Stefano C. Chiovelli, David V. Darling, Alan G. Glover, DavidJ. Horsley
NOVA Corp. of Alberta
Calgary
Combined effects of a preexisting weld defect, settlement of adjacent branch piping, and less than critical line pressure caused the rupture and fire Jan. 8, 1992, on the Western Alberta main line.
The line is operated by NOVA Corp. of Alberta's Alberta Gas Transmission Division, Calgary.
The failure site is approximately 60 miles (100 km) north of Calgary.
The fire at the adjoining James River interchange meter station rendered inoperable the exchange facility between NOVA's eastern and western systems (Fig. 1).
The subsequent metallurgical investigation concluded that the rupture origin ed at a preexisting hydrogen crack located at the toe of a hot-tap stub weld on the 36-in. carrier pipe.
Brittle fracture propagation in both directions from the tee resulted in a rupture length of approximately 1,225 ft (420 m).
Crack propagation was consistent with properties of the 1960s vintage 36-in. pipe material which had met American Petroleum Institute (API) requirements at installation.
Fracture analysis showed that all three conditions - weld defect, piping settlement, and line pressure - were necessary for the failure to occur.
The failure pressure was in fact less than the maximum allowable operating pressure (MAOP).
INTERCHANGE
NOVA's Alberta Gas Transmission Division owns and operates a network of pipelines which form the main system for transporting natural gas in the province of Alberta.
It carries about 78% of an marketed Canadian gas each year.
Gas is delivered through more than 11,400 miles (18,400 km) of pipeline to points in the province and to its borders for export.
About 82% of the gas transported is for export and delivered through three major exit points: Empress, McNeill, and Alberta-B.C. (Fig. 1).
These three points deliver a total of 2.77 tcf/year (78.5 billion cu m/year; bcmy). The Alberta-B.C. exit point is fed by the Western Alberta main line and the Foothills Pipeline (Alta.) Ltd. system (operated by NOVA).
Empress and McNeill are fed in part by NOVA's Edson main line and the Foothills eastern leg.
These two main systems meet at the James River in change.
This interchange includes metering and facilities transfer gas from the Edson main line to the Western Alberta main line.
On Jan. 8,1992, a rupture occurred in the Western Alberta main line at the interchange. The rupture initiated at a hot tap tee, propagated by brittle fracture in two directions from the tee, and arrested in new pipe material.
The Western Alberta main line had been in service since 1962 and was constructed of 36-in. OD x 0.406-in. W.T. (914 mm x 10.31-mm W.T.) API 5LX Grade X-52 pipe material.
Mechanical testing confirmed that the 36-in. main line material met the specification requirements of the day, API 5LX Grade X-52.
The line's MAOP of 845 psi (5,826 kPa) gives a hoop stress of 71.9% specified minimum yield strength (SMYS).
A hot tap was installed in 1981 by NOVA welders with a standard NOVA procedure for welding branch connections to operating lines.
The full encirclement reinforcing saddle was specified as MSS Standard Practice 75 WPHY 60 Grade 414. Nominal specified dimensions of the saddle were in. long x 0.625 in. thick.
The 24-in. stub was specified as CSA Z245.2 Grade 414 with nominal dimensions of 24-in. OD x 0.5-in. W.T. Additional construction at the James River interchange meter station occurred during September 1991.
Part of this construction included installation of 490 ft of 30-in. piping onto the 24-in. branch piping from the hot-tap tee (Fig. 2).
INVESTIGATIONS
Examination of the site evidence indicated that the rupture of the 36-in. Western Alberta main line initiated at the hot tap tee fitting adjacent to valve APX-1, where a 24-in. pipeline lies into the main line.'
Examination of the fracture surfaces on the 36-in. pipe material in the trench revealed chevron markings and points of crack arrest at both sides of the tee. These indicated the fracture had propagated away from the tee.
Extensive brittle fracture was observed on the fracture faces of the pipe (Figs. 3 and 4).
The brittle fracture arrested at a tee assembly 360 ft north of the initiation site.
The southern-end crack arrest occurred in 1980s vintage pipe 1,015 ft south of the initiation site.
