Engineering Failure Analysis xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Engineering Failure Analysis

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

Effect of severe operation conditions on the degradation state

of radiant coils in pyrolysis furnaces
M. Santos a,b, M. Guedes c,e, R. Baptista c,f, V. Infante d,f,⇑, R.A. Cláudio c,f
a
REPSOL POLÍMEROS, Sines, Portugal
b
Setúbal School of Technology, Instituto Politécnico de Setúbal, 2910-761 Setúbal, Portugal
c
Department of Mechanical Engineering, Setúbal School of Technology, Instituto Politécnico de Setúbal, 2910-761 Setúbal, Portugal
d
Department of Mechanical Engineering, Instituto Superior Técnico, University of Lisbon, Av. Rovisco Pais, 1049-001 Lisboa, Portugal
e
CeFEMA, Instituto Superior Técnico, University of Lisbon, Av. Rovisco Pais, 1049-001 Lisboa, Portugal
f
LAETA, IDMEC, Instituto Superior Técnico, University of Lisbon, Av. Rovisco Pais, 1049-001 Lisboa, Portugal

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

Article history: This work arose from the need to understand the degradation phenomena taking place in
Received 30 September 2014 the radiant coils of pyrolysis furnaces of the petrochemical plant of REPSOL POLÍMEROS in
Received in revised form 25 February 2015 Sines, Portugal, in order to extend their service lifetime.
Accepted 27 February 2015
The radiation coils were studied through the finite element method, rendering stress and
Available online xxxx
displacement analysis and calculation of the Larson–Miller parameter in severe operation
conditions. The studied operation parameters were temperature increase, pressure
Keywords:
increase, net coil weight increase through coking. Additionally, microstructural analysis
Carburization
Creep
of samples of H39WM and H46M alloys in as-supplied and used conditions was carried out.
Failure analysis Attained results suggest that the operating lifetime of similar coils can be extended by
Radiant coils tuning their temperature profile and especially the amount of coke weight loaded, both
Pyrolysis furnaces during current operation and on decoking. Frequent and standardized temperature read-
ings are mandatory to assure a temperature profile below project temperature, determin-
ing the coke weight loaded at a given time would also allow a more accurate knowledge of
the remaining furnace creep life.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction

In ethylene production the hydrocarbon feeds are thermally cracked in the presence of steam at low pressure and process
temperatures of 788–843 °C. While the shield section, the lower convection section, the outlet transfer line and the quench
unit of the ethylene cracking furnace operate at relatively low temperatures, the radiant section of some of the furnaces
operate at end-of-run tube metal temperatures up to 1150 °C. This is in practice the upper temperature limit for available
heat resistant alloys [1]. Steam cracking furnaces thus provide one of the most aggressive settings to which alloys can be
exposed, combining high temperatures with a very aggressive chemical environment, which includes oxidizing and nitriding
flue gases at the outside, carburizing atmosphere at the tube inner surface, and severe start/stop and decoking cycles [2,3]. As
a result of oxidation and nitriding the material outer surface becomes glazy and spalls from the tube wall, decreasing tube
wall thickness [4]. Carburization results in a carbon layer build up inside the hotter sections of the coils during operation,

⇑ Corresponding author at: Department of Mechanical Engineering, Instituto Superior Técnico, University of Lisbon, Av. Rovisco Pais, 1049-001 Lisboa,
Portugal. Tel.: +351 21 8417643; fax: +351 21 8417915.
E-mail address: virginia.infante@tecnico.ulisboa.pt (V. Infante).

http://dx.doi.org/10.1016/j.engfailanal.2015.02.010
1350-6307/Ó 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Santos M et al. Effect of severe operation conditions on the degradation state of radiant coils in pyrolysis
furnaces. Eng Fail Anal (2015), http://dx.doi.org/10.1016/j.engfailanal.2015.02.010
2 M. Santos et al. / Engineering Failure Analysis xxx (2015) xxx–xxx

