An exponentially converging observer for the composition profile of distillation columns is presented. Its design is based on the structural properties of observability of distillation column models. A new canonical form that makes it possible to take supplementary measurements (the temperature of sensitive trays) into account is introduced. The proposed approach makes it possible to build a new high-gain extended Kalman filter on a more general canonical form. The observer considered has improved robustness to bounded modeling errors.
展开▼