The total ruptured pipe length was approximately 1,225 ft.
Gas was ignited and continued to bum at the northern end crack-arrest site on the 36-in. main line, the rupture-initiation site (adjacent valve APX-l), at a 24 in. meter-run inlet, and the control-valve assemblies at the James River interchange meter station. The fires continued to burn for approximately 12 hr.
No evidence of sustained fire at the southern end crack-arrest point was observed.
METALLURGY
The tee consists of the assembled fitting of the 36-in. carrier pipe, the 24-in. stub, and a fall-encirclement reinforcing saddle (Fig. 5). The metallurgical investigation was to analyze all material evidence that could indicate the cause of the 36-in. main line rupture.
OBSERVATIONS
NOVA carefully disassembled the tee to identify more accurately the fracture-initiation point. The fracture surface at the origin of the rupture was heavily oxidized by the heat of the fire at valve APX-L.
Fortunately, the mating fracture surface was not oxidized as it was ejected from the tee and buried in the soil (Fig. 6).
Visual examination of the fracture surfaces on the ejected section of 36-in. carrier pipe revealed chevron markings, from two directions, pointing back to an area which had a distinctly different fracture appearance (Fig. 7).
These observations led to the conclusion that the 36-in. main line rupture originated in this area.
Two small areas at the fracture initiation area appeared flat and displayed no chevron markings. NOVA concluded that these two areas were preexisting defects with dimensions of approximately 2.60 in. and 2.68 in. long separated by 1.26 in. and roughly 0.08 in. deep at the deepest point.
The lack of deformation on all fracture surfaces indicated brittle fracture.
The pre-existing defect lay at the toe of the stub weld on the 36-in. carrier pipe. Viewed from the inside of the carrier pipe, this defect was located at approximately the 1:00 to 3:00 o'clock position.
The appearance of the cap pass of the stub weld indicated that one side, from 6 to 12 o'clock, was welded with a weave technique vertically up. On the side containing the pre-existing defect, however, a stringer-bead technique was used, welding having been vertically down from the 1 to 5 o'clock positions.
The stringer bead passes were made at the weld toes on the 36-in. pipe and 24-in. branch side.
This observation suggested that this area had not been welded according to the specified vertical-up welding procedure.
The sleeve was examined to determine the mode of failure on the attachment bar fillet welds. The full-encirclement reinforcing saddle failed through the fillet welds on both attachment bars.
Failure at the fillet welds was by tensile shear.
The quality of the weld was generally good. The hole in the pipe associated with the hot tap appeared to be concentric with the 24-in. stub. Deformation of the reinforcing saddle prevented accurate measurements of the fit of the saddle on the carrier pipe.
The length of the saddle was measured to be 84.25 in. and the thickness 0.843 in. These measured dimensions did not comply with the specified dimensions.
As will be discussed later, however, they did not contribute to the failure.
FRACTOGRAPHY
NOVA performed a fractographic examination of the 36-in. main line's fracture origin both optically and with a scanning electron microscope to determine the nature of the preexisting defect, mode of fracture from it, and any evidence of subcritical crack growth from the defect.
No fractographic examination could be performed on the failed fillet welds on the reinforcing saddle because of heat damage from the fire at A.PX-I.
Examination of the fracture surface associated with the origin of the rupture revealed vertical chevron marks which originated from the preexisting defect located on the outside surface.
The small shear lip on the interior surface of the pipe confirmed the direction of crack propagation. Pop-in features, which indicate low toughness and rapid, single-event fractures, appeared immediately below the preexisting defect.
No evidence of progressive crack growth on the through-wall fracture surface was apparent, and this fracture surface was relatively clean compared to the preexisting defect.
NOVA took several microsamples from the rupture's initiation site and examined them with the scanning electron microscope.
The fracture surface in Fig. 8 shows several distinct features. Regions A and B indicate the preexisting defect. Region C identifies the region of radial through-wall propagation.
Region A corresponds to weld metal in the area of the toe. Region B coincided with the heat-affected zone (HAZ) as shown in Fig. 9. The end of Region B corresponded with the end of the visible HAZ.
