TUGraz OPEN Library

Steyrl, David

Krausz, Gunther

Koschutnig, Karl

Edlinger, Günther

Müller-Putz, Gernot

Reference layer adaptive filtering (RLAF) for EEG artifact reduction in simultaneous EEG-fMRI

Knöbelreiter, Patrick

Reinbacher, Christian

Shekhovtsov, Alexander

Pock, Thomas

End-to-End Training of Hybrid CNN-CRF Models for Stereo

Winkler, Margit

Winkler, Christoph K.

ORIGINAL PAPER
Trametes versicolor carboxylate reductase uncovered
Margit Winkler1 • Christoph K. Winkler2
Received: 17 November 2015 / Accepted: 3 December 2015 / Published online: 1 March 2016
 The Author(s) 2016. This article is published with open access at Springerlink.com
Abstract The first carboxylate reductase from Trametes
versicolor was identified, cloned, and expressed in
Escherichia coli. The enzyme reduces aromatic acids such
as benzoic acid and derivatives, cinnamic acid, and
3-phenylpropanoic acid, but also aliphatic acids such as
octanoic acid are reduced.
Graphical abstract
Keywords Bioreduction  Carboxylate reductase 
Carboxylic acids  Aldehydes  Enzymes
Introduction
The reduction of carboxylic acids is a challenge both for
chemical transformations and biotransformations. From the
thermodynamic viewpoint, the carboxylic acid moiety is in
an energetically favored state, shows little reactivity and
needs a high level of activation to participate in chemical
reactions. Due to the high stability of carboxylates,
examples for their direct reduction are rare and usually an
initial activation step (e.g., to the carboxylic ester or to the
acyl halide) is required. The reduction of the COOH group
or its derivatives yields the respective aldehyde in the first
step; because aldehyde is more reactive than the carboxylic
acid, the aldehyde does not accumulate but is typically
reduced further to the respective primary alcohol. Certain
reaction conditions may promote the accumulation of
aldehydes instead of alcohol: the well-known examples are
the reduction of acid chloride with lithium tri-t-butoxy
aluminum hydride [1] or DIBALH [2]. In the past two
decades, several selective chemical reductions to alcohols
or aldehydes using hydrosilanes have also been described
[3]. In laboratory practice, however, the acid is often fully
reduced to the alcohol and then re-oxidized to the aldehyde
[4], e.g., via the classical Swern or Dess–Martin oxidation
[5]. Other oxidation protocols use chromate or manganese
reagents, but also a large variety of methods using
molecular oxygen in combination with immobilized tran-
sition metal catalysts have recently been developed [6]. For
example, ruthenium or platinum on carbon in organic
solvents serves as the catalyst for the oxidation of aromatic
alcohols to the corresponding aldehydes [7, 8]. The use of
organic solvents, toxic reagents, undesired reaction condi-
tions or the high demand of reducing equivalents, however,
induces a strong wish for one step alternatives to the
selective preparation of aldehydes from acids.
A green substitution to chemical methods is the use of
biocatalysts with carboxylic acid reduction ability. The first
biocatalyzed reduction of carboxylic acid substrates was
accomplished by the white-rot fungus Trametes versicolor
[9]. A significant number of subsequent reports followed
that described the reduction of carboxylic acids also by
& Margit Winkler
margit.winkler@tugraz.at
1 Institute of Molecular Biotechnology, Graz University of
Technology, Graz, Austria
2 Department of Chemistry, Organic and Bioorganic
Chemistry, University of Graz, Graz, Austria
123
Monatsh Chem (2016) 147:575–578
DOI 10.1007/s00706-016-1676-z
other fungi such as Fomes fomentarius [10], Pycnoporus
cinnabarinus [11], Aspergillus sp. [11, 12], Bjerkandera sp.
[13], Mucor sp. [14], and many others [15, 16]. However,
to date, the amino acid sequence of only one single fungal
carboxylate reductase has been elucidated from Aspergillus
terreus [17, 18]. Herein, the sequence and substrate scope
of the first carboxylate reductase from T. versicolor have
been explored.
Results and discussion