increasing tube net weight and promoting carbon diffusion into the alloy matrix with increasing time and temperature
[3,5,6].
Carburization is a primary cause of ethylene furnace tube replacement [7], wherein most cases there is a combination of
factors leading to coils failure, e.g. carburization and creep ductility exhaustion [2,8–10]. This results in bulging, bending and
ovalization of the tubes. Also, brittle fracture during furnace trips can result in large, longitudinal cracks on many tubes in the
furnace. Creep ductility exhaustion is determined by the number of cycles (start/stop and decoking cycles) and the nature
and severity of those cycles.
Radiant coils of pyrolysis furnaces are usually designed for a normal life of 100,000 h (11.4 years) of service at an operat-
ing temperature of 900 °C [7,11–15]. Their actual service life, however, varies from 30,000 to 180,000 h, depending on service
conditions and on the quality of materials [16]. After 10,000 h in service, the tubes crack length increases and their weldabil-
ity decreases. Also, creep resistance and ductility are reduced and the material becomes more prone to fracture, particularly
during thermal cycle.
Pyrolysis tubes failure can be prevented by a combination of proper furnace operation, materials selection, regular inspec-
tions and appropriate design [2,8–10,17–19]. Each pyrolysis plant performs with different and specific operating conditions
and philosophies; hence each plant has characteristic causes for radiant coil failure [2]. The analysis and understanding of
those typical mechanisms is thus of great significance, allowing the operators to carry out the most adequate selection of
tube materials and furnace operation parameters.
In this context, radiation coils from the petrochemical plant of REPSOL POLÍMEROS in Sines (Portugal) was studied
through the finite element method. Stress and displacement analysis and Larson–Miller parameter calculation were accom-
plished in severe operation conditions: increase of temperature, coke weight and gaps in the guides. Individual and compara-
tive characterization study of samples of H39WM and H46M alloys in as-supplied and used tubes was carried out. Attained
results allowed to establish a relationship between the operating conditions and the degradation state of the samples.

2. Materials and methods

Two different locations in a radiant section of a coil tube from the REPSOL POLÍMEROS plant were studied. Coil behavior
was analyzed with Finite Element Analysis (FEA) (Section 3). Also, the metallurgy of the corresponding alloys was studied.
For that purpose tube samples were collected and analyzed in as-supplied and used (after failure) condition. Fig. 1 shows the
correspondent coil configuration and the location of sample collection.
Samples were cut in the radial direction, both in as-supplied and in used condition tubes. The collected samples, labeled S1
and S2 for convenience, correspond to different service conditions and lifetime, which are slightly more severe for S2 (Table 1).
The samples were first grinded and polished to a 1 lm finish, and etched with glyceregia solution [20]. Microstructural fea-
tures of the alloys were then assessed using field emission gun scanning electron microscopy (FEG-SEM) (JSM-7001F, JEOL),
coupled with energy dispersive spectroscopy microanalysis (EDS) (Inca pentaFETx3, OXFORD INSTRUMENTS).
The tube portion where S1 was collected was manufactured in high temperature alloy H39WM, while the S2 portion was
manufactured in high temperature alloy H46M. These alloys correspond to the main materials used nowadays in centrifu-
gally cast tubes for steam cracker furnaces, especially for the hot outlet tubes [21]. Mechanical properties of both high tem-
perature alloys are rendered in Table 2. The H39WM alloy presents high creep strength and high carburization resistance
[21]. It is a fully austenitic alloy, with high Cr contents, which is essential for the formation of a protective Cr2O3 layer that
enhances oxidation and carburization resistance at high temperature [22]. Alloying elements such as Cr, Nb, Mo, Ti and W act
as carbide formers and retard the rate of softening at high temperature [1]. Ni, Si, Al and Mg strengthen ferrous alloys
through solid solution strengthening and grain-boundary control [1,23]. Higher nickel grades such as H46M were developed
in the 1980s aiming to improve chemical resistance up to higher temperatures, with sufficient creep strength for ethylene
tubes [21], while further increasing resistance to carburizing atmospheres and carbon diffusion [6,21]. Additionally the high
percentage of nickel improves resistance to thermal fatigue and maintains an austenitic structure so the alloy remains
ductile [22,23]. In both alloys increased creep performance results from the introduction of Nb additions: Nb precipitates
eutectic niobium-carbides, predominantly at grain-boundaries and also among the chromium carbides [21]. The high
melting point niobium carbides are insoluble up to 1250 °C and remain in their positions without volume change during
alloy exposure to high temperatures [24]. Grain growth is inhibited and the formation of creep voids during coalescence
of chromium carbides is transported to longer times or higher stresses and/or higher temperatures [21,25].

