Engineering Failure Analysis 18 (2011) 1675–1682

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier .com/locate /engfai lanal

Examination of a failure detected in the convection zoneof a cracking furnace

V. Mertinger a, M. Benke a,⇑, Sz. Szabó b, O. Bánhidi c, B. Bollo b, Á. Kovács a

a University of Miskolc, Institute of Material Science, Hungaryb University of Miskolc, Department of Fluid and Heat Engineering, Hungaryc University of Miskolc, Institute of Chemistry, Hungary

a r t i c l e i n f o

Article history:Available online 12 February 2011

Keywords:Stainless steelCracking furnaceCorrosionFluid mechanical modelPhase transformation

1350-6307/$ - see front matter � 2011 Elsevier Ltddoi:10.1016/j.engfailanal.2011.02.003

⇑ Corresponding author.E-mail address: marton_benke@hotmail.com (M

a b s t r a c t

Heat-transfer investigation and a failure analysis of a cracking furnace are presented. Thecracking furnace, in which thermal decomposition of hydrocarbons occurs in its radiantsection, represents a plug-flow reactor placed in a firebox in point of fluid mechanicsand heat-transfer processes. The reacting mixture that consists of hydrocarbons and dilu-tion steam is heated up by means of natural gas in the burners of radiant section.

The aim of the investigation was to characterize the mechanical, chemical, and corrosiontransformation processes occurring inside the convection zone and to examine a specificfailure (leakage) process and damage. The geometrical model of the convection zone ofthe furnace was established by FLUENT software as well as the fluid mechanical modelof the heating stage which covered the calculation of the flow characteristics and the tem-perature field resulted by the corresponding heat flow processes.

Samples were taken from the different positions of the pipe made of different types ofsteel (A106, A335, A312, B407). The causes of the failure were investigated on the baseof the results of composition and fine structure examinations. The outer surfaces of thesamples of the pipes were examined using optical emission spectrometry (ARL 3460 OESinstrument), while the main metallic components were determined with atom absorptionspectrometry (device: PYE UNICAM PU 9100).

After taking the samples signs for failure and changes in the structure were looked for.For these investigations, optical microscopy, scanning electron microscopy, point or smallarea microprobe (EDAX) and X-ray diffraction methods were used.

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction

The cracking furnace producing stock for petrol chemistry is a gas–gas type counter flow heat exchanger. The gas to becracked enters the pack of four tubes at the top part of the convection zone of the furnace. The gas flows downwards in thetubes where it is heated up by the counterflow flue gas flowing between the tubes. The upper zone is the so called preheatingsection. At the bottom of this section the tubes exit the furnace so the technological steam and the admixture containingsulphur can be added. The tube containing the hydrocarbon-steam mixture returns into the furnace where it keeps heatingthe mixture flowing downwards. As the mixture leaves the convection zone, its temperature reaches the 700 �C, which is thetemperature the cracking process starts at. However, cracking is avoided in this segment of the furnace, because the materialof the tubes in this section is not suitable for the temperature of the cracking process. The most common failure of the

. All rights reserved.

. Benke).

1676 V. Mertinger et al. / Engineering Failure Analysis 18 (2011) 1675–1682

furnace is the leakage of the tubes. Reparation of this kind of failure requires shutdown of the complete technological processand replacement of the damaged tube part inside the furnace is also very complex. Reports mark the operation at too hightemperatures as the reason for failure most frequently, which can be originated from designing or operating problems[1–6].

The sections of the tubes can be classified as straight or elbow (bend) parts by geometrical property. The materials of thetubes are A106, A335, A312, B407 alloys in the direction of increasing temperature. The aim of our examinations wasthe characterization of the flow mechanical, heat transfer, chemical, corrosion and transformation processes occurring inthe tube system and to reveal the failure mechanisms and their risks in the different alloys of the tubes. For that, the flowmechanics and heat transfer numerical simulation of the convection section was performed. Samples were taken fromdifferent sections of the tubes during the planned service period when the tubes were replaced, and they were comparedto the results of our calculations. Furthermore, a specific pinning was also examined. The failure process of the alloyA312 is demonstrated in the present paper.

2. Results

2.1. Flow mechanical and heat transfer simulation

Constrained flow occurs both inside the tubes and between the tubes, meaning that heat transfer occurs through con-strained convection and heat radiation. The housing was considered as a three-layer housing during the model establish-ment. The outer 5 mm thick steel housing was drawn with zero thickness but in the heat transfer model its heatpermeability was set to be equivalent with the 5 mm thick steel housing. The steel beams were neglected. The rock cottoninsulation inside the steel housing was taken into account. The inner heat resistant housing was simplified to two differentmaterials, heat resistant perlite concrete and heat resistant brick.