Trametes versicolor (alias: Polystictus versicolor) was
reported to reduce benzoic acid derivatives to mixtures of
aldehydes and alcohols [9]. The enzyme responsible for the
first reduction step from the acid to the aldehyde had not
been identified to this date, nor has a particular enzyme
been purified and characterized. To discover the enzymatic
activity on sequence level, the fungal carboxylate reductase
enzyme sequence and three sequences of homologous
bacterial ATP and NADPH-dependent carboxylate reduc-
tases were used as templates to search within the non-
redundant protein sequences on NCBI with the restriction
to the organism T. versicolor. The hits from this search
were all annotated as acetyl-CoA synthetase-like proteins.
Aspergillus terreus ATEG_03630 [17] is classified as CaiC
for the adenylation domain and Thioester-redct for the
reduction domain by the multi-domain model, whereas the
CARs from Nocardia iowensis [19], Mycobacterium mar-
inum [20], and Segniliparus rotundus [21] are classified as
FAA1 and Lys2B. From the most significant hits of the
search, one sequence was classified to consist of an FAA1
adenylation domain in combination with a Thioester-re-
ductase domain. Therefore, it seemed most likely that this
sequence would display carboxylate reductase activity.
Since homologous bacterial carboxylate reductases were
shown to require the attachment of a phosphopantethein
moiety [22], the protein was expressed from the pET-
DUET1 vector that harbored simultaneously Escherichia
coli phosphopantethein transferase for post-translational
modification of the carboxylate reductase and the coding
sequence of the putative CAR enzyme with an N-terminal
fusion tag. Despite codon optimization of the fungal gene
sequence for expression in E. coli, the amount of soluble
protein was disappointingly low and the majority of the
protein was found in the insoluble fraction. The protein was
purified via affinity chromatography, yielding T. versicolor
CAR (TvCAR) in enriched form. As known CAR enzymes
have been well expressed heterologously in E. coli in their
apo-form [22], the reason is unlikely insufficient post-
translational phosphopantetheinylation. Optimization of
the expression conditions, such as the choice of a weaker
promoter, the use of chaperones [23], or ultimately a fungal
expression system, will likely result in significant
improvements with respect to enzyme yield.
In comparison to published CAR enzymes, T. versicolor
CAR is most similar to Aspergillus terreus CAR, albeit
with only 24 % identity according to UniProtAlign,
whereas N. iowensis CAR and M. marinum CAR exhibit
17 % and S. rotundus CAR 16 % identity, respectively.
Carboxylate reductase enzymes rely on ATP and
NADPH as cosubstrates. As typical for adenylating
enzymes, the presence of Mg2? is important due to its
interaction with ATP [24]. TvCAR preparation was assayed
using the conditions as described in the experimental sec-
tion. In addition, control reactions without the addition of
substrate (E)-cinnamic acid (1a), ATP, or MgCl2 were
carried out. NADPH was not oxidized in the absence of
ATP or substrate. When the reaction was carried out in
TrisHCl buffer without MgCl2, the carboxylate reduction
rate was diminished by approximately 30 %. It needs to be
noted that the enzyme preparation contained 10 mM
MgCl2, resulting in 0.5 mM MgCl2 end concentration in
this particular reaction.
TvCAR was subjected to an NADPH depletion assay in
the presence of different carboxylic acids. Enzymatic
activity was observed for the substrates listed in Table 1. In
the presence of, e.g. phenylacetic acid, mandelic acid,
2-pyrrole-, 2-pyridine-, 3-indole-, and 2-pyrazinecar-
boxylic acid as well as short chain aliphatic acids levulinic
acid, pentanoic acid and hexanoic acid, NADPH was not
oxidized by TvCAR. The fact that 1a, benzoic acid (2a),
3-hydroxy- and 3-methoxybenzoic acid (3a, 4a) as well as
4-methoxybenzoic acid (5a) are reduced, but phenylacetic
acid is not, is consistent with the observations of Farmer
