BRENNEN

Industries / Energy, water and waste / Oil and gas extraction / Offshore drilling

Well control

Well control is keeping oil, gas and water in the rock from flowing up the well, with heavy drilling fluid and cement as the first barriers and the blowout preventer, a stack of valves and rams on top of the well, as the last.

5

Milestones achieved, from 1969 to 2025.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1969

    Short casing, cracked seafloor

    On 28 January 1969 Union Oil's Platform A, about five miles off Summerland, California, was pulling pipe from a 3,500 ft well when gas blew out the drilling fluid, split the casing and cracked the seafloor around the well. The US Geological Survey had let Union Oil set shorter casing than federal standards required. Oil flowed for eleven days until mud sealed the cracks, by which time about three million gallons had escaped, according to UC Santa Barbara geographers. Federal and state drilling rules were strengthened.

    Lessononce flow escapes outside the casing, closing valves on the well cannot stop it, so casing depth is a safety decision.

    people.geog.ucsb.edu

    Credit

    Union Oil (operator); Keith Clarke and Jeffrey Hemphill (UC Santa Barbara, record)

  2. 1977

    Bravo blows out

    At 22:00 on 22 April 1977 well B-14 on Phillips Petroleum's Ekofisk Bravo platform in the Norwegian North Sea blew out during a workover, a job to pull and replace the production tubing. A downhole safety valve had not been properly attached, and the lowest part of the blowout preventer had been fitted upside down. NOAA puts the release at about 202,380 barrels and calls it the first major oil release in the North Sea; the well was capped on 30 April. The government commission found human error to be the main cause and called for better work routines, and the regulator then required the preventer to be fully assembled before workovers.

    Lessonwell control equipment that is installed wrong, or not yet installed, is not a barrier.

    ekofisk.industriminne.noincidentnews.noaa.gov

    Credit

    Phillips Petroleum (operator); the Bravo commission of inquiry

  3. 2009

    One untested barrier

    On 21 August 2009 the Montara H1 well, drilled from the West Atlas jack-up rig in about 77 m of water off northwest Australia, blew out; oil and gas flowed for more than ten weeks until a relief well brought it under control on 3 November. PTTEP Australasia's own standards called for two proven barriers, but the only one in place was a cemented casing shoe whose integrity had never been proven. Of two planned pressure-containing caps, one was never fitted and the other had been removed hours before. It was the first such blowout in Australian waters for 25 years.

    Lessona barrier counts only if it has been tested.

    openresearch-repository.anu.edu.au

    Credit

    Montara Commission of Inquiry; Jan Hayes (Australian National University, analysis)

  4. 2023

    Preventer failures counted, then falling

    Since the 2016 US rule, every failure of well control equipment in US waters must be reported to SafeOCS, a confidential system run by the Bureau of Transportation Statistics. Its 2023 report, covering 2017 to 2023, counted 337 events in 2023, the fewest since reporting began, or 20.6 per 1,000 days of blowout-preventer use. In subsea systems, wear and tear was the leading root cause since 2017, at 46.4%, and nearly all external leaks were of hydraulic control fluid.

    safeocs.govsafeocs.gov

    Credit

    Bureau of Transportation Statistics; BSEE

  5. 2025

    Gas trapped behind casing

    On 23 September 2025 crews permanently plugging a Troll field well from Odfjell Drilling's semisubmersible Deepsea Bollsta, with 114 people aboard, cut a 13⅜-inch casing 160 m below the seabed. A log had been read to show sound cement and no gas behind the casing, so the cut was made without closing the blowout preventer. Water and gas gushed onto the drill floor and into the shaker room until the crew closed the preventer; one person was slightly hurt. Norway's Ocean Industry Authority (Havtil) found a failed outer barrier, logging technology not qualified for checking barriers and no barriers in place during the cut, and said the gas could have ignited. Equinor, the field operator, put the release at about 930 kg of gas and now requires the preventer to be closed during shallow cuts.

    Lessonwhen a barrier cannot be proven, take the cautious choice and close the well before cutting into it.

    havtil.nooedigital.com

    Credit

    Havtil (investigation); Equinor (operator); Odfjell Drilling (rig)

Watch for

  1. All-electric blowout preventer

    COSL Drilling Europe says it is working with NOV, HMH and SLB on a blowout preventer driven entirely by electricity instead of hydraulic fluid, which it says could cut total system weight by about 50% and be in use "within a few years." HMH separately markets an all-electric preventer it says shears pipe about four times faster than industry requirements. Neither gives outside test results.

