Вы находитесь на странице: 1из 11
Seediscussions,stats,andauthorprofilesforthispublicationat: https://www.researchgate.net/publication/254498988 A combined

Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/254498988

Article in IOPConferenceSeriesMaterialsScienceandEngineering·July2012

DOI:10.1088/1757-899X/33/1/012026

CITATIONS

10

4authors:

9 PUBLICATIONS 43 CITATIONS

NationalUniversityofIreland,Galway,Ireland9 PUBLICATIONS 43 CITATIONS SEEPROFILE MingmingTong 34 PUBLICATIONS 275 CITATIONS SEEPROFILE

34 PUBLICATIONS 275 CITATIONS

READS

52

30 PUBLICATIONS 168 CITATIONS

132 PUBLICATIONS 1,143 CITATIONS

Someoftheauthorsofthispublicationarealsoworkingontheserelatedprojects:

Phasetransformationsduringprocessingofmetallicalloys. Viewproject Viewproject

ECFP7MINTWELDprojectwithTataSteelasleaderofindustrialpartners Viewproject Viewproject

AllcontentfollowingthispagewasuploadedbyWajiraU.Mirihanageon03July2017.

Theuserhasrequestedenhancementofthedownloadedfile.Allin-textreferencesunderlinedinblueareaddedtotheoriginaldocument

andarelinkedtopublicationsonResearchGate,lettingyouaccessandreadthemimmediately.

A combined enthalpy / front tracking method for modelling melting and solidification in laser welding

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2012 IOP Conf. Ser.: Mater. Sci. Eng. 33 012026

(http://iopscience.iop.org/1757-899X/33/1/012026)

Download details:

IP Address: 173.232.20.173

This content was downloaded on 03/07/2017 at 12:21

Please note that terms and conditions apply.

You may also be interested in:

G Duggan, M Tong and D J Browne

T. Koseki, H. Inoue, Y. Fukuda et al.

W U Mirihanage, D J Browne, L Sturz et al.

L Sturz and G Zimmermann

M A Martorano and V B Biscuola

G Zimmermann, L Sturz, B Billia et al.

Zhiye Chen, Sonja Arnsfeld and Dieter Senk

T Carozzani, H Digonnet and Ch-A Gandin

MCWASP XIII

IOP Publishing

IOP Conf. Series: Materials Science and Engineering 33 (2012) 012026

doi:10.1088/1757-899X/33/1/012026

A combined enthalpy / front tracking method for modelling melting and solidification in laser welding

G Duggan 1 , WU Mirihanage 2 , M Tong 1 , DJ Browne 1

1 School of Mechanical and Materials Engineering, University College Dublin, Belfield, Dublin 4, Ireland

2 Department of Physics, Norwegian University of Science and Technology, Trondheim, Norway.

E-mail: gregory.duggan@ucd.ie

Abstract. The authors present an integrated meso-scale 2D numerical model for the simulation of laser spot welding of a Fe-Cr-Ni steel. The melting of the parent materials due to the applied heating power is an important phenomenon, leading to the formation of the weld pool and the subsequent conditions from which solidification proceeds. This model deals with the dynamic formation of the weld pool whereby melting may be occurring at a given location while solidification has already commenced elsewhere throughout the weld pool. Considering both melting and possible simultaneous solidification in this manner ensures a more accurate simulation of temperature distribution. A source based enthalpy method is employed throughout the calculation domain in order to integrate the melting model with the UCD front tracking model for alloy solidification. Melting is tracked via interpolation of the liquidus isotherm, while solidification is treated via both the tracking of the advancing columnar dendritic front, and the nucleation and growth of equiaxed dendrites using a volume-averaging formulation. Heterogeneous nucleation is assumed to take place on TiN grain refiner particles at a grain refiner density of 1000 particles per mm 2 . A mechanical blocking criterion is used to define dendrite coherency, and the columnar-to-equiaxed transition within the weld pool is predicted.

1. Introduction The solid-to-liquid phase transformation, i.e. the melting of the parent material in the initial stages of welding, leads to the formation of the weld pool. The geometry and temperature distribution within the weld pool, once the heat input has stopped, have a significant influence on the subsequent microstructural development (e.g. columnar-to-equiaxed transition - CET) as solidification proceeds.

