ht. J. Heat Moss Transfer. Vol. 16, pp. 1825-1832. Pergamon Press 1973.

Printed in Great Britain

NUMERICAL SOLUTION OF PHASE-CHANGE

PROBLEMS

C. BONACINA and G. COMINI

lslltuto di Fisica Tecnica dell’IJniversit8 e Laboratorio per la Tecnica de1 Freddo de1 C.N.R.,
Padova. Via Marzolo. 9. Italy

A. FASANO and 111.PRIMICERIO

lstituto Matematico “Ulisse Din? e Gruppo Nazionale per la Fisica Matematica de1 C.N.R.,
Universita di Firenze, Italy

(Received 26 October 1912 and in revisedform 15 January 1973)

Abstract-A three-time level implicit scheme, which is unconditionally stable and convergent, is employed
for the numerical solution of phase-change problems, on the basis of an analytical approach consisting in
the approximation of the latent heat effect by a large heat capacity over a small temperature range. Since
the temperature dependent coefficients in the resulting parabolic equations are evaluated at the inter-
mediate time level, the complication of solving a set of nonlinear algebraic equations at each time step is
avoided. The numerical results thus obtained are satisfactorily compared with the available analytical
solutions.

NOMENCLATURE time step [s]:

C, specific heat [J/kgK] ; step function;
C = cp, volumetric heat capacity rJ/m3K1; heat of phase change per unit volume
local order of accuiacy iI?/s] ; - [J/m31;
Y? surface temperature [“Cl ;
enthalpy per unit volume [J/m31 ;
difference operator:
:‘J+‘+f + l;Th-+
H,
k, thermal conductivity [W/mK]; W,Th =
2
K thermal conductivity integral [W/m]; Thf’ _ Th-’
slab thickness [m] ; =

$Ax2), the remainder in Taylor series expan-

phase front position [m] ;
2 ;
sion and of the order (8x2);
4 temperature: solution of the heat con- density [kg/m31 ;
duction equations [“Cl; time [s] :
7: temperature: solution ofthe numerical initial temperature distribution [“Cl.
equations [‘Cl;
x, y, z, position coordinates [m]. Subscripts
a. surface ;
Greek letters .L phase change;
YX’ difference operator: i, lattice parameter in a one-dimensional
r,q = q_+- q_+; grid: x = iAx;
6, Dirac “function”: reference :
ix, grid spacing in the x direction [m] ; ;, initial ;
At, temperature semi-interval across t F 1, solid region;
WI:
1

2, liquid region.
1825
1826 C. BONACINA, G. COMINI, A. FASANO and M. PRIMICERIO

Superscripts numerical calculations in [6, 9, 11. 221, while

h, time level: z = hAz: a rigorous analysis, based on the concept of
-1 equivalent; “generalized (weak) solution” and including a
* approximated. convergence proof, has been given in [12, 13,
15,201. Nevertheless this procedure has not been
INTRODUCTlON used as much as its simplicity would suggest.
A LARGE number of current applications of This is undoubtedly due to the difficulties
engineering interest involve solution of the involved in the solution of the resulting heat
equation describing heat conduction with a conduction equation which has critically tem-
change of phase. However, while many im- perature dependent coefficients.
portant theoretical results are available about The intent of the present work is to show how
existence uniqueness, and properties of classical a three-time level difference method, which has
solutions (see e.g. [7, 8]), the literature surveys been recently proposed for the solution of heat
[2, 3, 19, 211 reveal that most actual analytical conduction problems when the thermophysical
solutions deal only with the one-dimensional properties depend on temperature [4, 51, can
geometry and very special boundary conditions be advantageously applied also to the solution
and cannot thus be extended to many practical of melting and freezing problems, provided that
problems. they are formulated as previously outlined.
Numerical methods have also been proposed
by several authors (see [lo, 211) but, especially FORMULATION OF THE PROBLEM
in the multidimensional case, their applicability Usually a problem with a phase transition
is limited by their great complexity. Therefore of a substance from one state to another, in the
it can be inferred that the conventional mathe- one-dimensional case under boundary conditions
matical description, considering an interface of the first kind, is mathematically formulated
surface which moves either into the solid region as follows [ 181
(melting) or into the liquid region (freezing) in
accordance with the relative magnitudes of the c,(r):=& k,(t); : 0 < x < 5(z),
temperature gradients on either side of it, does ,[ 1
z > 0; (1)
not yield to a general mathematical formulation
which combines good accuracy with computa- t,(x,O) = q,(x) < ty 0 < x < ((0): (2)
tional ease.
Fortunately another model is possible which t,(O, z) = f,(z) < tf: z > 0, (3)
has a sound theoretical basis and leads to a where subscript 1 refers to the “solid” region,
much more compact and convenient analytical and :
formulation.
Most phase-change processes involve sub-
stances which, like the foodstuffs or the metal
C,(t); = ;
[k,(t);
1
;
((7) < x < L, z > 0; (4)
alloys, are not pure and, therefore, the latent
heat effect can be expressed by a finite temperature t& 0) = q,(x) 3 t/: C(O)< X < L; (5)
dependent heat capacity, since it occurs over a
1,(L. z) = f&Z) > t/ z>o (6)
temperature range [4,23].
This feature suggests that the true latent heat where subscript 2 refers to the “liquid” region.*
effect in pure substances can be approximated ___-
by a large heat capacity over a small temperature * Each of (3) and (6) can be replaced by a condition of
prescribed x-derivate on .x = 0, x = L respectively. In
range. Such an intuitive approach, proposed in (lH6) one assumes 0 < i(O) < L; trivial modifications are
[ 1. 241, has been employed in recent years for needed in this scheme for the cases t(O) = 0 or t(O) = L.
 NUMERICAL SOLUTION OF PHASE-CHANGE PROBLEMS 1827

