Welcome to the Glass Age

119 all can contribute to efficient operation. A proportional-integral-derivative (three-term) controller (PID) employs a control loop using feedback. They have been widely used in applications requiring continuously modulated control but often offer limited success because: (1) round the clock PID control by a single operator was demanding and unreliable; (2) furnace temperatures were slow to react; and (3) responses to change were subject to long dead times. MPC strategies using dynamic algorithms offer an alternative. They capture process behavior with minimal intervention while maintaining optimum quality, lowest emissions and minimal operational costs. MPC typically works with furnace inputs such as gas, crown and bottom temperatures (Figure 7.7A-C). Mathematical models are created using software such as Expert System III TM and this historical data. These linear models predict the future response of a furnace. The next step requires more complex inter-relationships to be understood. One is how temperatures relate to glass quality. Should furnace temperature be increased for better glass quality, or lowered as would be the case for re-boil or refractory reactions? Such questions showcase areas where artificial intelligence offers more than straightforward linear models. Figure 7.8. Furnace image using near IR camera. Source: Glass Service, a.s.