Failure Analysis Case Studies II Episode 5 pps

128 k Fig. 14. Starboard bilge keel 8 Brittle fracture 9 Brittle fracture [l in. (25mm) in centre ductile] 10 Brittle fracture [into drainhole 1 in. (25 mm) ductile, then brittle] 11 Brittle fracture 12 Brittle fracture 13 Ductile Generally, even where full 45" shear fracture was evident, the reduction of area accompanying the fracture was small. An inspection of the port bilge keel remaining on the aft section of the ship revealed cracks in two weld details similar to the bilge keel initiation site on the port side. Extraction of these samples was requested. The heating coils used to keep the oil in the tanks at 140 OF (60 "C), are evident in many of the figures. It must be pointed out, however, that the initiation site in the port bilge keel would have been surrounded by sea-water, and, as such, could be considered to be at the water temperature of -0.7 "C. Two meetings were held between the interested parties on 16 and 17 April 1979, respectively. At the first it was agreed by all concerned that TWI should carry out any tests required to ascertain the reasons behind the failure. At the second meeting, the five test samples required for this investigation selected by the DOT and TWI representatives during their inspections were described to the interested parties, who agreed to the extraction of these samples and their shipment to the TWI. Details of the samples required were left with the ship owners, and a Lloyds' representative was appointed to supervise the extraction of the samples. The positions of the required test samples were then marked on the vessel by the DOT and TWI representatives. 3. DISCUSSION OF THE RESULTS OF THE FAILURE INVESTIGATION 3.1. Introduction Three of the samples identified during the dry dock inspection of the vessel described above contained initiation sites. At the start of the mechanical and metallurgical test programme described in [l], each of these samples were treated with equal importance. However, as the test programme developed, various aspects became clearer, and the port bilge keel sample was identified as the primary sample. To keep this section of the paper reasonably concise, rather than outline the gradual progression of the failure investigation, the sequence of events leading to the failure of the MV 129 Kurdistan, deduced from the investigations of [I] are presented in chronological order below, drawing on the test data where necessary to justify the conclusions reached. A summary of the mechanical test results is given in Table 1. 3.2. The initiation of fracture The port bilge keel detail extracted in the principal test sample was the primary initiation site. Crack initiation by a cleavage mechanism occurred from a defect situated in the ground bar butt weld contained in this sample. This defect was formed by incomplete penetration due, in part, to the lack of an edge preparation on the ground bar plate (Fig. 15). In addition, during assembly of this detail the bulb bar plate was almost certainly attached by intermittent welds (Fig. 16) prior to the completion of the ground bar weld. This procedure resulted in an area of no weld on the outer lower edge of the ground bar joint, which, together with the lack of penetration defects present in the double-sided section of the weld, formed the defect shown in Fig. 17. This defect constituted an effective stress raiser, which was extended in service by fatigue, as illustrated in Figs 18 and 19. Three separate areas along the length of the defect showed evidence of fatigue damage, as shown in Fig. 20. This fatigue crack growth had the separate effects of increasing the overall defect size and enhancing the acuity at the notch tip. The initiation toughness of the weld metal in this region was believed to be low at the sea temperature at the time of the incident (- 1 "C) at low strain rates, and very low under medium and high rates. This conclusion was reached from toughness tests carried out on the corresponding weld detail from the starboard side of the vessel (Table 1). Metallurgical examination showed that this detail had similar microstructure, composition and hardness to the ground bar weld metal in which the initiation site was located, and is thus considered to have very similar toughness properties. The weld detail of the ground bar weld of the starboard sample differed from the port sample in that it was double-sided over its entire length, although, due to the lack of edge preparation, it contained a comparable lack of penetration defect. No evidence of any fatigue crack