The Failure of an Offshore Platform Pifra ame iiteko a ron toms rsa ae Cer (photo courtesy of Aker Norwegian Con- ptereLed ame Pc he challenge of extracting oi and gas from beneath th Noth Sea, one of the world’ most hostile ocean environ= ments, Ted to the development of Condeep platforms. Standing: in water Gepths of up t0 300 m (980 f), these el- ant reinforced concrete structures are pressive feats of structural engine g that have advanced the art of con- crete design and construction. As such, they are worthy of comparison with the Roman Pantheon, a much of-the-art conerete structure, which also had as its Key components a con- rele dome on top of a concrete eylin der Fig, 1), The construction of atypical Condeep platform starts in a large dry Gock where the lower domes and part of the cylindrical walls of the cluster of buoyaney cells are cast, After flooding of the drydock, the partially completed structure is floated out and anchored at a deep-water site ina sheltered Norwe- giun fjord. As the slipformed construc tion extends the structure upwards. solid ballast and water batlast are added to the buoysaney cells to lower the base of the stricture deeper into the water. Usually three or four cells are extended ls to form the shafts, which will support the deck and provide conduits for the drilling and the oil pipes, When the conereie structure is completed, ad- ditional water ballast is added until the top of the concrete structure is nearly submerged, At this stage the top deck of the plat form, which provides accommodations for about 200 people and supporss the rilling equipment and process equip- iment, all of which may weigh about 40,000 tonnes (4,000 tons), is floated Cover the top of the concrete structure. Ballast water is then pumped out of the buoyancy cells and us the concrete siruclure rises it mates with and lifts the deck structure. After deck mating, the completed structure is towed to ils off- shore site and lowered to its final desti- nation on the sea floor A critical factor in the design of deep-water concrete platform is the thickness of the walls, Ifthe walls are too thin, they may fail under the very high water pressures to which they are subjected during deck mating. Howey er, unlike the situation for a typical land-based structure, the de: not have the option of gr ing the wall thickness 10 conservative design. Ifthe slier state- walls are too 28ter once the tops of the cells were submerged. Because of this. the walls ofthe tricells had to resist a substantial hydro static pressure thick, the structure will not float, or will not be hydrostat cally stable during the tow to the field, These severe const mean that Forthese weight-sensi tive structures, rather low fac eee (On August 23, 1991, ballast tors of safety are employed. As as : water was being pumped into a consequence, great care is re- peck the buoyaney cells to cause the quired in all pects of ign —— structure to descend at arate of aind construction. Ea about 1 m (3 f) every 20 min- ‘On August 23. 1991 the con- ee utes. It was intended 10 lower crete ise structure for the “ the structure until its base was Sleipner A platform was being 104 m G44 ft) below the su lowered into Gandsfjond in fuce. However, when a depth preparation for deck mati In ly of 99 m (325 ft) was reached a comparison to the 11 previous: loud rumbling noise was heard Condeeps, Sleipner A was arel- from one ofthe (wo drill shafts, atively. small structure’ planned Water could be heard pouring fora water depth of 82.5 m 271 into the drill shaft from a loew- ft) Gig. 2), During deck mating tion that was estimated to be a Condeep structure is about 2( about 2 m (6.6 ft) above the mm (66 {t) deeper in the water surface of the ballast water than itis during operation, Thus, F ). Afier a few minutes, fora structure planned fora 145 the structure was sinking at a m G76 fh) deep site deck mat- tate of about one meter every ego EE RS reRES 3 minute and therefore had to be ‘at the base by about 14 percent Fig.1 — A Condeep offshore oil platform and the Roman inpendloncds Arey gan wea at ind would inerease the pressure. Pantheon: {Wo slate-of-the-art concrete structures. fer jit disappeared below the Se Ra einen Details of the structure and buoyancy cells collapsed. A local scis- Oe et tata, of the collapse ecopisnlon exited beech ufing deck mating was to be about °° The concrete gravity use structure of magnitude 3.0 earthquake, ate sor Pei Fis Boyan coi