Using wavelet neural network model product quality of continuous casting furnace and hot rolling mill

Abstract

The industrial process of continuous casting furnace and hot rolling mill is very complicated, and the number of parameters, which determine the final product quality, is more than 30, so there is no mathematical model available. This paper models the product quality using wavelet neural network based on large number of data acquired from production. In accordance with the industrial process consisting of several procedures, this paper proposes a new architecture of the high-dimension input wavelet neural network. In the presented neural network, some input variables are connected directly to the second hidden layer or more later hidden layers according them to make action being early or late in work procedures; the key input connected to not only all nodes in the subsequent layer but also the output nodes directly. To avoiding fall in local minimum, this paper present a new global optimization algorithm based on a filled function. Experimental results indicate that the developed methodology is efficient and has a high accuracy in application to establishing production quality model.