Академический Документы
Профессиональный Документы
Культура Документы
DOI 10.1007/s00170-012-4102-7
ORIGINAL ARTICLE
Received: 14 May 2011 / Accepted: 13 February 2012 / Published online: 26 April 2012
# Springer-Verlag London Limited 2012
Abstract In the present work, a mathematical model was Generally, porosities are formed by two mechanisms
developed based on finite difference method to predict the being gas segregation and solidification shrinkage which
microporosity distribution in A356 aluminum alloy casting. occur concomitantly [1–4]. Liquid entrapped in the cast part
Heat, mass, and gas conservation equations were solved in during the solidification is potentially able to form porosity.
this model. Moreover, Darcy’s equation was considered in Since dissolved gas ejects into these zones, the melt density
the mushy zone. Results show that the distribution and decreases during the solidification and entrapped liquid
concentration of microporosities in cast parts vary with both cannot be fed which results in porosity formation.
cooling rate and initial gas content. Simulation results were Porosities are classified by their size and dominant mech-
compared with experimental data where proportionally anism of formation. In connection with size, micro- and
good agreement with experimental results was found. Final- macroporosities can be used. Microporosities nucleate in
ly, a complex cast part was simulated presenting the ability entrapped liquid between dendrite arms (they are formed
of the model to predict the porosities in industrial cast parts. because of gas precipitation and lack of feeding), and mac-
roporosities occur in bulk of entrapped liquid either inside
Keywords Modeling . Simulation . Microporosity . Gas the cast part known as closed shrinkages or top of the
segregation . Dendrite arm spacing surface of the cast part called pipe shrinkages. Based on
formation mechanism, they are classified into gas and
shrinkage porosities. Gas porosities are spherical and
shrinkage porosities are dendritic.
1 Introduction The stages of microporosity formation in long freezing-
range alloys are explained schematically in Fig. 1. In the
It is well- known that the mechanical properties of castings first stage, the pore nucleates on heterogeneous nucleation
such as ultimate tensile strength and fatigue strength are sites (such as an oxide particle surface), then it grow spher-
mainly affected by the existence of porosities [1]. So, un- ically by gas precipitation. In the next stage, growing mi-
derstanding the location and amount of porosities is very crostructure limits the pore growth. Hence, the pore is
useful to avoid sudden drop of mechanical properties. deformed into a dendritic-shaped one. After the mentioned
stages, pores grow adjacent to eutectic phase due to density
A. Bahmani (*) : N. Varahram : P. Davami
increase (volume shrinkage) of the eutectic phase.
Department of Materials Engineering, Porosity models are classified into two groups: indirect
Sharif University of Technology, and direct methods. Indirect models [5–9] are based on a
Tehran, Iran parameter representing the pattern of solidification; they
e-mail: ahmad.bahmani@gmail.com
evaluate the probability of microporosity formation in dif-
N. Hatami ferent zones of the casting. These models, whose critical
Razi Metallurgical Research Center, values are found by experiments, consider just a part of
Tehran, Iran solidification concept. Although indirect models do not
completely consider the principles of solidification phenom-
M. O. Shabani
Materials and Energy Research Center (MERC), enon, due to simple programming and analysis, casting
Tehran, Iran simulation packages usually use some of these parameters
1314 Int J Adv Manuf Technol (2013) 64:1313–1321
(b) If gS tends to 1, permeability will be zero and this gravity vector and PL is the local pressure of liquid set initially
shows a conflict with Darcy equation. by Eq. 6.
To solve this problem, Kubo and Pehlke [13] used two Eμ
thresholds of solid fraction as 0.01 and 0.7. Outside this μ ¼ μ0 exp ð5Þ
RT
range, solid fraction is fixed. For example if gS 00.8 then gS
will be replaced by 0.7 and if gS 00.005 then gS is set to
0.01. Sabau and Viswanathan [10] believe that for compu- Pl ¼ Pa þ Pm ð6Þ
tation of Eq. 1 from stereological studies Sv can be written as
Sv 04/λ2. For A356 aluminum alloy, an isothermal ternary where μ0, Eμ, and R are the viscosity of liquid at very
eutectic reaction occurs at a 0.05 (liquid fraction) and co- high temperatures, activation energy and universal gas
herency point for the alloy is 0.4 of solid fraction. They constant, respectively. In modeling, these values are set
introduced the permeability equation as: based on Table 1. Here, T is the temperature in Kelvin. Pa
( and Pm are also atmospheric and metallostatic pressures,
Sv ¼ ð10:6
gs
Þ l2 ; gs > 0:4
4 respectively.
