TSINGHUA SCIENCE AND TECHNOLOGY ISSN 1007-0214 05/20 pp142-146 Volume 13, Number 2, April 2008

Melt Quality Evaluation of Ductile Iron by Pattern Recognition of Thermal Analysis Cooling Curves
LI Zhenhua (), LI Yanxiang (),**, ZHOU Rong ( )
Kunming University of Science and Technology, Kunming 650093, China; Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China Abstract: The melt quality of ductile iron can be related to the melts thermal analysis cooling curve. The freezing zone of the thermal analysis cooling curve was found to indicate the melt quality of the ductile iron. A comprehensive difference parameter, , of the thermal analysis cooling curves was found to be related to the properties of ductile iron melts such as composition, temperature, and graphite morphology. As approached 0, the thermal analysis cooling curves were found to come together with all the properties indicating melt quality about the same. A database of thermal analysis cooling curves related to the properties of the ductile iron melts was set up as a basis for a method to accurately evaluate the melt quality of ductile iron by pattern recognition of thermal analysis cooling curves. The quality of a ductile iron melt can then be immediately determined by comparing its thermal analysis cooling curve freezing zone shape to those in the database. Key words: melt quality; ductile iron; evaluation; thermal analysis; cooling curve

Introduction
Ductile iron is one of the most important engineering materials. The melt quality of ductile iron is determined by the melting and later treatment processes which determine the microstructure and mechanical properties of ductile iron castings. The melt quality includes indexes such as the chemical composition, temperature, efficiency of inoculation and nodularization, and the tendency of chill and shrinkage. The melt quality should be evaluated before pouring. Thermal analysis technique for evaluating the melt iron quality came into being in the early 1960s. Re searchers at the British Cast Iron Research Assocation (BCIRA) first used a thermal analysis method to measure the carbon equivalent (CE) of the iron melt[1].
Received: 2006-10-19; revised: 2007-10-15

To whom correspondence should be addressed.

E-mail: yanxiang@tsinghua.edu.cn; Tel: 86-10-62773640

About 10 years later, Ford developed a special CE cooling curve computer[2]. Later, special instruments for quick measurements of C%, Si%, and CE of iron melts were applied on foundry floors. These techniques are still widely used in foundries. Many researchers have tried to establish relationships between the characteristic temperatures on thermal analysis cooling curves and indexes of the melt iron quality using regression analyses or artificial neural networks to predict melt iron quality[3-8]. However, these relationships have been obtained for specific experimental or production conditions so they are less sensitive to casting conditions and not easily adapted to production conditions in different foundries. This paper describes a method to accurately evaluate the melt quality of casting alloys using pattern recognition analyses of cooling curves. This method uses the freezing zone of the cooling curve to evaluate the melt quality of the cast alloy rather than characteristic temperatures. The method has been applied to evaluate the

LI Zhenhua () et alMelt Quality Evaluation of Ductile Iron by

143

quality of iron melts[9-12] and Al-7Si alloy melts[13]. The principles for evaluating ductile iron melts are discussed in detail in this paper.

S= [(TiT)2/(n1)]1/2

(3)

T= (Ti)/n (4) where Ti and Ti are the temperatures along two cooling curves at the same time from the liquidus temperature to the end of eutectic solidification, and n is the number of comparison points on the two cooling curves. has the unit of Celsius degree (). In Eq. (1), the first term of Ti/n indicates the distance between two cooling curves, while the second term S indicates the variation of this difference. Therefore, the two thermal curves will be similar when is very small. The minimum will identify the cooling curve in the database that is the most close to the current melt with similar properties. (5) After the database of thermal analysis cooling curves and melt quality indexes of melts has been set up, the melt quality can be assessed by comparing the thermal analysis cooling curves freezing zones. The quality of an unknown melt can be determined by comparing to a melt in the database whose thermal analysis cooling curve freezing zone shape is very close to that of the unknown melt. In previous research, the method has been preliminarily applied to evaluate the quality of ductile iron melts[9-12]. The results have shown a relationship between and the features of ductile iron melts such as the graphite morphology of solidification samples. There are still many other indexes that indicate the melt quality of ductile iron such as the composition and melt temperature, which need further investigation.