et al. [9]. However, 2-hydroxy- and 2-methoxybenzoic
acids were not reduced by TvCAR under the conditions
used. This supports the idea that T. versicolor harbors more
than one CAR enzyme and CAR (XP_008043822.1) is
partly but not fully responsible for the reactions observed
in 1959. As an analog of 1a, 3-phenylpropanoic acid (6a)
was also reduced. Surprisingly, also aliphatic carboxylates
from C7 to C9 (7a–9a) were converted, whereas short
chain acids and longer chain acids were not accepted as
substrates under these conditions (Table 1).
To investigate the identity of the reduction product, (E)-
cinnamic acid (1a) was subjected to a biotransformation
reaction in the presence of the TvCAR enzyme and excess
of co-factors. The reaction was analyzed after extraction
and derivatization of the remaining substrate 1a by GC/MS
analysis, revealing cinnamaldehyde (1b) as the sole pro-
duct in a mixture with methyl cinnamate (1d) (Scheme 1).
576 M. Winkler, C. K. Winkler
123
Conclusion
The currently available portfolio of carboxylate reductase
enzymes is limited to a handful of protein sequences. With
the aim to enlighten new biocatalysts, we identified a
putative acetyl-CoA synthetase-like protein from T. versi-
color, expressed it in E. coli and confirmed that this protein
indeed shows carboxylate reductase activity. The enzyme
exhibits a promising substrate scope; however, its
heterologous expression is a clear limitation at the moment.
Experimental
Substrates, product references, DTT, and trimethylsilyl-
diazomethane (solution, 2 M in diethyl ether) were
purchased from Sigma Aldrich and used without further
purification. ATP was obtained from Roche Diagnostics.
NADPH and MES were purchased from Roth, IPTG from
Serva, and MgCl2 from Merck.
Cloning and expression
Four literature known carboxylate reductase enzyme
sequences (Q6RKB1.1, WP_012393886.1,
WP_013138593.1, and XP_001212808.1) were used as
search templates for a blastp search against non-redundant
protein sequences with the restriction to the organism T.
versicolor (taxid:5325). A gene, coding for the protein
sequence of NCBI accession code XP_008043822.1 was
ordered as synthetic gene with optimization of the genetic
code for expression in E. coli. The protein was expressed
from the pETDUET1 vector with the E. coli pptase (NCBI
accession code CAQ31055.1) cloned into the first multiple
cloning site and N-terminally HIS-tagged CAR sequences
in the second multiple cloning site. The plasmid was
transfected to E. coli BL32(DE3) Star and colonies selected
on LB/Amp. The enzyme was expressed using standard
conditions. The cells were harvested by centrifugation and
stored at -20 C. After thawing, the cells were disrupted
by sonication and the protein purified by nickel affinity
chromatography, using the gravity flow protocol. The
protein containing fractions were pooled and dialyzed
overnight at 4 C into 50 mM MES buffer, pH 7.5, con-
taining 10 mM MgCl2, 1 mM EDTA, and 1 mM DTT.
Aliquots of the resultant, slightly turbid protein solution
were shock frozen in liquid nitrogen and stored at -80 C.
Substrate screening
The capability of the protein to reduce carboxylic acids was
determined using a photometric NADPH depletion assay.
Therefore, potential carboxylic acid substrates were dis-
solved in KOH (0.1 M in water) or DMSO. The assay
composition was as follows: The substrates (10 mm3 of
100 mM stock solution) were added to 160 mm3 of
TrisHCl buffer (100 mM, pH 7.5, containing 10 mM
MgCl2). Subsequently, 10 mm
3 of NADPH (10 mM in
water), 10 mm3 of ATP (20 mM in water), and 10 mm3 of
enzyme preparation (3.85 mg/cm3 with an estimated purity
of 50 %) were added. Each reaction was measured four
times in parallel in UV-Star 96 well plates (Greiner). The
depletion of NADPH was followed on a Synergy Mx
Platereader at 340 nm and 28 C for 10 min. Blank
Scheme 1
OH
O
H
O
1a 1bATP
NADPH
TvCAR preparation
Table 1 Substrate scope of Trametes versicolor carboxylate reductase
Entry Substrate Specific activitya/mU mg-1
1 (E)-Cinnamic acid (1a) 41 ± 4
2 Benzoic acid (2a) 30 ± 9
