Materials Transactions, Vol. 59, No. 10 (2018) pp.

1578 to 1584
© 2018 Japan Foundry Engineering Society

Prediction of Shrinkage Cavity in Heavy-Section Ductile Cast Iron

Using CAE Considering Volume Change during Solidification+1
Yutaka Miyamoto1,+2 and Haruki Itofuji2
1
Ube Steel Co., Ltd., Ube 755-0067, Japan
2
Adstefan Casting Solution Center, Tohoku University, Sendai 980-8579, Japan

A new analysis parameter, which is the volume balance during solidification in heavy section ductile cast irons, was adopted to predict
shrinkage cavities by computer simulation. To realize higher precision and quantify the behavior of expansion/contraction during solidification,
the temperature and solidification ratio of each reaction stage from the start to completion of solidification were determined by the tangent line
method. The expansion/contraction amounts of each reaction were calculated from chemical compositions of carbon and silicon and the initial
temperature of the test material. Finally, the expansion/contraction degree was calculated by dividing the amounts of the expansion/contraction
at each reaction by the solidification ratio. The quantified value was input into the casting simulation software as the expansion/contraction
amounts. The result showed better matching compared to the actual shrinkage phenomenon. [doi:10.2320/matertrans.F-M2018833]

(Received May 7, 2018; Accepted July 3, 2018; Published September 25, 2018)
Keywords: ductile cast iron, heavy-section, CAE, simulation, shrinkage

