Bounded Adaptive Multivariable Control of Cement Raw Material

Abstract This paper deals with the adaptive control of raw material in cement manufacturing. The process studied is a MIMO linear time-varying system with inequality constraints on the control variables. A non-linear objective requires a transformation leading to a new linear objective. The on-line identification is provided by estimation of the time-varying composition of raw material, using Kalman filters. A control law, based on linear programming optimization, takes explicitly into account the control bounds. The adaptive control scheme is therefore indirect. We present simulation results related to an existing cement factory, including the delays and the time constants of the real plant. The results point out a robust and quite satisfactory behaviour of the controller. Moreover, the tuning is straightforward by choosing only one parameter. The implementation of the proposed control law is planned for the near future.