Industries / Mining and materials / Steel and metals
Casting and rolling
Casting and rolling turn liquid steel into solid slabs, billets or strip, then press it between rolls, hot or cold, into plate, sheet, bar and wire.
5
Milestones achieved, from 1783 to 2009.
1
To watch for: announced, not yet confirmed outside.
1
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1783
Grooved rolls and puddling
Henry Cort patented steam-powered grooved rolls in 1783 and a puddling furnace in 1784. In puddling, molten pig iron is stirred to remove impurities; the ball of iron is then hammered and rolled into bars. Cort built on Peter Onions' puddling process.
Credit
Henry Cort
1949
Continuous casting of steel
In continuous casting, molten metal is poured into a cooled mould with no bottom and a solid strand is drawn out without stopping, instead of filling one ingot at a time. In May 1949 Irving Rossi carried the method over from non-ferrous metals to steel, and in 1954 he founded Concast in Zurich to sell the machines. Continuously cast steel was 59.15% of world output in 1990, by the same account, and 97.4% in 2025, by worldsteel's.
Credit
Irving Rossi
1989
Thin slab casting
Nucor's mini mill at Crawfordsville, Indiana, was the first in the world to make quality flat-rolled steel from thin cast slabs, by Nucor's account. Nucor's timeline files the entry under 1992 but says that by 1992 the plant had three years of production, which puts the start in 1989.
Credit
Nucor
2002
Strip cast from liquid
The Castrip plant at Nucor's Crawfordsville mill started up in May 2002. It casts steel directly into thin strip between two rolls, and made sheet down to 0.84 mm thick. Nucor's own history says the technology "never achieved commercial viability."
Lessona process can work at full plant scale and still not pay.
Credit
Nucor, BlueScope Steel and IHI
2009
Endless casting and rolling
Acciaieria Arvedi's ESP line at Cremona, Italy, began endless operation in summer 2009, making hot-rolled coil straight from liquid steel in one unbroken line, down to 0.8 mm thick. Online measurements at the plant, reported by the Association for Iron & Steel Technology, showed 45% less energy than conventional casting and rolling.
Credit
Acciaieria Arvedi
Watch for
A faster endless line
Primetals says the Arvedi ESP line at Zhongshou Special Steel in Hebei, China, made its first coil on 13 May 2026, five days after its first cast. It calls the line the most powerful and productive Arvedi ESP line to date, and says it is designed for strip from 0.7 to 12.7 mm thick.
Achieved whenthe mill or an outside party publishes production figures that show the claimed output, with 0.7 mm coil in regular production.
Claimed by
Primetals Technologies, May 2026
Open
01
Open breakout warning benchmark
A public set of mould temperature records from continuous casters, labelled with sticker breakouts and false alarms, and a scoring rule. A breakout happens when the thin solid shell of a strand tears below the mould and liquid steel runs out; thermocouples in the mould walls are watched for the warning pattern.
The briefCredit · unclaimed
Open 01 · the brief
Open breakout warning benchmark
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of mould temperature records from continuous casters, labelled with sticker breakouts and false alarms, and a scoring rule. A breakout happens when the thin solid shell of a strand tears below the mould and liquid steel runs out; thermocouples in the mould walls are watched for the warning pattern.
Why it is still open
Published breakout predictors are built and tested on one plant's private records. A 2022 paper tested at Bokaro Steel Plant in India, with 13 breakouts in its field trial, and its data statement reads "No data were used to support this study."
Built with what exists
Anonymized thermocouple logs from partner casters; Python time-series tools; Zenodo for hosting.
First step
Write to the authors of five published breakout papers and ask whether they can release anonymized records under a shared licence.
Done looks like
At least 50 labelled breakouts or near-breakouts and at least 1,000 hours of normal casting from at least two casters, public, with a baseline detection rate and false alarms per 1,000 heats.