Service by real hours, not by a rule someone typed once.
Every machine has a service interval on paper. What actually determines when it needs work is how hard it has been run. Harvinfra reads the second number and flags service in time.
Three signals per unit.
No black box. The reasoning is deliberately narrow so a fleet manager can override the recommendation and understand exactly what changed.
The primary signal. Hours to date against the manufacturer service interval for that unit.
Whether the machine has been run near peak load or moderate load - not just how long, but how hard.
When it was last serviced, what was done, and how the machine behaved in the weeks after that service.
Generator GN-02 - flagged 12 hours before service window closes.
Load pattern accelerated the burn.
GN-02 has been running yard power at Sector 4 near peak load for eleven days straight. The load pattern signal told the model the effective wear was ahead of the raw hour count - the flag arrived seven days earlier than a pure hour rule would have.
What we predict well, and what we do not.
- Scheduled interval services on well instrumented machines
- Filter and fluid replacements tied to real running hours
- Overrun risk when a unit is being run near peak load for extended shifts
- Component level failure prediction on machines with no sensor coverage
- Root cause diagnostics from vibration or acoustic signatures
- Anything on a machine still on paper only in its first month with us
Tell us the shape of your fleet. Watch the board light up.
Pick the equipment you actually run. Set a rough count. The board rebuilds live at your scale, with an illustrative mix of healthy, watch and service due units - and a sample reallocation suggestion.
An illustrative view based on typical patterns - not a live read of your actual fleet. Real data appears once your equipment is connected.