BRENNEN

Industries / Mining and materials / Chemicals and petrochemicals / Specialty chemicals / Adhesives and sealants

Application equipment

Application equipment is the tools and machines that put an adhesive or sealant in place: hand guns and cartridges, mixing nozzles for two-part products, heated melters and hoses for hot melts, and dispensing valves carried by robots.

4

Milestones achieved, from 1894 to 1999.

1

To watch for: announced, not yet confirmed outside.

3

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1894

    A gun for window putty

    Theodore Witte of Chilliwhack, British Columbia, was granted US patent 512,930 on 16 January 1894 for a "Puttying-Tool": a tube that holds putty, a cap with a three-cornered nozzle that shapes the putty to fit "the corners where the glass and sash meet," and a lever with a ratchet and gear that pushes a toothed rod and piston forward a step at a time. Its purpose was to apply putty quickly "without the use of a knife and without the necessity of touching the putty with the fingers."

    patents.google.com

    Credit

    Theodore Witte

  2. 1960s

    Hot glue on packing lines

    Hot melts are adhesives that are solid at room temperature, melted to a liquid to apply, and set as they cool. Eric and Evan Nord founded Nordson in Amherst, Ohio, in 1954 to make airless spray equipment, after buying patents for a "hot airless" way of spraying paint. Nordson says that in the mid-1960s its experience with heated coatings led it to build equipment to apply hot melts "for case sealing, carton manufacturing and product assembly operations."

    nordson.comfundinguniverse.com

    Credit

    Eric and Evan Nord, Nordson Corporation

  3. 1966

    A mixer you throw away

    A two-part adhesive, such as most epoxies, comes as a resin and a hardener that start to cure only when mixed. Constantine D. Armeniades, William C. Johnson and Raphael Thomas of the consultancy Arthur D. Little patented a static mixer, granted on 22 November 1966: a tube holding a row of short twisted elements, alternately left- and right-handed, that cut and fold two flowing liquids together with no moving parts. Their examples dispensed two-part epoxy and polyester resins. The elements could be stamped from one metal strip or made of cheap plastic such as polyethylene, and since resin left inside would set and ruin the tube, the patent notes that after a run "it may be advantageous to discard the device rather than attempting to clean it."

    patents.google.com

    Credit

    Constantine D. Armeniades, William C. Johnson and Raphael Thomas, Arthur D. Little Inc

  4. 1999

    A standard for windshield beads

    A replacement windshield is held in by a bead of adhesive that the installer lays in the frame, often outside a factory. The Auto Glass Safety Council, formed in the late 1990s by glass makers, carmakers, adhesive companies and glass retailers, writes that before it the replacement trade "had no standard for windshield installation" and that this "resulted in lives lost and serious injuries." By 1999 it had developed the Auto Glass Replacement Safety Standard (AGRSS) with the American National Standards Institute, which retailers follow voluntarily. The 2018 edition requires the bead's cross-section to match or exceed the carmaker's, forbids an adhesive that will not reach its "minimum drive-away strength" by the time the car may be driven, requires the owner to be told the minimum drive-away time, the time the adhesive needs to reach that strength, and requires adhesive lot numbers to be traceable to each job; the strength is tied to US federal vehicle safety standards 208 and 212.

    LessonA bead laid by hand in a shop carries the same load as one laid in the factory, so its cure time has to be handed to the driver and its materials traced to the job.

    mgaleg.maryland.gov

    Credit

    Auto Glass Safety Council and its AGRSS Standards Committee

Watch for

  1. Dispensers that fill their gaps

    Coherix says that four 3D laser sensors mounted around a dispensing nozzle give a full 360° view of the bead, that AI-driven control keeps bead volume and placement within limits, and that an "auto-repair" program re-runs the robot's path and dispenses only where it measured too little adhesive, such as a gap left by air in the hose. Its chief executive, Dwight Carlson, writes that bead application is often checked on just one part in every 500, and that an unnamed major carmaker cut unscheduled downtime at one assembly plant by five hours a week with the system. In June 2026 Kawasaki Robotics said it would show the system with Coherix at the Automate trade show in Chicago, applying sealant to a Ford F-150 door skin while measuring and adjusting bead placement up to 400 times a second.

    Achieved whena named manufacturer or an independent party publishes before-and-after counts of bad and repaired beads, scrap and downtime from a production line.

    adhesivesmag.comrobotics247.com

    Claimed by

    Coherix, January 2026; Kawasaki Robotics, June 2026

Open

  1. 01

    A cheap arm lays beads

    A low-cost robot arm, taught from a few dozen human demonstrations, that lays a bead of sealant along curved paths on parts it was not trained on, watches the bead with a camera as it goes, and is scored by measuring every bead it lays.

    The brief
    Credit · unclaimed
  2. 02

    Glue beads with measured sizes

    An open set of images of adhesive and sealant beads in which every bead has been measured, with width and height taken from a laser profile or calipers at marked points, and flaws labeled: gaps, thin spots, excess, bubbles and beads off the path. It would let anyone test whether a camera model can judge a bead's size and not only spot a defect.