Sv ð gs ; l2 Þ ¼
Sv ¼ l42 ; gs < 0:4
8
< K ¼ ð1gs Þ2 ; g < 0:05 ð2Þ
s 2.3 Gas conservation
K ð gs ; l2 Þ ¼ 5Sv 2
: K ¼ 0:0522 ; gs 0:4
5Sv
The gas solubility in the solid is much lower than in liquid
Kurz and Fisher [24] demonstrated that secondary dendrite and therefore during the solidification, dissolved gas in the
arm spacing (λ2) is proportional to the local solidification time solid is ejected into the remaining liquid. The growth rate of
with power of 0.33, Eq. 3. This was confirmed by other the pores depends on the amount of gas diffusion into the
researchers who reported a modified version of this correla- liquid. Considering the fact that hydrogen diffusion is at
tion based on experimental findings [4, 10, 13, 23–25]. least three orders of magnitude higher than other gases [1]
and since hydrogen is the most important source of micropo-
l2 ¼ atfb ð3Þ rosity evolution in aluminum alloys [2–4, 10, 12, 13, 23, 27],
gas conservation equation can be written as follows:
Where, tf is the local solidification time, and “a” and “b”
are composition-dependent constants. agp Pg
ρL ½H0 ¼ ρs gs ½Hs þ ρL gL ½HL þ ð7Þ
T
2.2 Pressure drop in mushy zone where, ρ, g, and [H], are the density, volume fraction
and hydrogen content respectively and 0, L, S, P, and g
Fluid flow through a porous medium can cause a pressure indices indicate initial, liquid, solid, porosity and gas
drop and a decrease in the fluid velocity. In mushy zone, respectively. α is the gas conversion factor and Pg is
dendrite arms act as a porous medium. This behavior can be the gas pressure.
described by Darcy’s equation [2–4, 11, 13, 26]. The relationship between [H]S and [H]L is as follows:
K [H]S 0kh[H]L, in which kh is a partition coefficient set using
ul ¼ ðgradPl ρl gÞ ð4Þ Table 1. [H]L is related to the gas pressure by Sievert’s
μgl
equation as follows:
Where K is the fluid permeability, μ is the dynamic viscos- pffiffiffiffiffi
ity of liquid defined in Eq. 5, ρL is the liquid density, g is the ½ H l ¼ Kl Pi ð8Þ
In this equation, Pi is non-dimensional gas pressure and is mushy zone and gas pressure applied by segregation of gas
calculated as Pi 0Pg/Pa, where Pa is the atmospheric pressure from solidifying liquid.
and KL is the hydrogen solubility. First, in order to estimate thermal distribution and solid-
Gas pressure (Pg) is related to the local pressure and ification pattern in the solution method, heat conservation
capillary effect through Eq. 9. equation is solved by an explicit scheme. In any time step,
the domain is counted in three directions. If any node’s
Pg ¼ Pl þ Pσ ð9Þ
temperature falls below the liquidus, the secondary dendrite
Where Pσ represents surface tension, and is calculated as arm spacing is calculated by Eq. 3 with “a” and “b” con-
Pσ 0σPL (1/r1 +1/r2), in which σPL is the surface tension stants of 10.2×10−6 and 0.33, respectively. Moreover, per-
between pore and liquid and r1 and r2 are principal radii of meability in the mushy zone and pore radius are calculated
curvature. Assuming the porosities spherical, we will have from Eqs. 2 and 11, respectively.
r1 0r2 0r and therefore Pσ can be calculated as: Here, the pressure and microporosity distribution in
the cast part are being obtained by developing Kubo’s
2σPl
Pσ ¼ ð10Þ 2D-algorithm [5]. In the present work, the Kubo’s 2D-
r algorithm has been extended to 3D. In addition, since
Radius of porosity is a function of microstructure. In the pressure distribution has a significant effect on the
majority of cases, this radius is expressed as a function of porosity distribution, a subroutine to check the pressure
dendrite arm spacing [2, 4]: convergence has been added to the calculation. Some
modifications have been also incorporated, such as dif-
l2
r¼ ð11Þ ferent thermo-physical properties (density, heat conduc-
4 tivity and specific hat capacity) for liquid and solid
In this equation, secondary dendrite arm spacing can be phases and variable metal/mold interface heat transfer
calculated from Eq. 3. coefficients.
