Вы находитесь на странице: 1из 8

The 2010 Sir Henry Bessemer Lecture

From trial and error to computer modelling of


thermomechanical processing
C. M. Sellars*
Since the 1950s, thermomechanical laboratory in Swinden Laboratories in this difference arises from the diffe-
processing in the iron and steel industry Rotherham, and US Steel had both rence in crystal structure, with bcc
has progressed from being skills based, Fundamental and Applied Research ferrite of high stacking fault energy and
using results from trial and error Laboratories in Monroeville on the high diffusion coefficient undergoing
development work, to being science outskirts of Pittsburgh, there was still a rapid dynamic recovery, leading to low
based, using computer modelling for gulf between production metallurgists flow stress and low stored energy,
process optimisation and control, and and laboratory research. This was partly which has the consequence that recry-
prediction of product properties. Some because of the complex conditions of stallisation is relatively slow. Conversely,
of the steps in this evolution are industrial processing and the fcc austenite, which has a relatively low
illustrated in the lecture, using hot simplifications that had to be made for stacking fault energy, undergoes only
rolling, as the main example, because systematic study in the laboratory. The slow dynamic recovery, leading to high
The University of Sheffield has been gap was first bridged by studies of flow stress and high stored energy. In
involved in research in this field over product properties and their dependence Fig. 2, the flow stress ratio is about 4 in
the whole of the period. As basic on microstructure, but before electron the high purity iron, but this ratio decrea-
understanding of the physics of the microscopy of thin foils, many critical ses with increasing carbon and other
microstructural changes, and the speed features could not be observed, and there alloy element content.
of computing have increased, there has was still much learning by rote in In Fig. 1b, the flow stress of the ferrite
been a continuing trend for more metallurgy degrees. The systematic study rises with strain to a steady state value,
complex models to be used for offline of the dependence of microstructure on which decreases somewhat with increa-
optimisation and for online control and the processing conditions during hot sing strain because of the temperature
property prediction. Such models are working was a relatively late rise produced by deformational heating
now accepted as an economically development, which for steels can be at the higher strain rates. There was
valuable tool for thermomechanical considered to start with the hot torsion discussion about the possible role of
processing. testing carried out by Claude Rossard at grain boundary migration, but reversed
IRSID and published in 1958.1 By 1969 in torsion testing of aluminium, which
a review of strength and structure under behaves in an analogous way to ferrite,
This lecture provides an opportunity for hot working conditions,2 there were still showed that the original grains still
me to summarise some of the research only 176 relevant references on steels existed to high equivalent strains of
that has contributed to the award of the and non-ferrous metals, but the physical about 2 (Fig. 3). The heavily distorted
Bessemer Medal, and to acknowledge metallurgy foundations on which later grains after forward twisting are clearly
the major contributions of more than 150 modelling was built had been established. returned to their original size and shape
people who have worked with me. These by the reversed twisting.
people include research students, post- Flow stress and microstructure In Fig. 1a, the flow stress for austenite
doctoral research associates, research The stressstrain curves for a carbon rises to a peak, then falls to a steady
visitors and colleagues in academia and in steel austenite and for a 25% state value, with oscillations in the stress
industry. They are too many to name, but chromium ferritic stainless steel are at low strain rates. These oscillations are
they know the contributions they have shown in Fig. 1. These display analogous to the oscillations in strain
made to the transition, which is the title important similarities and differences rate during creep tests at constant
of this lecture. between austenite and ferrite during stress, e.g. on nickel, which behaves in a
I start in the 1950s, because this is deformation under hot working similar way to austenite during hot
the time when I studied for my degree conditions. The flow stress for both deformation (Fig. 4). These oscillations
in metallurgy at the University of steels is highly sensitive to strain rate, had been shown to arise from the
Sheffield. It was a period of rapid and other tests in the systematic study occurrence of repeated dynamic recry-
advances in the steel industry, but showed a similar sensitivity to the stallisation. There was lively debate
these were mainly generated from deformation temperature, indicating about whether dynamic recrystallisation
work trials and empirical data. Although that the flow stress is determined by could be fast enough to take place at the
steel companies carried out basic workhardening balanced by thermally higher strain rates of interest in hot
research, e.g. the United Steel activated dynamic softening processes. working, and whether the observed re-
Companies had a major research It is clearly evident by comparison of crystallised grains had formed statically
Fig. 1a and b that, despite its high alloy during cooling after deformation rather
content, the ferritic steel has less than than dynamically during the deformation.
Department of Materials Science and
Engineering, The University of Sheffield, Mappin half the strength of the austenite under The observation (Fig. 5) of tangled
Street, Sheffield S1 3JD, UK the same deformation conditions. dislocation subgrains inside the
*Corresponding author, email c.m.sellars@ Tests on pure iron at the transformation recrystallised grains in nickel established
sheffield.ac.uk temperature (Fig. 2) established that that the grains had been formed during

