632/1998 February 1998

Catastrophic Failures of Steel Structures in Industry: Case Histories

By: B Hayes and R Phaal, TWI

Contents
Executive Summary 1. Introduction 2. Hasselt Bridge 2.1 Summary Details 2.2 Background 2.3 Causes of Failure 2.4 Lessons Learnt 2.5 References 3. Schenectady T2 Tanker 3.1 Summary Details 3.2 Background 3.3 Causes of Failure 3.4 Lessons Learnt 3.5 References 4. Fawley Crude Oil Storage Tank 4.1 Summary Details 4.2 Background 4.3 Causes of Failure 4.4 Lessons Learnt 4.5 References 5. World Concord Tanker 5.1 Summary Details 5.2 Background 5.3 Causes of Failure 5.4 Lessons Learnt 5.5 References 6. The Kings Bridge 6.1 Summary Details 6.2 Background 6.3 Causes of Failure 6.4 Lessons Learnt 6.5 References

7. Sizewell Boiler 7.1 Summary Details 7.2 Background 7.3 Causes of Failure 7.4 Lessons Learnt 7.5 References 8. John Thompson Pressure Vessel 8.1 Summary Details 8.2 Background 8.3 Causes of Failure 8.4 Lessons Learnt 8.5 References 9. Cockenzie Power Station Boiler Drum 9.1 Summary Details 9.2 Background 9.3 Causes of Failure 9.4 Lessons Learnt 9.5 References 10. Typpi Oy Ammonia Plant Water Coolers 10.1 Summary Details 10.2 Background 10.3 Causes of Failure 10.4 Lessons Learnt 10.5 References 11. Robert Jenkins Pressure Vessel 11.1 Summary Details 11.2 Background 11.3 Causes of Failure 11.4 Lessons Learnt 11.5 References 12. M V Kurdistan Tanker 12.1 Summary Details 12.2 Background 12.3 Causes of Failure 12.4 Lessons Learnt 12.5 References 13. Alexander L Keilland Accommodation Platform 13.1 Summary Details 13.2 Background 13.3 Causes of Failure 13.4 Lessons Learnt 13.5 References 14. Union Oil Amine Absorber Tower

14.1 Summary Details 14.2 Background 14.3 Causes of Failure 14.4 Lessons Learnt 14.5 References 15. Ashland Storage Tank 15.1 Summary Details 15.2 Background 15.3 Causes of Failure 15.4 Lessons Learnt 15.5 References 16. Summary and Conclusions 17. Acknowledgements

Executive Summary
A series of fourteen case studies are described, where failure has occurred by fracture, and the lessons which can be learnt from these failures are discussed.

Background
Fortunately, catastrophic structural failures are rare. However, when significant structures such as pressure vessels, storage tanks, bridges or ships fail, the ramifications can be extensive, in terms of human injury, together with loss of capital revenue. In order to reduce the incidence of costly failures in the future, it is important to learn from past events. These lessons are incorporated into industrial and national codes of practice, and are valuable to practising engineers responsible for the management of critical structural plant.

Objectives
To describe the events relating to significant structural failures, together with the associated failure investigations. To summarise the lessons which can be drawn from these failures.

Approach
This report describes a series of fourteen industrial failures, where the mechanism of failure was by catastrophic fast fracture (for example, see Fig. 10 in main text). Failure case studies are presented for six pressure vessels, three ships, two bridges, two storage tanks and one offshore rig, in chronological order. For each case study, the failure events are described, together with an account of the main contributing factors and failure mechanisms. (Fracture mechanics analyses of the failures are not included in this report, see TWI Members' Report 512). The lessons which can be learnt from these failures are discussed.

Discussion
A range of issues have been identified which have contributed significantly to the structural failures described in this report: Fracture toughness is of particular importance for welded fabrications, where fracture toughness is dependent on chemical composition, microstructure, joint configuration, loading rate and temperature. Low fracture toughness was a factor which contributed to most of the failures which have been discussed in this report. All of the failures which have been described in this report were associated with welds. Weldments are associated with a higher risk of fracture due to the combination of complex metallurgy, welding residual stresses, stress concentrations and higher constraint associated with the joint configurations, together with the inherent flaws which are present in all welds.

Many of the failures occurred at relatively low temperatures (-20 to 13(C). Ferritic steels undergo a transition from ductile behaviour at higher temperatures to brittle behaviour at lower temperature. Many of the failures occurred during hydrotests. The environmental/service conditions to which critical structures are exposed are important. These include any factors which could lead to embrittlement of the component materials during its anticipated lifetime. The effective management of fracture control in critical structures implies an ongoing commitment to effective maintenance, inspection and quality assurance, with regard to issues relating to management of fracture risk.

Main conclusions
The lessons which are learnt from structural failures influence the industrial and national codes of practice for design, fabrication and operation of critical plant. Structural failures have also resulted in the development of 'fitness-for-purpose' assessment methods, such as BSI PD 6493:1991 'Guidance on methods for assessing the acceptability of flaws in fusion welded structures'. These methods are based on fracture mechanics principles, and allow the significance of weld flaws to be assessed in terms of structural integrity assessment. PD 6493-type methods are used extensively, on an international basis, for many applications, including pressure vessels, pipelines, storage tanks, ships, bridges, buildings and other structural components.

Recommendations
In order to reduce the incidence of costly failures in the future, it is important to learn from past events. These lessons need to be incorporated into industrial and national codes of practice, and are valuable to practising engineers responsible for the management of critical structural plant. Top of page

Catastrophic Failures of Steel Structures in Industry: Case Histories

By: B Hayes and R Phaal

1. Introduction
Fortunately, catastrophic structural failures are rare. However, when significant structures such as pressure vessels, pipelines, bridges or ships fail, the ramifications can be extensive, in terms of human injury, together with loss of capital and revenue. In order to reduce the incidence of costly failures in the future, it is important to learn from past events. These lessons are incorporated into industrial and national codes of practice, and are valuable to practising engineers responsible for the management of critical structural plant. This report describes a series of fourteen industrial failures, where the mechanism of failure was by catastrophic fast fracture. Failure case studies are presented for six pressure vessels, three ships, two bridges, two storage tanks and one offshore rig, in chronological order. For each case study, the failure events are described, together with an account of the main contributing factors and failure mechanisms. The lessons which can be learnt from these failures are discussed. For fracture mechanics analyses of some of the failures described in this report, the reader is referred to Part III of the TWI Members' Report 512/1995 by Challenger, Phaal and Garwood (see section 16 for full reference).

2. Hasselt Bridge
2.1 Summary Details
Failed structure: Date: Place: Conditions: Failure mode: Cause: Vierendeel truss welded steel bridge

14 March 1938 Hasselt, Belgium Light load but cold weather, about -20C Brittle fracture Low toughness steel subject to multiaxial restraint and residual stresses in the presence of weld defects Consequences: Complete collapse of bridge

2.2 Background
The Hasselt Bridge over the Albert Canal in Hasselt, Belgium collapsed at about 8.20am on the 14 March 1938 (see Fig. 1). Witnesses saw a crack open in a lower chord of the bridge (between the 3rd and 4th verticals) accompanied by a loud report. This failure of the lower chord transferred the load to the upper chord and six minutes later the bridge fell into the canal in three sections taking with it a tramcar and a number of pedestrians. All of the people on the bridge survived. The bridge was one of approximately 50 Vierendeel truss bridges built across the Albert Canal. The Fig. 1 Hasselt Bridge bridge consisted of straight lower chords, supporting failure the deck, and two parallel curved upper chords. The upper and lower chords were connected by vertical girders, the bridge having no diagonal members. The bridge, of span 75m, was designed to carry road and light railway traffic. It was erected between 1935 and 1936 and commissioned in January 1937. The structure was all-welded, the upper and lower chords being box girders made from welded plate I-beams. The girders of the upper chords were 1050mm wide and 1090mm deep with a maximum material thickness of 55mm. The lower chord girders were deeper (1290mm) and the maximum steel thickness was 45mm. The verticals were made from welded I-beams joined by a plate at the centre of the webs. The steel was a Belgium St-42 grade non-killed Bessemer steel with specified tensile strength between 365 and 435N/mm2.

2.3 Causes of Failure

The collapse was the first major failure of an all-welded structure and it was the subject of great interest and several investigations, not all of which reached the same conclusions. All the fractures in the bridge, both through parent plate and welds were brittle. Some of the fractures were thought to have occurred as a result of the impact of the fall. Analysis of the steel found it to be of expected chemical composition albeit with rather high sulphur and phosphorus contents. Tensile test results were acceptable but the notch toughness measured by Izod impact tests was low. Radiographic examination of the welds showed their quality to be poor. Cracks in the roots of important welds were discovered and the phosphorus level in the weld metal was found to be high.

Fig. Hasselt bridge (fracture 2 initiation)

The fractures in the lower chord, which was the first part of the structure to fail, were associated with connections to verticals (see Fig. 2). Fracture initiation is thought to have occurred in the weld joining the flanges of a vertical to those of the lower chord girder. Residual stress levels measured in similar undamaged joints were very significant. These, combined with the high stress concentrations at the weld resulting from poor joint design and welding procedures, the inherent rigidity of the bridge, the low material toughness at the ambient temperature and the presence of defects, led to failure of the lower chord and subsequent collapse of the bridge.

