It is very long for Mechanical properties testing cycle of ShouGang competitive bar products plant which affect whole manufacture of alloy constructional steel product. So we should research Mechanical Properties Prediction System for Alloy Constructional Steel of ShouGang Corporation which replace original testing pattern with prediction system. If succeeded, it will be remarkable for process adjustment, organism of production and sales, where could increase testing cycle and reduce producing cost.This paper makes a comprehensive design of development and application of mechanical properties prediction system for alloy constructional steel product of ShouGang Corporation. We construct BP neural network model by analysis of primary aspect, which including confirmation of I/O variable, treatment of sample data and selection of the number of the Hidden Layer Nodes. We choose the LM calculation method by comparison, of which improve ability of network generalization. We developed non-linear approximation similarity calculation method by SPSS statistical analysis software for which accuracy of properties prediction of non-sample data Accuracy is unsatisfactory. If the above-mentioned methods are used together, it will improve accuracy of properties prediction evidently.In the base of the theory of modal, we develop the mechanical properties prediction system for alloy constructional steel product of ShouGang Corporation using Maltlab. All of prediction result is satisfactory by variance analysis. Now the prediction system has put into service for production testing. It will be engaged in other product in the future. |