2011 Institute of Materials, Minerals and Mining


Published by Maney on behalf of the Institute
250 DOI 10.1179/030192310X12706305605817 Ironmaking and Steelmaking 2011 VOL 38 NO 4
NEWS AND VIEWS

1 Inuence of strain rate on stressstrain curves derived from hot torsion data by Rossard and Blain1 on a 0.25%C steel
in the austenitic condition and b 25%Cr ferritic stainless steel2

nucleation and limited growth. From temperature creep, it was found that
these, and many other equivalent ob- flow stress s at a constant value of
servations, the basic physical metallur- strain, the peak flow stress or the steady
gy of deformation under hot working state flow stress could be closely corre-
:
conditions can be summarised as lated with the strain rate e and the (instan-
shown in Fig. 8. taneous) absolute temperature T, over the
whole range of interest in hot working
using only two experimentally determined
Quantification material dependent parameters: a, a
With this basic understanding, it was strength coefficient, and Q, an activation
possible to describe the flow stress by energy for deformation. An example of
2 Temperature dependence of hot equations that are physically reasonable, such a correlation is shown in Fig. 9.
strength of ferrite and austenite in even though the fundamentally correct The whole stressstrain curve could
zone rened iron: dashed curve form of the equations was not known. then be modelled for any combination
estimated from the Ac transforma- Following work by Garofalo on high of strain rate and temperature. This is
tion temperatures3

deformation, and modelling (Fig. 6)


demonstrated how overlapping cycles of
dynamic recrystallisation could result in
the single peak in stress observed at
high strain rates. Sakai and Jonas7
showed that the conditions for over-
lapping cycles of recrystallisation arose
when the recrystallised grain size is less
than half the original grain size, so that
recrystallisation takes place as a series
of necklaces of new grains, which start
forming at the grain boundaries.
Under these conditions, it was found
that the recrystallised grain size did not
change during recrystallisation (Fig. 7)
and that the grain size is uniquely
related to the steady state flow stress,
and therefore to the strain rate and
temperature of deformation. Thus in
contrast to classical static
recrystallisation, which proceeds by
nucleation of the new grains followed a 5z5 rev., 62; b 525 rev., 62; c 5z5 rev., 620; d 525 rev., 620
by their continuous growth, dynamic 3 Effect of reversal in direction of twisting during torsion testing of aluminium
recrystallisation proceeds by repeated at 400uC4