3. Finite element analysis

FEA was carried out using the ABAQUSÒ program, in order to simulate radiant coil behavior and response during various
stages of the furnace service and run length.
The radiant coils of pyrolysis furnaces are suspended vertically. Depending on their arrangement, coils are combined into
a transfer line, TL, whose output is at the top of the combustion chamber. Each coil has tabs on the top and bottom. The upper
guides, which extend through the roof of the combustion chamber, are fixed outside the furnace, using a spring suspension
system. The left part of the coil (due to this being approximately symmetrical) was drawn and modeled (Fig. 1a) to determine
the stress and displacement distributions arising from the loading conditions (Fig. 1b).

Please cite this article in press as: Santos M et al. Effect of severe operation conditions on the degradation state of radiant coils in pyrolysis
furnaces. Eng Fail Anal (2015), http://dx.doi.org/10.1016/j.engfailanal.2015.02.010
 M. Santos et al. / Engineering Failure Analysis xxx (2015) xxx–xxx 3

Fig. 1. Schematic representation of the analyzed coil and position of sample collection: (a) coil model; (b) boundary conditions used in FE simulation. S1:
H39WM alloy and S2: H46M alloy.

Table 1
Service conditions and nominal composition of the alloys [5,6].

Location S1 S2
Operating temperature (°C) 1075 1100
Operating pressure (MPa) 0.3 0.4
Service life (years) 5 9
Material H39WM (25/35/1/0.4-Cr/Ni/Nb/C) H46M (GX45NiCrSiNb45-35)
Composition (wt%) C 0.4 0.4
Si 1.8 0.8
Mn 1 1
Cr 35 25
Fe 15 35
Nb 1 1
Ni 45 35
Ti + Zr Microadditions Microadditions

The materials were considered homogeneous isotropic, with linear elastic behavior. It was assumed that all materials
have Poisson’s ratio of 0.33. Table 2 identifies the characteristics and physical properties of the coil used in the finite element
model. The coil is characterized by four steps (tubes 1/2/3/4) of two parallel tubes (A/B), followed by two steps that converge
to an outlet pipe of higher diameter (TL).
Tested run parameters were net weight, internal pressure, thermal expansion and coke weight, expected to introduce
stresses and strains upon coil structure. The study was carried out assuming end-of-run service conditions (Table 3). Coke
deposition was uniformly distributed throughout the coil, with a load value of 400 kg. The temperature distribution in
the coil was obtained previously considering the end-of-run [27] since it corresponds to the most severe condition.
Fig. 1b shows the location and directions of boundary conditions and the loads of model coil.

Please cite this article in press as: Santos M et al. Effect of severe operation conditions on the degradation state of radiant coils in pyrolysis
furnaces. Eng Fail Anal (2015), http://dx.doi.org/10.1016/j.engfailanal.2015.02.010
4 M. Santos et al. / Engineering Failure Analysis xxx (2015) xxx–xxx

Table 2
Characteristics and physical properties of the materials in different coil steps.

Tube 1/2/3(A/B) 4(A/B)/5/6 TL Guides

Material 25Cr35NiNb-Micro (H39WM) 35Cr45NiNb-Micro (H46M) 20Cr32NiNb G-NiCr28W
Density (kg/m3) 7970 7970 7940 8200
Young’s Modulus (GPa) 20 °C 158.5 176.5 155 190
800 °C 108.2 127.5 105.5 127
1000 °C 95.1 109 95.8 116.5
1100 °C 92.5 100 – –
Coefficient of thermal expansion (°C1) 20–100 °C 15.5  106 15.6  106 14.5  106 14.5  106
20–800 °C 18.9  106 18.8  106 17.5  106 17.0  106
20–1000 °C 19.4  106 19.2  106 18.5  106 17.5  106
20–1100 °C 19.7  106 19.5  106 – –

Table 3
Operation conditions of the furnaces in start-of-run and end-of-run [26].

