The Kielland Rig Was Already Cracked When It Left the Yard

Chapters

1. 0:00 Ekofisk, 18:30
2. 1:01 A bad weld, and a second hole
3. 1:41 A drilling rig that never drilled
4. 2:32 Brace D-6, and two holes in it
5. 3:17 One was structure. One was outfit.
6. 4:02 Six millimetres against ten to eighteen
7. 4:43 One to seven percent
8. 5:18 Paint inside the crack
9. 6:16 From a bracket weld to a broken brace
10. 7:13 The calculation that was never made
11. 8:04 Nobody had looked for them
12. 9:05 A hundredth of a second
13. 10:02 A column was not on the list
14. 10:53 Twenty minutes
15. 11:50 What changed
16. 12:46 The label travelled

Transcript

Ekofisk, 18:30 0:00

The North Sea, Ekofisk field, March 27th, 1980. Half past six in the evening. Wind at sixteen to twenty metres a second, waves of six to eight metres. Ordinary weather for the place. The Alexander L. Kielland is a floating hotel for oil workers, anchored beside a production platform. Two hundred and twelve people are aboard. There is a bang, then a second one, and a shaking. Some hear a sound like metal tearing. One of the five legs is gone. The platform heels to thirty-five degrees almost at once, and turns over in about twenty minutes. One hundred and twenty-three people die. The crack that did this started in a weld six millimetres wide, holding a bracket for a sonar. The sonar carried no load at all. And the crack was there, painted over, before the platform ever left the yard.

A bad weld, and a second hole 1:01

The story everyone tells about the Kielland is a bad weld. A fatigue crack grew from it, a brace broke, a leg fell off. Every word of that is in the report. It is also where the report begins, not where it ends. Because the weather that evening was nothing. The stress in that brace was a fifth of what the steel could take. And on the same brace, in the same plate, less than a foot from the sonar hole, there was a second hole. That one was welded properly. The question this episode is about is why one plate carried two holes, welded two different ways.

A drilling rig that never drilled 1:41

The Kielland was a Pentagone rig, a French design. Five pontoons, five columns eight and a half metres across, a hundred and one metres from end to end. It was built at Dunkirk and delivered in July 1976 as a drilling rig. It never drilled a well. It was used as accommodation instead. It came with eighty beds. In four rounds of adding living modules on deck, that became three hundred and forty-eight. The design goal, in the words of the report, was minimum weight. As few braces as the design could take. High strength steel, worked hard, especially in the braces. The braces are what hold the five columns together. Under the waterline, they are the structure.

Brace D-6, and two holes in it 2:32

Brace D six was a horizontal tube, two point six metres across, with a wall twenty-six millimetres thick. Two holes had been cut into that wall. The first was a lightening hole, an oval opening, eight hundred and seventy by three hundred and seventy millimetres. The second, two hundred and seventy millimetres away, was round, three hundred and twenty-five millimetres across. Into it went a short steel tube with a flange, the holder for a hydrophone. A hydrophone is a listening device. On a drilling rig it tells the crew where the rig is over the well. This rig never drilled. The instrument in that holder had nothing to do all day.

One was structure. One was outfit. 3:17

Here is the first thing the common version leaves out. The lightening hole was treated the way ship practice treats a hole in a loaded plate. Its edge was reinforced with a ring, and the ring was joined with a K weld, fused through the full thickness. The hydrophone holder, in the same plate, was fixed with fillet welds. Six millimetres. The report says why. At the design office, the hydrophone arrangement was handled by the people who dealt with equipment. The commission calls it surprising that two holes placed so close together were never looked at as one problem. They were not one problem. One was structure. One was outfit.

Six millimetres against ten to eighteen 4:02

What six millimetres means is a number. For a fillet weld to be as strong as the plate it joins, its throat has to be forty to seventy percent of the plate thickness. On a twenty-six millimetre plate, that is ten to eighteen millimetres. The drawing said six. The weld as made was worse than the drawing. Where the metal failed to fuse, the effective throat was about five. The weld had barely bitten into the holder at all. Penetration into that plate was under two millimetres. None of this was hidden. It was what a weld looks like when nobody has asked what it carries.

One to seven percent 4:43

The steel of the holder was worse than the weld. Loaded through its thickness, the way this weld loaded it, a good plate contracts twenty percent before it tears. That was the figure in the Veritas guidance from 1973. The plate in the hydrophone holder gave between one and seven. Its impact toughness was less than half the minimum in the building specification. And the specification had no requirement for through thickness properties on that plate. Nobody had classified it as a plate that would be loaded that way.