Ironmaking and Steelmaking 2011 VOL 38 NO 4 251


NEWS AND VIEWS

pass reduction, to give e, the roll speed


:
to give e and the mean temperature to
give T for each pass. The problem was
to know the mean temperature,
because, in industry, pyrometers can
only measure the surface temperature,
which changes continuously. In the
laboratory, an internal temperature
could be measured using embedded
thermocouples, but a finite difference
computer model had to be developed to
interpret the results.13 A heat transfer
4 Creep curves for nickel under constant stress in compression, showing coefficient to the rolls could then be
repeated dynamic recrystallisation5 found, which gave good matching
between the modelled and measured
thermocouple temperatures, as
illustrated in Fig. 12. In this figure, the
short time during a pass gives an almost
vertical line with an increase in
temperature at the centre of the slab
because of deformational heating, but
severe chilling at the surface. In the
1970s, the model had to be run over-
night on the main frame computer of the
university, but now essentially the same
model runs in a second or so on a PC!
The model enabled mean temperatures
to be computed for any roll passes.
a e51.6361022 s21, e57.0; b e54.9561021 s21, e54.5 Validation of the model for industrial
5 Electron micrograph of thin foil of pure nickel deformed in torsion to true rolling was obtained by comparing
strain of 7.0 at 934uC and strain rate of 1.661022 s21 (Ref. 6) predicted surface temperatures with
measured pyrometer temperatures, as
illustrated in Fig. 10 for ferrite, when equation for workhardening and dyna- shown for example in Fig. 13. The roll
dynamic recovery is the only softening mic recovery, leading to an extrapola- loads and torques could now be
process balancing workhardening. For ted steady state stress se at higher computed for any pass for which the
this situation, there is a continuous strains. Beyond the critical strain, the stock entering the pass is in a fully
exponential decay in the rate of rise of additional softening Ds is then modelled recrystallised condition.
flow stress from an initial value so to from the kinetics of dynamic
the steady state value. When dynamic recrystallisation, as illustrated in Fig. 11. Interpass microstructural
recrystallisation also takes place after From the quantitative relationships changes
some critical strain, which is less than for the stressstrain curves, the mean
the strain to the peak flow stress ep flow stress required to compute the The high temperature of hot rolling
(Fig. 11), the initial part of the curve can rolling load and torque can then be means that recovery and
still be modelled by the exponential computed from the roll geometry and recrystallisation of the deformed

6 Modelled stressstrain curves for hot deformation when dynamic recrystallisation takes place at a low and b high
strain rates6

252 Ironmaking and Steelmaking 2011 VOL 38 NO 4


NEWS AND VIEWS

7 Dependence of dynamically recry-


stallised grain size on fraction re-
crystallised in commercially pure
nickel deformed at 880uC and
1.161021 s21 (Ref. 8)

microstructure can take place rapidly in


the time interval between passes.
Because there is no concurrent
deformation, these processes are
referred to as static to distinguish 9 Correlation between ZenerHollomon
them from the dynamic changes that parameter Z and sinh function of 10 Comparison of experimental and
take place during the passes. These peak ow stress for 0.25%C austenite modelled stressstrain curves for
static processes cause significant from data such as those in Fig. 110 pure iron in the ferrite tempera-
softening, as illustrated in Fig. 14 for a ture range of workhardening and
notional six-pass rolling schedule. The dynamic recovery11
upper curve shows the situation if
static recrystallisation is too slow to
start in the time between passes, and
the minor effects of static recovery are
neglected, so that the strain effectively
accumulates from pass to pass. This
means that the original equiaxed grain
structure becomes progressively more
elongated and develops a crystallo-
graphic texture. This situation is
common in aluminium alloys, which
means that modelling microstructural
evolution under rolling conditions is
primarily concerned with controlling
texture, because of its important
effects on properties and surface
quality. The same situation can also
occur in ferritic stainless steels, but for
steels that are single phase austenite
under rolling conditions, static
recrystallisation is rapid, leading to 8 Summary of microstructural changes resulting from a workhardening and
complete softening between passes, dynamic recovery, and b workhardening, slow dynamic recovery and
as shown by the lower curve in Fig. 14. dynamic recrystallisation9
The recrystallisation also leads to grain
refinement of the austenite, and to
subsequent refinement of the product
microstructure formed by the
transformation on cooling after hot
rolling, and so grain refinement is a key
feature of modelling the rolling
conditions for austenite.