In January 1940, two other Vierdendeel truss bridges over the Albert Canal failed but did not collapse. In the first at Herenthals-Oolen, failure initiated in the lower flange of the lower chord at the joint to a wind brace. At the time the bridge was unloaded and the ambient temperature was -14C. The bridge was subsequently blown up during the German invasion of Belgium. The second bridge was at Kaulille and was built of rolled I-beams rather than plate girders. Cracking occurred in the lower chord and was not associated with any welds. This factor pointed to the importance of using steels with adequate notch toughness regardless of the possibility of welding defects or not.

2.4 Lessons learnt

As the first major failure of an all-welded structure, the collapse of the Hasselt Bridge sent shock waves around the engineering community. The main lesson learnt was the importance of joint design and welding procedure and quality. The probable influence of residual stresses was also recognised.

2.5 References
Author Bondy O Shank M E Busch H, Reuleke W Reeve L Title 'Collapse of an all-welded bridge at Hasselt, Belgium'. Engineering, 145, 17 June, 1938, pp.670-671, 682. 'A critical survey of brittle failure in carbon plate steel structures other than ships'. Welding Research Council Bulletin No.17, January 1954. 'Investigation of failure in a welded bridge'. Welding Journal, 25 (8), August 1946, pp.463s-465s 'Examination of welded steel specimens from the Hasselt Bridge'. Quarterly Trans Inst Welding, Vol.3, January 1940, pp.3-13.

Top of page

3. Schenectady T2 Tanker
3.1 Summary Details
Failed structure: World War II T2 type welded steel tanker Date: Place: Conditions: Failure mode: Cause: Consequences: 16 January 1943 Portland, Oregon, USA Still water at ~4C, air temperature ~-3C Brittle fracture Poor weld quality combined with stress concentration Temporary loss of ship

3.2 Background
At 11pm on the 16 January 1943, a few days after completing sea trials, the 152m long T2 tanker 'Schenectady' broke in two amidships while lying at the outfitting dock in the constructors yard in Portland, Oregon, USA (see Fig. 3). The temperature of the harbour water was about 4C and the conditions were still. The air temperature was approximately -3C and the winds were light.

Fig. Schenectady T2 tanker 3 failure The failure was sudden and accompanied by a report that was heard a mile away. The fracture extended through the deck, the sides of the hull, the longitudinal bulkheads and the bottom girders. The vessel jack-knifed, hinging on the bottom plate which had remained intact. The central part of the ship rose clear of the water so no flooding of the hull through the fracture occurred. The Schenectady was built by the Kaiser Company as part of the huge World War II emergency ship building programme. This programme produced 2580 Liberty ships, 414 Victory ships and 530 T2 tankers over the years 1941-1946. Although fractures in the emergency programme ships had been reported, the Schenectady was the first catastrophic failure, made all the more impressive by the still conditions under which it occurred. Then, in March 1943, a sister ship to the Schenectady' the 'Esso Manhattan', broke in two at the entrance to New York harbour in sea conditions described as very moderate.

3.3 Causes of Failure

The US Coast Guard, who were responsible for the safety of merchant vessels, requested the setting up of a Board of Investigation into the design and construction of welded ships. The Board was set up in April 1943 and co-ordinated a major research effort into the fracture of ships. The failure of the Schenectady initiated on the deck between two bulkheads. A defective weld was present in a region of stress concentration arising at a design detail. The nominal tensile stress in the deck was calculated to be 68N/mm2. Poor welding procedures were cited by the committee investigating the failure as contributory, however, at the time, the problems were not fully understood. The importance of weld quality was dramatically illustrated by the experience of the T2 tankers in which 50% of fractures initiated in welds not associated with design discontinuities. The investigation into the 'Schenectady' also questioned the adequacy of steel specifications for all welded ship hulls. The steel used to build the Schenectady was of a quality which was known to be acceptable for riveted ships. The final Report of the Board of Investigation was published in 1946. It considered 4694 welded steel merchant ships built in the emergency ship building programme, of which 970 sustained fractures. The report concluded that the fractures were due to the presence of notches in steels which were notch sensitive at the operating temperature and that the specifications current at the time were 'not sufficiently selective to exclude' such steels. Research into ship failures continued with the Charpy V notch properties of casualty ship plates being investigated. The absorbed energy in the Charpy V notch test, one of the few standardised fracture tests then available, was found to correlate well with the observed crack initiation, propagation and arrest behaviour of the ship steels. By the early 1950s the 15ft lb or 20J Charpy transition temperature was being used as a reference as it appeared to define the highest temperature at which brittle fracture initiation would occur in ship quality steels. However research showed that the critical temperature for brittle fracture initiation corresponded to higher Charpy energy values when modifications to alloying elements, grain size, deoxidation methods and normalising heat treatments were made. Hence the approach to brittle fracture avoidance could not be based on a simple fixed reference Charpy energy level.

3.4 Lessons learnt

The failure of the Schenectady and other war-time ships gave a significant impetus to the study of brittle fracture. These failures highlighted the influence of temperature on material toughness and the need to specify toughness requirements for welded ships. As for the Hasselt Bridge, the importance of joint design and weld quality was recognised.

3.5 References
Author Anon Title 'The design and method of construction of welded steel merchant vessels'. Final Report of a Board of Investigation ordered by the Secretary of the Navy, Government Printing Office, Washington DC, 1947. (Reprinted in part in Welding Journal, 26, No.7, July 1947).

Williams M L 'Investigation of structural failures of welded ships'. Welding Journal, and Ellinger October 1953, pp.498s-527s. GA Brown D P Hodgson J and Boyd G M Tipper C F Pellini W S 'Observations on experience with welded ships'. Welding Journal, September 1952, pp.765-782. 'Brittle fracture in welded ships - an empirical approach from recent experience'. Quarterly Trans INA, Vol. 100, No. 3, July 1958. 'The brittle fracture story'. Cambridge University Press, 1962 'Guidelines for fracture-reliable design of steel structures'. The Welding Institute, Abington, Cambridge, UK, 1983.

Top of page

4. Fawley Crude Oil Storage Tank

4.1 Summary Details
Failed structure: Date: Place: Conditions: Failure mode: Cause: 42m diameter welded steel oil-storage tank 12 February 1952 Fawley Hampshire Hydrotest with water temperature of +4C Brittle fracture Small defect associated with repair weld which probably produced strain ageing embrittlement in surrounding material

Consequences: Loss of tank

4.2 Background On 12 February 1952 a large all-welded oilstorage tank collapsed during hydrotest at the Esso Petroleum plant at Fawley in Hampshire (see Fig. 4). Hydrotesting had commenced on 30 January following completion of the tank, but was halted when a 0.6m long vertical crack appeared in the bottom two strakes. The tank was emptied and the crack repaired. When the hydrotest was recommenced on the Fig. Fawley crude oil storage tank 11 February, the air temperature was near 4 failure freezing and the water temperature +4C. The tank split when the water reached 90% of the tank height, a continuous vertical fracture running through the parent plate of every strake. The shell was torn from the tank bottom and collapsed on the surrounding band, leaving the roof lying on the base. The cylindrical tank was 42m in diameter and 16m high. The bottom was conical with a 0.6m fall at the centre and roof was a detached fully floating pontoon. The tank shell consisted of nine strakes made from butt welded plates measuring 1.8m x 7.2m. The strakes were progressively thinner from bottom to top, being aligned to produce a flush internal surface. The bottom strake was 28mm thick and the top 6mm. The construction of the tank was according to API Code 12C. The material used was a BS 13 steel with specified tensile strength in the range 430 to 510MPa, equivalent to ASTM A7 or A283 steel. Plate edge preparation for welding was carried out prior to rolling the plates to the required radius. The shell welds were full penetration double or single V welds, depending on the plate thickness. No-preheating was used except to dry the plates or remove frost. Boat

shaped samples were cut from the welds of the lower courses for inspection, leaving grooves that were repair welded. All boat samples, bar one, were satisfactory.

4.3 Causes of Failure

The 0.6m long crack which occurred during initial hydrotesting originated from a repaired boat-sample site. The brittle crack which caused the collapse of the tank also initiated at a repaired boat sample position in the circumferential weld between the lower two strakes. A very small cavity had been left at the bottom of the boat sample groove when it was repaired. This defect was found to be much smaller than others detected in the shell welds after the failure. The weld quality was in fact quite variable although this had not been revealed by the inspection during fabrication. Tests on the plate material showed it to meet the specification. Its Charpy impact transition temperature, however, was in the approximate range 0C to 15C, hence the tank material did not have good toughness at the hydrotest temperature. The existence of defects which were significantly longer than the one from which the fracture initiated perturbed the investigators. As no evidence of shock or impact loading which could have triggered the collapse was found, the investigation into the failure did not reach a conclusion regarding the cause of fracture initiation. Approximately one month after the failure of the crude oil tank, a neighbouring gas oil tank failed during hydrotest. This tank split vertically but remained in one piece. The tank was 45.7m in diameter and 14.6m high built, like the crude oil tank, of BS 13 steel to API 12C. The water temperature was +4C and the air temperature +9C at the time of failure. Examination of the fracture faces revealed that the failure initiated at a partially repaired crack in a vertical weld in the bottom shell course. The surfaces of the crack were blackened indicating that the crack had gone through a heating cycle due to a nearby welding operation. Subsequent studies indicated that the probable cause of failure was the presence of very low toughness material in the region of the initiating defects. These regions of low toughness would have resulted from dynamic strain-ageing embrittlement at the tip of the flaws during repair welding (or subsequent heat cycling). This type of strain ageing embrittlement, which is intensified at crack tips, is a potential problem associated with repair welds, particularly in coarse grained non-aluminium treated steels.