Numerical method for solving the porosity distribution
is implemented as shown in Fig. 2. Here, the numerical
2.4 Continuity method to solve the pressure distribution and porosity
fraction includes three steps. In the first step, porosity
Continuity or mass conservation equation can be expressed has not been formed while in the second, it has been
by Eq. 12 [28]. formed and is also fed by liquid. But in the third step
which shows the growth, while porosity has been formed
@ρ
þ divðρuÞ ¼ 0 ð12Þ it is not fed by liquid.
@t
After calculating the permeability, secondary dendrite
In this equation, ρ is the average density and can be arm spacing and pore radius for nodes located on mushy
written as ρ0ρsgs + ρlgl + ρpgp and u is the velocity vector. zone, the program checks the volume fraction of porosities
Considering no motion of solid and pore phases, u can be (gp). The positive value of a porosity fraction indicates the
written as glul. Knowing that gs + gp + gl 01 and substitution formation of porosity. With no porosity formed, the gas
of Eq. 4 for “u” (Eq. 4 into Eq. 12), continuity equation is pressure is solved by Eq. 9, considering the local pressure
re-written as follows: of previous time step. Then, the new amount of porosity is
calculated from Eq. 7. If the porosity is formed, when the
@ρ @gp k
ρl þ ρ1 div ðgradPl ρl gÞ ¼ 0 ð13Þ interdendritic fluid flux is positive, the porosity content will
@t @t μgl
be obtained using the same method.
Where ρ ¼ ρl gl þ ρs gs But when the porosity has been formed and the flux of
interdendritic fluid is negative, porosity is calculated from
Eq. 13 by the explicit scheme using local pressure of previous
time step. Here, gas pressure is calculated based on Eq. 7.
3 Numerical method However, according to the algorithm, if initially the
porosity or the pressure must be calculated, the Eq. (13)
The present model of microporosity formation is based on should be solved. This equation can be discretized as
five different pressures that affect the formation of this follows:
defect comprising the atmosphere and metallostatic pres- h o
P iþ1;j;k þPo i1;j;k 2Po i;j;k
gp n ¼ gp o þ ρ ρρ ρl KΔt
n o
sures (or local pressure that preliminary applied via the fluid μ ΔX 2 þ
l
i ð14Þ
flow modeling), the pressure applied due to surface tension P i;jþ1;k þP i;j1;k 2P i;j;k
o o o
P o
þP i;j;k1 2P i;j;k
o o
ΔY 2 þ i;j;kþ1 ΔZ 2
between a pore and neighboring liquid, pressure drop in the
Int J Adv Manuf Technol (2013) 64:1313–1321 1317
To compare of the simulation results with real cast parts, two Fig. 3 Cup-shaped model used to validate the simulation model Zhu et
models have been employed being a cup-shaped casting and al. [4]. Numbers are in millimeters
1318 Int J Adv Manuf Technol (2013) 64:1313–1321
1.E-04
2 = 16 µm
1.E-06
Permeability (m-2)
2 = 35 µm
1.E-08
2 = 48 µm
1.E-10
1.E-12
1.E-14
1.E-16
0 0.2 0.4 0.6 0.8 1
Solid Fraction (-)
60 Permanent Mold
Melt
50
Secondary Dendrite Arm
40
Spacing (µm)
30
20 Present Simulation
H0 = 0.10 cc/100 g Al
10
H0 = 0.137 cc/100 g Al
0
0 20 40 60 80
Distance from the Chill (mm) Insulator
Fig. 5 Secondary dendrite arm spacing along the part centerline in-
cluding present simulation and data of Zhu et al. experiments [4] Fig. 7 Solid model of the car wheel showing the casting parts
Int J Adv Manuf Technol (2013) 64:1313–1321 1319
3 1
To evaluate the permeability inside the casting, the per- second stage, the permeability will be decreased by an
meability of three different points as bottom, center and top upwind curvature until the coherency point. Here, the rate
of the casting with different microstructures have been plot- of permeability drop decreases. At the third stage, that is to
ted in Fig. 6. For a constant solid fraction, permeability in say after the solid fraction of 0.4, due to coherency of
fine microstructures is smaller than that of coarser struc- dendrites, the curvature turns to downwind and the drop
tures. This shows that in fine structures the liquid flow rate increases. It shows that the flow of melt through the
through the mushy zone is quite difficult. It may be imag- dendrites in this stage will get into difficulty by increasing
ined that due to difficult feeding of fluid through the mushy the solid fraction. According to Wisvanathan [23], the final
zone, the potential of porosity formation in finer structures is eutectic transformation takes place at the solid fraction of
higher. But, the time of diffusion of gas in the entrapped 0.95, so the permeability at the end of solidification is
liquid in finer structures is more limited than coarser struc- constant for each cooling rate.
