Condition trends with the time left attached, baselines set per machine, and the rules that turn a prediction into a job somebody actually does.
The category
Predictive maintenance watches the condition of a machine and estimates when it is likely to fail, so the repair can be planned before it happens. The useful output is a date and a confidence, not a red light.
Three terms get used as though they were the same thing. Preventive maintenance runs on a fixed interval, every 500 hours or every six months, whether the machine earned it or not. Condition based maintenance acts when a reading crosses a threshold. Predictive maintenance reads the trend of those numbers to estimate when the threshold will be crossed, which is the part that buys you time to schedule.
All three depend on knowing the condition of the machine. In heavy industry that is exactly where the category tends to break down.
The gap
The same failure mode keeps recurring and nobody has costed what it actually takes out.
Alerts fire often enough that the team has stopped reading them.
Services run on a fixed interval whether the machine worked that month or not.
The machines most likely to fail are the ones with no sensors on them.
The fourth one is the hard one, and it is the reason most predictive maintenance projects cover a fraction of the fleet and stall there.
Capabilities
Nine areas, weighted toward what happens after a prediction rather than toward the model that produced it.
Temperature, pressure, vibration and current tracked over time, with the projection out to the alarm limit shown next to them. The useful output is not that a reading is high. It is that it will be out of range in nine days.
A failure window is only actionable next to the operation calendar. The job gets booked into the last quiet slot before the window opens, rather than into whichever morning the planner happened to be looking at.
Each unit is compared against its own history. A 2004 crane running warmer than a 2023 one is not news. The same crane running warmer than it did last month is. Fleet-wide thresholds generate alerts nobody trusts, and untrusted alerts get ignored.
Rules written in your terms: risk high and spare in stock raises the job now, risk high with no spare orders the part and derates the unit, critical to the berth plan escalates to the planner. The prediction is the easy half. This is the half that changes an outcome.
Most heavy fleets are part connected and part not, and the unconnected part is usually older and closer to failure. Field reports count as condition data here, so coverage is the whole fleet instead of the sensor-equipped share of it.
Condition, criticality and consequence ranked together, so the reliability team can argue from evidence. A minor fault on the only unit that can service a berth outranks a serious fault on a spare.
The same repair costed both ways: four hours on day shift with a part from stock, against thirty-one hours with a call-out crew, air freight and two other units standing idle. The parts bill is rarely what moves. The other three lines are.
Telematics, gateways, PLC and SCADA tags all arrive at the same machine record instead of a separate dashboard nobody opens. The point of connecting a machine is what ends up in its history.
A prediction that does not become a work order is an observation. Opsima raises the job in your CMMS, SAP or Maximo with the reasoning attached, and writes the outcome back so the next prediction is better informed.
The difference
Every predictive maintenance platform works well on a connected asset. Readings arrive, the model trends them, an alert fires. That is a solved problem, and it is why the demos are convincing.
Then you look at the actual fleet. The 2023 units are instrumented. The 2004 units are not, because retrofitting sensors onto them costs more than they are worth, and those are the units closest to failing. A predictive programme that covers the newest third of the fleet is solving the easiest third of the problem.
Opsima treats what crews say as condition data. An operator reporting that a spreader is sluggish, a fitter noting a hydraulic weep on a walk-around, a machine pulled out of service twice in a fortnight for reasons explained out loud at handover. Those are observations about condition, and they arrive daily. Written to the machine record and trended alongside telemetry, they give the unconnected two thirds of the fleet a signal it never had.
In the field
Recorded equipment status changes went from roughly 1,000 a month to roughly 14,000 over the same period, which is what made the condition history dense enough to be worth trending. Taavura's Earth Moving Division runs the same layer across quarries and infrastructure sites.
Delivery
Your CMMS, SAP or Maximo stays the system of record. Opsima adds condition trending, the policy rules and coverage of the unconnected machines, then raises the resulting jobs in your system. IT reviews and signs off the rollout in staging before it goes live.
When a predictive pilot covered a handful of assets and never went further, usually because extending it meant instrumenting everything, Opsima builds the replacement around the fleet you actually have. It runs on your servers or ours. Working software ships in weeks, and you pay once it earns its place.
Goes deeper
Where the predicted job actually gets raised, assigned and closed.
Lifecycle cost, parts buying and compliance across every site.
Every category Opsima builds, and how this fits alongside them.
The interval-based approach this builds on top of.
FAQ
Bring the failure that keeps costing you a shift. We build it on your real history and run it with you. You pay only after it goes live.