4.4 Lessons learnt

These failures raised concern over the weld inspection method specified in the API Code which relied on taking boat samples from the welds. In the case of the crude oil tank, the failure initiated from a poorly repaired boat sample site and in the case of the gas oil tank a significant defect was missed by the inspection method. These concerns led towards the use of radiography for weld inspection in storage tanks.

The failures also highlighted the importance of material toughness for storage tanks, and the introduction of the use of materials with minimum Charpy V properties greatly improved the safety of these structures.

4.5 References
Author Anon Feely F J and Northup M S Anon Title 'Failure of a large welded oil-storage tank', British Welding Journal, June 1955, pp.254-263. 'Why storage tanks fail'. Oil and Gas Journal, Vol.1, February 1954, pp.73-77. 'Learing from experience. Industry reviews fracture avoidance practices for large tanks: Part 1 and Part 2'. Metal Construction, December 1987 and January 1988. 'Locally intensified strain ageing of C and C:Mn steels and weld metals'. OMAE 95 Conference, Copenhagen, June 1995.

Dawes M.G and Francis-Scrutton N Top of page

5. World Concord Tanker

5.1 Summary Details
Failed structure: Date: Place: Conditions: Failure mode: Cause: 199m long all-welded steel oil tanker 27 November 1954 ~30km southwest of the Smalls in St George's Channel Gale force winds (8-9 on Beaufort scale), 10m wave Brittle fracture Combination of low toughness material, moderately high stress levels and superimposed shock loading

Consequences: Partial loss of ship

5.2 Background
The World Concord left the Mersey on the afternoon of the 26 November 1954 and set off southward down the Irish Channel. The vessel was in 'Winter Ballast Departure Condition', as recommended by the builders, carrying about 18000 tonnes water ballast. In the evening, warnings of severe south-westerly gales were received. The master took on more ballast and reduced the engine speed. By midnight the wind force was 8-9 on the Beaufort scale and waves were about 10m high. The engine speed was reduced further. In the early hours of 27 November, two very large waves hit the ship. The master estimated that the crest of the first wave was under the centre of the ship when the second wave broke over the fore of the ship. There was a loud rumbling noise and the vessel broke in half (see Fig. 5). Both parts remained afloat, although they did collide. No casualties were reported. Fig. World Concord 5 tanker failure The World Concord was built by Vickers Armstrong Ltd in Barrow-in-Furness. It was a single deck, all-welded, single screw steam turbine oil tanker, 199m long with a registered tonnage of 11700 tonnes. At the time of its failure, it was the largest tanker in the world. The vessel framing consisted of longitudinals approximately 0.8m apart and transverse webs positioned about 3m apart. Transverse bulkheads were located every 12m along the vessel's length. There were ten cargo tanks each with starboard, centre and port compartments. The steel plating for the hull and deck was 20 to 31mm thick. The bilge strakes, deck stringers and shear strakes were of 'special quality' steel meeting the requirements of Section 7 of the Rules for Quality and Testing Materials used by Lloyd's Register in

1950 with the extra condition of 0.23% maximum C content. The remaining plates were of 'ordinary shipbuilding quality' by early 1950s standards. The World Concord was repaired and operated under another name until 1974.

5.3 Causes of Failure

The break in the vessel occurred approximately amidships at the position of a transverse bulkhead. The T-shaped bottom longitudinal stiffeners were joined to the transverse bulkheads by welded vertical brackets. Also, as the longitudinals were scalloped, filling-in flat bars were welded between the bottom plates and the longitudinals at the bulkhead connections. This resulted in two regions of stress concentration at the bulkhead position, one at the vertical bracket and one at the filling-in plate. It was thought that the fracture initiated at one of these two stress concentration regions associated with a bottom longitudinal near the starboard side. The fracture propagated in a brittle manner in two directions along the line of the transverse bulkhead. In one direction it ran across the bottom of the ship, up the port side and back across the deck to the starboard side and in the other direction along the bottom of the vessel and up the starboard side. The final separation of the two halves of the tanker took place at the starboard deck to hull angle where the sheer strake and deck stringer plate were severely distorted. The investigation into the failure revealed that the Charpy V notch properties of much of the hull plate were poor at the casualty temperature (~12C) with absorbed energies less than 27J. (Most of the 'special quality' plates had Charpy energies above 40J at 12C). Furthermore the ballast condition of the ship, which was considered to be correct at the time, did in fact result in high stress levels in the bottom plating. Although the quality of welding was good with no significant defects, the combination of low toughness material, high stress levels and shock loading due to wave action resulted in conditions that the structure could not withstand.

5.4 Lessons learnt

This failure, and others, led to the changes in classification society rules regarding notch toughness of steels used in welded ships, the design of structural details and control of welding quality. In 1957 Lloyds Register incorporated mandatory minimum Charpy requirements into its Rules. These requirements were for at least 47J absorbed energy and 30% fibrous fracture appearance at 0C. Maximum permitted levels of carbon, silicon, manganese, phosphorous and sulphur were also introduced to ensure weldability.

5.5 References
Author McCallum J Top of page Title 'A Case Background - The World Concord'. RINA Spring Meetings 1981.

6. The Kings Bridge

6.1 Summary Details
Failed structure: Date: Place: Conditions: Failure mode: Cause: Consequences: Welded plate girder bridge with concrete deck 10th July 1962 Melbourne, Australia Passage of 48 tonne vehicle on a winter's day Brittle fracture from pre-existing defects Poor design and improper welding procedures for the restricted weldability steel used in construction Temporary closure of the bridge

6.2 Background
On the morning of the 10th July 1962, after 15 months in service, the steel girders of a span near one end of the Kings Bridge in Melbourne, Australia failed under the weight of a passing 48 tonne vehicle and dropped by 30cm (see Figures 6 and 7). Complete collapse of the span was prevented by the concrete deck catching on the vertical wall slabs under the bridge. The bridge was closed to traffic pending investigations and repair.

Fig. 6 - Kings bridge failure Fig. 7 - Kings bridge failure (detail)

The Kings Bridge, which has three sections, was built between 1957 and 1961. The failure occurred in the 1km long high-level section which was constructed of alternate cantilever and suspended spans. The failed span had four suspended steel plate girders approximately 30m in length which were supported by cantilevers extending from the adjacent spans. Each girder was made up of two flange plates welded to a web plate.

Vertical stiffeners were welded to the web at regular intervals. To increase the thickness in the region of maximum stress, cover plates had been welded to the lower flanges of the girders over their central portions.

6.3 Causes of Failure

The 365N/mm2 yield strength steel (to BS 968:1941) used in construction was known to have restricted weldability and the specification for the bridge required precautions to be taken in welding. The investigation into the failure found that the combination of improper welding procedures, inadequate preheat and the failure to dry the lowhydrogen electrodes properly in addition to local stress elevation due to design, led to the formation of hydrogen cracks at the toes of the transverse welds at the end of the cover plates in over 50% of the girders fabricated for the bridge. In the span which failed, the hydrogen cracks had penetrated the flange in three of the girders. During the winter of 1961 one of the inner girders failed but this had remained undetected. The crack in the second inner girder extended by fatigue along the web until, in the winter of 1962, the conditions of crack length, load and temperature for rapid brittle fracture were met. The outer two girders of the span failed at the same time in a brittle manner. The Royal Commission into the failure of the Kings Bridge found that the inexperience of the fabricator in welding low-alloy steel and the highly variable quality of the steel used were major factors in the failure. The steel supplied was not adequately tested and the toe cracks were not detected by either the fabricator or the Country Roads Board inspectors. The lack of clear and precise specifications for manufacture, testing and inspection was contributory as was the failure to fully investigate the proposed type of steel before construction was undertaken.

6.4 Lessons learnt

A full knowledge of the weldability of the steel to be used in construction is required. This allows appropriate welding procedures to be specified. The need for proper supervision and inspection during fabrication must not be underestimated.

6.5 References
Author Title

Burren W H et al 'Report of the Committee of Investigation on the failure of Kings Bridge'. Welding Fabrication and Design, Vol.6, No.2, October 1962. Anon Madison R B and Irwin G R Top of page Report of Royal Commission into the Failure of King's Bridge, Victoria, 1963. 'Fracture analysis of Kings Bridge, Melbourne'. Journal of the Structural Division, ASCE, Vol.97, No.ST9, September 1971.

