Predicting mechanical properties of hot-rolling steel by using RBF network method based on complex network theory

Recently, producing high-precision and high-quality steel products becomes the major aim of the large-scale iron and steel enterprises. Because of the internal multiplex components of products and complex changes in the production process, it is too difficult to achieve precise control in hot rolling production process. In this paper, radial basis function neural network is used to complete performance prediction. It has the advantage of fast training and high accuracy, and overcomes shortcomings of BP neural network used previously, such as local minimum. When determining the center of radial basis function we make use of complex network visualization which can clearly figure out the relationship between input vectors and receive the center and width according to the relationship of the nodes. Experiments show that the method that is combining community discovery algorithm and RBF enjoy high stability, small training time which means to be suitable to analysis large-scale data. More importantly, it can reach high accuracy.