Engineering Failure Analysis 14 (2007) 1093–1102

www.elsevier.com/locate/engfailanal

Analysis and redesign of failed jacket in a petrochemical reactor

a
Failure Analist, GIE S.A. Camusso 30, Mar del Plata, Argentina
b
Department of Materials Engineering, University of Mar del Plata, J.B. Justo 4302, Mar del Plata, Buenos Aires, Argentina

Received 12 October 2006; accepted 29 November 2006

Abstract

The root cause of repeated water leaks due to cracking in a circumferential weld of the heating jacket of a petrochemical
reactor was traced to cyclic steam pressure, not accounted for in the design. Fractographic and metallographic analyses of
the jacket base and weld materials revealed widely open cracks, initiation and propagation being due to larger than allow-
able cyclic stresses. There were no indications of environment induced cracking, neither creep nor stress corrosion crack-
ing. Appendix 2 of API RP 579 was used to verify alternative designs for the substitution of the heating jacket. Due to time
restraints, the substituted part had to be built with a lower strength steel. Reconstruction was successfully done using an
austenitic SA 316L stainless steel.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Petrochemical reactor; Failure analysis; Redesign; Modelling

1. Introduction

The original scope of the failure analyses presented in this paper is related to repeated water leaks due to
cracking in the upper circumferential weld of the heating jacket of a petrochemical reactor. Fig. 1 shows an
overview of the hemispherical lower part of the 2 m diameter reactor. Fig. 2 shows a sketch of the configura-
tion of the reactor. The hemispherical lower part is made of a clad structural steel, inside which the catalyst is
located. To maintain constant temperature, water flows within the outer jacket. Several inlet and outlet pipes
get across both reactor and jacket walls.
The failed jacket is made of a duplex stainless steel, and used to control the temperature in the reactor by
means of hot water (sometimes mixed with steam). Table 1 shows chemical composition of original jacket
material. The leakages had not been a serious problem in the past, the leaking circumferential weld was repeat-
edly repaired by welding, see Fig. 3.
The objective for the company was to develop a repair method that would ensure an extended four-year
service live before the following plant stop. The scope of this study was therefore the redesign of the failed

*
Corresponding author. Tel./fax: +54 223 4748300.
E-mail address: jotegui@fi.mdp.edu.ar (J.L. Otegui).

1350-6307/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2006.11.058
1094 A. Marquez et al. / Engineering Failure Analysis 14 (2007) 1093–1102

Fig. 1. Overview of the hemispherical lower part of the 2 m diameter reactor.

R1
01
2,5

6.3
953

20 (16.8+3.2)

Fig. 2. Configuration of the reactor.

Table 1
Chemical composition of original jacket Duplex material
Material Fe Al Si Cr Ni Mo
Base metal 66.9 24.2 4.4 4.5
Weld root 66.4 0.7 20.6 9.5 2.8
Weld reinforce 64.5 0.6 0.8 20.5 10.2 3.2

section and its repair. To do this, it was decided to assess the repair according to the procedures by API RP
579, Fitness for Service [1]. What makes this analysis interesting is its operative constrain. At the time of the
analysis the operator was about to begin the plant stop and the repair of this jacket had not been foreseen with
sufficient time in advance. This lead to the following additional constraints:

1. The redesign and reconstruction of the failed part of the jacket should ensure a lower stress state and an
extended life.
 A. Marquez et al. / Engineering Failure Analysis 14 (2007) 1093–1102 1095

Fig. 3. The leaking circumferential weld was repeatedly repaired by welding.

2. In order to avoid major repair work in inlets and outlets, the repair would only include removal of the
upper part of the jacket.
3. No duplex steel would be locally available at the time of reconstruction, only an austenitic SA 316L stain-
less, which has a much lower strength.
4. The repair procedure should not interrupt the ongoing maintenance jobs.

2. Failure analysis

Nital and Picral etchings were used to reveal the duplex ferrite-austenite microstructures of the failed jacket.
Fig. 4 shows the section of the original weld; which joins the upper cone or skirt to the jacket plate. All

Fig. 4. Section of the original weld joining the upper cone skirt to the jacket plate.
1096 A. Marquez et al. / Engineering Failure Analysis 14 (2007) 1093–1102

Fig. 5. (a, b, ·5) Polished and etched cross sections showing cracks widely open with large plastic deformations.

Fig. 6. (·50) Secondary cracks parallel to the main crack of Fig. 4b.

