A multivariable predictive control with internal model of a laminated windshield bending furnace

Laminated windshield bending furnaces have always been controlled manually. Previously, several attempts of automatic control have been tried unsuccessfully. This is essentially due to the fact that this kind of plants is a strongly coupled multivariable system, with nonlinearities and a large time delay. An other serious problem of the furnace to be controlled is the fact that it produces successively and periodically different types of windshields. So the plant is a time-varying system with periodic coefficients and disturbances. In this paper, an adequate combination of modern automatic control techniques, such as predictive control, the internal model of Morari and Smith predictor, have been used to obtain an original and efficient controller scheme for periodic plants. It has been tested on a real plant and has given satisfaction. An increase of the bending precision from 30% to 100% relatively to the manual control has been obtained.<>