Industries / Mining and materials / Steel and metals
Steelmaking
Steelmaking turns molten iron, sponge iron and scrap into steel by burning off excess carbon and impurities and then adjusting the chemistry.
5
Milestones achieved, from 1856 to 1969.
1
To watch for: announced, not yet confirmed outside.
1
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1856
Air blown through iron
Henry Bessemer told the British Association at Cheltenham that blowing air through molten pig iron, the high-carbon iron from a blast furnace, could make iron and steel without fuel. Early metal failed: phosphorus in ordinary pig iron stopped it making good malleable iron, and Robert Mushet saved over-oxidized "burnt" metal by remelting it with spiegeleisen, a manganese-rich iron. When makers turned the process down, Bessemer opened his own works in Sheffield in 1858.
Lessona process proven on one iron can fail on another, because the chemistry of what goes in decides what comes out.
Credit
Henry Bessemer; Robert Mushet
1878
Phosphorus removed in converter
Sidney Gilchrist Thomas and his cousin Percy Gilchrist, a chemist at an ironworks in Blaenavon, Wales, announced in 1878 that they had removed phosphorus from iron in a Bessemer converter. Edward Windsor Richards took the method up at Bolckow, Vaughan and Co in Middlesbrough. Now called the basic process, it was most valuable in continental Europe, where phosphoric iron was common, and its slag was sold as phosphate fertiliser.
Credit
Sidney Gilchrist Thomas; Percy Gilchrist
1900
Electric arc furnace
Paul Héroult made the first industrial use of the electric arc for steel at La Praz, France, on 9 October 1900. An arc furnace melts its charge with electric current jumping from carbon electrodes. By 1939 it was still kept for quality and special steels.
Credit
Paul Héroult
1952
Pure oxygen converter
VÖEST tapped the first 30 tonnes of steel from a converter blown with pure oxygen at Linz, Austria, on 27 November 1952. A second plant started at Donawitz on 22 May 1953, which gave the process its name, LD. Primetals dates the Linz start-up to 26 November 1952. Worldsteel reports that 69.4% of the world's crude steel came from oxygen converters in 2025.
Credit
VÖEST
1969
Scrap-based mini mill
Nucor's first mini mill, at Darlington, South Carolina, went into production in 1969. A mini mill melts scrap in an electric arc furnace and casts and rolls it on one small site, instead of starting from ore. Electric furnaces made 30.3% of the world's crude steel in 2025, by worldsteel's figures.
Credit
Nucor
Watch for
Steel by electrolysis
Boston Metal says it has commissioned an industrial-scale molten oxide electrolysis cell in Woburn, Massachusetts. Molten oxide electrolysis makes liquid metal from iron ore with electricity instead of coal. Its chief executive said "tonnage steel is flowing" from the cell; no quantity or outside measurement was given.
Achieved whenan independent party reports a measured tonnage of steel made this way, or a customer confirms receiving it.
Claimed by
Boston Metal, March 2025
Open
01
Open arc furnace heat records
A public, anonymized record of several thousand electric arc furnace heats (one heat is one furnace load from charge to tap). Each record holds the charge by scrap grade and weight, electricity and oxygen used, tap temperature, final chemistry and time, plus a baseline model that predicts electricity per tonne.
The briefCredit · unclaimed
Open 01 · the brief
Open arc furnace heat records
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public, anonymized record of several thousand electric arc furnace heats (one heat is one furnace load from charge to tap). Each record holds the charge by scrap grade and weight, electricity and oxygen used, tap temperature, final chemistry and time, plus a baseline model that predicts electricity per tonne.
Why it is still open
A 2024 survey of AI work in the steel industry found no paper that made its data public as an appendix, and named the confidentiality of company data as the likely reason. It found only three public datasets dedicated to steel production.
Built with what exists
Records from a partner mill; standard anonymization; Python with pandas and scikit-learn for the baseline; Zenodo for open hosting.
First step
Write a one-page anonymization plan and a data-sharing agreement, and send them to five electric-furnace operators and two university steel research centres.
Done looks like
At least 5,000 heats from at least one mill under an open licence, a field-by-field description, and a baseline error in kilowatt-hours per tonne on a held-out set of heats.