Sleipner A-1 Gravity Base Structure
23 August 1991 in the Gandsfjord outside of Stavanger, Norway
Ksenia Tretiakova -Integrated BAE/MS Candidate (Dec 2012) in the Department of Architectural Engineering, Pennsylvania State University
Please note: All images in this article are generated by the author; In-text figure references include credit for any resources referenced during the illustration.


Figure 1: Simplified rendition of a typical Condeep platform built for natural gas drilling in the North Sea
The original Sleipner A (SLA-1) platform was to be used for oil and natural gas drilling in the Sleipner gas field in the North
Sea. It was built in the fashion of typical Condeep platforms (as seen in Figure 1[1][7]), but sank during a controlled ballast test, unlike other platforms with similar designs. It was found that a combination of poor geometry andinadequate design were the causes of the platform failure. The failure mechanism was concluded to be a shear failure that split open several walls in one of the platform shafts, which led to rapid intake of water.

Typical Condeep Platform

Condeep platforms are reinforced concrete structures meant to float in water up to 300 meters (~1000 feet) deep. These platforms consist of of a number of buoyancy cells that serve as the floating mechanism. Water ballast is pumped in and out of the buoyancy cells to control the depth of the lower portion of the structure (the Gravity Base Structure or GBS) in the water. Several of the buoyancy cells extend upward ("shafts") to provide structural support to the deck, which contains all of the buildings, machinery and amenities needed for several hundred people to live on the platform and drill for natural gas. This deck typically weighs around 55,000 tons[3].

Figure 2: Partially completed buoyancy cells to be floated to a fjord

The construction process for a Condeep platform starts on a dry dock, where all of the buoyancy cells and shafts are cast without enclosing the cell tops. The dock is then flooded, and the conglomeration of buoyancy cells is floated out to a sheltered location with deep water ("fjord"), where the buoyancy cells that will not be support shafts are capped off, and the GBS is slipformed upward (see Figure 2[1][8]). As more of the structure is cast, water ballast is pumped into the buoyancy cells to sink the structure and keep the construction close to the water. Once the GBS is completed, the platform structure must undergo a controlled ballast test to test the mechanical equipment under service loads and check for minor leaks. A controlled ballast test requires the platform to be sunk until the GBS is submerged, as it would be during deck-mating, which would follow at the end of the test and any necessary repairs. During deck-mating, the GBS is sunk until submerged once more, and the 55,000 ton deck is floated above the GBS; ballast is released from the buoyancy cells, and the GBS lifts the deck into the air. The completed platform can then be floated to its final location. For deck-mating, the GBS typically has to be about 20 meters (66 feet) deeper than it would be at service load.

As a result of the need to float a very heavy structure using concrete, extreme care has to be taken in design. One of the critical design factors for a deep sea platform GBS is the buoyancy cell wall thickness: if the walls are too thin, they will rupture during deck mating; if the walls are too thick, the structure will simply not float, or will prove difficult to move to its final destination. To deal with the fine line between "too thick" and "too thin", low factors of safety are used during the design, requiring both precise load estimation and precise analysis of buoyancy cell geometry[1].

The Sleipner A-1

The Sleipner A-1 platform was an example of a typical Condeep platform that "did not deviate significantly from earlier platforms"[2]. It was the 12th platform built by Norwegian Contractors (now Aker Oil & Gas Technology, Inc.)[5]. SLA-1 was designed using the same software that Norwegian Contractors (NC) applied to the other 11 deep sea platforms.
Figure 3: Plan view of SLA-1 buoyancy cells, with shafts and tri-cells highlighted

Construction for SLA-1 started in July of 1989, just under 2 years before the failure of the GBS[6]. As can be seen in Figure 3[2][3][4][7][8], the SLA-1 platform had 24 buoyancy cells, four of which extended into the shafts that supported the deck. Two of the shafts served as "drill shafts" while the remaining two served as riser and utility shafts. The GBS was 110 meters (361 feet) tall[1][7], and designed to operate in 82 meters (269 feet) of water[2][7][9]. The deck that would be mated to the SLA-1 GBS weighed approximately 57,000 tons[9].