the Sleipner A platform, which was 110 vey of the bottom of the 220 m (722 1) ie epieonune mae cells, When iy) (361 ft) high, consisted ofa cluster of deep fjord revealed that no debris lurz~ the structure was about 5m (16f) from 34 cers, four of which extended up- er than 10 m (33 ft) remained. the planned deck-mating depth @ cell Wards 1o form the shafts (Fig. 3 and 4). The loss of the structure was ultribut- I failed, allowing water to rush into Wife the exterior walls of the cells ed to the failure of the wall of Tricell 23 the drill shaft. The emergency debal- yore circular. with a radius of 12m (39 adjacent to drill shaft D3 (ig. 3). At lasting pumps could not keep up with Fy), the interior walls, which separate failure this 550 mm (22 in.) thick wall the water flow and hence, the structure he celfs, were straight. Atthe intersec- wats resisting a 65 m (213 #0) head of ank. As it went deep into the lord, the gion poiats of these interior walls. a sea water. resulting in a pressure of buoyancy cells imploded, totally de smatt triangular void called a tricell about Nin? (13.7 kips/tt?)_ as stroying the $180 million structure. AIL ag formed. There were a total of 32 shown in Fig, 6. Ashe clea span othe as apileof rubble atthe such tricells. Because these ticells had) wall was 4.478 m (14.36 0 the shear at bottom of the fjord. ‘Openings al the top, they filed with wa- each end of the wall must have been In the weeks following the Sleipner WPL = 655 x 4.3782 = 1434 uceident a number of investiga- {N/m (98.3 Kip). Ifthe flex tions were launched with the : HPP ural ssiiness ofthe tricell wall ssim of identifying the cause of 4 alt J Was unitorm along its span the © fixed-end moment would be pie/12. Flexural eracking at the M ends of the wall. would have one such investigation, whi | = caused some redistribution of studied the shear strength of the ' : these moments. For this mem- w ber the ACT Code! would sug- how this strength was influ- Maly =: gest that this redistribution will enced by reinforcement detil- reduce the “negative moment” The results of the study i by about 17 percent. resulting indicated that the current ACI 5 in an end moment of about 870 code provisions for the shear i kNn/m (196 kip-fvf. sirength of members subjected If an individual buoyancy fo high axial compression can Fig 9 A comparison at €ome Condesp platform (photo. El! Was separated from the rest cviymemeane ign Ae aueneane 0 of the structure and was sub- —_—_————— eee 30 Concrete Internationalups stopped just below the failure location (Fig. 5), The other reinforcing detail that is nportant to note in Fig, 7is the T-headed bar placed across the throat of the tricell joint, This 25 mm diameter bar was about 1m G fi) long and had steel plates welded on its ends to provide anchorage. Fig. 8 is a photograph of the reinforcing details near the wi- call joint taken in June 1990, when the construction had ached the top of the ticell walls. As can be seen, at this lo- cation there were stirrups in the tricell walls and three T-headed ars across the throat of the join ‘The specilied concrete quali- ty for Sleipner A was C6: ‘which implies that the charse teristic concrete cube stre is at least 65 MPa (9425 psi) at an age of 28 days. It is estimat- ced that atthe time of the fail the concrete cylinder strength ‘was at least 60 MPa (8700 psi) jected to an external pressure of p. the circurnferential compres- sion in the walls of the cell, away from the end dome would be pR, where & is the r ius to the outside fice, Thus, for a pressure of 655. kNin (13.7 kips/f@)and a radius of 12.5 m (41 fy) the axial com- pression would be 8189 kNém (361 kips/ft), When the 24 cells are joined together, the determi- nation of the axial compres- sions in the different walls is a more complex problem. The duces the diameters ofthe cells. However, at the top and bowom of the cells, the horizontal stiff ress of the domes will prevent the overall. dimensions of the cluster from being substantially reduced. As aresult, the vertical | centerlines of the exterior cells will bend inwards towards the center of the cluster. This so-called “caisson effect” will reduce the axial compression in the walls of the cells, with the reduction becoming greater to- The specified reinforcement wards the center of the cluster. grade was K5QOTS, which im- Based on a finite element anal- Plies a “minimum” — yield ysis ofthe total structure itises- Fig, — Geometric details of the