7. Sizewell Boiler
7.1 Summary Details
Date: Place: Conditions: Cause: 31 May 1963 Sizewell Nuclear Power Station Hydrotest at 13C Probably shock loading of vessel

Failure mode: Brittle fracture Consequences: Financial loss

7.2 Background
On Friday 31 May 1963, boiler No.2A at the Sizewell Nuclear Power Station failed during hydrostatic testing (see Fig. 8). The full test pressure of 31 bar had been almost reached, the test pressure being set at 1.5 times the design pressure. The water temperature at the time of failure was approximately 13C. Fig. 8 Sizewell boiler failure The failure affected two strakes near one end of the cylindrical portion of the boiler. The cracking in these strakes was fairly extensive and such that a section of the vessel wall about 5.5m long was pushed out by the escaping water and left attached by a relatively small ligament. Witnesses reported hearing a loud bang and seeing the failed end of the boiler lift off the test frame by about 5cm before falling back. The overall length of the boiler was 28.6m and the internal diameter of the central cylindrical shell was 6.85m. It was made from a Mn-Cr-Mo-V steel, 57mm thick in the shell section. The plates for the shell were supplied in the normalised condition and, after welding, they underwent a post-weld heat treatment at 600C. The specified minimum tensile strength of the steel was 525N/mm2 and the Charpy V notch requirement was for 27J absorbed energy at -10C.

7.3 Causes of Failure

Examination of the fracture faces (see Fig. 9) found that the crack propagation along the lower fractured edge of the section of material pushed out from the vessel wall was typically brittle. Three fracture origins were identified on the lower fractured edge, each associated with thermal sleeve attachment points. The fracture appearance of the upper fractured edge indicated that here, the final failure had been due to bending.

The thermal sleeves were set-through pipes of approximately 100mm outside diameter. These were fillet welded to the boiler shell on the inner and outer surfaces. Investigation of the thermal sleeves and attachment welds in the regions of fracture initiation revealed no significant defects. Non-destructive testing of all other thermal sleeves in the boiler indicated that these were also free of defects. Fig. Sizewell boiler failure 9 (fracture face) The investigators into the failure were satisfied that the material and fabrication procedures met the required standards and so the stress conditions in the vessel were investigated. Strain gauge measurements were made on another of the four boilers for the No.2 reactor. These indicated that the maximum stress in the region of the failure was about 300N/mm2. Such a stress level was too low to account for the failure in the absence of significant defects. The metallographic examination of the material, both in the region of the failure and distant from it, showed the presence of twinning bands in considerable numbers. This was interpreted as evidence of shock loading of the vessel. The vessel had been supported during hydrotesting on eight piers made of steel I beams. Curved chocks were fitted to the top of each pier to support the vessel and the position of the chocks was adjusted via jacking screws. After the failure, chocks from the three piers at the fractured end of the boiler were found to be dislodged. Witnesses had seen the chocks from the two piers closest to the end of the vessel fall off when the boiler lifted and fell back during the failure. It was postulated that the third dislodged chock, which had been located at the centre of the failed section, had been insufficiently tightened and was pushed out under the weight of the water-filled vessel. The sudden collapse of support effectively caused shock loading of the boiler as the loads were redistributed. The material, although defect free and of acceptable quality, could not withstand these shock loading conditions. This was considered to be the only explanation of the failure which fitted all of the observations.

7.4 Lessons learnt

Unforeseen loadings can have disastrous consequences on otherwise acceptable structures.

7.5 References
Author Title West of Scotland Iron and Steel Institute: 'Special report on the failure of a boiler during hydrostatic test at Sizewell Nuclear Power Station', 1964. Smith N and Hamilton I G 'Failures in heavy pressure vessels during manufacture and hydraulic testing'. West of Scotland Iron and Steel Institute Journal, Vol.76, 1968-69.

Top of page

8. John Thompson Pressure Vessel

8.1 Summary Details
Failed component: Date: Place: Conditions: Failure mode: Cause: Consequences: 18.2m long Cr-Mo-V steel pressure vessel, 150mm wall thickness 22nd December 1965 Wolverhampton Hydraulic proof test at 10C Brittle fracture initiating from pre-existing defects Poor toughness and residual stresses due to inadequate heat treatment One minor casualty; financial loss

8.2 Background
In December 1965 a large pressure vessel being manufactured by John Thompson (Wolverhampton) Limited for the ICI Immingham plant fractured during a hydraulic test (see Fig. 10). Damage to the vessel was extensive with four large pieces being thrown from the vessel. One of these, weighing approximately 2 tonnes, went Fig. John Thompson pressure through the workshop wall and landed some 46m 10 vessel failure away. One minor casualty was reported. The vessel consisted of a plain shell of 1.7m inside diameter made from cylindrical strakes of 150mm thick silicon killed Mn-Cr-Mo-V steel. The end closures were forged, one end being a flange with a flat cover. The forging material was similar to the shell steel, however, a higher carbon content of 0.2% was used in the flange forging in order to meet the strength requirements in the thickest section. The forgings were supplied in a normalised and tempered condition. The overall length of the vessel was 18.2m and it weighed 167 tonnes. It had been destined for use as an ammonia converter with a design pressure of 35N/mm2 at 120C. The proof test requirement was for 48N/mm2 gauge pressure at ambient temperature (not less than 7C) but the testing of the vessel was troubled by leaks from the bolted flange joint and several re-pressurisations were required. At the first attainment of 34N/mm2 pressure, the vessel failed accompanied by 'a kind of dull thud'. No one present noticed anything unusual before the failure. The ambient and water temperatures at the time were determined to be less than 10C. The failure occurred at the flange end of the vessel. The flange forging was cracked through in two locations, the first two shell strakes broke into several pieces and cracking extended into the third strake.

8.3 Causes of Failure

Investigation revealed two fracture initiation sites. These were small pre-existing cracks in the heat affected zone (HAZ) of the submerged arc weld joining the flange end forging to the vessel shell. The cracks were located on the forging side of the weld about 15mm below the outer surface in regions of segregation where the carbon and alloying element contents were locally increased. This segregation would have increased the susceptibility of the material to hydrogen cracking, the probable cause of the original crack formation. The welding procedure was such that the preheat was discontinued immediately on completion of the weld, thus not allowing the reduction of hydrogen levels in the faster cooling surface regions. The failure occurred by the extension of these pre-existing cracks into the adjacent weld metal which had poor toughness properties due to inadequate heat treatment. The toughness of the forging and shell plate, although meeting the requirements, was not sufficient to arrest a running crack of size equal to the weld cross section. The forged flange and first strake sub-assembly had been furnace post-weld heat treated. The specified conditions were 620-660C for six hours, however high hardnesses measured on the casualty material indicated that the temperature of the sub-assembly had not reached this level. The furnace temperature had been monitored by pyrometers lowered from the roof. From subsequent temperature measurements made on similar components in the furnace, it was estimated that the actual temperatures achieved in the circumferential weld between the forging and the first strake were between 520-610C, depending on the position around the weld. The Charpy V notch requirements for the forging, plate and weld metal were 38J absorbed energy at +20C. The weld metal did not meet these requirements and the absorbed energies measured at +7C were in the range 12-25J which was considered the lower shelf for this material. Re-heat treating the casualty material at 650C for six hours considerably improved the Charpy properties of the weld metal but only at temperatures of 20C and above. Residual stresses were also considered a contributory factor to crack initiation at the relatively low applied stress level as the heat treatment conditions had not been sufficient for full relief of the residual stresses.

Lessons learnt
The British Welding Research Association report of the investigation into this failure proposed that fracture mechanics principles be used when setting fracture avoidance criteria for thick high strength steels. It also recommended carrying out pressure tests at temperatures above the ductile-brittle transition temperature of the vessel material in order to reduce the risk of failure.

8.5 References
Author Title

Anon

'Brittle fracture of a thick walled pressure vessel'. BWRA Bulletin, Vol.7, No.6, June 1966.

Top of page

9. Cockenzie Power Station Boiler Drum

9.1 Summary Details
Failed component: Date: Place: Conditions: Failure mode: Cause: Consequences: 22.8m long, 140mm thick Mn-Cr-Mo-V steel power station boiler drum 6 May 1966 Cockenzie Power Station, Scotland Near full hydraulic test pressure (270 bar) at 7C Brittle fracture Presence of significant defect, exact origin of which is unknown Financial penalties

9.2 Background

On the 6 May 1966, a boiler drum manufactured by Babcock and Wilcox Ltd failed during hydrotest at the Cockenzie Power Station in Scotland (see Fig. 11). The boiler was at approximately 270 bar gauge pressure, about 96% of the full test pressure, when the failure occurred. Fig. Cockenzie power station The ambient temperature was 7C. One end of the 22.8m long boiler drum split, with two 11 boiler drum failure longitudinal brittle fractures extending about 4.9m. The connecting pipework prevented the pieces of failed plate from being thrown any distance and no one was injured. The boiler drum was made of six cylindrical courses of 1.7m inside diameter, fabricated in 140mm thick Mn-Cr-Mo-V steel (Ducol W30). The drumheads were pressed from the same steel but slightly thicker (152mm). Production of the vessel involved manufacturing two halves consisting of three courses plus drumhead each. The two halves were furnace post-weld heat treated before being joined together. The closing circumferential weld was then locally post-weld heat treated. Each course had four set-through nozzle attachments which were welded by the manual metal arc process. These welds were tested internally and externally using magnetic particle inspection. One weld was found to contain a crack and, as the associated nozzle was also found to be deeply seamed, the whole nozzle was replaced. An angle bracket attachment close to the nozzle was cut down to a stub to improve access during welding of the replacement nozzle and subsequently reattached. The repair was carried out prior to the furnace post-weld heat treatment (PWHT). The boiler drum was inspected during its fabrication by both the manufacturer and independent insurance company surveyors, but no non-destructive testing was

performed after the final heat treatment. The manufacturers carried out an hydraulic pressure test at 11/2 times the design pressure in June 1964 prior to delivery to the power station in October 1964. The first on-site pressure tests were undertaken in March 1966 once the boiler erection was sufficiently advanced to allow it. Faulty welds in the superheater area remote from the drum were found and repaired. The boiler was fully pressurised on 1 April and passed by the surveyors except for a number of tube defects. The boiler was pressurised again on the 22 April to check that the repairs, which were not associated with the drum, were satisfactory. The failure of the drum occurred during the final pressure test on the 6 May.

