ART-2 neural network usage to determine moment of slag discharge during steel teeming process

Abstract

Using adaptive resonance theory (ART-2) neural network, a method is proposed to process a signal of power spectral density of surface acceleration of a steel ladle protective pipe manipulator in order to determine the moment preceding the slag discharge from the steel ladle. We compare the spectrum amplitude analysis, the power spectrum analysis, and the power spectral density analysis of the vibration signal from the protective pipe manipulator surface to choose the best approach to create training set for an ART-2 network. It is proved by modeling that the neural network is able to determine the teeming process state preceding the slag discharge from the steel ladle under real production conditions. The results of the research prove the effectiveness of using an ART-2 neural network based classifier to solve the considered problem.