The region showing initiation of the through-wall fracture propagation from the preexisting defect can be seen in Fig. 10.
Quasi-cleavage fracture features were present on the surface of the preexisting defect with a small zone of plastic deformation (stretch zone) at the tip of the preexisting defect.
Cleavage and river patterns associated with brittle fracture appeared on the fracture surfaces from the stretch zone to the shear lip on the ID surface. No evidence of progressive crack growth appeared on the fracture surfaces either immediately before or immediately after the stretch zone.
The small plastic zone observed at the tip of the preexisting defect typically occurs upon loading and immediately before brittle fracture. Secondary cracks were associated with the through-wall fracture propagation, but no such features appeared on the preexisting defect. Based on this information, NOVA concluded that the through-wall fracture had initiated from the preexisting defect on the OD surface and propagated as a single event.
The maximum depth of the preexisting defect was measured to be 0.087 in.
MICROHARDNESS
NOVA performed metallography to determine the microstructure of the weld and pipe material and to locate the preexisting crack with respect to the weld.
Microhardness measurements were taken to check for excessive hardness around the preexisting defect. The microstructure of the pipe material was typical of a 0.25 wt % carbon, hot-rolled steel fabricated according to industry specification API 5L of early 1960s vintage (Fig, 11).
Results from a chemical analysis of the 36-in. main line material are shown in Table 1. The carbon equivalent of this material is 0.434.
The through-wall fracture had initiated from a defect which was partially in the HAZ and partially in weld metal very near the toe of the reinforced groove weld on the 36-in. carrier pipe (Figs. 9 and 12).
Microhardness surveys (Vickers microhardness 500 91 15 s) taken in the HAZ and weld metal adjacent the origin of failure revealed values in the range of 518-546 HV in the HAZ with corresponding hardnesses of 390-440 HV in the weld. Average hardness of surrounding parent metal was 210 HV.
Unfortunately, heat damage from the fire at APX-L prevented determining the extent of this high-hardness region around the stub at the toe of the weld.
It could only be concluded that the unaffected material associated with the vertically down stringer weld had excessively high hardness in the HAZ. The microstructure observed in the regions of the HAZ that was associated with the high hardness was lath martensite (Fig. 13).
The appearance of the. weld beads in profile show one stringer pass was made on the toes of the weld. The balance of the weld appears to have been welded with a weave technique.
Hardening curves were developed for this material to establish the heat input required to result in the observed microstructure and hardness. This testing determined that a low heat input in the range of 7.6 to 10. 2 kJ/in. (0.3-0.4 kJ/mm) would be required to result in the observed microstructure and hardness.
The sensitivity of the 1960s vintage pipe material to martensitic transformation with low heat inputs is characterized by an arc strike on the 36-in. carrier pipe near the weld disclosed by metallographic examination.
Hardness measurements taken in these areas varied from 400 to 610 HV. The region contained several small cracks.
FRACTURE ASSESSMENT
A lower bound Charpy V-notch energy of 13.3 ft-lb at 23 F. (18 J, -5 C.) was measured with an average Charpy V-notch energy of 24.4 ft-lb at 23 F.
In addition, from the drop-weight tear tests (DWTT) results, the nil ductility temperature was inferred from the temperature at which the percentage shear begins to increase from less than 5%.
The nil ductility temperature of the 1960s vintage pipe material was approximately 41-50 F. with an average of 3% shear at 23 F. These tests led to the conclusion that this material was not resistant to crack propagation.
The test results were consistent with the brittle fracture observed in the site investigation.
No toughness properties were required at the time this line pipe material was purchased. Current CSA Z245.1-M90 standard, however, requires that DWTT and Charpy V-notch tests exhibit a fracture appearance shear area of 60% minimum for any test with no individual test specimen exhibiting less than 50% shear area taken at 23 F.
Absorbed energy requirements for Charpy V-notch testing is specified as 30 ft-lb minimum. The lower bound cracked-tip opening displacement (CTOD) of the 36-in. main line base material was determined to be 0.00134 in. at 23 F.