9.3 Causes of Failure

The origin of the fracture was found to be a 330mm long surface breaking arrested brittle crack with a maximum depth of approximately 90mm (see Fig. 12). The crack had originated in the internal weld of the replacement nozzle and extended into the plate between the nozzle and the adjacent angle bracket. However, no defect was found to explain the initiation of this brittle crack. The surfaces of the arrested brittle crack were Fig. Cockenzie power station 12 boiler drum (fracture face) blackened and it was concluded that the crack had formed during the initial stage of the final PWHT, as a crack of such dimensions would have been detected during prior inspection. The parent material in the region of the failure had room temperature Charpy V notch energies in the range 60-80J meeting the requirements of BS 1113-1958. Although it was questioned whether this level of Charpy toughness was in fact sufficient, the material had withstood several pressurisations in the presence of a significant crack. There was no evidence of any crack extension during these earlier pressurisations. The argument was made at the time of the investigation into the failure, that if the pressure test had been conducted at a higher temperature, the failure may not have occurred and the vessel would have entered service in a severely cracked condition. The investigation into this failure also highlighted the importance of design details. Two main negative effects could be associated with the proximity of the angle bracket to the nozzle: the presence of complex residual stress and stress concentration patterns, and the difficulty of inspection of such a configuration.

9.4 Lessons learnt

The finding that failure was initiated from a crack formed during PWHT was instrumental in modification of standards to include mandatory non-destructive testing after heat treatment. It was also recommended that the heating rates at the start of stress relieving should be carefully controlled taking into account the shape and thickness of the vessel although there was no evidence of excessive heating rates in this case.

9.5 References
Author Smith N and Hamilton I G Anon Title 'Failure in heavy pressure vessels during manufacture and hydraulic testing'. West of Scotland Iron and Steel Institute Journal, Vol.7, 1968-69. 'Report on the brittle fracture of a high-pressure boiler drum at Cockenzie Power Station'. Welding Research Abroad, Vol.XIII, No.8, October 1967.

Top of page

10. Typpi Oy Ammonia Plant Water Coolers

10.1 Summary Details
Failed: Date: Place: Conditions: Set of four high pressure forged and welded water coolers 19 March 1970 Oulu, Finland Normal process conditions (230 bar pressure) with ambient temperature -3C Hydrogen embrittlement of the material under proces

Failure mode: Brittle fracture Causes: Consequences: Loss of vessels; temporary shut-down of plant

10.2 Background
A set of four high pressure heat exchangers at the Typpi Oy ammonia plant in Oulu, Finland failed on the 19 March 1970 (see Fig. 13). The plant had been restarted after a two-week shut-down and had been running at the process pressure of 230 bar (58% of the hydrotest pressure) for about an hour when the head chambers of the water coolers fractured suddenly. Pieces of two of the head chambers (A and B) were thrown up to 250m and the third chamber (C) showed extensive cracking around the nozzles. The fourth cooler (D) appeared undamaged. Personnel working near the heat exchangers did not hear Fig. Typpi Oy ammonia or see any signs of leakage prior to the failure and the 13 plant water cooler records showed the process conditions to be normal. The inlet temperature of the effluent was +10C and its outlet failure temperature +3C. The ambient temperature was 3C below zero. The head chambers were forged slightly oversized from a creep resisting Ni-Mo-V steel and normalised at 920C. Following rough machining to close to the final dimensions (outside diameter 1090mm, length 1100mm with thickness 85-150mm), the forgings were heat treated at 950C for four hours, oil quenched, tempered at 675C and finally air cooled. Following visual and ultrasonic inspection, the final machining was carried out. A mild steel overlay was deposited on the tube-plate face using manual metal arc (MMA) welding. The last weld metal to be deposited was around the circumference of the tube-plate, where it joined the chamber barrel. Following welding of other attachments to the chamber (except the tubes), the chamber was given a post-weld heat treatment at 560-580C. The weld overlay around the circumference of the tubeplate was then skimmed and the whole area inspected with a dye penetrant prior to the

attachment of the tubes. Final inspection by ultrasonics and dye penetrant testing was carried out before leak testing and hydrostatic testing of the assembled water cooler.

10.3 Causes of Failure

Investigation of the failure concentrated on head chambers A and B which fractured completely around their circumferences, in a brittle manner (see Fig. 14). The origin of the fracture in chamber A was in the heat affected zone (HAZ) of a nozzle attachment weld but was not associated with a defect. In chamber B, the fracture started from a small oxidised crack in the toe of the weld overlay around the circumference of the tube-plate. From consideration of the deformation of the plant and positions of the broken pieces, it was determined that chamber B failed first. Fig. Typpi Oy ammonia Tests on the material of the forged chamber showed that 14 plant water cooler the chemical composition was within specification. The results of Charpy V-notch impact tests were, however, (fracture face) much lower than those shown on the certification test records (average of 12J at 0C compared to 80-180J at the same temperature). This was determined to be due to the much slower cooling rate from the hardening temperature (950C) of the massive head chamber compared to the test material ring used for the original Charpy tests. Metallographic studies of the chamber steel revealed an upper bainite microstructure instead of the desired tempered martensite. Upper bainite microstructures typically exhibit good strength but poor ductility. The nil-ductility transition (NDT) temperature of the head chamber material (as measured by the Pellini drop weight test) was +20C, confirming its low toughness. Regions of high hardness were found in weld HAZs which had not been appreciably softened by the post-weld heattreatment (PWHT). The PWHT was also shown to have been insufficient to cause stress relaxation. The evidence indicated that the small original defect was formed during PWHT due to an excessive heating rate for a steel of this composition. Then during the year between the hydrotest and the commission of the water coolers, the defect extended by a stress-corrosion mechanism in the presence of high hardnesses and residual stresses. Finally, while in operation, the already low toughness material was embrittled by hydrogen from the process environment so creating the critical conditions for brittle fracture.

10.4 Lessons learnt

This failure illustrated the importance of acceptance tests being made on material typical of the structure and of correct post-weld heat treatment conditions being specified. It also demonstrated how the benefit of a hydrotest at a higher pressure than the working pressure could be removed by crack extension in service.

10.5 References
Author Moisio T Title 'Brittle fracture in a failed ammonia plant'. Metal Construction and British Welding Journal, January 1972, pp.3-10.

Top of page

11. Robert Jenkins Pressure Vessel

11.1 Summary Details
Failed component: Date: Place: Conditions: Failure mode: Cause: Consequences: 6 November 1970 Rotherham Hydraulic proof test at 12C Brittle fracture A liquation crack extended by hydrogen cracking to a critical size Loss of vessel; considerable damage to workshop due to release of water

11.2 Background
A large pressure vessel manufactured for export by Robert Jenkins and Co Ltd in Rotherham failed during pressure testing at the company's works on the 6 November 1970 (see Fig. 15). At the time of failure, the water temperature was approximately 12C and the gauge pressure was 29 bar, 85% of the required value. The vessel had been under pressurisation for 7 1/2 hours and contained 171,000 litres of water which were released into the workshop causing considerable damage. The vessel was 34m long in two sections which were joined by a flanged and bolted connection. One section was ~4.5m in diameter with a wall thickness of 41mm. The second Fig. Robert Jenkins section, in which the failure occurred, was 2.7m in diameter 15 Pressure vessel with a wall thickness of 28mm. The vessel was constructed failure according to ASME Section VIII. The larger diameter section was post-weld heat treated, while the failed section was in the as-welded condition. The failure initiated in the vicinity of the fillet weld of a manhole compensating plate on the main shell. The fracture, which was brittle, extended in two directions through the strake containing the manhole into the adjacent strake on one side and the dished end on the other before arresting. The overall length of the fracture was 4.3m with a subsidiary crack of 1.5m long present on the other side of the manhole. The vessel material was a C-Mn ASTM 515 Grade 70 (1967) steel for intermediate and high temperature use produced in the UK to coarse grained practice and supplied in the as-rolled condition. The yield strength of the material was in the region of 300N/mm2 and the Charpy properties of the strake in which the fracture initiated were

poor: 8-20J longitudinal and 11-14J transverse at 10C with fracture appearances of at least 90% crystallinity.

