BRENNEN

Industries / Industrial equipment and services / Industrial services / Commercial diving

Deep diving

Deep diving is work below about 50 m, where divers breathe helium mixed with oxygen and, for long jobs, live for days or weeks at the pressure of the worksite in chambers on deck, riding a diving bell down to work and decompressing only once at the end (saturation diving).

7

Milestones achieved, from 1963 to 2013.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1963

    Living at pressure

    Saturation means the body's tissues have taken up all the gas they will hold at a given pressure, so the time needed to decompress stops growing however long the stay. Following George Bond's Genesis project of animal and chamber tests, the first human saturation dive in helium and oxygen, to 30 m of seawater, was made at the US Navy Experimental Diving Unit in 1963, by the account of Jean-Pierre Imbert and colleagues. Undersea habitats such as Sealab and Conshelf came next, and then chambers on deck, which made it possible to leave a site in bad weather.

    pmc.ncbi.nlm.nih.gov

    Credit

    George F. Bond, US Navy; Albert Behnke is credited with the idea

  2. 1965

    Saturation goes to work

    In 1965 the Undersea Division of Westinghouse Electric, led by Jerry O'Neill and Alan Krasberg, cleared the trash rack (the grille that keeps debris out of the water intakes) of Smith Mountain Dam in Virginia at 61 m, in what Imbert and colleagues call the first commercial saturation project. Its Cachalot system paired a large decompression chamber with a personnel transfer capsule, a small bell that could be locked onto the chamber under pressure.

    pmc.ncbi.nlm.nih.gov

    Credit

    Jerry O'Neill and Alan Krasberg, Westinghouse Electric Corporation

  3. 1981

    Nitrogen calms the tremors

    Below about 160 m, divers breathing helium and oxygen develop the high pressure nervous syndrome (HPNS): tremors, jerking, nausea, sleepiness and changes in brain waves. Peter Bennett's laboratory at Duke University found from 1973 that adding a little nitrogen, which is itself mildly narcotic, eased these signs. In its 1981 Atlantis III chamber dive, three divers were compressed over 6 days 8 hours to 650 m and then to 686 m (2,250 ft) for 24 hours, where HPNS tremors were controlled and they completed all their tasks, with a 20% to 30% loss of concentration.

    LessonHow fast divers are compressed matters as much as the gas: faster compressions in earlier dives cost 40% to 50% of performance on the first day.

    ebi.ac.ukebi.ac.uk

    Credit

    Peter B. Bennett and colleagues, F. G. Hall Laboratory, Duke University Medical Center

  4. 1983

    A clamp opened under pressure

    On 5 November 1983, on the drilling rig Byford Dolphin at the Northeast Frigg field in the Norwegian North Sea, the clamp that joined a diving bell to the chamber system was opened by mistake while the inner door between the connecting trunk and the chambers stood open. The chambers, at about 9 atmospheres (the pressure at 90 m), lost their pressure at once: four divers inside and a tender outside were killed, and another tender was badly hurt. The system had no device to stop the clamp opening while there was pressure behind it; Det Norske Veritas (DNV), which certified it, had asked for one, but it had not been fitted. A month later DNV and the Norwegian Petroleum Directorate made such interlocks an absolute requirement on all chambers, and shift plans for divers, stricter limits on time in the bell and limits on overtime for surface crews followed.

    LessonA procedure that one wrong move can defeat needs a mechanical lock behind it.

    snl.noebi.ac.uk

    Credit

    Norway's expert commission (NOU 1984:11); autopsy study by J. C. Giertsen and colleagues; account by Store norske leksikon

  5. 1988

    Hydrogen at 534 m

    From 22 February to 22 March 1988, off Marseille, six divers, four from Comex and two from the French Navy, worked from the diving support vessel Orelia at 520 m and 534 m breathing hydreliox, a mix of hydrogen, helium and oxygen. In 27 hours underwater they joined two 12-inch pipe sections with a flanged spool and connected a flexible pipe. Comex's research since 1983 had found that divers breathing hydrogen mixes showed no clinical signs of HPNS and breathed far more easily. Its later Hydra 10 chamber dive, to a simulated 701 m, needed 15 days of compression and 23 of decompression.

    LessonA record depth did not become a working depth: Comex dropped hydrogen after Hydra 12 in 1996, as companies turned to crewed underwater vehicles and helium saturation, by MIT Technology Review's account.

    cordis.europa.eutechnologyreview.com

    Credit

    Comex (Compagnie Maritime d'Expertises), with European Community funding

  6. 1990

    Norway sets one frame

    In the 1970s companies stamped their saturation manuals "Secret", because faster decompression won contracts, and in the 1980s Norwegian divers' unions warned that competition could push companies to shorten decompression. In 1984 the Norwegian Petroleum Directorate hired Val Hempleman of the Royal Navy to review the procedures in use; regulators, companies and divers' unions then agreed common frame conditions for saturation to 180 m, built from the most conservative features of existing tables, and these went unchanged into the first NORSOK U-100 standard in 1999. Havtil, Norway's offshore safety regulator, reports that saturation diving on its shelf has followed standardised frames since 1990, that the last death in saturation diving there was in 1987, and that no case of decompression sickness in saturation was reported in the 20 years to 2023.