Modelling methodology
Modelling must therefore compute the
kinetics of recrystallisation, the
recrystallised grain size produced, and
the rate of subsequent grain growth
(coarsening) in order to characterise the 11 Comparison of experimental and modelled stressstrain curves for Ti bear-
microstructure entering each pass. ing CMn steel austenite tested at 900uC and strain rate of 4.9 s21 (Ref. 12)

Ironmaking and Steelmaking 2011 VOL 38 NO 4 253


NEWS AND VIEWS

14 Effect of static recrystallisation


between passes on ow stress of
micoalloyed steel during deforma-
tion at 900uC and a strain rate of
5 s21 (Ref. 15)
16 Comparison of predicted evolu-
empirical equations, based on analysis of tion of austenite grain size with
available experimental data had to be observations (points) by Sekine
derived. Validation that these equations and Maruyama during experimen-
are satisfactory for single phase austenite tal plate rolling of vacuum melted
12 Temperature change during experi- is shown in Fig. 16, for which the rolling carbonmanganese steel17
mental rolling of a 19 mm slab of variables of reduction and temperature
type 304 stainless steel13 were measured and are shown in the austenite errors in prediction of micro-
figure. The modelled reductions in aus- structure after one pass tended to be
Equations for these variables can then tenite grain size after each pass for this compensated in the next pass, so they
be used to model the evolution of plate rolling schedule are compared with were not cumulative, and the predicted
microstructure pass by pass from the experimentally measured values, to give microstructures tended to converge to a
initial reheated microstructure of the surprisingly good agreement. grain size dependent on the rolling
slab/billet and the rolling parameters for In order to generalise the microstruc- conditions. This favourable situation is
the passes, as shown schematically in tural evolution model to other rolling considered to be a true reflection of the
Fig. 15. In this figure, S represents all conditions, it was necessary to connect behaviour of austenite, which minimises
the parameters required to describe it with other submodels, which generate the effects of cast to cast variations in
grain structure and S9represents all the the necessary data flow, as illustrated in composition and initial microstructure,
parameters required to describe any Fig. 17. This particular interconnection and greatly benefits the steel industry.
second phase present. In an ideal of submodels formed the basis of However, when the model was used to
model, these would be internal state collaborative work between research try to predict the variation in microstruc-
variables, which might include grain groups in the University of Sheffield and ture through the thickness of an experi-
size, grain shape and their distributions, the University of Leicester, which led to mentally rolled slab of stainless steel, it
dislocation density and its distribution, the development of a commercial model
proved to be totally inadequate, as shown
texture, volume fraction of second for hot rolling, SLIMMER (Sheffield
in Fig. 18.
phase, particle size, distribution, etc., Leicester Integrated Model for Micros-
Finite element modelling was used
because all are known to influence the tructural Evolution in Rolling).13 This
to compute the redundant shear strains
subsequent evolution of microstruc- model greatly facilitated study of the
ture, but, once again to make progress, effects of the rolling variables on micro-
physically reasonable, but essentially structure and flow stress.
Sensitivity analysis using the model
showed that in multipass rolling of

17 Schematic diagram of submodels


within model for industrial hot
13 Computed temperature changes deformation processes, showing
during industrial of 230 mm slab 15 Schematic diagram showing prin- information ow: s5micostruc-
to 11 mm plate (lines) compared ciples of modelling microstruc- ture, t5time, W5work done, T5
with pyrometer readings of sur- tural evolution during multipass temperature, e5equivalent strain,
:
face temperature (points)14 rolling16 e5strain rate, s5ow stress18