On the interface it is are obtained

an d
r,Wr), r] =

k,(t)
t,[t(7),

2 - k,(t) 2 =Iz2
73 = ts (7)

(8)
lim I, = lim
E’O E-+0

;+e
s
r-c
r+e
i?(t) ; dx

= lim C(t) 4 dx
where I is the heat of phase change per unit E-+0[S
volume.
However, based on a physical approach, a
different mathematical model can be tried.
Boundary condition (8) in fact, stems from the
“jump” of magnitude L which the enthalpy per
unit volume H undergoes at the temperature
of the phase transition tf. Therefore, from the
enthalpy definition r-E
H(t) = j C(t) dt + $(t - t,); - k(t); _
IX-<-E
Since I, = I,, condition (8) can be immediately
1. (13)

derived. Therefore, if a numerical solution is

where f < tf is arbitrarily chosen, an “equiva- tried, instead of starting from the conjugated
lent” heat capacity per unit volume c(t) can also problems (1) and (4) coupled through condition
be defined [6, l&20,22] (8), a single equation (11) can be considered.
Actually this step is not a straightforward one:
dfilft)
C(t) = --&- = C(t) + 126(t- t,); equation (11), in fact, involves a delta-function
in the definition of c(t) and must be regarded
C,(t), t < t/ thus as a “generalized” formulation for problem
C(t) =
t > ts (10)
C,(t), (l), (4) and (8) to be dealt with on the basis of
where ii(t - tf) is the Dira~-function. the theory of distributions. However, due to the
Substituting in equations (1) and (4) the practical intent of this work, only the theoretical
following expression is derived results which are needed to further proceed with
a meaningful numerical solution will be men-
tioned :
t?(t); = ; k(t); ;
-[ 1 1. Under suitable assumptions (unessential for
practical purposes) on the data and the coeffic-
k,(t), t< tr
k(t) =
t> (11) ients, there exists a generalized solution to the
k,(t), tf
problem (ll), (2) (3), (5) and (6)t which is
It can be shown that every sufficiently smooth unique and, by means of elementary calcula-
solution of (ll), such that the regions t > tf, tions, can be shown to coincide with the classical
t < tf are separated by a unique continuously solution to the original problem (l)-(8) when-
differentiable curve x = t(r), verifies condition ever the latter exists.
(8). In fact, let us consider the integrals 1, and 1, 2. Such a solution can be obtained as the limit
of the left and right members of equation (11) in of a uniformly convergent sequence of classical
a small layer ([ - E < x G { + a) symmetrical solutions to approximating problems, deduced
with respect to the interface and then let us take by smoothing the coefficients in equation (11)
the limits as c + 0. The following relationships following a few general rules:
1828 C. BONACINA. G. COMINI. A. FASANO and M. PRIMICERIO

(a) The delta-function in the definition of C(t) Thus the approximating problems are form-

?t
must be replaced by a deltaform function ulated as follows:

