Engineering Failure Analysis 12 (2005) 376–386

www.elsevier.com/locate/engfailanal

Analysis of cracking in LPG Horton spherical vessel

R.K. Pandey *

Department of Applied Mechanics, Indian Institute of Technology, Hauz khas, New Delhi, 110 016, India

Received 6 May 2004; accepted 8 May 2004

Available online 28 October 2004

Abstract

Occurrence of cracking in the Horton spherical vessel made of a 537 class I steel used for containing liquefied petro-
leum gas has been analysed. The cracks were located using ultrasonic technique and their propagation behaviour was
studied through scanning electron microscope examinations. All the cracks were found to be of embedded type, con-
fined in the heat affected zone (HAZ) and located around mid-thickness region of the vessel. Tensile testing, hardness
measurement, Charpy test, bend test and microstructural studies were conducted to determine the causes for incidence
of cracking. The cracking is apparently hydrogen induced which has been assisted due to unfavourable micro-structure
of the HAZ, high residual stresses and an unsuitable composition of the alloy. The remedial action to avoid cracking
has been suggested.

2004 Elsevier Ltd. All rights reserved.

Keywords: Hydrogen-assisted cracking, Welds; Heat-affected zone; Pressure–vessel failures; Microstructures

1. Introduction

Cracking was reported in the weld region of Horton spherical vessel of 17.5 m diameter and 35 mm
thickness. The vessel was used for storing liquefied petroleum gas (LPG). The cracks were reported to
be embedded cracks located in the weld region. The vessel was made of a 537 Class I type structural steel
and a normalizing heat treatment had been given. The vessel was designed for a pressure of 1.5 kg/cm2 and
had been subjected to a hydrostatic pressure of 1.5 times the design pressure during proof stressing. The
cracks were noticed after five years of service while routine inspection of vessel was being conducted. It

*
Tel.:+91 011 26591751; fax: +91 011 26581119.
E-mail address: [email protected] (R.K. Pandey).

1350-6307/$ - see front matter
2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2004.05.006
 R.K. Pandey / Engineering Failure Analysis 12 (2005) 376–386 377

Table 1
Chemical composition and tensile properties of steel
Carbon Manganese Phosphorus Sulphur Silicon
Composition (wt%)
0.24 0.7–1.6 0.035 0.040 0.15–0.50
Yield strength (minimum) Tensile strength (MPa) Percentage elongation (in
(MPa) 50 mm gauge length)
Tensile properties
310 450–585 22

was reported that when repair welding was done after removing material from cracked region, the cracks
reappeared in the rewelded material.

2. Material and weldment

The composition and prescribed tensile properties of the vessel material, etc. i.e. 537 class I structural
steel is given in Table 1. The vessel was known to be welded by manual metal arc welding (MMAW) using
standard electrodes and double V groove type edge preparation had been employed.

3. NDE inspection of cracked vessel

To pinpoint exact location of cracks in the vessel a portion of the vessel was dissected and examined
using Ultrasonic technique (UT) by employing 70 angle probe with 2.5 MHz frequency. The probe
was placed on the inner surface of the dissected part of the vessel over the weld zone and was moved
along the weld direction covering total weld zone as well as nearby regions. The location of cracks is
shown in Fig. 1 in three parts i.e. AA–BB, BB–CC and CC–DD. Following observations were made
from Fig. 1.

(a) A number of cracks were noticed of varying length i.e. from 25 to 100 mm.
(b) All the cracks were found to be present within the weld zone where the weld zone had a width of about
26 mm.
(c) Majority of cracks were present within a depth of 15–20 mm from the inner surface of the vessel.
(d) Almost all the cracks were present near the weld zone boundary over a width of less than 10
mm.

The cross-section of weldment is shown in Fig. 2 as revealed by etching method. The cross-section
has a double V configuration where root of V is located at a depth of 22 mm from the surface of the
vessel. The width of weld zone reduces to 7 mm at the V-root as compared to 25 mm at the inner
surface. On the outer surface, the weld zone width was found to be of 21 mm. The typical width of
HAZ was found to be about 2.5 mm throughout the thickness. Based on UT observations the zone
of actual location of majority of cracks is also shown in Fig. 2. It may be noticed from the figure that
most of the cracks are located either in the heat-affected zone (HAZ) or in the base metal close to the
HAZ. If cracks were present at all in the weld zone, these were mostly concentrated near the weld zone
(WZ)-HAZ boundary.
378 R.K. Pandey / Engineering Failure Analysis 12 (2005) 376–386

Fig. 1. Approximate location of cracks observed by UT.

