Materials Science and Engineering A 393 (2005) 310314

On-line prediction of carbon equivalent on high-nickel

austenitic ductile iron
Qin Hua

, Yuhui Zhang, Yongshen Yan

Department of Material Science and Engineering, Shanghai University, Shanghai 200072, Peoples Republic of China
Received 14 July 2004; accepted 25 October 2004
Abstract
In this paper, experiments have been made on high-nickel ductile iron for controlling the property by the computer-aided thermal analysis
system. The experimental results have been analyzed with statistics and applied to on-line predicting and controlling carbon equivalent,
which obtained satisfying result. The experiments show that the relationship between the carbon equivalent of high-nickel ductile iron and its
liquidus temperature is linear, which can be expressed as: CEL=15.7826 0.0096575 T
L
. In order to ensure the tensile strength greater than
400 MPa with the probability up to 99%, the liquidus temperature of high-nickel austenitic ductile iron must be in the range of [12031226

C].
2004 Elsevier B.V. All rights reserved.
Keywords: Carbon equivalent; High-nickel ductile iron; On-line prediction; Cooling curve
1. Introduction
Despite its cost, high-nickel austenitic ductile ironis one of
the most widely utilized materials in corrosive environments
due to its excellent heat and corrosive resistance [1]. An ab-
normal type of graphite may occur in high-nickel austenitic
ductile iron with nickel contents ranging from 13 to 37%.
This type of graphite forms as ne ake-like chunks in the
most slowly cooled portions of casting. It was reported that
the carbon, silicon and nickel contents in austenitic ductile
iron are adjusted according to the formula: [2].
TC%+0.2 Si%+0.06 Ni% 4.4
(TC% = total carbon %)
If not, the presence of chuck graphite can be detected
in many cases. According to ANSI/ASTM A43989, D5-S
high-nickel austenitic ductile iron contains: C%2.30; Si%
4.905.00; Ni% 34.037.0; the left-side value of formula is
easily over 4.4. The lower the carbon content, the longer is the

Corresponding author.
E-mail address: qhua@mail.shu.edu.cn (Q. Hua).
dendrite arm and castings will exist much shrinkage [3]. So,
it is important to control the level of carbon equivalent for the
high quality of high-nickel austenitic ductile iron castings.
The thermal analysis used to study Mg-treated nodu-
lar graphite iron that can be traced to the early 1970s [4].
Computer-aided thermal analysis system can provide infor-
mation about the composition of alloy and determine degree
of modication and grain rening on aluminum alloy and
graphite morphology on cast iron [5,6].
The main objective of the research effort presented in this
paper has been to set up the relationship between solidi-
cation model and carbon equivalent, in order to control the
composition and property of high-nickel austenitic ductile
iron castings.
2. Experimental procedures
The composition of metal charge was C%: 1.52.0, Si%:
4.55.0, Mn%: 0.150.20, Ni%: 34.037.0, Cr%: 1.62.0,
P%0.02, S0.02, which was melted by coreless induc-
tion furnace of 150 kg capacity. The melt was superheated
at 16001620

C. After superheating the melt was treated in

the ladle with nodulizing alloys (magnesiumnickel ferrosil-
0921-5093/$ see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2004.10.029
Q. Hua et al. / Materials Science and Engineering A 393 (2005) 310314 311
Fig. 1. Structure of thermal analysis system.
icon) by the trigger method. After the reaction was completed
and slagged off, post inoculation was performed with 75%
ferrosilicon at 15301550

C.
The shell cup of 50 mm in diameter and 60 mm in height,
into which a type of RtRh thermocouple was inserted and
poured and a cooling curve was recorded by computer-aided
analysis system(Fig. 1). At the same time other samples were
poured for microstructure and mechanical property.
3. Experimental result and discussion
3.1. Typical thermal analysis curve of high-nickel
austenitic ductile iron
Fig. 2 shows a cooling curve, a rst-derivative curve also
called cooling rate curve and a second-derivative curve also
called cooling acceleration curve, which contain a great deal
of informationabout solidicationof cast iron. Twoplatforms
at the cooling curve, T
L
(temperature of liquidus arrest) and
T
EU
(temperature of eutectic undercooling), are widely ap-
plied at pouring station for quality control [7]. Two peaks
at the rst-derivative curve, which are called austenite peak
and eutectic peak, can be used to calculate the amount of
austenite and eutectic [8]. Though the cooling acceleration
curve was much undulant, it may help to determine the critical
point.
3.2. The effect of carbon equivalent on the character of
cooling curve
As known to all, cast iron with different carbon equiv-
alent has a different form of cooling curve. In general, the
lower of carbon equivalent the higher the temperature of
austenite arrest and the lower eutectic action, the same result
also take place in high-nickel austenitic ductile iron shown in
Fig. 3.The carbon equivalent of high-nickel austenitic ductile
iron can be calculated as the following formula: [9]
CEL = C%+0.33(%Si) +0.047(Ni%)
0.0055(%Ni)(%Si)
In this paper, the carbon equivalent was also calculated as
the formula and the data in Table A.1 were the result of this
experiment (see Appendix A).
3.3. Linear regression
Suppose that the relationship between carbon equivalent
andliquidus temperature is linear, bythe least squares method
Fig. 2. Typical thermal analysis curves.
312 Q. Hua et al. / Materials Science and Engineering A 393 (2005) 310314
Fig. 3. Cooling curves of different carbon equivalent.
a regression linear formula as following is derived from the
data of Table A.2 (see Appendix A).
CEL = A +B T
L
where
B =

