BRENNEN

Industries / Industrial equipment and services / Industrial services / Welding and fabrication

Cutting

Cutting turns plate, sheet, pipe and beams into parts of the right size and shape, with a gas flame, an electric arc, a laser beam or a jet of water.

6

Milestones achieved, from 1903 to 2015.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1903

    Steel burned in oxygen

    A gas torch cuts steel by heating it and then burning it away in a jet of pure oxygen, not by melting it. Thomas Fletcher said in 1890 that his 1888 hydrogen torch cut 6 mm plate at 75 mm a minute, but it worked by melting, and the Belgian engineer F. Jottrand's first torch did the same. In 1903 Ernst Wiss built a hydrogen-oxygen cutting torch, and in 1904 Jottrand and an engineer named Lulli built one with an extra tube of pure oxygen. C. Picard and E. Fouché, who built the first acetylene torch in 1901, adapted it for cutting three years later, in the account of Anna Pocica of Opole University of Technology. Cutting torches kept that layout: a nozzle for the heating flame and a separate one for the oxygen stream. Messer's company timeline calls Wiss's 1903 torch "the first oxyfuel cutting torch".

    LessonThe early torches that melted steel were the wrong idea; cutting worked once a separate stream of pure oxygen burned the hot metal out of the cut.

    pspaw.plin.messer-cutting.com

    Credit

    Ernst Wiss at Griesheim-Elektron; F. Jottrand and Lulli; C. Picard and E. Fouché

  2. 1911

    A machine guides the torch

    A torch held by hand shakes, which spoils the cut, and early fixes were simple carriages and guides. In 1911 the Davis-Bournonville company in New York began selling the Oxygraph, which Pocica calls the first cutting machine. A pantograph arm followed the outline of a drawing with a small motor-driven wheel, and the torch at its other end copied the path. It cut steel 75 mm thick at 150 mm a minute.

    pspaw.pl

    Credit

    Davis-Bournonville Company

  3. 1957

    Arc squeezed into a jet

    Plasma cutting forces an electric arc and a stream of gas through a narrow nozzle, which makes a jet hot enough to melt metal and fast enough to blow it out of the cut. Robert Gage of Union Carbide filed for a patent on his "arc torch and process" in July 1955, and it was granted in September 1957. The patent says the torch "has proved to be exceptionally suitable for cutting (severing) metals" when the workpiece is one of the electrodes. In 1968 Dick Couch and his professor from Dartmouth's Thayer School of Engineering found that injecting water around the arc made it narrower. Hypertherm, the company they founded, says this "virtually eliminated" two problems that had plagued plasma cutting: dross (metal that sets again on the bottom of the cut) and double arcing.

    patents.google.comhypertherm.com

    Credit

    Robert M. Gage, Union Carbide; Dick Couch and his Dartmouth professor, Hypertherm

  4. 1967

    Lasers cut with oxygen

    In May 1967 Peter Houldcroft of The Welding Institute (TWI) in Cambridge, England, cut steel sheet 1 mm thick with a focused carbon dioxide (CO2) laser beam and a stream of oxygen blown along the beam through a nozzle he had designed. The 300 W laser belonged to the Services Electronic Research Laboratory in Harlow, where Arthur Sullivan ran it, and the two published "Gas-jet laser cutting" in the British Welding Journal in August 1967. A 2002 TWI paper calls it the first experiment in laser materials processing that grew into an important industrial process. The same paper estimated that about 20,000 laser sheet-cutting systems had been installed worldwide since 1980, more than 65% of them in Japan.

    twi-global.com

    Credit

    Peter Houldcroft (TWI) and Arthur Sullivan (SERL)

  5. 1984

    Sand in the water jet

    A jet of water at very high pressure, first used by the forestry engineer Norman Franz in the 1950s, could cut only soft materials such as cardboard, paper and insulation. Mohamed Hashish, who was writing a doctoral thesis on waterjet cutting, added abrasive grit to the jet in a mixing chamber; Flow says the new jet could cut thick metal, glass and stone, and that the first commercial systems went on sale in 1984. The patent, filed by Hashish, Michael Kirby and Yih-Ho Pao for Flow Industries with priority from October 1982 and granted in March 1987, says the jet "could cut even very hard materials such as steel or glass at a rapid rate".

    flowwaterjet.comflowwaterjet.compatents.google.com

    Credit

    Mohamed Hashish, with Michael Kirby and Yih-Ho Pao, Flow Industries

  6. 2015

    Fiber lasers outsell CO2

    A fiber laser makes its beam inside an optical fiber, and its source needs little maintenance. Fiber lasers were offered for sheet cutting from about 2005 but sold slowly until several large machine builders showed fiber machines at the EuroBLECH fair in 2010; Amada says that in 2010 it became the first machine maker to develop its own fiber laser source. Industrial Laser Solutions reported in 2016 that in 2015 fiber lasers outsold CO2 lasers for the first time, and quoted a 6 kW fiber machine cutting 1/4 inch (6.35 mm) mild steel at 200 inches a minute, against 110 to 120 for a 6 kW CO2 laser. Bystronic's UK arm said it had not sold a CO2 cutting machine in Britain or Ireland since August 2015.

    laserfocusworld.comlaserfocusworld.comamada.co.jp

    Credit

    machine builders including Amada and Bystronic, and the makers of fiber laser sources

Watch for

  1. Cut edges tuned by AI

    Trumpf says its Cutting Assistant, offered for TruLaser machines of 6 kW and more, has an operator photograph a cut edge with a hand scanner. Software grades the edge on criteria such as burr (a sharp ridge of metal left on the edge), suggests new cutting settings, and gives "perfectly cut component edges" after about five rounds. Fraunhofer IPA, which developed the method with Trumpf, wrote in June 2025 that the AI reuses data already collected and can include criteria from the cut-quality standard ISO 9013, but it gave no figures for trial cuts saved or for accuracy.

