Engineering Failure Analysis xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Failure investigation of 321 stainless steel pipe to flange weld joint

H.M. Shalaby
Petroleum Research Center, Kuwait Institute for Scientific Research, P.O. Box 24885, 13109 Safat, Kuwait

AR TI CLE I NF O AB S T R A CT

Keywords: Failure investigation was conducted on a refinery pipe-to-flange weld joint that suffered cracking.
Stress corrosion cracking Both the pipe and flange are made of AISI 321 stainless steel. The flange was circumferentially
321 stainless steel welded to the pipe which is seam welded. The investigation revealed that both the circumfer-
Delta ferrite ential and seam welds were in sound conditions, namely no evidences of sensitization, lack of
Weld joints
weld penetration, and voids or porosities. Thus, welding practices were not suspected to be the
cause of failure. The failure of the weld joint was found to have started at the δ-ferrite phase in
the flange material and propagated through the circumferential and seam welds. The failure
mode was concluded to be chloride stress corrosion cracking synergized by the presence of H2S.
The presence of corrosive compounds in the refinery stream and the residual stresses at the weld
joint triggered active anodic dissolution of the δ-ferrite precipitates, resulting in cracking of the
material.

1. Introduction

Austenitic stainless steels (SSs) are engineering alloys with good corrosion resistance in environments containing various com-
pounds of sulfur, normally experienced in refinery process streams. However, austenitic stainless steels are susceptible to stress
corrosion cracking (SCC) under certain conditions involving high stress, changes in metallurgical structure due to high temperature
operation, and the presence of specific chemicals that promote cracking. The chemicals present in refinery streams known to induce
SCC in these alloys are chlorides and polythionic acids [1]. However, recent reported failures and research of annealed austenitic
stainless steels under sour gas conditions with insignificant amount of chlorides suggested the contribution of H2S in promoting SCC
[2–6]. Cracking is derived from either H2S enhanced hydrogen absorption or H2S enhanced breakdown of the passive film by
synergistic action between H2S and Cl−. Due to the H2S/Cl− synergism, the presence of H2S extends the environmental domain of
Cl− SCC to lower temperatures and lower Cl− concentrations [7].
Traditionally austenitic stainless steels are joined with welding electrodes that contain 5–10% residual δ-ferrite in the inter-
dendritic boundaries. This retained metastable ferrite phase is believed to be influential in reducing hot cracking and microfissuring
of the weld metal. However, the presence of δ-ferrite ferrite in austenitic stainless steel weldments can be both beneficial and
detrimental depending upon the amount and distribution of δ-ferrite, the morphology of ferrite, the steel composition, and the
corrosion conditions [8–10].
The present paper presents the results of a failure investigation carried out on a refinery pipe to flange weld joint that suffered
stress corrosion cracking. The cracks were found to have initiated at the δ-ferrite phase in the flange material.

E-mail address: hshalaby@kisr.edu.kw.

http://dx.doi.org/10.1016/j.engfailanal.2017.06.047
Received 20 March 2016; Received in revised form 15 June 2017; Accepted 21 June 2017
1350-6307/ © 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: Shalaby, H.M., Engineering Failure Analysis (2017),
http://dx.doi.org/10.1016/j.engfailanal.2017.06.047
H.M. Shalaby Engineering Failure Analysis xxx (xxxx) xxx–xxx

Fig. 1. Photograph showing location and appearance of the crack found in the weld joint at the internal surface.

2. Background

A leak was observed at the external surface of a pipe to flange weld joint located in the downstream of a refinery heat exchanger
tube side outlet. The pipe is 12″ in diameter and is seam welded. It is circumferentially welded to the flange. Both the pipe and flange
are made of 321 SS. The leak was in the form of a pinhole at the heat affected zone (HAZ) of the circumferential weld. When dye
penetrant test was applied, a crack was found in the internal surface at the weld joint (Fig. 1). The internal surface was covered with
an adherent black scale. The crack propagated across the circumferential weld and extended towards the pipe seam weld. Table 1
shows the composition of the fluid stream going through the pipe. The operating temperature and pressure through the pipe are
192 °C (377 °F) and 1780 psig, respectively. The hydrogen pressure is 2850 psig.