The critical design detail in the SLA-1 platform was the connection of the buoyancy cells to one another, called a "tri-cell" (as shown in Figure 3[2][3][4][7][8]). In Figure 4[2][3][4], the geometry of a "representative" platform tri-cell (Gullfaks B[2][3][4]) can be observed and compared to the SLA-1 tri-cell geometry. The reduced thickness of the walls in SLA-1 tri-cells as compared to Gullfaks B is attributed to an effort by the NC designers to optimize the concrete volume and tricell geometry[3]. Figure 5.1[2][3][4][7][8] magnifies one of the joints in the tri-cells used for SLA-1 so that the reinforcement details can be observed[2][3][4]. Additional reinforcement detailing is shown in Figure 5.2[1][7]: a section of an elevation of one of the SLA-1 shaft walls[1]. The key items to note in Figure 5[1][2][3][4][7][8] are highlighted in red: the T-headed bar in the joint (5.1[2][3][4][7][8]) and the location of the last stirrup in the shaft (5.2[1][7]).

The initial cost to build the SLA-1 GBS was approximately $180 million (US).

Figure 4: Tri-cell geometry of a typical platform (4.1) compared to the geometry of SLA-1 (4.2)
Figure 5: Tri-cell joint reinforcement detailing (5.1) and shaft wall reinforcement detailing (5.2) with key features highlighted

The Accident

Sleipner A-1 was scheduled to undergo deck-mating on 1 September 1991. A second controlled ballast test was administered on the morning of Friday, 23 August 1991 in the Gjandsfjord near Stavanger (southwest coastal Norway, in the North Sea) after the platform underwent repairs for minor leaks discovered during the first test. The platform began taking on water uncontrollably at a depth of 97.5 meters[2][8][9] (320 feet)-- about 5 meters (16 feet) from the anticipated deck-mating depth[1]. The platform took between 17[3] and 18.5[2] minutes to become fully submerged. The 14 people on board at the time of the accident were successfully rescued with no injuries[2][4].

Summary of Conditions

The initial intake of water was denoted with a very "deep bang-like sound"[2] from one of the drill shafts, D3 (pointed out in Figure 3[2][3][4][7][8]). An eyewitness account indicated that water ingress could be seen in the lower portion of the shaft, at an elevation of 49 meters (160 feet) or less. The estimated size of the opening required to sink the SLA-1 platform in the the observed time was approximately 0.4 square meters (4.31 square feet)[2][9], which translates to a calculated crack length of about 7 to 8 meters (23 to 26 feet). At the time of failure, the D3 shaft wall was experiencing a differential water pressure of about 67 meters (220 feet) waterhead.

Undersea probes were released after the failure to explore the sea floor and observe the wreckage. The debris of the SLA-1 platform was scattered over an area of roughly 90,000 square meters (approximately 97,000 square feet). The top of one of the shafts could be identified in the debris. 192 pieces of SLA-1 were mapped across the sea bed, indicating a total collapse of the concrete GBS. As a result of the complete demolition of SLA-1, no physical evidence could be used to indicate the cause of the failure.

The timeline below outlines key events and actions that took place.


Morning of 23 August 1991
  • 05:48:30 - Platform is at a depth of 97.5 meters; a "deep bang-like sound" is heard, followed by strong vibrations and the sound of rushing water.
  • 05:49:00 (30 seconds later) - another, slightly weaker, bang is heard. Preliminary investigations follow; Drill shaft D3 is identified as the location of the failure, and emergency deballasting is initiated.
  • 05:56:30 (8 minutes after the first bang) - The call is given for the 14 personnel on board to abandon the platform.
  • 05:59:30 (13 minutes after the first bang) - The platform is fully evacuated.
  • 06:07:00 (20 minutes and 30 seconds after the first bang) - The platform completely disappears underwater.
  • 06:07:45 (21 minutes and 15 seconds after the first bang) - A magnitude 3 (Richter) event is registered at several local seismological stations.

It was observed that D3 sank first, corroborating the assumption that D3 was the only shaft to experience rapid intake of water.
Note: Summary of Conditions and Timeline summarized from Jakobsen (1992) [2]

Likely Causes

It was concluded that the tri-cell walls and supports at the cell joints were the weakest points in the platform; Calculations based on final design geometry showed that the load on these parts was at or very close to their maximum capacity. The investigators from NC concluded that "final failure was believed to take place as crushing of the concrete, presumably at the intersection between the tri-cell wall and the cell joint."[2] The investigators working for Statoil agreed and elaborated on this conclusion, reasoning that "shear failure occurred due to lack of transverse reinforcement."[3]
Figure 6: Probable failure location at tri-cell joint (6.1) and probable failure location on shaft wall (6.2)

Ultimately, the most probable failure mechanism for the SLA-1 GBS was found to be concrete crushing and shear failure that occurred in the portion of a tri-cell wall located above the transverse reinforcement, close the the joint. Figure 6[1][2][3][4][7][8] illustrates the probable location of the failure mechanism.