3 3-Hydroxybenzoic acid (3a) 22 ± 8
4 3-Methoxybenzoic acid (4a) 23 ± 5
5 4-Methoxybenzoic acid (5a) 33 ± 2
6 3-Phenylpropanoic acid (6a) 34 ± 7
7 Heptanoic acid (7a) 34 ± 7
8 Octanoic acid (8a) 35 ± 4
9 Nonanoic acid (9a) 33 ± 1
a One activity unit is defined as the amount of enzyme preparation catalyzing the oxidation of 1 lM of NADPH per minute
Trametes versicolor carboxylate reductase uncovered 577
123
reactions without enzyme preparation were carried out in
parallel.
Biotransformation
1a (10 mm3 of 100 mM stock in DMSO) was added to
120 mm3 of MES buffer (50 mM, pH 7.5, containing
10 mM MgCl2, 1 mM EDTA, and 1 mM DTT). Subse-
quently, 10 mm3 of NADPH (100 mM in water), 50 mm3
of ATP (20 mM in water), and 60 mm3 of CAR enzyme
preparation (4.5 mg/cm3 total protein, containing approxi-
mately 50 % of CAR) were added. The reaction proceeded
in a Thermomixer at 28 C and 300 rpm. After 2 h and
overnight incubation, the reaction was acidified by the
addition of aqueous HCl (4 N, 50 mm3) and the products
were extracted with ethyl acetate (2 9 0.3 cm3). The
extracts were dried over Na2SO4 and TMS-diazomethane
solution (2 mm3, 2 M in diethyl ether) was added for the
derivatization of 1a to the corresponding methyl ester for
GC analysis. GC–MS analyses were performed on an
Agilent 7890A Series GC system equipped with a 5975C
MS system, using a (5 % phenyl) methylpolysiloxane
capillary column (HP-5MS, 30 m, 0.25, 0.25 lm film) with
He as the carrier gas. The method was 100 C hold
0.5 min, 10 C/min to 300 C. Retention times were: 1b:
6.10 min; cinnamic alcohol (1c, a potential product derived
from over-reduction): 6.47 min; methyl cinnamate (1d):
7.44 min.
Acknowledgments Open access funding provided by Graz
University of Technology. The Austrian science fund is kindly
acknowledged for financial support (Elise-Richter fellowship V415-
B21).
Open Access This article is distributed under the terms of the
Creative Commons Attribution 4.0 International License (http://
creativecommons.org/licenses/by/4.0/), which permits unrestricted
use, distribution, and reproduction in any medium, provided you give
appropriate credit to the original author(s) and the source, provide a
link to the Creative Commons license, and indicate if changes were
made.
References
1. Brown HC, Rao BCS (1958) J Am Chem Soc 80:5377
2. Chandrasekhar S, Kumar MS, Muralidhar B (1998) Tetrahedron
Lett 39:909
3. Addis D, Das S, Junge K, Beller M (2011) Angew Chem Int Ed
50:6004
4. Kotha S, Ravikumar O (2015) Beilstein J Org Chem 11:1259
5. Dess DB, Martin JC (1983) J Org Chem 48:4155
6. Ba¨ckvall J-E (2004) Modern oxidation methods. Wiley-VCH
Verlag GmbH and Co. KGaA, Weinheim
7. Mori S, Takubo M, Makida K, Yanase T, Aoyagi S, Maegawa T,
Monguchi Y, Sajiki H (2009) Chem Commun 34:5159
8. Korovchenko P, Donze C, Gallezot P, Besson M (2007) Catal
Today 121:13
9. Farmer VC, Henderson MEK, Russell JD (1959) Biochim Bio-
phys Acta 35:202
10. Ishikawa H, Schubert WJ, Nord FF (1963) Arch Biochem Bio-
phys 100:140
11. Arfmann HA, Abraham WR (1993) Z Naturforsch C 48:52
12. Motedayen N, Ismail MBT, Nazarpour F (2013) African J
Biotechnol 12:6618
13. Lauritsen FR, Lunding A (1998) Enzyme Microb Technol 22:459
14. Ma L, Liu X, Liang J, Zhang Z (2011) World J Microbiol
Biotechnol 27:2133
15. Napora-Wijata K, Strohmeier GA, Winkler M (2014) Biotechnol
J 9:822
16. da Silva Amaral L, Rodrigues-Filho E (2015) J Mol Catal B
Enzym 113:90
17. Wang M, Zhao H (2014) ACS Catal 4:1219
18. Wang M, Beissner M, Zhao H (2014) Chem Biol 21:257
19. He A, Li T, Daniels L, Fotheringham I, Rosazza JPN (2004) Appl
Environ Microbiol 70:1874
20. Akhtar MK, Turner NJ, Jones PR (2013) Proc Natl Acad Sci
110:87
21. Duan Y, Yao P, Chen X, Liu X, Zhang R, Feng J, Wu Q, Zhu D
(2015) J Mol Catal B Enzym 115:1
22. Venkitasubramanian P, Daniels L, Rosazza JPN (2007) J Biol
Chem 282:478
23. Weis R, Winkler M, Schittmayer M, Kambourakis S, Vink M,
Rozzell JD, GLieder A (2009) Adv Synth Catal 351:2140
24. Schmelz S, Naismith JH (2009) Curr Opin Struct Biol 19:666
578 M. Winkler, C. K. Winkler
123