3. Experimental details

Two samples were cut from the failed pipe/flange section. The first sample contained the cracked portion of the circumferential
pipe to flange weld and pipe seam weld, while the 2nd sample was taken at 90° angle away from the failure location for comparison
purposes. This later sample did not contain the seam weld, but contained pipe to flange circumferential weld. Fig. 2 shows the
locations of the cut samples.
Small sections were machined from the two cut samples for optical and scanning electron microscopic (SEM) examinations. The
flange/pipe direction was engraved on each section. The examinations were made before and after metallographic preparations. The
metallographic preparations included grinding, polishing, and etching. The metallographic preparations were made on two per-
pendicular surfaces; namely, the surfaces parallel to the pipe axis and the cross-sectional surfaces. For general microstructure, etching
was made in villela's reagent through immersion for 1 min in 5 ml HCl + 1 g picric acid in 95% ethanol. To reveal grain boundaries,
the material was electrolytically etched in 10% oxalic acid at 6 V for 1 min. Electrolytic etching was also made in 20% NaOH at 20 V

Table 1
Constituents of fluid stream going through the pipe.

Component % V/V

Hydrogen 80.62
Hydrogen sulfide 2.88
Ammonia 0.28
Methane 8.15
Ethane 0.48
Propane 0.37
Isobutan 0.07
N-Butan 0.18
C5 & heavier 1.3
Water 5.67

2
H.M. Shalaby Engineering Failure Analysis xxx (xxxx) xxx–xxx

Fig. 2. Photograph showing locations of samples cut for failure investigation.

for 20 s to preferentially outline and color δ-ferrite tan. All electrolytic etching was conducted at room temperature using SS cathode.
During the SEM examinations, elemental energy dispersive spectroscopic (EDS) analysis was made on the internal surface of the joint.

4. Results

4.1. Stereo microscopic examinations

Stereo microscopic examination of the as-received internal surface of the weld joint revealed that the crack existed in the material
of the flange and crossed the circumferential and seam welds (Fig. 3a). On the other hand, examination of the weld cross-section after
metallographic preparations revealed the existence of multiple cracking in the flange material near the circumferential weld and
away from it (Fig. 3b). It can be seen from the later macrograph that the cracks nearer to the circumferential weld propagated
through the weld and coalesced into one major crack with fine branches. In addition, no gross metallurgical defects were observed in
the weld. Thus, the weld microstructure appeared in sound condition.

4.2. Optical microscopic examinations

Fig. 4 shows an optical micrograph of the base metal of the pipe/circumferential weld area. The micrograph reveals the wrought
structure of the pipe base metal and the cast structure of the circumferential weld, respectively. Once again, no gross weld defects
were observed. Moreover, the austenitic microstructure of the base metal of the seam-welded pipe appeared relatively clean (i.e., did
not contain large size inclusions).

Fig. 3. Stereo macrographs, showing: (a) as-received internal surface of the weld joint at the area containing the crack; and (b) polished and etched weld joint cross-
section revealing that the cracks initiated in the flange material and propagated into the weld (etched in Villela's reagent).

3
H.M. Shalaby Engineering Failure Analysis xxx (xxxx) xxx–xxx

Fig. 4. Optical micrograph of the base metal of the pipe/circumferential weld area, showing absence of gross weld defects (10% oxalic acid etch).