During the model establishment of the tube system, the wall thickness was also not drawn, but it was taken into accountduring the calculations, similar to the case of the housing.

In the present case the natural wall of the convection zone is the border of the calculations as well. In the case of theflowing media the borders are the zone entering and zone leaving cross sections. The number of elements used for thenumerical modelling is above 3 million (3,275,297 pieces). The boundary conditions of the calculations were establishedby the data given by the procurer. The calculations were performed on ethane as feed of the furnace. Realizable k-epsilon(rke) model was used because this one is the optimal for cylindrical flows, like flows in tubes. Because of the largetemperature differences, the heat radiation is significant in the convection zone. Radiation DO-model was used to model this.The CFD model was validated by global data. The results of the simulations were compared to the empirical results given bythe procurer. The calculated and measured data were identical within the error of the measurement.

The simulation of the whole heating process was performed using a non-steady model of the FLUENT system. The numberof the mathematical steps was enormous because of the large number of the geometrical elements and the heat transfer andflow mechanical equations. Furthermore, the calculation was stable and convergent using a very small (s = 0.001 s) timestep.

The real time s = 8.2329 s of the examined phenomena was continuously calculated during the process. It must beconsidered during the evaluation of the results. Accordingly, the presented data do not belong to a stable stationaryoperation, but to the �2/3 part of the heating stage. However, the retrieved data are sufficient to present the fine structureof the evolved flow mechanics and heat transfer conditions in the convection zone of the cracking furnace and to draw therequired conclusions. The examined states are the following:

� Both the ethane and the steam flows to the tubes in the full length.� The walls are continuously heated, but the increase is only 4 �C. Notable amount of the heat of the steam heats the wall.� The cooling of the steam and the slow increase (�0.95 �C/day) of the exit temperature of the ethane-steam mixture

flowing in the tube occurs.

The ethane and the joining gas are heavily heated as they pass through the tubes. The ethane is heated to 250–300 �C,where it encounters the steam with the temperature of 230 �C. Their mixture is further heated. The temperature of theethane-steam mixture leaving the convection zone was not experienced to reach the 680–710 �C, it only reached 473 �C tillthe examined time. However, the results show significant, more than 30 �C differences between the temperatures of thetubes. By heat equilibrium the estimated temperature difference between tubes is �50 �C. The ethane entering with18.8 m/s and steam the entering with 16.3 m/s reach the average velocity of 71.2 m/s during their heating, but locally, espe-cially in the lower elbow bends their velocity exceeds the 90 m/s. After reaching the final temperature, this value exceeds the100–110 m/s. Because of this numerous velocity, the contaminations together with the corrosion and erosion products in thegas, cause further significant erosion.

The temperature and the velocity distributions in the cross section of the furnace are shown in Figs. 1 and 2.Henceforth the fine structure of the evolved flow was investigated based on the results of the simulation. The sections of

the tube cut for the material investigations were examined separately. Based on the examination of the temperature and

Fig. 1. Temperature distribution in the furnace (�C).

Fig. 2. Velocity distribution in the furnace (m/s).

V. Mertinger et al. / Engineering Failure Analysis 18 (2011) 1675–1682 1677

velocity distribution in these sections it is found that the sections were under notable thermal stress and because of the highvelocities the incidental corrosion and erosion products were potential failure sources especially in the lower elbow bendsections.

Fig. 3. Macroscopic degradations (a) pitting on the inner surface of the elbow bend sample, (b) asymmetric decrease in wall thickness of a straight sample,rough pitting.

1678 V. Mertinger et al. / Engineering Failure Analysis 18 (2011) 1675–1682

2.2. Macrostructure investigations

The degradation of the inner surfaces of the tube sections were detected in some locations during the macro-structuralvisual examinations. An example is shown in Fig. 3, where serial pitting arranged in the direction of the gas flow can beseen together with a surface with rough pitting and asymmetric decrease in wall thickness can be seen on a sample cutfrom a straight section right after an elbow bend section. The results of the flow mechanic model were examined at thelocations of these samples and they confirmed the degradations. The temperature of the wall with decreasing thickness inthe straight section is elevated and the temperature and velocity distributions in this cross section are asymmetric(Fig. 4).

The fine structure of the flow varies highly depending on the location inside the elbow tube consequently so does thelocation of the failure. Fig. 5 shows a strongly asymmetric wall temperature distribution and a failure in agreement with this.The austenitic sample exhibits strong ferromagnetic sign in the vicinity of the failure. The pinning inside the furnace wasrepaired through welding.

