ARTICLE IN PRESS

ICHMT-02110; No of Pages 6

International Communications in Heat and Mass Transfer xxx (2010) xxxxxx

Contents lists available at ScienceDirect

International Communications in Heat and Mass Transfer

j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / i c h m t

A. Abbas Nejad a, M.J. Maghrebi a, H. Basirat Tabrizi b,, Y. Heng c, A. Mhamdi c, W. Marquardt c

4
5
6

a
b
c

Department of Mechanical Engineering, Shahrood University of Technology, P.O. Box 3619995161-316, Shahrood, Iran
Department of Mechanical Engineering, Amirkabir University of Technology, P.O. Box 15875-4413, Tehran, Iran
AVT-Process Systems Engineering, RWTH Aachen University, D-52056 Aachen, Germany

i n f o

a b s t r a c t

RO

a r t i c l e

A transient nonlinear inverse heat transfer problem arising from alloy solidication processes is considered.
In practice, the solidus and liquidus interface motions and thus the mushy zone thicknesses are pre-given to
control the material quality. To achieve the desired front motions, the required time-dependent boundary
conditions have to be predicted on both mold sides simultaneously. In this study, the enthalpy method is
used for the derivation of governing equations. Hence, the inverse problem will be solved only in a single
spatial and temporal domain. The conjugate gradient method with adjoint equation is applied for the
resulting minimization problem. The method is applied as comparison for pure material with other previous
studies. Then, alloy material with different front velocities is set up to investigate the solidication process.
The obtained results show a close agreement between the desired and computed front motions and mushy
zone thickness.
2010 Elsevier Ltd. All rights reserved.

Available online xxxx

DP

Keywords:
Inverse heat Transfer
Alloy solidication
Conjugate gradient
Mushy zone thickness

TE

8
9
10
12
11
13
14
15
16
17
18

33
32

1. Introduction

35

Heat transfer problems involving melting and solidication appear

in many important material processing technologies such as welding,
casting and crystal growth. The scale of microstructures, crystallographic growth morphology and grain orientation in a solidication
process are closely related to the macroscopic features of the product.
During the solidication of pure material, the interfacial microstructures and solidied material quality largely depend on interface
velocity and the solidliquid side temperature gradient [1]. Hence, the
desired material structures with specied properties and qualities can
be achieved by controlling the interface velocity and temperature
gradient. Also the solidus and liquidus front motions and thus mushy
zone thickness are important factors which dene the crystallographic growth morphology, the scale of microstructures and the freckle
formation of alloys [1]. For example, freckles form in regions with high
local solidication times (LST). In other words, it happens in slow
solidifying processes with large mushy zone thickness. Therefore, the
thin mushy zones are needed to avoid freckle formation [2].
This research aims to study the control of front motion and mushy
zone thickness of alloys during solidication to reach a desired
material quality. Inverse heat transfer methods are applied to get
time-dependent boundary conditions on both sides of the mold which
lead to the desired front positions or velocities.

44
45
46
47
48
49
50
51
52
53
54
55
56

RR

42
43

CO

40
41

UN

38
39

Several numerical and experimental studies investigated the

control of solidication. Zabaras et al. [3] proposed an integral
method with boundary element approach to nd the required
boundary heat ux for a desired interface velocity. Zabaras et al. [4]
solved the inverse Stefan's design problem using the spatial
smoothing and modication of Beck's future information method
using nite element technique. Zabaras and Yuan [5] used a dynamic
programming methodology for the inverse design of Stefan's problem.
Zabaras and Kang [6] found the time history of boundary heat ux for
controlling the interface movement in a 2D geometry using adjoint
method. The effect of natural convection in liquid phase was
considered in several research works including Zabaras et al. [710].
All foregoing studies were based on front xing and front-tracking
nite element method which distinguish the two phases and solve the
problem for every phase, separately.
Frankel and Keyhani [11] and Hale et al. [12] proposed a weighted
residual method for the independent control of the interfacial
temperature gradient and velocity during solidication. Xu and
Naterer [1315] used experimental and numerical techniques to
control velocity and acceleration of a pure material. Their numerical
approach relies on an enthalpy formulation using a control volume
based nite element method and a non-gradient-based iterative
inverse solution. Voller [16] presented an enthalpy method based on a
xed domain. The discretized equation was obtained by a nite
volume formulation with future time-stepping. Recently, Okamoto
and Li [17] used sequential and whole domain with Tikhonov
regularization methods to predict the boundary heat ux. They used
a deforming nite element method to solve the governing equations.

