Optimal trajectory profile generation of continuous processes using multimodel framework

Optimal transition from one operating grade to another can give profitable operation of continuous processes like polymer, paper and pulp etc. Design of optimal trajectories governing these transitions requires relatively accurate process model and the inherent models are typically non-linear. This paper proposes the use of a multimodel framework for optimal trajectory profile generation for such processes using only routine operating data instead of rigorous first principle models. Local linear models developed at the operating points are interpolated to describe the nonlinear dynamics of the process. The multi-model developed is the basis for trajectory optimization. The novelty of the work lies in the usage of plant archived data for training the multimodel framework. Further a simplified weighting strategy is proposed for interpolation of local model dynamics during transition. Simulation results on a polystyrene process illustrate the capability of the proposed approach.