Hook cracks are one of several defect types that are characteristic of electric-resistance weld (ERW) seams, whose occurrence traces to the use of dirty steel, as elaborated later in this QR Segment. The seams considered in this discussion include low frequency (LF) contact processes, the flash-weld (FW) process[1], and high frequency (HF) welds made by a contact (HFC) or induction (HFI) process. For those interested in other types of ERW seam defects, such details can be found via the links: and The content of these links is the basis for much of what follows.

API 5T1 [2] defines hook cracks (or ‘upturned fiber imperfections’) identically for the FW and ERW processes as “separations resulting from imperfections at the edge of the skelp, parallel to the surface, which turn toward the inside diameter (ID) or outside diameter (OD) pipe surface when the edges are upset during welding.” Such imperfections develop due to through-thickness weakness that can exist in the microstructures of ‘dirty steel’. Such defects were more prevalent until steel cleanliness was recognized as critical for higher toughness at lower transition temperatures in the 1960s. Even so, steel can be sourced today for applications specifying ISO 3183 / API 5L PSL2 pipe [3] that can and have led to hook cracking.

Hook cracks are the primary source of defects in the upset pipe wall and heat-affected zone (HAZ) that form when making an ERW seam. The terms ‘upset’ in regard to the pipe wall and ‘HAZ’ due to skelp-edge heating are illustrated in reference to the cross-section through a typical hook crack, as shown in Figure 1. This image is similar to that used in the Material and Construction (M&C) Defects Chapter to illustrate such cracking, except that more of the cross-section is shown and it is annotated. As the FW process involves arcing prior to bumping (upsetting) the seam, the heat input was high, which led to a broad HAZ. The upside was that the weld cooled slowly which generally avoided undesirable martensitic microstructures.

Figure 1. View of ERW seam identifying the upset wall and the HAZ.

API 5T1 [2] identifies several types of defect that are specific to upset autogenous welds. It makes reference to contact marks and related arc burns “resulting from the electrical contact between the electrode ….. and the pipe surface.” Such features, however, tend to occur beyond the upset and HAZ due to ERW seam production. API 5T1 also makes reference to inclusions that are “foreign material or non-metallic material entrapped in the metal during solidification” that as such lie within the bondline. In this context hook cracks are the only recognized defect that forms within the upset wall and HAZ for typical ERW seams. While less a concern for FW seams, the other seam processes can be prone to form locally “hard” microstructures that can be sites for environmentally assisted cracking (EAC) in service, which if severe could promote a hydrotest failure. EAC can be an issue where locally hard microstructures form if they remain in the pipe due to a process upset during the post-weld heat treatment (PWHT) if such was used. Process upsets are known due to inadequate heating, as well as due to incorrect positioning of the PWHT relative to the location of the seam, rendering it ineffective.

The hook crack shown above developed in the upset of a well-trimmed FW seam, which given the differing structures that are evident either side of the bondline was made using split-skelp. The macrostructure evident for this weld also shows that it had been produced with a quite dirty steel. The term ‘split skelp’ refers to the slitting of the coil-supplied skelp to produce a strip whose width developed the mismatch required to upset the wall as the pipe passed through the pipe forming cage. For those not familiar with the ERW process and these terms, primers on pipe making and supporting videos can be found through the following links:

Video: general pipe making

Video: ERW-seamed smaller diameter pipe

Low Quality (Dirty) Steel is Central to Forming Hook Cracks

Dirty steel is central to the formation of hook cracks because they initiate and grow between the bands of pancaked/strung-out Manganese Sulfide inclusions and/or other impurities and oxides, such as Aluminum Oxide. These microstructural bands with strung-out inclusions and oxides turn toward the ID or OD pipe surface as the upset forms, creating the characteristic hook shape feature that underlies the name of this defect. The interface along such inclusions/impurities is weak. That interface separates, leading to crack initiation on the ID or the OD where the interface intersects the scarf’d surface of the upset and is closest to perpendicular to the pressure-induced hoop stress, as shown in more detail for another hook crack in a FW seam.

a) section through upset in a “dirty steel”

b) fracture surface for a “hook crack” origin

Figure 2. Two views characteristic of hook cracks.

