Predictive maintenance with sensors: a practical guide
- Predictive maintenance decides when to intervene according to the real condition of the equipment, measured with sensors or inspections.
- It pays off on critical equipment or equipment that is costly to stop, where a breakdown costs far more than the sensor.
- Start with few variables and simple rules; an alarm that keeps going off gets ignored.
Predictive (or condition-based) maintenance consists of monitoring the real condition of a piece of equipment with sensors or inspections and acting when the data indicates that a failure is approaching, instead of intervening by calendar. It does not replace preventive maintenance: it complements it on equipment where stopping is expensive and wear can be measured.
What can be measured
| Variable | What it may indicate |
|---|---|
| Temperature | Bearings with friction, overload, cooling failures. |
| Vibration | Imbalance, misalignment or mechanical wear. |
| Pressure | Blockages, leaks, clogged filters. |
| Electrical current | Overloaded motors or motors with insulation problems. |
| Hours or cycles of use | Accumulated wear, to schedule the intervention. |
How it works in practice
- The sensor or device sends readings to a central system (for example, through an API with its own key).
- The readings are stored linked to the machine, like a historical meter.
- A rule evaluates the value: “if the temperature exceeds X for Y minutes”.
- If it is met, an alarm is triggered and a work order is opened automatically.
- The technician intervenes and records the result.
Avoid false alarms
The biggest enemy of predictive maintenance is the alarm that goes off and resets every few seconds around a threshold. It is avoided with hysteresis: the alarm is triggered when a value is exceeded and only cleared when it drops below a lower value. It also helps to require the condition to last a minimum time before alerting.
When does it pay off?
- The equipment is critical and stopping it is very costly.
- The breakdown can be anticipated with a measurable variable.
- Repairing in time is much cheaper than repairing after the failure.
If the equipment is cheap, redundant or fails without warning, classic preventive maintenance is usually more profitable.
How to start without complicating things
- Choose 1 or 2 critical pieces of equipment.
- Measure one variable that you already know anticipates its failure.
- Define a simple rule with hysteresis.
- Review the results for a month and adjust the thresholds.
- Expand only when the first rule has proved its worth.
Sensors, rules and alarms with automatic orders
Mantenplan lets you register devices with channels, store their readings in each machine’s meters, define rules with alarms and open work orders automatically.
Try MantenplanFrequently asked questions
What is the difference between preventive and predictive maintenance?
Preventive maintenance intervenes according to a calendar or a usage set in advance. Predictive maintenance intervenes according to the real, measured condition of the equipment, when the data indicates deterioration.
Do I need expensive sensors to start?
Not necessarily. Often it is enough to measure a simple variable (temperature, hours of use) on a critical piece of equipment. What matters is that the measurement really anticipates the failure.