Evaluation technology and application of nonequili

  • Detail

Evaluation technology and application of non-equilibrium state of casting alloy melt

in the equilibrium state, the characteristics of liquid substances only depend on the two parameters of composition and temperature (assuming constant pressure). However, the liquid metals encountered in actual production and scientific experiments are often far from equilibrium. Therefore, it often occurs that the microstructure and properties of liquid metals with the same composition and temperature are very different under the same solidification conditions. This raises the question of characterization and evaluation of liquid metal quality under non-equilibrium conditions. If we define the quality of liquid metal as a characteristic that affects the solidification process, solidification structure and properties, it must be affected by the following factors:

· raw materials used for smelting, that is, the so-called heritability

· trace elements refer to trace impurity elements that are not generally analyzed or are difficult to be accurately analyzed

· smelting process, including the maximum temperature and high temperature residence time experienced in the smelting process

· smelting method, blast furnace, resistance furnace, induction furnace, electric arc furnace, etc

· furnace pretreatment, refining, modification, spheroidization, inoculation, electric and magnetic treatment, etc

the common feature of all these influencing factors is that they are dynamic unbalance parameters. Therefore, the mass characteristics of liquid metals cannot be expressed by one or several equilibrium parameters

for decades, people have adopted a series of methods whose degradation rate is affected by many factors to evaluate the quality of liquid metals, such as metallographic examination under the same solidification conditions, measurement of electronic transport characteristics of liquid metals, thermal analysis, etc. Among them, thermal analysis method has a series of advantages, the most research and the fastest development

with the rapid development of electronic technology, the thermal analysis method has gradually realized electronization, digitization and computerization. The advent of electronic thermal analyzers and computerized thermal analyzers has greatly promoted the development of thermal analysis. Abroad, computer and thermal analyzer are combined to develop a computer-aided thermal analysis technology (computer3. graphene computer-aided thermal analysis), referred to as cata, or computer-aided cooling curve analysis (ca-cca). Using this technique, one can evaluate the inoculation effect of molten iron. The interface circuit of microcomputer thermal analysis design connects the thermal analyzer with the microcomputer. Through the microcomputer, the mathematical model and its coefficients can be easily modified according to the on-site production conditions and test conditions, thus greatly improving the adaptability and test effect of the thermal analyzer

In 1996, a system based on thermal analysis and artificial intelligence was developed for grey cast iron and nodular cast iron. This system can analyze the solidification process of the sample and predict the possibility of various casting defects according to the stability system, and can also estimate the physical properties. In this system, the computer-aided thermal analysis technology is used to compile a software program to evaluate the microstructure and inoculation effect of molten iron. The researchers selected 10 characteristic values from the thermal analysis curve as control parameters, and defined thresholds for each parameter. If all 10 points meet the threshold requirements, the molten iron is considered qualified. The method is simple, convenient and clear at a glance

however, even with the development of thermal analysis, the number of selected eigenvalues is from less to more, and the evaluation of molten iron quality by selecting eigenvalues is still affected by subjective factors to a certain extent. This is because the application of thermal analysis technology is limited to more clearly and intuitively displaying the eigenvalues on the cooling curve, and establishing a certain regression relationship between these eigenvalues and the prediction parameters of molten iron. Smithers Pira, a British market research company, pointed out in the report "2020 global packaging future" recently released that the accuracy of finding eigenvalues of thermal analysis and the regression accuracy of mathematical models are limited. Even if the first-order differential and the second-order differential are added, considering the analysis of crystallization latent heat, the selection of eigenvalues is still based on existing experience and subjective factors. Moreover, the analysis of differential curve and cooling curve requires very professional knowledge, which increases the difficulty of analysis and the role of subjective factors

at present, artificial intelligence neural network has been used to predict the properties of gray iron castings in China. Artificial neural network is an artificial intelligence pattern recognition method established by simulating the method of biological nerve transmitting information. It has the advantages of parallel and strong adaptability. As the basic element of neural network, neuron is widely used because of its fast computing speed. Due to the good self-learning function of the neural network model, the system has strong adaptability with the continuous increase and update of patterns to samples. Therefore, a dynamic comprehensive database can be designed and established, in which a large number of pattern pairs are stored. With the help of the system, buildings will not collapse immediately, and new fact samples will be stored continuously with the operation of the system, so as to make the neural network self-learn as a new pattern pair, So as to continuously improve the adaptability of neural network model and the prediction hit rate. The pattern recognition method can be applied to the production process of gray cast iron affected by many factors; The self-organizing artificial neural network is used for pattern recognition and classification of the production process controlled by multiple factors. According to the spatial distribution structure of the representative points of the production state, the relationship between them and the control objectives is found. The indescribable functional relationship between input and output is transformed into the classification and discrimination of pattern recognition. A computer intelligent expert system can be established to identify the qualified and unqualified states of gray iron quality, So as to predict the brand of sample performance. Zhangzhaochun and others from Shanghai Metallurgical Research Institute used the least square method and combined with the S-test of the sum of squares of predicted residuals to extract several main factors (CE, Mn, Cr, Sn, SI) that affect the casting performance from several main aspects (such as the chemical composition of molten iron and pouring process parameters), and used the inverse mapping method to determine the change trend and optimization range of these variables on the basis of the existing production process. Taking these main variables which affect the properties of castings as the input characteristics of artificial neural network, the two indexes of castings can be predicted by using the known sample set training

recently, the author and his collaborators developed a new thermal analysis system for liquid metal quality assessment by using pattern recognition method. The basic principle of the system is that the results of self thermal analysis, composition analysis, metallographic analysis and mechanical properties form a database with a certain structure. As long as enough records of different types, brands, solidification conditions and quality indicators are accumulated in the database, the evaluation basis can be formed. When a new thermal analysis test result is input, the computer can quickly find out the closest one to the cooling curve in the existing database, and use the composition, metallographic structure and mechanical properties of the found curve to evaluate the current measured melt. This method uses pattern recognition technology instead of the original differential and multi cup simultaneous measurement methods to characterize the cooling curve; By using database instead of linear regression and artificial neural network analysis. This method can comprehensively evaluate various indexes related to melt quality, and has self-learning function for different conditions. The research results are currently being used in the research and development of hot metal quality evaluation and control technology for large section ductile iron and the research of key technologies for direct production of high-quality castings with short blast furnace hot metal process. It is playing a key role in the production of 100 ton extra large section ductile iron in China and the direct production of high-quality castings (gray cast iron engine block, vermicular cast iron, ductile iron pipe, etc.) in the short process of blast furnace molten iron. (end)

Copyright © 2011 JIN SHI