A key concern then is the accurate modelling of the weld pool formation as this provides the starting

point from which solidification commences. This paper focuses primarily on the development of a dynamic phase transformation scheme within the weld pool. To this end, fluid flow within the weld

pool is not considered at this time but it is acknowledged that this will be an important step to be taken

in future work to allow the results to be compared to experiments.

The welding process may be split into 3 stages. Initially there is pure melting of the parent material as a result of the heat input. This is followed directly by a period whereby both melting and solidification may occur throughout the weld pool simultaneously, as is discussed in further detail in section 4. Lastly there is pure solidification within the weld pool when the melting front has ceased to advance into the parent material. The first and last stages will be outlined in sections 2 and 3 followed by the combined melting and solidification phase in section 4.

Published under licence by IOP Publishing Ltd

1

MCWASP XIII

IOP Publishing

IOP Conf. Series: Materials Science and Engineering 33 (2012) 012026

doi:10.1088/1757-899X/33/1/012026

2. Melting

The field of research on melting is a relatively new one when compared to research on solidification. There is a tendency to try to adapt knowledge directly from the much more extensively developed research on solidification. However it has been shown in experiments that there exist asymmetries between melting and solidification. To the authors knowledge, there has yet to be a melting model developed based on the kinetics of the process although there is research being carried out in the area [1-3]. As this is the case, a widely used form of the enthalpy scheme, developed by Swaminathan and Voller [4], is implemented to model the initial melting stage of the welding process.

2.1. Enthalpy Scheme

Using an enthalpy scheme allows the calculation domain to be considered as a continuum, avoiding the necessity to explicitly address the discontinuity associated with phase change at the solid/liquid interface. The apparent specific heat capacity method is used to model the melting phase transformation that occurs during the heating of the work-piece via the laser heat input. The laser heat input is modelled as a Gaussian distribution.

3. Solidification

In order to model solidification the energy equation, shown in the discretised form, in equation 1, is solved explicitly over a fixed grid to model heat transfer and provide the temperature profile across the domain. A source term, shown in equation 2, is included to formulate the release of latent heat. A truncated Scheil equation is used to model the change in solid fraction due to the growth of grains and the thickening of the mushy zones.

where ρ is density, C p is the specific heat, ∆x and ∆y are the grid spacing, t is time, T is temperature, ∆a is the area the flux q is acting upon i.e. ∆x or ∆y in the 2D case. E is the source term associated with the release of latent heat, V m is the captured volume/ dendritic mushy volume in the current control volume (CV) while g s is the solid fraction within this mushy volume, the subscripts m and s stand for mushy zone and solid phase respectively. The superscripts n, n+1 and n-1, refer to the current time step, the next time step and the previous time step respectively. The 1 st term at right hand side of equation 2 may be considered to be the latent heat release due to the thickening of the mushy zone (E t ) while the 2 rd term is the latent heat release due to the advancement (E a ) of solidification front.

3.1. Columnar Front Tracking

The front in question is defined at a mesoscopic level; it encompasses the growing dendrite tips. The front is defined by massless markers at the points where the interface intersects the underlying fixed orthogonal grid. These markers are advanced into the liquid melt, as solidification proceeds, based on the local level of undercooling. The Burden and Hunt model [5] is used to calculate the dendrite tip

growth velocities (alternative growth laws may be easily incorporated). The details of this front tracking scheme are fully outlined elsewhere [6,7].

3.2. Equiaxed Grain Growth

The model for equiaxed grain growth as fully outlined elsewhere [8] is based on a volume averaging approach. Key to the model is the nucleation of equiaxed grains from inoculants within the undercooled liquid ahead of the columnar front. These inoculants are assumed to follow a log-normal size distribution having a uniform density distribution. As in the columnar front tracking case, the equiaxed envelope is assumed to grow via a dendrite tip growth kinetics model based on the level of

2

MCWASP XIII

IOP Publishing

IOP Conf. Series: Materials Science and Engineering 33 (2012) 012026

doi:10.1088/1757-899X/33/1/012026

undercooling in the liquid. Growth of the equiaxed envelopes as well as thickening of the inter- dendritic mushy zones is incorporated into the overall latent heat calculation for both columnar and equiaxed growth.

