BRENNEN

Industries / Mining and materials / Steel and metals / Tool and die

Repair

Repair keeps dies and molds working: servicing them, stopping the press before a fault wrecks the tool, and rebuilding worn or broken steel so the tool can go back into production.

6

Milestones achieved, from 1960s to 2020.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1960s

    Arc welding rebuilds dies

    In the late 1960s toolmakers began restoring worn tool steel with gas tungsten arc welding (GTAW, or TIG), which melts thin filler wire onto the damaged surface under a shield of argon or helium. Car makers repairing their stamping dies were the first users. Over 30 years the power supplies improved until an arc could be held at 0.1 amp, but the work stayed a hand skill: in 2002 Michael Francoeur of Joining Technologies estimated that fewer than 20 craftsmen in North America could micro-weld complex tool steel parts, so most damaged tools had to be sent out to one of them.

    LessonA repair method that rests on a few people's hands makes every broken tool wait for one of them.

    laserfocusworld.com

    Credit

    automotive die shops; account by Michael Francoeur, Joining Technologies

  2. 1980s

    Sensors stop the press

    A die protection system puts sensors in and around the die and stops the press when something is wrong, such as a strip that has not moved forward (a misfeed) or a part left in the die. The first systems used mechanical switches that opened the press's stop circuit, and they could not tell when a switch or its wiring had failed. Solid-state electronics in the 1980s, mature by the 1990s, made sensors small and tough enough to sit inside a working die. By 2001 a double-thickness sensor could detect a stripper plate out of place by 0.001 inch (0.025 mm).

    thefabricator.com

    Credit

    makers of die sensors and press controls; account by Harry J. Walters in The Fabricator, 2001

  3. 1985

    Electron beam repair tried

    Joining Technologies of East Granby, Connecticut, began rebuilding tool steel with an electron beam in 1985. Filler was tacked onto the worn spot and the beam melted it inside a vacuum chamber, leaving clean welds with no sink that toolmakers liked. But the first chambers were 9-inch (23 cm) cubes, the repair work crowded out the shop's better-paid production jobs, and by 2002 the company wrote that electron beam welding was rarely used on tool steel because of its cost.

    LessonA repair process that needs the tool to fit inside a vacuum chamber cannot serve large dies.

    laserfocusworld.com

    Credit

    Joining Technologies

  4. 2002

    Lasers rebuild mold edges

    Pulsed laser welding melts filler wire as fine as 0.003 inch (0.08 mm) onto a mold with very little heat, so corners, engravings and small holes can be rebuilt without warping nearby features. The method came from joining heat-sensitive parts in the US nuclear weapons industry, which had used it since the 1970s, and laser cladding of tool steel was already in use in Europe. Precision Joining Technologies wrote in 2003 that it had repaired more than 3,500 molds for mold shops in the past five years. At the Moldmaking 2002 show in Chicago, Trumpf let visitors try its PowerWeld laser welder, and most were laying weld passes on tool steel within minutes.

    moldmakingtechnology.comlaserfocusworld.com

    Credit

    Joseph J. Kwiatkowski, Precision Joining Technologies; Trumpf

  5. 2007

    Rebuilding beats replacing

    University of Michigan researchers compared building new dies and molds with remanufacturing old ones by direct metal deposition (DMD), in which a laser melts metal powder onto the tool layer by layer under closed-loop control. In one case POM Group of Auburn Hills, Michigan, turned a truck maker's 2002-model prototype stamping die into the 2003 model by adding about 1.5 mm of metal to a small area, where a new die would normally have been built. The study found that laser remanufacturing cut cost and environmental impact together, more so as tools got larger, while noting that remanufacturing of tooling "is rarely performed".

    public.websites.umich.edu

    Credit

    W. R. Morrow, H. Qi, I. Kim, J. Mazumder and S. J. Skerlos, University of Michigan; POM Group

  6. 2020

    Cameras check every stroke

    Schuler's Visual Die Protection uses two cameras and software to look into the die after each stroke, and it stops the press if a part sits crooked or a punched slug is left behind. Alzner Automotive, a car parts supplier, has run it on three Schuler presses since the end of 2020, and its owner, Hans Martin Tekeser, says the risk of damage to its dies has fallen "to virtually zero".

    andritz.comthefabricator.com

    Credit

    Schuler (now ANDRITZ Schuler); Alzner Automotive

Watch for

  1. Repairs planned by AI

    Fraunhofer ILT says that in its AI-SLAM project a line laser and camera scan a worn tool, software compares the scan with the original CAD model and works out the deposition paths and layer thickness, and AI watching the melt pool corrects the process while it runs, so the operator need only "issue the start command". The demonstration was on a tool from the mining industry; Fraunhofer says die-casting mold repairs could be made reproducible and standardized the same way.