EC

34

36
37

OF

Optimal operation of alloy material in solidication processes with inverse

heat transfer

Communicated by W.J. Minkowycz.

Corresponding author. Tel.: +98 21 64543455; fax: +98 21 66419736.
E-mail address: hbasirat@aut.ac.ir (H. Basirat Tabrizi).
0735-1933/$ see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.icheatmasstransfer.2010.03.002

Please cite this article as: A. Abbas Nejad, et al., Int. Commun. Heat Mass Transf. (2010), doi:10.1016/j.icheatmasstransfer.2010.03.002

19
20
21
22
23
24
25
26
27
28
29
31
30
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84

ARTICLE IN PRESS
2

A. Abbas Nejad et al. / International Communications in Heat and Mass Transfer xxx (2010) xxxxxx

T1:1

Nomenclature

T1:2
T1:3
T1:4
T1:5
T1:6
T1:7
T1:8
T1:9
T1:10
T1:11
T1:12
T1:13
T1:14
T1:15
T1:16
T1:17
T1:18
T1:19
T1:20
T1:21
T1:22

a
b
c
d
e
fs
H
k
L
q
s
Ste
t, tf
T
Tf
Tliq
Tsol
T0
V
x

T1:23
T1:24
T1:25
T1:26
T1:27
T1:28
T1:29
T1:30
T1:31
T1:32
T1:33

Greek symbols

thermal diffusivity [m2/s]

search step size

small variation

threshold parameter

Lagrange multiplier

conjugate coefcient

adjoint variable

measurement error
s[q(t)] objective function gradient

T1:34
T1:35
T1:36
T1:37
T1:38
T1:39

subscripts
i
initial value
l
liquid phase
RMS
root mean square
s
solid phase

T1:40
T1:41
T1:42

superscripts
k
iteration number

88
89
90
91
92
93
94
95
96
97
98
99
100
101

122

A schematic representation of the unidirectional diffusion-dominated

solidication of alloy material is illustrated in Fig. 1. There are two moving
interfaces namely solidus and liquidus which need to be controlled to
achieve the desired material quality. The governing equations of the
problem in enthalpy form are non-dimensional form as follows:

123
124

DP

RO

OF

2. Formulation of the control problem

x=

104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121

125
126
127

x
t
TTf P
H
;
;H =
; t = 2l ; T =
Tf T0
a
cl Tf T0
a
L
qt a
k
;q=
;k =
cl Tf T0
kl Tf T0
kl

TE

Ste =

RR

EC

where, a is the mold length, Tf, Tsol and Tliq are fusion, solidus and 128
129
liquidus temperatures and L is the latent heat of the fusion. l, Cpl and 130

CO

86
87

102
103

Hinze and Ziegenbaly [18] controlled the evolution of the interface

using a steepest descent method.
Most of the studies cited in the literature used a deforming grid and
solved two sets of equations for liquid and solid, separately. Only a few
studies were based on a xed grid solution (enthalpy method) using
Beck's method. All the investigations devoted to the pure materials which
have a distinct phase change at a specied phase change temperature and
an interface between solid and liquid phases. Some of the previous
researchers studied the solidication of eutectic alloy material where the
mushy zone was ignored. They solved inverse problem for solid and
liquid zones with one interface between them [7,8].
In this study, the solidication control problem of alloy material is
formulated as a nonlinear inverse heat transfer problem. The enthalpy
formulation does not consider phases, separately. The control
problem is solved in a single spatial and temporal domain. Three
zones including solid, liquid and mushy with two different interfaces
solidus line and liquidus line are considered for alloy material.