propagation comparable to that found in the port sample was observed. Table 1. Summary of mechanical test results Plate tensile properties (primary sample) at + 20 "C uy = 243-258 N mm2 au =41748 Nmm2 Elongation = 32-34.5% Pellini NDT temperatures Plate No. NDT 1A IC 2A 3A >O"C -5°C +5"C 0°C Weld metal Charpy results (starboard bilge keel sample) Test temperature ("C) CV energy (J) Crystallinity (%) - 10 -1 +5 + 10 + 20 8 95 6, 10 95,85 14,30 60,60 28 70 5, 8, 48, 80 95,90, 50, 35 Weld metal CTOD results (BS 5762) at - 1 "C Test rate (mms-I) CTOD (mm) Type of result 0.01 1 .o 4.50 0.10 0.11 0.04 0.16 0.07 0.03 0.04 6U 6m 6C 6C 6c 6C 6C 130 Fig. 15. Ground bar butt weld in double-sided region. Note incomplete penetration. x 7 Fig. 16. Intermittent lap fillet weld in sample 1 between ground bar and bulb bar, seen from above Although the lack of penetration defect of the starboard sample must have seen comparable loading to that experienced by the detail which initiated in the port sample, this was not a critical defect. The presence of a complete double-sided weld was obviously sufficient to reduce the overall stress and local stress concentration effects experienced by this weld to levels below that at which any significant fatigue crack propagation had occurred. Because of the lower overall stress and local stress concentration levels, and the presence of a smaller and blunter defect in the starboard detail, fracture initiation from this site was much less likely. As noted in [l], if the ground bar weld detail 131 Fig. 17. Close-ups of port bilge keel in the port sample had been two-sided, stress levels approaching ultimate tensile strength of the plate material would have had to have been experienced under static conditions to cause failure. The defect that actually existed in the ground bar detail in the port sample was greater than the maximum defect size, which it is believed could have been tolerated safely at stresses likely to be experienced under normal operating conditions at low strain rates (R= 60 N mm-3'2 SKI). From visual inspection of the fracture surface, it was deduced that initiation occurred from the fatigue cracked areas situated in the ground bar weld metal (Fig. 19). The intermittent weld and ground bar to shell plate weld details in this area were not considered to be prime initiation sites since: (a) crack tip opening displacement (CTOD) tests have shown that these regions were likely to have had a significantly higher initiation toughness than the partial penetration butt weld detail; 132 Fig. 18. Fracture surface of ground bar butt weld in sample 1 in single-sided region, close to inboard edge of bulb bar. Note smooth, featureless region in root of weld x 4 Fig. 19. Fracture surface of ground bar butt weld in sample 1 in single-sided region, close to outboard edge of ground bar. Note smoother fracture surface in weld root, from which some chevron patterns (arrowed) emerge. x4. (b) evidence from chevron patterns showed that the crack ran through the bilge bar to shell plate (c) the notch acuity of the defect was lower in these areas due to the absence of a fatigue crack. The directions of propagation from the initiation sites in the port sample are believed to have been as shown in Fig. 20. weld into the shell bilge plate; 133 IVU' - 90. 80- 6 70- 60. 5 50- 30. 20. IO x al $ 40. 0- propagation i up port shell plate Shell plate (1C) Bilge bar (IA) WLld metal (6W) 27 J level - ' Bilge strake shell plate Bulb / bar Crack propagated from ground bar butt weld via ground bar to shell weld into bilge strake plate Brittle crack propagation running into bottom shell Fig. 20. Crack initiation in ground bar weld and subsequent propagation into shell-port bilge keel detail. 3.3. Propagation Once the crack had initiated in the ground bar weld, it was able to propagate into both the bulb bar and the shell, since the dynamic toughness of the shell to ground bar fillet weld, and the bulb bar to ground bar intermittent lap weld were insufficient to arrest a running crack at the sea-water temperature (- 1 "C), as evidenced by the Charpy data obtained on representative samples (Fig. 21). In addition, there was no crack arrest hole in the ground bar butt weld, the presence of which may have prevented the crack from propagating into the hull. Furthermore, the dynamic toughness of the Grade A plate used in the hull and the bulb bar was also inadequate to arrest a running crack, as discussed in [I]. Although the oil cargo was believed to be heated to around 60 "C, it is surmised that the shell plate below the water-line