Breakout forecasting system based on multiple neural networks for continuous casting in steel production

In the continuous casting process, breakouts are a serious problem that can cause an extended shutdown. A neural network has been applied to breakout forecasting. A breakout is preceded by an abnormal and subtle change in temperature at the wall of the mold. Analysis of temperature patterns showed that multiple, time and spatial neural networks could recognize the characteristics of breakouts. An off-line simulation showed that the forecasting performance of this system was far better than that of the conventional system. An on-line test conducted at a steelmaking plant of Nippon Steel Corporation showed that the system could forecast breakouts with an accuracy of nearly 100 percent