Engineering Failure Analysis 16 (2009) 1825–1836

Contents lists available at ScienceDirect

Engineering Failure Analysis

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

Common root causes of recent failures of flanges in pressure vessels

subjected to dynamic loads
J.L. Otegui a,*, P.G. Fazzini b, A. Márquez a
a
INTEMA – University Mar del Plata – CONICET, Welding and Fracture Division, J.B. Justo 4302, (B7608FDQ) Mar del Plata, Argentina
b
GIE S.A. Galicia 52, (B7608AUB) Mar del Plata, Argentina

a r t i c l e i n f o a b s t r a c t

Article history: Present rules in fabrication codes are aimed to reduce the probability of fatigue cracks in
Received 27 August 2008 flanges welded to pressure vessels subjected to cyclic pressure or vibrations. Yet, several
Accepted 29 August 2008 leaks and ruptures have recently occurred at flanges in pressure vessels and pipes. A review
Available online 17 September 2008
of three cases is presented, which involved five failures; their common root causes are dis-
cussed, and the influences of manufacture and operation conditions on crack initiation and
Keywords: propagation mechanisms are highlighted. Some cracks initiate from the outer surface, but
Fatigue failures
many cracks initiate in the outer half of the thickness of the reinforcement, from very dif-
Pressure vessels
Flanges
ficult to avoid weld defects. Ultrasonic testing, with an adequate procedure, can be reliably
Reinforcement saddles used to detect these defects before they become leaks.
Weld defects Common aspects of these failures are crack initiation in weld metal and heat affected
zone of welds, all related to an inadequate design of the reinforcement and poor execution
of the welding procedure. Weld inadequacy is in one case a result of an increase in thick-
ness as an attempt to increase safety; which also increased cyclic stresses due to excessive
weight of the vessel. Fillet welded reinforcements induced rigidity inconsistencies within
the flange joint, and concentrated deformations in the failed regions. In another case, pre-
operational hydrotest could have been detrimental.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Fatigue cracks in flanges welded to pressure vessels subjected to dynamic loads due to cyclic pressure or vibrations
transmitted through the vessel and pipe bodies have been very frequent in the past [1,2]. Present rules in fabrication
codes are aimed to reduce their probability. Yet, it has been found that several leaks and ruptures have occurred in
pressures vessels in the gas and petrochemical industries [3]. The derived costs of some of these failures have been large,
for what it is interesting to determine the influence of constructive and operation aspects, and possible mitigation
alternatives. In this paper, a review of some of these cases is presented and their common root causes highlighted
and discussed. The analysis is focused in the five following failures, all related to a crack in the reinforcement saddle
of a pipe flange welded to the vessel:

 Case 1: Two discharge vessels at an alternating-compressor in a petrochemical plant.

 Case 2: 1200 collector pipe in a natural gas turbine-compressor plant.
 Case 3: Suction vessels in two natural gas alternating-compressor plants.

* Corresponding author. Fax: +54 223 481 0046.

E-mail address: jotegui@fi.mdp.edu.ar (J.L. Otegui).
URL: http://www.giemdp.com.ar (P.G. Fazzini).

1350-6307/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2008.08.027
1826 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836

These cases were studied using the usual techniques for analysis of mechanical failures:

 Analysis of operation and construction conditions.

 Visual and NDT inspections.
 Fractographic analyses.
 Mechanical and chemical analyses of plate reinforcement and saddle materials.
 Microstructural characterization.

in order to get conclusions related to:

 Characterization of crack propagation into vessel material.

 Influence of manufacture on crack initiation and propagation mechanisms.
 Influence of operation conditions, and possible mitigation measures.