27
i=1
S
xy

27
i=1
S
xx
= 9.6575 10
3
S
xy
= (T
Li

T
L
)(CEL
i
CEL)
S
xx
= (T
L

T
L
)
2
A = CEL B

T
L
= 15.7826
Thus, the rst linear regression formula can be described as:
CEL = 15.7826 0.0096575 T
L
In Fig. 4, its shape is shown and the points are the data of this
experiment.
Fig. 4. The regression line for carbon equivalent and liquidus temperature.
3.4. Signicant test
According to the theory of statistics, if
B

S
xx
> t
/2
(n 2)
the linear regression is signicant, where:
=

Q
e
/n 2, a variance of point estimation;
Q
e
=
n

i=1
(CEL CEL)
2

i=1
(T
L

T
L
)
2
,
a residual sum of squares;
n, times for experiment;
t
/2
(n 2), Students distribution with n degrees of free-
dom against (1 ), which is a condence level.
Thus,
=

Q
e
/n 2 = 3.0106 10
2
,
regarded as the error of carbon equivalent
So,
B

S
xx
= 25.656 > t
/2
(n 2) = 2.787
It is clear that the linear regression is considerable condence.
The t
/2
(n 2) can be resulted from the table of probability
points of t distribution [10] and the condence level (1 )
is 0.99.
3.5. Carbon equivalent predicting and controlling
In order to predict the carbon equivalent of high-nickel
austenitic ductile iron accurately, it is necessary to determine
a reasonable interval of carbon equivalent for a certain liq-
uidus temperature, T
L0
. According to statistical law, the prob-
ability of coverage of carbon equivalent against certain T
L0
is as follows:
P{C

EL (T
L0
) < CEL
0
< C

EL +(T
L0
)} = 1 ;
where
(T
L0
) = t
/2
(n 2)

1 +
1
n
+
(T
L0

T
L
)
2
S
xx
;
C

EL = A +B T
L0
.
If 1 =0.99 and 1190

C<T
L0
<1240

C, then the cover-

age of CEL
0
is shown in Fig. 5. When T
L0
<1199

C the
probability of carbon equivalent within [4.11, 4.29] is 99%.
Because,
CEL
0
= A +B T
L0
t
/2
(n 2)

1 +
1
n
+
(T
L0

T
L
)
2
S
xx
,
Q. Hua et al. / Materials Science and Engineering A 393 (2005) 310314 313
Fig. 5. Predicting range of carbon equivalent.
If the value of T
L0
is very close to

T
L0
and n is bigger
enough, then

1 +
1
n
+
(T
L0

T
L
)
2
S
xx
1;
Thus,
CEL
0
= A +B T
L0
t
/2
(n 2) .
In order to guarantee the tensile strength of casting is greater
than 400 MPa, the carbon equivalent must be controlled in-
side [3.86, 4.25%], since the relationship between carbon
equivalent and tensile strength is shown as Fig. 6 in this ex-
periment.
In order to ensure the carbon equivalent within [3.86,
4.25%], an interval of liquidus temperature must be deter-
mined, soliquidus temperature is controlledinside [T
L1
, T
L2
].
Thus,
T
L1
= CEL
1
A +t
/2
(n 2) /B = 1203
T
L2
= CEL
2
A t
/2
(n 2) /B = 1226
here
CEL
1
= 3.86% CEL
2
= 4.25%.
If the range of liquidus temperature is within[12031226

C],
the carbon equivalent is certainly inside [3.86, 4.25%], that
Fig. 6. The relationship between carbon equivalent and tensile strength.
is to say, the tensile strength of high-nickel ductile cast iron
is greater than 400 MPa with the probability up to 99%.
4. Conclusions
(1) It is obvious that the relationship between the carbon
equivalent of high-nickel ductile iron and its liquidus
temperature is linear, which can be expressed as: CEL =
15.7826 0.0096575 T
L
.
(2) In order to ensure the tensile strength greater than
400 MPa with the probability up to 99%, the liq-
uidus temperature must be inside the range of [1203
1226