1. Introduction FCD, can be evaluated using various analysis software,11­15)

It is known that sound cast iron can be produced from
heavy-section ductile cast iron (large FCD) under riserless
conditions owing to the expansion pressure and volume of
eutectic graphite.1­4) Large FCD does not have any shrinkage
cavities. However, while producing plate-type cast iron,
shrinkage cavities may be generated under a riserless
condition; thus, the riserless technique cannot be applied to
the formation of some shapes. A practical calculation of the
riserless safety index, based on the modulus of casting (Mc)
and/or the shape factor, is used to determine whether the
riserless design is applicable to create a desired shape.5­7)
When it is determined that a riser is necessary, the risers
are designed by a similar method to that used for designing
cast steel. In addition, the use of heat balancers, which would
not undergo shrinkage themselves, has been proposed to
minimize the size of the riser needed.8,9) However, in heavy-
section FCD, the shape of cast iron is more complicated,
therefore it is more difficult to predict the portions that will
generate shrinkage cavities. If weld repairs are not feasible
due to the quality specifications, the part must be produced
again (even if it is a few tons of cast iron), which would result
in a significant loss. A design with excessive risers is
employed to reduce the risk of such losses. However, this
can reduce the work efficiency and/or product yield in some
cases due to the difficulty in predicting the shrinkage cavities
formed in FCD by computer-aided engineering (CAE).
Various prediction methods can be used depending on the
material and manufacturing method,10) including the Niyama-
criterion method, temperature-gradient method, and closed-
loop method. The accuracy of these analyses has been
established for cast steel and cast aluminum alloy, among
others, which are formed in a skin-like solidification
process by continuing contraction. However, the mushy-type
solidification, accompanied by graphite expansion in cast
+1
This Paper was Originally Published in Japanese in J. JFS 90 (2018) 175­
181.
+2
Corresponding author, E-mail: 34271ubs@ube-ind.co.jp
but the prediction accuracy is inefficient with these
approaches.16) To achieve an appropriate prediction accuracy,
there is a need for analysis software that takes the FCD-
specific solidification process and volumetric changes into
consideration.
To predict the volumetric changes upon solidification, the
proposed approaches include measuring the expansion and
contraction of a cast material in a measurement mold or
calculating the expansion and contraction of a material from
its chemical composition. However, even when the chemical
composition of a material is known, the volumetric balance is
theoretically positive owing to the generation of shrinkage
cavities. Therefore, it is used only as a referential index to
assess the tendency toward the generation of shrinkage
cavities after casting.17­20) Such issues may occur because, as
mentioned above, the generation of shrinkage cavities varies
depending on the size of the cast material and also causes
variations in the timing, ratio, and amount of expansion/
contraction during the solidification process. However, there
have been a few attempts to simulate solidification that
involved the quantitative study and characterization of these
factors. Accordingly, in this study, we explored a method for
predicting the generation of shrinkage cavities, involving the
determination of the theoretical volumes of contraction and
expansion associated with solidification. Then we conducted
calculations by substituting these volumes for the corre-
sponding areas that were divided at the flexion points of a
cooling curve during solidification.
2. Experimental Procedure
2.1 Shrinkage cavity analysis
To simulate the expansion and/or contraction that occur
during the solidification process, a solidification analysis is
conducted to predict the formation of shrinkage cavities.
This analysis involved the quantitative determination of
volumetric changes, which are then used as input values
for the simulation software. First, using the experimentally
measured cooling curve from the solidification process, as
 Prediction of Shrinkage Cavity in Heavy-Section Ductile Cast Iron Using CAE Considering Volume Change during Solidification
1600
䐟Liquid contraction
1540 䐟Liquid contraction
+ 䐠Primary austenite contraction or graphite expansion
䐟Liquid contraction
1480
Temperature,K
䐢Austenite contraction