2. Discharge vessels at an alternating-compressor in a petrochemical plant

Original and replacement vessels failed the same way. They were located in a polymerization line, subjected to the char-
acteristic conditions of pressure and vibrations of an alternating-compressor [4]. The original vessel was designed in 1979
according to ASME PVBC code, section VIII Div. 1[5], was built in 1991, and failed four years later. Operation conditions
are: temperature 110 °C, pressure 5 bar, and gas flow 4000/4500 Kg/h. The inner fluid is Ethylene plus a small quantity of
solvent and serums coming from the low pressure recycle. The compressor has a rotation frequency of 375 rpm. The vessels
are supported by means of elastic supports located in its inner part (grey arrow in Fig. 1), which were never subjected to non-
destructive tests.
Material from saddle and reinforcement plate are common C–Mn ferritic pearlitic structural steels, plate material is 0.10 C
and 0.66 Mn, while saddle material is 0.14 C and 1.33 Mn. Sulfur and phosphorus levels are average 0.014–0.022%, less than
the 0.03% upper limit required for these steels. Material micro-hardness results for base and weld materials close to the area
of the crack is around 20 HRc, while 30 HRc hardness values were found in some spots in the saddle HAZ. Hard spots were
confirmed by course grained martensitic microstructures in HAZ.
Leaks developed where indicated by white arrow in Fig. 1, at the end of a longitudinal reinforcement of the pipe inlet. Pipe
and reinforcement are 6.7 mm thick, saddle and vessel are 12.8 mm (design thickness of the vessel shell was 9.5 mm). Excess
heat input caused undercuts in HAZ of the thinner plate. Brittle HAZ developed from too high cooling rates in saddle mate-
rial. Discontinuous fillet welds were specified for the vertical reinforcement of the pipe inlet, these welds created high stress
raisers at the ends of weld, and especially at the end of the reinforcement plate.
The crack initiated at the root of the fillet weld, and grew parallel to the surface of the vessel wall, on the outer surface of
the saddle, see Fig. 2a–c. The crack path followed the weakest section in the fillet weld, near the end of the reinforcement of

Fig. 1. Case 1: Elastics supports of discharge vessels.

 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836 1827

Fig. 2. The cracked area.

the pipe inlet, in the course grained HAZ of the saddle plate. The white arrows in Fig. 2c schematically indicate the direction
of crack advance. The fracture surface shows the typical flat morphology indicative of propagation by mechanical fatigue.
Very delicate and spaced beach marks are observed, indicating that the amplitude of the cyclic stresses were mostly constant
during the whole crack propagation process.
Crack propagation was due to small amplitude stress cycles, during a very large number of cycles. A change in the rough-
ness of the fracture surface is apparent in the upper part of the crack in Fig. 2c, with a river pattern marks parallel to the crack
growth direction, indicative or faster crack growth. The fracture surface grows inclined with respect to the thickness, along
the perimeter of the vessel, with a butterfly wing shape that completely penetrates the thickness of the saddle and partially
penetrates the vessel wall. The deepest point of the crack is 92.3% of the vessel thickness, at the position of the weld root. The
remaining ligament was of only 1 mm. A gas leak was about to occur if no remedial action were taken.
The spring-loaded supports located at the bottom of the vessel (grey arrow in Fig. 1) include frictional dampeners, which
showed indications of excessive wear. All signs of malfunction are consistent with the excessive weight of the vessel, due to
being much thicker than designed.
After the first failure, the plate reinforced flange was replaced by a weldolet – type welded flange, see Fig. 3. However, this
new flange connection was also built with a lengthwise reinforcement welded with discontinuous fillet beads. A new failure
occurred, this time in a shorter time, at the same position, see insert in Fig. 3. As expected, a finite element model of stresses
due to dead weight and internal pressure (Fig. 4) shows that the end of the weld between reinforcement plate and saddle is
the highest stressed region.
The failure of this partial redesign was due to the incomplete understanding of the stresses that were being generated at
the initiation site. The stress concentration at the end of the pipe reinforcement was left unresolved. In this case crack
growth occurred solely through the vessel wall, which led to an even reduced fatigue live. This study reveals that:

 The initiation of the crack that gave origin to the failure was located in the heat affected zone of the plate to saddle fillet
weld, and is related to an inadequate design of the reinforcement and of the welding procedure. High hardness of HAZ and
lamination defects in saddle material contributed to generate weak spots for crack initiation.
 Weld inadequacy is mostly a result of the large difference in thickness between the actual vessel and saddle (12.8 mm)
and the design thickness (9.5 mm). Subsequent fatigue propagation of the cracks is related to a large number of cyclic
loads, most probably due to vibrations coming from the compressor.
 Fatigue stresses were larger than expected also because the excessive weight of the vessel was not properly balanced by
the elastic foundation, designed for a thinner, 25% lighter vessel.
1828 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836

Fig. 3. Welded flange in the second vessel, replacement of the failed vessel.