Figure 2b provides a three-dimensional (3D) view of the fracture surface tracking along the outbent banding, with the origin nucleating at the surface along this feature and the adjacent side of this banding, and the propagating into the wall. This cracking origin developed in 12.75-inch (324 mm) diameter x 0.250-inch (6.35 mm) X46 (Gr 317) nonexpanded pipe made by Republic Steel circa 1958 using a LF ERW process. Figure 3 is a typical view of a banded microstructure, which includes a pancaked oxide inclusion.

Figure 3. Typical banded microstructure, with a pancaked oxide inclusion.

Figure 4a indicates that the cracking in Figure 2b ran about 40% into the wall before it bent around as it began to track this banding back parallel to the pipe surface, at which point its growth along the hook path arrested. This transition in cracking planes is evident in Figure 2b. Because the operation of this hazardous liquid pipeline at a peak stress = 67% of SMYS, included significant pressure cycling, this relatively deep long feature eventually initiated fatigue cracking, which grew and led to an in-service leak. Figure 4b provides an overview of this cracking, which shows sequential steps in the oxidation of this fracture surface as this fatigue cracking grew over time and eventually grew stably through the wall, leading to a leak.

a) initially tracking the outbent banding

b) fatigue crack growth at a “hook crack” origin

c) overview along the fatigue origins from the hook cracking at cracking secondary to the leak

Figure 4. Fracture surfaces at hook cracking in liquid pipeline service (t=0.250” / 6.35mm).

As ‘dirty steel’ underlies hook cracking, such cracking was most problematic until the 1960s when steel cleanliness was recognized as a critical factor for higher toughness at lower transition temperatures. While in that context they were more an issue in vintage systems, centerline segregation has led to hook cracks forming and failing in modern line pipe made using continuous cast (concast) steel. Figure 5 illustrates such segregation in pipe steel sourced from 2003 concast production. This section was cut about one foot (33mm) removed from the end of a hook crack that leaked during hydrotesting of this smaller-diameter HFI ERW line pipe. In this context hook crack defects can be found in modern pipe, just as they were in the earlier ERW seams. They are however less frequent due to improved steel quality practices and better specifications. The cause for such cracking traces to the steel, not the process, such that improvements to the HF processes will not alleviate this concern. It should be emphasized in this context that while centerline segregation promotes hook cracking that tracks to this feature, the presence of a banded microstructure structure will support hook and other forms of cracking at any point through the wall, which depends on where in the upset the cracking originates.

Figure 5. Hook feature associated with centerline segregation in a PWHT HFC seam.

Additional Images and Discussion of Hook Cracks

Figures 6a and b show macro-etched sections of hook cracking. Figures 6a is similar to Figure 1 as it shows initiation from one side of the wall (ID) whereas Figures 6b is a view from both sides, as happens occasionally. In this case that more or less symmetrically, although such behavior depends on the local circumstances. The reference to ‘flow lines’ in these figure titles and later relates to the inelastic flow that occurs as the pipe wall is physically upset in making the seam.

Depending on the number of sites that initiate such cracking over the length of the hook crack, and the number of near-parallel planes of inclusions that are sufficiently weak and become involved in the failure, cross sections made through hook cracks can vary greatly in complexity. This is evident in comparing Figures 6a, 6b, and 6c. For example, the image in Figures 6a reflects a situation where a single dominant origin formed and grew until the feature reached a critical size, and ruptured while in liquid service. In this case, the origin tracks the dominant flow line and then turns as expected onto a plane perpendicular to the hoop stress. In contrast, the path for the feature shown in Figures 6c is quite complex, with an unetched section used to clearly illustrate this complexity. Related microscopy shows that this hook crack’s origin lies to the right of the bondline, and that after briefly tracking the flow line of its origin the crack path jumps onto and/or across other planes of inclusion stringers, eventually crossing the bondline and then tracking back to it to fail in the bondline.