between eutectic cells
1420
䐡Contraction of austenite
and expansion of graphite
during eutectic solidification
1360
0.0 Solidification ratio
1300
0 20 40 60
Time,min
Fig. 1 Critical points of volumetric change in solidification curve of
300 © 300 © 150 mm sample casting, for example.
shown in Fig. 1, the flexion points were determined by a
tangent-line method and temperature and solidification
ratio were determined for each point. We assumed that
the change in the solidification ratio takes place when the
pouring was completed and that the respective reactions
occurred in the following order: (1) liquid contraction ¼ (1)
liquid contraction and (2) proeutectic reaction ¼ (1) liquid
contraction ¼ (3) eutectic reaction ¼ (4) austenite contrac-
tion between eutectic cells at the last stage of solidification.
Thereafter, the obtained values for the C and Si contents
and pouring completion temperature were substituted into
eq. (1) to determine the theoretical volumetric change.5)
TV ¼ Sl þ Epg ðor Sp£Þ þ Eeg þ Se£
TV = Volumetric change
Sl = Liquid contraction (vol%)
Epg = Expansion of proeutectic graphite (vol%)
Sp£ = Contraction of proeutectic austenite (vol%)
Eeg = Expansion due to the crystallization of eutectic
graphite (vol%)
Se£ = Contraction due to the crystallization of eutectic
austenite (vol%)
In the above equations, Epg was used when the chemical
composition was hypereutectic and Sp£ was used when
the composition was hypoeutectic. These values may be
determined by the following equations:
Sl ¼ ððTi  1423Þ=100Þ  1:5
Epg ¼ ðCx  CeÞ=ð100  CeÞ  3:4  100
Sp£ ¼ ðCe  CxÞ=ðCe  C£Þ  3:5
Eeg ¼ ðð1  SlÞ=100Þ  ðð100  CxÞ=ð100  CeÞÞ
 ððCe  C£Þ=ð100  C£ÞÞ  3:4  100
Se£ ¼ ðð1  SlÞ=100Þ  ðð100  CxÞ=ð100  CeÞÞ
 ðð100  CeÞ=ð100  C£ÞÞ  3:5
Ti = Initial temperature of the molten metal in the mold (K)
Ce = Carbon content at the eutectic point (mass%)
Cx = Carbon content of the molten metal (mass%)
C£ = Carbon solid solution content in austenite (mass%)
Here, the liquid contraction was assumed to be 1.5 vol% per
1579
Table 1 Physical properties and heat transfer parameters for plate test
blocks.
1.0
100 K.1,5,21,22) Furthermore, Ce and C£ were determined by
80 100 120
the following equations:
Ce ¼ 4:27 þ Si=3 ð7Þ
C£ ¼ 2:045  0:178  Si ð8Þ
Si = Silicon content in the molten metal (mass%)
Dividing the expansion and contractions volumes in the
reactions by respective solidification ratios, the degrees of
expansion and contraction were calculated. The actual
calculation results will be reported in Chapter 3.
For the solidification analysis, ADSTEFAN Ver. 2016
from by Hitachi Industry & Control Solutions, Ltd. was used.
The solidification analysis was conducted without conducting
a fluidity analysis before and assuming that the temperature
of the molten metal was consistent in the respective areas of
the mold.
The predicted expansion and contraction behavior were
entered into the software in the form of numerical values for
the solidification shrinkage rates. These rates were used to
ð1Þ
determine the amount of solidification shrinkage15) by taking
the flow of the molten metal into account and multiplying
the increase in the solidification ratio by the solidification
shrinkage rate. In the analysis, the flow of molten metal was
assumed to occur in partially flowable regions, in which the
solidification ratio was greater than the critical solid fraction,
and it was assumed that the supply of molten metal occurs
only from adjacent elements. When the solidification ratio
was smaller than the critical solid fraction, the area was
assumed to be a free-flow area and was divided such that
the supply occurs from the elements within that free-flow
area. The analytical predictions regarding the solidification
shrinkage were expressed as indices, which are known as
ð2Þ “filled ratios”. The physical properties and boundary
conditions, used as the initial values, in the solidification
ð3Þ analysis were representative of those in practical situations
ð4Þ and are shown in Table 1. Moreover, to reproduce the
expansion and shrinkage behavior, the flow-critical solid
ð5Þ fraction was set to 1.0. The splitting mesh size, in the three-
dimensional model, was set to 5 mm.
ð6Þ 2.2 Preparation of test blocks for measuring the
temperatures and shrinkage cavities
It is known that there are various factors, other than the
solidification characteristics, that lead to the generation of
shrinkage cavities. In this study, disregarding the noise
factors, the tendency toward cavity generation due to
1580 Y. Miyamoto and H. Itofuji