Load (ton/h) Steam/hidrocarb P/M CIT/COT (°C) CIP/COP (bar)

Start-of-run 21.8 0.50 0.45 585/851 2.84/1.77
End-of-run 21.8 0.50 0.45 615/851 3.03/1.83

CIP: coil inlet pressure; CIT: coil inlet temperature; COP: coil outlet pressure; COT: Coil outlet pressure; P/M: polypropylene/methane ratio.

Table 4
Summary of maximum stress and displacement obtained from FEA.

Maximum stress (MPa) Maximum displacement (m)

Component weight 4.7 6.0  102
+ Internal pressure 4.8 6.0  102
+ Thermal expansion 9.4 2.1  101
+ Coke weight 7.8 2.0  101

The study began with analysis of the coil when subjected to its net weight at room temperature. The maximum stress
obtained for this condition is located at the bottom of the furnace inlet pipes, which coincides with the region that is sub-
jected to the most bending (maximum stress 4.7 MPa). The influence of internal pressure was then added to coil simulation.
As a result the maximum stress value slightly increases to 4.8 MPa, while displacement remained approx. the same (Table 4).
The maximum stress occurs on a low temperature area, but with the internal pressure it is possible to see higher stress levels
on the hotter outlet pipes, with the maximum stress level reaching 1.6 MPa.
The following step consisted in the introduction of the temperature profile as loading condition. As a result stresses
increased significantly. The coil region submitted to greater stress is located in the outlet section, with the stress level reach-
ing 9.4 MPa. Displacement increases substantially, reaching its maximum at the lower end of the coil.
Finally, coke weight (400 kg) was considered. Fig. 2 shows stress and strain distributions in the coil resulting from
simultaneous application of net weight (Fig. 2a), internal pressure (Fig. 2b), temperature (Fig. 2c) and coke weight
(Fig. 2d). The coil location submitted to greater stress was again the bottom of the furnace inlet pipes, but the actual maxi-
mum value has now decreased to 7.8 MPa. This location is not critical for creep failure, as the temperature is lower than on
the outlet section. Table 4 summarizes the numerical results obtained from FEA.
FEA of coil when submitted to different load conditions shows that temperature and coke weight are the operating
parameters more relevant to coil performance degradation. The greatest influence in coil displacement was observed for
the thermal expansion condition, although stresses thus caused are within the acceptable range.
As for the coke weight influence on stress distribution, it results that adding a 400 kg distributed load actually decreased
the overall maximum stress level on the furnace. In order to better understand this behavior a sensitivity analysis was car-
ried out, considering the coke weight as a variable ranging from 0 to 1200 kg. After analyzing the stress and temperature
distributions, the most critical point was considered to be located 1.5 m below the furnace main outlet junction (Fig. 1).
A temperature of 1060 °C and the higher stress level at this point lead to a lower creep life. Fig. 3 shows that a 9 MPa von
Mises stress is obtained when no coke weight is added to the furnace. Then the stress level on the critical point decreases
with the amount of coke weight that is loaded to the furnace, reaching a minimum of 1 MPa for 750 kg of coke weight.
The stress level will then begin to increase again. To understand this behavior the pipe section loads, axial force and bending
moment, are also shown in Fig. 3. While the axial force is proportional to coke weight, the bending moment decreases until
the loaded coke weight reaches 750 kg. As the normal stress is the most important stress component on the pipe, and is cal-
culated using Eq. (1), it is suggested that the bending moment is the term responsible for the stress level behavior.

Please cite this article in press as: Santos M et al. Effect of severe operation conditions on the degradation state of radiant coils in pyrolysis
furnaces. Eng Fail Anal (2015), http://dx.doi.org/10.1016/j.engfailanal.2015.02.010
 M. Santos et al. / Engineering Failure Analysis xxx (2015) xxx–xxx 5

Fig. 2. Simulations of von Mises stress in the coil considering: (a) net weight; (b) net weight and internal pressure; (c) net weight, internal pressure and
temperature distribution; (d) net weight, internal pressure temperature distribution and coke weight.

Please cite this article in press as: Santos M et al. Effect of severe operation conditions on the degradation state of radiant coils in pyrolysis
furnaces. Eng Fail Anal (2015), http://dx.doi.org/10.1016/j.engfailanal.2015.02.010
6 M. Santos et al. / Engineering Failure Analysis xxx (2015) xxx–xxx

Fig. 3. Coke weight influence on the critical point: von Mises stress level, pipe section axial force and bending moment.