Optical microscopic examinations of a cross-sectional surface revealed the presence of branched major cracks close to the cir-
cumferential weld and away from it (by ≈1 cm) in the austenitic structure of the flange (Fig. 5a). The cracks were similar to those of
stress corrosion cracking (SCC) of austenitic SSs. At higher magnifications, short-segmental microcracks were seen in the flange
material in areas containing major cracks and away from them (Fig. 5b). The short-segmental cracks were more in number in areas
surrounding the major cracks. The segmental cracks were observed as far as the end of the examined surface (i.e., ≈ 1 cm away from
the weld). The segmental microcracks were noticed to exhibit intergranular/transgranular paths, but were more of intergranular
nature (Fig. 5b and c). The segmental microcracks were found to have initiated at dispersed stringer-type second phase precipitates
(Fig. 5d). These precipitates were observed at the austenite grain boundaries, twin boundaries, and across both the austenite and twin
boundaries. However, more precipitates were seen along the grain and twin boundaries. Thus, the observed microcrack morphology
reflected the shape of the second phase precipitates, as the cracks were the result of active anodic dissolution of the precipitates. The
same type of dispersed stringers was also seen in the flange material 90o away from the failure location. However, no cracks were
found associated with these latter precipitates. Thus, it was concluded that this precipitated phase is an inherent part of the flange
metallurgy (i.e., was present in the as-received condition). It is worth stating that no excessive grain growth of the austenite phase
was observed in the flange material.
Optical microscopic examinations of the circumferential weld close to the flange after electrolytic etching in 20% NaOH revealed
typical microstructure of weld zone, namely dendritic and columnar grains structure (Fig. 6a). One major crack was seen in the weld
with considerable branches emanating from it. Some of the cracks appeared to have propagated along the δ-ferrite phase (see Fig. 6b).

Fig. 5. Optical micrographs showing: (a) overall view of cracking; (b) higher magnification revealing scattered short segmental intergranular/transgranular cracks in
the flange material; (c) short cracks initiated at a second phase in the form of stringers; and (d) details of the initiation of cracks at the second phase stringers (10%
oxalic acid etch).

4
H.M. Shalaby Engineering Failure Analysis xxx (xxxx) xxx–xxx

Fig. 6. Optical micrographs showing: (a) circumferential weld microstructure and cracks; and (b) crack branching, possibly propagating along the δ-ferrite phase in the
circumferential weld near the flange (20% NaOH etch).

Fig. 7. Optical micrographs showing the microstructure and crack morphology in the circumferential weld metal near the seam weld (10% oxalic acid etch).

Fig. 8. SEM micrographs of the as-received internal surface of the weld joint, showing: (a) cracks in the flange/circumferential weld area; and (b) cracks in the seam
weld area.

As can be seen in Fig. 7, oxalic acid etching was capable of outlining the carbides present in the area of the circumferential weld
metal close to the seam weld. The micrographs of Fig. 7 indicate that the crack links the precipitated carbides in the weld along the
austenite boundaries.

4.3. SEM examination and analysis

SEM examination of the as-received internal surface of the weld joint revealed that the flange material was relatively rough,
containing parallel grooves or bands in the axis direction (Fig. 8a). On the other hand, the surface of the seam weld was relatively
smooth (Fig. 8b). Increased surface roughness was also observed at the fusion zones of both welds. The internal surface of the joint
was observed to contain a small amount of scattered scales flaky in appearance (Fig. 8a). The deposited scales were more at the
surface of the flange than at the circumferential and seam welds.
EDS analysis of the internal surface of the weld joint below the circumferential weld provided spectrum exhibiting strong peaks of
Fe, Cr, and S (Fig. 9a). On the other hand, spot analysis of deposits observed inside a major crack yielded spectrum with strong S peak
and small peaks for Fe, Ni, Cr, and Ca (Fig. 9b). The presence of strong sulfur peak in both spectrum suggests the deposition of sulfides
at the internal metal surface and inside the cracks.

5
H.M. Shalaby Engineering Failure Analysis xxx (xxxx) xxx–xxx

(caption on next page)

6
H.M. Shalaby Engineering Failure Analysis xxx (xxxx) xxx–xxx

Fig. 9. (a) Area EDS analysis of the internal metal surface below the circumferential weld; and (b) spot EDS analysis of deposits observed inside a main crack.

Fig. 10. Line scanning of Cr, Ni and S across one of the observed stringer-type second phase precipitate in the flange material.

