On-line measurements using neural networks and soft computing for the control of glass production furnace
Liquefied petroleum (LP) gas is used as a backup energy system for glass production furnace. LP gas is mixed with air at a desired ratio in order to get a proper gravity. The previous research was conducted to improve a performance of monitoring and diagnosis of glass production furnace. The objective of this research is to apply adaptive-network based fuzzy inference system (ANFIS) and counterpropagation neural networks (CPN) for on-line monitoring and measurements of LP gas for glass production. Three inputs, air inlet absolute pressure (PSIA), air/mixed differential pressure (PSID), and propane/mixed differential pressure (PSID), were selected for on-line measurements. Three to five ANFIS membership functions were used for each input of ANFIS for on-line measurements of the specific gravity of the LP gas. The ANFIS using Generalized Bell membership functions yielded the best performance with an average absolute error of 0.23%, a maximum error of 1.78%, and a minimum of 0%, while an average error of 1.7%, a maximum of 4.58%, and a minimum of 0%, were obtained using 3x12x1 CPN.