Download files
Abstract
This thesis presents a number of new developments on decentralized analysis and control design of process networks - defined in this thesis as the interconnections of chemical process systems and their controllers - based on the theory of dissipative systems. In the proposed approach, each subsystem of a network is viewed as an input-output mathematical operator that can be dynamic and nonlinear depending on the physics of
each process system and the configuration of its controllers. Each operator is further characterized by a dissipativity property that describes the behavioural constraints between its input and output. Quadratic dissipativity properties are particularly studied in this thesis to facilitate scalable applications to large-scale networks. Based on the dissipativity property of subsystems and the network topology, the overall dynamic performance of the network (in terms of H-infinity gain) can be determined using simple algebra. In this thesis, the use of physics (thermodynamics), geometric nonlinear techniques, and optimal control theory are investigated to characterize the dissipativity property of open-loop and closed-loop subsystems. All the above developments, which depend on the classical theory of dissipative systems, are sensitive to the equilibrium values of systems. Therefore, prior to the proposed study, a centralized steady-state analysis has to be completed to obtain this particular information. It also follows that if there are changes to the equilibrium values, both the centralized and dissipativity-based studies will have to be redone. In this view, the concept of incremental dissipativity is introduced to potentially enable free-equilibrium dissipativity-based study so that the analysis and control design tasks can be further decentralized (i.e., such that they can be completed with less centralized information). Each development in this thesis is illustrated using a simple network example.