Process yield optimisation — Steel
Models that recommend set-points to lift yield and cut scrap. Delivered for a steel client in Dahej.
- Location
- Dahej
- Year
- 2020
- Duration
- 24 weeks
- Team
- 4 engineers
- +2.5%
- yield
- −21%
- scrap
- 11,807
- batches modelled
The challenge
Conditions on site — heat, dust, vibration and patchy connectivity — had defeated two earlier attempts at this in steel.
What we built
A phased rollout: a measured pilot first, then scale-out. The core runs on XGBoost with Bayesian optimisation, integrated into existing systems through Grafana.
Delivery timeline
Phase 1
Discovery sprint and site survey
Phase 2
Prototype built and tested in the Motera lab
Phase 3
Pilot on site with before/after measurement
Phase 4
Scale-out, training and handover
Highlights
- Runs on-premises; no data leaves the site unless agreed
- Documentation, drawings and source handed over at close
- Operator training delivered in Gujarati, Hindi and English
- Integrated with existing PLC, ERP or GIS systems — no rip-and-replace
Related AI projects
AILiveChemicals · Delhi NCR · 2024
Process yield optimisation — Chemicals
Models that recommend set-points to lift yield and cut scrap. Delivered for a chemicals client in Delhi NCR.
- +6.7%
- yield
- −16%
- scrap
- 729
- batches modelled
AILivePharmaceutical · Pune · 2021
Process yield optimisation — Pharmaceutical
Models that recommend set-points to lift yield and cut scrap. Delivered for a pharmaceutical client in Pune.
- +5.8%
- yield
- −22%
- scrap
- 13,996
- batches modelled
AIPilotPlastics · Surat · 2020
Process yield optimisation — Plastics
Models that recommend set-points to lift yield and cut scrap. Delivered for a plastics client in Surat.
- +6.8%
- yield
- −43%
- scrap
- 13,719
- batches modelled
AILiveGlass · Motera, Ahmedabad · 2023
Process yield optimisation — Glass
Models that recommend set-points to lift yield and cut scrap. Delivered for a glass client in Motera, Ahmedabad.
- +5.1%
- yield
- −24%
- scrap
- 18,212
- batches modelled
Have a problem that needs a machine built for it?
Tell us the constraint — line speed, accuracy, environment, deadline. An engineer will read it and reply with a straight answer, including when the answer is that you don't need us.
We reply to every enquiry within one business day.