Fig. 4. Finite element model of stresses due to dead weight and internal pressure.

 The discontinuous fillet welded longitudinal reinforcement plate increases bending rigidity of the flange joint, and con-
centrates deformations in the failed region. It is not a mechanically very efficient design.
 Changes in elastic supports and NDT verification for possible cracks and hard spots in reinforcement welds were under-
taken to avoid recurrence of cracking in this vessel and other similar in the plant. Thicknesses and support systems were
set according to manufacturer specifications.
 Eliminating the reinforcement plate saddle in this case was non conducive; the next case studies, however, show how
improper construction of these saddles and the difficulty to detect bad procedures once the vessel is built could severely
reduce reliability.

3. Collector pipe in a natural gas turbine-compressor plant

Dehydrated natural gas at 63 bar flows through this discharge piping system at a natural gas turbine-compressor plant
[6], see Fig. 5. A longitudinal fracture occurred in the weld joint between a 300 mm (1200 ) diameter collector and a
150 mm (600 ) diameter venting pipe, see arrow in Fig. 5, after many years of service. Four hours before, discharge pressure
of the compressors increased to 77 bar, for less than half an hour. This was due to a faulty valve. The section fractured in
 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836 1829

Fig. 5. Case 2: Location of element failed in collector pipe.

two sections, both fractures initiated in the welded joint of the 600 tube, a fillet welded flange with a circular reinforcement
plate fillet welded to the collector.
Collector material has a SMYS of 400 MPa, probably an API 5L X56, SMYS for tube material only gets to 312 MPa. The
upper end of the tube section adhered to the weld shows around 20% reduction of thickness due to plastic deformation.
The root of the tube-to-reinforcement weld shows small tears and fractures that indicate lengthwise traction of the tube.
Fig. 6a shows the failed section. The severe deformation after the failure, typical of pressurized gas containers following a
blow up, makes it difficult to pinpoint the path for crack growth. The physical model in Fig. 6b allows defining fracture
initiation.
As seen in Fig. 7, ‘‘fracture 1” is approximately longitudinal to the collector, and continues longitudinal in the tube. This
‘‘fracture 2” becomes circumferential and arrests when coming near to the bolted jacket of the tube, after crossing a distance
of 150 mm. The other fracture in the collector is approximately circumferential. This ‘‘fracture 4” curves in the weld between
reinforcement and tube, and continues circumferential in the tube. This ‘‘fracture 3” seens to cross fracture 2 and continues
20 mm, until arresting in tube material. Fracture 1 from point B until the reinforcement-to-collector weld is mode one (ten-
sion), ductile with shear lips. When crossing the weld between reinforcement and collector, the fracture is brittle, with chev-
ron marks, see Fig. 8, and immediately becomes ductile again. Fracture 3 (sections B–E), took place in mode three (anti-plane
shear), with signs of tear by loads normal to the surface of the plate.
Fig. 9 shows in detail zone B of initiation. Fracture initiation occurred at severe defects in the weld between reinforcement
and pipe. Although somehow masked by corrosion, Fig. 9 shows a fish eye type of crack, named ‘a’ in the figure, which is
typical of a hydrogen crack (cold crack during welding). Several pores coalesced to form another small crack, named ‘b’ in
the figure. Fractographic analysis showed that these two cracks eventually coalesced to form a single surface crack, indicated
by the dotted line. This first propagation was brittle, coalescence occurred either during the preoperational hydrostatic pres-
sure test, or during the overpressure event.
To further elucidate this, a mechanical model of static stresses at the intersection due to pressure was carried out accord-
ing to the procedures of section 9 of API 579 [7]. Initial crack depth a is 5.18 mm, the depth of the lately evolved crack b is
6.2 mm (crack ‘‘a” coalesced with pores). In a tube of fine wall internal pressure creates zero radial stress and a hoop stress
according to the equation of Barlow. The presence of the 600 hole generates a stress concentration. The reinforcement reduces
the stress to levels similar to which would be without the hole, but previous pressurization of the reinforcement rises stres-
ses again, so that a stress concentration of 2 is considered. For an inner diameter D = 300 mm, a thickness of reinforcement
t = 7.2 mm, and an internal pressure of 6 MPa, the nominal circumferential stress is