1693
Flow off, 30
b) ∼1723K
1603
∼1633K
300㽢220㽢70 Pad,70㽢220㽢70 d)

c)
1.0% Cover
a) 0.3% Inoculant
(a) Basic shape (Riserless) (b) Pad 1.2% Spheroidizer 1370 ∼1673K

1) Slag off 3) Liquid treatment Transfer 4) Pouring

Runner, 2) Super heatingӍ1773K,5min Within 5min
Chillers, 100㽢50㽢50
(55-60)㽢70 Sprue,
Two pieces at the top and bottom
60㽢500 Fig. 3 Temperature and time schedule for melting, liquid treatment and
Closed riser,
240㽢240 Gate, pouring.
100㽢15㽢150

composition was analyzed after the spheroidizing treatment

e) Sprue gating desgin
(c) Chillers (d) Riser
Fig. 2 Casting design for plate test blocks. (Unit; mm)
shrinkage, under the manufacturing conditions of the heavy-
section FCD, was characterized experimentally in four types
of plate test blocks. The shapes and casting designs of the
test blocks are shown in Fig. 2. A riserless design with a
contraction rule of 8/1000 (220 mm wide, 300 mm deep, and
70 mm in thickness), as shown in Fig. 2(a), was used as the
basic shape. For the pad design shown in Fig. 2(b), a pad
having dimensions of 220 © 70 © 70 mm (width © depth ©
thickness) was installed on the top part of the test block at
the center in the longitudinal direction. For the chillers design
shown in Fig. 2(C), two chillers with dimensions of 50 ©
100 © 50 mm (width © depth © thickness) were installed on
the top and bottom of the test block. For the riser design
shown in Fig. 2(d), a neck-down riser having dimensions of
240 © 240 mm (diameter © height) was installed on the top of
the test block. A flow-off of ¤ = 30 mm was installed in one
position at the center of the top surface of each test block. In
addition, to minimize the variables in the casting conditions,
such as chemical composition and pouring temperature, the
casting mold was made by connecting the four types of test
blocks via runners within one casting frame. The casting mold
was prepared by kneading silica sand, 0.8 mass% furan resin,
and 40 mass% (relative to the resin) curing agent, to obtain a
mold strength of 4.5 MPa or greater, and prevent movement
of the mold wall. Alcoholic MgO-type material was used as a
mold-coating material. After drying the mold, by firing, it was
further air-dried for 24 hours or more to fully recover the mold
strength.
The method of melting and pouring the metal is shown in
using an emission spectrophotometric analyzer. Setting the
target chemical composition as hypoeutectic component(s),
another identical mold was prepared with a K thermocouple
at the center of the thickness of the plate section of each type
of test block to measure the cooling curve during solid-
ification.
3. Experimental Results
3.1 Casting of 70 mm-thick test blocks
(1) Pouring results
Table 2 shows the obtained chemical compositions of the
test blocks upon completion of the pouring; the composition
was equivalent to that of FCD 450 and yield of Mg was
satisfactory.
Figure 4 shows the cooling curves measured during the
solidification of the cast test blocks. The eutectic temperature
increased, and the solidification completion time increased
from the chiller deign to the riserless design to the pad
design to the riser design. These findings confirm that the
temperature measurements were successfully conducted.
Table 2 Chemical composition of plate test blocks. (mass%)
1600
Riserless
1550
Pad
Chiller
1500
Riser
Temperature,K

Fig. 3. A low-frequency induction furnace was used to melt

the return materials from production. After melting to a 1450
temperature of 1,723 K, the composition was adjusted to
the desired ratios. After superheating for 5 minutes at a
1400
temperature of 1,770 K or higher, the melt was allowed to
cool gradually and the molten metal was tapped. For the
1350
spheroidizing treatment and inoculation, 1.2 mass% of an
alloy of Fe­45 mass% Si­5.8 mass% Mg, 0.3 mass% of an
alloy of Fe­50% Si, and 1.0 mass% of a cover material were 1300
0 1000 2000 3000 4000 5000 6000
put in a ladle sequentially and a sandwich method was
Time,sec
employed. The time between the end of the reaction and
completion of pouring was five minutes. The chemical Fig. 4 Cooling curves during the solidification of plate test blocks.
 Prediction of Shrinkage Cavity in Heavy-Section Ductile Cast Iron Using CAE Considering Volume Change during Solidification 1581

1,650

1,600
(a) Basic shape (b) PadPad
Riserless (c) Chillers
Chllers (d) Riser
Riser
(Riserless)
1,550
Temp,K

1,500 0.08
0.09 0.07 0.07
0.10 0.18 0.12
0.14
1,450 0.33 0.32
0.35 0.96 0.97 0.97
0.50
0.96
1,400

1,350
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Solidification rate Solidification rate Solidification rate Solidification rate

Fig. 5 Relationship between temperature and solidification ratio at each reaction of plate test blocks.

Table 3 Theoretical volumetric change of plate test blocks.

10.0
䐡Eutectic austenite contraction
and expansion of graphite
5.0
䐟Liquid
+3.27
-3.19 䐡 +3.16 䐡 +3.27 䐡 +3.17
Expansion/Contraction

contraction
0.0
䐟 -1.37
-2.48 䐟 -2.28 䐟 -2.63
-5.0
-0.19
䐢Austenite contraction
䐢 -0.15 䐠 -0.27 䐢 -0.23 䐢 -0.17
between eutectic cells
-10.0
䐠 -0.27 䐠-0.27
-15.0 -0.27
䐠Primary austenite (a)Basic shape
(b)Pad (c)Chillers (d)Riser
contraction (Riserless)
-20.0
0 0.2 0.4 0.6 0.8 10 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 10 0.2 0.4 0.6 0.8 1
Solidification ratio Solidification ratio Solidification ratio Solidification ratio

Fig. 6 Relationship between expansion and contraction at each reaction of plate test blocks.