4. Microstructural observations

The cross-section of weldment was polished, etched and observed on a Scanning electron microscope
(SEM, Cambridge Stereoscan-360) at magnifications varying in the range of 550–1200·. The microstruc-
tures of base metal, weld metal and heat affected zone (HAZ) were observed and are presented in Fig. 3.
The microstructure of the base metal (Fig. 3(a)) exhibits typical ferrite–pearlite structure. The banding of
pearlite was evident along a direction perpendicular to the weld direction. The grain diameter of ferrite was
found to lie in the range of 10–12 lm. The HAZ microstructure is shown in Fig. 3(b) and (c). Microstruc-
ture consisted of equiaxed grains of ferrite and pearlite. In addition, some lath martensite type structure
could also be seen in the HAZ-WZ boundary region (Fig. 3(c)).
 R.K. Pandey / Engineering Failure Analysis 12 (2005) 376–386 379

Fig. 2. Idealized presentation of WZ, HAZ, base metal.

The microstructure of the weld zone is presented in Fig. 3(d)–(e). It consisted of elongated as well as
polygonal ferrite grains, pearlite and fine acicular ferrite in different zones. The carbides were found to
be distributed in the matrix.

5. Hardness distribution

The hardness distribution was obtained over the cross-section in the base metal, HAZ and the weld zone
(Fig. 4). The hardness was measured on Vickers Hardness tester using 5 kg load and the average hardness
values are reported as follows:

Base metal – 176

HAZ – 240
Weld metal – 200

6. Evaluation of mechanical properties

The following types of test specimens were machined from the supplied cracked section of the vessel.

(a) Tensile test specimen from unwelded region.

(b) Charpy V notch specimen from unwelded region.
(c) Tensile test specimen with weld zone located in the center of the gauge length.
(d) Charpy V notch specimen from the weld zone.
(e) Charpy V notch specimen from the HAZ.
(f) Bend specimens.

Specimens were machined after carefully isolating the cracking zones so that no specimen did contain
cracks. The longitudinal axis of the tensile specimen and the Charpy specimens was kept transverse to
380 R.K. Pandey / Engineering Failure Analysis 12 (2005) 376–386

Fig. 3. Typical SEM microstructures: (a) Base metal; (b) HAZ; (c) HAZ; (d) Weld metal; (e) Weld metal.

the weld direction. All weld tensile specimens could not be machined due to shortage of material consisting
good weld region. Therefore, only transverse weld type specimens were machined. All the tensile specimens
were of round bar type with gauge length = 50 mm and diameter = 7.0 mm. The gripping diameter was
about 10 mm and total length = 200 mm. Charpy specimens employed were provided with
 R.K. Pandey / Engineering Failure Analysis 12 (2005) 376–386 381

Fig. 4. Hardness distribution in the joint.

10 mm · 10 mm cross-section and 55 mm length. The bend specimens were machined with 10 mm thick-
ness, 37.5 mm width and 150 mm length and had an orientation same as the other specimens.

6.1. Tensile test results

The tensile test was conducted on a 25 ton MTS machine in displacement control mode and about 3–4
tensile specimens were tested in each case. The average tensile properties are reported in Table 2. It was
noticed that specimen containing the weld region did not really break in the weld zone. Rather all the spec-
imens failed in the base metal indicating higher tensile strength of the weld metal as compared to the base
metal.

6.2. Charpy test results

The Charpy test was conducted at ambient temperature (
25 C) employing 4–5 specimens in each case.
The results of Charpy energy for base metal, weld metal and HAZ are presented in Table 3.

Table 2
Tensile properties of base metal and welded part of the vessel
Specimen condition Yield strength (MPa) Tensile strength (MPa) Percentage elongation Percentage reduction in area
Base metal 365 548 26 66.5
Base metal-weld combine 378 560 17 62
382 R.K. Pandey / Engineering Failure Analysis 12 (2005) 376–386

Table 3
Charpy V-notch energy of the vessel weldment
Condition Charpy energy (kg m) Average energy
(kg m) (ft lb)
Base metal 19.2 18.95 137
19.4
19.0
18.2
Weld metal 17.0 14.75 107
14.0
13.6
14.4
HAZ 11.8 11.80 85
11.4
12.5
11.6

6.3. Bend test results

Using the standard setup, the transverse bend test of the weld was conducted in two conditions: (a)
root bend condition and (b) face bend condition. Under the face bend condition, the weldment re-
vealed satisfactory performance and no cracks were noticed in the bend region even after 180 bending.
On the other hand cracks and fissures lying between 1 and 3 mm sizes were noticed in the weld zone
on root bend test specimen after bending. The percentage elongation was found to lie in the range of
9–11%.