rc ¼ 2  P  D=2t ¼ 2  120 MPa ¼ 240 MPa ð1Þ

In absence of effective post weld heat treatments, maximum residual stress is around half of yield strength of base mate-
rial (150 MPa) [8]. So maximum stress in the tube wall is

rc þ rres ¼ 240 þ 150 ¼ 390 MPa ð2Þ

1830 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836

Fig. 6. (a) Section 12 of the collector, fractured into two sections. (b) Fracture initiation.

For a semicircular crack in the pipe wall with a depth of 80% thickness, the applied stress intensity factor K is

K ¼ 1:2  390  ð3:1  0:006Þ1=2 ¼ K capl ¼ 73 MPa m11=2 ð3Þ

Taking into account the poor quality of manufacture, toughness of reinforcement material is likely low, conservative esti-
mation is

K IC ¼ 100 MPa m1=2 ð4Þ

Crack driving force to toughness ratio is
Kr ¼ K capl =K IC ¼ 0:73 ð5Þ
The applied stress to the section in the case of plastic collapse is mainly the Von Mises equivalent due to pressure. Using a
factor for thickness reduction of section of 1.5:
rc ¼ 1:5  240 MPa ¼ rVM ¼ 360 MPa ð6Þ
 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836 1831

Fig. 7. Details of the zone box in Fig. 6a.

Fig. 8. Brittle fracture with chevron marks in the collector material.

So the applied stress to flow stress ratio is

Sr ¼ rVM =rf ¼ 0:76 ð7Þ
The point (Sr = 0.76, Kr = 0.7) falls in the safe part of the level 2 FAD of API 579, although near the safety limit.
The propagation of the failure from the preexisting weld defects can therefore be reasonably explained, only considering
nominal internal pressure. Longitudinal stresses in the tube produced during assembly or during the previous overload could
have generated larger stresses. Operation data allows to explain why the failure was 4 h delayed after the overpressure. It
was found that the overload occurred in the late afternoon of a winter day, after dark, the low room temperature could have
reduced material toughness, leading to crack instability and final failure.
Some of the root causes related to weld defects can be seen in this Fig. 9, but these were not the only construction flaws.
The cross sections depicted in Fig. 10, for example, show numerous aspects of weld manufacture that do not match standard
quality requirements: large pores, cracks, undercuts, weld toes of different size, etc. The welds between tube and collector
and between tube and reinforcement would have to be rejected if evaluated under the requirements of, for example, API
1104, due to the presence of a great amount of large defects [9]:
1832 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836

Fig. 9. Details of Zone B of initiation.

 The hole in the collector and the tube end were flame cut.
 The tube-to-collector and tube-to-reinforcement welds show great amount of pores and large undercuts.
 The end of the tube does not cross the hole in the collector pipe, but tops its outer surface; this is an error in the design of
the joint.
 The fillet weld shows legs of different size, their sections in some cases are below the thickness of the welded elements.
 Arc strike marks were found throughout the surface of the reinforcement.
 The purge hole of the gap between collector and reinforcement was filled with a weld of very bad quality; the weld
detached during failure and remained at one side of the fracture.
 The inadequate welding sequence caused craters at weld start and stop regions, and cold cracks.

Fig. 10. Several cross sections of defective reinforcement.

 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836 1833

Fracture mechanics analysis shows that these initial defects, however gross, were stable during subsequent service, for
many years, until a small increase in pressure followed by cold weather re-activated propagation.