tures and hence porosity nucleation is lower.
Considering Fig. 4, at the initial stages of solidification 4.2 Automotive wheel casting
(for a solid fraction about 0.015 for 16 μm SDAS and 0.01
for 48 μm), during the formation of preliminary solid nuclei, To evaluate the model in complex parts, an automotive
the permeability of liquid decreases abruptly. Here, the wheel was simulated by the present program. The present
permeability values decrease by106 times. This sudden model was coupled with SUTCAST simulation software.
change causes to a sudden drop in pressure and may cause Here, the enmeshment and fluid flow results including pres-
the pressure convergence to get into difficulty during the sure and temperature distribution were exported by the
program solution. By evolution of solidification front in the SUTCAST software. These exported data are imported by
Microporosity Location
1320 Int J Adv Manuf Technol (2013) 64:1313–1321
the present program as initial conditions. The cross section porosities are finer and their concentration decreases.
of the above-mentioned part is shown along with its casting This is because of limitation of gas diffusion and pore
system in Fig. 7. The molten metal was A356 poured against growth.
the gravity vector with 720°C into the steel H13 mold. 4. The present model can be applied to complex parts with
Figure 8a and b, present the positions of porosity forma- a large number of meshes where a good agreement
tion in real casting and simulation. between simulation and experiments can be achieved.
As it can be seen that the simulation (Fig. 8b) predicts the
position of the microporosities with a high precision in this
sample. The position marked by number 1 in Fig. 8a illus- Acknowledgment This project was supported by Razi Metallurgical
trates the porosity that its percentage is certainly higher than Research Center (RMRC), SUTCAST Simulation team, and Sharif
University of Technology. The authors are especially thankful to Ms.
predicted (Fig. 8b). It should be noticed that the sizes and K. Asgari at RMRC for her helpful supports.
volume percents of these pores are high enough to be named
as macropores. In spite of large contents of porosity pre-
dicted in this location, the predicted values are not large
enough as it seems in real part. This is occurred because of References
not taking the formation of macropores (which form in a
bulk entrapped liquid) in simulation into account. In fact in a 1. Campbell J (2000) Casting. Butterworth Heinemann, United
bulk liquid entrapped by a mushy zone front, a macropore Kingdom
2. Zou J, Shivkumar S, Apelin D (1990) Modeling of microstructure
will be formed. In these mushy zones a lot of micropores evolution and microporosity formation in cast aluminum alloys.
also nucleate and grow beside a macropore. So the predicted AFS Trans 98:897–904
position just presents the microporosities formed near a 3. Pequet Ch, Gremaud M, Rappaz M (2002) Modeling of micropo-
macropore. rosity, macroporosity, and pipe-shrinkage formation during the
solidification of alloys using a mushy-zone refinement method:
To describe how the porosities have been formed in these
applications to aluminum alloys. Met Mat Trans A 33:2095–2106
locations, the pressure distribution at two time sections is 4. Zhu JD, Cockcroft SL, Maijer DM (2006) Modeling of micropo-
given in Fig. 9. As it can be seen, there is a high-pressure rosity formation in A356 aluminum alloy casting. Met Mat Trans
gradient at that position. Also the pressure of that region is A 37:1075–1085
5. Hanzel JG, Keverian J (1965) The theory and application of a
higher than that of the neighbors. Due to this reason, this
digital computer in predicting solidification patterns. J Met 5:83–
position acts as a feeder resulting in liquid leakage. 91
6. Imafuku I, Chijiwa K (1983) A mathematical model for shrinkage
cavity prediction in steel casting. AFS Trans 91:527–540
7. Pellini WS (1953) Factors which determine riser accuracy and
5 Conclusions feeding range. AFS Trans 61:61–80