Logically, where the cracking evident in a cross section is simple, the fracture surface appears equally simple, whereas when such cracking tracks a complex path the fracture surface shows a quite complex appearance. Hook cracking and their fracture surfaces run from the simple views in Figures 6a, 6b, and 6d, to more complex than evident in Figures 6c and 6e. While sizing the origin in Figures 6d poses little problem, sizing origins like that in Figures 6e that vary in depth is much more difficult. Features such as that in Figure 6e have been termed ‘woody fractures’ in some reporting.

a) ID origin: simple crack path tracking macro-flow lines

b) ID/OD origins: simple macro-flow cracking

c) OD Origin: complex crack path  tracking planes of inclusions

FW – t = 0.281” (7.14 mm)

FW – t = 0.281” (7.14 mm)

LF ERW – t = 0.375” 9.53 mm)

d) fracture surface for a hook crack originating in and running along a dominant macro flow plane the defect exposed in a hydrotests of LF ERW pipe – t = 0.344” (6.20 mm) is clearly evident

e) portion along the facture surface of more complex hook crack failure – the origin is about the same depth as in part d), but uneven in depth, being somewhat deeper than apparent in the section shown in part c) – exposed in hydrotesting of LF ERW pipe – t = 0.375” (9.53 mm)

Figure 6. Macro views typical of hook cracking – OD is shown topside.

As for bondline defects, hook cracks tend to be axial, and can have planar components that can lie near normal to the pipe wall, with segments of the crack’s surface also tracking the flow lines. Such was the case for Figures 6a, 6b, and 6d. However, as shown above, cracking that originates in a hook crack can be complex, as was the case in regard to Figures 6c and 6e. As evident in Figures 6b, cracking with a hook crack origin can initiate from both the ID and the OD at the same axial location, and grow along the flow line that originates in prior to turning onto a plane perpendicular to the hoop stress. As evident in Figures 6b, both ID and OD origins turned and grew to almost the same depth, after which both tracked along a flow plane before turning again onto a plane perpendicular to the hoop stress, where they coalesced causing a rupture. In scenarios where such growth is stable TW, hook cracks can develop to a depth equal to the full thickness of the pipe, rather than just 50% of the wall as some have reported.

Hook cracks initiate, and grow in depth and length, and fail due to pressure-induced loads across the seam, just as occurs for bondline defects, and like those defects could also fail due to ovalization if its effects were focused in the vicinity of the seam. Failure for a hook crack depends on the length, depth, and the axial and TW continuity of the defect, the properties of the interface and HAZ, and the magnitude of the load across the seam. Either fracture or plastic-collapse can control this failure, with the failure mode as brittle versus ductile versus mixed depending on the transition versus service temperature, and both the consequences as leak versus rupture, and the axial extent of axial propagation depending on the properties (and also the transported product). QR Segments are available dealing with these topics, which can be accessed by scanning over the embedded links. Accurate predictions of failure pressure can be achieved for such cracking, but as always this requires knowledge of the defect’s size and shape and the local properties. These and related topics are considered in detail, which can be found through the following link:

See also Failure Mechanisms & Pipeline Properties and Collapse and Fracture Controlled Failure.

Occasionally, because of process problems during the PWHT smaller hook-crack origins and origins in other seam defects fail in a brittle manner, whereas they might have remained stable if the steel was more fracture resistant. Where the origin is due to mill process issues such as alignment or edge defects, such origins are identified by those terms, often in conjunction with a generic label like mill defect. As becomes evident below, poor vertical alignment of the edges (i.e., mismatched or offset edges) can cause asymmetric / skewed welds, which can lead to hook cracking. Similar issues follow due to poor edge preparation. The degree to which such issues develop depends on the extent of the mismatch or edge-preparation issues, and the amount of upset force. Larger mismatch coupled with higher upset forces reflect the worst case scenario, which when coupled with scarfing can cause thinning of the wall along the weld – which in some cases is already the weakest link for that joint of pipe.

The cross section in Figure 7 shows etched views of LF ERW that had a nominally 0.250-inch (6.35 mm) thick wall, where the effects of offset edges combined with the upset force to produce a skewed bondline. It is apparent that the OD was over-scarf slightly to the left side of the bondline, but due to the vertical offset this over-scarf was magnified on the right side. This wall-thickness reduction adjacent to the seam was present over most of the length of the joint that in this case initiated a seam split in a hydrotest. The seam was skewed the order of 12, which led to ~15% net reduction in the pipe wall due to the combined effects of the OD and ID scarf. The extent of the over-scarf in this case was compounded by a within-tolerance but less than nominal thickness pipe wall.