Sleipner concrete base strength (lowest 5 percent) of limated that the axial structure 500 MPa (73 ksi) and means compression in the walls of Tri- that the average yield strensth cell 23 atthe location of failure would have been about was 5000 KN/m (343 kips/b. MPa (80 ksi). ‘The reinlorcement details in the iret! walls near the failure location are described in Fig. 7. Nonlinear finite Acrid ollberizonial undivertical element analyses of bars was provided near each the tricell face of each wall, The bars had To develop a better understand a diameter of 25 mn (1 in.) and ing of the factors influencing were spuced at 170 mm (6.7 the failure of the ticell wall in), eenter to center. On the in- series of nonlincsr finite ele- side face of the tricell walls ad- ‘ ‘ment snalyses were conducted ditional horizontal. ars were 4 using the SPARCS program.*° pluced near the ends of the * his program, which was de- walls. Thus, at these locations " ; veloped at the University of there were two bars every 170 Toronto, is formulated to mod- ‘mum, In addition to the grids of S el the three-dimensional re horizontal and vertical 25. mm sponse of reint onerete diameter bars, the ticell walls E structures. It uses a secant stiff also contained 12 mm (0.5 in.) Je . 7 neys based solution scheme that diameter stirrups. For shout the Ja ‘ wolves. substituting succe: botiom third of the height of the E ively better estimates of mate tricell walls these stirrups were i rial stiffness into a linear elastic spaced 170 mm upart horizon- 4 finite element algorithm. The tally and 170 mm apart vertical- brick, wedge, und truss el ly, Near mid-height of the walls 4 a ‘ments are bused on linear di the spacing was 170 mm apart Fig. 4—The nearly complated concrote base structure placement functions. These horizontally and 340 mm (13.4 floating in Gandsfiorxsphoto courtesy of Aker Norwegian low powered elements, when in.) apart vertically. These stir- Contractors) used in sufficient quantity, can ‘August 1997 31Fig. 5 — Details of dil shaft D3 at the time of failure. predict accurately the Joad-dleforma- tion response of reinforced concrete structures provided — thatthe stress-strain relationships for the cracked concrete are modeled accurate ly. In SPARCS, these constitutive rela- tionships are derived from the modified ‘compression field theory." The finite clement model of the tri- cell is described in Fig. 9. Because of symmetry, only one-sixth of the trill needed to he modeled. The one element mesh contained 342 brick ele- ‘ments, 338 wedge elements and 1172 nodes. Roller supports were used to ‘model the symmetry conditions at mid span of the tricell wall and along the centerline of the 800 mim (31 in.) thick cell wall, Roller supports were also Used to restrain the vertical movement of the bottom layer of nodes. een ERE Fig. 7 — Reinforcement details for tricell 23. In the analyses the horizontal a force in the 550 mm (22 in.) thick ti cell wall was held constant at 5000 Nim (343 kips/fi) while the hydrostat- ic pressure on the inmer Face ofthe wall was increased until failure was pre cd to occur. In addition, a vertical com pressive stress al 7 MPa (1015 psi) was applied to both walls und was held con- stunt as the hydrostatic pressure in- creased, The first run of SPARCS program, predicted that the as-built structure ‘would fail when the applied water pres sure on the inner faces of the tticell reached 625 kN/m? (13 kips/{t?). ‘This ‘of sea water 62 m (203 1) hig agreement with the estimated 65. m (213 ft) head that caused Slefpner to fail. The predicted pattern of deflee: tions at fuilure is shown in Fig, 10, Note that failure is associated with a di- agonal band of exteemely distorted ete- ments near cach end of the tricell’s walls. Also note that the thickness of the walls is substantially increasing near their ends. This “bulging” of the section combined with the diagonal pattern of damage, indicates « shear failure of the wall The designers of the structure were interested in how the strength of the tri cell wall would haye changed if the stirrups, which were used just below the failure location, had been continued higher up the wall. They also wanted (0 Know how the length of the T-headed bar influences the failure. ‘To answer these questions a total of 14 different nonlinear finite element analyses were conducted, For one set of 7 analyses the walls of the tricell were assumed 10 contain 12 mm (0.5 in.) diameter stir- rups