Fig. 10 shows line scanning of Cr, Ni, and S across one of the stringer-type precipitates. It can be seen that the precipitate contains
less Ni and S than the surrounding matrix. The depletion of Ni and S in the second phase precipitate could be attributed to elemental
segregation of these elements away from the precipitate. Since Ni is an austenite stabilizing element, its depletion suggests that the
second phase stringers are δ-ferrites.

5. Discussion

The present results indicated that the crack observed at the internal surface of the weld joint propagated through the cir-
cumferential and seam welds in a direction parallel to the pipe axis. This directionality of the crack strongly discounts sensitization of
the material as the cause of the observed cracking. Indeed, no grain boundary precipitations were seen after Oxalic acid etching. In
addition, both the circumferential and seam welds appeared in sound conditions. Thus, welding practices were not suspected to be

7
H.M. Shalaby Engineering Failure Analysis xxx (xxxx) xxx–xxx

the cause of this failure.

The internal surface of the flange in its as-received condition was noticed to contain more dispersed flaky scales than the cir-
cumferential and seam welds. This suggests that the material of the flange suffered more corrosion than the circumferential and seam
welds. Since the seam weld contained the least amount of scales, it can be safely assumed that it was exposed to a minimum amount
of active dissolution. The flaky scales and the precipitates inside the cracks were found to be sulfides, which are the usual corrosion
products in such a refinery stream.
Fig. 9 exhibited strong sulfur peaks in the EDS analyses made at the metal surface and inside the cracks. The presence of sulfur
suggests that the observed scales are sulfides. Several investigators [11–13] indicated that sulfur-containing compounds, such as
sulfurous acid, thiosulfuric acid, thiosulfate, tetrathionic acid, ploythionic acid, etc., can be formed during refinery unit shutdown
due to the interaction of the sulfide scale with moisture and oxygen at ambient temperatures. Some of these chemicals were found to
cause SCC of sensitized austenitic SSs through attacking metal inhomogeneity.
Optical microscopic examinations revealed the existence of considerable cracks inside the flange material near the weld and away
from it. These cracks appeared to be structurally dependent on the existence of stringer-type precipitates. Both the cracks and the
precipitates appeared to assume intergranular/transgranular nature, although the intergranular component was more dominant.
Similar precipitated phase was observed in the flange material 90o away from the failure location. However, no cracks were found
associated with these latter precipitates. Thus, the fact that the cracks developed near the weld joint and not away from it is simply
due to the increased weld residual stresses at such a junction.
Line scanning of one of the observed precipitates in the flange showed that Ni and S are depleted in these precipitates. Since nickel
is an austenite-stabilizing element, its depletion indicated that the stringer-type precipitates are δ-ferrite. Previous work [10,14] has
shown that this elemental partitioning can result in galvanic interaction between the δ-ferrite and austenite phase. The ferrite phase
has been shown to be more active electrochemically than the austenite. On the other hand, several investigators [8,15] attributed the
attack at δ-ferrite to the formation of carbides, which is accompanied by “ferrite decay” in which the chromium content of δ-ferrite
decreases as it contributes to the formation of carbides at the austenite/δ-ferrite interface. Nazdezhdin, et al. [16] stated that if the
environment is capable of inducing SCC, the propagation of the initiated cracks could be retarded by the discontinuous islands of δ-
ferrite. On the other hand, stringer-type δ-ferrite may provide a continuous path, if present above a certain level.
The present results suggest that the failure of the weld joint was due to chloride SCC that initiated at the δ-ferrite phase in the
flange material. The chloride SCC was synergized by the presence of H2S in the refinery stream. Thus, the presence of corrosive
compounds in the refinery stream and the residual stresses at the weld joint triggered the active anodic dissolution of the δ-ferrite
precipitates. Consequently, the formation of microgalvanic cells at the precipitates resulted in the formation of microcracks replacing
the stringers. The linkage of the microcracks led to the formation of the observed large size cracks. It is only natural that the cracks
closest to the circumferential weld would grow through the weld and progress into the seam weld. This could be attributed to the
increased stress concentration, the higher weld hardness, the presence of dispersed carbides associated with the weld ferrites, and
finally it is easier for the cracks to grow in the δ-ferrite phase of the weld than the austenite.

