Connect to the trend data you already log.
BMS trend logs, DCIM, EPMS or historian exports. Access is read-only. The data path, including whether Prevantec runs inside your own environment, is agreed with your IT and security team in Phase 0.
Prevantec AI reads the trend data your BMS, DCIM or EPMS already logs. It flags a chiller, pump, fan or generator when it drifts from its own normal behaviour, before it trips or alarms. You schedule the repair into a maintenance window instead of responding to a thermal event. No new sensors. No changes to your BMS.
Cooling and power plant is built N+1 or 2N so that one unit can fail without dropping load. When a unit fails without warning, you run at N until it is repaired, and a second failure in that window takes load down. Calendar-based PM does not change this, because it services units on a schedule regardless of condition. Fixed BMS alarms only fire once a value has crossed its limit. Lead time before the failure lets you repair the degrading unit while redundancy is still intact.
Typical inputs are the tags you already trend: motor current, supply and return temperatures, differential pressure, flow, run hours, and vibration or bearing temperature where it's instrumented.
UPS systems and switchgear can be assessed where trend data is already logged. Whether they're in scope is confirmed during Phase 0.
BMS trend logs, DCIM, EPMS or historian exports. Access is read-only. The data path, including whether Prevantec runs inside your own environment, is agreed with your IT and security team in Phase 0.
Each unit is compared against its own history, not a generic threshold.
Every alert shows the tag and the trend that triggered it, so your facilities engineers can check the reasoning before they raise a work order in your CMMS.
Prevantec was validated on a Canadian ethanol producer's dryer system using their existing PI historian data. It detected dust collector degradation ahead of 7 confirmed mechanical shutdown events, with a mean lead time of 79 hours.
It has not yet been deployed in a data centre. The failure modes it detected, fan, motor and bearing degradation on rotating equipment, are the same ones found in data centre cooling plant. MASE AI is a young company proving the system one asset at a time, and we're looking for a first data centre site to prove it on.
Phase 0 takes roughly 4 to 6 weeks. You get an audit of your BMS, DCIM or EPMS trend data, a walkthrough of your mechanical and electrical schematics, a criticality ranking of your cooling and backup power assets, and a written scope with a go or no-go recommendation. If your data can't support reliable detection, the report says so.
If you run cooling or backup power plant and already trend it, a 20-minute data review will tell you whether Phase 0 is worth doing.