Fig. 7. (·200) Duplex weld, HAZ and base metal microstructures are normal for this type of material.

investigated cracks are widely open, indicative of large plastic deformations during crack growth. The polished
and etched cross sections of Fig. 5 (a, b, ·5) show typical cracks found in the weld between the thin jacket
plate (to the right) with the thicker skirt. Fig. 5a shows a crack initiating at a large porosity defect at the root
 A. Marquez et al. / Engineering Failure Analysis 14 (2007) 1093–1102 1097

of the weld. Fig. 5b shows a crack that lead to a leak, its initiation site is at the stress concentration at the weld
root. Cracks initiated also in previously repaired parts of the weld.
Vickers Micro hardness testing was carried out on these cross sections. No clear variations were found,
however, minimum hardness values are in weld and Heat Affected Zone (HAZ) materials. Secondary cracks
parallel to the main crack of Fig. 5b reveal a high susceptibility of the material to microstructurally controlled
cracking; see Fig. 6 (·50). The columnar weld microstructure is seen on the top left of Fig. 7 (·200), the two-
phase base metal is seen on the lower right. In the HAZ the duplex structure is heavily distorted, due to the
heat treatment induced by the weld passes. All microstructures are normal for this type of material.
Fractographic and metallographic analyses reveal that all cracks initiate from geometrical discontinuities at
the weld joint. The apparent reason for the initiation and propagation of these cracks is a large traction stress
in the longitudinal direction of the vessel. No indications of intergranular crack propagation were found, that
could be related with environment induced cracking, such as thermal (creep) or chemical (stress corrosion
cracking).
The larger than expected stresses were traced to a service condition: steam pressure had been used for many
years for heating the water within the jacket. This apparently created pressure cycles inside the jacket, which
was designed according to ASME Boiler and Pressure Vessel Code, Section VIII. Division 1 [2]. Presently,
ASME BPVC Sec. VIII Div. 1 does not list any rules for fatigue evaluation of component. When fatigue eval-
uation of a component is required in accordance with UG-22 or U-2(g) of ASME BPVC VIII-1, the general
practice is to use ASME BPVC Sec. VIII Div. 2 fatigue criteria as a guidance up to the temperature limits of
VIII-2 [3].
These cyclic overloads, however, could not be eliminated, since the heating system could not be replaced
before the initiation of the next four-year service period. This adds an additional constraint to the redesign.

3. Redesign of jacket

In order to verify the mechanical strength and capabilities of new alternative designs for the substitution of
the heating jacket, and to eliminate the causes of these recurrent failures, numerical models were carried out
using three-dimensional continuum elements. A linear elastic behaviour was assumed, pseudo elastic finite ele-
ment models were carried out using the ALGOR FEM software [4]. Fig. 8 shows the axisymmetric model of
the jacket. Daily temperature and pressure fluctuations due to the batch process were considered. Typical
amplitudes considered in designing the jacket repair are: pressure 8.5 bar (120 psi), temperature 150 C. Pres-
sure frequencies are in the order of one cycle per day.
Stress allowances were verified in accordance with the specifications of ASME BPVC Sec. VIII. Div. 2 for
new vessels [5], which are similar to those by Appendix B (Stress Analysis Overview for a Fitness for Service
Assessment) of API RP 579. This document is applicable when assessing repairs and re-rating of existing

Fig. 8. Axisymmetric model of the jacket.

a b
Stresses in
Dist. [mm] vertical
7 direction
Stress
Classification
Line (SCL) 6
Y FEM
5
membrane
4

3 bending+
membrane
2

0
-400 -200 0 200 400 600
Stress [Mpa]

Fig. 9. (a, b) linearization of vertical stresses along the SCL in the failed weld.

Fig. 10. Originally proposed redesign.

Table 2
Ratios between applied stresses and allowances for the originally proposed redesign (Fig. 10)
API 579 App. 2 stress ratios APIMembrane Pm APIBending PL + Pm Peak PL + Pm + Q
Section MPa Ratio MPa Ratio MPa Ratio
Weld 1 60 0.6 90 0.6 220 0.7
Weld 2 70 0.7 220 1.5 300 0.9
Weld 3 35 0.3 170 1.1 310 1.0
Base metal 160 1.6 170 1.1 290 0.9