This failure mechanism manifested because of several inconsistencies in the initial conditions defined in the design software[1][2][8][9][10]. The sinking was partially caused by the inappropriate use of finite element (FE) -code NASTRAN[10] with regards to the global analysis of the finalized design[2][9]: the finite element mesh used to analyze the tri-cells was too coarse to predict the shear stress accurately. The flawed analysis and post-processing of the tri-cell design led to shear forces being underestimated by some 45%[1][2][9].

Another flaw with the use of the software was the human error involved: the design software involved considerable complexity, which led to a a very high perception of precision by those who used it. One of the features of this software was the way it ran calculations: it would only flag certain sections that were deemed critical by the software presets, and indicate those as sections that the engineers on the design team needed to check. The probable failure point of the SLA-1 GBS was not flagged.

Additionally, the supports for the tri-cell walls in SLA-1 were designed to only resist lateral forces indirectly, which meant that the detailing for the tri-cell joints had to be very carefully designed and analyzed[2]. It was found that the T-headed reinforcing (highlighted in previous figures) in the tri-cell joints was not long enough to resist the forces on the tri-cell walls. It was also found that the transverse reinforcing (also highlighted in previous figures) did not extend far enough up to resist the forces on the tri-cell walls[2][3][9].


NC began an investigation into the failure the same day that the SLA-1 GBS sank, as did Statoil (the company that would have operated the finished platform). The groups that carried out the investigation attempted to fit potential physical causes for the GBS failure to eyewitness accounts using a "Sequential Times Events Plot" system that diagrammed in-depth witness descriptions along a time axis to pinpoint significant events and their locations[6]. The investigative teams then followed up with analytical calculations to verify their conclusions.

Additionally, the company SINTEF carried out testing of small and full scale models to support the findings of the investigation team. These models included versions of the existing (collapsed) structure, as well as modified versions to correspond to investigative theories regarding the initial inadequate reinforcing at the joint details and in the shaft[2][4][8][9]. The models tested with the "improved" design provided evidence to suggest that, had the reinforcing been appropriately designed, the load capacity of the tri-cells would have been about 70% higher[2][3], and the SLA-1 GBS would not have been a total loss.

The findings among the NC investigation team, the Statoil investigation team, and the SINTEF investigation team coincided very well. As a result of the consistency of the findings in each independent investigation, Statoil chose to rebuild the Sleipner A platform[9], with some modifications to assure safety and structural stability.

Reactions and Outcome

Figure 7: Tri-cell geometry of SLA-2 (7.1) compared to the geometry of SLA-1 (7.2)

The fact that two simples changes to the reinforcement detailing would have prevented an extremely expensive structural failure has resounded widely among structural designers. The total loss of the Sleipner A-1 platform is an unfortunate, but excellent, example of the need for extreme care and detail in design. The SLA-1 accident demonstrates the importance of having experienced engineers verify computer-generated design work to ensure the proper use of analysis and design techniques[9][11].

The initial cost for the SLA-1 GBS was about $180 million. Statoil estimated their losses without the SLA-1 platform came in at about $1 million per day. It took over two years to redesign the Sleipner platform and bring it to fully operational status[4].

Revised Design Philosophy

The design philosophy after the SLA-1 GBS collapse was focused on establishing additional, extensive risk analysis criteria that includes analysis for impact damage, and design for platform stability in the case that one of the buoyancy cells or other compartments is damaged[3].

Additionally, the revised design philosophy called for greater forgiveness in the initial design in order to account for any potential future changes that might occur later in the design process. Additional load factors were applied to all new platform designs. Once the design was completed, the new philosophy required qualified engineers to check every pre-identified critical section-- not as flagged by any software in use, but as identified to be a potential weak section in any design by those who investigated the SLA-1 GBS collapse[3].

