| Predicting mechanical property of steel material accurately according to researches on processing parameters and chemical composition is always main objective for steel maker. It is difficult to predict the performance of materials accurately according to a pure quantitative calculation of various metallurgical phenomena, because steel material has many components and complex processing parameters. Designing the overall structure of the system according to the mechanical properties of steel demand forecasting system analysis. The research contents are introduced as follows:(1) Preliminary data preprocessing:In view of large amount of technological parameters and field data in rolling process and also poor data quality, based on the data pre-preprocessing, this paper adopt algorithm combined PCA with subtract clustering and rough set data reduction model. Field test data show that:Based on system-wide coherence of the rough set attribute reduction methods can effectively improve the mechanical properties of the prediction accuracy.(2) The systems are compared the network prediction performance about genetic algorithms and particle swarm optimization BP network weight value and threshold value. After all, BP network is chosen to simulate on the experimental data.(3) Forecasting System Data Management:Relying on Steel production line of mou sheet plant, we established the database and management system of Hot-rolling microstructure and property prediction system. And realize the corresponding data batch import and export、real-time data storage of prediction system and data input, delete, query functions of the production process data. Dynamic display the training error curve and real-time predict the steel mechanical properties through the main console interface.Compared with actual production data, this software has a higher precision. The accuracy of the system:Tensile strength from10.3to7.8, the yield strength from11.2to7.5and the specific elongation from8.5to5.2. |
PDF Full Text Request