Model based process control for glass furnace operation

Since the eighties of the 20th century supervisory control systems have been used in the glass industry. Model Predictive Control (MPC) technology, which was developed in the chemical industry, was introduced in the glass industry for control of forehearths and crown temperatures in glass melting tanks. The mathematical (black box) models used in these controllers were determined from process data. To this end tests were performed on the process so that the time dependent behavior of the process could be observed and the model parameters could be estimated. In the last ten years application of Computational Fluid Dynamics (CFD) models has become possible through model reduction. Via Proper Orthogonal Decomposition the relevant flow and temperature characteristics in the glass melt can be found for the normal operating range of the process. Depending on the operating range linear or nonlinear reduced models can be chosen to make fast predictions of the process behavior. This paper presents some recent results of industrial applications of MPC based on reduced CFD models, called the RMPC approach. Results will be shown for a container glass forehearth. The forehearth controller was designed for two glass colors and two models were needed to cover the working range. In both applications the variation of the relevant temperatures was reduced by 50% or more