pressure vessels. Unlike ASME BPVC, API 579 gives the choice of using Tresca or Von Mises stress equiva-
lents in assessing allowable stresses. Both documents require the distribution of local membrane (Pm), bending
(Pl) and peak stresses (Q) along a stress classification line (SCL). A SCL should therefore be defined across the
thickness in each of the potentially most critical regions of the component. Membrane stresses Pm are then
compared with the stress allowance for the material, (Pl + Pm) is compared with 150% of the stress allowance,
and (Pl + Pm + Q) is compared with 300% of the stress allowance.
As an example, Fig. 9 shows the linearization of vertical stresses along the SCL in the failed weld in its
original configuration. Fig. 9a shows the model, Fig. 9b shows the distribution along the SCL of vertical nor-
mal stresses at the failed weld, which are the largest in this region. Not surprisingly, stresses are above allow-
able for the original duplex steel.
As mentioned before, due to time restraints the substituted part had to be built with a lower strength,
austenitic SA 316L steel plate, which has an allowable stress of 101 MPa. The original duplex steel 2205
has an allowable stress of 150 MPa. A Nickel base weld material was specified, match as much as possible
the mechanical and chemical resistance of the original duplex steel.
Fig. 10 shows the originally proposed redesign. Table 2 shows the ratios between the applied stresses and
allowances for this geometry, the reinforcing weld between jacket and vessel was the most critical point. Ratios
are above 1.5 in two regions; therefore this proposed redesign does not work. New designs were devised, based
on the analysis of the origin of the vertical tensile stresses in the critical regions. These stresses are due to the
sum of dead weight and bending moments due to the cyclic pressure within the jacket. The first approach (M1)
was to avoid removing the cone skirt, made of the strong duplex steel. This way maximum stress ratios are
1.25, still not acceptable.
It was noticed that the upper region of the jacket is subjected to bending moments in opposite directions.
There is a clockwise bending near the insertion of the jacket in the reactor, due to its weight, and a counter
clockwise bending at a lower region, due to pressure. Noting that dead weight is a major contributor to the
stresses, it is possible to reduce bending stresses by reducing the angle formed by the jacket in its insertion to

Fig. 11. Design M2 optimizes the geometry of the toroid to be replaced.

Table 3
Maximum stress ratios for configuration M2 are 1.0 (Fig. 11), acceptable in all sections
APIAPI 579 App. 2 stress ratios APIMembrane Pm APIBending PL + Pm Peak PL + Pm + Q
Section MPa Ratio MPa Ratio MPa Ratio
Weld 1 90 0.9 130 0.9 300 1.0
Weld 2 70 0.7 140 0.9 180 0.6

the reactor wall. On the other hand, it is possible to relocate the weld between repair and jacket far from the
highly bent region.
Design M2 resulted after several exchanges with the committing firm in order to optimize strength and ease
of repair, and is seen in Fig. 11. This design optimizes the geometry of the toroid to be replaced. It calls for an
8 mm thick, constant radius toroid, inserted to the reactor wall at a 30 angle. Table 3 shows that maximum
stress ratios for this configuration are 1.0, acceptable in all sections.

4. Repair procedure and results

The geometry for the repair defined in Fig. 9 allowed achieving conformity with stress allowances. The
failed jacket was cut to the dimensions defined by the model. Reconstruction of the jacket was successfully
done using a preformed toroid with a radius of 200 mm, made with an 8 mm thick SA 316L steel plate.
The toroid was then cut in 8 sections, which were welded to the upper part of the jacket.
NDT inspections were carried out on the outer surface of the reactor, before welding the toroid to the reac-
tor wall. Multiple shallow cracks were found on the outer surface of the hemispherical part of the reactor,
these cracks were concentrated around the original jacket to reactor weld. Ultrasonic inspection around all
inlets from the inner surface of the reactor allowed concluding that this cracking was only present in the region
of the repair. The position and appearance of the cracks made it possible to speculate a mechanism of corro-
sion fatigue, possibly due to the presence of steam in the upper part of the jacket at some time during the heat-
ing cycles. These cracks were ground up to total elimination; stresses in the ground conditions were checked.

