Structural Optimization And Design For Process Control Systems

Modern industrial processes are developing more and more complex and integrating, which complicates the task of process control systems and reqires more sophisticate control strageties. The aim of control structure design (CSD) is to provide some "structural" instructions imposing on the overall control systems, so that the process operation can improve product quality and increase profit margin more easily, while the safety assurances are guaranteed. The phase of CSD is between process design and controller installation, playing an important role in integrating design and implementation. This dissertation studies the two main problems of CSD, namely selection of variables and establishing the relationships bewteen them, mainly focusing on the selection of controlled variables (CVs) and the design of decentralized control structure, which are summarized as follows:1. The problem of selection of CVs is studied and a new strategy for CV selection is proposed. The relationship between CVs selection and optimal operation of process is elaborated and the importance of CVs selection is highlighted, the idea of self-optimizing control (SOC) is introduced, together with existing CVs selection methods developed based on SOC. By employing necessary conditions of optimality (NCO), a new CVs selection method is proposed. In the new method, NCOs are suggested as CVs and online unmeasurable NCOs are approximated by measured varaibles. New method overcomes two main limitations of exsisting methods:(1) based on linearized model (2) only linear combinations of measured varailbes are considered.2. Based on interaction analysis for MIMO processes, a new loop pairing method is proposed. By exploring the statistical properties of gain fluctuation in respective channel due to interaction, the interaction analysis has been extended to two arbitrary subsystems. A new cirterion evaluating interaction effect of control system is proposed, which is then used as a cirterion for decentralized control structure design. New method overcomes the limitations of traditional pairing tools such as the RGA. Furthermore, the problem of computation complexity, which may arise along with new developed cirterion, is addressed. The algorithm is developed based on genetic algorithm, and improved accordingly. This allows the pairing process for high dimensional systems can be accomplished in affordable time. 3. The integrity problem for decentralized control systems is explored, a new algorithm and a new measure are proposed. Firstly, an algorithm for screening decentralized control structure is proposed based on branch an bound method. Compared to brute force approach, proposed algorithm can effectively judge integrity and screen appropriate control structure. Secondly, a quanlified measure for integrity, namely integiry degree, is proposed. This new measure is devised to overcome the conservativeness of original concept of integrity. Hence integiry degree serves as another design index for decentralized control structure. Finally, by combining the interaction measure, the VI-EID pairing rule is summarized.4. The disturbance rejection ability of decentralized control structure is further studied and a new method is proposed. RGA based controllability analysis tools fail to include the impact of disturbances, which is a main purpose of introducing control systems. A new measure is proposed to evaluate the disturbance rejection ability of decentralized control structure by considering the scenarios when controllers are in different statuses. New measure serves as a third design index for decentralized control structure in this dissertation. As a complement, a NSGA-â…¡originated algorithm is developed for multi-objective pairing, based on the three cirterions for decentralized control structure design.5. A complete CSD procedure is summarized, which includes operational objective establishment, variables selection, regulatory control layer, variables pairing and so forth. The CSD procedure is illustrated on an evaporator process and a reactor, separator and recycle process. The purpose of this chapter is to demonstrate the effective usage of methods proposed in this dissertation and show what roles they are playing in the whole CSD progress.Finally, the conclusions are drawn, and the remain challenges of CSD and its possible directions are given.