Modeling, simulation and control of glass feeding

Abstract

This paper presents a nontraditional modeling approach to a glass production process in order to facilitate control design. A real distributed-parameter system (glass feeder) is simplified to a lumped-parameter system and then bond graphs are used to develop a model whose parameters are identified from physical insight and historical records. The adopted modeling approach yields a reduced-order nonlinear model, which is much less computationally demanding than CFD models. Moreover, it is valid in a wider range of temperatures compared to black-box linear models often used in control design. As an example, an LQ controller based on a linearized model is designed and validated on the nonlinear model.