3.3. Inoculant particles The development of effective grain refiners for steel alloys, in order to improve mechanical properties, has been recognized as an area of research with great potential for cost savings. New applications call for ever increasing levels of control over the mechanical properties of steel. Research into the use of grain refiners in cast steels has been undertaken sparingly in comparison to the extensively studied field of grain refiners in aluminium alloys. Some of the earliest research pertaining to steels was carried out by Bramfitt [9] in 1970, who evaluated the effectiveness of various powders as inoculant particles in steel based on the level of undercooling required to initiate solidification. TiN, TiC and SiC were found to be the most effective inoculant particles. In recent years this area is becoming the focus of intensive research. TiN grain refiners are the most commonly studied for a variety of steels [10-13]. In welding an equiaxed grain structure is desirable as opposed to the typical coarse grained, low- ductility, fully columnar grain structure found in many welds. However, fully columnar structure is typical of most welds due to the very high thermal gradients and the speed of the solidification front. By including inoculant particles within the weld material, heterogeneous nucleation may be promoted in the undercooled liquid ahead of the columnar front, thus a CET may be initiated. Work carried out by Koseki et al [14], on the weld solidification of Fe-Cr-Ti alloys in the presence of TiN grain refiners, showed that CET may be initiated with the correct level of grain refiner density present in the weld material. The nucleation and growth of equiaxed grains was shown to result from the combined effects of the heterogeneous nucleation sites provided by the TiN inoculant particles as well as the increased undercooling due to the soluble titanium in the weld material. The present work, although not considering the effects of soluble titanium in the weld material, uses a similar grain refiner density and size distribution of TiN particles as those used in Koseki et al [14] and Park [15].

4. Integration of melting and solidification routine

Once the heating power of welding is switched off, the free surface of liquid metal will begin to cool quickly due to radiative/ convective heat loss to the surroundings. This results in solidification commencing at the outermost parts of the weld pool. While this is occurring, heat transferred from superheated liquid to the solid phase is still proceeding resulting in further melting. This means melting and solidification of the parent material co-exist for some finite period once the heating power is switched off as shown in figure 1.

once the heating power is switched off as shown in figure 1. Figure 1. Melting and

Figure 1. Melting and solidification occurring simultaneously within the weld pool.

The phase transformation model has to be able to track both melting and solidification simultaneously. The original front-tracking model for solidification as outlined in section 3, has been updated in order to consider melting. To facilitate the application of a single set of energy equations to the calculation domain to treat both melting and solidification simultaneously, a source term based enthalpy method is employed to replace the apparent specific heat capacity method previously outlined in section 2.

3

MCWASP XIII

IOP Publishing

IOP Conf. Series: Materials Science and Engineering 33 (2012) 012026

doi:10.1088/1757-899X/33/1/012026

The energy equation for the front tracking formulation, shown by equation 1, is still valid in the source term based enthalpy method. The source term E is split into 2 components E a and E t . In the case of melting, E t becomes the absorption of latent heat due to the change of volume fraction of solid phase. The contribution of latent heat due to advancement of solidification front E a does not exist in

melting. Therefore, E is calculated in equation 3 by appropriately selecting the source terms E a and E t as shown in equations 4 and 5, based on whether melting or solidification is occurring within a given

CV.

(3)

(4)

(5)

where V cv is the volume of a CV. The existing numerical scheme of assigning and advancing computational markers [6,7] during the solidification process cannot be applied in the case of melting. Therefore a new numerical scheme for marker assignment and advancement to track the boundary of the weld pool with a Lagrangian grid of computational markers, has been developed as follows. Once the heat input is stopped, the initial front is calculated by assuming it is given by the liquidus isotherm (T L ). The temperatures along the walls of every CV in the simulation domain are scanned using bilinear interpolation. Once T L is encountered, a computational marker is assigned. This scanning procedure is performed along both the horizontal and vertical directions. Using linear interpolation along the correct sequence of markers, a proper front is defined as illustrated in figure 2. This procedure is only implemented once, at the moment the heating input ceases.

implemented once, at the moment the heating input ceases. Figure 2. Assigning initial markers on grid

Figure 2. Assigning initial markers on grid lines (vertical gridlines 2 x for clarity).

Once the front is defined, during the following process of the simulation, the temperature of each marker can be calculated by interpolating temperature values from the surrounding node temperatures.

