Development of materials database system for cae system of heat treatment based on data mining technology

GU Qiang¹, ZHONG Rui¹, JU Dong-ying²

1. Graduate School of Engineering, Saitama Institute of Technology, Fusaiji 1690, Fukaya, Saitama 369-0293, Japan;

2. Computer Science Deptartment, Saitama Institute of Technology, Fusaiji 1690, Fukaya, Saitama 369-0293, Japan

Received 10 April 2006; accepted 25 April 2006

Abstract: Computer simulation for materials processing needs a huge database containing a great deal of various physical properties of materials. In order to employ the accumulated large data on materials heat treatment in the past years, it is significant to develop an intelligent database system. Based on the data mining technology for data analysis, an intelligent database web tool system of computer simulation for heat treatment process named as IndBASEweb-HT was built up. The architecture and the arithmetic of this system as well as its application were introduced.

Key words: materials database; CAE system; heat treatment; data mining

1 Introduction

Numerical simulation on materials processing, such as heat treatment and casting, is an important method to evaluate complex physical behavior with coupled fields of temperature, microstructure and stress/strain. However, this simulation needs a lot of data containing the parameters of the mechanical and thermal properties as well as the diagram related to phase transformation. On the other hand, parameters of inelastic constitute equation and kinetics model of phase transformation can not be directly obtained from experiment. Therefore, it is significant to develop an intelligent system based on data mining.

Based on the above strategies, the following technique was employed to construct the frame of the database. The development tools including. NET Framework, ASP.NET, VB.NET, C＃, CGI, XML techniques, are used to develop the intelligent database system which posseses the following functions: data search, data intelligence processing,  data statistics, data mining, data conversion and data export.

The original data of material properties obtained from the experimental results and technical references are classified and searched by steel grade, technical test and materials properties. Functional modulus with regression and clustering algorisms has been built to generate the function curve and output the using equations. The accumulated data were analyzed by the data mining technique integrated in this system to determine the potential functional elastic coefficient and the strain/stress relationship for heat treatment simulation. This system provides an effective and efficient tool for the materials simulation researchers.

In this paper, the research strategies for supporting computer simulation of heat treatment process　are to construct the database based on the following points:

1) Web-based system for easy remote access and administration;

2) Knowledge-based database to analyze the functional relationship and to discover the potential rule of the data by its intelligence;

3) Interfaces to allow experts to inspect and modify the algorism of the core modular.

According to the above strategies, the following techniques were employed to construct the frame of the database:

a) Browse/Server mode to work on networks and transform platform;

b) Component design for extensibility;

c) Security and authorization;

d) Date mining technique;

e) Data intelligence processing.

Based on these requirements, we developed the intelligent database system, which is shown in Fig.1. This system includes the following kinds of functions: data search, data intelligence processing, data statistics, data mining, data conversion and data export. The function map is shown in Fig.2.

Fig.1 Architecture of this system

Fig.2 Function map of modulus

2 Metallo-thermo-mechanics

Fig.3 shows a three-way coupling of temperature, stress and metallic structure which is often encountered so that the task of simulating the process accurately becomes quite complex, The theory was proposed by INOUE T(1981). It is a metallo-thermo-mechanics theory for considering the coupling effects.

Fig.3 Metallo-thermo-mechanical coupled analysis and effect of chemical composition.

① Thermal stress—The thermal expansion caused by such a temperature gradient is restricted by the shape of a soild body, thus generating thermal stress.

② Heat generation due to deformation—When stress/strain is applied to a soild, the energy is partially discharged as heat in the case of inelastic deformation.

③Temperature-dependent phase transformation start time. However, in the case of diffusion-type transformations of ferrite, pearlite, and bauxite, the temperature history also affects the phase transformation.

④ Latent heat due to phase transformation— Latent heat generated in the course of phase transformation affects the temperature field.

⑤ Stress (or strain)-induced transformation—The phase transformation behavior is also affected by stress/strain existing in the solid. For example, pearlite transformation time is shortened under tensile stress, and vice versa. Martensite transformation is generated even though a material is processed at temperature higher than the martensite transformation temperature under the applied stress or strain.

⑥Transformation stress and transformation plasticity—Volume dilatation in the work is caused by the phase transformations. When this volumetric dilatation is inhomogeneous depending on the complicated shape of the body, stress and strain are induced, defined as transformation stress and strain. The level of such induced stress is comparable to the thermal stress. The effect of transformation plasticity is sometimes important.

3 Build of scientific database

3.1 Data source and classification

Table1 shows that the original data are obtained in various ways such as from experimental results and technical references. The data are classified by steel grade concerning alloy and carbon content; by technical test and by materials property. The expansibility is enhanced by the adoption of component design. The system can get external data and export data sheet by ODBC. We design and implement the XML-style presentation of model and integrate the database system with data mining technique.

3.2 Data search

Three types of engines are built by steel name, chemical compositions and examining test. User can select the combined search mode and output statistical information about the data. As an example, the searched page of steel SCr420 is shown in Fig.4.  And the detailed result enclosed the original data and corresponding graphs is shown in Fig.5.

