BRENNEN

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

Forming and fitting

Forming bends, rolls and punches cut parts into shape; fitting sets the parts in their drawn positions and holds them with clamps or small tack welds so they can be welded.

5

Milestones achieved, from 1882 to 1983.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1882

    The cornice brake

    Before bending machines, sheet metal was hammered by hand over stakes, anvils and molds. Steve Benson writes in The Fabricator that the first "brake" as we know it was the cornice brake, patented in 1882, in which a hand-operated leaf bent a clamped sheet along a straight line. Its descendants are the leaf brakes, box-and-pan brakes and folding machines still in use. Powered press brakes, which press a punch into a die, followed in the early 1920s with flywheel drives, then hydromechanical and hydraulic brakes in the 1970s and electric brakes in the 2000s, by Benson's account.

    thefabricator.com

    Credit

    the 1882 patentee is not named in the source; history as told by Steve Benson, The Fabricator

  2. 1970

    Two halves that missed

    At 11:50 am on 15 October 1970 a 112 m steel span of Melbourne's West Gate Bridge, span 10–11, collapsed during construction and killed 35 men. The contractors had built the span's box girder as two half-girders, jacked them onto the piers and then joined them, a method that, by Donald Charrett's review of the commission's report, had not been used anywhere else under similar conditions. Where the halves met, their camber (the planned upward curve) differed by 110 mm. Ten 8-tonne blocks of kentledge (dead weight) were placed to force one side down, the thin steel buckled, and bolts were then removed to ease the buckle. The royal commission, chaired by Justice E. H. E. Barber, reported in July 1971 that the span's design was the principal factor, followed by the "unusual" erection method, and that "error begat error".

    LessonWhen large fabricated pieces do not meet, the fix is itself an engineering decision; forcing them together with weights and loosened bolts turned a misfit into a collapse.

    createdigital.org.auresearchdata.edu.aumtecc.com.au

    Credit

    a failure; investigated by the Royal Commission into the Failure of West Gate Bridge

  3. 1971

    Punching under number control

    A turret punch press carries many punches and dies and punches holes and shapes into sheet one stroke at a time. Amada says that in 1971 it set up a technology development company in Seattle, Washington, and developed a numerically controlled turret punch press, which it says created a new market in Japan, where the short name "tarepan" became common.

    amada.co.jp

    Credit

    Amada

  4. 1977

    A machine bends whole panels

    On 19 September 1977 Guido Salvagnini, working near Milan, was granted his first patent, for a "Machine for the production of rectangular sheet metal panels with bent edges". His panel bender used one bending tool, a pair of counterblades, a manipulator that moves the sheet, and a mechanical programmer to coordinate them. The Fabricator's 2017 report says it brought flexibility and automation to bending. Salvagnini Maschinenbau in Austria built its 4,000th panel bender in late 2025, for the lift maker CMAlifts.

    thefabricator.comfabricatingandmetalworking.com

    Credit

    Guido Salvagnini

  5. 1983

    The brake corrects springback

    Bent steel springs back a little when the punch lifts, so press brake operators found the right depth by "making several trial bends". A team at the Massachusetts Institute of Technology (MIT) patented a control, filed in 1980 and granted in October 1983, that senses the ram's position and the force on the sheet during the stroke and works out where to reverse, removing "trial and error procedures". In 2001 Wim Serruys described adaptive bending in The Fabricator: the machine measures the bend angle during forming, for example with a laser, and corrects the stroke, because sheets vary in thickness and strength and the same depth does not give the same angle.

    patents.google.comthefabricator.com

    Credit

    David C. Gossard, David Hardt, Kim A. Stelson and colleagues at MIT

Watch for

  1. Robot bending without teaching

    Robots that load and bend parts at a press brake usually have to be taught their moves. LVD says the software for the robotic bending systems it launched in September 2025 programs both the robot and the press brake, chooses where the gripper holds the part and finds a bend sequence free of collisions, so that "no robot teaching or programming experience is required".