2.3. Micro-structural investigations

Table 1 contains the results of the local chemical compositional examinations on the major components performed onboth inner (from the gas flow) and outer sides of the same tube section to understand the fine structural processes. Sample2 is taken from the location of the previously mentioned decrease in wall thickness, sample 1 is taken from the same crosssection of the tube but from a location with original wall thickness, sample 3 is taken from a location closer to the zone ofelevated temperature, a location with around 100 �C higher temperature. It can be seen, that moving forward the C contenthighly increases in the inner side of the tube, meaning that C diffuses from the gas into the tube.

The same phenomenon occurs in the case of S (one hand S comes from in the hydrocarbons, on the other hand from thetechnological additives). In the content of the major alloying elements, Cr and Ni and the minor microalloying elements thereare no notable differences along the tubes and between the inner and outer surfaces.

The following features were found based on the microstructure examinations:

Fig. 4. (a) Temperature distribution (�C) of the wall (b) temperature distribution (�C) of the gas flowing through the cross section (c) velocity distribution(m/s) of the gas flowing through the cross section of the sample with asymmetric decrease of wall thickness.

Fig. 5. (a) Asymmetric wall temperature distribution of the elbow sample (�C) (b) Tube pinning with a subsequent welding.

Table 1Compositions of the samples [m/m%].

Sample C inside C outside S inside S outside Cr inside Cr outside Ni inside Ni outside

1.(6/2) 0.16 0.05 0.017 0.014 17.4 17.1 10.2 10.52.(6/3) 0.31 0.05 0.073 0.014 17.7 17.3 10.4 10.63.(9/2) 0.82 0.10 0.054 0.014 17.5 17.0 10.6 10.3

V. Mertinger et al. / Engineering Failure Analysis 18 (2011) 1675–1682 1679

� The change of the structure of the surface chromium-oxide:

The surface chromium-oxide protective layer was pinned, tapered, damaged at many locations. Iron-oxide appeared atsome locations instead of the chromium-oxide. The chipped pieces were flocked together in cemented form at some loca-tions. This is shown in Fig. 6a. The selective area EDAX examinations revealed that the composition of areas marked with5 and 1 are identical to the matrix, while area 2 is enriched in Cr and O, and area 3 is enriched in Cr, O and S, while area4 is depleted in Ni.

The chromium-oxide was revealed to appear not on the surface, but beneath the spongy corrosion product in many cases.This spongy corrosion product on the inner (gas flow) side is marked with 1 in Fig. 6b. The solid chromium-oxide is markedwith 2, while 3 marks the Cr depleted matrix. The morphology of the chromium-oxide definitely shows that the growth rateis smaller on the matrix side and it is formed not during the production, but the operation. The most convenient period forthat is when the tubes are cleaned by steam flow.

� Presence of S content phases:

More and more S containing phases appear near to the surface towards the hotter section of the tube. They appearpredominantly parallel with the surface. This is shown in Fig. 7 area marked with 2 is enriched in S. The orientation andmorphology of the phase accelerates the decrease of wall thickness.

Fig. 6. Backscattered electron (SEM) image of the cross section of the tube, inner (gas flow) side.

Fig. 7. Composite (SEM) image of the cross section of the tube, inner (gas flow) side.

1680 V. Mertinger et al. / Engineering Failure Analysis 18 (2011) 1675–1682

� Chromium-carbide precipitation at the grain boundaries:

Chromium-carbide appears at grain boundaries and in some cases inside the grains at distinct locations, predominantlybeneath the heavily damaged chromium-oxide layer. The amount of the revealed phase is obviously larger than the C con-tent of the alloy would cause. This fact and the close locations of the phase to the edge of the sample together suggest that Ccomes from the gas and Cr comes from the matrix near to the surface. An example for this is shown in Fig. 8. Area 1 is iden-tified as corrosion product, area 2 is a chromium-carbide precipitation and area 3 is a region depleted in Cr.

Fig. 8. Backscattered electron (SEM) image of the cross section of the tube, inner (gas flow) side.

Fig. 9. Composite image (SEM) of the location of pinning, inner (gas flow) side.

Fig. 10. Selective area X-ray diffraction patterns from two locations of decreasing wall thickness in the vicinity of pinning.

Fig. 11. The amount of austenite calculated from selective area X-ray diffraction measurements in the vicinity of pinning.

Fig. 12. The variation of hardness from the inner (gas flow) side of the tube towards the outer side in the vicinity of pinning.