UN

85

mold length [m]

non-dimensional mold length
Heat capacity [J/kg K]
direction of descent
error
solid fraction
enthalpy [J/kg]
thermal conductivity [W/m K]
Latent heat of fusion [J/kg]
heat ux [W/m2]
objective function
Stefan number Ste = L/cl(Tf T0)
time, nal time [s]
temperature [K]
melting point [K]
liquidus temperature [K]
solidus temperature [K]
reference temperature [K]
front velocity
spatial coordinate [m]

To control the front motion, the temperature must be the fusion

temperature at the desired solid liquid interface for pure material.
Also the solidus and liquidus temperatures must maintain at the
desired solid mushy (solidus) and mushy liquid (liquidus) front
locations at desired times. Hence, an objective function is dened for
the control problem based on the difference between the fusion,
solidus and liquidus temperatures and computed temperatures at the
desired front positions. The conjugate gradient method is applied to
minimize the objective function. The optimal boundary heat uxes on
both sides of the mold are computed simultaneously to achieve
the desired mushy zone thickness. Several case studies for different
interface motions are sought.
The paper is organized as follows. In Section 2, the mathematical
problem formulation in enthalpy form is given and the control task is
stated in terms of an inverse heat transfer problem. In Section 3, a
conjugate gradient based optimization approach for the solution of
the formulated control problem is preformed. In Section 4, two
simulation case studies for pure and alloy material are set up to
validate and assess the performance of the solution approach. Finally,
in Section 5 some concluding remarks are given.

Fig. 1. Schematic problem representation for alloy material.

Please cite this article as: A. Abbas Nejad, et al., Int. Commun. Heat Mass Transf. (2010), doi:10.1016/j.icheatmasstransfer.2010.03.002

ARTICLE IN PRESS
A. Abbas Nejad et al. / International Communications in Heat and Mass Transfer xxx (2010) xxxxxx

133
134

kl are thermal diffusivity, heat capacity and thermal conductivity of

the liquid phase, respectively. The general non-dimensional form of
the governing heat transfer equations where the sign over the nondimensional parameters neglected are as follows [19]:



H

T
=
kT
t
x
x

135
136

T
x

137
138

T
x

x=0

= qs t

0 b x b b 0 b t tf ;

or H = Hi

0 b t tf ;

1c

0xb t = 0:

1d

h
i
k
k
di t = s qi t ; for k = 0;
h
i
k
k
k k1
di t = s qi t + i di t ; for other k;

Solid region
Mushy region
Liquid region

!
TTsol
:
Tliq Tsol

149
150

where, is the specied tolerance.

TE

kT = fs T ks + 1fs T kl :

163
164
165
166
167
168
169
170
171
172
173
174
175
176

sqs ; ql  =
177
178
179
180

t
2
0f Tsol T xs ; t; qs ; ql  dt

t
0f

h
i2
Tliq T xl ; t; qs ; ql dt: 5

Now, the conjugate gradient method is used to determine the

unknown heat uxes by minimizing the objective functions Eq. (5).