11.3 Causes of Failure

Examination of the fracture faces found evidence of an initial crack 115mm long at the fillet weld toe of the compensating plate on the external surface of the shell. This defect extended approximately halfway through the plate thickness. Fracture mechanics analyses carried out as part of the failure investigation showed that the applied stress intensity factor at the defect was close to the measured values of fracture toughness obtained at 12C. The origin of the initiating defect was thought to be hydrogen cracking in the HAZ of the fillet weld. Regions of martensite with hardnesses in the range 500-515 VPN were found near the toe of the fillet weld. A hardness survey of the area of the toe of the weld measured hardnesses from 393 to 496 VPN. (In comparison, the hardness of the parent plate was 190 VPN). With a carbon content of 0.33%, the plate was considered to be liable to hydrogen cracking in the high hardness regions. Although the manufacturer's recommendations for drying the electrodes were followed (150C for 1 hour), it was suggested by the investigators that this was insufficient given the hardness levels present in the HAZ. Tests on the failed plate showed the chemical composition in the initiation region to be susceptible to liquation cracking. It was postulated that the initiating defect may have formed as a liquation crack which extended by hydrogen cracking under the conditions of high restraint and residual stresses present at the fillet weld toe. An investigation was made into the effects of post-weld heat treatment on the welds in the as-welded failed section of the pressure vessel. It was found that little improvement in toughness properties was obtained in the range of temperatures associated with hydrotesting although the peak welding residual stresses were reduced which would be beneficial. This failure arose from a material problem: the toughness of the parent plate was low and it was susceptible to liquation cracking and formation of high hardness zones in the HAZ, leading to probable hydrogen cracking. The selection of a fine grained steel with superior toughness properties meeting specified levels would give greater protection against this type of failure during hydrotest or during service should the temperature be reduced below normal operating temperature.

11.4 Lessons learnt

The effect of temperature on material toughness is critical. Material selection should take into account all possible service temperatures, including pressure testing temperatures. The possibility of hydrogen cracking needs to be considered when establishing welding procedures.

11.5 References
Author Banks B Title 'Pressure vessel failure during hydrotest'. Welding and Metal Fabrication, January 1973.

Top of page

12. M V Kurdistan Tanker

12.1 Summary Details
Failed structure: Date: Place: Conditions: Failure mode: Cause: All welded tanker built to construction category Ice Class I 15 March 1979 South of the Cabot Strait, off Nova Scotia Moderately high seas, air temperature near 0C, cargo temperature ~60C Brittle fracture Presence of defect in bilge keel welds combined with high thermal stresses

Consequences: Loss of ship by intentional sinking

12.2 Background
On the morning of 15 March 1979 the motor tanker M V Kurdistan left Point Tupper in Nova Scotia bound for Sept-Isles, Quebec. The tanker was carrying a heated cargo of oil for the first time. The weather conditions were not good and the ship was rolling heavily. At about 12.30 the Kurdistan came to the edge of an ice field but, after travelling 2.5km into the ice, the ship was brought to a halt. The ship was turned around and headed back towards the open sea. At 13.50 the Kurdistan cleared the edge of the ice belt and put full ahead. Almost immediately there was a thud and a shudder during a downward pitch of the vessel. (The sea conditions were described as 'very heavy swell'). Oil started to escape from a vertical crack in the sides of No.3 wing tanks. The crack came up to about 3.6m below the main deck level.

Fig. 16

M V Kurdistan tanker failure

To reduce the loss, the transfer of oil from No.3 wing tanks to the No.4 tanks was undertaken `while the ship continued on its course. At 18.40 a second shudder was felt and the transfer of oil was stopped. The weather conditions had improved and the wave height was 2m. At 21.30 the ship broke in two: a shudder was felt and the bow rose, hinging about the deck at the No.3 cargo tanks before finally separating from the stern. Almost eight hours had elapsed between the initial fracture of the vessel's shell and its breaking in

two (see Fig. 16). The Kurdistan was built to construction class 'Ice Class I' and completed in 1973. The vessel was longitudinally framed except for the sides where the framing was transverse. With six cargo tanks, each divided into two wing tanks and a centre tank,

the overall length of the ship was approximately 182m. The Kurdistan was built almost entirely in Grade A steel (no Charpy requirements). The bottom shell was 19.5mm thick and the bilge strake 14.7mm. The bilge keel over a length of ship including the region failure consisted of 125 x 11mm ground flat bars butt welded end to end and overlapped on the underside by 300 x 13mm bulb plates, attached by intermittent welding. The bilge keel was connected edge- on to the bilge strake by continuous fillet welds above and below. The design of the keel called for a 25mm crack arrest hole to be drilled in each butt weld joining the ground bars.

12.3 Causes of Failure

Examination of the fracture faces revealed that the initial fracture through the bottom and side shell plates was brittle. The origin of the crack was a defective butt weld in the port bilge keel (see Fig. 17). There was lack of penetration in the butt weld and, where the bulb plate overlapped the underside of the ground bar, there was no weld at all. The bulb plate was misaligned and the crack arrest hole was Fig. M V Kurdistan failure (fracture missing. This region of the bilge 17 initiation) keel had been damaged in 1975 and repaired in 1977. Areas of fatigue crack growth along the lack of penetration at the weld root were found. The inquiry into the failure of the Kurdistan did not establish precisely the sequence of failure of the ship's longitudinal structure, which showed both brittle and ductile fracture. Given that the ship's shell plates were found to have 27J Charpy transition temperatures of between 5 and 20C, the steel in contact with the sea water was close to or below its transition and that in contact with the heated cargo was above. The displacement of oil by water entering the cargo tanks lowered the steel temperature to below its ductile/brittle transition. Calculations of the thermal stresses in the ship resulting from the carriage of a warm cargo in a cold sea indicated that a high tensile stress level would have been present in the shell and bilge keel. It is thought that the stresses due to the impact of a wave on the bow, superimposed on the high thermal stress and the stresses due to the moderate wave bending moments, triggered the fracture of the Kurdistan's bilge keel. The toughness of the shell plate was insufficient to arrest the propagating crack and complete failure ensued. The initiation of the fracture was due to the classic combination of poor weld metal toughness and high stresses in the presence of a defect.

12.4 Lessons learnt

This failure showed two important failings of the requirements for ships of the size of the Kurdistan built as First Year Ice Class vessels. Firstly that the ship could be built entirely of Class A steel with no notch impact requirements and, secondly, that no calculation of thermal stresses was required for cargoes at temperatures below 65C. Furthermore, this failure showed how critical the quality of workmanship can be even for a detail of apparently little significance such as the bilge keel.

12.5 References
Author Corlett E CB Title 'Kurdistan - The Anatomy of a Marine Disaster'. The Royal Institute of Navel Architects, Spring Meetings 1987. 'The use of yielding fracture mechanics in post failure analysis'. ASME Pressure Vessel and Pipeline Technology, 1985, p.20. Top of page

13. Alexander L Kielland Accommodation Platform

13.1 Summary Details
Failed structure: Date: Place: Conditions: Failure mode: Cause: Pentagone type semi-submersible rig 27 March 1980 Ekofisk field, North Sea Bad weather, ~60-75km/hr wind speeds, ~6-8mm wave height Fatigue failure followed by brittle fracture in one brace and ductile overload in remaining adjacent braces Fatigue crack growth from a weld defect

Consequences: Loss of 123 lives and platform

13.2 Background
On the evening of 27 March 1980, one of the five columns of the 'Alexander L Kielland' accommodation platform anchored in the North Sea, broke off (see Fig. 18). (The five columns were the principal buoyancy elements of the platform). The platform immediately heeled over to an angle of 30-35 and then continued to heel and sink Fig. Alexander L Kielland accommodation slowly. Twenty minutes after 18 platform the loss of column D, the platform capsized. Of the seven lifeboats on board, only two were launched successfully albeit with great difficulty in part due to bad weather conditions (one landed upside down in the water). Some inflatable rafts launched themselves due to the listing of the platform. A massive international air and sea rescue operation was undertaken. Of the 212 men on board the platform when it failed, 123 died. The Alexander L Kielland was a semi-submersible mobile rig of the Pentagone type, a design which had been developed in France. The rig was built between 1973 and 1976 in France for an American operator. Although it was designed as a drilling rig, it was only ever operated as an accommodation platform during its four years in service. The platform had five columns, of overall height of 35.6m, mounted on 22m diameter pontoons. The columns were positioned at the apexes of a pentagon with braces running between adjacent columns and the deck or hull. Accommodation units and a drilling tower were mounted on the deck.

