Professional Documents
Culture Documents
Silva 2007
Silva 2007
www.actamat-journals.com
Carburization alters the mechanical and magnetic properties of tubes used in pyrolysis furnaces. In this work, a giant magneto-
resistive sensor was used to measure the magnetic flux density at the surface of samples with different carburization degrees. The
magnetic measurements were correlated with the volumetric fraction of chromium carbides in each sample. A micro-genetic algo-
rithm was set up to solve the inverse problem, which estimates from the magnetic measurements the spatial distribution of the vol-
umetric fraction of the chromium carbides through the wall.
2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Pyrolysis furnaces are used in the petrochemical alloy and the operational characteristics of the furnace,
industry for the production of ethylene through thermal notably temperature. When the protective layer cracks,
cracking of hydrocarbons such as naphtha and ethane, the carbon from the gas reacts with chromium and with
in the presence of steam. The serpentine tubes of these other alloying elements, forming a layer of carbides [3].
furnaces are made of the HP family of steels with a high Due to the high atomic mobility of carbon in these
content of Cr and Ni (see Table 1). Some manufacturers layers, it permeates the alloy, causing extensive carbide
introduce Al, Nb and Ti into their alloys. The metallur- precipitation through the tube wall [1]. Near the internal
gical alterations associated with carburization end up surface of the tubes, the majority of the carbides are of
embrittling the tube itself, compromising its reliability, the M7C3 type [4]. As the distance from the internal sur-
turning it more susceptible to corrosion and reducing face increases, more of the carbides are of the type
its useful life [1–3]. Carburization also reduces the rup- M23C6. This distribution depends on the relation be-
ture limits, the ductility and the weldability of the mate- tween the carbon entry velocity through the protective
rial [5]. oxide layers, the rate of diffusion of this element through
The temperatures required for the cracking process in the matrix and the interdendritic boundaries of the
a pyrolysis furnace give rise to temperatures greater than material.
1000 C at the surface of the tubes in determined regions The degree of carburization of a tube can be deter-
of the serpentines. Such conditions, together with the mined by measuring the volumetric fraction of chro-
characteristics of the fluids present, ease the penetration mium carbides and following their through-wall
and diffusion of carbon into the tube walls through the distribution. Usually this is done by means of destruc-
internal surface, causing the detrimental effects known tive techniques, such as optical microscopy. We propose
as carburization. a non-destructive technique that determines the distribu-
The carbon comes from reactions involving carbon- tion of the volumetric fraction of the chromium carbides
rich gases that are present inside the serpentines. During through the wall.
normal operation a superficial layer of protective oxides High pressure (HP) steel tubes when not carburized
are formed inside the tubes. The nature and the stability are paramagnetic and have an austenitic micro-struc-
of this layer depend on the chemical composition of the ture, but with the increase of carburization they gradu-
ally become ferromagnetic [6–9], due to chromium
carbide precipitation [10,11]. As a result of carburization
* Corresponding author. E-mail: ivanco@metalmat.ufrj.br the magnetic permeability increases, starting from the
1359-6462/$ - see front matter 2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.scriptamat.2006.09.004
318 I. C. Silva et al. / Scripta Materialia 56 (2007) 317–320
Table 1. Basic composition of HP alloy used (wt.%) mium carbide through-wall volumetric fractions, which
C Cr Ni Si Mn Mo P S had been determined experimentally from their mag-
0.35–0.75 24–28 33–37 2.50 2.00 0.5 0.04 0.04
netic responses. Not only has this type of algorithm been
used extensively for optimizing variables of complex
problems, it has also been used for solving inverse prob-
internal surface of the tube towards the external surface lems related to non-destructive testing [15,16].
[12]. Inspections were made in three test specimens manu-
The technique used in this work was to measure the factured in HP steel, taken from different pyrolysis fur-
tangential component of the magnetic flux density over naces, presenting diverse degrees of carburization (Table
the surface of the tube. The specimen was exited by a 2). Measurements of the magnetic flux density were
small cylindrical ferrite magnet 3 mm long, with 3 mm made at reading points every 20 mm along the perimeter
diameter. The magnet had its N–S axis aligned with of different sections of the test specimens. Afterwards
the length of the tube and was placed at 10 mm from five points were chosen for the extraction of samples
the surface. The magnetic flux density was measured and volumetric fraction measurements. Measurements
with a giant magnetoresistance sensor [13,14] with its were carried out by scanning electron microscopy
sensitivity axis in the direction of the length of the tube. (SEM) at selected points of the samples in accordance
The value of the magnetic flux density was subtracted with the ASTM E 562 standard. For each sample, a
from the magnetic field measured in the absence of the curve was traced showing the percentage of chromium
tube, thereby obtaining the magnetic contribution of carbides through the walls of the tube, and the area un-
the material (see Fig. 1). der the curve was calculated. The magnetic measure-
The magnetic contribution of the material was mod- ment results obtained were then compared with the
eled by a distribution of equivalent magnetic dipoles measurements of the areas under the curves of the volu-
[15]. The thickness of the carburized tube was divided into metric fraction in such a way as to establish a correla-
L layers, containing N dipoles pointing in the direction of tion between them.
