Virtual Reference Feedback Tuning for PID Controller Design
Model-based techniques have been the predominant controller design methods that have received much research interest in the past several decades. For example, PID tuning rule based on the first-order-plus-dead-time (FOPDT) model and ITAE performance index, direct synthesis design method, and Internal Model Control (IMC) design. These transfer function models like FOPDT and second-order-plus-dead-time (SOPDT) models are practical for industrial applications, as it is very easy to obtain such models from the standard identification experiments. These models are commonly employed in the direct synthesis and IMC design methods, which employ two-step design procedures, where the first step involves the identification of the best model among pre-specified model structures that gives the best modeling accuracy, followed by the controller design based on the model thus obtained. However, the use of these models in modelbased PID design suffers several drawbacks. First, these models may not carry sufficient information for the process under control and thus the performance of the resulting PID controller may become poor if the discrepancy between the process and model is too large. Even when these models have acceptable modeling accuracy, the model with better modeling accuracy does not necessarily guarantee that the resulting model-based PID controller design would give better control performance. As a result, model-based PID controller design methods inherently need to resort to a trial and error procedure to cope with such interplay between the model identification and controller design. Moreover, if one desires to design a PI/PID controller by the direct synthesis and IMC design methods, some approximations of the time-delay in the FOPDT and SOPDT models are necessary, which would inevitably lead to the degradation of the best achievable control performance. Lastly, the tuning parameter in these design methods, e.g. IMC filter time constant, is usually designed on-line by trial and error procedure to yield the best achievable control performance. This ad-hoc design approach places additional burden on the users of these model-based design methods.To alleviate the aforementioned drawbacks, a data-based PID design method is developed by using the Virtual Reference Feedback Tuning method in this paper. Based on the process input and output data measured in an open-loop experiment, a systematic one-step design procedure for PID controller is discussed. Unlike the model-based design methods, the proposed method can be carried out without the need of a process model and can do without the trial and error procedure in designing the controller parameters. Extensive simulation results show that the proposed PID controller can provide better or comparable control performance than the model-based IMC-PID and Maclaurin-PID controllers, which have been tuned on-line to give their respective best control performances.
PID controller Model-based Controller Design Virtual Reference Feedback Tuning (VRFT)
Bu Xu Min-Sen Chiu
Department of Chemical and Biomolecular Engineering National University of Singapore 4 Engineering Drive 4,Singapore 117576
国际会议
西安
英文
2007-08-15（万方平台首次上网日期，不代表论文的发表时间）