Trametes versicolor carboxylate reductase uncovered

Pereira, Joana

Ofner, Patrick

Schwarz, Andreas

Müller-Putz, Gernot R.

Discriminating goal-directed from nongoal-directed 
movements and its potential impact for BCI control 
J. Pereira1, P. Ofner1, A. Schwarz1, G. R. Müller-Putz1*
1Institute  of Neural Engineering, Graz University of Technology, Graz, Austria
*Stremayrgasse 16/IV, 8010 Graz, Austria. E-mail: gernot.mueller@tugraz.at
Introduction: Differences in the electroencephalographic (EEG) recordings between the execution of goal-
directed and nongoal-directed movements have been recently shown in [1]. Such differences can be of interest 
for brain-computer interfaces (BCIs) control, when combined with information on the kinematic level (e.g. 
velocity decoding), since this combination mirrors the hierarchic way one plans a movement. In this study, we 
show that the time-domain differences between these movements are discriminable in a single-trial classification.  
Material, Methods and Results: Ten healthy, right-handed subjects participated in the experiment. Subjects were 
presented a small red ball on the monitor (Goal) or a red screen (No-Goal).  After 2 seconds and only when the 
stimuli color changed from red to purple, subjects were instructed to reach-and-touch the ball (Goal Movement) 
or to decide on their own where to touch (No-Goal Movement). 72 trials per condition were recorded. EEG 
signals were recorded using 60 passive electrodes and sampled at 512 Hz. 
Independent component analysis (ICA) was performed for artefact removal: components representing eye 
movements and muscle activity were rejected. To extract relevant low-frequency time-domain features, data were 
down sampled to 16 Hz, common average referenced and band-pass filtered from 0.3 to 3 Hz with a zero-phase 
4th order Butterworth filter. Classification was done using a random forest binary classifier; accuracies were 
calculated for each time-point and validated using 10x5-fold cross-validation. To score significantly above the 
chance level, 64.7% had to be reached (p=0.01, Bonferroni corrected for multiple tests over the trial length).  Fig. 
1 shows the time-course of the classification accuracies when discriminating Goal Movement and No-Goal 
Movement. Accuracies rise above the chance level after both first and second cues. After the GO cue (second 
cue), the average accuracy peaks immediately after movement onset with 67%. Here, 4 out of 10 subjects show 
accuracies above 80%, and all subjects are above the chance level. Also interestingly, 3 of the subjects show high 
accuracies even 2 seconds after movement onset. 
Figure 1. Classification accuracies when discriminating Goal and No-Goal Movements, time-locked at movement onset 
(t=0s). The first 2 vertical lines correspond to the average time-points when the 1st and the 2nd cue appeared, in 
respect to movement onset. The thick black line corresponds to the grand-average accuracy. 
Discussion: Our results show that there are differences between goal-directed and nongoal-directed movements 
when time-locking at movement onset. Namely, the motor-related cortical potentials – after the second cue- show 
different amplitudes between conditions.  These differences are discriminable in a single-trial classification. 
Future work will be to investigate whether similar results are obtained with neuroprostheses end-users and 
movement imagination (MI). If so, this information could be useful to establish activation thresholds, or even by 
instructing the subjects to imagine the kinesthetic MI associated with a target. We hypothesize that this 
instruction, combined with movement decoding at the kinematic level, could additionally improve classification 
accuracies.  
Significance: The results contribute to the goal of our research: a naturally-controlled BCI neuroprostheses. 
Furthermore, we encourage the BCI community to explore the neural correlates behind goal-directed movements 
and how recent neurophysiological findings in action planning (e.g. [2]) can be of practical interest for BCIs. 
Acknowledgements: This work is supported by the EU ICT Programme Project H2020-643955, “MoreGrasp”
and the ERC Consolidator Grant “Feel Your Reach”. 
References 
[1] Pereira J, Ofner P, Muller-Putz G.R. Goal-directed or aimless? EEG differences during the preparation of a reach-and-touch task. In 
Proceedins of the 37th Annual International Conference of the IEEE/EMBS, 1488-1491. IEEE, 2015.  
[2] Aflalo T, Kellis S, Klaes C, Lee B, Shi Y, Pejsa K, Shanfield K, Hayes-Jackson S, Aisen Heck C and Liu C. Decoding motor imagery 
from the posterior parietal cortex of a tetraplegic human. Science, 348(6237), pp.906-910,2015.  
DOI: 10.3217/978-3-85125-467-9-67Proceedings of the 6th International Brain-Computer Interface Meeting, organized by the BCI Society
Published by Verlag der TU Graz, Graz University of Technology, sponsored by g.tec medical engineering GmbH 67


Discriminating goal-directed from nongoal-directed movements and its potential impact for BCI control

Singhal, Arpit

Cloete, Schalk

Radl, Stefan

Amini, Shahriar

Multiscale modelling of heat transfer from arrays of spherical particles

Braud-Santoni, Nicolas

Bloem, Roderick

Jacobs, Swen

Synthesis of Self-Stabilizing and Byzantine-Resilient Distributed Systems

Wu, Mingqiu

Khinast, Johannes

A Model to Predict Liquid Bridge Formation between Wet Particles Based on Direct Numerical Simulations

Chen, Xiaojun

Xu, Lu

Yang, Yue

Egger, Jan

A semi-automatic computer-aided method for surgical template design

Weingrill, Leonhard Andreas

Nasiri, Mohammad Bagher

Enzinger, Norbert

10th International Conference on
Proceedings
Organizing Society
American Welding Society (AWS)
Japan Welding Society (JWS)
http://trends2016.org
October 11 - 14, 2016
Hitotsubashi Hall,
National Center of Sciences Building
Tokyo, Japan
&
9th International 
Welding Symposium of
Japan Welding Society 
(9WS)
Trends 
in Welding 
Research
Numerical simulation of Pearlite formation during welding of rails 
Weingrill Leonhard, Nasiri Mohammad Bagher, Enzinger Norbert 
Institute of Material Science and Welding, Graz University of Technology, Graz. Austria 
leonhard.weingrill@tugraz.at, +43 (316) 873 - 1671 
 