Intelligent Evaluation of Melt Quality

Although some researchers have recognized that thermal analysis cooling curves can be used as an alloys fingerprint, traditional thermal analyses have just used some characteristic temperatures to evaluate the melt quality rather than all of the information in the cooling curves. Then they have tried to use more characteristic temperatures and artificial intelligence methods to improve the precision of the traditional thermal analysis method[7,8]. Li and Hu[9] then proposed a method to evaluate the melt quality of cast alloys based on the following principles[10-13]: (1) The shape of a cooling curve measured by a thermocouple mounted in a thermal analysis sample cup reflects the solidification process of the melted cast alloy for the given solidification conditions. All the factors influencing the solidification process, such as the chemical composition and trace elements, the inoculation and nodularization treatment efficiency, and the efficiency of modification, will influence the cooling curve shape. The freezing zone of the cooling curve, which is the segment of the thermal analysis curve from the liquidus temperature to the end of the eutectic solidification, accurately represents the entire thermal analysis cooling curve. (2) The freezing zone of a thermal analysis cooling curve can serve as a fingerprint of the melt quality. Tiny differences in the melt quality will give rise to changes of the freezing zone shape, so similar freezing zones will indicate similar melt quality. (3) The melt quality of cast alloys can be accurately evaluated by measuring the shape of the freezing zone by comparing to the freezing zone shapes of known melts for the specialized solidification conditions. (4) The difference between thermal analysis cooling curves can be expressed by the parameter which is defined as

Experiments

T
n

+S

(1) (2)

Ti = Ti Ti

Thermal analysis cooling curves of over 60 ductile iron melts were recorded by therocouples mounted in thermal analysis samples 30 mm in diameter and 50 mm in height. The chemical composition of the melts varied from 3.50%-3.86% C, 1.74%-3.56% Si, 0.27%-0.33% Mn, 0.03%-0.05% P, 0.006%-0.06% S, and 0.018%0.155% Mg, with various amounts of nodularizers and inoculants. Samples 80 mm in diameter and 200 mm in height were also solidified for 6 h from these melts in a high precision thermal simulation system for analyses[14]. The compositions of the melts were analyzed by a spectrometer and the graphite morphologies of all the samples were analyzed by using a quantitative metallographic analysis method.

144

Tsinghua Science and Technology, April 2008, 13(2): 142-146

Results and Discussion

Figure 1 shows the relationship between and the chemical compositions of the ductile iron melts for various carbon content (CC), silicon content (CSi), magnesium content (CMg), and rare earth content (CRe). approaches zero as CC, CSi, CMg, and CRe all approach zero. The relationships between and the contents of other elements in the melts were the same. Thus, when two ductile iron melts have similar thermal analysis cooling curves ( is small enough), they must have similar chemical compositions.

difference of the maximum temperature in the thermal analysis cooling curve (Tmax). approaches zero as Tmax approaches zero. Thus, when two ductile iron melts have similar thermal analysis cooling curves, they must have similar temperatures.

Fig. 2 Relationship between and Tmax of ductile iron melts

Figures 3 and 4 show the relationship between of the thermal analysis cooling curves and the graphite morphologies of the thermal analysis samples and samples solidified for 6 h as indicated by the nodularity difference (Q) and the graphite area ratio difference (A). As Q and A approach zero, approaches zero. Thus, when two ductile iron melts have similar thermal analysis cooling curves, they must have similar qualities and similar solidified microstructures not only in thermal analysis samples solidified for several minutes, but also in samples solidified for 6 h.

Fig. 1 Relationship between and the chemical composition differences of ductile iron melts

Fig. 3 Relationship between and graphite morphology differences of thermal analysis samples

Figure 2 shows the relationship between of the thermal analysis cooling curves and the temperature difference of the ductile iron melts, represented by the

These results are further confirmed by the comparison of the freezing zones of the thermal analysis cooling curves of two ductile iron melts shown in Fig. 5.

LI Zhenhua () et alMelt Quality Evaluation of Ductile Iron by

145

enough, the melt quality of the two ductile irons will be very similar and the graphite morphology of samples solidified for short or long times will be very similar. Therefore, the melt quality and the graphite morphology of ductile irons in the database can be used to evaluate the melt quality and predict the graphite morphology of new melts based on for their cooling curves. The database of thermal analysis cooling curves and the ductile iron features related to the melt quality will be expanded in the future to improve the pattern recognition results for the thermal analysis cooling curves.