193
194
195
196
197
198
199
200

9
201
202

EC

RR

161
162

In a direct problem, the temperature distribution in the domain

and interface locations can be determined by using known boundary
heat uxes. The considered control problem corresponds to the
inverse problem, where the heat uxes qs(t) and ql(t) are unknown
functions and should be determined to impose a desired interface
motion (velocity or acceleration).
From the physical behavior of the alloy solidication, we know
that the temperatures are solidus and liquidus temperatures at the
solidmushy and mushyliquid interfaces, respectively. This is the
fusion temperature at solidliquid interface for pure material.
Therefore, controlling the interface motion requires that the temperature at the desired front locations remain at the fusion, solidus or
liquidus temperatures (Tf, Tsol, Tliq).
For given qs, ql, let T(xs, t; qs, ql) and T(xl, t; qs, ql) denote the
temperature eld which is determined from the solution of the direct
problem at the desired solidus and liquidus locations, respectively.
Thus, the solution of the inverse problem to determine the boundary
heat uxes qs and ql can be obtained by minimizing the L2-norm of the
difference between the calculated temperature at the desired solidus
and liquidus interfaces (xs, xl) and solidus and liquidus temperatures.
In other words, it can be obtained by minimizing the following
objective function.

CO

159
160

At each iteration step, two quantities need to be evaluated namely

optimal search step size (ki ) and the gradient of boundary heat uxes
(s[qki (t)]). Their evaluation requires solving the sensitivity and
adjoint problems. The derivation of the sensitivity and adjoint
problems is presented in the following sections.
The following criterion is chosen to stop the iterative procedure:

h
i
k + 1
t b ;
s q

UN

157
158

n h
i
h
io h
i
t
0f s qki t s qk1
t s qki t dt
i
n 
:
=
2 o
t
dt
t
0f s qk1
i

The thermal conductivity k(T)can be written as a function of solid

fraction or temperature:

153
154
155
156

0i = 0
ki

with Hs = cs T, Hl = cl T + Ste, cs;l = cs;l = cl , k (T) = (T)/kl and Tsol

and Tliq are the solidus and liquidus temperatures, respectively and fs
is the solid fraction obtained from following formula:

fs = 1

151
152

189
190

The conjugate coefcient i is computed by the following 191

expression:
192

DP

147
148

TbTsol
Tsol TTliq
T N Tliq

RO

The enthalpy-temperature relations are dened in Eq. (2) as

follows:
8
< cs T
H x; t = fs Hs + 1fs Hl
:
cl T + Ste

145
146

where k denotes the iteration number. The iteration is started by using 185
186
q0i (t) = 0. The direction of descent dki (t) is updated from the following 187
formula:
188

141
142
143
144

k k

t = qi t i di t ; i = s; l; t N 0

1b

139
140

T = Ti

k + 1

qi

= ql t x = b

x=b

For the computation of the unknown heat uxes qs(t) and ql(t) by 182
the conjugate gradient method, the following general iteration 183
process is performed.
184

1a

x = 0 0 b t tf ;

181

3. Conjugate gradient algorithm

OF

131
132

3.1. The sensitivity problem

203

The sensitivity problem is obtained by perturbing the temperature

and enthalpy T(x, t) and H(x, t) by a small amounts [19,20].
Replacing T(x, t) by [T + T], H(x, t)by [H + H], qs by [qs + qs] and
ql by [ql + ql] in the direct problem (Eqs. (1a)(1d)) and subtracting
the original direct problem from the resulting expressions, one
obtains:

204

C T T
2
= 2 kT T
t
x

kT
x

kT
x

0bxbb

0 b t tf ;

205
206
207
208
209

10a
210
211

x=0

= qs t x = 0 0 b t tf ;

10b
212
213

x=b

= ql t x = b

0 b t tf ;

10c

T = 0 0 x b t = 0 :

10d

214
215
216
217
218

From Eq. (2):

8
cs
>
>
>
<

f
C T = fs cs + 1fs cl + s cs Tcl TSte
>
T
>
>
:
cl

TbTsol

Solid region

Tsol TTliq

Mushy region

T N Tliq

Liquid region

Please cite this article as: A. Abbas Nejad, et al., Int. Commun. Heat Mass Transf. (2010), doi:10.1016/j.icheatmasstransfer.2010.03.002