4. Suction vessels in natural gas alternating-compressor plants

Repeated cracking in plate reinforced welded flanges of several suction vessels in identical natural gas compressor
plants presumed design, construction or operation problems [10]. In two of such failures investigated, cracks in the
reinforcement saddle of the flange connections were detected before gas leak. Cracks initiate at the toe of the pipe
to saddle weld, see Fig. 11, somehow similar to those found in Case 1 (Fig. 2). The root cause analysis of the problem
included:

 Verification of design of the compressing facilities.

 As built drawings verified in field.
 Study of vibrations and pulsations [11].
 Failure analyses.
 Analysis of design assumptions of dampeners.
 Inspection procedure with indication of defect acceptance/rejection.

Evidences proved that, as expected, the cracking mechanism was mechanical fatigue, controlled by the cyclic variations of
stress in the welded joints, with origin in variations of internal pressure and in dynamic phenomena related to vibrations and
pulsations. The two analyzed failures have common characteristics, as depicted in Fig. 12: they initiated in the same zone in
both large bottles; in both cases cracks started off large defects in the welds; the quality of the welded joints was found
deficient.
In the outside, the reinforcement looked very well. Since according to ASME VIII Div. 1 these welded connections do
not require NDT inspection, embedded defects stayed undetected. Cross sections (Fig. 13a) revealed many weld defects,
in one case a weld bead over the entire surface of the reinforcement was made to comply with thickness specification
(Fig. 13b).
Fatigue propagation of cracks require an initiator and a driving force, these are in this case some gross weld defects (hard
spots, inclusions, and voids) and vibrations/pulsations of the system, respectively. A small crack in the outer surface of the
container would have a very short fatigue life, further propagation would occur not only in the reinforcement but also in the
vessel thickness, so immediate replacement of such flanges was required. Subsurface indications could be repaired by means
of the replacement of the reinforcement or removal of indications and repair weld. The final solution to this recurrent prob-
lem was to repair the defective joints, using material and procedures according to standard best practices, including 100%
inspection of installed large bottles, and to reduce the amplitude of vibrations in the bottles and associated pipes. This
required some re-engineering of the system, as per ANSI-API Std. 618 [11].
In order to detect any other manufacture defects, particularly those preexisting defects from which cracks could initiate,
non-destructive inspection of all potentially dangerous welds in all similar equipment was required. In all cases, critical

Fig. 11. Case 3: Crack initiation in the reinforcement saddle of the flange in suction vessel.
1834 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836

Fig. 12. Characteristics of the two analyzed failures in suction vessels.

regions for detailed inspections were pinpointed: those in the outer half of the reinforcement thickness, near the weld to the
tube.

Fig. 13. (a) Cross section, note slag inclusions and the size of the heat affected region. (b) Further details of the cross section of the reinforcement welds.
 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836 1835

5. Conclusions

Present rules in fabrication codes are aimed to reduce the probability of fatigue cracks in flanges welded to pressure ves-
sels subjected to cyclic pressure or vibrations. Yet, several leaks and ruptures have recently occurred at flanges in pressure
vessels and pipes. A review of some of these cases is presented, and their common root causes highlighted and discussed:

 Discharge vessels at an alternating-compressor in a petrochemical plant.

 1200 collector pipe in a natural gas turbine-compressor plant.
 Suction vessels in two natural gas alternating-compressor plants.