The New Sleipner A Platform

The new Sleipner platform (SLA-2) was designed entirely using hand calculations. This was the result of a decision by Statoil to continue through with the design while investigators continued troubleshooting the software used to design all of the NC platforms in the North Sea. The stipulation behind the SLA-2 design was that, in addition to hand calculation and design, all of the buoyancy cells in the platform were modeled and analyzed before they left the dry dock for further construction[3].

The designers of SLA-2 changed the cell wall geometry to minimize stress concentrations (see Figure 7[2][3][4] for the differences in geometry). The criteria for the new design included a 10% higher load factor when accounting for water pressure. The formula for shear capacity was revised as well[3].

Overall, the total time spent on verification of the design was quadrupled in comparison to the design of the SLA-1 GBS. The SLA-2 platform was up and operation in October of 1993, more than 2 years after the initial SLA-1 GBS collapse. The overall estimated cost of the SLA-1 GBS collapse totals to $700 million, after taking into account the downtime losses and cost to redesign the platform[1][3].


Works Cited

[1] Collins, M., Vecchio, F., Selby, R., Gupta P (2000) "Failure of an Offshore Platform," Canadian Consulting Engineer - Structures, pp 43-48, March/April
  • Journal Article. Provides a wealth of information on the typical construction and design of a Condeep platform; Also describes the Sleipner A-1 design, failure mechanism, failure causes, and the failure aftermath (including information on the construction of Sleipner A-2). Discusses estimated costs and lessons learned as well.

[2] Jakobsen, B. (1992) “Loss of the Sleipner A platform,” The proceedings of the International Offshore and Polar Engineering Conference, pp. 1-9
  • Conference Proceedings. Written by members of the Norwegian Contractors' investigation team. Describes investigations into the platform collapse; Discusses eyewitness accounts that helped locate the probable point of failure, as well as the most likely causes for failure and collapse (and why); Also describes some model tests that have been performed to verify the conclusions in the paper.

[3] Rettedal, W. (1993) “Design of concrete platforms after Sleipner A-1 sinking,” Proceedings of the International Conference on Offshore Mechanics and Arctic Engineering – OMAE, pp. 309-316
  • Conference Proceedings. Discusses new design methods for offshore concrete platforms; Describes full-scale testing of certain joints to identify failure modes, as well as the revised design philosophy (including discussion of loads, design criteria, etc.). Also covers the design of the platform that replaced Sleipner A-1, and the verification of that design. Additionally, discusses risk analysis and some forethought on future designs.

[4] Jakobsen, B. (1994) “The Sleipner Accident and its Causes,” Engineering Failure Analysis, Vol. 1, No. 3, pp. 193-199
  • Journal Article. Written by members of the Norwegian Contractors' investigation team. Describes the initial collapse as well as the factors that determined the probable cause of the failure. Also discusses the additional model tests to verify failure modes on the joints. Compares results from model tests to eyewitness accounts to verify the actual failure mode.

[5] (1996) "Norwegian Companies in USA - - Aker,", n.d., Web
  • Periodical. News article from the official site of Norway in the United States. Regarding the acquisition of Norwegian Contractors (NC) by Aker Oil & Gas Technology, Inc. Discusses company holdings and history, and includes contact information for Aker.

[6] Ynnesdal, H., Berger, F. (1994) "The Sleipner Accident," Proceedings of the Second International Conference on Health, Safety & Environment in Oil & Gas Exploration & Production, pp. 715-716, 25-27 January
  • Conference Proceedings. Provides a general summary of the entire story on behalf of the Society of Petroleum Engineers, emphasizing the usefulness of a "Sequential Times Events Plot" (STEP) in pinpointing the probable cause and location of the failure based on witness interviews. Briefly covers the base facts of the accident, then reviews the methodology of the investigations, involving the "STEP" system to track the course of events of the accident.

[7] Schlaich, J., Raineck, K-H. (1993) "Die Ursache fur den Totalverlust der Betonplattform Sleipner A," Beton- und Stahlbetnbau, Vol. 88, pp. 1-4
  • Journal Article. Article is in German. Title translates to "Causes for the total loss of the Sleipner A concrete platform." Primary usefulness of this article for this wiki page is the availability of clear illustrations for portions of offshore platform geometries to reference in figure generation. The article recounts the accident, describes the conditions during the accident (with great diagrams), includes diagrams for the finite element meshing of various portions of the GBS, also shows the loading conditions for the tri-cells.