    LessonNorway's record improved after table speed stopped being a way to win work, though diving activity also fell in the mid-1990s and the figures do not separate the causes.

    pmc.ncbi.nlm.nih.govhavtil.no

    Credit

    Norwegian Petroleum Directorate; Val Hempleman

  7. 2013

    Divers win on information

    About 350 to 400 Norwegian divers worked in the North Sea's "pioneer period" of 1965 to 1990, and many later had health problems. After a commission led by Judge P. A. Lossius reported in 2002 (NOU 2003:5), Norway set up a state compensation board by royal decree on 4 June 2004. On 5 December 2013 the European Court of Human Rights ruled, by five votes to two, that Norway had violated the divers' right to private life (Article 8 of the Convention) by failing to ensure that they received essential information about the risks of the decompression tables in use, which companies had kept confidential to stay competitive. It awarded each applicant €8,000.

    LessonDecompression tables are health information, and keeping them from the people who dive them breached their rights.

    globalhealthrights.org

    Credit

    Dag Vilnes and six other former divers; European Court of Human Rights

Watch for

  1. Speech recognition for heliox

    Helium makes a diver's voice high and hard to understand over the intercom, and engineers have built electronic "unscramblers" for it since at least 1966. A team at Donghua University in Shanghai adapted OpenAI's Whisper-large-v3 speech model with LoRA, a cheap way to retrain a small part of a model, and reports a character error rate of 4.7% at 12 m and 7.2% at 25 m on recordings made in a ship-mounted saturation system. The depths are far shallower than most saturation work, and the recordings, held by the PLA Naval Medical Center, are not public. The site's Note "Fewer word errors" covers ElevenLabs' Scribe v3, a general speech-to-text model not tested on heliox speech.

    Achieved whenan outside group reports the error rate of this or a similar model on heliox speech recorded below 100 m.

    pmc.ncbi.nlm.nih.gov

    Claimed by

    Mao and colleagues, Scientific Reports, 2026

  2. Learned control for chambers

    A decontamination decompression chamber has several compartments whose pressures must stay within set differences of each other while divers wash and decompress. Zhang, Lin and Jiang propose a controller that plans with a learned policy and adapts to faults such as an offset valve, a failing actuator or a leak, and report higher rewards, faster convergence and better robustness than conventional methods; the results come from their own experiments, not from a working chamber with people inside.

    Achieved whenthe controller runs a real, unoccupied multi-compartment chamber with faults injected on purpose, and a published test shows the pressure differences staying within limits.

    ebi.ac.uk

    Claimed by

    Zhang, Lin and Jiang, Sensors, 2025

Open

  1. 01

    Saturation tables as open code

    A free program that computes the final decompression from saturation under the published US Navy (2016) and NORSOK rules, step by step, so that anyone can compare procedures. It is marked as not for planning real dives.

    The brief
    Credit · unclaimed
  2. 02

    Norway's dive record as numbers

    A public table of Havtil's diving statistics for every year from 1985: man-hours in saturation, diving support vessel days, incidents by type, cases of decompression sickness and deaths.

    The brief
    Credit · unclaimed

Open 01 · the brief

Saturation tables as open code

Credit · unclaimed

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

What to build

A free program that computes the final decompression from saturation under the published US Navy (2016) and NORSOK rules, step by step, so that anyone can compare procedures. It is marked as not for planning real dives.

Why it is still open

Imbert and colleagues compared the saturation procedures of 11 companies in 2024 with their own program, anonymised them, and found a 25.7 h spread between the slowest and fastest decompression from 150 m; no code was released. Their 2026 review notes that the US Navy never published how its saturation decompressions were computed. GitHub searches in November 2026 for saturation, heliox and excursion decompression found only two 2020 stub repositories and a game; open dive planners cover scuba (Bühlmann ZH-L16 libraries and a transcription of the 1973 DCIEM programs), not saturation.

pmc.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.gov

Built with what exists

Python; the saturation chapter of the US Navy Diving Manual, revision 7 (2016); NORSOK U-100; the decompression rates, oxygen levels and rest stops tabled by Imbert and colleagues; an AI coding agent to turn each published table value into a unit test.

  1. First step

    Enter the NORSOK decompression rates and rest stops into a spreadsheet and reproduce the 156 h figure for 150 m.

  2. Done looks like

    The program reproduces the NORSOK durations in Imbert's Table 5 (78, 104, 130 and 156 h from storage depths of 60, 90, 120 and 150 m) within 0.5 h, with every rule linked to the page it came from and the code and tests published.

Open 02 · the brief

Norway's dive record as numbers

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 Havtil's diving statistics for every year from 1985: man-hours in saturation, diving support vessel days, incidents by type, cases of decompression sickness and deaths.

Why it is still open

Havtil's DSYS database is the only one covering offshore commercial diving, and companies do not share their decompression sickness figures, Imbert and colleagues wrote in 2026; in 2024 they noted that a rate per dive cannot be worked out because exposure is given in man-hours. The DSYS reports, in Norwegian, show the 1985–2025 series only as chart images, with a few yearly figures in the text, such as 111,388 man-hours in saturation in 2023 and 53,396 in 2025.

pmc.ncbi.nlm.nih.govhavtil.nohavtil.no

Built with what exists

PyMuPDF to pull the chart images from the DSYS PDFs; WebPlotDigitizer; a vision-language model to read bar heights, checked by hand; Python with pandas; a written request to Havtil for the yearly number of saturations and bell runs.

  1. First step

    Download the DSYS reports for 2018, 2023 and 2025, digitise the saturation chart in each, and check that the years they share agree.

  2. Done looks like

    A CSV of each series for all 41 years from 1985 to 2025, every value tied to a report and figure, agreeing within 2% with every yearly figure stated in the reports' text, with 10% of points read independently by two people, and Havtil's answer to the request published.