This lower-bound value indicates this material had poor resistance to fracture initiation.
NOVA performed a fracture analysis using Appendix K of CSA Z184-M 86.2 The length of the defect used in this assessment was based on interaction rules between the defects given in the standard.
A graph of strain vs. defect length (Fig. 14) predicted that the strain required to initiate a failure from a defect 6.54 in. long and 0.087 in. deep for the measured CTOD of 0.00134 in. was approximately 0.11%.
Based on previous work on predicted vs. actual failure strain,3 NOVA applied an empirically derived factor of 2 to establish the critical strain to failure.
The strain required to propagate the observed preexisting defect was 0.22%. This implies that the local stress at the toe of the stub weld must have exceeded the yield stress of the material just before failure.
STRESS ANALYSES
To indicate the stress-strain state at the hot-tap location for the estimated loads at the time of failure, stress analyses were performed.
These used both two- dimensional shell analysis and full three dimensional finite-element analyses.
Internal pressure and external loads were the loading effects analyzed.
The internal pressure analysis considered the fit between the saddle and the carrier pipe, as well as the installation sequence. The external load analysis considered vertical settlement from new 1991 summer construction, thermal loads, and transient pressure surge.
Results from these stress analyses led to the following conclusions:
- The hot-tap installation practices would not cause failure.
- The most likely cause of the failure was settlement of adjacent piping.
- The amount of settlement required was sufficiently small that such settlement was likely to have occurred.
- The failure would not have occurred in the absence of the preexisting weld defect, under the loads identified.
RESULTS
The metallurgical investigation showed that the rupture of the 36-in. main line originated at a pre-existing defect located at the toe of the stub weld on the hot-tap adjacent to APX-L.
The metallography, hardness testing, and scanning-electron-microscope examination led to the conclusion that the preexisting defect was most probably a hydrogen crack which occurred during the installation of the stub.
The high carbon content of this material in conjunction with low heat input stringer-bead welds at the toe of the stub weld created a microstructure which was extremely susceptible to hydrogen cracking, even with the use of low hydrogen electrodes.
Chemical analysis of the stub weld's stringer beads was consistent with a low hydrogen E8018-C2 electrode. The visual examination confirmed that the stringer beads were welded vertically down.
Welding vertically down with a low hydrogen electrode requires a fast travel speed. The high travel speed and small stringer-bead size is consistent with a low heat input.
These observations indicate that the welding which was performed would yield a very high cooling rate which would be necessary to form the susceptible martensitic microstructure observed.
The fracture analysis concluded that the strain required to propagate the preexisting hydrogen crack must have been in the order of 0.22%. This indicates that the local stress at the toe of the stub weld must have been approaching the yield strength of the 36-in. material. Additionally, the 36-in. material had very poor resistance to fracture propagation. Once the fracture initiated, therefore, there would have been little or no resistance to fracture propagation. The high toughness and greater wall thickness of the material at the northern and southern crack-arrest sites are consistent with fracture arrest at these locations.
Charpy V-notch and DWTT results (CV 100 J, -5 C. and DVM 4,700 J, 0 C.) from the northern and southern crack-arrest sites clearly exceeded both CSA and NOVA specifications for toughness.
FAILURE FACTORS
The analysis identified two key factors which could result in failure: internal pressure of the 36-in. main line or external loads applied to the tee adjacent APX-L.
NOVA analyzed each scenario to determine if either could have resulted in the observed failure.
The possibility of static pressure overload was eliminated. The pressure at the time of failure was 782 psi (5,392 kPa), less than MAOP. The 36-in. main line adjacent APX-l was known to have sustained pressures as high as 860 psi as recently as Sept. 1, 1991, that is, before completion of the additional construction at the site.
Between that date and the time of failure, the pressure did not exceed 782 psi, although it did approach that level in December 1991.
Inability of the saddle to reinforce the tee was considered as a possible cause of failure. Chemical analyses and sizing of the fillet weld on the attachment bars indicated that the saddle would have had to have seen an internal pressure greater than 1,060 psi for the fillet welds to fail in shear, as observed.