11

ARTICLE IN PRESS
4

where cl = 1. To obtain the optimal search step size, the solution of

Eqs. (10a)(10d) for Tis used as follows.
8
F1 A22 F2 A12
>
>
>
< A A A2 ; i = s
11 22
12
k
i =
>
> F2 A11 F1 A12
>
;i = l
:
A11 A22 A212

13d

tf 
tf 


k
k
k
k
F2 = Txs Tsol Tl;xs dt + Txl Tliq Tl;xl dt;
0

13e

Here Ts and Tl are the solutions of the sensitivity problems

obtained by setting q s = d ks , q l = 0 and q l = d kl , q s = 0,
respectively.

237

3.2. The adjoint problem

238

To derive the adjoint problem, rst multiply the direct governing

equations by Lagrange multiplier(x, t), then integrate over spatial
and temporal domains and add the resulting expression to the cost
functional (Eq. (5)).

0f 10 x; t k

T
H
dxdt:

t
x2

i2
Tliq T xl ; t; qs ; ql dt +

14

243
242

246
247
248

The following adjoint problem is obtained by replacing T by [T + T],

qs by [qs + qs], ql by [ql + ql] and s[qs, ql] by [s[qs, ql] + s[qs, ql]] in
Eq. (14), subtracting the result from Eq. (14), using the boundary and
initial conditions and further allowing terms containing T(x, t) to
vanish,:

CO

244
245

252
251

UN

264
265
266
267

In this section, rst we introduce the method used for the solution
of direct problem for the validation purpose with some other
numerical simulations. Then the performance of the proposed inverse
approach is investigated by the solution of the inverse control
problem for a pure material which has one moving interface between
solid and liquid phases. The obtained solution is compared with
previous analytical [21] and numerical [3,16] solutions. Then, the
problem involving alloy solidication with two moving boundaries is
considered. The obtained solution is validated by comparing the
desired and computed front positions due to the lack of any other
numerical and experimental studies for controlling solidus and
liquidus front positions or mushy zone thickness.
The compact third order Runge Kutta along with central nite
difference method is applied to solve the direct, sensitivity and adjoint
equations. Temporal and spatial grid dimensions are assigned
according to the front velocities. Detailed discussion about the
approach and the results can be found in Abbas Nejad et al. [22].
To evaluate the performance of the direct solver, the solution of
direct PbSn alloy solidication problem with known boundary
conditions is compared with an existing numerical approach [23].
The comparison of the temperature history at x = 0.37 in Fig. 2 and the
solidus and liquidus front positions are illustrated in Fig. 3.
A special case of pure material solidication with constant velocity
at the interface with Ti = Tf = 0, Ste = 0.5 is considered to study the

271
272

x; t
x; t
+ kT
C T
+ 2fTTsol xxs +
t
x2
0bxbb
h
i
TTliq xxl g = 0
250
249

18b

270

EC

RR

"

t
0f

18a

4. Results and discussion

TE

235
236

t
2
0f Tsol T xs ; t; qs ; ql  dt

17

The above equations are used to determine the direction of 268

descent and conjugate coefcient from Eqs. (7) and (8), respectively. 269

tf 
tf 


k
k
k
k
F1 = Txs Tsol Ts;xs dt + Txl Tliq Ts;xl dt;

sqs ; ql  =

sqs t  = 0; t ;

13c

258
259

260
261

sql t  = b; t :

tf

sq = 0f qsqt dt:

OF

13b

A12 = Ts;xs Tl;xs dt + Ts;xl Tl;xl dt;

241

16

Comparing the last two equations yields to the expressions for 262
calculation of the boundary heat ux gradients:
263

tf 
tf 
2
2
k
k
A22 = Tl;xs dt + Tl;xl dt;

tf

239
240

13a

227
226

232
233
234

where, is the Dirac delta function. After eliminating the terms 255
256
consisting T(x, t), only the following integral term is left.
257

since q(t) a L2(0, tf), one can write:

231
230

15d

where

229
228

254
t = tf ; 253

sqs ; ql  = 0f 0; t qs dt + 0f b; t ql dt:

tf 
tf 
2
2
k
k
A11 = Ts;xs dt + Ts;xl dt;

225
224

0xb

RO

223
222

12

=0

DP

219
220
221

A. Abbas Nejad et al. / International Communications in Heat and Mass Transfer xxx (2010) xxxxxx

0bttf ;

15a

=0
x

x = 0 0bttf ;

=0
x

x = b 0bttf

15b

15c

Fig. 2. Comparison of temperature history at x = 0.37.

Please cite this article as: A. Abbas Nejad, et al., Int. Commun. Heat Mass Transf. (2010), doi:10.1016/j.icheatmasstransfer.2010.03.002

273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294

ARTICLE IN PRESS
A. Abbas Nejad et al. / International Communications in Heat and Mass Transfer xxx (2010) xxxxxx

297

effectiveness of the proposed inverse approach. An inverse analytical

solution for the boundary heat ux can be obtained from the general
solution [21]:
2

V t

q0; t = 0:5 Ve

RO

295
296

OF

Fig. 3. Comparison of solidus and liquidus front positions.

19

298
299

310
311
312
313
314
315
316
317
318
319
320
321
322
323

DP

TE

308
309

Fig. 5. A. Desired and computed front positions for Vl = 2, Vs = 1.25. B. Desired and
computed front positions for constant mushy zone thickness.

and thus the desired mushy zone thickness is achieved. As mentioned

earlier, because the quality of solidied alloy material is largely
dependent on the solidus and liquidus front positions or mushy zone
thickness, the desired material quality for a certain application is
attained. As indicated in Fig. 5, there is a deviation from the desired
positions at the beginnings (t 0.06) which may be related to the
initial temperature of the liquid alloy.
A comparison of the higher and lower heat uxes (ql, qs) are shown
in Fig. 6 for different mushy zone thicknesses. The gure shows, the
lower heat ux qs decreases with an increase in the solidus velocity.
But any decrease in the lower heat ux leads to a fast movement of
both solidus and liquidus interfaces. Therefore, only heating is
required to damp the high liquidus velocity resulting from the high
cooling rate at the bottom boundary (qs) and maintain it at a desired
value as illustrated for high solidus velocities.

EC

306
307

RR

304
305

CO

302
303

The required boundary heat ux for V = 2 is simulated numerically

and is shown in Fig.4. For comparison purpose, the analytical [21],
semi-analytical [3] and numerical [16] solutions are drawn. The
results reveal the accuracy of the present method except at the start of
solidication. This is due to the difculty of the solution of an optimal
control problem at initial times and initial temperature of the liquid.
Now, to evaluate the performance of the proposed method for
alloy material, the solidication of Pb10%Sn alloy is investigated with
the following non-dimensional thermo-physical properties: Ste =
0.86, Tf = 0, Tsol = 0.31, Tliq = 0.15, Ti = 0, ks = 1.5, kl = 1, cs = 1.1,
cl = 1.
It is assumed that the mold is initially lled with a liquid alloy having the melting point temperature above the liquidus
temperature.
Considering different solidus and liquidus front velocities, different mushy zone thicknesses may be taken. Different cases with thick,
moderate and thin mushy zone thicknesses are studied. The liquidus
velocity is chosen as Vl = 2 with different solidus velocities Vs = 0.25,
1.25, 1.75, where Vs and Vl are the solidus and liquidus front velocities
respectively. In order to have a constant mushy zone thickness with
Vl = 2, assume that Vs = 1for t [0, 0.06] and Vs = 2 for t N 0.06.
The desired and computed front positions are shown in Fig. 5A, B
for two different mushy zone thicknesses. It is obvious that the front
position values are asymptotically approaches to the desired values

UN

300
301

Fig. 4. Desired and computed heat ux for pure material.