    Achieved whena customer or test lab publishes how many cuts it took to reach a stated edge quality on several materials and thicknesses, with the edges measured independently.

    trumpf.comfertigungstechnik.de

    Claimed by

    Trumpf, March 2025; Fraunhofer IPA and Trumpf, June 2025

  2. Plasma that tracks its wear

    The electrode (cathode) and nozzle of a plasma torch wear out and must be replaced. Kjellberg Finsterwalde says the digital condition monitoring in its Q-series plasma systems measures the cutting current and voltage in real time, shows green, amber or red for the state of these parts, and stops the cut before a failing cathode can damage the torch. It claims an average of 30% longer life for cathode and nozzle. It presented the monitoring with its new iQ series at the Schweißen & Schneiden fair in September 2025 and again at EuroBLECH in October 2026.

    Achieved whena user or test lab publishes cathode and nozzle life and torch failures over months of cutting, with and without the monitoring.

    kjellberg.dekjellberg.de

    Claimed by

    Kjellberg Finsterwalde, September 2025

Open

  1. 01

    Plasma edges graded by photo

    A public set of plasma-cut and flame-cut edges on steel plate 6 to 50 mm thick, each photographed with a phone and measured for the two qualities that, according to the KIT and Trumpf researchers below, define a cut edge under ISO 9013: roughness and perpendicularity (how far the edge leans from square). Add a dross score and a model that grades a new edge from its photos.

    The brief
    Credit · unclaimed
  2. 02

    Edge-photo tuning for budget lasers

    An open program that suggests new cutting settings for a fiber laser from photos of test-cut edges, for the low-cost machines many small shops buy, and a public record of how many test cuts it needs to reach a target edge.

    The brief
    Credit · unclaimed

Open 01 · the brief

Plasma edges graded by photo

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public set of plasma-cut and flame-cut edges on steel plate 6 to 50 mm thick, each photographed with a phone and measured for the two qualities that, according to the KIT and Trumpf researchers below, define a cut edge under ISO 9013: roughness and perpendicularity (how far the edge leans from square). Add a dross score and a model that grades a new edge from its photos.

Why it is still open

The work so far is on laser edges. Researchers at Karlsruhe Institute of Technology (KIT) and Trumpf trained a neural network on 3,336 edges of 3 mm stainless steel from one Trumpf laser and estimated roughness from color images with a mean error of 3.6 µm, and Trumpf's Cutting Assistant grades laser edges on its own machines. For plasma, Laith Al Juboori's 2026 paper predicted dross from the cutting settings, not from images of the edge, with 84.7% accuracy over 500 plates; a 2017 conference paper by Flemmer and Tang is titled "Computer vision for automatic accurate plasma arc cutting measurement". Searches of Crossref and Zenodo in November 2026 found no public set of thick plasma or flame-cut edges labeled to ISO 9013.

publikationen.bibliothek.kit.edulink.springer.comapi.crossref.org

Built with what exists

A phone with a clip-on macro lens and a ring light; a contact roughness tester or a hired optical profilometer; an engineer's square and feeler gauges for perpendicularity; Label Studio for labels; a convolutional network trained the way the KIT paper describes; Google's Gemini Decision 1, which returns a probability for each answer on a fixed list instead of writing text, to place each edge in an ISO 9013 range (see the Note "A thousand questions, one pass").

  1. First step

    Ask a plate-cutting shop or a welding school with a CNC plasma table for 50 offcuts cut at recorded settings, then photograph and measure them.

  2. Done looks like

    1,000 edges from at least three plasma systems and one oxy-fuel machine at two or more shops, measured and released under an open license. On edges from a shop the model never saw, its range matches the measured range for at least 80% of edges, with the full table of hits and misses published.

Open 02 · the brief

Edge-photo tuning for budget lasers

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open program that suggests new cutting settings for a fiber laser from photos of test-cut edges, for the low-cost machines many small shops buy, and a public record of how many test cuts it needs to reach a target edge.

Why it is still open

Trumpf's Cutting Assistant does this only for Trumpf TruLaser machines of 6 kW and more. Philipp Wagner and colleagues at Fraunhofer IPA published a method in 2025 that reuses data from earlier cutting jobs to reach a low burr in a few trial cuts, but they built and tested it on plate 50 to 150 mm thick cut by 6 to 12 kW lasers, and no code link was found in the paper. KIT's image model was trained on edges from one Trumpf machine and one material. Kardan's 2023 conference paper calls judging a cut edge "a non-trivial task" and says a versatile image-based method is still "a significant challenge". Searches in November 2026 found no open tool, and no public count of trial cuts, for the low-cost thin-sheet lasers many small shops run.

trumpf.comarxiv.orgapi.crossref.org

Built with what exists

A 1.5 to 6 kW fiber laser cutter at a partner shop; a phone with a macro lens; the edge model from the item above; Meta's open-source Ax or BoTorch libraries for Bayesian optimization, a method that picks the next settings to try from the results so far; a contact roughness tester to check the result.

  1. First step

    Cut 30 coupons of 3 mm mild steel on one machine at settings spread across speed, focus position and gas pressure, then photograph each edge and record its settings.

  2. Done looks like

    On five combinations of material and thickness, the program reaches a stated roughness target, checked with a profilometer, in a median of five test cuts or fewer, with an experienced operator's count on the same jobs published beside it.