1
fi(t - tS’ At) which assumes large but finite
iit ?
values in the semi-interval At across tf where it C*(t)iiz = ii.u k*(t);l.y (17)
is different from zero. Consequently a smoothed [
heat capacity C*(t) can be defined with initial, and boundary conditions given by
c,w. t < t/. - At (2). (3), (5) and (6). Each of these problems
c,w, t > tf + At
c*(t) = (14) possesses unique classical solution [ 161 which,
according to points 1 and 2 and to the quoted
and results of [15. 201, as At -+ 0 tends to the
t,+Al generalized solution of the Stefan problem.
j C*(t)dt = I_ + y C,(t) dt The actual smoothing can be performed in
L~-AI If-AL
many ways. Several shapes have been tested
l, +A1
numerically and the choice does not seem to be
+ s C,(t) dt (15)
11
critical.
A definition of C’*(t) and K*(t) which works
which retains the enthalpy variations.
(b) Similarly a smoothed thermal conductivity well and, owing to its simplicity, turns out to be
k*(t) has to be introduced, subjected to the very suitable for practical applications is des-
cribed here.
conditions
Suppose that C,, C,. k, and k, do not depend
k,(t), t < tf - At
k*(t) = (16) on t. then in the interval: tJ - At < t B Ts + At
k,(t), t > tf + At' the following definitions may be assumed

& +c, + c,
.; A generalized solution to the problem (11). (2). (3). (5)
and (6) in the rectangle: 0 < x < I,, 0 : r < 0 is every c*(t) = 2 (18)
measurable function t(x, r) satisfying the integral identity
(I 1.
^
?’
and

//
.iii *
0
ff[t(X,r)] ; Ffx. r) + K[t(u. t)] G F(x, r)
I
dx dr
k -k
k*(t) = k, + w[f - (ts - At)]. (19)
1. B
c3
+
_L 0
H[dx)] Rx. 0) dx +
J‘i
,I
K[f;(T)] - KY, r)
c’u ~1-0
At this point it is worth mentioning that an
important feature of the procedure outlined here
is that the coefficient smoothing is carried out
where with respect to t and so it does not depend on
K(t) is defined as: whether the problems are one-dimensional or
K(t) = I‘ k(t)dt
multi-dimensional. The extension to the multi-
I, dimensional case therefore is straightforward
cp(x) is given by and yields to equations which are formally
analogous to equation (17). For example, in the
three-dimensional case the equation correspond-
and F(x, r) is any function which is continously differenti- ing to equation (17) is
able, twice with respect to x and once with respect to L and
such that
F(0. r) = F(L. r) = F(X. 0) = 0. (IV)
The results of [IS] and [20] concerning similar cases can
be still utilized to show the existence and uniqueness of the
(20)
generalized solution thus defined.
 NUMERICAL SOLUTION OF PHASE-CHANGE PROBLEMS 1829

THE DIFFERENCE SCHEME It can be shown that the three-time level

Only the numerical solution of equation (17) implicit scheme described here is unconditionally
under boundary conditions of the first kind will stable and convergent [4, 171. By a Taylor’s
be considered in detail here. However, since the series expansion the local order of accuracy of
three-time level implicit scheme employed here equation (23) can be written as
f4, 171 has been extended to the multi-dimen-
sional case and to different boundary conditions $=
[5], it can be said that, in principle, all the phase-
change problems can be solved in this same way.
-
By using central difference operators the
following approximation for equation (17) can
be obtained + = O(AT*+ Ax*). (25)
(21)
RESULTS
where (7’1) is the average temperature:
The numerical method described in the
(7’;) = f(T;+’ + T: + T;-I). (22) previous sections has been checked against the
results of available analytical solutions of one-
Equation (21) can be rewritten as
dimensional freezing probIems [18]. A Fortran
C*~T~)~T~+ ’ - Tf-‘) IV program, based on formula (23) has been
written to calculate the temperature distribution
in a slab with variable thermophysical properties.
The thermophysical properties used in the
+ U-f,, - T;) + CT;;,’ - T;-‘)] computations were those of the system water-ice.
Constant values of the heat capacity and of the
- k-[(T:+’ - T;_f;) + (Tf - Tf_ J
thermal conductivity were assumed in the solid
+ (T;-’ - T;:;)]j (23) and in the liquid region; in the phase-change
where zone the heat capacity and the thermal conduct-
ivity were smoothed according to def~itions (18)
k+ = k*(TF+$ ")
k*(Tf++)g and (19). Since freezing problems were treated,
(24) the phase front was taken to coincide with the
isotherm (tf - At), corresponding to the lowest
k- = k*(T;_,) g k” (T’+$).
temperature end of the freezing zone.
Reference was made to a layer subjected to
The expressions (24) do not alter the order of the following boundary conditions
accuracy of formula (23) and involve values of T
at grid points only. Angther important feature t,(O, z) = tp = const: z>o (26)
of formula (23) is that the coefficients are
computed at the intermediate time level: resort- (27)
ing to iterations is thus avoided, since the
tridiagonal system of difference equations to First a simplified initial condition was con-
be solved at each time step is linear. sidered assuming the temperature of the water
When only data at T = 0 is given, in order to to be at the freezing point, with no ice present:
start the calculations it is necessary to obtain
t&O) = t, = tf = const;
the data at z = AZ from a two level formula of
comparable accuracy. r(o) = 0 < X < L. (28)
 C. BONACINA, G. CUMINI, A. FASANO and M. PRlMlCERiU