 
Abstract 
In this work, a thermo-metallurgical coupled simulation model of welding of R350HT pearlitic rail steel was 
realized to estimate the mechanical properties in the HAZ. The applied approach was to estimate the hardness of 
pearlitic phase fractions by calculating the interlamellar spacing. This was realized in a methodology of multiple 
calculation steps. Therein, MATLAB®-routines are used to calculate parameters for a metallurgical model to be 
used in SYSWELD®, based on a simulated TTT-diagram from JMatPro® and an experimentally determined 
welding- CCT-diagram of R350HT rail steel. This so established metallurgical model was verified by a simulation 
of dilatometry tests in SYSWELD® and subsequent quantitative comparison to experimental results. In a final 
step three more MATLAB®-routines are used to calculate phase fractions, interlamellar spacing and hardness for 
a practical example. Therefore, measured temperature curves from instrumented flash-butt welding (FBW) 
experiments of standard rails were used as input. The results of the simulation are compared to real welds’ phase 
fractions and hardness values in the HAZ and showed very good agreement. 
 
 
Introduction 
Due to various reasons – such as increased traffic, higher market shares of high speed trains, a general increase in 
traveling speeds and axle loads, but also increased safety regulations – the requirements for railways have 
increased in recent decades [1] und thus have tightened the cost situation. Consequently, the intention of rail 
operators to extent maintenance intervals and overall lifetime of tracks in order to cut down operation cost is more 
vital as ever. In answer to this development manufacturers seek to provide rails with improved properties against 
wear and rolling contact fatigue (RCF) [2], [3]. However, when welding modern rail steel grades undesirable 
microstructures are formed within the HAZ, which deteriorates the designed properties locally [4], [5]. As a 
countermeasure in today’s common rail welding processes thermal cycles are adjusted to reduce the dimension of 
this so called soft zone. However with further increasing mechanical properties of newly development rail steel 
grades also requirement for correspondingly optimized welding processes is given. 
 
Interlamellar spacing is an informative aspect in order to characterize the mechanical properties of pearlitic steels 
such as strength, hardness and thus wear-resistance [6]–[8]. The pearlitic phase transformation is influenced 
beside others by the thermal history of the material [9]. Thus, by adapting the thermal cycle during welding also 
the properties of the weld joint can be controlled. 
 
Table 1: Chemical composition and Brinell hardness at running surface of investigated pearlitic rail steel [10]. 
 
steel grade 
acc. to [10] 
main alloying elements in weight -%  HBW 
on running surface C Si Mn P max. S max. Cr 
R350HT 0,72- 0,80 
0,15- 
0,58 
0,70- 
1,20 0,020 0,025 ≤0,15 350- 390 
 
 
Work content 
The objective of this work was to have a tool to estimate the influence of microstructure changes in the HAZ of 
R350HT rail steel – s. Table 1 – on the mechanical properties as a function of the thermal history due to welding. 
Therefore a simulation model was developed which consists of numerical calculations, but also uses experimental 
methods for data gathering and validation. It calculates pearlitic phase fractions and the interlamellar spacing, as 
well as the hardness. 
 
Approach 
The thermo-metallurgical numerical simulation was realized by using MATLAB® routines in combination with 
SYSWELD®. The general approach is schematically depicted in a flow chart in Fig. 1 and calculation steps are 
correspondingly described in the following passages. 
 
S32-1
599
October 13th (Thursday)
In the first preparatory step (i) required material and process data are gathered. This is done with the help of:  
 Simulation of a TTT-diagram of R350HT rail steel in JMatPro®, based on measured chemical 
composition and austenitization conditions comparable to welding. 
 A welding-CCT-diagram of R350HT rail steel, obtained by dilatometry experiments. 
 Instrumented flash-butt welding experiments to capture T(t)-curves (Tij) in the HAZ at the rail head at 
different distances (points of measurement Pij) from the weld centerline (WCL). 
These so retrieved data are than preprocessed to the corresponding format and provided for further deployment. 
 
In the subsequent calculation step (ii) MATLAB®-routine (a) is used in order to determine parameters of the 
metallurgical model of the type SYSWELD® uses, s. [11]. Johnson-Mehl-Avrami-Kolmogorov (JMAK) type 
kinetics model for diffusion controlled but non-isothermal transformations, as also described in [12], is used. For 
martensitic transformations Koistinen-Marburger type model is used [13]. For both, the routine initially calculates 
parameters based on the data from the simulated TTT-diagram and experimental welding CCT-diagram. 
Subsequently an iterative algorithm, based on additivity rule, as also described in [14], calculates the final phase 
fractions Pf of Martensite, Bainite and Pearlite for each cooling rate . In order to optimize the results the 
parameter  is determined by comparing numerical and experimental results and recomputing Pf at every 
repetition cycle until sufficient agreement is reached. Finally, all determined parameters are put together to 
establish the metallurgical database file for SYSWELD®. 
 