4

MCWASP XIII

IOP Publishing

IOP Conf. Series: Materials Science and Engineering 33 (2012) 012026

doi:10.1088/1757-899X/33/1/012026

Markers are designated as “Enthalpy Markers” if their temperature is greater than or equal to T L , while makers with temperatures below T L i.e. undercooled markers, are designated “Front Tracking Markers”.

4.1. Updating temperature field

Two types of CV designations are used in order to ensure the correct source terms are used when updating the temperature field. A CV will be designate as an “Enthalpy CV” if it is intersected by an “Enthalpy Marker” where the next marker downstream of this is also an “Enthalpy Marker” as illustrated in figure 3.

is also an “Enthalpy Marker” as illustrated in figure 3. Figure 3. Assigning of a CV

Figure 3. Assigning of a CV types based on the type of markers intersecting it.

A CV containing “Front Tracking Markers” will be designated a “Front Tracking CV”, while CV’s which do not contain markers are by default set to “Enthalpy CV”. The mushy volume fraction i.e. the volume fraction captured by the front in a CV, is calculated based on the location of intersecting markers for CV’s designated “Front Tracking CV” only. Details of the way to calculate the volume fraction of mushy zone can be found in the work of Browne and Hunt [6]. This mushy volume fraction is incorporated in the source terms as outlined in equation 4. Once the temperature field has been updated the marker temperatures are updated as well as the marker types and CV types accordingly. The model then updates the time and repeats the process again until the specified end time.

4.2. Advancing markers The Burden and Hunt dendrite tip growth kinetics model [5] is used to advance markers appropriately based on level of marker undercooling (T L - T marker ). Only markers that are designated “Front Tracking Markers” are advanced in this manner. A marker will advance (assuming it is undercooled) into the liquid at the local normal to the front. Once markers have been advanced, the position of new markers are redefined following the procedures outlined in [6,7]. The temperatures of the new markers are redefined and the markers are designated Enthalpy / Front Tracking appropriately. The CV types are also updated at this time The new locations of Enthalpy markers are also calculated at this stage by interpolating the new position of the T L isotherm. If melting is occurring at a marker location, it will move outwards to the new grid/ T L intersection point. Enthalpy markers will proceed in this manner as the melting begins to slow until a point where the marker becomes undercooled as its type will be switched to “Front Tracking Marker” and it will proceed to grow into the liquid as described previously.

5. Columnar to Equiaxed Transition Columnar and equiaxed volume fractions are calculated separately in each CV. In order to predict CET, three blocking criteria are implemented [8]:

5

MCWASP XIII

IOP Publishing

IOP Conf. Series: Materials Science and Engineering 33 (2012) 012026

doi:10.1088/1757-899X/33/1/012026

1. If the average equiaxed dendrite diameter is less than the primary columnar inter-dendrite arm spacing the columnar front will continue to grow. The primary columnar inter-dendrite arm spacing is calculated according to the empirical relation outlined in [16].

2. If the average equiaxed dendrite diameter is greater than the primary columnar inter-dendrite arm spacing, but no coherent equiaxed dendrite network has formed, then equiaxed dendrites are expected to grow simultaneously ahead of the columnar front

3. A coherent equiaxed dendrite network has formed; this will act as a barrier to physically block the columnar front, i.e. CET will occur.

6. Details of case study

The sample size is 20 mm by 5mm. The heat flux from the laser is assumed to follow a Gaussian distribution. The power of the laser is 1404 W, the radius is 2.5 mm and the heating duration is 0.80 s. Heat loss due to convection and radiation is assumed through the boundaries of the sample. An emissivity value of 0.5 is assumed and a heat transfer coefficient of 8 W/m -2 K [17] is assumed

between the air and the sample. The air temperature is assumed to be 300 K. The alloy used is Fe-15.9 wt% Cr -14.1 wt% Ni; thermophysical properties used are listed in table 1. The grain refiner density was typically taken to be 1000 particles per mm 2 [14], while this value was varied in order to assess the effectiveness of the grain refiners in promoting CET.