0, I .________Analyticol
solution
-

FIG. 1. kIterfaCe histories.

In the calculations 41 equally spaced nodal the particular case: r, - I, = 20 K. The accuracy
points were used (AX = 0.0125 m: L = 0.5 m) of the results plotted in Figs. 1 and 2 is always
with time steps variable from 200 to 600 s but better than 1 per cent.
constant within each run. The phase change was A phase change problem with the initial
usually assumed to take place jn a temperature temperature of the water above the freezing
interval of 05 K, but the results obtained were point was then considered. The initial and
rather insensitive to the particular choice boundary conditions were taken to coincide
operated. In Fig, 1 the computed phase front with the analytical solution at z = 72000 s
position curves are compared with the corres- yielded by the following boundary and initial
pond@ analytical sofution for different values conditions :
of the initial “jump” (t, - te) in the external t,(O,t) = r, = -20°C: r>O (29)
surface temperature. In Fig. 2 the calculated
t,ix,O) = t, = 10°C; ((0) = 0 SSX d L. 130)
tem~rature distributions at different time values
are compared with the analytical solutions For Tat& 1
___ ...._- .-__ ....-_.. _.- .._~_
e(m)
5 x 1()-4(S)--.~~~.-~...--l_-~ _-_- _._
Analytical SC& Numerical sol.
-_____ ~.-~~~~_-_-_.- .-.-
7,2 0.123 0123
9.0 0.138 0138
10.8 0.151 0,156
12.6 0,163 0.166
14.4 0.175 0*180
16.2 0.185 0189
i8.0 0.195 0199
19.8 0.205 020s
21.6 0.214 0216
23-4 0.223 0.225
25.2 a231 0233
2TO 0239 0239
2RR 0,247 0247
FIG. 2. Temperature distribution at different time values. -- .__.__-.._ ....-.- ---- -A
 NUMERICAL SOLUTION OF PHASE-CHANGE PROBLEMS 1831

In order to use in the comparison the analytical tion does not depend on the geometry con-
solution for a semi-infinite body, calculations sidered, extension to multidimensional cases is
were stopped at z = 288000s as soon as the straightforward.
temperature t(L, z) started to change appreciably. As a final remark it is worth indicating that
When two phases are simultaneously present, the assumption of pure conduction in the liquid
as in this case, the choice of the temperature phase is not always strictly correct, but ex-
interval in which freezing is assumed to take perience has shown that for design calculations
place becomes critical. Numerical calculations an equivalent thermal conductivity dependent
have shown that good results are obtained only on the Rayleigh number can be used when
if the phase change interval embraces at least convection is of importance [14], avoiding thus
2-3 nodes of the network along the x-axis [22]. changes in the problem formulation.

Table 2

t(x. T) (“C)

T x 10-4(s) .Y= 0.1 m x = 0.2 m x = 0.4 m

An. sol. Num. sol. An. sol. Num. sol. An. sol. Num. sol.

