漏鋼預報是冶金連鑄生產過程中的關鍵環節，需要非常可靠的安全性。提出的二值數據窗前置處理法，不但實現了數據的歸一化，而且利用數據窗技術只需截取發生漏鋼時溫度變化異常的8 個采樣點，這樣大大減少了網絡的輸入個數，提高了網絡運算速
度。實際應用表明，該系統可直接用于生產現場，具有預報速度快、精度高、魯棒性能好等優點，有著廣泛的應用前景。
關鍵詞：連鑄 漏鋼預報 二值數據窗
Abstract：Being one of key sections of continuous casting process，breakout prediction requires safety with excellent reliability. A binary data window preprocessing technique is proposed. It not only implements data normalization，but also employs the data window technique to acquire eight sampling points where an abnormal temperature change occurs when breakout takes place，thus reducing the number of network input and increasing network operation speed. Practical applications show that this system features fast prediction speed，high accuracy，excellent robustness when directly used in the production field，and has a wide range of applications in future.
Keywords：Continuous casting Breakout prediction Binary data window