| Breakout is the most catastrophic incident in the continuous casting process, especially the slab sticker breakout. To avoid or decrease breakout, there are usually two ways:the first one is to research the form theory on sticker breakout deeply, so as to the producing conditions on sticker of elimination; the second one is to develop slab continuous casting breakout prediction system, which is to detect the breakout symptom in the mold, then take certain measures such as reducing the casting speed.First, according to the interpretation on the sticker form theory this paper puts forward the mathematical model of steel shell growth when leakage occurs, and thus obtains the method that through to the thermocouple temperature curve pattern recognition to detect the breakout of the sign. Based on the theory of the steel shell growth, this paper optimized embedding spacing of thermocouples.Second, in order to distinguish the curves that had the characteristics of leakage bonding we established the prediction models based on logic judgment and the artificial neural network respectively and simulate them with the offline test data from The Baosteel. The mathematical model which based on the artificial neural network is optimized by LM Algorithm. The simulation result shows that it is better than the traditional gradient BP algorithm in training time and prediction accuracy. This paper also made a comparative analysis on the two models, which laid a theoretical foundation for the prediction system’s further research and improvement.Finally the system is developed with MATLAB and VC on mixing programming for the system design. The proposed algorithm first optimize in MATLAB, and then combine the VC with the interface package. This system contain the functions both of logic forecast and the neural network forecast. It can reflect the thermocouple temperature changes in the visual display and the position of dangerous thermocouple. Achieve good interaction between human and machine. |
PDF Full Text Request