    Achieved whenan all-electric subsea preventer is qualified by a classification society and used on a well, with closing and shearing test results published.

    oedigital.comhmhw.com

    Claimed by

    COSL Drilling Europe, November 2025

  2. Preventers watched from shore

    DNV, a classification society, has granted Technology Qualification to the real-time monitoring of blowout preventers in ModuSpec's Argus platform, which gathers rig data, flags problems and logs what was done about them. ModuSpec says it was piloted with operators on the Norwegian shelf but gives no rig counts or results.

    Achieved whenoperators or a regulator publish what the monitoring caught across a fleet, such as failures found before a well needed the preventer.

    oedigital.com

    Claimed by

    ModuSpec (MR Group), January 2026

Open

  1. 01

    Answer key for negative tests

    A public set of simulated negative pressure tests with known answers. In a negative pressure test, the crew lowers the pressure inside a newly cemented well to check that the cement and seals hold before the heavy fluid that keeps the well shut is removed. Each simulated test gives drill pipe and kill line pressures over time for a sound well, a leaking cement job, a leaking preventer seal and heavy fluid left below the preventer, with a scorer that tests rule-based checkers and AI models on them.

    The brief
    Credit · unclaimed
  2. 02

    Re-test shear formulas openly

    A machine-readable table of published tests in which a blowout preventer's shear rams cut drill pipe, starting with the manufacturer data printed in a 2004 study for the US regulator, and an open computer model of the cut that is scored against those tests beside the standard formula.

    The brief
    Credit · unclaimed

Open 01 · the brief

Answer key for negative tests

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 simulated negative pressure tests with known answers. In a negative pressure test, the crew lowers the pressure inside a newly cemented well to check that the cement and seals hold before the heavy fluid that keeps the well shut is removed. Each simulated test gives drill pipe and kill line pressures over time for a sound well, a leaking cement job, a leaking preventer seal and heavy fluid left below the preventer, with a scorer that tests rule-based checkers and AI models on them.

Why it is still open

A 2018 model for reading these tests, written after a US court found that "the misinterpretation of the negative pressure test was a substantial cause" of the Deepwater Horizon blowout, was quantified with expert judgment, not with recorded tests. Searches for this file found patented methods for automated checks but no public set of tests with known answers, and no GitHub project that simulates these tests.

iapsam.org

Built with what exists

Python; a published model of fluid compression and flow in a well; the National Academies' Macondo appendix for the pressure difference at the start of the test (reservoir pressure 11,892 psi); time-series models and vision-language models to be tested.

nationalacademies.org

  1. First step

    Rebuild the start-of-test pressure difference for Macondo from the National Academies appendix in a short Python script, then add a leaking-cement case.

  2. Done looks like

    At least 2,000 labeled tests across at least 10 well designs, with the Macondo test rebuilt as one case; at least three checkers scored on missed leaks and false alarms; data and scorer published.

Open 02 · the brief

Re-test shear formulas openly

Credit · unclaimed

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

What to build

A machine-readable table of published tests in which a blowout preventer's shear rams cut drill pipe, starting with the manufacturer data printed in a 2004 study for the US regulator, and an open computer model of the cut that is scored against those tests beside the standard formula.

Why it is still open

The 2004 study by WEST Engineering Services for the Minerals Management Service evaluated 214 drill pipe shear data points, compared them with the formula the industry uses (the Distortion Energy Theory) and fitted its own equations, but printed its data only as tables in a PDF report. In 2013 Lloyd's Register began finite element modeling to predict shear forces more accurately, noting that newer high-strength pipe grades had raised doubts about the formulas and that a shear test usually cost $200,000 to $400,000; no public model from that work was found. Since 2016, US rules have sent new shear test results to a BSEE-approved verification organization, not to the public.

bsee.govdrillingcontractor.orgfederalregister.gov

Built with what exists

The 2004 report's appendix tables; a vision-language model to transcribe them, checked by two people; CalculiX, an open finite element solver; published drill pipe material properties; Python.

  1. First step

    Transcribe the first 20 rows of the 2004 appendix table by hand and compute the standard formula's error on them.

  2. Done looks like

    The 196 data points with full material data from the 2004 appendix in an open table checked by double entry; the standard formula and the open model scored on the same points, with the model within 10% of the measured force on at least 20 points held out from its tuning; code published.