Prediction Of Mechanical Properties Of Hot Rolled Strip Based On BP Neural Network Optimization

Nowadays,people are more and more strict with the quality of the iron and steel products,especially in mechanical properties,pursuiting the higher stability and precision.In the long-term production process,a lot of production-related information have been accumulated by many steel mills.with these data of production can provide a reliable reference for future production.Based on these data,a high quality prediction model was established based on the mechanical properties of the hot rolled products(yield strength,tensile strength and elongation,respectively),the chemical composition content and the rolling parameters in steel production.The study of the mechanical properties of hot-rolled strip has a great effect on the production of hot-rolled products,the improvement of production efficiency and the optimization of the process.BP neural network algorithm has been relatively mature,and its application is also quite extensive.In this paper,BP neural network is used to predict the mechanical properties of hot-rolled strip.However,there are some shortcomings of BP neural network,such as slow convergence and easy to fall into local minimum point.In order to solve these problems,LM(Levenbegr-Maruqart)algorithm is introduced to improve its convergence rate and Genetic algorithm is used to optimize its weight and threshold to achieve satisfactory prediction results.In this dissertation,the mechanical properties of hot-rolled strip(represented by tensile strength)will be analyzed with the background of the production technology model of double-stove steckel mill in a steel mill.Based on the established mechanical performance model,the dissertation first uses the BP algorithm to predict the research,and the prediction effect is generally good.However,the accuracy and convergence speed of the model need to be further optimized for the current high precision requirements of the steel producers.Then the paper introduces the algorithm(LM algorithm)and GA genetic algorithm to optimize the BP model.The introduction of these two algorithms,through the comparison of data we can clearly found:,the BP algorithm's convergence rate has been greatly improved as LM algorithm does not change the accuracy of the case;GA genetic algorithm not only improves the prediction accuracy of BP algorithm,but also improves its convergence.In summary,the BP algoritm is very effective in predicting the mechanical properties of the strip based on the double-rack steckel mill,and it is proved that the innovative application on the double-rack steckel mill is feasible.