Predictive maintenance for oil and gas: pumps, compressors and rotating equipment

The machines whose failure stops production or forces an expensive call-out. That usually means reciprocating and screw compressors at compressor stations, centrifugal and positive displacement pumps on transfer and injection service, rod lift and electric submersible pumps on wells, and the drivers, gearboxes and cooler fans around them.

Each gets models for its own failure modes. A compressor valve going bad looks nothing like a worn pump bearing, so we model each mode separately, using vibration spectra, bearing and discharge temperatures, suction and discharge pressures, rod load where it is measured, and run hours.

Yes, and that is where most projects start. Years of tagged data often already hold the early signs of past failures; the work is aligning tags with the maintenance records that say what actually broke, then testing whether a model would have warned in time.

Gaps are normal. Polling rates set for operations can be too slow for vibration, so where a failure mode needs higher-frequency data we add a sensor or edge device rather than pretend the historian has it. The plumbing behind this is covered on our data infrastructure page.

Remote sites run the models at the edge. Data is processed on location, alerts go out over whatever link the site has, whether cellular, radio or satellite, and records are stored and forwarded when the link drops.

Any device placed in a classified location, such as a Class I, Division 1 or Division 2 area, or a zoned area under the zone system used in the United States and internationally, must carry the right rating and go through your permit-to-work process. We design sensor placement, ratings and installation with the operator's safety team and qualified electricians, and nothing goes on equipment until that team has signed off. Our energy, oil and gas industry page covers the rest of our work in the sector.