6. Conclusions

1. The failure of the weld joint was concluded to be due to chloride SCC synergized by the presence of H2S in the refinery stream.
2. Cracking started at the δ-ferrite phase in the flange material and propagated through the circumferential and seam welds.
3. The conjoint action of residual stresses at the weld joint, the δ-ferrite phase in the flange material and the presence of corrosive
compounds in the refinery stream was the triggering cause of cracking.

Acknowledgement

Thanks to W.T. Riad and F. Al-Sabti for conducting the optical microscopy work and Dr. N. Tanoli for carrying out the SEM and
EDS work.

References

[1] A.J. Brophy, Stress corrosion cracking of austenitic stainless steels in refinery environments, Mater. Perform. 13 (5) (1974) 9–15.
[2] Y.M. Liou, S.Y. Chiu, C.L. Lee, H.C. Shih, Electrochemical pitting behaviour of type 321 stainless steel in sulphide-containing chloride solutions, J. Appl.
Electrochem. 29 (1999) 1377–1381.
[3] V. Singh, Performance of austenitic stainless steels in wet sour gas, part 1, Mater. Perform. 47 (8) (2004) 52–55.
[4] V. Singh, Performance of austenitic stainless steels in wet sour gas, part 2, Mater. Perform. 47 (9) (2004) 46–50.
[5] Y.Y. Chen, Y.M. Liou, H.C. Shih, Stress corrosion cracking of type 321 stainless steels in simulated petrochemical process environments containing hydrogen
sulphide and chloride, Mater. Sci. Eng. A 407 (2005) 114–126.
[6] Y.Y. Chen, H.C. Shih, Y.M. Liou, L.H. Wang, J.C. Oung, Stress corrosion cracking of type 321 stainless steel under simulated petrochemical conditions containing
thiosulfate and chloride, Corrosion 62 (2006) 781–794.
[7] P.R. Rhodes, Environment-assisted cracking of corrosion resistant alloys in oil and gas production environments: a review, Corrosion 57 (2001) 923–966.
[8] K. Prasad Rao, S. Prasanna Kumar, Electrochemical studies on corrosion behavior of austenitic stainless steel weld and clad metals, Corrosion 41 (1985)
234–242.
[9] K. Prasad Rao, S. Prasanna Kumar, Corrosion behavior of austenitic weld and clad metals in accelerated boiling acid tests simulating passive conditions,
Corrosion 42 (1986) 1–10.
[10] Z. Fang, Y. Wu, R. Zhu, Stress corrosion cracking of type 304 stainless weldments in the active state, Corrosion 50 (1994) 171–175.
[11] S. Ahmad, M.L. Mehta, S.K. Saraf, I.P. Saraswat, Stress corrosion cracking of sensitized 304 austenitic stainless steel in petroleum refinery environment,

8
H.M. Shalaby Engineering Failure Analysis xxx (xxxx) xxx–xxx

Corrosion 38 (6) (1982) 347–353.

[12] C.H. Samans, Stress corrosion cracking susceptibility of stainless steels and nickel base alloys in polythionic acid and acid copper sulfate solution, Corrosion 20
(1964) 236t–262t.
[13] R.L. Piehl, Stress corrosion cracking by sulfur acids, Proc. API, vol. 44[III], 1964, pp. 189–197.
[14] T.G. Gooch, J. Honeycombe, P. Walker, Welding Institute Report MICS 23/7, Materials Performance, Cambridge, U.K., 1970 (July).
[15] R. Castro, J.J. de Cadenet, Welding Metallurgy of Stainless Steel and Heat Resisting Steels, Cambridge University Press, Cambridge, 1975, p. 87.
[16] A. Nazdezhdin, K. Cooper, G. Timbers, Effect of ferrite on cast stainless steels, Mater. Perform. 33 (9) (1994) 62–65.