Latent variable modeling for energy savings in an industrial flat glass production process

The Thesis aims at developing strategies for reducing both energy consumption and production costs in the glass melting process. A comprehensive analysis of the flat glass manufacturing process identifies the key variables that most significantly impact energy usage in the furnace. A Partial Least Squares model is constructed to establish correlations between specific energy consumption and key process variables. Based on these findings, targeted interventions are implemented on specific variables, with the objective of achieving overall process energy savings. Additionally, the thesis explores two distinct phases of the raw material preparation process: preheating and moisturizing. These stages are of critical importance for the efficient melting of glass and require a precise control of temperature and humidity through the introduction of hot water and steam. A systematic methodology is proposed to adjust batch heating and moisturizing based on ambient temperature, thus further enhancing energy efficiency by minimizing excess energy usage. For the glass melting process, targeted interventions such as adjustments to glass temperature, canal temperature, batch charger speed and the relocation of the combustion air filter, led to a total estimated annual energy savings of 981358 Sm³, corresponding to 745831 €/year in cost reductions. In the batch preparation process, the study established a reference value for steam use, indicating that it is necessary when the ambient temperature falls below 14°C. A predictive model was also developed to guide operations, with an error margin of 10%, offering a reliable tool for daily decisions and future improvements of the process.