7. Study of crack configuration on polished section

The cracked piece from the Horton sphere was carefully sectioned at the appropriate depth from the in-
ner surface based on the information available from the ultrasonic test. The surface was polished, and
examined using Magnaflux method to detect the surface cracks, if any. This was continued in steps till
the embedded crack got exposed to the surface. The schematic view of cracks observed after sectioning
at 20 mm depth from the inner surface is shown in Fig. 4. The cracks were found to be lying fully in the
HAZ.
The cracked surface was examined under the SEM at low magnification i.e. 20–50· and the crack tra-
jectory obtained is shown in Fig. 5. A weld defect (incomplete fusion) lies at the WZ-HAZ boundary
and the main propagating crack is contiguous at both the ends of this defect. The cracks are found to
be quite zig-zag in nature and all the time traveling wholly through the HAZ.

8. Microscopic observation of crack trajectory

The polished sections were etched and examined under the SEM to reveal the microstructure
through the trajectory of cracks as shown in Fig. 6. Fig. 6(a) shows the propagating crack through
the fine grained ferrite structure of the HAZ. The crack propagation is transgranular through the
 R.K. Pandey / Engineering Failure Analysis 12 (2005) 376–386 383

Fig. 5. A view of the crack trajectory on a polished section.

HAZ. The main propagating crack is found to produce number of branches (i.e. secondary cracks as
evident from Fig. 6(c)–(e)), which give rise to delta formation at times. Some intergranular cracking
may also be seen giving rise to complete failure of grains leading to widening of crack. Formation
of sister cracks parallel to the main crack was also noticed in some cases (Fig. 6(f)). Some irregular
zig-zag nature of crack branching were also seen which on further magnification revealed fine scale
intergranular cracks. No detrimental precipitation was noticed along the grain boundaries when the
structure in the vicinity of cracks was examined under EDAX.

9. Discussion

9.1. Mechanical properties of the cracked vessel

From the results of tensile test as shown in Table 2, it is noted that the yield strength for all the two types
of test specimens (i.e. base metal, and the specimen containing weld metal) lies in the range of 365–378
MPa. The tensile strength for the above two types of specimens is found to lie between 548 and 560
MPa. The prescribed yield and tensile strength of the vessel are 310 MPa (minimum) and 450–585 MPa,
respectively. Thus the strength properties of the fabricated vessel are within the prescribed range. Since,
no failure of tensile specimen was noticed in the weld region, the weld metal is considered to possess a
higher tensile strength than the base metal. The ductility of the weld tensile specimen on the other hand,
is significantly lower than that of the base metal as indicated by percentage elongation values in Table 2.
The base metal has an elongation of 26% as against weld specimens possessing elongation of 17%.
The Charpy toughness results as presented in Table 3 provide the Charpy energy values for the base me-
tal, weld metal and the HAZ as about 190, 148 and 118 J, respectively. This shows that the Charpy tough-
ness of the weld metal is decreased by 22% and that of the HAZ by 38% with respect to the base metal
toughness. On the other hand, the hardness of the HAZ is the highest amongst the three i.e. 240 VHN
against the base metal and the weld metal which possess corresponding values as 176 and 200 VPN. Thus
the highest hardness combined with the lowest Charpy toughness of the HAZ makes it more vulnerable to
cracking as compared to the weld metal and the base metal.

9.2. Microstructural analysis of the cracking region

As detailed in the earlier sections, the microstructure of the base metal consists of ferrite–pearlite struc-
ture with banding of pearlite in rolling direction. The HAZ possesses ferrite–pearlite structure however, lath
martensite type structure is also noticed. The weld metal is marked by mixed type of structure consisting of
384 R.K. Pandey / Engineering Failure Analysis 12 (2005) 376–386

Fig. 6. (a)–(f) Crack propagation paths through the microstructural features.

grain boundary elongated ferrite, acicular ferrite, polygonal ferrite and carbides The higher hardness of the
HAZ is apparently the result of its microstructure consisting of fine grained ferrite and pearlite and also
some lath martensite. This has the effect of decreasing ductility and toughness of the HAZ significantly
as compared to the base metal and the weld metal.
 R.K. Pandey / Engineering Failure Analysis 12 (2005) 376–386 385

It is interesting to note that the cracks in the vessel are mostly confined in the HAZ of the weldment
and appear to develop on planes closer to the root region of the weldment. The tendency of cracking is
obviously favored by higher hardness, lower ductility and reduced Charpy toughness of the HAZ. Also
the cracks are found to be zig-zag and discontinuous in nature. Both the transgranular and intergranular
type of cracks are noticed. In addition, significant amount of branching and secondary cracking are also
found.