Fig. 12. Model of ground outer surface of reactor, after cracks were found near the jacket weld during repair.
 A. Marquez et al. / Engineering Failure Analysis 14 (2007) 1093–1102 1101

In order to ensure that ground areas stay away from future chemical degradation, the reactor to jacket weld
was placed at a lower region, see Fig. 12.
Once the repair was completed, the reactor was placed into operation. After the rush, there was time for
final calculations and the elaboration of recommendations to ensure reliable long-time operation. It is cleat
that the origin of the cracking problem in the jacket is an inadequate design for cyclic operation. The cyclic
pressures generated by the heating method induce cyclic pressures and steam formation within the water
jacket. A redesign of the heating method was recommended, to avoid possible corrosion fatigue and caustic
attack degradation during future service, both in jacket and reactor materials.
The amplitudes of equivalent stresses that consider thermal and peak stresses were assessed from pressure
and temperature diagrams, representative of the current and future operation of the reactor. Stress histograms
at the most relevant sections were this way constructed, and fatigue life assessed. Equivalent stresses for each
point in the histogram were calculated in each assessment location. With this information, a load amplitude
histogram was created, which was grouped in ranges of effective stresses using the rainflow method [6].
Then the stress tensor was computed at the beginning and the end of the kth cycle in the previous histo-
gram, for each stress range. Using these data the stress amplitudes, that is, the difference between the stress
components at initial and final points of each cycle, were defined. These values are designated. Drki;j . Then
the Tresca effective intensity of stresses was computed, for the range of each particular cycle:
DS krange ¼ max½jDr1  Dr2j; jDr2  Dr3j; jDr3  Dr1j:
The effective cyclic stress intensity Salt was determined for each cycle (k-th, for example), with the equation
S kalt ¼ 0:5K ke DS krange
Ke is the knock-down factor (Table B.4, API RP 579).
The acceptable number of cycles for the stress intensity Sk was then determined, for all the ranges of the
stress histogram, using the curves of appendix F of API RP 579. Alternatively, those of appendix 5 of ASME
BPVC Sec. VIII div. 2 could also be applied [7]. Finally, the fatigue damage for each load block was deter-
mined, and damages from all blocks added using the Miner approach [8,9]. This calculation allowed conclud-
ing that it is not possible to ensure that the jacket would remain reliable during the whole service life of 4 years.
Since reconstruction of the jacket was done using an austenitic SA 316L stainless steel, which has also
poorer corrosion properties than the original duplex material, it will also be necessary to carry out some
in-service NDT inspections, to assess possible fatigue, corrosion fatigue or SCC damage in future operation.

5. Conclusions

This paper presents the results of a study done during the repair of the cracked heating jacket of a petro-
chemical reactor. The failed jacket was originally made of a duplex stainless steel, and is used to control the
temperature in the reactor by means of hot water (sometimes mixed with steam). In the original jacket con-
figuration, cracks and leaks were repeatedly reported in the circumferential weld between jacket and vessel.
The main reason for the initiation and propagation of these cracks was cyclic steam pressure.
API RP 579 was successfully used during the design the repair, to verify the mechanical strength and capa-
bilities of alternative designs. Allowable stresses were achieved in all base metal and welded sections, in accor-
dance with ASME BPVC Sec. VIII. Div. 2 and Appendix 2 of API RP 579, even when, due to time restraints,
the substituted part had to be built with a steel of lower strength than the original material.
Calculated fatigue life of repaired jacket did not allow ensuring a reliable life of 4 years, as required. Cor-
rosion fatigue is the most probable degradation mechanism in future service, unless the heating method is
replaced and the source of cyclic pressures within the jacket eliminated.

References

[1] Recommended Practice API RP 579. Fitness for Service. 1st ed. USA: American Petroleum Institute; Jan. 2000.
[2] ASME Boiler and Pressure Vessel Code. Section VIII. Division 1. USA: American Society of Mechanical Engineers; 1999.
[3] Farr JR, Jalad MH. Guidebook for the design of ASME section VIII pressure vessels. 2nd ed. USA: ASME Press; 2001.
1102 A. Marquez et al. / Engineering Failure Analysis 14 (2007) 1093–1102

[4] ALGOR Finite Element Analysis and Event Simulation Software, USA,1994- update; 1997.
[5] ASME Boiler and Pressure Vessel Code, Section VIII. Division 2, Appendix 4 Design Based on Stress Analysis. USA: American
Society of Mechanical Engineers; 1999.
[6] ASTM E 1049-85 (reapproved 1990), Standard Practices for Cycle Counting in Fatigue Analysis. USA: American Society for Testing
and Materials; 1998.
[7] ASME Boiler and Pressure Vessel Code, Section VIII. Division 2, Appendix 5 Design Based on Fatigue Analysis. USA: American
Society of Mechanical Engineers; 1999.
[8] Anderson TL. Fracture mechanics, fundamentals and applications. 2nd ed. USA: CRC Press; 1998.
[9] Harvey JF. Theory and design of pressure vessels. Reinhold: Van Nostrand; 1991.