Nxx Mzz y M yy z
rxx ¼  þ ð1Þ
A Izz Iyy

(where rxx is the section normal stress, Nxx is the axial force, Myy and MZZ are the bending moments, and A, Iyy and Izz are
the section geometrical properties). Bending is thus expected to result from the horizontal coil displacement produced by
the temperature distribution. Therefore by adding coke weight to the furnace the level of horizontal coil displacement will
be reduced, as for the bending of the furnace, until coke weight reaches 750 kg. The stress level will then decrease as the
weight is loaded, until finally starting to increase again, as the bending also starts to increase.
The degradation phenomena taking place in the radiant coils of pyrolysis furnaces at REPSOL POLÍMEROS frequently
result in failure at service times much shorter than the theoretical values rendered by the supplier. The service life
corresponding to actual operating conditions was thus calculated. The rupture time of the component, for a given tempera-
ture and pressure, was determined using the Larson–Miller creep parameter (Eq. (2)):

PLM ðrÞ ¼ Tðlog t þ CÞ ð2Þ

(where PLM is the Larson–Miller parameter, T is the temperature in K, t is the stress-rupture time in hours and C is a constant,
usually of order 20). Using Eq. (2) and the simulations rendered above, it is possible to calculate the rupture time of the
furnace. Combining the temperature and stress data allowed to assess the risk of creep failure on the H39WM alloy.
Using the information collected for the critical point (Fig. 1) on the H46M alloy, the rupture time can be calculated as a
function of coke weight loaded into the furnace. Fig. 4 shows the mean rupture times calculated, as well as the maximum
and minimum values accordingly to the data presented in [5,6]. It results that if the furnace is not loaded with coke the
mean rupture time is around 42,660 h, while a 400 kg load would increase the rupture time to 382,700 h. Considering
365 days  24 h  120 h (corresponding to 5 maintenance days/year)  135 h (corresponding to 9 decoking cycles/years),
the furnace yearly working time is around 8505 h. This means that the furnace rupture time would be between 5 and

Fig. 4. Coke weight influence on the mean rupture furnace time.

Please cite this article in press as: Santos M et al. Effect of severe operation conditions on the degradation state of radiant coils in pyrolysis
furnaces. Eng Fail Anal (2015), http://dx.doi.org/10.1016/j.engfailanal.2015.02.010
 M. Santos et al. / Engineering Failure Analysis xxx (2015) xxx–xxx 7

45 years, while the design stress-rupture life is of 12 years. Therefore the coke weight loaded into the furnace is the most
important parameter on the creep failure life of the furnace. If not well adjusted, this parameter can lead to early coil rupture.

