Frequency Based Model Predictive Control of a Manufacturing System

Abstract

Manufacturing scenarios have to incorporate many dynamic influences such as urgent orders or unavailability of resources. To better address this, tools of control theory and dynamic modelling should be employed. In this paper, we present a novel multi-product dynamic model inspired by an existing bond graph model for manufacturing systems, and apply a Model Predictive Controller (MPC) to steer the machines to a reference state to fulfil a given product demand. In previous related works, the capacity adjustment of the machines is based on local information only and the existing model can only represent a fixed manufacturing configuration. Thus, the contribution of this approach is twofold: it allows applying optimal control methods, generating a centralized solution based on global information; it depicts the machines as autonomous elements, enabling to represent more flexible configurations of production systems. We show in simulations that an appropriate cost function and a suitable prediction horizon length are important factors for the control of the system.