In the next step (iii) the accuracy of the model is verified. Therefore, the beforehand conducted dilatometry 
experiments are simulated using the thermo-metallurgical coupled calculation option of SYSWELD®. Simplified 
sample geometry and artificially adapted cooling parameters are used. The cooling curves from the experimental 
welding CCT diagram are accurately reproduced. Phase fractions of Martensite, Bainite and Pearlite are 
quantitatively compared to the experimental results in order to verify the established metallurgical model. 
 
 
Fig. 1 Schematic flow chart of the used calculation approach. 
 
In step (iv) the calculation is finalized by using three more MATLAB® routines. First, routine (b) is fed with two 
T(t)-curves measured in the HAZ during welding, gathered during experiments in a recently published work [15]. 
Based on these so defined thermal cycles the metallurgical model established in step (ii) is used to calculate the 
phase fractions of Martensite, Bainite and Pearlite in distinct points. Subsequently, MATLAB®-routine (c) 
calculates the interlamellar spacing S of the computed pearlitic phase fraction and corresponding T((t)-curve, 
respectively the corresponding cooling rate. The calculation of S is based on an empirical expression proposed in 
[16] and depicted in Equation 1. Thereafter the interlamellar spacing in eutectoid steel can be predicted for 
isothermal transformation as a function of specific alloying elements’ content, the temperature T, at which pearlite 
600
October 13th (Thursday)
formation happens, and the eutectoid temperature TE. In order to take into consideration the non-isothermal 
formation of the Pearlite during welding, a weighted average of the interlamellar spacing is calculated 
subsequently for each cooling rate  based on Equation 2 subsequently. It integrates the interlamellar 
spacing S(T) over the derivatives of formed phase fractions dP at the corresponding calculation step for a given 
cooling rate and finally sets it in relation to the final phase fraction Pf. Finally MATLAB®-routine (d) calculates 
the hardness based on an empirically determined equation proposed in literature [6], s. Equation 3. 
 
Equation 1 [16]: 
 
Equation 2: 
 
Equation 3 [6]: 
 
 
 
Results 
First of all the results of step (iii) from the dilatometry simulation in SYSWELD® are compared in Fig. 2 to 
experimental results from the welding CCT-diagram of R350HT. It can be derived that the simulation shows very 
good agreement to the experimental data for all three relevant phases and in all investigated cooling rate ranges. 
The calculation of formed phase fractions based on the new metallurigcal model of this work and two 
experimental T(t)-curves in step (iv) show fully pearlitic transformation in the HAZ. This is in very good 
agreement with metallography data provided by rail steel manufactrurers [17]. Table 2 shows the final result for 
computed interlamellar spacing at two distinct points in the HAZ – represented by their distances from WCL– 
together with computed hardness. Accordingly, the interlamellar spacing inside the HAZ of FBW rail decreases 
with increasing distance from the WCL resulting in an increased hardness. This can be understood as a result of  
the higher cooling rate. For comparison reasons the computed values are converted and compared to values from a 
literature source [4]. The computed hardness values are in very good agreement to experimental results. Thus, the 
model is able to very accurately calculate the hardness based on interlamellar spacing as a function of the cooling 
rate. 
 
 
 
Fig. 2. Calculated vs.measured phase fractions as a functions of cooling rates for R350HT rail steel. 
 
Table 2: Calculated interlamellar spacing and corresponding calculated hardness for 2 exemplary thermal cycles 
in the HAZ for R350HT 60E1 rail and comparison to experimental values. 
 
distance 
from weld 
centerline 
(WCL) 
 in mm 
average 
cooling 
rate from 
800°C to 
500°C 
peak 
temperature 
in °C 
calculated 
average 
interlamellar 
spacing  
in µm 
calculated 
hardness 
HV10 
HV10 
converted to 
HRC acc. to 
[18] 
experimental 
HRC of FBW 
of rails acc. to 
[4] 
1 -0,9321 1355 0,1078 357 36,6 38,5 
10 -1,1565 854 0,0809 388 39,5 39 
601
October 13th (Thursday)
Conclusions 
A methodology of combining MATLAB® routines, SYSWELD® and JMatPro®, as well as data from 
experimental work was developed to implement a multistep thermo-metallurgical simulation of phase 
transformations in the HAZ during welding of pearlitic rail steel R350HT. With the help of the determined 
metallurgical model phase fractions and interlamellar spacing was calculated as well as the hardness was 
calculated for different distances from the WCL in the HAZ at the rail head for a practical example. The 
comparison of calculated and experimental results showed that very good agreement was obtained. Thus, it was 
shown that the proposed approach to estimate the hardness in the HAZ by calculating the interlamellar spacing as 
a function of thermal history in a numerical welding simulation is a very practicable method. 
 
