Model predictive control tools for evolutionary plants

Abstract

The analysis and design of control system configurations for automated production systems is generally a challenging problem, in particular given the increasing number of automation devices and the amount of information to be managed. This problem becomes even more complex when the production system is characterized by a fast evolutionary behaviour in terms of tasks to be executed, production volumes, changing priorities, and available resources. Thus, the control solution needs to be optimized on the basis of key performance indicators like flow production, service level, job tardiness, peak of the absorbed electrical power and the total energy consumed by the plant. This paper proposes a prototype control platform based on Model Predictive Control (MPC) that is able to impress to the production system the desired functional behaviour. The platform is structured according to a two-level control architecture. At the lower layer, distributed MPC algorithms control the pieces of equipment in the production system. At the higher layer an MPC coordinator manages the lower level controllers, by taking full advantage of the most recent advances in hybrid control theory, dynamic programming, mixed-integer optimization, and game theory. The MPC-based control platform will be presented and then applied to the case of a pilot production plant.