Table 1. Thermophysical properties used in the case study

Property

Value

Unit

Source

Eutectic temperature (T e) Liquidus temperature (T L) Specific heat (C p) Density (ρ) Thermal Conductivity (K ) Latet heat of Fusion (L) Diffusivity of Cr in liquid (D l) Solid/liquid interfacial energy (γ) Volumetric entropy of fusion (∆S v ) Gibbs Thomson coefficient (Γ) Liquidus Slope (m L ) Partition coefficient (α) Burden and Hunt growth coefficient (C)

1679

K K J/kgK kg/m 3 W/mK J/kg m 2 /s J/m 2 J/m 3 K mK K/wt% - m/K 2 s

[18]

1710

[18]

780

[13]

7250

35

1.854x10 5

1.8x10 -9

[13]

0.299

[18]

7.71x10

5

[18]

3.88x10

-7

[18]

-2.9

[18]

0.86

[18]

0.8x10 -5

[18]

7. Results and Discussion

7.1. Melting For the initial melting stage, the laser heat input lasts for a duration of 0.8s. During this period the weld pool develops to a width of 4.8mm and a depth of 1.12mm. As is apparent in figure 4, the high rate of heating results in a very narrow (yellow) mushy zone between the T s and T L isotherms. The low thermal conductivity of the alloy (relative to a typical value in laser spot welding of an aluminium alloy [19]) means the weld pool retains heat very well and it is slow to dissipate through the free surface and the base material. This results in a wide shallow weld pool. The initial position of the front is shown in both figures 5 and 6 via the solid blue line, while the advanced position of the front at two subsequent time steps are depicted via the red dashed lines. Melting at the base of the weld pool is evident while solidification proceeds along the upper sides. There is a significant amount of continued melting for a further 0.06s from when the laser is switched off. The model accurately reproduces this important stage of the process whereby the weld pool is allowed to fully develop before solidification proceeds at all points.

6

MCWASP XIII

IOP Publishing

IOP Conf. Series: Materials Science and Engineering 33 (2012) 012026

doi:10.1088/1757-899X/33/1/012026

33 (2012) 012026 doi:10.1088/1757-899X/33/1/012026 Figure 4. Weld pool at 0.8 s, heat input just turned off.

Figure 4. Weld pool at 0.8 s, heat input just turned off. Narrow yellow, mushy region apparent.

input just turned off. Narrow yellow, mushy region apparent. Figure 6. Further development of the weld

Figure 6. Further development of the weld pool as melting persists at the base of the weld pool at 0.84 s.

as melting persists at the base of the weld pool at 0.84 s. Figure 5. Advancement

Figure 5. Advancement of melting front at base of weld pool, while solidification proceeds at the sides at 0.82 s.

pool, while solidification proceeds at the sides at 0.82 s. Figure 7. Undercooled liquid region (white)

Figure 7. Undercooled liquid region (white) apparent ahead of the columnar front at 0.86 s.

7.2. Solidification The progression of the columnar front is plotted in figures 7-9. The undercooled liquid region is illustrated via the white area ahead of the columnar front in figure 7. As can be seen this undercooled region is initially quite narrow, this is due to the low conductivity of the alloy which maintains a high thermal gradient ahead of the columnar front. While this is the case, any equiaxed grains that do nucleate ahead of the columnar front are quickly overgrown as per criterion number 2 in section 5. These equiaxed grains are given little opportunity to grow and reach coherency in order to block the columnar growth. The slow dissipation of heat within the weld pool maintains high thermal gradients until at a later stage, the thermal gradients have fallen enough to allow a significant undercooled liquid region to develop. Figures 8 and 9 show the contours of solid fraction within the weld pool, blue (1.0) being fully solid. The position of the columnar front (white dashed line) illustrates the level of

7

MCWASP XIII

IOP Publishing

IOP Conf. Series: Materials Science and Engineering 33 (2012) 012026

doi:10.1088/1757-899X/33/1/012026

equiaxed growth in the undercooled region ahead of the columnar front. The conditions within the weld pool are now such that the equiaxed grains can grow to the point of coherency and block the columnar front. CET occurs at approximately 0.95s. The remaining liquid within the weld pool from this point on solidifies with an equiaxed microstructure. The location of the columnar front at the point of CET is shown in figure 9 via the white dashed line.

point of CET is shown in figure 9 via the white dashed line. Figure 8. Equiaxed

Figure 8. Equiaxed growth apparent ahead of columnar front at 0.92 s.

Equiaxed growth apparent ahead of columnar front at 0.92 s. Figure 9. Undercooled liquid at 0.96

Figure 9. Undercooled liquid at 0.96 s, while the position of CET is indicated in the solid (note change in legend).

