Process and systems based methodologies related to control structure selection
This thesis is concerned with an important aspect of process control design, that is, the synthesis of the control structures. A review of the rapidly growing process methodologies' literature is presented and this leads to the identification of wider issues and new problems which are referred to as global instrumentation and forms the main subject of this thesis. The main objective has been the integration of existing process based tools and methodologies with a much more general approach of a systems and control theory character. The problem of Global Process Instrumentation concerns the selection of systems of measurement and actuation variables, found during the synthesis/design and operation of large-scale industrial processes/systems. The role of traditional instrumentation was considered but the emphasis has been on the systems aspects. In fact, instrumentation leads to the shaping of the final system and thus, is crucial in defining the control quality properties and operability characteristics of the final design. The development of these system aspects led to the emergence of an integrated framework for Global Instrumentation. An attempt was also made to abstract some results and formulate generic issues and problems, that would provide a wider scenario for activities in the future. Development of CAD to support the selection of control structures has been a major task undertaken here. The system aspects of Global Instrumentation are demonstrated by studying two specific problems that involve the study of the structural properties of interconnected systems as a function of local selection of sensors and actuators and the problem of well-conditioning badly structured transfer functions. The role of selection of inputs and outputs, on the overall shaping of composite structure properties, at the subsystem level, was examined, and the significance of an assumption related to interconnections, referred to as the completeness assumption, was investigated. Specifically, the significance of the deviations from the completeness, was the subject of the investigation. Matrix Pencil Theory was used to examine the controllability, observability and zero structure related properties of composite systems under partial or total loss of inputs/outputs at the subsystem level. Selecting subsets of the original sets of inputs, outputs to guarantee full rank transfer function, was also an issue that was examined. The above problems were presented as part of an integrated design philosophy that aims to explore the system structure. An integrated approach to the overall problem of control structure selection was formulated and open issues and problems were identified. It was based on the assumption that there exists a progenitor model of the linear type for the process, which, however, may not be well defined. Structural analysis of the system theoretic framework, the interaction measures and the results for evaluation of alternative decentralisation schemes were then used, to specify a step by step approach to the control structure selection. The problem of handling alternative criteria was also considered and basic elements of a system procedure were given. There are many open issues, which were identified and are still open and thus the proposed structural approach should be considered as the first step to the development of an integrated methodology that involves the following major steps: (a) Classification of system model variables and definition of well structured progenitor model. (b) Definition of effective input, output structure based on operability, controllability criteria. (c) Determining the structure of the control scheme by evaluation of alternative decentralised structures. An important part of the integrated methodology for control structure selection is the - so called - interaction analysis. It consists of a number of diagnostics and structural tests that help to restrict the choice of the best scheme. Several of these tests/methodologies were reviewed and some of them were further expanded. The outcomes obtained by these methodologies provided promising results. These results gave the motivation for the construction of a complete CAD package, the "Interaction Analysis Toolbox", written in MATLAB®t. This Toolbox provides many tools and diagnostics that can be applied during the design stages, for the evaluation of the various alternative control structures.