(2) Characterization of the expansion and contraction temperature of the molten metal in the mold upon completion
behavior of pouring varies, the cooling curves were measured during
Figure 5 shows the relation between the temperature and solidification and the values were calculated from each curve
solidification ratio, in the respective reactions during the independently. The data shows that the volumetric balance
solidification process, as determined by the experimentally was net positive by the time the solidification was completed
measured cooling curves. The periods for which the length in all samples. This shows that, theoretically, the chemical
of the solidification completion time of the test block was composition used here does not generate shrinkage cavities,
correlated with the solidification ratio of the respective even in the riserless design. Generally, a eutectic composition
reaction. This included only the eutectic reaction period and that results in minimal shrinkage cavities should be used.
the inter-cell austenite shrinkage period generated in the last However, a hypoeutectic composition may be employed
stage of the solidification. for some heavy-section FCD as a measure to control the
In samples with long solidification times, the eutectic generation of chunky graphite, which frequently causes
reaction period was large and period of inter-cell austenite problems.23­25) Figure 6 shows the behavior of expansion and
shrinkage was small. Table 3 shows the theoretical volu- contraction during the solidification process. The degrees of
metric changes in the different test blocks. Since the initial expansion and contraction were calculated by dividing the
1582
(a) Riserless Pad
(b)
(c)
Fig. 7 Result of CAE analysis for plate test blocks.
pffiffiffiffi a) Shrinkage distribution at cross section of test blocks. (The right side of the photos is
the gate) b) Filled ratio 99.9³0.0%. c) G/ R 0.7³0.0.
calculated theoretical volumetric changes for the respective
reactions by the solidification ratio. On comparing the
degrees of expansion and contraction for the respective
reactions in the test blocks having different solidification
times, it was found that the degree of contraction of the inter-
cell austenite in the last stage of the solidification process
decreased as the time required to complete the solidification
increased.
(3) Results of the shrinkage cavity analysis
Figure 7(a) shows the result of a penetrant examination of
the center cross-sections of the different test blocks. In the
riserless design, there was a positive indication that shrinkage
cavities were generated near the center of the thick portion. A
similar finding was observed in the pad design near the center
of the thick portion. In the chiller design, there was a positive
indication that shrinkage cavities were generated near the
center of the thick portion on both ends. Moreover, the area
of the shrinkage cavities was larger on the gate side. In the
riser design, there was no indication on the plate portion, but
there was an indication inside the riser that shrinkage cavities
were generated. Figure 7(b) shows the analytical results of
this study in terms of the filled ratio; the behavior of
expansion and contraction was represented as the solid-
ification contraction rates in the software and the amount
of solidification contraction was determined.pffiffiffiffi Figure 7(c)
shows results obtained by the G/ R method. The
observation points were at the same positions as those on
the cross sections where the shrinkage cavities were observed
Y. Miyamoto and H. Itofuji
Chiller Riser
experimentally. The filled ratio is expressed as an index of
the amount of solidification contraction and threshold of the
shrinkage cavities was set as not greater than 99.9%. This is
based on the assumption that a filled ratio of 100% represents
the completely filled state and that the smaller shrinkage
cavities are generated when the filled ratio is less than 100%.
It was observed that the positions where shrinkage cavities
were actually generated corresponded closely to the predicted
positions of the shrinkage cavities according to the filled ratio
measurements. However, the actual positions of the shrinkage
cavities
pffiffiffiffi were slightly higher than those that predicted by the
G/ R method. In addition, the results show that the final
solidification positions were shifted upward from the center
of the thick portions in the actual casted metal blocks.
3.2 Prediction and actual results of large heavy-section
test blocks
FCD is known for not generating shrinkage cavities when
the Mc is high enough and shape is close to a cube.
Figure 8(a) shows the result of the penetrant examination of
cross sections from the 600 mm cubic test blocks (hereinafter,
Mc = 10 cm) cast with a hypoeutectic composition and cut
at the center. Although these samples had hypoeutectic
compositions, similar to those of the plate test blocks,
there was no indication of shrinkage cavities in the center.
pffiffiffiffi
However, according to predictions made by the G/ R
method, the portion at the center of the cube, which solidifies
last, was always indicated as an at-risk area. Thus, it was
 Prediction of Shrinkage Cavity in Heavy-Section Ductile Cast Iron Using CAE Considering Volume Change during Solidification 1583