254 Ironmaking and Steelmaking 2011 VOL 38 NO 4


NEWS AND VIEWS

21 Interaction of recrystallisation and


start of strain induced matrix pre-
cipitation as function of niobium
18 Predicted and measured distribution content of microalloyed steels
of fraction recrystallised through rolled by 15% reduction at mean
thickness of slabs of type 316 strain rate of 10 s21: RLT shows
stainless steel rolled in experimen- lower temperature limit for full
tal mill at 1000uC, quenched and (85%) of recrystallisation and RST
20 Stressstrain curves at 850uC and shows upper temperature limit for
then annealed at 1000uC for 86 and
strain rate of 5 s21 for reduction stoppage (5%) of recrystallisation
142 s19
schedules of 30/15/15/15% with in subsequent passes23
various time intervals between
that occur during rolling at all positions deformations for X70 steel com-
away from the centre, and these were pared with carbonsiliconmanga-
then combined with the compression nese steel22 gave good agreement with the mean
of the rolling pass to obtain the local flow stresses calculated from rolling
value of equivalent strain e for substi- loads, measured by Tom Hope of Davy
predict the local variations in microstruc-
tution into the recrystallisation model to Mckee, up to pass 6, but for subsequent
tural evolution in rolling and forging when
compute the local fraction recrystallised. passes, the rolling loads are significantly
there are major changes in strain path.20
The large discrepancy between the higher than expected. It was speculated
predicted and measured fractions that this is a consequence of
recryallised at all positions away from Microalloyed steels precipitation hardening by the very fine
the centre shows that, while equivalent Modelling the microstructural evolution carbonitride particles.
strain is an adequate surrogate for the of this family of steels has been particu- Subsequent experimental plane strain
microstructural state variables for the larly significant in the overall development compression testing of an X70 steel
monotonic compression in rolling, it is of models for hot rolling, because the (Fig. 20), confirmed the contributions
totally inadequate when there are steels are used in the hot rolled condi- from solid solution strengthening and
changes in strain path. A new genera- tion and their mechanical properties are workhardening, and showed that
tion of models is still being developed to enhanced by optimising and controlling significant additional strengthening
the rolling variables. The key to the could accumulate pass by pass at
enhanced properties is control of strain temperatures where strain induced
induced precipitation of the alloy precipitation takes place. The amount of
carbonitrides. In the niobium microalloyed strengthening is sensitive to the pass
steels, the precipitation of Nb(C,N) takes reductions and to the time intervals
place at low temperatures in the austenite between passes, as shown in Fig. 20.
range and has a major effect on the hot Modelling the microstructural evolu-
strength, as illustrated in Fig. 19. tion of microalloyed steels therefore
In this figure, the crosses at the required additional submodels to those
bottom show the flow stress expected for single phase austenite. The first
for coarse grained reheated austenite was to predict when significant precipi-
of low carbon steel, and the circles tation would occur before recrystallisa-
show the increase in strength to be tion is expected to start in the absence
expected from the grain refinement of precipitation. This model, based on
produced by static recrystallisation the thermodynamic and kinetic equa-
after each of the earlier passes. In the tions for the precipitating species,
X60 niobium microalloyed steel, the together with empirical equations for
niobium in solid solution and then the density of nucleation sites in the
precipitation of fine particles of Nb(C,N) deformed austenite led to C curves for
is expected to prevent static recrys- the start t0.05 of precipitation shown in
tallisation between the finishing passes, Fig. 21. Clearly, the kinetics are
leading to a major increase in flow stress sensitive to the niobium content of the
19 Analysis of contributions to mean because of the retained workhardening. steel, and the model also showed the
ow stress as function of nish- There is also a contribution from solid sensitivity to the rolling variables, which
ing pass number during industrial solution strengthening by the niobium in can be compared with the sensitivity of
plate rolling of X60 steel21 solution. The sum of these contributions the kinetics of recrystallisation in the

Ironmaking and Steelmaking 2011 VOL 38 NO 4 255


NEWS AND VIEWS

22 Predicted precipitation strength-


ening during three-pass rolling of
0.06%Nb steel at 950uC with 15 s
between passes25