4. Microstructural analysis

Tube condition has a significant effect on microstructure and on property changes, and the hottest zones (such as posi-
tions S1 and S2) are expected to be exposed to the maximum probability of damage and microstructural change [1].
Fig. 5a illustrates the microstructure of the H39WM alloy (S1) in as-supplied condition. Accordingly with its composition
the microstructure presents austenitic matrix, with lamellar eutectic primary carbides at the grain boundaries. EDS elemen-
tal analysis confirmed the presence of a chromium-rich carbide phase (dark grey in Fig. 5b) and of a niobium-rich carbide
(light grey in Fig. 5b, in smaller amount).
Fig. 6 shows the microstructural features of H39WM tube cross section after failure (Fig. 6a), together with X-ray elemen-
tal analysis maps. At least four regions are distinguishable in the microstructure near the internal diameter, labeled I, II, III
and IV in Fig. 6a. EDS elemental mapping shows that the region immediately next to the coke deposit is rich in Cr and O
(Fig. 6d and g, top), corresponding to the presence of the inherent Cr2O3 layer (I). When this film is present carbon diffusion
into the tube is retarded; however decoking operations result in Cr2O3 removal and regeneration [22] resulting in the
formation of the somewhat discontinuous and irregular Cr2O3 layer seen in Fig. 6d. A thin layer of Si2O (Fig. 6g and h,
top) follows (region II).
In region III the austenitic matrix is chromium depleted in an extension of approx. 100 lm (Fig. 6d, middle), as a result of
chromium diffusion to the surface. The coil metal becomes carburized by contact with coke and with the feedstock, resulting
in carbon enrichment from the inner diameter of the radiant tube (i.e., at the process gas side) [3]. This accelerates carbon
diffusion into the tube material [22]. Service at high temperatures may result in re-dissolution of primary carbides and
rearrangement of secondary carbides by reaction of diffusing carbon with carbide-forming elements [1], both in the super-
saturated matrix and at grain boundaries. As a result (region IV and Fig. 7) a continuous precipitates network is formed in the
grain boundaries, together with second-phase precipitation in the matrix. EDS results show that the precipitates in the
matrix have a complex composition including Fe, Cr, Si, Ti, Ni and Nb (r-phase), and their presence is expected to results
in loss of ductility and creep strength during coil cooling to room temperature [3,22].
Fig. 8a shows the microstructure of the H46M alloy (sample S2) in as-supplied condition. Again the alloy presents
approximately equiaxed grain structure, with carbide distribution along grain boundaries. EDS elemental analysis confirmed
the presence of a chromium-rich carbide phase (dark grey in Fig. 8b) and of a niobium-rich carbide (light grey in Fig. 8b), in
smaller amount.
Microstructural features of H46M alloy after failure in the tube inner diameter vicinity are quite similar to those of H39WM,
with the presence of a chromium oxide surface layer followed by an inner layer of silicon oxide and a chromium depleted layer
(with approximately 160 lm). From that layer onwards a continuous precipitate network is formed in the alloy grain bound-
aries (Fig. 9), mainly a chromium carbide (dark grey) and also r-phase (light grey). Precipitation in the matrix is quite less
abundant than in the H39WM alloy. Pore coalescence at grain boundaries is visible (Fig. 9b), confirming creep damage.
Results attained for both alloys suggest that microstructural degradation is related to the decreased coil service life at
REPSOL POLÍMEROS. Although the metallurgy of the used alloys confirms their ability to reduce carburization, it cannot
be totally eliminated. Coke formation is inherent to the cracking process and several damaging effects outcome from its
deposition on tube wall. On the one hand it contributes both to the decrease of maximum stress and to the increase of coil
displacement (Table 4). On the other hand carbon diffusion into the alloy results in microstructural and property changes:
the properties of the carburized metal are different from those of the uncarburised part of the tube wall, with different creep
strength, ductility and thermal expansion, which can eventually lead to damage in the tube wall by a variety of mechanisms
[5,6]. In particular, formation of carbides [7] and of r-phase precipitates [3,7,22] is expected to result in loss of room tem-
perature ductility and creep strength, imposing additional failure risk during cooling at decoking and stop-start operations.

Fig. 5. As-supplied H39WM (S1 sample): (a) low magnification backscattered electron image; and (b) detail of the carbide phases.

Please cite this article in press as: Santos M et al. Effect of severe operation conditions on the degradation state of radiant coils in pyrolysis
furnaces. Eng Fail Anal (2015), http://dx.doi.org/10.1016/j.engfailanal.2015.02.010
8 M. Santos et al. / Engineering Failure Analysis xxx (2015) xxx–xxx

Fig. 6. X-ray elemental map of the inner surface of S1 sample collected after failure: (a) FEG-SEM image; (b) Fe distribution; (c) C distribution; (d) Cr
distribution; (e) Nb distribution; (f) Ni distribution; (g) O distribution; (h) Si distribution.

5. Recommendations

Attained results suggest that the shorter than expected service life of radiant coils in pyrolysis furnaces at REPSOL
POLÍMEROS at Sines, Portugal, outcomes mainly from coke deposition. Considering the findings of FEM analysis, a few
recommendations are suggested. Experimental determination of the material creep curve and coke weight deposition, in
order to accurately calculate the coils creep rupture time, should be carried out. With this equation it will be possible to opti-
mize the decoking intervals and the overall furnace creep life. The material creep curve is more accurate then the Larson

Please cite this article in press as: Santos M et al. Effect of severe operation conditions on the degradation state of radiant coils in pyrolysis
furnaces. Eng Fail Anal (2015), http://dx.doi.org/10.1016/j.engfailanal.2015.02.010
 M. Santos et al. / Engineering Failure Analysis xxx (2015) xxx–xxx 9

Fig. 7. Precipitates network ahead of the chromium depleted zone in H39WM (sample S1) after failure: r-phase (light grey) and chromium carbide (dark
grey).