9.3. Properties of weldment versus cracking behaviour

An attempt was made to assess the characteristics of the weld joint at this stage. A cracking parameter,
PNB for the HAZ is computed as follows [1]:

PNB ¼ C þ Si=20 þ Mn=10 ¼ 0:24 þ 0:5=20 þ 1:6=10 ¼ 0:425

The cracks normally do not occur in the HAZ if the factor PNB is lesser than 0.25. A higher value of
PNB in the HAZ of the vessel may result in cold cracking in presence of higher hydrogen concentration
(i.e. H cracking). The carbon equivalent, Ceq for the weld joint was further calculated to ascertain the weld-
ability of the steel. The Ceq is given as [2]
C eq ¼ C þ Mn=6 þ ðCr þ Mo þ VÞ=5 ¼ 0:51
The recommended limiting value of Ceq for good weldability is 0.40. A higher value of carbon equivalent
would promote cold cracking i.e. H-induced cracking or martensite cracking in the HAZ.
The thickness of the vessel is fairly large i.e. 35 mm. This is likely to generate higher level of thermal
residual stresses during welding. If no proper heat treatment is carried out like preheating or post-weld
treatment, the tensile residual stresses would further aggravate the tendency for cold cracking i.e. H-crack-
ing of the HAZ [1]. It is interesting to note that cracks were mostly noticed in near mid thickness planes
where high constraint is likely to be present.

9.4. Overall causes of failure

At this stage an attempt was made to compute the stress level to which the vessel was subjected during
operation or during hydrostatic pressurization. Taking pressure P = 1.5 kg/cm2, the hydrostatic pressure
which is 1.5 times the operational pressure becomes = 2.25 kg/cm2 = 0.225 MPa.
The hoop stress rh for the vessel = PD/2t where, P = pressure, D = vessel diameter and t = vessel
thickness.
Therefore,
rh ¼ ð0:225  17:5Þ=2  35  103 ¼ 56 MPa:
Since the yield strength of vessel = 365 MPa, the safety factor on yielding is more than six. Therefore, the
vessel is elastically stressed only even in the presence of some mild stress concentrators.
From the crack trajectory vis a vis microstructural observations it is noticed that the nature of
cracking is intergranular as well as transgranular along with significant extent of crack branching.
The cracks are very fine also. The intergranular cracks are presumably located along original austenitic
grain boundaries. From the observations made in the earlier section and the nature of cracking, it is
sufficiently clear that the cracking which is confined mostly in the HAZ of welded vessel is typically
a case of hydrogen cracking. The grain boundaries of the austenite are the potential sites for trapping
of hydrogen. Thus the formation of H-induced cracking along the grain boundaries of austenite is very
386 R.K. Pandey / Engineering Failure Analysis 12 (2005) 376–386

much favored [3]. The increased H-concentration has also led to decrease in ductility and Charpy
toughness of the HAZ.
The banded pearlite structure is not considered to be congenial as this leads to microsegaration of Mn
and C which in turn increases the carbon equivalent and the cracking parameter value [4]. This may give
rise to undesirable microstructure like martensite which has been found indeed at some places. The pearlite
bands are also believed to act as trapping sites for hydrogen and enhance the tendency for hydrogen crack-
ing. In fact the cracks have been noticed to move preferentially through such zones. The weld metal-HAZ
boundary defects do serve as H-trapping sites and accentuate the tendency of cracking.

10. Conclusions

The following conclusions are drawn based on the present study:

1. The cracking of LPG horton sphere is confined in the HAZ of weld joint in the near mid thickness region
i.e. within 15–20 mm depth from the inner surface.
2. The harder microstructure of the HAZ, the high weld residual stresses of the thick vessel (
35 mm) and a
banded ferrite–pearlite structure have apparently decreased HAZ toughness and increased its hardness
which have favored the hydrogen cracking of the HAZ.
3. The higher value of cracking parameter (=0.425) for the HAZ and a higher carbon equivalent (=0.51),
have also enhanced the tendency for cracking.
4. The cracking of the vessel could be avoided by proper precaution during welding and control of welding
parameters to minimize entry of hydrogen into the weld metal, pre-heating and post-weld heat treatment
and possibly choosing a low C, micro-alloyed steel.

References

[1] Easterling K. Introduction to the physical metallurgy of welding. 2nd ed. Oxford: Butterworths-Heinmann; 1992.
[2] Lancaster JF. Metallurgy of welding. 3rd ed. George Alien and Unwin; 1980.
[3] Raghavan V. Physical metallurgy. New Delhi: Prentice-Hall; 1997.
[4] Kumar AN, Pandey RK. Phenomenon of crack branching. Trans. Indian Inst. Metals 1983;436–441.