Optimal On line Decision Making for Food

In this contribution, we present a distributed decision-making architecture to optimally command thermal processing operation, despite process uncertainty or sudden process disturbances. The network combines and links in a synchronous way modelling and simulation environments with efficient dynamic optimization and system identification tools and methods. The simulation environment takes the place of a virtual plant providing a complete dynamic representation of the system including the evolution of temperature and pressure in the retort unit as well as temporal and spatial distribution of temperature and quality or safety parameters within the product. Such virtual representation will be regularly confronted with plant measurements to quantify the degree of discrepancy (uncertainty) between real plant and models and react accordingly when such discrepancy becomes unacceptable by re-estimating plant parameters either during the cycle or from batch to batch. The virtual plant will be also accessed by the regulatory system as well as the dynamic optimization module. In the first instance to estimate unmeasured states related with the product status under feed-back control. In the second, to continuously re-compute optimal cycle profiles so to respond to unexpected disturbances or deviations from the prescribed safety constraints while maximizing quality attributes. Experimental evidences of the resulting complete system performance will be given on a pilot plant scenario.