What sensors are required?

For basic prediction, vibration sensors (e.g. Bosch CISS or similar, €150–300 each) and temperature sensors are sufficient. For advanced analysis, electrical current measurement can also be used.

Does it work without sensors, using only historical service records?

Yes — to start, the system works with manually entered operational values and service history. Prediction accuracy is lower but improves as more data accumulates.

Which machines can the system monitor?

Rotating machinery (motors, compressors, pumps, CNC spindles, gearboxes), as well as hydraulic systems, HVAC units, and other equipment with measurable parameters.

AI Machine Failure Prediction: Zero-Downtime Manufacturing

An unplanned manufacturing machine breakdown lasts an average of 6–12 hours. Factor in: lost production + emergency service call-out + express spare parts + knock-on downtime. On average, one breakdown costs a mid-sized manufacturer €8,000–25,000.

The Physics Before a Failure: Measurable Signals

Machines show measurable changes before they fail:

Bearings (70 % of all failures): elevated vibrations in specific frequency bands 2–6 weeks before failure
Electric motors: increased current draw = greater mechanical resistance (worn bearings, insufficient lubrication)
Gearboxes: shift in vibration frequency spectrum = damaged teeth
Hydraulics: pressure drop, elevated temperature = worn seals or pump

AI reads these signals continuously and detects anomalies before the operator notices them.

How Modulario AI Predicts Failures

Data Collection (every 15–60 seconds)

Sensors send vibration, temperature, and current data via IoT gateway to Modulario
Data is stored with a timestamp per device, per component

Baseline + Anomaly Detection

First 4–8 weeks: the system learns normal values for every machine in every operating mode
After that: detects statistical deviations and trend anomalies

Prediction with a Horizon

AI predicts:

Which component will fail
With what probability (%)
Within what time horizon (days)

Alert and Service Work Order

⚠️ PREDICTIVE MAINTENANCE — 25 Apr 2026

Equipment: CNC Mazak VCN-430 (spindle)
Component: Front spindle bearing SKF 7010
Probability of failure within 14 days: 84 %
Signal: Vibrations 2× above baseline since 20 Apr, trend increasing

Recommended action: Planned bearing replacement
Spare part: SKF 7010 ACDGA/P4A (2 units in stock)
Estimated repair time: 3 hours

Maintenance Strategy Comparison

Strategy	Cost	Downtime	Risk
Reactive (after breakdown)	Highest (emergency repair)	6–48 hours	High
Preventive (calendar-based)	Medium (unnecessary replacements)	2–4 hours/year	Medium
Predictive (AI)	Lowest	<1 hour/year	Minimal

AI Machine Failure Prediction — Breakdowns Stopped Before They Happen