    The brief
    Credit · unclaimed
  3. 03

    Tape applied by robot, scored

    A low-cost arm that peels the liner from a strip of double-sided foam tape and presses the strip onto a part at a marked spot, with a public score for placement error, trapped air and how well the strip sticks afterwards.

    The brief
    Credit · unclaimed

Open 01 · the brief

A cheap arm lays beads

Credit · unclaimed

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

What to build

A low-cost robot arm, taught from a few dozen human demonstrations, that lays a bead of sealant along curved paths on parts it was not trained on, watches the bead with a camera as it goes, and is scored by measuring every bead it lays.

Why it is still open

Published work on teaching robots to glue uses costly arms and stops short of measured beads. D. Schäle, M. F. Stoelen and E. Kyrkjebø of Western Norway University of Applied Sciences (2023) compared learning from demonstration with paths planned in CAM software for gluing the groove of an industrial sensor housing on a Franka Emika Panda arm, using a mock-up glue gun; they found the learned approach "strongly dependent on the quality of the demonstrations." Closed-loop bead correction is sold for industrial robots (see Watch for), and searches of the web and GitHub in November 2026 found no public tally of bead quality from a learned policy on a low-cost arm.

mic-journal.no

Built with what exists

The open-source SO-101 arm and Hugging Face's LeRobot software (see the Notes "No download" and "Five demonstrations per task"); an open robot model such as NVIDIA GR00T 2 or Runway's Praxis-1 (see the Note "Open robot model weights"); a battery caulking gun on the wrist, loaded with water-based acrylic caulk, which needs no heat; a USB camera and OpenCV; calipers and a low-cost laser line module to measure bead width and height.

  1. First step

    Mount a caulking gun on an SO-101, record 20 teleoperated demonstrations of a bead along one curved plastic part, and measure the first beads by hand.

  2. Done looks like

    50 beads on 10 part shapes not used in training, each bead's width measured at 20 marked points, published with video, the trained policy and a tally of points within 10% of the target width and of gaps, set beside the same arm replaying a hand-programmed path.

Open 02 · the brief

Glue beads with measured sizes

Credit · unclaimed

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

What to build

An open set of images of adhesive and sealant beads in which every bead has been measured, with width and height taken from a laser profile or calipers at marked points, and flaws labeled: gaps, thin spots, excess, bubbles and beads off the path. It would let anyone test whether a camera model can judge a bead's size and not only spot a defect.

Why it is still open

The public glue-defect sets found on Roboflow Universe in November 2026 are small and unmeasured: one holds 42 photos in a single "defects" class, another 52 photos in a single "Glue" class. A student project at the French engineering school CESI, published on GitHub in 2025, measured glue fillets on phone casings from photos but posts code and sample images, not a measured set. Coherix's chief executive writes that bead application is often checked on only one part in 500.

universe.roboflow.comuniverse.roboflow.comgithub.comadhesivesmag.com

Built with what exists

A phone or USB camera on a fixed stand; a low-cost laser line module, or a structured-light 3D scanner, and calipers; caulks, silicones, hot-melt sticks and two-part epoxies in several colors; CVAT or Label Studio for labels; an open segmentation model fine-tuned with PyTorch.

  1. First step

    Lay 100 beads of two caulks on a printed grid, measure them with calipers, photograph them, and settle the labels.

  2. Done looks like

    At least 3,000 beads of at least 10 materials, each with images, width and height measured at five points and flaw labels, published under an open license, with a baseline model whose width estimates on materials held out of training are scored against the measurements.

Open 03 · the brief

Tape applied by robot, scored

Credit · unclaimed

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

What to build

A low-cost arm that peels the liner from a strip of double-sided foam tape and presses the strip onto a part at a marked spot, with a public score for placement error, trapped air and how well the strip sticks afterwards.

Why it is still open

Open robot data covering tape stops at simple tasks. LeRobot's "aloha_static_tape" dataset holds 50 recorded episodes of the task "Cut a small piece of tape from the tape dispenser then place it on the cardboard box's edge," with no measure of where the tape landed. Xiao Liang, Michael Yip and colleagues (arXiv, October 2025) built a robot that peels up and lays adhesive tape for wound dressing, learning the peeling from teleoperated demonstrations and planning placement for wrinkle-free tape across different body surfaces. No public tally of liner peeling and placement accuracy on parts was found.

huggingface.coarxiv.org

Built with what exists

An SO-101 arm with LeRobot and its ACT policy (Action Chunking with Transformers, a model that learns short runs of motion from demonstrations); a printed fingernail or wedge tool for lifting the liner; strips of double-sided foam tape; printed targets and a camera with OpenCV to measure placement; a hand force gauge for a 90° peel test of the applied strips. The Note "A real production line" reports a robot model's success rate on a car-parts packing line, the kind of trial this would copy at small scale.

  1. First step

    Record 30 teleoperated demonstrations of peeling and placing one 50 mm strip on a flat panel, and train a first ACT policy in LeRobot.

  2. Done looks like

    100 attempts on five part shapes, each counted a success only if the liner is fully removed, the strip lands within 1 mm and 2° of its target and no bubble is wider than 2 mm; a published tally with video; and peel forces of robot-applied strips set beside strips applied by hand.