Fig. 6. Comparison of qs and ql for different mushy zone thicknesses.

Please cite this article as: A. Abbas Nejad, et al., Int. Commun. Heat Mass Transf. (2010), doi:10.1016/j.icheatmasstransfer.2010.03.002

324
325
326
327
328
329
330
331
332
333
334
335
336
337
338

ARTICLE IN PRESS

343
344
345

error =
347
346
348
349
350
351

xdesired xcomputed
100;
xdesired

20

where x is the solidus or liquidus front position. The percentage of

error remains within 5% after passing the transient times (t 0.06).
Further, to evaluate the difference between the desired and the
computed front positions, a relative root mean square error, eRMS, is
dened.
r


eRMS =

1 = M M
m = 1 xdesired;m xcomputed;m
r


1 = MM
m = 1 xdesired;m

21

100;

365

5. Conclusion

366

In this work, we have formulated the solidication control

problem of alloy material as a nonlinear inverse heat transfer
problem. Instead of considering the solution for solid and liquid
phases separately, the proposed enthalpy formulation allows the
simplication of the solving procedure only in a single spatial and
temporal domain. It helps to overcome the difculty of connecting the
separate solid and liquid solutions at the interface and expected
computational time. The arising minimization problem has been
efciently solved by using the conjugate gradient method with adjoint
equation. Two moving solidus and liquidus front motions and thus
mushy zone thicknesses are controlled to achieve the desired material
quality. As the literature review indicates, this is the rst time that the
mushy zone thickness is controlled by estimation of two boundary
heat uxes (ql, qs) using inverse heat transfer methods. Simulation
case studies for pure and alloy materials have been set up and the
results indicated a good performance of the proposed solution
approach for the different front motions and even for constant
mushy zone thickness.

361
362

367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
442

t1:1

t1:2
t1:3
t1:4
t1:5
t1:6
t1:7

EC

360

RR

358
359

385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441

CO

356
357

[1] M.C. Flemings, Solidication Processing, McGraw-Hill, New York, 1974.