C].
(3) By the computer-aided thermal analysis system on-line
prediction of carbon equivalent on high-nickel ductile
ironis available, the predictingprecisionof carbonequiv-
alent is within 3.01%.
Acknowledgments
The author is grateful to Professor Zhenghua Zhu and
Zhenghua Pang of Shanghai University, for their encourage-
ment and suggestions over the years on the thermal analysis
in the foundry. Much appreciation is also expressed to Yimin
Wang, Dewei Yao, Gouhua Cai and Liuqian Jiang, who of-
fered great help in this experiment.
314 Q. Hua et al. / Materials Science and Engineering A 393 (2005) 310314
Appendix A
See Tables A.1 and A.2.
Table A.1
Composition, carbon equivalent and liquidus temperature
No. of furnace C (%) Si (%) Ni (%) CEL T
L
S1 1.60 4.87 34.0 3.89 1234
S2 1.77 4.93 34.2 4.08 1211
S3 1.57 5.26 38.6 4.00 1218
S4 1.50 4.40 34.9 3.75 1245
S5 1.41 4.46 34.9 3.67 1250
S6 1.48 4.97 35.3 3.81 1236
S7 1.73 3.89 36.0 3.94 1230
S8 1.54 4.98 35.0 3.87 1234
S9 1.97 4.78 34.7 4.27 1187
S10 1.78 4.98 34.6 4.10 1209
S11 1.91 4.70 35.2 4.21 1198
S12 1.82 5.01 34.2 4.14 1209
S13 1.53 5.01 36.8 3.90 1229
S14 1.80 4.75 34.9 4.10 1208
S15 1.90 4.93 35.1 4.22 1199
S16 1.81 4.96 34.6 4.13 1209
S17 1.74 4.87 35.0 4.05 1218
S18 1.60 4.74 34.2 3.88 1232
S19 1.82 5.01 34.2 4.14 1208
S20 1.69 5.18 34.0 4.03 1215
S21 1.76 4.82 35.0 4.07 1208
S22 1.78 5.00 35.4 4.12 1207
S23 1.87 5.00 34.7 4.20 1203
S24 1.70 4.98 34.1 4.01 1223
S25 1.88 4.56 35.3 4.16 1208
S26 1.93 4.49 35.2 4.20 1195
S27 1.84 4.89 35.3 4.16 1206
Table A.2
Data of regression
No. of
furnace
CEL T
L
S
yy
S
xx
S
xy
1 3.89441 1234 0.0212 327.61 2.63518
2 4.07697 1211 0.00137 24.01 0.18114
3 4.0033 1218 0.00135 4.41 0.07707
4 3.74772 1245 0.08543 846.81 8.50535
Table A.2 (Continued )
No. of
furnace
CEL T
L
S
yy
S
xx
S
xy
5 3.666 1250 0.13987 1162.81 12.7533
6 3.81427 1236 0.05095 404.01 4.53708
7 3.93548 1230 0.01092 198.81 1.47373
8 3.86975 1234 0.02899 327.61 3.08152
9 4.26604 1187 0.05109 835.21 6.53247
10 4.10191 1209 0.00383 47.61 0.42715
11 4.20548 1198 0.02738 320.41 2.96209
12 4.13832 1209 0.00967 47.61 0.6784
13 3.89888 1229 0.01992 171.61 1.84872
14 4.09604 1208 0.00314 62.41 0.4427
15 4.22486 1199 0.03417 285.61 3.12419
16 4.12911 1209 0.00794 47.61 0.61487
17 4.05463 1218 2.14E-04 4.41 0.03071
18 3.88001 1232 0.0256 259.21 2.5759
19 4.13832 1208 0.00967 62.41 0.77672
20 4.02874 1215 1.27E-04 0.81 0.01013
21 4.06775 1208 7.70E-04 62.41 0.21923
22 4.1203 1207 0.00645 79.21 0.71467
23 4.19665 1203 0.02454 166.41 2.02079
24 4.0121 1223 7.78E-04 50.41 0.19808
25 4.15858 1208 0.01406 62.41 0.93675
26 4.19684 1195 0.0246 436.81 3.27787
27 4.16341 1206 0.01523 98.01 1.22172
Average 4.040217 1215.889 0.022936 236.9137 2.28799
Summation 113.1261 34044.89 0.619259 6396.67 61.7759
References
[1] D.W. Zeng, C.S. Xie, K.C. Yung, Mater. Sci. Eng. A 333 (2002)
223231.
[2] I. Karsay, R.D. Schelleng, AFS Trans. 69 (1961) 725730.
[3] S. Zeji, T. Ji, S. Guiqiao, Foundry Technol. 24 (2) (2003) 9193.
[4] P. Zhu, R.W. Smith, AFS Trans. 103 (1995) 601609.
[5] K.G. Upadhya, D.M. Stefanesuc, K. Lieu, D.P. Yeager, AFS Trans.
97 (1989) 6166.
[6] S.L. Backerud, G.K. Sigworth, AFS Trans. 97 (1989) 459464.
[7] I.-G. Chen, D.M. Stefanescu, AFS Trans. 92 (1984) 947964.
[8] E. Fras, W. Kapturkiewicz, A. Burbielko, H.F. Lopez, AFS Trans.
101 (1993) 505511.
[9] R.D. Schelleng, AFS Trans. 68 (1960) 301303.
[10] K. Chen, Applied Probability and Statistics, Tsinghua Universitys
Publishing House, 2000, pp. 360362.