was close to the sea temperature of - 1 "C, since the results of the Pellini drop weight tests indicate that the nil ductility transition (NDT) temperature was around 0 "C (Table I). Had the shell plate been at a higher temperature, it is likely that shear lips would have been observed on the shell plate fracture surface, and no such evidence for ductile fracture was observed . Once the crack had entered the shell, it propagated in two directions: (a) Up the port side until it arrested at an indeterminate point at least 3m above the bilge keel. Due to extensive mechanical damage to the fracture surface, the precise point of arrest was not evident. (b) The crack also ran across the entire breadth of the bottom shell plate and up the starboard side 134 Secondary initiation associated with - 6 in. crack Region in which fracture face was destroyed by mechanical damage, and where major crack is thought to have arrested during first incident Secondary initiation associated with ,6 ft crack Similar region to that seen on starboard fracture face. Crack is thought to have arrested here during first incident Primary initiation (Fig. 20) W Fig. 22. Extent of crack propagation after the first reported incident. General view. with no visible interruption, until it arrested at least 3 m above the bilge keel, in another area which had suffered extensive mechanical damage. The path of the fracture initiating from the primary sample is shown in Fig. 20. As the crack propagated along the bottom shell, it entered the longitudinal girders and bulkheads to the degree shown in Figs 22-25. In several of these, especially the bulkheads, the crack was arrested, due to higher temperatures of the shell caused by the heated cargo, and to the thinner plate in the bulkhead material. As discussed in [l], it was believed that the initiation sites found in two of the other samples removed at St John could not have occurred except as a consequence of the extensive bottom and side shell fracture described above, and fracture would have initiated at the sites during, or immediately following, the major shell fracture. This conclusion is reached due to there being no weld defects present, the high toughness of the weld metal, and the fact that these sites would normally be After the first incident: Port longitudinals 8, I1 and 12 are thought to have fractured completely, No. 9 is likely to have been partially intact. The brittle fracture ran into a drain hole in No. IO and arrested. The longitudinal bulkhead and No. 13 are considered to remain intact. No. 13 in fact failed in a fully ductile manner in a vertical plane forward of the bottom shell fracture path. Brittle fracture appeared continuous along the length of the bottom shell fracture. level Fig. 23. Crack propagation-port bilge keel region during the first incident. 135 After the first incident: The centre girder is likely to have experienced 1-2 in. brittle fracture which arrested leaving the girder essentially intact. The starboard bulkhead experienced 3 in. brittle fracture which then arrested. Longitudinals 1-7 starboard of the centre girder are thought to have fractured. Longitudinals 1 and 2 (port) may have seen crack arrest in the drain holes. Port longitudinals 3,5,6 and 7 fractured. No. 4 remained intact. Brittle fracture was apparently continuous along the length of the bottom shell. brittle fracture Continuous brittle fracture Fig. 24. Crack propagation during the first incident-bottom shell fracture. Brittle crack propagation running into region where fracture face was destroyed by mechanical damage - crack is thought to have I 1 After the first incident: apart from a 3in. brittle crack extending from the bottom shell the longitudinal bulkhead remained intact. Longitudinals IO and I I are thought to have fractured but 8, 9, 12 and 13 probably experienced arrest of the brittle crack arrested before water- as it entered the drain holes and thus remained partially intact. Brittle fracture was apparently continuous along the bottom shell and through the starboard Continuous brittle fracture Brittle crack propagated through starboard bilge keel Fig. 25. Crack propagation during the first incident-starboard bilge keel region. situated in a low-stress region. Thus, it is considered that these were secondary, and not primary initiation sites. The extent of propagation from these cracks could not be determined, but cracks initiating from these sites were thought at the time of the investigation likely to correspond with the location of oil leaks witnessed by the crew. After the first incident, it would appear that the vessel was held intact by partially fractured longitudinal bulkheads and bottom longitudinals, the upper regions of the hull sides, and the deck plate and its associated longitudinals. The bottom shell plate crack could have arrested either due to a rise in temperature, and hence toughness of the steel, or to its running out of driving stress. 