spaced at 170 mm (6.7 in.) hori- zontally and 340 mm (134. in vertically, while for tricel = nay Fig. 6 —Detais of the geometry and loading for tricell 23, the other set the walls were assumed to contain no stirrups. Within each set the length of the T-headed ar was varied from zero (i... no bar) to the maximum le length, which was 1.5 m (3 ft) 7). ts ofthe additional finite el cement analyses are summarized in Fig. 11, which shows both the predicted failure pressures and the predicted de- flected shapes at failure. For the two ceases with no T-headed bars, failure is predicted to occur at a pressure of about (0.6 MPa (87 psi) — a 60 m (197 ft) head of water — by yielding of the re- inforcing bars crossing the throat of the tricell, For the cases where there are no stirrups, adding 08 m (26 f) long T-headed bars to the throal docs not in- crease the failure pressure but changes the predicted mode of failure from a ‘lexual failore in the throat region to a shear failure near the ends of the tricell Fig. 8— Placement of T-headed bars across the throat of the 22 Concrete International| Aeteee 0.8 mi 1.5 m5 fh) only: marginally. in- otersuppons | creases the strc \ | as the failure be- comes govemed by the flexural capacity ae of the throat region ae Tiitt Tt (ig. 11). The results Net iydronaie pressure | of these studies indi- cate that the tricell cea 0 bye seses could have resisted Teottieen 3 edge cements bout an additional 20 m (66 £1) of water head if either the stir- ones: Retotorcement ran nial bee tens fxooura 4 580Mra tinued up the wall or ‘wall. As the length of the '-headed ba is increased beyond 0.8 m, the predict- ed failure pressure increases and the zone of shear failure moves outwards When the T-headed bar is 1.5 m long, the predicted failure pressure is about 0.85 MPa (123 psi) — an 85 m (279 fi) head of water — and the failure mode ‘changes back to s Nexural failure of the throat region. When the tricell walls do not con: stirrups the T- n ally increase the strength of the tricell until the length of these bars is long enough to penetrate about three-quar- {ers of the way into the cell wall (ic. @ length of 1.3'm [4.3 ft). However, if the cell walls contain. stirrups. the ‘T-headed bars significantly increase the strength of the tricell once the burs are long enough to penetrate one-quar ter of the way into the cell wall (ic. length of about 0.8 m [2.6 It). For this cease, extending the T-bar length from if the T-headed bars in regions with no stirrups had been about half a meter longer. Shear strength calculations using ACI and AASHTO The tricell wall that fuiled did not co tain stirrups because the global fini element analysis performed as part of the design seriously underestimated the magnitude of the shear at the ends of the wall, while the sectional design pro- cedure used! seriously overestimated the beneficial effects of the axial con pression on the shear strength of the ‘wall The design procedures used to es- timate the shear strength of the wall were those contained in the 1977 No wwegian concrete code,? which had been influenced by the shear provisions of the 1971 ACT building code.6 These ACT shear provisions were formulated in the period following the Air Force Warehouse collapses,” failures which were believed t0 be due to the detti- mental effect of axial tension on shear sirength, Thus, itis not surprising that Fig, 10—Dellected shape of the tricell at failure as predicted by tho finite element model, ese provisions, wnica remain un- ‘changed in the current ACI coxte,! su gest that axial tension. substantially reduces shear capacity while axial substantially increases compression! shear capa The failure shear-sxial load imerac- ‘ion diagram for the Sleipner tricell wall section calculated using Equations (IL-5) to (11-8) of ACI 318-95! is shown in Fig. 12. Also shown in this figure is the fuilure shear-axial load in. leraction diagram calculated using Sec- tion $.8.3 of the 1994 AASHTO LRED Specificutions.