    Achieved whena fabricator or an outside study publishes, for a run of new part numbers, the time from CAD file to first good part and the share of programs that ran without manual correction.

    machinery-market.co.uk

    Claimed by

    LVD, September 2025

Open

  1. 01

    Fit-up checked against CAD

    A phone or tablet app that scans a tacked-up weldment before it is welded, matches the scan to the CAD model and flags any part that is missing, out of place by more than a set distance, or turned the wrong way, so the fitter can fix it before welding.

    The brief
    Credit · unclaimed
  2. 02

    First bend right, open data

    An open log of real press brake bends from many shops, each with the sheet's material certificate (thickness and strength), the punch and die, the programmed ram depth and the measured angle, and a model that predicts the first-bend angle on brakes that have no angle sensor.

    The brief
    Credit · unclaimed

Open 01 · the brief

Fit-up checked against CAD

Credit · unclaimed

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

What to build

A phone or tablet app that scans a tacked-up weldment before it is welded, matches the scan to the CAD model and flags any part that is missing, out of place by more than a set distance, or turned the wrong way, so the fitter can fix it before welding.

Why it is still open

Comparing a 3D scan with the design is routine in measurement labs, and it has been tried on large welded steel: Zhu and colleagues (Buildings, 2026) used laser scans for a virtual trial assembly of welded box-section bridge arch ribs and found assembly conflicts and weld gap anomalies. The handheld tool found for shop joints, Servo-Robot's Wiki-Scan 2.0, measures single joints (root and face opening, gap, mismatch and bevel angle) and finished weld beads, not where each part sits in an assembly. Anthony Franze's 2024 PhD thesis at Queensland University of Technology set out a framework for augmented reality in custom fabrication shops from interviews and workshops, without a measured test of error catching. Searches in November 2026 found no published trial that counts how many fit-up errors a scan-to-CAD check catches on small tacked weldments before welding.

api.crossref.orgthefabricator.comeprints.qut.edu.au

Built with what exists

A handheld structured-light 3D scanner such as a Revopoint or Creality model, or photogrammetry from phone photos; the weldment's STEP file; Open3D to register the scan to the CAD model and measure each part's offset; a welding school or fabrication shop that supplies tacked test pieces.

  1. First step

    Get the CAD file and one tacked sample of a simple bracket from a local shop, scan it, and measure how far the scan-to-CAD match puts each part from its drawn position.

  2. Done looks like

    50 tacked weldments of 5 to 20 parts each, with errors seeded by a third party (a part shifted 3, 5 or 10 mm, flipped, or left off). The app's catch rate and false alarms per weldment are published beside an experienced fitter's tape-measure check of the same pieces, and the scans and CAD files are released.

Open 02 · the brief

First bend right, open data

Credit · unclaimed

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

What to build

An open log of real press brake bends from many shops, each with the sheet's material certificate (thickness and strength), the punch and die, the programmed ram depth and the measured angle, and a model that predicts the first-bend angle on brakes that have no angle sensor.

Why it is still open

Brakes with adaptive bending measure and correct the angle themselves, but many shops find the depth by trial bends. The published prediction models found were trained on finite-element simulations, not on shop bends: Guo and Tang's 2017 neural network and Khadra and El-Morsy's 2016 kriging model. Searches of Zenodo and Crossref in November 2026 found no public log of shop bends.

api.crossref.orgzenodo.orgthefabricator.com

Built with what exists

A digital protractor, or a camera angle reader such as sam3-angle, an open-source project that measures press brake bend angles in photos with Meta's SAM 3 model; a shared spreadsheet or small web form for each shop; the numbers from each sheet's mill test certificate; gradient-boosted trees (XGBoost) as a baseline model.

  1. First step

    Log 200 bends on one press brake over a month, with the material certificate for each sheet, and plot the angle error against sheet thickness.

  2. Done looks like

    5,000 logged bends from at least ten shops and three steel grades, released under an open license, and a model whose first-bend predictions on bends from a shop it never saw fall within 0.5° of the measured angle for at least 80% of bends, compared with the brake's own table.