the tube length (see Fig. 1). The dipoles are uniformly The GAs were used to determine the equivalent mag-
distributed within each layer. The radius of the curvature netic dipole moments of the carburized samples in such
of the tube was taken into account in order to calculate a way as to produce a decline profile similar to that of
the distance between each dipole and the measurement the volumetric fraction of chromium carbides. It is
position on the external wall. The resulting magnetic flux known that the carburization process alters magnetiza-
density of the distribution can be expressed as tion, and that this phenomenon is related to the volu-
Bx ðx;y;zÞ metric fraction of the chromium carbides formed in
8 9 consequence of this process. The fitness function is
L X
X N >
< 2ðx x Þ2 ðy y Þ2 ðz z Þ2 > = defined by Eq. (2), where Bexp is the magnetic flux den-
l 0 n‘ n‘ n‘
¼ m‘ h i ; sity subtracted from the magnetic field measured in the
‘¼1 n¼1
4p > : ðx xn‘ Þ2 þ ðy y Þ2 þ ðz zn‘ Þ2
5=2 >
; absence of the tube.
n‘
2 3
ð1Þ XL X N
l 2ð x x Þ 2
ð y y Þ 2
ðz z Þ 2
0 6 n‘ n‘ n‘ 7
Fitness ¼ Bexp m‘ 4h i 5 :
4p 5=2
where l0 is the permeability of the free space, m‘ is the ‘¼1 n¼1 ðx xn‘ Þ2 þ ðy y n‘ Þ2 þ ðz zn‘ Þ2
magnetic dipole moment of the layer, (x, y, z) is the sen- ð2Þ
sor position coordinates and (xn‘, yn‘, zn‘) are the coordi-
nates of the magnetic dipoles in each layers. The maximum number of generations accepted with-
A genetic algorithm (GA) was used in this work to out alteration in individual improvement aptitude was
solve the inverse problem of distribution of the chro- 20 per population. Since the GA here is an algorithm
of the micro-genetic type, successive populations of a
small number of individuals were created and elimi-
nated, while preserving the best fitness. The stop criteria
established was the stagnation of improvement of the
best individual after five successive populations. In this
way it was possible to establish values of the magnetic
moments for each layer of model that tended to follow
the same decline profile as the volumetric through-wall
fractions. The parameters used in the genetic algorithms
and in the dipole model are shown in Table 3.
Figure 4. Comparison of the volumetric fraction of the chromium carbides with the profile of the dipole moments obtained by GA.
Table 4. Magnetic flux density measurements (lT) [3] D.M. Stewart, K.J. Stevens, A.B. Kaiser, Review of
Sample Experimental Theoretical (GA)
Quantitative Nondestructive Evaluation 20 (2001) 1732.
[4] K.J. Stevens, A.J. Tack, C.W. Thomas, D. Stewart,
2 200 190 Journal of Physics D: Applied Physics 34 (2001) 814.
3 300 300 [5] M. Zergoug, S. Lebaili, A. Boudjellal, A. Benchala, NDT
4 360 330 and International 37 (2004) 65.
5 500 310 [6] T. Klumper-wetkamp, P. Mayr, W. Reimche, K. Feiste,
M. Bernard, F. Bach, in: Proceedings of International
Symposium on Nondestructive Testing in Civil Engineer-
density and the total area of chromium carbides through ing 2003 (NDT CE-2003). BAM, DGZfp, Berlin.
the tube wall. This characteristic was exploited with the [7] Y. Cao, F. Ernst, G.M. Michal, Acta Materialia 51 (2003)
objective to develop a non-destructive method based on 4171.
magnetic flux density measurements. The results ob- [8] L.V. Biehl, D.Sc. Thesis, Rio Grande do Sul, PPGEM/
UFRGS, 2002.
tained confirmed the adequacy of the method as a tool
[9] M. Shindo, H. Nakajima, Journal of Nuclear Materials
to characterize the degree of carburization of the tubes. 144 (1987) 20.
The curves of the magnetic dipole moments, although of [10] G. Palasantzas, H.J. Kooij, J.T.M. DeHosson, Applied
different magnitudes, obtained by the GA held similari- Surface Science 229 (2004) 190.
ties with the volumetric fraction distribution of the chro- [11] K.J. Stevens, A. Parbhu, J. Soltis, Current Applied
mium carbides through the wall. Physics 4 (2004) 304.
[12] C.C. Spinosa, J.A. Marengo, M.C. Ruch, J.O. Garcı́a, in:
The authors would like to thank CNPq, CAPES, Proceedings of Third Pan American Conference for Non
FAPERJ and ANP for the scholarships and grants given Destructive Testing, June (2003) Rio de Janeiro.
[13] B. Wincheski, M. NamkungReview of Quantitative Non-
to support this work, and TSEC LTDA for the HP car-
destructive Evaluation, vol. 21, American Institute of
burized samples. Physics, Melville, NY, 2002, p. 465.
[14] J.R. Reitz, F.J. Milford, R.W. Christy, Foundations of
Electromagnetic Theory, fourth ed., Addison-Wesley,
[1] T.F. Silveira, Carburização em Tubos de Aço Inoxidável Reading, MA, 1993, p. 210.
Austenı́tico da Famı́lia HP Fundidos por Centrifugação. [15] R. RemptReview of Progress of Quantitative Nondestruc-
Rio de Janeiro, COPPE/UFRJ, M.Sc. Thesis, 2002, p. tive Evaluation, vol. 21, American Institute of Physics,
206. Melville, NY, 2002, p. 1771.
[2] H.J. Grabke, I. Wolf, Materials Science and Engineering [16] F. Kojima, N. Kubota, S. Hashimoto, Journal of Mate-
87 (1987) 23. rials Processing Technology 108 (2001) 263.