Paint inside the crack 5:18

Now the paint. The brace was fabricated at a works in the south of France, then shipped to the yard at Dunkirk to be joined to the rig, and painted. After the accident, the fracture surfaces were examined in the laboratory. Along seventy millimetres of the inner fillet weld, on the broken faces, there was paint. Two layers of it. Chocolate brown, then black. The yard's paint. You cannot paint the inside of a crack that does not exist yet. So a crack at least seventy millimetres long was open in that weld before the platform was painted, before it floated, before anyone lived on it. The Veritas surveyor said every weld on the holder had been checked visually and with dye penetrant. There is no record of an ultrasonic test. The crack may have opened after the inspection, on the truck to Dunkirk. Either way, it went to sea.

From a bracket weld to a broken brace 6:16

The crack did not sink the platform by itself. It changed the geometry. Early in the rig's life, the fillet welds cracked around three quarters of the holder. The holder stopped carrying anything. So the hole in the brace was now just a hole. And the stress at the edge of a hole in a loaded plate is about three times the stress in the plate around it. At two points on the rim, opposite each other, the load rose by a third to three quarters. From those two points, two fatigue cracks set off around the circumference of the brace. For the first two or three hundred millimetres on each side they grew the way fatigue cracks grow, ring by ring, with every wave. Then faster, in jumps. By the evening of March 27th they covered two thirds of the circumference. The rest broke in a hundredth of a second.

The calculation that was never made 7:13

Was any of this calculated? The documentation submitted for approval contained no fatigue calculation. None. Veritas classed the rig and did not ask for one, although its own rules named fatigue as a design criterion. The designers state that stress calculations at the important joints were considered. They were not carried out. The commission is direct about what such a calculation would have shown. A relatively short fatigue life, even with the holder intact and the weld perfect. In a hundred year storm, this brace ran at half to seventy percent of the yield stress of its steel. On the evening it broke, a fifth. The strength of the platform as a whole had been documented. The strength of its details had not been examined at all.

Nobody had looked for them 8:04

Then there is what was inspected in service. Veritas surveyed the Kielland once a year, three times in its life. Each survey was visual, by spot check, of the upper outside parts of the platform, mostly the deck joints. It took about one day. The braces under the water were not inspected at all. The four year survey, the one where cracks get found, was still to come. And no inspection handbook existed for Pentagone rigs. At the last annual survey, in the autumn of 1979, the cracks were probably still small. Perhaps not yet through the wall. After the accident, other Pentagone rigs in the North Sea were examined properly. Cracks were found at hydrophone holders and lightening holes on many of them. The report says they were probably there at earlier surveys too, and were not found, first and foremost because nobody had looked for them.

A hundredth of a second 9:05

When D six let go, the whole leg went with it. Here is why. The final break took between one and ten thousandths of a second. The load it had carried was thrown onto the five other braces holding column D, and thrown so fast that the shock multiplied it, by a factor of one and a half to two and a half. Without D six, the report says, there was hardly strength in the structure to carry the platform's own loads in still water. The five other braces failed one after another. The column tore off and floated away on its own. A few more braces would have been enough to survive the loss of one. The designers left them out, for operations and for weight. And no rule made them. The rules required a rig to stay afloat with two tanks flooded. They said nothing about strength after a member had failed.

A column was not on the list 10:02

The same shape shows up in the stability approval. The calculations that were approved assumed the deck would never go under water. That assumption came from the list of damage cases the rules asked about, and the loss of a column was not on the list. So watertight doors were only required at the top of the columns. The deck had ordinary doors, and ventilators. With column D gone, the platform lay at thirty to thirty-five degrees, and the sea came in through the lift shafts of two other columns and through the deck. Flooding through the drains alone would have taken more than an hour. It turned over in twenty minutes. An event outside the categories was not analysed, so nothing was designed for it. Exactly the logic of the hydrophone holder, one level up.

Twenty minutes 10:53

Two hundred and twelve people had twenty minutes. The main generator tripped almost at once and the lights went out. The alarm sounded for a minute and a half, then stopped. There were seven covered lifeboats, fifty seats each. Their release hooks could not be let go while the boat hung under load, and in that sea every boat was under load. Three boats were smashed against the platform. One got away with twenty-six people, after a man reached the aft hook by hand, through a hole where the wheelhouse had been crushed. Of everyone who ended up in the water and did not reach a boat or a raft, fourteen were picked up alive. The standby vessel arrived after nearly an hour. It rescued no one. Eighty-nine people survived. Of the two hundred and twelve on board, seventy-six had ever attended a safety course.

What changed 11:50

The commission reported in April 1981. Norway had already, that autumn, required a survival suit for every person offshore. The rules for floating rigs changed to require that a unit stay afloat with a column gone, which is done by making part of the deck a buoyant box. The report asked for design against damage, not only against load. For fatigue calculation of details, including fillet welds. For inspection aimed at the critical areas, and a thorough survey of every new rig after its first winter. The wreck was righted in September 1983 and sunk in a fjord. One more line from the report. On a later Pentagone rig, the owner's office asked for the hydrophone holder to be welded with full penetration. For some reason, the report says, this was not followed up during construction.