[8] Arnold, D. (2009) "The sinking of the Sleipner A offshore platform," Institute for Mathematics and its Applications (IMA) at University of Minnesota, 7 September 2009, Web.
  • Website. Produced by the director for mathematics at the University of Minnesota. Provides an excerpt of one of the elusive SINTEF reports and provides a general summary of the platform, the failure, the circumstances of failure, and where to look for additional information.

[9] Jakobsen, B., Rosendahl, F. (1994) "The Sleipner Platform Accident," Structural Engineering International, IABSE, Vol. 3, pp. 190-193
  • Journal Article. Produced by members of the NC investigation team. Summarizes the accident and discusses lessons learned. Mentions several important dates for the rebuilding of the structure, and reasons for choosing to build SLA again. Published in the "Lessons from Structural Failures" section of the journal.

[10] Huckle, T. (2011) "Collection of Software Bugs," Institut fur Informatik, TU Munchen, 7 November 2011, Web
  • Website. Produced by a professor at the university. A large and relatively comprehensive list of accidents and disasters that occurred as a result of an error in the design software. Provides names, dates, summaries of the failures, and reasons for failure. Also includes additional resources to find more information on each accident. SLA-1 is #15 in the list.

[11] Holand, I. (1996) "Structural analysis of offshore concrete structures," IABSE Congress Report, IABSE, Vol. 15, pp. 879
  • Technical Report. A report made available to the public on the web by the International Association for Bridge and Structural Engineering. Discusses how the Sleipner accident is a good example of the necessity for verification, risk analysis, and quality assurance.

Works Read

Stead, B. (1992) “Importance of fabrication engineering in the early phases of the Sleipner a development,” Proceedings of the International Offshore Mechanics and Arctic Engineering Symposium, Vol. 1, pp. 543-548
  • Conference Proceedings. Describes the conclusions drawn that led to the initial design choice for the offshore platform (discussions include topographical factors that led to the choice of a gravity-only system, costs for construction as well as the various hookups required for the typical offshore platform, and considerations leading to the choice of concrete)

Vecchio, F. (2002) “Contribution of Nonlinear Finite-Element Analysis to Evaluation of Two Structural Concrete Failures,” Journal of Performance of Constructed Facilities, pp. 11-115, August
  • Journal Article. Describes 2 failures, one of which is Sleipner A; Recounts the event itself, summarizes what happened, includes some discussion of loads and dimensions in the summary. Discusses the nonlinear finite-element modeling analysis the author performed, and how it accurately portrays load capacity and failure modes: justifies use of computer modeling as a way to analyze a structure.

Wackers, G.(2004) “Resonating Cultures: Engineering Optimization in the Design and (1991) Loss of the Sleipner A GBS,” Center of Technology, Innovation and Culture, University of Oslo, May, Web
  • Web-based Discussion Paper. This is a work in progress paper, with detailed discussions of… the accident itself, as well as some discussion of the theory behind design parameters. Also includes discussions of the appropriate way to approach investigations for failure modes and mechanisms. Extrapolates to some cultural stuff, and then continues with an evaluation of changes to the design of the new platform. Also talks about different forms of analysis, as well as the engineering of the platform design.

Thompson, M., Jirsa, J., Breen, J. (2006) “Behavior and Capacity of Headed Reinforcement,” ACI Structural Journal, Vol. 103, No. 4, pp. 522-530, July/August 2006
  • Technical Report. Mainly discusses design recommendations for headed reinforcement, and why they are recommendations in the first place. Uses the Sleipner A failure as a model for the importance of appropriate design of headed reinforcement.

Additional Reading

Holand, I., Lenschow, R. (1996) “Research Behind the Success of the Concrete Platforms in the North Sea,” Proceedings of Mete A. Sozen Symposium – A Tribute From His Students, ACI SP-162, pp. 235-272

Selby, R., Vecchio, F., Collins, M. (1996) “Analysis of Reinforced Concrete Members Subject to Shear and Axial Compression,” ACI Structural Journal, Vol. 93, No. 3, May/June

Bea, R. (1994) “The Role of Human Error in Design, Construction, and Reliability of Marine Structures,” Ship Structures Committee

Gudmestad, O., Holand, I., Jerson, E. (2000) “Design of offshore concrete structures,” 1st ed., Spon Press

The SINTEF reports on Sleipner A (STF22): A97725, A97854, A97857, A97859, A97861, A97833 (there are 13 more available somewhere)