(a) (b) 1600

Riser
1550 Pad
Riserless
1500
Chiller
Mc=10cm

Temperature,K
1450

100mm
1400
Fig. 8 Cross section of Mc = 10 cm block and analysis result. a) Shrinkage
distribution. b) Filled ratio 99.9³0.0%.
1350

considered as impossible to produce cast metal without 1300

shrinkage cavities with a riserless design. However, the
proposed results reveal that it is actually possible to produce
cast metal without shrinkage cavities. This is because, when
the solidification time is long, eutectic cells of graphite and
austenite grow and fill the shrunk portions during the last
stage of the solidification process. Therefore, it is also
necessary to predict the absence of shrinkage cavities in
addition to the generation positions of shrinkage cavities in
order to make accurate predictions for heavy-section FCD.
As shown in Fig. 8(b), the solidification analysis conducted
on Mc = 10 cm test blocks indicated that no shrinkage
cavities would be generated following the final solidification.
Figure 9 shows the cooling curves of the plate test blocks
and the Mc = 10 cm test blocks during solidification. The
solidification time of the Mc = 10 cm test blocks was
sufficiently long (9 hours or longer). Figure 10 shows the
solid phase ratio and the amount of expansion and shrinkage
in the Mc = 10 cm test blocks. The solidification shrinkage
area and amount of shrinkage of the eutectic inter-cell
austenite were smaller than that of the plate test blocks.
Assuming that the mechanism of generating a shrinkage
cavity is partially related to the growth of eutectic cells upon
solidification, this behavior can be reproduced by the
proposed method. Therefore, it can be considered that the
tendency of generating shrinkage cavities was lower even in
the analytical results.
0 10000 20000 30000 40000 50000
Time,sec
Fig. 9 Cooling curves during solidification of plate test blocks.
4. Discussion
In this study, the expansion and shrinkage of the metal
during solidification were considered. In the prepared test
blocks, factors influencing the generation of shrinkage
cavities, such as gas generation and the movement of the
mold wall, were excluded. Therefore, both the test blocks and
analytical results can be considered to have simulated the
shrinkage cavity generation occurring due to solidification.
According to the observed shrinkage cavities in the 70 mm-
thick test blocks, although the theoretical volumetric balance
was positive in the view of chemical composition, shrinkage
cavities were generated in the final solidification positions.
Even in the analytical results, the positions of the generated
shrinkage cavities were at the same positions and it was
possible to set a reasonable and easily understandable
threshold stating that shrinkage cavities are generated when
the filled ratio is 99.9% or less. Furthermore, the areas of the
shrinkage cavities generated on the left and right sides of the
test blocks were not equal in the chiller design. This may take
place due to the influence of the gate. However, if we assume

1,650 10.0

1,600 5.0
Mc=10
Expansion/Contraction

䐡 +3.05
1,550 0.0
䐟 -2.10
Temp,K

1,500 -5.0
0.11 䐢 -0.09
1,450 -10.0
0.98
0.07
0.30
1,400 -15.0 䐠 -0.17 Mc=10

1,350 -20.0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Solidification ratio Solidification ratio