The label travelled 12:46

Last episode, a crack in a hangar had no category to be reported under, so it was never reported. This episode, a bracket had no category to be analysed under, so it was never analysed. And the label travelled. Not structure meant no fatigue calculation. No calculation meant no weld procedure to approve. No procedure meant no production test, no ultrasonic check, no line on the surveyor's list. Nobody decided that weld did not matter. They decided the hydrophone was equipment, and the weld went wherever the hydrophone went. Two years later, in the North Atlantic, another floating rig goes down in a storm, with everyone aboard. February 15th, 1982. The Ocean Ranger, off Newfoundland. Eighty-four people, no survivors. A window breaks in the ballast control room. Seawater reaches the control panel, and the valves that trim the rig begin opening on their own. The crew fights the panel for hours. The failure is not the storm.

Description and sources

On March 27th, 1980, in ordinary North Sea weather, the accommodation platform Alexander L. Kielland lost one of its five columns at the Ekofisk field and turned over in about twenty minutes. Of the 212 people aboard, 123 died.

Brace D-6 was a tube 2.6 metres across with a 26 mm wall. Two holes had been cut into it: an oval lightening hole, treated as structure with a reinforced edge and a full-penetration K weld, and 270 mm away a round hole for a hydrophone. At the design office the hydrophone was handled by the equipment side. Its holder got 6 mm fillet welds where a load-bearing fillet on that plate would need 10 to 18; as made, the effective throat was about 5 mm. The holder's steel contracted 1 to 7 percent through its thickness where the Veritas guidance asked for 20 — and the specification required nothing, because nobody had classified that plate as loaded.

Along 70 mm of the inner weld the fracture faces carried two layers of the Dunkirk yard's paint: the crack was open before the platform was painted. No fatigue calculation of any detail was submitted for approval. Three annual surveys were visual spot checks of the upper deck, about a day each; the braces under water were never inspected. After the accident, cracks were found at hydrophone holders on many other Pentagone rigs — not found earlier, the report says, because nobody had looked.

When D-6 broke, the five other braces on column D failed one after another; a few more braces would have carried the loss, and no rule required them. The stability approval assumed the deck would never go under water: losing a column was not among the damage cases the rules asked about.

PRINT-READY, FROM THIS CHANNEL

The Failure Atlas, Vol. 01 — Tacoma Narrows · Citicorp Center · Millennium Bridge · Apollo 13 · the 2003 blackout · Hyatt Regency

https://therepository.gumroad.com/l/failure-atlas

PRIMARY SOURCES

• NOU 1981:11, «Alexander L. Kielland»-ulykken — report of the Norwegian commission of inquiry (Næsheim, Moan, Bekkvik, Kloster, Øveraas), Universitetsforlaget, 1981; Norwegian with an English summary (pp. 207–216). Read page by page from the scanned original in the National Archives: https://media.digitalarkivet.no/view/105511/1
• Norwegian Petroleum Museum, "What caused the Alexander L. Kielland accident?" — for the survival-suit rule, the buoyancy rule and the fate of the wreck: https://ekofisk.industriminne.no/en/what-caused-the-alexander-l-kielland-accident/

ARCHIVE PHOTOGRAPHS — the real structure, with title, author, date, archive and licence. CC BY items are used under that licence, cropped and reframed for the screen.

• Alexander L Kielland and Edda 2-7C — Norsk Oljemuseum — 1980 | Wikimedia Commons | CC BY 3.0
• Alexander L Kielland Column D drifting in the water — Norsk Oljemuseum — 1980 | Wikimedia Commons | CC BY 3.0
• Column D of Alexander L Kielland details — Norsk Oljemuseum — 1980 | Wikimedia Commons | CC BY 3.0
• Alexander L Kielland Column D with fractures — Norsk Oljemuseum — 1980 | Wikimedia Commons | CC BY 3.0
• Alexander L Kielland Head of Column D after accident — Norsk Oljemuseum — 1980 | Wikimedia Commons | CC BY 3.0
• Edda 2-7C and the capsized Alexander L Kielland — Norsk Oljemuseum — 1980 | Wikimedia Commons | CC BY 3.0
• Edda 2-7C and rescue boats after capsizing of ALK — Norsk Oljemuseum — 1980 | Wikimedia Commons | CC BY 3.0

The 27 technical plates in this video are illustrations generated for the channel by a diffusion image model, styled to match its cyanotype identity. They are diagrams of the system — the photographs above are the archive, the drawings are ours, not photographs of the hardware, and no person is depicted in any of them.

Root Cause investigates why engineered systems fail, using official investigation reports and primary technical literature. Sources for this episode are linked above.

Charts and dimensioned comparisons are drawn from the sources listed above; photographs of the Alexander L. Kielland are archive images, credited on screen and above.

#engineering #offshore #rootcause