These results agree well with those observed in [14] where CET was seen to occur during the welding of a similar alloy with the same grain refiner density. It is also evident that reducing the grain refiner density results in CET occurring later in the process. This model directly predicts the microstructure of the weld and the presence of CET without having to rely on plots of the thermal gradient (G) against the local solidification velocity (v). Typically G/v plots have been used in conjunction with the Hunt Diagram [20] in order to infer the type of microstructure present e.g. [21]. The model presented here provides a more direct representation of how the microstructure changes throughout the weld pool during solidification, based on the thermodynamics and kinetics of the process. At this stage these numerical results have not been validated against experimental laser spot welding results as a fluid flow model will need to be incorporated to generate meaningful results as regards the solidification kinetics, to compare to experiments.

8. Conclusion An integrated melting and solidification model of 2D laser spot welding is presented. The front tracking method has previously been successfully extended to 3D [22] with good results. The front tracked in this case consists of triangular surfaces as opposed to line segments. To carry out the extension to 3D in this case would require considerable programming effort but primarily a considerable amount of geometrical calculations would be required. The grid of surfaces defining the front would need to be continuously refined / coarsened in order to ensure a level of accuracy that would capture the important features of the weld pool. This paper focuses on developing the phase transformation model and thus the extension to 3D is left to future work. An enthalpy scheme is used to model the initial melting stages of the process, while a newly adapted UCD front tracking method is used to track subsequent melting and solidification. Equiaxed nucleation upon TiN grain refiner particles is considered and their subsequent growth to coherency is tracked in order to predict CET. A mechanical blocking criterion is employed to predict CET, which is

8

MCWASP XIII

IOP Publishing

IOP Conf. Series: Materials Science and Engineering 33 (2012) 012026

doi:10.1088/1757-899X/33/1/012026

seen to occur in the late stages of the process, when the thermal gradient has reduced enough to allow a significant undercooled liquid region to develop and promote the growth of equiaxed grains.

Acknowledgments This research work is supported by the European Commission as part of the FP7 programme, as the

project, Modelling of Interface Evolution in Advanced Welding; contract number no. NMP3-SL-2009-

229108.

References

[1]

[2]

Rettenmayr M, 2009 International Materials Reviews 54 1-17

[3]

[4]

Swaminathan CR and Voller VR, 1993 International Journal of Numerical Methods for Heat &

[5]

Fluid Flow 3(3) 233-44 Burden MH and Hunt JD, 1974 Journal of Crystal Growth 22 109-16

[6]

Browne DJ and Hunt JD, 2004 Numerical Heat Transfer B 45 395-419

[7]

McFadden S and Browne DJ, 2009 Applied Mathematical Modelling 33(3) 1397-1416

[8]

Mirihanage WU and Browne DJ, 2009 Computational Materials Science 46 (4) 777-84

[9]

Bramfitt BL, 1970 Metallurgical Transactions 1 1987-1995

[10]

[11]

Poole WJ and Weinberg F, 1998 Metallurgical and Materials Transactions 29A 885-861

[12] Wang M, Cheng G, Qiu S, Zhao P and Gan Y, 2010, International Journal of Minerals,

Metallurgy and Materials 17(3) 276-281 Löser W and Herlach DM, 1992 Metallurgical Transactions 23A 1585-1591

[13]

Materials 4 183-195 [15] Park JH, 2011 CALPHAD: Computer Coupling of Phase Diagrams and Thermochemistry 35

455-462

[16]

[17] Mohammad Hossenini H, Mirzaei M, Azari S, 2006 Modelling of Casting, Welding and

El-Bealy M and Thomas BG, 1996 Metallurgical and Materials Transactions 27B 689-693

Advanced Solidification Processes XI 903 Koseki T and Flemings MC, 1995 Metallurgical and Materials Transactions 26A 2991-2999

[18]

[19] Duggan G, Tong M and Browne DJ, 2011 IOP Conference Series : Materials Science and Engineering 27 012077

[20]

Hunt JD, 1984 Materials Science Engineering 65 75-83

[21]

Villafuerte JC, Pardo E and Kerr HW, 1990 Metallurgical Transactions 21A 2009-2019

[22] Seredynski, M, 2009 Micro-macroscopic model of binary mixture solidification, PhD thesis,

Warsaw University of Technology (in Polish)

9