Fig.4 Result of search function

3.3 Data analysis

This system can generate the function curve and output the equation. Functional modulus with regression and clustering algorisms has been built. The interfaces allow experts to select and inspect the algorism and scope for data mining.

Fig.5 Detail of database

3.4 Data conversion  

This system can converse the mining results into special data format for commercial simulation software; user can also choose the conversion mode.

3.5 Data intelligence processing  

This system provides process data in an intelligent up-loading mode. The user preserves the data obtained from the experiment in the data base. According to the data of material preserved, the system can automatically arrange and distinguish the classification. After examining, the manager opens it to the public. The sample is shown in Fig.6. Moreover, when the user read and requested original data by the data chart, we thought it was very inconvenient. So the system offers software that reads data from a special data chart, and has improved the convenience in using extremely.

4 Data mining system

Data mining, sometimes called knowledge discovery or data exploration, is an analytical tool of

Table 1 Data source and classification

finding correlations or patterns among dozens of fields in large relational database. It allows users to analyze data from different aspects, categorize, summarize and identify the unknown relationship. In this system (the flow of data mining system is shown in Fig.7), the following data mining technique were employed.

Fig.6 Data uploading and intelligence processing

Fig.7 Flow of data mining system processing

a) Sequence model. Use regression method to find the potential rule in the data.

b) Classification. Classify the data to suitable data scope.

By using the data mining technology, for a metallic material in the heat-treatment process, this system can determine it about the parameter of the phase transformation’s physical properties value, the heat physical properties, the mechanics physical properties, and the mechanical properties, also can retrieve the element and the standard of the material, and display the characteristic chart of each material. By the high-speed Internet, it is united with the heat-treatment simulation system intimately, and integration is achieved. The relation between them is shown in Fig.8.

5 Application of database

As an example, the accumulated data were analyzed by the data mining technique integrated in this system to determine the potential functional elastic coefficient and the strain/stress relationship for heat treatment simulation. The measured materials properties by tensile test are recorded as numerical value in series of information and data sheets. The information sheet describes the detailed testing conditions and specimen. The data sheet concerns the original experimental results shown in Table 1 as part of active webpage. The original data are first analyzed by clustering and classification tools to release the noise and determine an optimized scope. The regression is used to get the functional relationship of the elastic coefficient with temperature by multinomial regression as shown as Fig.9. In summary, this system provides an effective and

Fig.9 Results of data mining generated

efficient database tool for the materials simulation researchers

6 Conclusions

This research has advanced the development of materials database system for CAE system of heat treatment based on data mining technology. This system is assumed to be an efficient support system for the heat-treatment simulation. The user, as a user of heat-treatment CAE, analyzes the retrieved characteristic material data from the database, and the data mining system that was able to use it between intranet and internet was developed. In the retrieval of a general material database, the retrieval result only displays the saved data as table and graph. But this system, by the heat-treatment theory, is able to analyze the retrieved data by the database. Moreover, it is possible to support it to heat-treatment CAE system.

References

[1] INOUE T, OKAMURA K, JU D Y. Material database for simulation of metallo-thermo-mechanical field [A].Proc 20th ASM Heat Treating Conf. on Quenching and Distortion Control [C]. ASM Inter, 2002: 753-760.

[2] TOTTEN G E, GERGELY M, SZILVIA S. Software tool for heat treaters and material engineers: EQUIST2000, the database of standard steels [A]. Proc 4th Inter Conf Quenching and Control of Distortion [C]. 2003: 193-200.

[3] CHEN N, ZHU D. Intelligent materials processing by hyperspace data mining, engineering applications of artificial intelligence [J]. 2000, 13: 527-532.

[4] INOUE T. Inelastic Constitutive Relationships and Applications to Some Thermo-mechanical Processes involving Phase Transformation, Thermal Stresses Ⅲ, North-Holland, 1988.

[5] ISOGAWA S, MORI I, TOZAWA Y. Determination of basic data for numerical simulation–analysis of multi-stage warm forging sequence for austenitic stainless steel I [J]. Journal of the JSTP, 1997, 38(436): 84-89.(in Japanese)

[6] TEKKAYA A E. Current state and future developments in the simulation of forming processes [A]. Proc 30th Plenary Meeting of the International Cold Forging Croup ICFG [C]. 1997: 1-10.

[7] NAGATA S, YANAGIMOTO J. Fundamental in Metal Forming [M]. Coruna Publishing Co, Ltd, 1997.(in Japanese)

[8] ANGEL T. Formation of martensite in austenitic stainless steels–effects of deformation, temperature, and composition [J]. J Iron Steel Inst, 1954, 177: 165-174.

[9] LECROISEY F, PINEAU A. Martensitic transformations induced by plastic deformation in the Fe-Ni-Cr-C system [J]. Metallurgical Transactions, 1972, 3: 387-396.

(Edited by PENG Chao-qun)

Corresponding author: D.Y.Ju; Tel: +81-48-595-6826; Fax: +81-48-585-5928; E-mail: dyju@sit.ac.jp