PROOF COPY 013302ISA PROOF COPY 013302ISA Comparison of several well-known controllers used in process control Weidong Zhang,* Xing He, Xiaoming Xu Department of Automation, Shanghai Jiaotong University, Shanghai 200030, People’s Republic of China Received 10 July 2001; accepted 29 April 2002 Abstract In this paper, several well-known design methods, PID control, Smith predictor, inferential control, internal model control, Dahlin controller, deadbeat control, and predictive control, are studied. A suboptimal Smith predictor is derived. The relationship among these methods is investigated. It is shown that these methods are equivalent to each other on certain premises. This explains why they are widely used in process control and provides insight into the work of bringing these techniques together to control system design. Examples are given to illustrate the result. © 2003 ISA—The Instrumentation, Systems, and Automation Society. Keywords: Linear system; Time delay; PID controller; Smith predictor; Dahlin controller; Deadbeat control; Inferential control; Predictive control 1. Introduction The control of processes involving time delay presents a continuing challenge to the control theorists. The nature of the time delay and the sig- nificant amounts of lag which can be introduced into the system response frequently make the use of conventional control algorithms a poor pros- pect. As early as 1953, Cohen and Coon 1ad- dressed the problem of controller design for sys- tem with time delay by correlating PID settings with model gain, time constant, and time delay. Low loop gain was required to avoid instability when time constant was small compared to the time delay, leading to poor system performance. Smith 2suggested a time delay compensation scheme for single input/single output systems, now referred to as Smith predictor. Smith predictor is a simple and powerful control tech- nique for processes with time delay. Its attractive- ness comes from the fact that the design can be performed by using techniques applied to pro- cesses with rational transfer functions. However, it is sensitive to model mismatch and has poor dis- turbance rejection capability. How to overcome these shortcomings is the subject of numerous studies 3–6. Other well-known techniques used in process control include continuous frequency domain i.e., s domaindesign methods, such as inferential control 7and internal model control IMC8, and discrete domain methods such as the Dahlin controller 9, deadbeat control 10, and predictive control 11. All of these methods have been applied to physical systems and achieved good response. Then, an interesting problem is what relationship exists among these methods? This problem was first presented by Zhang 12and this paper will give a detailed dis- cussion. *Corresponding author. Tel: +86.21.62933329; fax: +86.21.62826946. E-mail address: wdzhang@mail.sjtu.edu.cn; URL: http://iic.sjtu.edu.cn/ wdzhang, http://wdzsjtu.home.chinaren.net ISA TRANSACTIONS ® ISA Transactions 42 20031–0 0019-0578/2003/$ - see front matter © 2003 ISA—The Instrumentation, Systems, and Automation Society. PROOF COPY 013302ISA