Multivariable end-point weighted generalized predictive control of a space crystal furnace

A multivariable end-point state weighted generalized predictive controller for the Bridgman crystal furnace TITUS is developed. A multivariable temperature prediction is done by a reduced dynamic representation using a matrix polynomial representation of the plant. The theoretical equivalence between the proposed multivariable control and Kalman filter state prediction is proven. It is shown, that the incorporation of the stochastic model improves the control performance.