* These AASIITO shear design provisions are based on the modified compression field theory*"” and reflect sudvances in understanding of shear behavior that have occurred in the last 25 years. It can be seen that while, for this section, the two sets of shear provisions lead to similar esti- mates of shear strength when the axial load is zero. the predicted changes in sheur strength with change of axiul load are very different. For the Sleipner tr- cell wall the axial compression on the wall increased as the shear fore in: creased. In this siluation the predicted ‘EAR FORCE (hm) Fig. 11 — The influence of stirrups and length of the T-headed ‘oars on the failure pressure and failure ‘mode of the tricell Tenton={=Coneesion. IAL LOAD Hm) Fig. 12— Shear force-axial oad interaction diagram for tho tricell wall of the Sleipner platform. ‘August 1997 33failure load is very sensitive to the assumed intersetion be- ‘oven shear strength and axial Toad, As can be seen from Fig. 12, for the Slefpner loading ra- tios the ACI interaction di gram predicts a failure load corresponding to a 120 m (394 fi) water hewd, while the AASHTO interaction diagram predicts a 45 m (148 f) water head at failure Wall elements subjected to combined axial compression and shear Inthe six years since the Sleipner fail- wea number of experiments!" hhave been conducted in the University ‘of Toronto’s shell element tester"? to study the response of reinforsed con- crete wall elements subjected to com. Fined axial compression and. shear Some of these experiments were con- ducted to assist the engineers involved in the design and construction of the re- placement Sleipner platform (Hig. 13). Most of the tests were aimed at ient- fying appropriate design techniques for teinforeed eonerete members subjected to compression and shear ‘The results of one series of tests in which three wall elements, PCI, C20 and PC2, were loaded at cif éni ratios of axial compression to she ane ilustited in Fig. 14. Also shown in this diagram are the shear strengths for this wall section predicted by the AC and the AASHTO shear design provi- sions. It ean be seen that the AASHTO prodictions are considerably more ac- curate. Further. while all three of the specimens failed in shear, the ACI pro- Fig. 14 — Shear force-axial compression interaction diagram for reinforced concrete wall. ‘rat coumnesion Fig. 13 — Engineers from the Sleipner project team test a spacimon representing a portion of the rebuilt platform. jons predict that specimens PC20 nd PC19 would fail in combined flex- ure and axial load prior to reaching their shear capacities, Specimen PC21 was subjected. to loading ratios that were reasonably comparable lo those experienced by the tricell wall of the Sleipner platform, The observed load-deformation re- sponse and crack development for this specimen are recorded in Fig. 15. Flex- ural cracking was predicted 10 begin near the ends of the specimen when the applied shear reached about 100 kN 2255 kips). As the eracking developed there was a substantial reduction in the stiffness of the clement resulting in a nonlinear Toa-deformation response By the time the shear reached 600 KN (135 kips) — loud stage 4 —a number of small diagonal cracks had formed and small expansions of the wall thick ness (i... bulging of the wall) were re~ corded. By load stage 6, at a shear of 770 KN (173 kips), the diagonal eracks had reached a width of about 1.5 mm (59 mils) and the wall had increased in thickness by about 0.75 mm (30 mits), pera toes Je, While the readings were being taken at load stage 6, the defor- mation of the specimen was held constant and as a result the load decreused. When the specimen was reloaded after oud stage 6, a loud thud was heard as a sudden shear failure of the wall occurred, A wide di- agonal crack formed at shout 25° to the longitudinal axis of the member (Fig. 16). It should be appreciated that the form tiom of such cracks at the ends of the tricell walls in the Sleip- ner platform would result in the sinking of the structure. Conclusion The Sleipner concrete gravity base structure, which was destroyed on Au- a 1991, had taken about thre Years to design and construct, During this period extensive use was made oF the sophisticated computer software that had been developed for the design of previous Condeep platforms. These alobal analysis and sectional design software tools enabled several thou sand locations on the structure to be checked for several hundred different load cases. It is indicative of the per- ceived precision of the design and con struction that the thickness of the curved extetior walls of the buoyancy cells wats speeitied to be 490 mm (19.3 in) rather than 500 mm (19.7 in.) as shown in Fig. 3. The software identi- fied rtical locations and. loadings Which the engincers could check man- ually. Unfortunately, because the up plied shear was underestimated by the lobal analysis and the shear