Predictive Maintenance on a Rolling Mill
Maharashtra, India · Metals & Steel
Overview
Vibration and thermal monitoring across 38 drives, with alerts on signature drift rather than calendar servicing.
The challenge
Unplanned stoppages were running at 71 hours a quarter, and the two largest failures in the previous year had each been preceded by weeks of measurable vibration drift that nobody was watching.
What we did
Existing PLCs were read over OPC UA with retrofit vibration and temperature sensors on uninstrumented drives. Edge gateways push outbound only, keeping the plant network segmented per IEC 62443. Alerts fire on deviation from each asset's own learned baseline, not a generic threshold — which is what stopped the false-alarm fatigue that killed their previous attempt.
Project data
| Assets monitored | 38 drives and motors |
|---|---|
| Unplanned downtime | 71 → 26 hours per quarter |
| Failures caught pre-breakdown | 9 in first year |
| False alarm rate | Under 4% after baseline learning |
| Energy per tonne | -6.2% from load balancing |
| Payback | 5 months |
Standards applied
- IEC 62443
- OPC UA
- ISO 50001
Scope supplied
- StancorOps
Related
- StancorOps