Solutions by industry / Metals & mining
Predict cutting-torch failures, +60% asset availability
A long-product steel producer uses Wizata to predict failures on the continuous caster and plan maintenance before a torch stops the line.
The challenge
Torch failures in the hardest part of the plant
The producer wanted to cut costs by avoiding production stoppages and making maintenance more efficient. It focused on continuous casting, its most challenging process, and on the 4 cutting torches at the end of the strands.
Torch failures are costly and hard to predict:
What the team set out to do
- Predict cutting-torch failures with AI instead of relying only on operators' experience.
- Detect critical events automatically, in real time.
- Plan maintenance ahead of failures rather than handle emergencies.
- Build a solution that other plants can reuse.
Our approach
From root causes to 76 models in production
Wizata and the producer's team worked through the failure history, then put the models to work on the live caster.
Four root causes
The analysis found the 4 root causes behind most failures, the ones with the highest cost/benefit ratio to address.
76 models in real time
76 models run on the Wizata platform, scoring the 4 strands in real time and triggering predictions.
Alerts, then other plants
Operators are notified by email and SMS before a failure. The solution is ready to be replicated to other plants with little adaptation.
The outcome
Fewer stoppages, planned maintenance
Failures are predicted, so useless downtime is avoided, maintenance costs fall and the production rate holds.
4 strands in real time
Every strand is connected live to the Wizata platform.
Alerts before failure
Operators get an email and SMS on a deviation, in time to avoid a stoppage or a last-minute intervention.
Planning, not emergencies
The operations team anticipates malfunctions and adapts its planning, which saves time.
ROI from day 1
The KPIs set at the start of the mission were met, with solid ROI from the first day.
Built on the Wizata platform
Explore the platform →Next step
Stop failures before they stop your line
You don't have to do it alone. Talk to one of our engineers, or spend one or two days with our team on your own data.