BRENNEN

Industries / Construction and property / Heavy civil / Marine construction

Cofferdams and caissons

This work builds foundations below water level by keeping the water out, either with a temporary wall around the site or by sinking a box and working inside it.

6

Milestones achieved, from 1739 to 2008.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1739

    Piers built in floating boxes

    For Westminster Bridge in London, the Swiss-born engineer Charles Labelye built each stone pier inside a caisson, a timber case floated to its spot and sunk into a bed dug in the river; the Survey of London calls it the first time the method was used on a large scale. The piers rested on the soil rather than on piles, and in 1747, with the bridge almost finished, one pier sank 16 inches (about 40 cm) and the arches beside it had to be rebuilt.

    LessonA caisson makes underwater masonry possible but does not make soft ground firm; it still needs piles or a hard layer beneath.

    british-history.ac.uk

    Credit

    Charles Labelye

  2. 1841

    Compressed air holds back water

    On a sandbank of the Loire in France, an engineer named Triger covered an iron tube about 1 m across with an airlock he had devised and pumped in compressed air to push the water out, so workers could dig at the bottom; the tube went down about 20 m. Pneumatic caissons, large boxes worked the same way, later carried the Brooklyn Bridge, the Eads Bridge at St. Louis and two of the Eiffel Tower's legs.

    daiho.co.jp

    Credit

    Triger, by the account of the Japanese caisson contractor Daiho Corporation

  3. 1871

    Caisson workers fall sick

    The piers of the Eads Bridge at St. Louis and the Brooklyn Bridge in New York were sunk in pneumatic caissons, and workers who spent shifts at pressures up to 55 psi (about 3.8 bar) above the atmosphere came out with caisson disease, now called decompression sickness. The Eads Bridge physician, Alphonse Jaminet, published his observations in 1871, and the Brooklyn Bridge physician, Andrew Smith, in 1873.

    LessonRaising air pressure to keep water out moves the danger into the workers' bodies; it has to be paid for with slow decompression or by taking people out of the chamber.

    ebi.ac.uk

    Credit

    Alphonse Jaminet and Andrew Smith, as reviewed by W. P. Butler in 2004

  4. 1990

    Caissons dug from the surface

    Kajima Corporation and Shiraishi Corporation in Japan built a system that runs the excavator inside a pneumatic caisson's pressurized working chamber by remote control from a room on the ground, with instruments that track the caisson's sinking. Tokyo Electric Power used it on a site, the partners reported it in 1990, and by 2000 Shiraishi reported more than 50 caissons built with remote-controlled digging.

    iaarc.orgiaarc.org

    Credit

    Kajima Corporation and Shiraishi Corporation

  5. c. 1990

    Concrete poured 60 m down

    The two tower foundations of the Akashi Kaikyo Bridge in Japan were built by sinking steel caissons 80 m across into dredged pits in the seabed and filling them with anti-washout underwater concrete, a mix with additives that keep it from washing apart as it falls through water. About 500,000 m³ went in within a year, in pours of about 10,000 m³ placed nonstop over three days, in tidal currents up to 8 knots and at depths down to 60 m, with robots cleaning the bedrock and the joints between pours.

    iaarc.orgice.org.uk

    Credit

    Honshu-Shikoku Bridge Authority

  6. 2008

    A cofferdam floor heaves

    A cofferdam is a temporary wall, often of interlocking steel sheet piles, that keeps water out while a foundation is built. In November 2008, at the Kabinakagami River Bridge on Highway 11 in Ontario, Canada, the sheet piles of the west abutment cofferdam moved inward near the bottom, cracks and sinkholes opened around it, and the province's Ministry of Labour stopped work. A third-party review found that the floor heaved and the wall toes then kicked out: the excavation appears to have gone to full depth before the bracing was in, no seepage control was used, and driving piles inside raised the water pressure in the soil.

    LessonWater pressure and soft layers below the floor decide whether a cofferdam stands as much as the walls do, and the order of digging and bracing is part of the design.

    members.cgs.ca

    Credit

    Exp Services Inc., third-party reviewer for Ontario's Ministry of Transportation

Watch for

  1. Caisson shovels drive themselves

    Oriental Shiraishi and the AI firm DeepX say three caisson shovels, the excavators that run on rails under the ceiling of a pneumatic caisson's working chamber, moved soil automatically for just over six hours, a full working day, on a real construction site, avoiding collisions and obstacles while a person monitored them.

    Achieved whenthe system digs whole caissons on several sites and an owner or outside party publishes output and tilt records alongside those of remote-operated digging.

    prtimes.jp

    Claimed by

    Oriental Shiraishi and DeepX, June 2026

Open

  1. 01

    A register of cofferdam failures

    A public table of every cofferdam failure with a published account, giving the date, place, type of cofferdam (sheet pile, cellular, earth), water depth, how it failed (floor heave, piping, overtopping, bracing failure), injuries and deaths, and a link to the account.

    The brief
    Credit · unclaimed
  2. 02

    A model caisson, openly logged

    A tabletop open caisson (a bottomless box that sinks as soil is dug from inside it) in a tank of wet sand, with load cells under its cutting edge, a tilt sensor and a camera, and a public log of every digging step and the tilt and sinking that followed, so automatic digging plans can be compared on the same data.

    The brief
    Credit · unclaimed

Open 01 · the brief

A register of cofferdam failures

Credit · unclaimed

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

What to build

A public table of every cofferdam failure with a published account, giving the date, place, type of cofferdam (sheet pile, cellular, earth), water depth, how it failed (floor heave, piping, overtopping, bracing failure), injuries and deaths, and a link to the account.

Why it is still open

Failures are written up one case at a time. The 2019 review of the Kabinakagami failure leans on single-case papers from 1953, 1970 and 1980, and searches for a compiled database or statistical review of cofferdam failures found none.

members.cgs.ca

Built with what exists

The free Crossref and OpenAlex search APIs; OSHA's published fatality investigation summaries and comparable national safety records; a vision-language AI model to pull the fields from scanned papers, quoting the page for each value; a shared spreadsheet.

  1. First step

    Pull every paper with "cofferdam" and "failure" or "collapse" in its title from the Crossref API and list them by year and country.

  2. Done looks like

    50 failures, each with a source link and a failure mode checked by a second reader, with the rate at which the two readers agreed published.

Open 02 · the brief

A model caisson, openly logged

Credit · unclaimed

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

What to build

A tabletop open caisson (a bottomless box that sinks as soil is dug from inside it) in a tank of wet sand, with load cells under its cutting edge, a tilt sensor and a camera, and a public log of every digging step and the tilt and sinking that followed, so automatic digging plans can be compared on the same data.

Why it is still open

Caisson sinking records stay with their owners. A 2025 Scientific Reports study of a large caisson at the Zhuchong pumping station in China says its data are available from the authors "on reasonable request", and the Oriental Shiraishi and DeepX trial ran on the contractor's own site data.

pmc.ncbi.nlm.nih.govprtimes.jp

Built with what exists

An acrylic tank and sand; a 3D-printed caisson model; load cells with an HX711 amplifier; a BNO085 orientation sensor; a Raspberry Pi; a small gantry, such as a converted 3D printer, to dig; Python.

  1. First step

    Build a 300 mm caisson model and log five sinkings dug by hand.

  2. Done looks like

    30 logged sinkings published with a baseline digging plan that keeps tilt under 1 in 100, and one run repeated by another lab within a stated error.