Fig. 8. As-supplied H46M (S2 sample): (a) low magnification backscattered electron image; and (b) detail of the carbide phases.

Fig. 9. H46M (sample S2) after failure: (a) precipitate network ahead of the chromium depleted zone (r-phase (light grey) and chromium carbide (dark
grey)); (b) magnification showing pores at grain boundaries.

Miller Parameter, enabling the material strain calculation over time. Coke weight deposition enables the stress and strain
calculation with the finite element model, which results in a single equation for the furnace life. Optimizing this value is thus
a mater of choosing the optimal decoking frequency.

6. Conclusion

Erosion caused both by fluid flow in the inner surface and by decoking operations destroys the dense, stable oxide layers
that hinder carbon diffusion into the matrix of the high temperature alloys in study. Carbon diffusion into the alloy results in
microstructural and property changes that can lead to damage in the tube wall by a variety of mechanisms.

Please cite this article in press as: Santos M et al. Effect of severe operation conditions on the degradation state of radiant coils in pyrolysis
furnaces. Eng Fail Anal (2015), http://dx.doi.org/10.1016/j.engfailanal.2015.02.010
10 M. Santos et al. / Engineering Failure Analysis xxx (2015) xxx–xxx

Results attained in this work suggest that the shorter than expected service life of radiant coils in pyrolysis furnaces at
REPSOL POLÍMEROS at Sines, Portugal, outcomes mainly from coke deposition. Finite element analysis results show that
the coil weight increase resulting from coking and the thermal expansion at high temperature, restricted by the gaps and
supporting springs, are the most determinant parameters for the mechanical behavior, resulting in severe microstructural
degradation and increasing both fracture and creep failure. Coking results in increased net weight, promoting creep. Also,
carbon proceeding from the coke layer into the tube metal alters the microstructure of the alloys near the inner diameter
region. Chromium depletion and formation of brittle precipitates lead to changes in the alloy properties, introducing lack
of creep resistance and low ductility mainly at low temperatures. Decoking operations remove the protective Cr2O3 surface
layer, exposing fresh metal surface. Carbon proceeding from ethylene cracking or from the decoking operation itself can thus
diffuse into the metal, promoting formation of a carbide network in the material grain boundaries (in both alloys) and matrix
(mainly in H39WM). These carbides decrease ductility during cooling to room temperature, accelerating failure under
thermal shock.
Accordingly, the operating lifetime of the coils can be extended by tuning their temperature profile, both during current
operation and on decoking. Frequent and standardized temperature readings are mandatory to assure a temperature profile
below project temperature, so that coke deposition is decreased.

