An Improved Elman Network and Its Application in Flatness Prediction Modeling

An improved Elman network, in which the self-gained vectors are added in the context units, is developed and the corresponding network structure and learning algorithm are presented. In the self-gained Elman network, the constant gain factor is replaced with the gain vector, so the power of the feedback units is strengthened. Therefore, the Elman network is provided with better approximating performance and dynamic characteristics. The model of flatness prediction for strip steel cold mill based on the improved Elman network is established. The simulation results show that it is a fast and precise model of flatness prediction.