, and extend the lifespan of machinery by using it more efficiently.