Figure 7. Rupture due to offset edges: LF ERW – t = 0.250” (6.35 mm)

Dirty Steels also can Promote Selective Seam Weld Corrosion

Steel cleanliness also can be a concern because dirty chemistries can lead to local corrosion cells that can focus corrosion in and along the REW seam, potentially leading to selective seam weld corrosion (SSWC) in the bondline. This selective attack leads to a vee-shaped groove that often has a uniform appearance on the OD with a near constant depth profile, which may lay on a largely uncorroded OD pipe surface, or may fall within a patch of corrosion. Visual traits that are typical of SSWC are illustrated in Figure 8.

a) typical view of uniform SSWC

b) typical x-section t = 0.312” (7.93 mm)

Figure 8. SSWC that here shows uniform attack absent significant surface corrosion.

This process can continue to deepen as a vee-notch, but also can transition to produce a focused crack-like attack. Further discussion and references can be found through the link:

Other types of defects can be manifest across the bondline and the upset reflect process upsets involving alignment/fit-up across what becomes the bondline. Discussion of these scenarios also can be found at the just-noted link.

Hook Cracks and Failure Pressure – Historic Trending

Absent in-service growth due to mechanisms like fatigue hook cracks generally fail at quite high pressures. This has been demonstrated in reference to the failure pressures trended over time, as evident in Figure 9, which presents results adapted from Reference 4.

Figure 9. Contrasting trends in failure pressure for hook versus cold-weld cracking.

Figure 9 presents the cumulative frequency of failure on the y-axis as a function of the failure pressure shown on the x-axis, with trends shown for both hook cracks and for cold welds (i.e. bondline failures). The underlying data represent archival reporting involving LF and HF ERW seam failures over the interval from ~1960 through ~2010. Of the almost 300 records available about 63% involved cold welds while about 27% were due to hook cracking. As such the trends shown in Figure 9 dominate the corresponding trend developed for the complete dataset, which also included results for stitched cracking and SSWC – both of which also occur in the bondline.

It is apparent from Figure 9 that the trends in failure pressure normalized relative to Psmys up to ~65% of SMYS are comparable for these quite different types of defects, whereas at higher pressures their trends are very different. While Reference 4 did not address causes for either the initially similar trends or their differences at higher pressures, the results evident in Figure 9 can be rationalized in regard to the role of fatigue cracking and the toughness gradient across the wall upset, as follows.

The longer, deeper features for both types of defects will fail at lower pressures prior to a plausible role for mechanism like fatigue becoming active. Accordingly, the worst-case features for both types of defect fail under comparable circumstances – which underlies their similar trends at lower pressures. As the initial sizes of the features becomes smaller, the failure processes for these defects is complicated by the effects of toughness and service conditions. As the failure pressure increases the features failing must be shallower and/or shorter, but what can fail at a given pressure is subject to effect of local toughness in the bondline versus the upset. The differing operational histories in gas and liquid transmission pipelines adds to this complexity.

Both types of defects can grow by fatigue in liquid pipelines, with fatigue much less likely for gas pipelines. The extent (significance) of fatigue crack growth, as well as the defect’s tolerance of pressure, are limited for both types of defects by the local toughness in the wall upset versus the bondline. Data indicate that bondline toughness can be much less than that of the upset, with the bondline also showing a higher ductile to brittle transition temperature. [5,6] On this basis, where fatigue can be active over a significant crack depth failures will be evident at lower pressures, as can occur for hook cracks but not at cold cracks when they lie in a less tough bondline. Regardless, as the failure pressure continues to increase, the limited toughness in the bondline will at some pressure promote a sharp upturn in the incidence of failures for the cold welds, but not for the hook cracks that lie in locally tougher steel, leading to a cross-over in their failure frequencies. Not surprisingly that cross-over is evident in Figure 9 at about 90% of SMYS, which is the minimum pressure for hydrotesting for the dataset in Figure 9. In contrast, the hook cracks in the tougher upset resist failing and so survive to higher pressures.