    Achieved whena toolmaker or die caster publishes a series of production molds repaired this way, with repair hours and tool life after repair, or an outside lab measures the repairs.

    ilt.fraunhofer.de

    Claimed by

    Fraunhofer Institute for Laser Technology ILT, June 2026

Open

  1. 01

    Die fault benchmark across presses

    A public set of force recordings from working stamping dies, each press stroke labeled with what the die shop later found: a misfeed, a pulled slug, a chipped punch, a worn cutting edge, or nothing wrong. The recordings come from several presses and dies, so a fault detector can be trained on some dies and tested on others it has never seen.

    The brief
    Credit · unclaimed
  2. 02

    Repair plans scored on scans

    A public set of 3D scans of worn and damaged stamping die and plastic mold sections, such as trim steels, draw radii, mold cores and gates, each paired with the tool's CAD model and with the repair a toolmaker actually made: what was cut away and where metal was added. Any repair-planning software, commercial, academic or AI, can then be scored on the same parts against the toolmaker's plan.

    The brief
    Credit · unclaimed

Open 01 · the brief

Die fault benchmark across presses

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 force recordings from working stamping dies, each press stroke labeled with what the die shop later found: a misfeed, a pulled slug, a chipped punch, a worn cutting edge, or nothing wrong. The recordings come from several presses and dies, so a fault detector can be trained on some dies and tested on others it has never seen.

Why it is still open

The public sets found each come from one lab machine. P2S, from TU Darmstadt, holds 2,264 strokes from one four-stage progressive die on one Bruderer press, labeled only normal or defect. RWTH Aachen's WZL set covers punch force and wear on one fine-blanking machine. Shops' own tonnage and sensor records stay private.

huggingface.codoi.org

Built with what exists

Tonnage monitors and in-die force sensors already fitted to shop presses; the shops' die repair tickets as labels; Python with tsfresh or PyTorch for a baseline detector; Hugging Face Datasets for publishing. No press work by the builder: the shops run their presses as usual and export data.

  1. First step

    Ask one stamping shop that already logs tonnage to export a month of records with its die repair tickets, and count how many tickets can be matched to strokes.

  2. Done looks like

    At least 50,000 labeled strokes from at least 3 presses and 5 dies in 2 or more shops, with at least 5 fault types of 30 or more examples each, published under an open license, and a table of the baseline detector's recall on dies left out of training.

Open 02 · the brief

Repair plans scored on scans

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 3D scans of worn and damaged stamping die and plastic mold sections, such as trim steels, draw radii, mold cores and gates, each paired with the tool's CAD model and with the repair a toolmaker actually made: what was cut away and where metal was added. Any repair-planning software, commercial, academic or AI, can then be scored on the same parts against the toolmaker's plan.

Why it is still open

Planning repairs from scans is being shown by single groups on their own parts, and the data stays private. Fraunhofer ILT demonstrated automatic scan-based path planning in June 2026. A University of Texas at El Paso thesis on scanning H13 forging dies for hybrid repair (December 2025) is embargoed until 2028. DMG Mori said in 2023 that scan-based automatic toolpaths would make its hybrid die repair for Toyota "significantly more adoptable". Deep, narrow cavities cannot be scanned directly and need a cast replica, as Marek Hawryluk and colleagues showed in 2020. A search of Zenodo found no dataset of worn-die scans.

ilt.fraunhofer.descholarworks.utep.eduengineering.compmc.ncbi.nlm.nih.gov

Built with what exists

A structured-light scanner such as a Shining 3D EinScan or a lab's ZEISS ATOS; silicone replica compound for deep cavities; CloudCompare for scan-to-CAD deviation maps; FreeCAD and Python with Open3D for repair volumes; retired dies donated by shops. Any metal deposition for the reference repairs is done by a qualified repair shop, not by the builder.

  1. First step

    Ask a die shop for three retired trim steels or mold cores with their CAD files and repair records, scan them, and publish the deviation maps.

  2. Done looks like

    At least 30 worn tool sections from at least 3 shops, each with scan, CAD model, the toolmaker's marked repair volume and photos, under an open license, plus a scoring script that reports how far a planner's repair volume is from the toolmaker's, in cubic millimeters and as percent overlap.