References

[1] Borjali S, Allahkaram SR, Khosravi H. Effects of working temperature and carbon diffusion on the microstructure of high pressure heat-resistant
stainless steel tubes used in pyrolysis furnaces during service condition. Mater Des 2012;34:65–73.
[2] Jakobi D, Gommans R. Typical failures in pyrolysis coils for ethylene cracking. Mater Corros 2003;54(11):881–7.
[3] Vekeman Johan, Dewaele Michel. Repair welding of HP40-Nb. Weld World 2010;54:34–42.
[4] Khodamorad SH, Haghshenas Fatmehsari D, Rezaie H, Sadeghipour A. Analysis of ethylene cracking furnace tubes. Eng Fail Anal 2012;21:1–8.
[5] Paralloy Limited. Paralloy H39WM – High strength, high temperature alloy for steam cracker furnaces and steam reformers furnaces; Technical papers.
[6] Paralloy Limited. Paralloy H46M – Excellent choice for the most difficult and severe pyrolysis conditions; Technical papers.
[7] Tillack D, Guthrie J. Wrougth and cast heat resistant stainless steels and nickel alloysfor the refining and petrochemical industries. Nickel Development
Institute, 1998.
[8] Ray Ashok Kumar, Sinha Samarendra Kumar, Tiwari Yogendra Nath, Swaminathan Jagannathan, Das Gautam, Chaudhuri Satyabrata. Analysis of failed
reformer tubes. Eng Failure Anal 2003;10:351–62.
[9] Bhaumik SK, Rangaraju R, Parameswara MA, Bhaskaran TA, Venkataswamy MA, Raghuram AC, et al. Failure of reform tube of an ammonia plant. Engi
Fail Anal 2003;9:553–61.
[10] Konosu S, Koshimizu T, Lijima T, Maeda K. Evaluation of creep-fatigue damage interaction in HK 40 alloy. J Mech Des Trans ASME 1993;115:41–6.
[11] Monobe L. Caracterização do envelhecimento da liga 20Cr32Ni+Nb fundida por centrifugação e de seu efeito sobre o comportamento mecânico a frio.
Dissertação de Mestrado, Escola Politécnica da Universidade de São Paulo; 2007 [in portuguese].
[12] Huber J, Jakobi D. Centricast materials for high-temperature service. In: International conference & exhibition, Dusseldorf, Germany, 21–24 February
2011.
[13] Ul-Hamid A, Tawancy H, Mohamemed A, Abbas N. Failure analysis of furnace radiant tubes exposed to excessive temperature. Eng Fail Anal
2006;13:1005–21.
[14] Silveira TF, Almeida LH, Moreira MF, Silveira TL. Microestrutura de tubos de forno de pirólise desativados após 91.000h de operação – resultados da
metalografia ótica. Trabalho apresentado na 6ªCOTEC – Conferência Sobre Tecnologia de Equipamentos Salvador, Agosto de 2002 [in portuguese].
[15] Swaminathan J, Guguloth K, Gunjan M, Roy P, Ghosh R. Failure analysis and remaining life assessment of service exposed primary reformer heater
tubes. Eng Fail Anal 2008;15:311–31.
[16] May IL, Silveira TL, Vianna CH. Criteria for the evaluation of damage and remaining life in reformer furnace tubes. Int J Prv Piping 1996;66:233–41.
[17] Shipley DG. Creep damage in reformer tubes. Int J Press Ves Piping 1983;14:21–34.
[18] Gong JM, Tu ST, Yoon KB. Damage assessment and maintenance strategy of hydrogen reformer furnace tubes. Eng Fail Anal 1999;6:143–53.
[19] Woo Nam Soo. Assessment of damage and life prediction of austenitic stainless steel under high temperature creep–fatigue interaction condition.
Mater Sci Eng A 2002;322:64–72.
[20] Dupont John, Lippold John C, Kiser Samuel D. Welding metallurgy and weldability of nickel-base alloys. John Wiley & Sons, Inc.; 2009.
[21] Kirchheiner R, Woelpert P. Niobium in centrifugally cast tubes for petrochemical applications. in: Proceedings of the Niobium science & technology
international symposium; 2001. p. 1041–54.
[22] Tari V, Najafizadeh A, Aghaei MH, Mazloumi MA. Failure analysis of ethylene cracking tube. J Fail Anal Prev 2009;9:16–322.
[23] Mankins WL, Lamb S. Nickel and nickel alloys. In: ASM Handbook, vol. 2 – Non-ferrous alloys and special purpose materials. ASM International, Ohio;
1990.
[24] Terekov KI. Alloying, structure and phase composition of heat-resistant nickel-chromium alloys containing niobium. In: Pridautsev MV, editor.
Structure and properties of heat-resistant metals and alloys, Keter Press; 1970. p. 137–142.
[25] Chou TC, Huang W, Paciej R. Stress corrosion cracking of pyrotherm reformer tube for steam-reforming hydrogen production. J Mater Sci
1997;32:67–72.
[26] Wimer. Additional Operation Manual, Recoiling/Revamp Naftha Furnaces F1005 and F1006. Linde AG. Gas and Engineering, September 2003.
[27] Mesquita J. Substituição das serpentinas de radiação e crossover da fornalha F1001. Engenharia: ISQ; 2008 [in portuguese].

Please cite this article in press as: Santos M et al. Effect of severe operation conditions on the degradation state of radiant coils in pyrolysis
furnaces. Eng Fail Anal (2015), http://dx.doi.org/10.1016/j.engfailanal.2015.02.010