An experienced pumper can hear when a compressor is wrong. The rhythm changes, a knock appears where there wasn’t one, the pitch drifts. That intuition — built over years of standing next to running equipment — has always been one of the most valuable diagnostic instruments in the field. It just couldn’t scale. One person can only stand next to one machine at a time.
Well Checked’s audio anomaly detection research began with a simple question: can that intuition be learned by a machine, and deployed to every site at once?
Born on the Compressor Skid
The work started with natural gas compressors — among the loudest, most acoustically complex environments in field operations, and also among the most expensive assets to lose to unplanned failure. The patented approach that emerged continuously captures audio and evaluates each sound multiple times through overlapping analysis windows, allowing genuine anomalies to be caught inside an environment that is, by nature, never quiet.
The key insight is that “normal” is site-specific and learnable. The system builds an acoustic fingerprint of each machine’s healthy operation — across load conditions, weather, and time of day — and then scores live audio against that baseline. A deviation isn’t just noise; it’s a signal that something in the machine has changed.
Minutes to Hours of Warning
Mechanical failures are rarely instantaneous. Bearings, valves, and rotating components typically degrade audibly before they fail functionally — often providing minutes to hours of acoustic indication before a shutdown event. That window is the difference between a planned intervention and a down unit: dispatch the right technician with the right parts, or discover the failure on the next scheduled visit.
This is predictive maintenance in its most direct form — not modeled from historical averages, but heard, live, from the machine itself.
From Research to Platform
That compressor research became the Sound layer of the Zensory.ai™ platform, extending beyond compressors to engines, pumps, fans, valves, and other facility equipment. Combined with vision and optical gas imaging, acoustic intelligence gives operators the third sense they actually use in the field — and applies it continuously, across every monitored site.
The underlying research — methodology, baselining approach, and field results — is the subject of ongoing publication. Visit our Research page for current and upcoming papers.