Manufacturing CompanyProduction and sensor data

45% less unplanned downtime by planning maintenance in time

A machine fails, the production schedule falls apart, and maintenance is once again behind the facts. A major manufacturer had 45% fewer machine failures by combining sensor data with a prediction model, so maintenance could be planned before a machine failed.

45% less unplanned downtime
by planning maintenance in time
  • 45%Less downtime
  • 3 weeksWarning time
  • 92%Detection accuracy
  • 340Machines monitored
Sector
Manufacturing
1Background

The challenge

Unplanned machine downtime disrupted production. Maintenance only after a failure led to expensive repairs and delays.

Key problems

  • 01Average of 12 unplanned downtimes per month
  • 02High costs for rush parts and overtime
  • 03Production schedules were constantly disrupted
  • 04No insight into machine health until it was too late
Lasser aan het werk in een metaalbedrijf
2Methodologie

Our approach

We built a system that analyzes the machines' sensor data directly and warns the maintenance team in time.

  1. 01

    Connecting the machines

    We connected 340 machines to one central store, with data on vibration, temperature, and power consumption.

  2. 02

    Recognizing deviations

    We taught models to recognize deviating patterns automatically, without anyone having to label them first.

  3. 03

    Predicting failures

    We developed models that estimate how long a part will last, per type of failure.

  4. 04

    Alerts and overviews

    The maintenance team gets overviews that update directly and automatic alerts.

3ImpactManufacturing

Results

The predictive maintenance system was rolled out within 4 months and delivered immediate results.

  • 45%Less unplanned downtimeFrom 12 to 7 downtimes per month
  • 3 weeksAverage warning timeTime to plan maintenance before failure occurs
  • 92%Failures recognized in advanceHigh accuracy minimizes false alarms
“We can now plan maintenance instead of reacting to problems. This has not only saved costs but also reduced the workload on our team.”
Plant ManagerManufacturing Company

Technologies

Tools & platforms used

  • Python
  • Apache Kafka
  • InfluxDB
  • TensorFlow
  • Grafana
  • Azure IoT Hub

Interested in this case?

Leave your email and we'll reach out to discuss how we can achieve similar results for your business.

Rutger Geerlings, solution architect at Stratalytic

Rutger Geerlings

Solution architect

What is it about?

No sales pitch, no newsletter. We only use your details for this conversation.

Thirty minutes to see where the biggest impact is.

An honest estimate in numbers, even if the answer is that it does not pay off.