Design of Hot Rolling Mill Plate Quality Control Model Based on WNN

WNN(Wavelet Neural Network) is a neural networks with activation function based on wavelet. Wavelet networks are usually limited to the low-dimensional input modeling by a single network. When the input space consists of several different classes of input data, it becomes very difficult to converge the network during the training phase. In this paper, a combined wavelet-based neural network modeling was introduced to resolve this problem. Based on a gating network, a combined network can divide a complex task into subtasks, and modeling each subtask with an single WNN. The performance of such networks in modeling hot plate mill production quality is examined and compared with that of single neural network.