Rela AIRela AI Docs
Use Cases

Case: Gradual Drift on an Industrial Compressor

How Rela AI caught progressive degradation on a compressor and coordinated two departments

Case: Gradual Drift on an Industrial Compressor

Context

  • Industry: metallurgy
  • Equipment: industrial air compressor, 200 HP
  • Sensors: discharge temperature, oil pressure, axial vibration
  • Criticality: high — feeds the entire plasma-cutting line

The previous problem

The compressor had sporadic failures every 4-6 months. Technicians couldn't predict them because the degradation was gradual — no obvious alarm, just a slow "drift" in temperature and pressure.

  • Each failure: 16-24 hours of downtime (diagnosis + waiting for parts).
  • Cost per event: ~$25,000 USD (lost production + parts + overtime).

What Rela AI did

Month 1: learning (Level 0-1)

Rela connected to the sensors via OPC UA from a Siemens PLC. During the first month it established baselines:

  • Discharge temperature: 85-92 °C (normal).
  • Oil pressure: 4.2-4.8 bar (normal).
  • Vibration: 2.1-2.8 mm/s (normal).

Day 35: drift detected

The model detected a gradual temperature drift: from an average 90 °C to 94 °C in 5 days. Not an alarm — a trend. AHI dropped from 78 (B) to 65 (C).

Day 38: AI diagnosis

The machine agent analysed the pattern and determined: "Gradual temperature increase with steady oil pressure suggests air-filter degradation or intake-valve wear".

Day 38: multi-department coordination

The agent created TWO work orders with a dependency:

  1. WO MEC-112: "Inspect air filter and intake valve" → assigned to the mechanical team.
  2. WO ELE-045: "Verify temperature sensors" → assigned to the electrical team (dependent on MEC-112).

Both technicians received a WhatsApp with full context.

Day 40: coordinated intervention

  • Mechanical: air filter 40% blocked + minor valve wear.
  • Electrical: sensors correct, confirmed the reading was real.
  • Total planned downtime: 3 hours.

Outcome logged

Work orders closed with intervention result: filter replaced, valve cleaned, failure confirmed.

Impact

MetricWithout RelaWith Rela
Detection typeCatastrophic failureGradual drift caught early
Downtime16-24 hours3 hours (planned)
Cost~$25,000~$4,000
Cost avoided$21,000
Departments coordinatedReactive, one at a time2 in parallel with dependencies
DiagnosisTrial-and-errorAI suggested the correct probable cause

What the model learned

  • The "temperature drift + steady pressure" pattern got associated with "clogged filter".
  • Alert threshold adjusted: trigger when temperature rises more than 3 °C in 7 days.
  • The compressor moved up to Level 2 (Predictive) after this first confirmed failure.

This case shows Rela's ability to catch gradual degradation (drift) that doesn't trigger conventional alarms, and to coordinate interventions across multiple departments with dependencies.

On this page