| Temperature control in the float glass production line is always the most important part. From the moment of the batch mixed by a certain percentage of the raw materials into the furnace, it will has a process of melting, clarification, forming, annealing and finally being turned into the float glass with the high optical quality, high level, non pollution. And each step will be adopted to track the temperature feedback control. In recent years, with the fierce competition of the float glass market and the demand for increasing the quality of products year by year, it is the pressing need to improve the quality of float glass and yield for more precise temperature control. While the traditional manually adjusting the temperature control mode have been unable to keep up the pace of the times, improving the automation of float glass production line is a feasible solution.Based on Shanghai Yaohua Pilkington Glass Co., Ltd. float glass production line, the paper focuses on an advanced hot-side temperature control system. The hardware part is composed of PLC, HMI, and communication. And the core part of the automation control, the components and the configuration are introduced specially. Besides, it includes network communication Profibus, industry Ethernet. The software of the system analyzes the Simens PLC software and the iFIX of PC configuration software in detail. Siemens STEP 7 is also specially introduced to combine field equipment to realize field control and remote control.After the complete update of the float glass production line hot end temperature automatic control system, it is to forecast the surface temperature of the liquid melted glass for using the soft sensor, which can't be measured when the production is running. It studies the forecasting methods based on Artificial Neutral Network. There are two forecasting methods, one based on BP ANN and the other based on RBF ANN, applying which to forecast the surface temperature of the liquid melted glass are discussed. And finally, we get the satisfactory results for the soft sensor model after analyzing and comparing the result from two methods. |