To many SMEs, “predictive maintenance” sounds like a luxury reserved for large corporations. In reality, it is often a very down-to-earth matter: avoiding downtime, planning for spare parts, and reducing the workload on staff.
Festo: Predictive Maintenance as a measurable lever
Festo reports that a pilot project was so successful that the system is being rolled out across all Festo plants. Furthermore, a significant increase in OEE is cited, along with the fact that the investment paid for itself in less than six months.
This is a rare example of a clear, quantified benefit. And it comes from an environment familiar to many SMEs: pneumatics, automation, manufacturing, and the pressure to find skilled workers.
Why it works: knowledge of deviations, not of disasters
Predictive Maintenance is not based on ‘major damage’. Rather, it is based on small, creeping deviations:
-
Temperature drifts.
-
Vibration changes.
-
Cycle time becomes erratic.
-
Pressure curve shifts.
AI is good at detecting precisely these patterns at an early stage. Humans are good at choosing the right course of action:
-
stop immediately,
-
repair during the next scheduled shutdown,
-
just monitor,
-
or replace the component.
The SME Transfer: A start can be this small
You don’t need 500 sensors. Start like this:
-
Select a bottleneck machine (the line determines this).
-
Measure 2–4 signals (e.g. vibration, temperature, pressure, current).
-
Define 3 fault modes (e.g. bearing, valve, leak).
-
Determine the cost of a shutdown (per hour).
-
Calculate: “If we avoid two shutdowns, the project pays for itself.”
This is precisely where AI-supported research combined with human assessment shines: it links technical data, spare parts history, shift logs, delivery times and costs to form a picture that enables decision-making.
What SMEs can take away from this
-
Predictive maintenance is a cash issue, not a technical gimmick.
-
The key is bottleneck thinking: start where downtime is most costly.
-
AI provides signals. People set priorities.
-
ROI becomes clear when downtime costs are accurately calculated.