3.4. Events leading toJinal fracture of vessel Due to the extensive damage to the shell structure which existed after the initial incident, separation of the two sections of the vessel is considered to have been inevitable, and thus the events leading to the final fracture of the vessel must be considered to be of secondary importance. Due to 136 the extent of the bottom shell fracture, the remaining intact and partially fractured structures would have experienced higher than designed stresses. This situation would have been aggravated by the entry of sea-water into the shell, which would, of course, have cooled the internal members. These factors, together with the presence of sharp crack tips, resulted in the progressive failure of these members by both ductile and brittle fracture mechanisms. It was unclear whether the side shell fractures emanating from the port bilge keel connected with those from sites in the secondary samples, or whether the cracks from these latter two sites were bypassed by subsequent fracture events. This would seem to be most likely, due to the complex interaction of cracks observed in these regions. Final separation of the vessel occurred when the deck plates and their associated longitudinals failed. 3.5. Possible causes of fracture The fracture mechanics calculations performed using PD6493 (1 980) procedures as described in [l] (Table 2) showcd that the defect situated in the port bilge keel detail of the primary sample exceeded the tolerable defect size at - 1 "C for normal operating conditions. These calculations showed that the combination of (a) the position of the bilge keel defect under the still water bending moment loading; (b) the influence of the thermal stresses caused by carrying a hot cargo in cold waters; (c) the effect of high tensile residual stresses, and (d) the wave loading on exiting the ice field, would have subjected the bilge keel defect to a high applied crack opening displacement. Hence, there existed some risk of fracture from this defect under normal operation at a sea-water temperature of - 1 "C or lower. It was believed that, under rough sea conditions, the strain rate in the defect region could have been elevated, by wave loading, to levels exceeding K = IO3 N rnm3!'s-'. Under these circumstances, there would be a decrease in the toughness of the weld in which the defect was situated, which, when accompanied by relatively high local stresses, would have meant that the defect in the primary sample exceeded the critical defect size at - 1 "C, and thus fracture would have been highly probable. Table 1 shows the decrease in CTOD toughness with applied strain rate. The medium test rate was roughly equated to the likely loading rates under rough sea conditions to predict the critical flaw sizes shown in Tablc 2. It was not considered likely that high strain rates (k > 10SNmm-3'2s ') would have been experienced under normal operating conditions. There was evidence that cleavage fracture occurred directly from the tips of the fatigue cracks in the weld metal of the defect region in the port bilge keel (Fig. 19). As some ductile extension was experienced in a comparable weld metal at low strain rates (Su value in Table I), this, combined with the defect assessment calculations, led to the conclusion that local strain rates above the lowest rate employed in the toughness evaluations were experienced by the port bilge keel region at the time of the incident. 3.6. Report of the public enquiry and further discussion The results of the failure investigation reported in April 1980 were considered in the light of detailed reports from the crew, and evidence from a range of experts at the public enquiry held over 51 days in London during 1981. The report of the court was presented on 12 November 1981 ([2]). This report fully describes the circumstances leading to the casualty, and, with assembled evidence, was able to adjudicate on the most likely timing of events for the failure, which had been the major area of disagreement between the parties represented at the enquiry. The court found that the weight Table 2. Critical defect assessment-half length of critical through thickness flaw size [2]* Applied load (N mm-2) Test