13.3 Causes of Failure

Following the accident, the platform and the separated column D remained afloat. Column D was towed to Stavanger and divers removed all the fracture faces from the capsized platform for investigation (see Fig. 19). The Commission responsible for the inquiry into the disaster concluded that the structural failure had occurred in the following stages:

Fig. Alexander L Kielland 19 accommodation platform (fracture face)

ii. iii.

i. Fatigue crack growth in brace D6 initiating from pre-existing cracks in the fillet welds between a hydrophone support and the brace Final, mainly ductile, fracture of brace D6 Subsequent failure of five remaining braces joining the column to the structure by plastic collapse

Brace D6 and the hydrophone support were both made from a C-Mn structural steel (equivalent to a Lloyds' ship steel Grade EH) with a minimum specified yield strength of 355N/mm2. The brace was 2.6m in diameter with a wall thickness of 26mm. The hydrophone support was 20mm thick with a diameter of 325mm and was set-through the brace. It was attached to the brace by two fillet welds, one on the outside of the brace and the other on the inside. Examination of these fillet welds revealed poor penetration into the hydrophone tube material and an unsatisfactory weld bead shape. Significant cracking was also found which was dated to the time of fabrication by the presence of paint on the fracture surfaces. Fatigue crack growth in brace D6 originated at the hydrophone support weld and extended, in the latter stages partly by ductile tearing, around approximately 2/3 of the circumference of the brace until final failure took place by brittle fracture. The chemical compositions of the brace and hydrophone material were within specification, as were the Charpy and in-plane tensile properties. The throughthickness ductility of the hydrophone material (which was not specified) was, however, poor. This, combined with its through-thickness tensile strength being lower than the in-plane strength of the brace material and with sub-standard welding, led to partial cracking of the fillet weld during fabrication.

13.4 Lessons learnt

Although material properties and welding quality played a significant part in this disaster, rig design was also a critical factor. Apart from the stability and buoyancy aspects which were inadequate, the design did not consider attachments to highly stressed braces such as D6 as important. The fatigue performance of the hydrophone attachment and its effect on the fatigue life of the brace were tragically overlooked.

13.5 References
Author Norges Offentlige Utredninger Moan T Title 'Alexander L Keilland' - ulykken'. Report NOU 1981:11, 1981. 'The progressive structural failure of the Alexander L Kielland platform'. Case Histories in Offshore Engineering, ed, G. Mauer, Springer-Verlag, 1985.

Almar-Naess A, Haagensen 'Investigation of the Alexander L Kielland failure P J, Lian B, Moan T and metallurgical and fracture analysis'. Proc. 14th Annual Simonsen T Offshore Technology Conference, Vol.2, 1982, p.79 Top of page

14. Union Oil Amine Absorber Tower

14.1 Summary Details
Failed structure: Date: Place: Conditions: Failure mode: Cause: 16.8m high monoethanolamine (MEA) absorber vessel 23 July 1984 Lemont, Illinois, USA Normal operating conditions: 1.4N/mm2 internal pressure, 38C Very low energy ductile failure (very little cleavage present) Hydrogen induced heat affected zone (HAZ) cracks combined with hydrogen embrittled material

Consequences: Loss of 17 lives; over $100 million of property damage

14.2 Background
On the evening of Monday 23 July 1984, the Union Oil Co refinery near Lemont, Illinois, USA was seriously damaged by an explosion and fire. Seventeen people working at the refinery were killed and the property damage was estimated to be over $100 million (see Fig. 20). The explosion was caused by the ignition of a large cloud of flammable gas (a mixture of propane and butane) which had leaked from a ruptured amine-absorber pressure vessel.

Fig. Site of Union Oil amine 20 absorber tower failure

An operator working near the absorber tower noticed gas escaping from a horizontal crack about 150mm long near the bottom of the vessel and tried to close off the main inlet valve. The crack grew to 600mm and he initiated evacuation of the area. As the company fire fighters arrived, the absorber tower cracked further and a large amount of gas was released. The gas ignited in a massive explosion which sent the upper part of the tower into the air, landing over a kilometre away. The explosion was felt over 20 kilometres away and the blaze which followed sent flames 150m into the sky.

The absorber tower first went into service in 1970. It was a cylindrical vessel 2.6m in diameter and of overall height 16.8m (see Fig. 21). The shell section consisted of six courses of 25mm thick ASTM A516 Grade 70 steel. These were joined by full penetration submerged arc welds in the as-welded condition. The vessel, built to ASME Section VIII, was designed to strip H2S from the propane/butane gas mixture passing through it. Monoethanolamine (MEA) was fed through the tower as part of this process. The operating conditions were 1.4N/mm2 internal pressure at 38C. The environment in the tower was corrosive. Soon after the amine absorber tower entered service, hydrogen blisters were found in the lower two courses of the shell and laminations were detected in the steel. The growth of hydrogen blisters continued and in 1974 the second course of the tower was replaced on site using manual metal arc Fig. Union Oil amine welding with no preheat or post-weld heat-treatment 21 absorber tower (PWHT). In 1976 a Monel liner to reduce corrosion was fitted in the bottom head and first course of the tower but it did not cover the repair section.

14.3 Causes of Failure

The investigation into the failure found that the tower fractured at the circumferential weld between the replacement ring and the lower course. Four large cracks in the heat affected zone (HAZ) had been present prior to the failure, originating at the inner surface of the tower and extending almost through the wall thickness. About 35% of the vessel circumference was affected. The location of the first leak observed corresponded to one of these HAZ cracks which was approximately 800mm long. Microhardnesses measured in the HAZ near the surface exceeded 29 HRC and peak hardnesses of 40 to 48 HRC were found near the fusion line. These facts, taken with the in- section appearance of the pre-existing cracks (straight in the HAZ near the surface and then zig-zagging through the base material at the limit of the HAZ), pointed to the cracks initiating by hydrogen cracking and then progressing by hydrogen-induced stepwise cracking (HISC). Tests according to a NACE standard procedure confirmed that the material was susceptible to HISC. The fracture ran around the HAZ of the circumferential weld at right angles to the axial stress of 35N/mm2. The fact that this stress level was so low and the crack did not change directions to run in a direction perpendicular to the higher hoop stress, indicated very low toughness material in the HAZ. Charpy V notch tests of the replacement course material and the weld between the replacement course and the upper part of the tower showed the weld metal and HAZ to have superior notch toughness to the base material. (20J transition temperatures: 0C for parent plate, 51C for weld metal, -40C for HAZ). Fracture toughness tests measuring crack tip opening displacement (CTOD) in the HAZ material gave much greater critical CTOD values than the applied CTOD in the tower at the time of failure, estimated ignoring any residual stresses, as 0.064mm. Tests on hydrogen charged specimens did,

however, reveal much reduced CTOD fracture toughness values in the range of approximately 0.070-0.080mm at 38C. A later fracture mechanics assessment of the tower found that when residual stresses were taken into account, failure was predicted at the level of CTOD measured in non-hydrogen charged specimens. Taking all of these findings into account, it can be concluded that this failure occurred because the welding procedure used when replacing a section of the vessel caused the formation of a hard microstructure in the HAZ of the weld. This hard region was susceptible to hydrogen assisted cracking resulting in growth of large cracks in the vessel. The uncracked material in the vicinity of the existing cracks had low toughness due to hydrogen embrittlement and failed at the applied CTOD in the vessel arising from the operating pressure and residual stresses associated with the weld.

14.4 Lessons learnt

For operation in corrosive conditions, the control of weld properties is critical. Welding procedures, particularly for field repair welds, need to be formulated to avoid the formation of high hardness microstructures for service in hydrogen environments. The significant contribution of welding residual stresses to the applied CTOD at a flaw present in a structure must not be overlooked.

14.5 References
Author McHenry H I, Read D T and Shives T R Anon Title 'Failure analysis of an amine-absorber pressure vessel'. Materials Performance, Vol.26, No.8, August 1987, pp.1824. 'Weld failure in pressure vessel generates revealing report'. Welding Journal, 1 April 1984, pp.57-60.

Top of page

15. Ashland Storage Tank

15.1 Summary Details
Failed structure: Date: Place: Conditions: Failure mode: Cause: 16000m3 capacity fuel oil storage tank 2nd January 1988 Floreffe, Pennsylvania, USA Filling of tank to near capacity with oil at 8C; ambient temperature of -3C Brittle fracture Very low toughness region (strain aged embrittled) associated with flaw

Consequences: Major environmental pollution for which Ashland Petroleum Co. assumed full financial responsibility; loss of tank

15.2 Background
On the 2 January 1988, tank No. 1338 at the Ashland Petroleum Company's Floreffe terminal in Pennsylvania was being filled to capacity with diesel fuel oil for the first time since its re-erection at this site the previous August. The temperature of the oil was 8C and the air temperature was -3C. At 5.00pm, when the oil level was almost at the operating maximum, the tank shell fractured vertically without warning. The tank shell parted from the bottom plate at the connecting welds and, under the force of the escaping oil, moved sideways about 35m. The tank roof to shell joint remained sufficiently intact for the roof to move with the shell. The escaping oil flowed over the surrounding dykes damaging an adjacent tank and passed through storm sewers into the Monongahela River and then the Ohio River. The total spillage was estimated at 15.2 million litres, causing severe harm to the environment and affecting the drinking water supply. The tank had been built originally at Whiskey Island in Ohio some time in the 1930s1940s. It was a 36m diameter cylindrical tank with a flat bottom and supported conical roof. The shell was approximately 14.4m high and consisted of six courses of welded plate, each plate being about 2.4m x 9.6m. The plate thickness in the bottom course was 21mm and 6mm in the top course. The thicknesses of courses 2 to 5 lay in between these. The tank capacity was 16000m3 or 16 million litres and until 1986 it had been used to hold distillate oils and heavier distillates. In 1986 the tank was taken down by oxyacetylene cutting adjacent to the original welds and then reassembled by welding in Floreffe, keeping the plates in the same order.