8. Niyama E, Morikawa M (1981) Prediction shrinkage in large steel
A 3D mathematical model was developed to trace the process casting from temperature gradient calculation. Int Cast Met J 6
of solidification and porosity formation. In the present model, (2):16–22
9. Lee YW et al (1990) Modeling of feeding behavior of solidifica-
effective parameters on porosity formation are considered tion Al–7Si–0.3Mg alloy plate casting. Met Trans B 21:715–722
including initial gas content in liquid and its distribution 10. Sabau AS, Viswanathan S (2002) Microporosity prediction in
during the solidification, metal shrinkage, liquid feeding in aluminum alloy castings. Met Mat Trans B 33:243–255
the mushy zone, local pressure, and pores’ surface tension. 11. Katzarov IH (2003) Finite element modeling of the porosity for-
mation in castings. Int J Heat Mass Transf 46:1545–1552
Simulation by the present model is compared with exper- 12. Atwood RC, Lee PD (2002) A three-phase model of hydrogen
imental results published earlier and also porosity distribu- pore formation during the equiaxed dendritic solidification of
tion was compared to results obtained from a real case (an aluminum–silicon alloys. Met Mat Trans B 33:209–221
automotive wheel part). 13. Kubo K, Pehlke RD (1985) Mathematical modeling of porosity
formation in solidification. Met Trans B 16:359–366
Finally, the principal conclusions from this work are 14. Apelian D, Flemings MC, Mehrabian R (1974) Specific perme-
summarized as below: ability of partially solidified dendritic networks of Al–Si alloys.
Metall Trans B 5:2533–2537
1. Evolution of solid increases the complexity of liquid 15. Murakami K, Okamoto T (1984) Fluid flow in the mushy zone
composed of granular grains. Acta Metall 32:1741–1744
path for feeding and decreases the permeability of the
16. Poirier DR, Ocansey P (1993) Permeability for flow of liquid
liquid severely. through equiaxial mushy zones. Matls Sci Eng A 171:231–
2. Increasing the cooling rate results in finer structures and 240
increases the resistance of mushy zone against the liquid 17. Brown SGR, Spittle JA, Jarvis DJ, Walden-Bevan R (2002)
Numerical determination of liquid flow permeabilities for equiaxed
flow.
dendritic structures. Acta Mater 50:1559–1569
3. Porosity distribution depends on cooling rate; in those 18. Mirbagheri SMH, Chirazi A (2006) Simulation of interden-
areas close to the chill where cooling rate is high, dritic liquid permeability for low and high solid fractions
Int J Adv Manuf Technol (2013) 64:1313–1321 1321
during the solidification of mushy alloys. Mat Sci Eng A 24. Kurz W, Fisher DJ (1998) Fundamentals of solidification, 4th edn.
427:51–59 Trans Tech, Switzerland
19. Bernard D, Nielsen O, Salvo L, Cloetens P (2005) Permeabil- 25. Shabani MO, Mazahery A, Bahmani A, Davami P, Varahram N
ity assessment by 3D interdendritic flow simulations on (2011) Solidification of A356 Al alloy: experimental and model-
microtomography mappings of Al–Cu alloys. Mat Sci Eng A ing. Kovove Mater 49:253–264
392:112–120 26. Mantaux O, Lacoste E, Danis M (2002) Numerical prediction of
20. Khajeh E, Mirbagheri SMH, Davami P (2008) Modeling of per- microporosity formation during the solidification of a pure metal
meability with the aid of 3D interdendritic flow simulation for within a porous perform. Compos Sci Technol 62:1801–1809
equiaxed dendritic structures. Mat Sci Eng A 475:355–364 27. See D, Atwood RC, Lee PD (2001) A comparison of three modeling
21. Santos R, Melo MLNM (2005) Permeability of interdendritic approaches for the prediction of microporosity in aluminum-silicon
channels. Mat Sci Eng A 391:151–158 alloys. J Mat Sci 36:3423–3435
22. Carman PC (1937) Fluid flow through granular beds. Trans Inst 28. Patankar SV (1980) Numerical heat transfer and fluid flow.
Chem Eng Lond 15:150–166 Hemisphere, USA
23. Viswanatan S et al (2001) Design and product optimization for cast 29. Bācherud L, Chai G, Taminnen J (1990) Solidification characteristics
light metals. CRADA final report of aluminum alloys, 2nd edn. Foundary Alloys, Norway