Fig. 4 Relationship between and graphite morphology differences of samples solidified for 6 h

Fig. 5 Freezing zones of two thermal analysis cooling curves with = 1.7

for these two curves is 1.7, which indicates that the curves are very similar. The compositions of the melts are listed in Table 1 with the graphite morphologies of the thermal analysis samples and samples solidified for 6 h shown in Figs. 6 and 7 all very similar. Thus, if of two thermal analysis curves is small

Fig. 6 Graphite morphology of thermal analysis samples used in Fig. 5

Table 1 Maximum temperatures and compositions of melts used in Fig. 5 Sample 1 2 Maximum temperature () C 1227.3 1210.8 3.65 3.67 Si 2.16 2.01 Mn 0.271 0.298 Composition (wt. %) P 0.0366 0.0364 S 0.0270 0.0268 Mg 0.0414 0.0587 Re 0.0154 0.0240

Conclusions

(1) The freezing zone of a thermal analysis cooling curve indicates the melt quality of ductile iron. The difference parameter that characterizes the cooling curves correlates well with the melt quality,

chemical composition, temperatures, and graphite morphologies of samples solidified for short or long times. (2) As approaches zero, the differences of all the indexes indicating the melt quality of ductile iron also approach zero. When of two thermal analysis curves

146

Tsinghua Science and Technology, April 2008, 13(2): 142-146 [3] Stefanescu D M, Loper C R Jr, Voigt R C. Cooling curve structure analysis of compacted/vermicular graphite cast irons produced by different melt treatments. AFS Transactions, 1982, 90: 333-348. [4] Frost J M, Stefanescu D M. Melt quality assessment of SG iron through computer-aided cooling curve analysis. AFS Transactions, 1992, 100: 189-199. [5] Zhu P, Smith R W. The prediction of the microstructure of cast iron using thermal analysis. Materials Science Forum, 1996, 215-216: 503-510. [6] Labrecque C, Gagne M. Interpretation of cooling curves of cast irons: A literature review. AFS Transactions, 1998, 106: 83-90. [7] [8] Sillen R V. Optimizing iron quality through artificial intelligence. Modern Casting, 1996, (11): 43-45. Calcaterra S, Campana G, Tomesani L. Prediction of mechanical properties in spheroidal cast iron by neural networks. Journal of Materials Processing Technology, 2000, 104: 74-80. [9] Li Yanxiang, Hu Xiao. Artificial intellectual recognition of cooling curves and new-type comprehensive evaluation of melt quality of casting alloy. Foundry, 1999, 48(7): 5-7. (in Chinese) [10] Li Yanxiang, Hu Xiao, Xu Xuerong. Pattern recognition of thermal analysis cooling curves and quality evaluation of melt cast alloys. Journal of Materials Science & Technology, 2001, 17(1): 73-74. [11] Li Yanxiang, Xu Xuerong, Wang Qiang. Evaluating melt iron features by computer aided recognition of thermal analysis cooling curves. International Journal of Cast Metals Research, 2003, 16(1-3): 41-45. [12] Li Zhenhua, Li Yanxiang. Evaluation of melt quality and graphite degeneration prediction in heavy section ductile iron. Metallurgical and Materials Transactions A, 2005, 36A(9): 2455-2460. [13] Wang Qiang, Li Yanxiang, Li Xiaochun. Grain refinement of Al-7Si alloys and the efficiency assessment by recognition of cooling curves. Metallurgical and Materials Transactions A, 2003, 34A(5): 1175-1182. [14] Li Zhenhua, LI Yanxiang, Geng Huiyuan, et al. Study on thermal simulation of solidification in heavy section ductile iron. Journal of Materials Science & Technology, 2003, 19(Suppl. l): 122-124.

Fig. 7 Graphite morphology of samples used in Fig. 5 solidified for 6 h

is small enough, the melt quality of the ductile iron samples is very similar. (3) The melt quality of ductile iron can be accurately predicted by comparison with a database of thermal analysis cooling curves and the features related to the melt quality of ductile iron melts. The properties are most accurately predicted by comparing to the freezing zone shapes of the thermal analysis cooling curves. Acknowledgements
The authors are grateful for the financial support and the help of technicians of the Ningbo Riyue Foundry Co., Ltd. Special thanks go to Mr. Geng Huiyuan, Mr. Chen Jianjun, and Mr. Zhou Chunyan for their help in the experimental work.

References
[1] Humphreys J G. Effect of composition on the liquidus and eutectic temperatures and on the eutectic point of cast irons. BCIRA Journal, 1961, 9(5): 609-621. [2] Warsinski R G. Ford develops CE cooling curve computer. Foundry Management and Trade, 1975, (3): 104-108.