V. Mertinger et al. / Engineering Failure Analysis 18 (2011) 1675–1682 1681

1682 V. Mertinger et al. / Engineering Failure Analysis 18 (2011) 1675–1682

� Formation of new iron based phases:

Examining the close vicinity of the pinning on the sample taken from the damaged section of the tube numerous phaseswere found in the originally austenitic matrix giving magnetic signs. Fig. 9 shows an area with such phases. Selective areaX-ray diffraction (diameter: 1 mm) was performed along the cross section of the wall. Fig. 10 shows two images taken fromthe cross section of the tube at the reduced wall thickness e martensite, ferrite with high alloying content was also detectedbesides the austenite at the inner (gas flow) side. The amount of the austenite was calculated based on the intensities of the{2 0 0} plane series of the austenite and ferrite phases given by the selective area X-ray diffraction measurements movingfrom the inner side of the tube to the outer regions (Fig. 11). It can be seen that it increases within the 1.5 mm distance to-wards the outer side. The remaining wall thickness in this region is 4.5 mm. In this thin region where the Cr content is higher,other metastable phases characteristic for these alloys like the r phase may also be present. However, the significant phasein this region is the ferrite, which gave the strong ferromagnetic signal. (The r phase is not ferromagnetic [7].) The mechan-ical characteristics of the wall vary through the formation of the new phases. To examine this, micro-hardness measure-ments were performed at the locations of the X-ray measurements through the cross section. According to the resultsshown in Fig. 12, the characteristics of the heterogeneous structure scatter strongly.

3. Conclusion

The corrosion resistance of austenitic stainless steels is due to the surface chromium-oxide layer and the solute chromiumcontent of the austenite. The primary failure of the examined component is the abrasion due to the high velocity (up to90 m/s) gas flow during the operation. Remarkable cyclic heat stress is contributed to the abrasion, leading to the damageand pinning of the surface layer loosing its protective purpose. Carbon and sulphur intake from the gas occurs at the elevatedtemperature regions leading to iron-oxide formation on the surface. Since the structure of the iron-oxide is not compact, butspongy, it can be easily chipped. Chipped iron-oxide particles help the abrasion of the tube. Chromium-carbide precipitationoccurs dominantly at the grain boundaries due to the change of composition. This causes the decrease of the amount ofsolute chromium actuating corrosion processes. Phases with sulphur content form with bad adhesion lead to surface damageand decreasing wall thickness. The periodic steam flow procedure would be a self-repairing procedure by intending toreform the surface chromium-oxide layer of the alloy. This procedure becomes a self-degradation process by further extract-ing the solute chromium content of the alloy. The surface composition of the tube is drastically changed by this procedure.Due to the change of the composition and the elevated temperature phase transformation processes occur leading to ferriteand e martensite formation. The mechanical characteristics of a heterogeneous structure are also heterogeneous, whichaccelerates the abrasion process. The decrease of the wall thickness reaches a critical value and the tube is pinned.

4. Summary

The cause of pinning of the tube system of a cracking furnace was investigated. Examinations revealed a complex rela-tionship between the flow characteristics in the pipe system, the thermal stress, erosion effects and the physical and chem-ical processes, for instance carbonisation, oxidation, sulphur uptake, and the caused phase transformations, which explainsthe failures in the pipe system.

Acknowledgements

The presentation was supported by OTKA NI 68129 and NKTH-OTKA 68207.The authors say thank you to Dr. Tivadar Gál and Gábor Barancsi, the employees of the Tisza Chemical Group Public

Limited Company for their professional assistance.

References

[1] Yoon KB, Yeong DB. Oxidation failure of radiant heater tubes. Eng Failure Anal 1998;6:101–12.[2] Guan Kaishu, Xu Hong, Wang Zhiwen. Analysis of failed ethylene cracking tubes. Eng Failure Anal 2005;12:420–31.[3] Ul-Hamid Anwar, Tawancy Hani M, Mohammed Abdul-Rashid I, Abbas Nureddin M. Failure analysis of furnace radiant tubes exposed to excessive

temperature. Eng Failure Anal 2006;13:1005–21.[4] Usman A, Nusair Khan A. Failure analysis of heat exchanger tubes. Eng Failure Anal 2008;15:118–28.[5] Ul-Hamid A, Tawancy HM, Al-Jaroudi SS, Mohammed AI, Abbas NM. Carburisation of Fe–Ni–Cr alloys at high temperatures. Mater Sci Pol 2006;24(2/1).[6] Shalaby HM. Failure investigation of a convection line elbow. Eng Failure Anal 2007;14:739–42.[7] Kobayashia DY, Wolynec S. Evaluation of the low corrosion resistant phase formed during the sigma phase precipitation in duplex stainless steels. Mat

Res São Carlos 1999; 2(4).