strength Wwas overestimated by the sectional DEFLEENON fon) Fig. 18 —Load-deformation response of reinforced concrete 34 Concrete Internationalanalysis, the ends of the tricell walls ‘were not identified as critical locations Alter the failure of the structure it was clear that there were major prob- lems with the previous design calcula- tions, ‘The trivell wall failed under water head of about 65 m (213 ft) whereas it should have been capable of safely resisting a water head of 70 m (230 f). ‘To give a factor of safety of 1,5 the wall shoutd not have failed until the water head reached 105 ) It was recognized that finding and cor recting the flaws in the computer anal ysis and clesign routines was be a major task. Further, wit come from the lost production of the tas field being valued at perhaps $1 million a day, it was evident that a re- placement structure needed to be d signed ancl builtin the shortest possible time A decision was made to proceed with the design using the pre-computer. slide-rule era techniques thal had been used for the first Condeep platforms designed 20 years previously. By the time the new computer results were available, all of the structure had been designed by hand and most of the stnic- ture had been built. On April 29, 1993 the new concrete gravity base structure ‘was successfully mated with the deck and Sleipner was ready 1o be towed (0 sea (See photo on title page). ‘The failure of the Sleipner hase struc- ture, which involved a total economic loss of about $700 million, was proba- bly the most expensive shear failu er. The accident, the subsequent investigations, and the successful rede- sign offer several lessons for structural engineers. No matter how complex the structure or how sophisticated the com- puller software it is always possible 10 Obtain most ofthe important design pa- rameters by relatively simple hand cal- culations, “Such calculations. should always be done, both to check the com- puter results and to improve the engi- neers’ understanding of the critical design issues. In this respect itis im portant to note that the desigm errors in Sleipner were not detected hy the ex- tensive and very formal quality assur- ance procedures that were employed. When desi for shear, even tor shear in walls, itis prudent to be gener ous with the use of stirrups. A portion of the tricell walls, about 15.m (49 1) in height, did not contain. stirups. [Tt ‘would have taken about an additional 70 tonnes (77 tons) of stirups to rein- force this height. ‘The analyses de- scribed in this paper indicate that if these stirmips had been provided the platform would not have failed. If an engineer is faced with des rein: forced concrete elements subjected to high compression and shear it would be unwise to use the shear provisions of Fig. 16 — Reinforced concrete wall affor failure, the current ACI building code as thes provisions can be dangerously uncor servative. In this situation the shear provisions of the AASHTO-LRFD specifications will result in a more cor servative and more accurate estimale of shear eapacity Reterences TACT Coaminee 318, “Bulking Code Requre- sft Stettel Conscte (ACT 31898) end Cone ary ACT 189095." autesan Const lastin 3, Sey, RG, “Tce dimensioal Coasieshe fons for Renfrcof Conese” PAD. desl. Ds ‘neat of Cll Eapacteg, Unicity eT, 1993 3 PP i Vecchio, Fi, and Colne MP, “The Moise Compression Feld Ts for Rasinool Chae Ed ects Sujet Shea 8. acl Commis fr Rlaored Shs, Deri 197.789. 7. Ekaic, RC, aid Hognesad. “Lab Asia Ft le ais. toe 188,11 9, Cog, MP. a Mi tree Srctres, Prone Hal Ea Acknowledgments ts pope a a foal Newark of Cents of Ex Testy, Cara : S Received and x ‘som iis, ‘ACI Fetiow F Michael P. Collins | & js Bahen-Tanen- ey baum Professor of | civil engineering the University of ‘ Toronto, Ontario, | s Canada. He is a member of ACI-ASCE Committee 44, Shear and Torsion, and Com- | mittee 358, Concrete Guideways. | fF ‘ACI Member Frank | J.Vecchio is a pr0- | {essorin the de- | By partment of evil | engineering atthe | University of Toron- | 46, Dy, to. Heisamember | Of ACI Committee | 441, Roinforced Concrete Columns, ‘and Commitice 447, Finito Element | Analysis. | Robert G, Selbyis | esearch engineer | with Morrison Her- shield Ltd of Tor- | onto, He received his BSfromtho | Ppp vrversity of Albor- taandhis PAD from the University of Toronto. Pawan. Guptais Pho candisten the depanmentot ivi engneerng | fhe Unverssy Toronto, Ontario, Canada. ‘August 1997 35