7.2 - 3.12 - 3.12 5.81 5.81 9.86 9.86

14.4 - 8.44 - 8.95 1.61 1.23 8.75 8.72
21.6 - 10.54 - 10.82 - 1.26 -1.81 I.24 7.18
28.8 -11.80 - 12.02 - 3.72 -4.13 5.81 5.13

Moreover too large temperature intervals are REFERENCES

not advisable since strong divergence from the 1. E. L. ALBASINY, The solution of nonlinear heat conduc-
tion problems on the pilot ACE, Proc. Inst. Electric
original problem would arise. In the example Engng. 103, Part B, Suppl. 1, 158-162 (1956).
considered here 81 equally spaced nodes were 2. S. G. BANKOFF,Advances in Chemical Engineering,
edited by T. B. DREW,J. W. Hoops and T. VERMEULEN,
used (Ax = O-0125 m; L = 1 m) and a phase- Vol. 5, pp. 75-150. Academic Press, New York (1964).
change interval of 2 K was assumed. The results 3. B. A. BOLEY, High Temperature Structuresand Materials,
obtained for the phase front histories and the Proc. 3rd. Symp. on Naval Structural Mech., edited by
A. M. FREUDENTHAL, B. A. BOLEYand H. LIEBOWITZ,
temperature distributions are reported in Table
pp. 260-315, Pergamon, Oxford (1963).
1 and in Table 2 respectively. The accuracy 4. C. BONACINA and G. COMINI,On a numerical method for
(r3 per cent) is still well within the limits the solution of the unsteady heat conduction equation
imposed for engineering calculations. with temperature dependent parameters, I.I.R., XIII
Int. Congr. of Refrig., 2-28 Washington (1971).
Run time for the Fortran program is about 5. C. BONACINAand G. COMINI,On the solution of the
20 s on a CDC 6600 computer with a grid of 41 nonlinear heat conduction equations by numerical
nodal points and 1000 time steps. methods, Znt. J. Heat Mass Transfer 16,581-589 (1973).
6. B. M. BUDAK,E. N. SOBOL’EVA and A. B. USPENSKII.A
CONCLUSIONS difference method with coefficient smoothing for the
solution of Stefan problems (in Russian), 211. Vychisl.
The finite difference method presented in this Mat. IMot. Fiz. 5. 82%X40 (19651.
paper has enabled to deal with melting and 7. J. R. CANNONand C. D. HILL,’Existence, uniqueness,
stability and monotone dependence in a Stefan problem
freezing problems. The results obtained have for the heat equation, J. Math. h4ech. 17. l-20 (1967).
been shown to be sufficiently accurate for 8. J. R. CANNONand M. PRIMICERIO, A two phase Stefan
engineering use. problem with temperature boundary conditions. Ann.
Mat. Pura Appl. 88(N), 177-192 (1971).
Reference has been made mostly to one- 9. E. J. COUCH,H. H. KELLERand J. W. WATTS, Perma-
dimensional cases but since the problem formula- frost thawing around producing oil wells, 21st Annual
1832 C. BONACINA, G. COMINI, A FASANO and M. PRIMICERIO

Technical Meeting of the Petroleum Society of CIM, 17. M. LEES, A linear three-level difference scheme for
Calgary (1970). quasilinear parabolic equations, Maths. Comp. -20,
10. A. FASANOand M. PRIMICERIO. Convergence of Huber’s 516-522 (1966).
method for heat conduction problems with change of 18. A. V. LUIKOV, Analytical Heat Diffusion Theory.
phase. 2. Anyew. Math. Mech. 53. 341-348 (19731. Academic Press, New York (1968).
II. A. K. FLEMING,The numerical caiculation of freezing 19. J. C. MUEHLBAIJER and J. E. SUNDERLAND,Heat conduc-
processes, Vol. 1, Thesis for the Technical Licentiate tion with freezing or melting Appl. Mech. Rev. 18.
Degree. TheTechnical UniversitvofNorwav. , Trondheim 951-9.59 (1965).
fi9ii). 20. 0. A. OLEINIK,A method of solving the general Stefan
12. A. FRIEDMAN,One dimensional Stefan problems with problem (in Russian), Dokl. Akad. Nauk SXSR 135,
ilonmonotone free boundary. Trans. Am. j~~aih. Sot. 1054-10.57 (1960).
133.89-t 14 (1968). 21. L. I. RUBINSTEIN.The Stefan Problem, Vol. 27. AMS
13. A. FRIEDMAN,The Stefan problem in several space Translations, Providence, Rhode Island (1971).
variables, Trans. A~)I. Math. Sot. 133, 51-87 (1968); 22. A. A. SAMARSKII and B. D. MOISEYNKO, An economic
Correction, ibid. 142. 557 (1969). continuous calculation scheme for the Stefan multi-
14. W. L. HEITZ and J. W. WEST~ATER.Extension of the mensional problem (in Russian). Zh. i’~hIsI. Mar. Mut.
numerical method for melting and freezing problems, Fiz. 5.816827 (1965).
Int. J. Heat Moss Transfer 13, 1371-1375 (1970). 23. R. H. TIE~Vand G. E. GEIGER,A heat-transfer analysis
15. S. L. KAMENOMOSTSKAJA, On Stefan’s problem fin of the solidification of a binary eutectic system. J. Heat
Russian), Mat. Sb. 53.489-5 14 (1961). Transfer 89, 230-235 (1967).
16. 0. A. LADYZENSKAIA, V. A. SOLONNIKOV and N. N. 24. A. N. TIHONOVand A. A. SAMARSKII, The Equations of
URAL’CEVA,Linear and Quasilinear Equations of Para- Mathematical Physics. 2nd., edited by GITTL. Moscow
botic Type. Vol. 23, p, 449. AMS Translations, Provi- (1953).
dence. Rhode Island (1968).