NDT inspections, fractographic and microstructural analyses, and mechanical and chemical tests are briefly discussed, in
the characterization of crack propagation into vessel material. Common root causes for the failures are investigated, and the
influences of manufacture and operation conditions on crack initiation and propagation mechanisms are highlighted.
Possible mitigation measures are discussed in each case, and common conclusions yielded. Fatigue propagation is, as
design expects, due to the large number of cyclic loads from gas pulsation and vibrations from the compressors. Some
conclusions allow increasing mechanical reliability and fitness for service of these flanges, especially before entering service.
The main aspect still is to assure an adequate dimensioning of the system to reduce the amplitude of cyclic stresses, but
there are also some manufacturing and inspection aspects to address.
Common aspects of these failures are crack initiation in heat affected zone of welds, all related to an inadequate design of
the reinforcement and of the welding procedure. Weld inadequacy is sometimes a result of large differences in thickness.
Increase in thickness as an attempt to increase safety is found to be detrimental, since larger than expected cyclic stresses
are sometimes due to excessive weight of the vessel. Improper foundation due to design errors or maintenance deficiencies
leads to higher than expected vibration amplitudes. The use of fillet welded reinforcements induces rigidity inconsistencies
within the flange joint, and concentrates deformations in the failed regions. In one case, it is believed that preoperational
hydrotest could have been detrimental, by coalescing individual small weld defects into a crack.
Most cracks initiate from the outer surface, so an adequate non-destructive inspection program during service helps pre-
vent failures. But other cracks initiate in the outer half of the thickness of the reinforcement, from very difficult to avoid weld
defects. Changes in the design of the reinforcements is required. Ultrasonic testing, with an adequate procedure, can be reli-
ably used to detect these defects.
Although some operation conditions have influenced the onset of crack growth, common root causes for these failures are
related to poor manufacturing practices. Flanges with welded reinforcements are not a mechanically very efficient design.
However, most relevant non-conformities are not related to inadequate design, but to bad workmanship and welding pro-
cedures. Gross weld defects were left in the welds between pipe and vessel. The reinforcement plate makes it impossible to
detect defects in the pipe to vessel weld, during post construction and in-service NDT testing. In line with this, construction
codes such as ASME PVBC section VIII Div. 1 do not specify a radiographic inspection.
Reinforcement plates are used due to reduced manufacturing costs, when compared to machined or forged transitions such
as weldolets. Reasonably good results are most often obtained with this reinforcement design due to some fortunate factors:

 Good workmanship and inspection during fabrication.

 Hard to inspect weld areas are close to mid – thickness, that is, close to the neutral plane of the section when subjected to
bending loads.
 Highest loads due to thermal transients and mechanical loads from the pipe and accessories fitted to it are of the bending
type, so that the region most severely weakened by the welding defects is not subjected to high stresses.

Owner specifications should aim to avoid manufacturing vices, especially in critical regions where fatigue conditions due
to vibrations and pulsations are foreseen, and where the consequences of a leak could be high. Inspection should play a key
role during manufacturing of these joints, since NDT inspection after construction would not reliably detect gross weld
defects.

Acknowledgements

This research work was funded by CONICET (Consejo Nacional de Investigaciones Cientıficas y Tecnicas) and SECyT, both
from Republica Argentina. Thanks are given to Dow Argentina, Petrobras, Gas Atacama and GIE S.A. (Argentina) for technical
and financial support and for permitting publication of proprietary information.

References

[1] Smith TA, Warwick RG. A survey of defects in pressure vessels in the UK for the period 1962–1978 and its relevance to nuclear primary circuits. Int J
Pres Ves Pip 1983;11:127–66.
[2] Davenport TJ. A further study of pressure vessel failures in the UK, international conference on reliability techniques and their application, Reliability
’91. London, UK; 1991.
[3] Hayes B. Six case histories of pressure vessel failures. Eng Fail Anal 1996;10:157–70.
1836 J.L. Otegui et al. / Engineering Failure Analysis 16 (2009) 1825–1836

[4] Report GIE 7101-04-04. Failure analysis of flange saddle in vessel V5707, LDPE plant. Dow Argentina; 2004.
[5] ASME Boiler and Pressure Vessel Code. Am Soc Mech Eng; 1999.
[6] Report GIE 7302-05-05. Análisis de falla de ducto de venteo en planta turbocompresora. Petrobras; 2005.
[7] API STD 579 - ASME FFS-1. Fitness for Service. USA; 2007.
[8] American Welding Society. Welding handbook. Miami; 1995.
[9] API 1104. Welding of pipelines and related facilities, 19th ed. USA; 1999.
[10] Report GIE 2201-11-05. Análisis de fallas de amortiguador de descarga en motocompresores de plantas compresoras. Gas Atacama SA; 2005.
[11] ANSI-API Std. 618. Reciprocating compressors for petroleum, chemical, and gas industry services, 5th ed. USA; 2007.