Fig. 10 Relationship between “temperature and solidification ratio” and “expansion and shrinkage” in each reaction of Mc = 10 cm block.
1584 Y. Miyamoto and H. Itofuji
that the tendency toward shrinkage cavity generation is
generally consistent and variations would be negligible for
estimating the shrinkage cavities, the precision seems to be
improved by quantitatively calculating the expansion and
shrinkage and incorporating a solidification simulation
compared with the predictions
pffiffiffiffi obtained by the theoretical
volumetric balance or G/ R methods.
5. Conclusion
In this study, the expansion and shrinkage that occur upon
solidification were calculated and compared with the values
obtained from the pouring tests. From the results, the
following conclusions were made:
(1) We developed a method of evaluating the formation
of shrinkage cavities based on the expansion and
shrinkage behavior calculated from the cooling curve
during solidification and theoretical volumetric balance.
(2) The precision of determining the positions of shrinkage
cavities was improved by considering the behavior of
expansion and shrinkage.
(3) By considering the amount of shrinkage of eutectic
inter-cell austenite, the predicted presence or absence of
generated shrinkage cavities was also consistent with
the experimental findings.
(4) We were able to confirm that the proposed approach
is useful as a practical method for predicting the
shrinkage cavities that will form in heavy-section FCD
manufactured with fran bonded sand.
REFERENCES
1) S.I. Karsay: Ductile Iron I-Production, (Quebec Iron and Titanium,
1992) pp. 158­161.
2) B. Chang: IMONO 55 (1983) 113.
3) M. Taffazzoli and V. Kondic: Foundry Management and Technology
104 (1976) 86.
4) H. Itofuji, K. Kawamura, N. Hashimoto and H. Yamada: AFS Trans.
98 (1990) 585­595.
5) B. Chang, K. Akechi and K. Hnawa: Spheroidal Graphite Cast Iron ®
Basis, Theory, Practice®, (Agne, 1983).
6) T. Yoshida, K. Yano and M. Kawabata: J. JFS 71 (1999) 104.
7) T. Kotani, K. Edane, K. Iwakado and H. Itofuji: Castech (2016).
8) H. Itofuji, M. Tamura, H. Ito, T. Nishimura and Y. Esashika: Mater.
Trans. 51 (2010) 103­109.
9) K. Ishikawa, T. Kotani, T. Ogino and H. Itofuji: The 12 AFC (2013) 46.
10) E. Niyama, T. Uchida, M. Morikawa and S. Saito: AFS Int. Cast Met. J.
7 (1982) 52.
11) N. Deki: J. JFS 86 (2014) 914.
12) T. Tashiro: J. JFS 86 (2014) 922­926.
13) A. Murakami and M. Takishita: J. JFS 86 (2014) 931.
14) T. Yonezawa: J. JFS 86 (2014) 940­945.
15) I. Takahashi, T. Uchida and K. Anzai: J. JFS 78 (2006) 661­667.
16) E. Niyama and K. Anzai: IMONO 67 (1995) 30.
17) C.E. Bates: AFS Trans. 85 (1979) 289.
18) S. I. Karsay: Ductile Iron ® The State of the Art, (1980) p. 26.
19) K. Ikawa: IMONO 41 (1969) 714.
20) T. Miyamoto, H. Narita and H. Kudoh: J. JFS 68 (1996) 843­851.
21) K. Su, I. Ohnaka, I. Yamauchi and T. Fukusako: J. JFS 58 (1986) 702­
706.
22) T. Yoshida and M. Kawabata: J. JFS 70 (1998) 336­342.
23) S.I. Karsay: AFS Trans. 78 (1970) 85.
24) S. Toyoshima: IMONO 67 (1995) 133.
25) H.W. Hoover: AFS Trans. 94 (1986) 601.