Because fillet welds are not designed to take bending loads, however, any bending load applied to the fillet weld would greatly reduce its efficiency. Bending loads would only be seen by these welds if the saddle had already begun to deform.
The dimensions and mechanical test results of the saddle base-plate material showed that the saddle did not conform to the original specification. The base material conformed to an ASTM A516 Grade 70 material, not Grade 414, as specified.
Additionally, the measured wall thickness was 0.871 in. as opposed to the specified 0.619 in.,The CSA Z184-M86 design requirements imply a wan thickness of only 0.615 in. would be required, based on the tested material properties.
NOVA therefore concluded that the saddle was adequately designed and was not the cause of failure.
NOVA considered as a possible cause improper saddle fit-up during installation resulting in failure of the saddle to reinforce. But if this were a factor, failure would have occurred after installation of the hot tap during periods of maximum operating pressure before 1992.
The only other possible failure scenario relating to pressure was the possibility of a pressure transient. A pressure transient in gas is the equivalent to the water-hammer effect which occurs in fluid piping systems.
A simulation of the 36-in. main line system ruled out a pressure transient as the cause.
Soil loading resulting from the new 1991 30-in. construction adjacent APX-L was considered. All aspects of improper compaction, post-construction soil settlement, and differential soil settlement were checked in the piping analysis.
Soil samples were also taken to determine accurate inputs for the piping analysis.
NO SINGLE FACTOR
NOVA determined that none of these factors, individually, caused the 36-in. Western Alberta main line rupture.
The following applied external loads were considered: loads created by the new 30-in. installation, the use of heavy equipment in the area, and any pressure or flow anomalies in the 30-in. piping. NOVA employed a piping analysis to determine the magnitude of the loads created by the installation of the 30 in. This information was used as input for the finite-element modeling.
It was determined that the stresses created by the new construction were insufficient to cause failure at the tee.
An evaluation of the effect of heavy equipment working in the area has determined it was not the cause of failure.
The effects of pressure and flow anomalies in the 30-in. line have been shown to be insignificant.
The key result from the finite-element modeling was the analysis of external bending on the hot-tap assembly, as would be experienced as a result of settlement in the branch-connection piping.
The results showed that settlement of as little as 0.50.6 in. of the new construction at the 30 in.-to-24 in. junction (Fig. 2), combined with internal pressure, was sufficient to cause the rupture.
The settlement of the piping was believed to have occurred gradually over time. But stresses at the preexisting defect increase rapidly once a certain deflection is reached.
The analysis is complicated by the effect of the unknown gap between the saddle and the pipe, thereby affecting friction at the interface, and thus the load transfer.
Finally, NOVA assessed thermal loading and longitudinal restraint and ruled them out as possible causes of this failure.
CAUSES
All analyses of the individual scenarios indicated that a combination of events led to the rupture. The primary event was the noncompliant procedure used to weld the 24-in. stub to the 36-in. carrier pipe, resulting in a hydrogen crack at the toe of the weld.
This alone was insufficient to cause failure.
The installation of new piping in the fall of 1991 led to gradual settlement and thus, over time, increasing stresses applied to the hot tap.
During the few hours preceding rupture, the internal pressure in the pipeline was increasing, although the pressure never exceeded MAOP. In fact, pressures nearly as high were recorded 2 weeks before the rupture.
Through-wall fracture occurred when the stresses on the preexisting defect, as a result of internal pressure and settlement, reached a critical value. This was immediately followed by rapid g brittle fracture of the 36-in. main line.
REFERENCES
- NOVA Corp. of Alberta, "36-in. Western Alberta main line Rupture and James River interchange Meter Station Fir"an. 8, 1992," submitted to Energy Resources Conservation Board, Alberta Province, Mar. 23,1992.
- CAN /CSA-Z] 84-M86, Gas Pipeline Systems," Canadian Standards Association, 1986.
- Glover, A.G., Coote, R.I., and Pick, R.J., "Engineering Critical Assessment of Pipeline Girth Welds," Conference on Fitness for Purpose Validation of Welded Constructions, London, November 1981.
Copyright 1994 Oil & Gas Journal. All Rights Reserved.