Outlook 
In a next step the model will be further extended in order to estimate the softening in the annealed zone of the 
weld. Therefore additional metallurgical phenomena which are believed to also have influence on the softening in 
the HAZ of pearlitic rail steel will be studied. Furthermore, additional T(t)-curves at the corresponding locations 
in the HAZ and proper material parameters are needed. Based on this, an extended approach will be developed 
and implemented to further enhance the here presented simulation model. 
 
 
Acknowledgments 
K-Project Network of Excellence for Metal JOINing is fostered in the frame of COMET - Competence Centers for 
Excellent Technologies by BMWFW, BMVIT, FFG, Land Oberösterreich, Land Steiermark, Land Tirol and SFG. 
The programme COMET is handled by FFG. 
 
 
References 
[1] European Commission, “Fourth report on monitoring development of the rail market,” pp. 1–43, 2014. 
[2] L. I. Xiao-fei, P. Langenberg, S. Münstermann, and W. Bleck, “Recent Developments of Modern Rail Steels,” 
HSLA Steels 2005, vol. 2, 2005. 
[3] R. Kuziak and T. Zygmunt, “A new method of rail head hardening of standard-gauge rails for improved wear 
and damage resistance,” Steel Res. Int., vol. 84, no. 1, pp. 13–19, 2013. 
[4] P. Micenko and H. Li, “Double Dip Hardness Profiles in Rail Weld Heat-affected Zone — Literature and 
Research Review Report,” Brisbane, Australia, 2013. 
[5] P. Mutton, J. Cookson, C. Qiu, and D. Welsby, “Microstructural characterisation of rolling contact fatigue 
damage in flashbutt welds,” Wear, pp. 1–10, 2016. 
[6] A. J. Perez-Unzueta and J. H. Beynon, “Microstructure and wear resistance of pearlitic rail steels.pdf,” Wear, 
vol. 162–164, pp. 173–82, 1993. 
[7] B. L. Bramfitt and B. S. Corporation, “Structure / Property Relationships in Irons and Steels,” Met. Handb. 
Desk Ed. Second Ed., pp. 153–173, 1998. 
[8] J. Toribio, B. Gonzalez, J. C. Matos, and F. J. Ayaso, “Role of the microstructure on the mechanical properties 
of fully Pearlitic eutectoid steels,” Frat. ed Integrita Strutt., vol. 30, pp. 424–430, 2014. 
[9] K. M. Wu and H. K. D. H. Bhadeshia, “Extremely fine pearlite by continuous cooling transformation,” Scr. 
Mater., vol. 67, no. 1, pp. 53–56, 2012. 
[10] “ÖNORM EN 13674-1: 2011 Bahnanwendungen - Oberbau -Schienen Teil 1: Vignolschienen ab 46kg/m;” 
Standards Institute Austria, Vienna, Austria, 2011. 
[11] “SYSWELD 2016 Reference Manual.” ESI Group, Paris, France, 2016. 
[12] P. Krüger, “On the relation between non-isothermal and isothermal Kolmogorov-Johnson-Mehl-Avrami 
crystallization kinetics,” J. Phys. Chem. Solids, vol. 54, no. 11, pp. 1549–1555, 1993. 
[13] D. P. Koistinen and R. B. Marburger, “A general equation prescribing the extent of the austenite-martensite 
transformation in pure iron-carbon and plain carbon steels,” Acta Metall., vol. 7, 1959. 
[14] J. B. Leblond and J. Devaux, “A new kinetic model for anisothermal metallurgical transformations in steels 
including effect of austenite grain size,” Acta Metall., vol. 32, no. 1, pp. 137–146, 1984. 
[15] L. Weingrill, J. Krutzler, and N. Enzinger, “Temperature field evolution during flash-butt welding of railway 
rails.” THERMEC 2016, Graz, Austria., 2016. 
[16] M. Takahashi, “Reaustenitisation from bainite in steels.” Ph.D. thesis, University of Cambridge, 1992. 
[17] “Interne Doku Abbrennstumpfschweißung R350HT Probe 6, Schweißstoß 624.” voestalpine Schienen GmbH, 
Leoben/Donawitz AUSTRIA, 2010. 
[18] “ÖNORM EN ISO 18265:2003 : Metallische Werkstoff - Umwertung von Härtewerten.” Standards Institute 
Austria, Vienna, Austria, 2004. 
602
October 13th (Thursday)


Graz University of Technology

Not a member yet