[2] M.J. Donachie, S.J. Donachie, Superalloys: A Technical Guide, 2nd ed.ASM
International, Ohio, 2002.
[3] N. Zabaras, S. Mukherjee, S. Richmond, An analysis of inverse heat transfer
problems with phase changes using an integral method, Journal of Heat Transfer
110 (1988) 554561.
[4] N. Zabaras, On the design of two-dimensional Stefan processes with desired
freezing front motions, Numerical Heat Transfer 21 (1992) 307325.
[5] N. Zabaras, Dynamic programming approach to the inverse Stefan design
problem, Numerical Heat Transfer 26 (1994) 97104.
[6] S. Kang, N. Zabaras, Control of the freezing interface motion in two-dimensional
solidication processes using the adjoint method, International Journal for
Numerical Methods in Engineering 38 (1995) 6380.
[7] R. Sampath, N. Zabaras, Adjoint variable method for the thermal design of eutectic
directional solidication processes in an open-boat conguration, Numerical Heat
Transfer 39 (2001) 655683.
[8] G.Z. Yang, N. Zabaras, The adjoint method for an inverse design problem in the
directional solidication of binary alloys, Journal of Computational Physics 140
(2) (1998) 432452.
[9] G.Z. Yang, N. Zabaras, An adjoint method for the inverse design of solidication
processes with natural convection, International Journal for Numerical Methods
in Engineering 42 (1998) 11211144.
[10] N. Zabaras, H.T. Enguyen, Control of the freezing interface morphology in
solidication processes in the presence of natural convection, International
Journal for Numerical Methods in Engineering 38 (1995) 15551578.
[11] J.I. Frankel, M. Keyhani, A new approach for solving inverse solidication design
problems, Numerical Heat Transfer Part B: Fundamentals 30 (2) (1996) 161177.
[12] S.W. Hale, M. Keyhani, J.I. Frankel, Design and control of interfacial temperature
gradients in solidication, International Journal of Heat and Mass Transfer 43 (20)
(2000) 37953810.
[13] R. Xu, G.F. Naterer, Controlled interface acceleration in unidirectional solidication, International Journal of Heat and Mass Transfer 47 (22) (2004) 48214832.
[14] R. Xu, G.F. Naterer, Deterministic physical inuence control of interfacial motion
in thermal processing of solidied materials, Experimental Thermal and Fluid
Science 29 (2) (2005) 227238.
[15] R. Xu, G.F. Naterer, Inverse method with heat and entropy transport in
solidication processing of materials, Journal of Materials Processing Technology
112 (1) (2001) 98108.
[16] V.R. Voller, Enthalpy method for inverse Stefan problems, Numerical Heat
Transfer Part B: Fundamentals 21 (1) (1992) 4155.
[17] K. Okamoto, B.Q. Li, A regularization method for the inverse design of
solidication processes with natural convection, International Journal of Heat
and Mass Transfer 50 (2122) (2007) 44094423.
[18] M. Hinze, S. Ziegenbalg, Optimal control of the free boundary in a two-phase
Stefan problem, Journal of Computational Physics 223 (2) (2007) 657684.
[19] J.V. Beck, B. Blackwell, C. Clair, Inverse Heat Conduction: Ill Posed Problems, John
wiley & Sons, New York, 1985.
[20] M.N. Ozisik, H.R.B. Orlande, Inverse Heat Transfer, Fundamentals and Applications, Taylor & Francis, New York, 2000.
[21] H.S. Carslaw, J.C. Jaeger, Conduction of Heat in Solids, 2nd ed.Oxford University
Press, Oxford, 1959.
[22] A. Abbas Nejad, M.J. Maghrebi, and H. Basirat Tabrizi. A High Order Time
Advancement Scheme for Prediction of Solidication Processes, in 5th International Conference on Diffusion in Solids and Liquids, Rome, 2009. (to be published
in Defect and Diffusion Forum).
[23] R. Swaminathan, V.R. Voller, A general enthalpy method for modeling solidication processes, Metallurgical and Materials Transactions B 23 (1992) 651664.

UN

354
355

384

TE

363
364

where, M is the total number of time steps and x is the solidus or the
liquidus front position. The values of RMS error for solidus and
liquidus positions, the number of iterations needed for minimizing the
objective function and the objective function value for the last
iteration are listed in Table 1.
Table 1 indicates that the present approach has a very good
convergence and accuracy in all cases. Also, it can be observed that the
solution exhibits a better convergence for the thick mushy zones with
two moving sensors far from each other.
From the obtained results, it can be observed that the desired
mushy zone thickness which leads to a desired product quality is
achieved.

353
352

References

OF

341
342

The high value of qs at the beginning may also be related to the

initial temperature of the liquid alloy which is higher than solidus
temperature. From the results of Fig. 6, one notes that the bottom
boundary requires cooling only after it converges to a desired value
but for the upper boundary both cooling and heating is required.
To investigate the performance of the proposed method, the
relative error is dened as follows:

RO

339
340

A. Abbas Nejad et al. / International Communications in Heat and Mass Transfer xxx (2010) xxxxxx

DP

Table 1
Comparison of the objective function and RMS error for different mushy zone
thicknesses.
Solidus
velocity

Liquidus
velocity

Objective
function

Iteration
numbers

Solidus
eRMS%

Liquidus
eRMS%

0.25
1.25
1.75
2

2
2
2
2

3.1e4
4.078e4
8.33e4
3.69e4

25
41
64
70

7.23
4.25
6.43
5.86

4.72
5.63
7.15
4.70

Please cite this article as: A. Abbas Nejad, et al., Int. Commun. Heat Mass Transf. (2010), doi:10.1016/j.icheatmasstransfer.2010.03.002