absence of precipitation. The start of


recrystallisation for the rolling condition
of the figure is shown by the lines labell-
23 Overview of modelling philosophy of IMMPETUS26
ed t0.05X, and the finish of recrystalli-
sation by the lines labelled t0.85X. It was challenging development, for which a In 1996, a multidisciplinary research
assumed that at temperatures above the physical basis was provided by detailed group, IMMPETUS (Institute for
point of intersection of the line for t0.85X electron microscopic studies of a Microstructural and Mechanical
with a C curve, there would be full microalloyed iron30% nickel Process Engineering: The University of
recrystallisation between passes if the austenite, which remains stable on Sheffield), was established by collafbo-
interpass times exceeded t0.85X. The quenching to room temperature. These ration between my group in the Depart-
intersection point was labelled the studies showed that strain induced ment of Engineering Materials (now
recrystallisation limit temperature. precipitation takes place on nodes in Department of Materials Science and
Above it the start of precipitation would the dislocation microband structure in Engineering), John Beynon of the
be severely retarded from the C cur- the deformed austenite.24 A conceptual Department of Mechanical Engineering
ves shown in the figure, because model based on these observations has and Derek Linkens of the Department of
recrystallisation removes the nucleation shown how incremental precipitation Automatic Control and Systems
sites for strain induced precipitation. can occur after sequential passes and Engineering. This was enthusiastically
Because full recrystallisation between lead to incremental strengthening, as supported by an Industrial Steering
passes in the roughing stage of rolling is shown by Fig. 22. Development of the Committee, showing the belief of its
required to refine the austenite grain model with more realistic kinetics is a member companies that modelling
size, the limit temperature represents topic of ongoing research. involving microstructural evolution is a
the lower limit of temperature for rough- valuable tool in optimising and control-
ing rolling, if mixed grain structures are ling thermomechanical processing
Conclusions
to be avoided. conditions.
This lecture has attempted to
At temperatures below the point of summarise some of the important The range of skills in IMMPETUS
intersection of t0.05X with the C curve, it steps in the transition from empirical meant that a broad philosophy of
is assumed that the strain induced skills based hot rolling of steel to modelling illustrated in Fig. 23 could
precipitates retard recrystallisation to computer modelling for optimisation be adopted. This has proved to be a
such an extent that it does not take and control of the process variables. It robust philosophy, which has enabled
place again for any interpass times in has concentrated on contributions the precision of data based black box
industrial rolling. This is labelled the made by the research group at the models to be combined in a hybrid
recrystallisation stop temperature and University of Sheffield to the basic approach with the physical metallurgy
represents the maximum temperature experimental work, on which the mo- based white box models to develop
for the start of finish controlled rolling dels are based, to the development of better understanding of new aspects of
so that deformation of the austenite the models, and to their experimental thermomechanical processing. Model
grains accumulates pass by pass. This and industrial validation. It has not tried predictions of product properties can
model successfully captured these to put this work in the context of con- now be more accurate, and provide
aspects of industrial rolling experience. temporary research by other groups, or more detailed information than sending
The second submodel is required to of the developments in computer sample specimens to the test house.
predict the kinetics and particle sizes of hardware and software that have been This marks the maturity, or at least the
strain induced precipitation in order to of major importance, but would require adolescence, of computer modelling of
compute the residual microalloy additional lectures to cover. Also, the thermomechanical processing.
element content in solution during modelling has not been brought up to
cooling and transformation, which is date, not because the research has Acknowledgements
important in controlling the mechanical stopped, quite the contrary, it is
properties of the product, and in order flourishing, but because I consider that The author is grateful to L. Sun, a
to compute the strengthening effects the transition period from trial and error research student in IMMPETUS, for his
on the flow stress during multipass to computer modelling had finished by assistance in the preparation of the
finish rolling. This is a much more the mid-1990s. figures for this lecture.

