Advanced modeling for small glass furnaces

ADVANCED MODELING FOR SMALL GLASS FURNACES Heath Morris One of the most pressing issues facing the glass industry is improving energy efficiency. The largest energy user in any glass company is the melting furnace or furnaces. While large float glass and container glass companies have developed sophisticated control systems, little work has been done until recently for small glass furnaces. This thesis extends the work of Holladay (2005), in which an observer was developed to estimate the temperature of glass in a small day-tank furnace. The current work eliminates the assumption of homogeneous glass melt and refractory temperatures, and develops a furnace model suitable for implementation with a real-time controller. A state space model of an end-fired furnace was developed in which the furnace was divided longitudinally into two zones. Zone 1 contains the burner flame “cylinder”, while Zone 2 is beyond the end of the flame cylinder. Separate states are identified for the temperatures of the refractory in the crown, the walls above the glass melt, the walls adjacent to the two primary melt zones, and the floor of the furnace. The furnace ends are also divided into similar zones constituting discrete states. The glass melt itself contains a thin, surface layer and two thicker layers of stratification. In all, 24 state variables are included in the model. The inputs are the net thermal power provided by the flame and the ambient temperature. Simulations were performed in Simulink and Matlab and were used to predict the temperatures of all 24 state variables. The results were verified using data collected from a similar tank furnace at Fenton Art Glass Company. The results showed a significant stratification in the vertical axis of the furnace but very nearly uniform temperatures in the length and width directions. The model was used to study various melting strategies. Preliminary results suggest that using the estimated glass temperature and feedback from thermocouples in the wall and floor of the furnace could lead to significant energy savings in the melt cycle. Suggestions are made for using the model within a real-time control system implementable on a small glass furnace.