15.3 Causes of Failure

Examination of the fracture faces in the tank shell showed them to be flat and perpendicular to the plate surfaces, with the characteristic chevron markings of brittle fracture (see Fig. 22). The chevron markings pointed back to a flaw below the weld between the 1st and 2nd courses, at the point where a vertical weld on the 2nd course met the circumferential weld. The flaw was described as being 'dime-size' or about the size of a 5 pence piece and its orientation was in the vertical direction of the tank. Metallographic studies of the flaw revealed it to be due to flame cutting, rather Fig. Ashland storage tank than welding but, surprisingly, not the flame cutting 22 (fracture face) of the dismantling procedure. The flaw had been present in the steel plate prior to being welded when the tank was originally built. Charpy V notch tests and drop-weight tests (Pellini) to measure the nil-ductility transition (NDT) temperature were performed on the shell plate. The parent material was an ASTM A10 steel, either rimmed or semi-killed. The NDT temperature was found to be +10C and at +3C, the estimated temperature of the tank wall at failure, the Charpy tests showed low energy absorption. However, engineering defect assessments using fracture toughness values measured at +3C indicated that the stress due to the hydrostatic pressure alone (approximately 80N/mm2) would not have been sufficient to trigger failure. Soil foundation analyses were carried out which ruled out subsidence as a contributory factor. Attention was then turned to the influence that the weld adjacent to the flaw may have had. Welding residual stresses may be as high as yield strength level. In the case of the Ashland tank this could have meant that the flaw was subject to a stress level of approximately 240N/mm2. Furthermore, the effect of the welding heat cycle on the material at the crack tip was thought to have caused locally intensified strain-ageing embrittlement to which steels of this type and vintage are susceptible. This was confirmed when low fracture toughness values were measured on shell plate samples simulating this form of embrittlement. It was finally concluded that the failure was due to the material immediately surrounding the flaw being of particularly low toughness, with crack initiation occurring under the combined effect of hydrostatic and residual stresses. As the tank was operating below the NDT temperature of the shell plate, the crack emerging from the locally embrittled area could not be arrested.

15.4 Lessons learnt

This failure highlighted the potentially serious problem of locally intensified strainageing embrittlement associated with re-welding and weld repairs of older steels. In these situations, the material toughness can be less than expected

15.5 References

Author Mesloh R E, Marschall C W, Buchheit R D and Kiefner J F Top of page

Title 'Battelle determines cause of Ashland tank failure'. Oil and Gas Journal, September 1988, pp.49-54.

16. Summary and Conclusions

A series of fourteen industrial failure case studies have been presented, including pressure vessels, ships, bridges, storage tanks and an offshore rig. For each case study, the failure events have been described, together with an account of the main contributing factors and failure mechanisms, and the consequence of the failure. A range of issues have been identified which have significantly contributed to the structural failures described in this report: i. Material properties The strength of a component is dictated by the geometry of the structure, stresses within the structure, and the properties of the materials which comprise the structure. Important material properties include the yield and tensile strength of the material, together with fracture toughness. Fracture toughness is of particular importance for welded fabrications, were fracture toughness is dependent on microstructure, joint configuration, loading rate and temperature. Low fracture toughness was a factor which contributed to most of the failures which have been discussed in this report. Careful consideration should be given to any factors which might reduce the fracture toughness of the materials in critical structures, such as reductions in temperature or strain ageing embrittlement (see Fawley and Ashland storage tank failures, for instance). Welds All of the failures which have been described in this report were associated with defective welds. For fracture to occur, there must be a detrimental combination of stresses, flaws and fracture toughness. Weldments are associated with a higher risk of fracture due to the combination of complex metallurgy, welding residual stresses and stress concentrations and higher constraint associated with the joint configuration, together with the inherent flaws which are present in all welds. To minimise the risk of fracture in critical structural components, special consideration must be given to welding during design, fabrication and operation. Additional consideration should be given to repair welds, were control of the welding process and post-weld heat treatment may be difficult. Attachment welds are also important, as running fractures can progress into the main structure (for example, the Alexander L Kieland platform and M V Kurdistan tanker). iii. Temperature

ii.

Many of the failures occurred at relatively low temperatures (-20 to 13C), including the Hasselt and Kings bridges, Fawley and Ashland storage tanks, Schenectady and M V Kurdistan tankers, Typpi Oy ammonia water coolers, together with the Sizewell, John Thompson, Cockenzie and Robert Jenkins pressure vessels. Ferritic steels undergo a transition from ductile behaviour at higher temperatures to brittle behaviour at lower temperature. The decrease in fracture toughness associate with this transition can be significant, over a fairly small temperature range. The absolute value of this fracture toughness transition temperature varies, depending on the steel composition, joint geometry, environment and loading rate. For these reasons, Charpy impact energy requirements in fabrication codes are typically specified at specific temperatures relative to the service temperature, to ensure that the material toughness is sufficient. iv. Proof testing Many of the failures occurred during hydrotest, including the Fawley storage tank and the Sizewell, John Thompson, Cockenzie and Robert Jenkins pressure vessels. It could be maintained that these hydrotests were successful, in that they prevented potential catastrophic failure in service. However, hydrotest failures are expensive, and appropriate lessons must be learnt to ensure procedures are modified to reduce this risk (i.e. restrictions on minimum temperature for testing, maximum pressurisation rates, improved inspection, etc., to facilitate repair prior to hydrotest). Nevertheless, as the Typpi Oy failure demonstrated, the benefit of a proof test may be removed by stress corrosion and other forms of crack extension in service. v. Environment/service conditions The environmental/service conditions to which critical structures are exposed is important. This includes any factors which could lead to embrittlement of the component during its anticipated lifetime (for example, the Union Oil amine absorber tower and Typpi Oy ammonia plant water coolers, were hydrogen embrittlement was a major contributing factor, and the Sizewell pressure vessel and World Concord ship, where dynamic loading was considered to be a factor. vi. Maintenance/inspection/quality assurance The effective management of fracture control in critical structures implies an on-going commitment for effective maintenance, with regard to issues relating to management of fracture risk (i.e. identification of critical components and joints, regular inspection, etc.). Poor maintenance was identified as a major contributing factor to the amine absorber pressure vessel failure. Particular attention should be given to these issues if a change of service conditions of life extension is planned. Table 1 indicates which of these six factors played a role in each of the failure cases described in this report.

The lessons which are learnt from structural failures influence the industrial and national codes of practice for design, fabrication and operation of critical plant. For instance, the amine absorber tower failure resulted in widespread changes to industrial codes for plant maintenance, while the Cockenzie boiler drum failure resulted in the requirement for inspection after post-weld heat treatment, and the specification of minimum temperatures for hydrotesting. The failure of the Fawley storage tank, and the Schenectady and World Concord tankers resulted in extensive research into fracture mechanics, leading to the requirements for adequate notch toughness in critical fabrications. Structural failures have also resulted in the development of 'fitness-for-purpose' assessment methods, such as BSI PD 6493:1991. These methods are based on fracture mechanics principles, and allow the significance of weld flaws to be assessed in terms of structural integrity assessment. PD 6493-type methods are used extensively, on an international basis, for many applications, including pressure vessels, pipelines, storage tanks, ships, bridges, buildings and other structural components.

16.1 References
Author Title BSI PD 6493:1991: 'Guidance on methods for assessing the acceptability of flaws in fusion welded structures', British Standards Institution, London, 1991. Challenger N V, "Appraisal of PD 6493:1991 fracture assessment procedures. Phaal R and Garwood Part III: assessment of actual failures", TWI Research Report SJ 512/1995. Table 1: Summary of factors contributing to failures Failure Case Hasselt Bridge Schenectady T2 Tanker Fawley Crude Oil Storage Tank(s) World Concord Tanker Kings Bridge Sizewell Boiler Factors Contributing to Failure i, ii, iii i, ii, iii i, ii, iii, iv i, ii, iii, v i, ii, iii, vi ii, iii, iv, v

John Thompson Pressure Vessel Cockenzie Power Station Boiler Drum Typpi Oy Ammonia Plant Water Coolers Robert Jenkins Pressure Vessel MV Kurdistan Tanker Alexander L Kielland Accommodation Platform Union Oil Amine Absorber Tower Ashland Storage Tank Top of page

i, ii, iii, iv ii, iv, vi i, ii, iii, v, vi i, ii, iii, iv, v i, ii, iii, v, vi ii, v, vi i, ii, v, vi i, ii, iii

17. Acknowledgements
The work described in this report was carried out within the TWI Core Research Programme, funded by the Industrial Members of TWI. The photograph in Fig. 22 is reproduced courtesy of Battelle Memorial Institute and Ashland Petroleum Company.

Link to related content, selected on the basis of keyword matches using the keyword set on the 'Advanced screen

TWI information:
Information and advice from TWI are provided in good faith and based, where appropriate, on the best engineering knowledge available at the time and incorporated into TWI's website in accordance with TWI's ISO 9001 accredited quality system. No warranty expressed or implied is given regarding the results or effects of applying information or advice obtained from the website, nor is any responsibility accepted for any consequential loss or damage.