Simulation And Prediction Of Microstructure And Properties Of Hot-rolled Steels Based On Physical Metallurgy And BP Neural Network

Nowadays, the steel products are facing more and more competition from other kind of materials. In order to improve the quality of steel products and reduce production cost, the Structure Properties Prediction and Control Technology (SPPC) in hot rolled steels have been become the research hotspot recently. The development of computer technology and automation has accelerated the application of SPPC in the steel plants. Based on the mathematical simulation of microstructure evolution in the hot rolling process and prediction of mechanical properties, the production costs could be reduced and the efficiency could be improved. Further, the exploitation of new products and new process could be speed up.The physical metallurgical models and BP neural network model, which describe the microstructure evolution and relationship between microstructure and mechanical properties in 1422mm hot rolling production line of Meishan steel Company, were established in the present work and the experiments of TMCP were carried out in laboratory. The contents include the following main parts:(1) After the thorough comprehension of data report in Level 2, the coil number of steel and the corresponding hot rolling processing parameters was extracted by programming and connected with chemical composition and mechanical properties by technology of database. It provides sufficient data for the establishment of physical metallurgical models and BP neural network model.(2) The physical metallurgical models describing the microstructure evolution which includes dynamic and static recrystalization, flow stress, dislocation density and phase transformation in hot rolling process was establish based on the work done by predecessors. Results of computer simulation have shown that good agreements have been achieved between measured and predicted values.(3) According to the knowledge of physical metallurgy and survey of production line, chemical composition and processing parameters with high correlation with mechanical properties are selected as the inputs of neural network. The prediction models of yield strength, tensile strength and elongation rate are built. The optimal hidden unites is determined by empirical equation and large quantities of experiments. High prediction precision is achieved.(4) For Q235 and X52, the experiments of TMCP were carried out in laboratory. The results were obtained as follows:the important parameters which determine the mechanical properties are optimized of combination of distribution of loads, finishing temperature and coiling temperature. Improving reduction rate behind the refined zone and decreasing finishing temperature and coiling temperature can improve the strength of products and keep the higher elongation percentage. Consequently supply the guidance for the composition design of new steel grade and the development of TMCP production technology.