BRENNEN

Industries / Industrial equipment and services / Industrial machinery / Fasteners

Cold forming

Cold forming squeezes a cut length of steel wire through a series of dies and punches at room temperature, pushing the metal into a head and a shank to make the blank for a bolt, screw, nut or rivet without cutting any metal away.

6

Milestones achieved, from 1806 to 1971.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1806

    Screw heads pressed cold

    Japy, then the largest screw maker in France, built a fly press (a hand-swung screw press) that shaped screw-head blanks cold. The Tools and Trades History Society calls it the first cold-worked press for the job.

    taths.org.uk

    Credit

    Japy

  2. 1854

    Pointed screws by machine

    Thomas Sloan's 1846 US patent covered making wood screws with a gimlet point, so the screw could start its own hole. The process was licensed in 1849 to the American Screw Company and the Eagle Screw Company, and John Sutton Nettlefold and Joseph Chamberlain bought the British rights in 1854. By 1876 The Engineer described their Birmingham works heading screw blanks from cold wire with a punch, on machines they designed themselves.

    taths.org.ukgracesguide.co.uk

    Credit

    Thomas J. Sloan; John Sutton Nettlefold; Joseph Chamberlain

  3. 1911

    Square sockets punched cold

    Peter Lymburner Robertson of Hamilton, Ontario, patented a heading tool with a square, slightly tapered punch that drives a socket into the end of a wire blank in two blows; the second blow also forms and faces the head. His patent says the metal spreads evenly on every side and fills the die "without causing defective distortion or crystallization." US patent 1,003,657 was granted in September 1911 from an application first filed in November 1907.

    patents.google.com

    Credit

    P. L. Robertson

  4. 1936

    The cross recess reaches cars

    Henry F. Phillips patented a screw with a tapering cross-shaped recess that keeps a power driver centred (US 2,046,837, granted 7 July 1936). Invention & Technology tells how engineers at the American Screw Company balked at making it until their president threatened to fire anyone who said it could not be done; the company then spent $500,000 developing a way to make it, and General Motors used the screws on its 1936 Cadillac.

    LessonA new head shape spreads only once someone works out how to make it at production speed.

    patents.google.cominventionandtech.com

    Credit

    Henry F. Phillips; American Screw Company

  5. 1944

    Blanks passed die to die

    National Machinery of Tiffin, Ohio, introduced its first cold nut former in 1944, with a transfer mechanism that turned the blank between operations. It completed a line of four-die boltmakers in 1956 and brought out five- and six-die cold formers in 1960. NASA's 1990 Fastener Design Manual says cold forming both the head and the thread is the most common way to make bolts up to 1 in (25 mm) in diameter.

    nationalmachinery.comntrs.nasa.gov

    Credit

    National Machinery Company

  6. 1971

    Six lobes replace the cross

    Bernard F. Reiland of Textron in Rockford, Illinois, patented a recess with six rounded lobes (US 3,584,667, granted 15 June 1971, from a 1966 priority filing). The patent says the corners of hexagon sockets become overstressed, reports that its test heads showed no cam-out (the driver riding up and out of the recess) at zero end load, and says punches for the rounded shape last longer than Phillips punches with their thin fins.

    patents.google.com

    Credit

    Bernard F. Reiland; Textron

Watch for

  1. Cameras inside the former

    Nedschroef Machinery showed several additions for its headers at the wire 2026 trade fair in Düsseldorf: a servo wire feed that it says holds the cut length to 0.02 mm, a "Fly-Over-Punch" transfer that starts moving the blank 25° of crank rotation sooner than a standard transfer, and a high-speed camera that keeps about four seconds of footage around each fault, which it says speeds diagnosis and cuts scrap. The report names no customer and gives no test data.

    Achieved whena fastener maker or independent party publishes scrap, speed or downtime figures from production use.

    umformtechnik.net

    Claimed by

    Nedschroef Machinery, March 2026

Open

  1. 01

    Die wear, start to finish

    A public data set that follows the dies of one multi-station cold former from new until they are replaced: the load curve of every stroke at every station, the measured size of parts at set intervals, a photo of each die face every few thousand strokes, and why each die was pulled. The data comes from a cooperating plant's own load monitors, so the person building the set handles only data.

    The brief
    Credit · unclaimed
  2. 02

    Forming sequences, scored

    A public benchmark of at least 40 cold-formed fasteners (flange bolts, socket screws, nuts, studs), each given as a drawing plus the progression its maker actually uses, meaning the shape of the blank after each station, so that any method, from rule-based programs to AI models, can be scored on whether it proposes a sequence with the same number of stations that simulates without folds or overloaded dies.

    The brief
    Credit · unclaimed

Open 01 · the brief

Die wear, start to finish

Credit · unclaimed

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

What to build

A public data set that follows the dies of one multi-station cold former from new until they are replaced: the load curve of every stroke at every station, the measured size of parts at set intervals, a photo of each die face every few thousand strokes, and why each die was pulled. The data comes from a cooperating plant's own load monitors, so the person building the set handles only data.

Why it is still open

The only public bolt-forming data set found, posted in 2020 by Haw-Ching Yang of National Kaohsiung University of Science and Technology, holds 386 readings per bolt from one pressure sensor for faults made on purpose (a blocked cavity, a broken core, too little lubricant, a wrong blank length); it does not follow die wear over time. Searches found vibration studies that classify which part a header is making and simulation-based die-life models, but no public record of a die's whole life.

ieee-dataport.org

Built with what exists

Load sensors of the kind fastener plants already fit to die holders for tonnage monitoring; a USB microscope camera for die photos; Python with scikit-learn and PyTorch for time-series models.

  1. First step

    Ask three fastener plants that run stroke monitors whether they keep stroke-level logs and would release one die's life with the part number removed.

  2. Done looks like

    At least three complete die lives from one machine, with every stroke logged, published under an open licence, and a model trained on two of the lives that predicts within 10% the stroke count at which the third die had to be pulled, compared with the plant's own replacement rule.

Open 02 · the brief

Forming sequences, scored

Credit · unclaimed

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

What to build

A public benchmark of at least 40 cold-formed fasteners (flange bolts, socket screws, nuts, studs), each given as a drawing plus the progression its maker actually uses, meaning the shape of the blank after each station, so that any method, from rule-based programs to AI models, can be scored on whether it proposes a sequence with the same number of stations that simulates without folds or overloaded dies.

Why it is still open

Programs that propose cold-forging sequences go back at least to the 1980s; a 1989 paper from the University of Padua and Teksid said such systems were "at the most, at the experimental prototype stage." Recent machine-learning work in cold forging predicts tool motion, energy or head height on a given process, not the sequence itself, and a search found no public set of fastener progressions with outcomes on which methods could be compared.

research.unipd.it

Built with what exists

Forming simulation software such as QForm or Simufact Forming, or the open-source solver OpenRadioss; a large AI model that writes simulation input files and reads the results back, for example an open-weight model such as Mistral Large 4 (see the Note "A trillion open parameters"); progressions that machine makers and papers publish as photos of the blank at each station.

  1. First step

    Collect 40 published progression photos from machine makers' brochures and trade-fair reports, and redraw each one as a set of profiles.

  2. Done looks like

    40 parts with their real progressions released openly, and at least two methods scored on how often the proposed sequence matches the real number of stations and simulates with no fold, no crack and no die load above the machine's rating.