SOLUTION NUMERIQUE DES PROBLEMES DE CHANGEMENT DE PHASE

R&am&-Un schema imp&cite a trois niveaux de temps qui est inconditionnellement stable et convergent,
est employt pour la solution numerique de problemes de changement de phase sur la base d’une representa-
tion analytique de l’effet de chaleur latente par une grande capacite caloritique sur un petit domaine de
temperature. Puisque les coefficients d&pendants de la temperature dans les equations paraboliques resul-
tames sont halt& au niveau intermtdiaire de temps, on peut eviter la resolution compliquee dun systeme
d’equations algtbriques non linbires a chaque pas de temps. Les &&tats numeriques ainsi obtenus sont
compares de maniere satisfaisante aux solutions analytiques connues.

NUMERISCHE LOSUNG VON PHASENANDERUNGSPROBLEMEN

Z~rn~nfa~~g-Ein Drei-Zeiten-E~nen-System. das absolut stabil und konvergent ist. wird fur die
numerische L&sung von Phasenlnderungsproblemen verwendet.
Fiir kleine Temperaturbereiche wird die latente Wlrme durch eine grosse Warmekapazitlt angenlhert.
Die’temperaturabhlngigen Koeffizienten in den resultierenden parabolischen Gleichungen wurden auf
einer Zwischenzeitebene abgeschfitzt, womit die Schwierigkeit, eine Reihe von nichtlinearen algebraischen
Gleichungen bei jedem Zeitschritt zu l&en, vermieden ist. Die so erhaltenen numerischen Liisungen
zeigten mit den verfiigbaren analytischei~ Lijsungen zufriedensteliende ~bereinstimmung.

YHCcnEHHQE PElIIEHME 3.4AAY 0 (DA30BbIX HEPEXO&~N

f~~HoTa~~-~~~ ~nc3le~noi 0 p~tne~nn
na;lau 0 $a30abl;x nepexonax mnoabapeTcfl aiico-
JIKtTHO YCTO~~~Ba~ El CXO~K~~a~CK HeRBHaR
CXeMa C TPeMR npeMe~i~ibIM~1~pOBI~n~31
(TpexcnotiHan CxeMa). Me’ro;r CO~TOATB TOM, 9TO CtipbITbIZtTeZIOBOli B#@eKT ElIIIIpOltCU-
MNpyeTCrl 6OJIhIlWvl 3Ha'lt'HIZeM TenJIOeMIiOCTIl B MajlOM TeMTIepaTppHOM WalENOHe.
nOf'KOJIb@ 3aBMClWfPle OT TeMIIepaTSpbI H(O3@@4~LIeHTbI B llapa60JIWieCK~IX ypaBHeH&WIX
[~~CCYATbIBaIOTC.R Ha RpOMWKyTOYHOM BpeMeHHOM YpOBHe, MOWHO 1236emaTb TPy~HOCTSI
peruerrm C35cTemI Henmie&ibIx anre6pameemx ypa3aemGi Iia Kaxzokf ispeMeKKonn yposse.
~O~~qeHHbIe TaKHM cna~o6o~x ~~C~eHH~e pe3gJlbTaTbI ~~O~~~erBO~3~Te~bHO COr~aC~~TC~ C
1tMe~~~~~I~C~ 3Ha~I~Ti~~eCK~~M~ipeU.IeHWiMI~.