256 Ironmaking and Steelmaking 2011 VOL 38 NO 4


NEWS AND VIEWS

The paper is based on the 2010 Sir grain growth, 167187; 1986, Andersen et al.), 1326; 1992,
Henry Bessemer Lecture of the Iron and Roskilde, Riso National Labor- Roskilde, Riso National Labo-
Steel Society of the Institute of Materials, atory. ratory.
Minerals and Mining, given at the 10. C. M. Sellars and W. J. McG. 19. A. J. McLaren and C. M. Sellars:
Institute, London, on 7 October 2010. Tegart: Int. Metall. Rev., 1972, 17, Mater. Sci. Technol., 1992, 8,
124. 10901094.
References 11. G. Glover and C. M. Sellars: Metall. 20. L. Sun, K. Muszka, B. P. Wynne
Trans., 1973, 4, 765775. and E. J. Palmiere: Scr. Mater., to
1. C. Rossard and P. Blain: Rev. Met.,
12. L. Leduc: Rolling of titanium be published.
1958, 55, 573594.
bearing steels, PhD thesis, The 21. B. Dutta and C. M. Sellars: Mater.
2. J. J. Jonas, C. M. Sellars and W. J.
University of Sheffield, Sheffield, Sci. Technol., 1986, 2, 146153.
McG. Tegart: Metall. Rev., 1969,
UK, 1980. 22. B. Dutta and C. M. Sellars: Proc. Int.
14, 124.
13. R. A. Harding: Temperature and Conf. on Physical metallurgy of ther-
3. D. M. Keane, C. M. Sellars and W. J.
structural changes during hot rolling, momechanical processing of steels
McG. Tegart: Deformation under hot
working conditions, Special report PhD thesis, University of Sheffield, and other metals (Thermec-88),
no. 108, Iron and Steel Institution, Sheffield, UK, 1976. Tokyo, Japan, June1988, Iron and
London, UK, 1968, 2128. 14. J. H. Beynon and C. M. Sellars: ISIJ Steel Institution of Japan, 261268.
4. M. M. Farag, C. M. Sellars and Int., 1992, 32, 359367. 23. B. Dutta and C. M. Sellars: Mater.
W. J. McG. Tegart: Deformation 15. C. M. Sellars: The Hatfield mem- Sci. Technol., 1987, 3, 197206.
under hot working conditions, Spe- orial lectures, (ed. P. Beeley), 24. W. M. Rainforth, M. P. Black, R. L.
cial report no. 108, Iron and Steel In- Vol. III, 143164; 2005, Cam- Higginson, E. J. Palmiere, C. M. Sellars,
stitution, London, UK, 1968, 6067. bridge, Woodhead Publishing I. Prabst, P. Warbicler and F. Hofer:
5. D. Hardwick, C. M. Sellars and Limited. Acta Mater., 2002, 50, 735747.
W. J. McG. Tegart: J. Inst. Met., 16. C. M. Sellars: Mater. Sci. Technol., 25. C. M. Sellars and E. J. Palmiere: Proc.
196162, 90, 2122. 1990, 6, 10721081. Int. Conf. on Microalloying for new
6. M. J. Luton and C. M. Sellars: Acta 17. C. M. Sellars: Proc. Int. Conf. on steel processes and applications, San
Metall., 1969, 17, 10331043. Hot working and forming pro- Sebastian, Spain, September 2005,
7. T. Sakai and J. J. Jonas: Acta cesses, (ed. C. M. Sellars and Trans Tech Publications Ltd, 314.
Mater., 1984, 32, 189209. G. J. Davies), 315; 1980, London, 26. C. M. Sellars: Proc. XVth Physical
8. J. P. Sah, G. J. Richardson and Metals Society. Metallurgy and Materials Science
C. M. Sellars: Met. Sci., 1974, 8, 18. J. H. Beynon and C. M. Sellars: Conf. on Advanced materials and te-
325331. Proc. 13th Riso Int. Symp. on chnologies, Krakow-Krynica, Poland,
9. C. M. Sellars: Proc. 7th Riso Int. Materials science: modelling of May 1998, Institute of Metallurgy and
Symp. on Annealing processes- plastic deformation and its engi- Materials Science, Polish Academy
recovery, recrystallization and neering applications, (ed. S. I. of Sciences, Vol. 3, 100107.

Ironmaking and Steelmaking 2011 VOL 38 NO 4 257

Вам также может понравиться