rate CTOD (mms-') (mm) 100 150 200 250 300 450 0.01 0.10 27.3 22.8 19.8 17.3 15.5 11.5 1 .o 0.04 10.8 9. I 7.8 7.0 6.3 4.8 450 0.03 8.3 6.8 6.0 5.3 4.5 3.5 *Equivalent defect size of actual flaw assessed as buried defect d = 8.6mm. 137 of evidence pointed to the brittle fracture initiation occurring as the vessel encountered head seas near the ice edge subsequent to the manoeuvring in the ice field, the failure initiation in the port bilge keel having been triggered by wave impact on the bow. As described in the paper by Corlett et al. [3], the period spent in the ice almost certainly led to the general cooling down of the longitudinals and the shell beneath the water-line, as the bunker oil solidified on the inner surface of the vessel in the calm conditions in the ice field. Without this general lowering of the temperature of the ship plate to its NDT (nil ductility temperature) value, the primary initiation induced in the port bilge keel would not have propagated with such disastrous consequences. As noted in the failure investigation [l], there was evidence of other bilge keel details which had cracked at some earlier occasion, but had not propagated into the ship’s structure. The fracture mechanics calculations performed in the failure investigation and referred to by the report of the court [2] were performed using the BSI PD6493 1980 procedures with an adjustment to remove the inherent factors of safety in the analysis to facilitate critical predictions as described in [4]. Subsequently, this analysis method was updated in 1991, and the Kurdistan casualty was re- examined as part of the validation exercise along with many large-scale laboratory tests and other well-documented failures [5]. The reanalysis is illustrated in Fig. 26 using the level 2 procedure of BSI PD6493: 1991 with weld metal CTOD values relating to the low and medium strain rates. Assessment points are drawn for stress inputs of 100, 150 and 200 Nmm-*, assuming a yield strength of 227 and 300 Nmm-2, relating to plate and weld metal, respectively. The reanalysis indicates that failure (indicated by points outside the failure locus) would, in fact, have been possible at low strain rates as the higher load level is approached. For the intermediate rate, CTOD toughness failure is predicted for the still- water condition of 100 N mm-2. This reanalysis confirms the criticality of the combination of high applied stresses (due to the combination of cargo loads, wave loads and hot cargo in cold seas), high residual stresses, the presence of a significant weld defect, and low toughness. The toughness of the weld was, of course, prejudiced by the incorrect weld procedure, but also 1.4 1.2 (a) - Unsafe - & 1.0 0.8 V -7 0.6 0.2 1 a I I 0 0.2 0.4 0.6 0.8 I .a I .2 - Safe - A CTOD = 0.04 mm, yield strength = 227 N mm-2 1 0 CTOD = 0.04 mm, yield strength = 300 N mrK2 A CTOD = 0.1 mm, yield strength = 227 N mm-* o CTOD = 0.1 mm, yield strength = 300 N mmS2 - - Sr 1.4 A 0 Unsafe (b) - 0 0.2 0.4 0.6 0.8 I .o Sr 2 Fig. 26. (a) MV Kurdistan level 2 assessments-mbedded defect assumption. (b) MV Kurdislan level 2 assessments surface defect assumption. [...]... Ganvood, S J and Harrison, J D., in Pressure Vessel and Piping Technology-l9 85 A Decade of Progress ASME, pp 1043-1 054 Challenger, N V., Phaal, R and Garwood, S J., Appraisal of PD 6493:1991 Fracture assessment procedures Part 111: assessment of actual failures TWI Research Members Report 51 2/19 95, June 19 95 Failure Analysis Case Studies If D.R.H Jones (Editor) 0 2001 Elsevier Science Ltd All rights reserved... maximum stress intensity factor is at point A (see Fig 5) and its value can be expressed as where P is the applied load and D is the bar diameter The coefficients C, are given in Table 4.A Table 4 Values of C, c u j=O 1 2 i=O i=2 i=3 1.118 1.4 05 3.891 8.328 -0.179 5. 902 -20.370 21.8 95 -0.339 -9. 057 23.217 - 36.992 i=4 3 0.130 3.032 -7 .55 5 12.676 m 152 Fig 5 Geometry and notation of the surface crack recent... [22'3 (5- t2)] 4 4 2 1 + 2(1 - 25) (5 - t 2 ) I l 2 +2- 'I3(< . (mms-') (mm) 100 150 200 250 300 450 0.01 0.10 27.3 22.8 19.8 17.3 15. 5 11 .5 1 .o 0.04 10.8 9. I 7.8 7.0 6.3 4.8 450 0.03 8.3 6.8 6.0 5. 3 4 .5 3 .5 *Equivalent defect. (J) Crystallinity (%) - 10 -1 +5 + 10 + 20 8 95 6, 10 95, 85 14,30 60,60 28 70 5, 8, 48, 80 95, 90, 50 , 35 Weld metal CTOD results (BS 57 62) at - 1 "C Test rate (mms-I). Meetings, Paper No. 8, 1987. 1043-1 054 . assessment of actual failures. TWI Research Members Report 51 2/19 95, June 19 95. Failure Analysis Case Studies If D.R.H. Jones (Editor) 0 2001