Key in this context is the observation that for the older pipelines that have experienced a prior hydrotest the worst-case larger defects that lie in lower-toughness pipes will have failed in that prior hydrotest. As those larger features get exposed, the test pressure to cause failure will rise.

Hook Cracks and Failure Pressure – Recent Trends

Recent experience with hydrotesting done on pipe joints removed from service as a part of rehabilitation motivated by inline inspection (ILI) data[2] have developed results consistent with the above speculation. For example, recent hydrotesting done by an operator as part of an EMATs tool calibration/tuning program for hook cracks, as well as testing done to more broadly quantify the failure behavior of ERW seams, have both developed results that are consistent with the historic patterns. In the case of the tool calibration/tuning program for hook cracks, three joints of pipe were excised for testing with a view to characterize failure in a ‘representative sample of features’. For the work directed at more broadly quantifying the failure behavior of ERW seams three joints of pipe also were selected from pipe excised in rehabilitation, but now with a view to characterize failure at ‘the worst-case features’ that were found via ILI, supported by laboratory non-destructive inspection. On average, the failure pressure in these tests exceeded 130% of SMYS, whereas relative to the UTS failure occurred at close to 90% of the collapse pressure for defect-free pipes. Such pressures fall well above that shown for the historic trend, which is anticipated as the pipelines these test pipes came from had been exposed to a previous high-pressure hydrotest.

It follows that the higher toughness in the upset wall versus that of the bondline means that failures in the upset will occur at relatively higher pressures than those in the bondline – all else being equal. For this reason, features identified via ILI as crack or crack-like that motivate field verification digs and/or other actions will have little value in managing risk when those feature lie in the upset wall. Improved probability of identification of hook cracks would therefore show clear value-added.


Discussion in this QR Segment has considered hook cracks, which can form in the upset wall / HAZ of the seam due to the use of dirty steel. Absent operational cycling or other factors that promote their growth, hook cracks that survive pre-service hydrotesting will remain stable thereafter.

The steep gradient showing higher toughness in the upset wall versus that of the bondline means that failures in the upset wall occur at relatively higher pressures than in the bondline. For this reason, features identified via ILI as crack or crack-like that motivate field verification digs and/or other actions will have little value in managing risk when those features lie in the wall upset. Improved probability of identification of hook cracks would therefore show clear value-added.


  1. Messier, R. W., Jr., Principles of Welding, Processes, Physics, Chemistry, and Metallurgy, Wiley-VCH Verlag, 1999.
  2. Anon., “API Bulletin on Imperfection Technology,” API Bul. 5T1 (R2010), 2010
  3. Anon., “Petroleum and natural gas industries – Steel pipe for pipeline transportation systems” ISO 3183 / Specification for Line Pipe, API 5L
  4. Leis, B. N. and Nestleroth, J. B., “Battelle’s Experience with ERW and Flash Weld Seam Failures: Causes and Implications,” Interim Report, Subtask 1.4, US DoT DTPH56-11-T-000003, September 20, 2013.
  5. Leis, B. N. and Nestleroth, J. B., “Compare – Contrast Analysis of Inspection Data and Failure Predictions versus Burst-Test Outcomes for ERW-Seam Defects,” Interim Report, Subtask 4.1, US DoT DTPH56-11-T-000003, April, 2013.
  6. Leis, B. N., “Characterizing the Fracture Resistance of ERW Seams,” Journal of Pipeline Engineering, Q3, 2018, pp. 165-181.

  1. Welding textbooks [e.g., see 1] indicate that although flash welding heats the faying surfaces by combined resistance and arcing it is classified a resistance welding process, and as such is included in this discussion. Large diameter FW line pipe was made only in the US by the A O Smith Corp. beginning circa 1928. The process relied on a method of economically producing large-diameter steel line pipe that helped to launch the transcontinental gas and oil pipeline business. A. O. Smith was a leading supplier of line pipe until it exited the business in 1972 according to details that can be found through the link: https: //
  2. The ILI anomalies were identified as crack or crack-like using current-generation crack-detection tools that were based on ultrasonic technologies (UT and EMATs).