CONTROL SYSTEM DESIGN FOR HIGH PERFORMANCE

Abstract

In today's economic climate where large capital investments in highly competitive or low-margin businesses are closely scrutinized, process companies need to deliver the most and the best for less and process operators are being asked to squeeze more performance from existing process units within the plant. This ever increasing competition, together with tougher environmental and safety regulations, has played a pivotal role in laying out more stringent plant product quality specifications and consequently, the performance requirements for process plants have become increasingly difficult to satisfy. With this in mind, the thesis is devoted to the development of new control design techniques to attain high process performance for both single variable and multivariable processes with more emphasis on the latter. For single variable control design, two control strategies, namely the Proportional-plus-Integral-plus-Derivative (PID) and the Internal Model Control (IMC), are investigated and the practical issue of control saturation is addressed. PID controllers are widely used in the process control industry and are adequate when the process dynamics are benign and the performance requirements are modest. An enhanced technique for the tuning of PID controller is developed which incorporates a new second-order plus delay modeling method and a gain selection strategy to achieve a user-specified sensitivity constraint. The proposed PID design demonstrates promising results in controlling an extensive class of processes. However, such controllers would become ineffective when the process deadtime is significant and advanced control strategies such as the IMC would have to be considered. In the thesis, a new IMC scheme is proposed. The proposed scheme has the advantage of using a fixed order controller, which eases the implementation. The design does not require an explicit transfer function model of the process and the parameters of the controller are given in simple formulae. Simulation shows that the proposed scheme gives consistent and satisfactory performance. In practice, all control systems face the problem of controller output constraints. Since the feedback loop is essentially broken down in such situations, the performance of the closed-loop deeply deteriorates. In the thesis, a new and effective method has been proposed to handle controller output saturation. It incorporates the controller output constraints into the controller design with the given specifications properly adjusted, so that the resultant system can have best achievable performance without the controller output running into saturation. This thesis devotes three chapters on the control of multivariable processes since most processes in the industries are inherently of the multivariable nature. The renowned difficulty in their control due to the multivariable interactions has posed a great challenge in the area of control design. Depending on the levels of complexity in the processes dynamics, the degree of multivariable interaction and the requirement, three control schemes or strategies, namely, multi-loop control, multivariable unity feedback control and multivariable internal model control, have evolved and they are investigated each in the details. For processes with modest interactions, multi-loop controllers are often more favored due to their simple structures. In the thesis, this simplest scheme is investigated with the extension of the original single-input single-output (S1S0) modified Ziegler-Nichols method and the dominant pole placement method to multi-loop versions for the control of two-input two-output (TITO) processes. The respective original scalar versions of the two methods have been proven effective for the control of single variable processes, and thus their extensions to the multi-loop case are highly valued. It is found that these two designs both lead to nonlinear problems and novel approaches are presented each to solve them. Both methods demonstrate performance improvement over the existing methods that emphasize on the suppression of interactions while our methods take the multivariable interaction into consideration and thus can speed up loop responses. The simple multi-loop controllers would rapidly become ineffective when the process complexity increases. The drastic increase in energy costs in the 1970s led to major redesigns of many new and old processes, using energy integration and more complex processing schemes. This increases control loop interactions and expands the dimension of control problems and high performance control of these complex multi variable processes with explicit suppression of interactions can only be achieved by a multivariable controller with its elements possibly in general transfer function forms, instead of simple PID forms. In the thesis, a systematic method is proposed for the design of general multivariable controller with the objective to attain a controller of lowest complexity and achieve fastest loops with acceptable overshoot and minimum loop interactions. For complex multivariable processes with significant time delays, an effective advanced control scheme is multivariable IMC control. In the thesis, a new approach to the IMC analysis and design is proposed for decoupling and stabilizing multivariable processes with multi-delays. All the stabilizing controllers which solve this decoupling problem and the resultant closed-loop systems are characterized in terms of their unavoidable time delays and non-minimum phase zeros. Such delays and zeros clearly quantify performance limitations for any multivariable IMC system which is decoupled and stable. A theoretical control design for the best achievable performance has also been presented based on the proposed characterization. Model reduction has been then exploited to simplify both analysis and design involved. The results presented in the thesis have very practical values as well as sound theoretical contributions. With necessary packaging, the finding in the thesis can be applied to industrial control systems. This has been evidenced by a number of following up applications, implementation of the results of this thesis in industrial control PCs and incorporation of the functionality into a Matlab toolbox. The author strongly believes that these results will contribute as useful modules towards the visionary high performance industrial controllers.