Research On Auto-Tuning Controlmethods For Dead-time System Based Relay Feedback Technique

In industrial process control, the introduction of auto-tuning capabilities to controllers has been recognized as one of the important means to improve the automation and intelligence of system. The relay feedback auto-tuning method has received a great deal of research attention due to its simplicity, reliability and low-cost. The researches are mainly focused on relay feedback identification and controller design.As to relay feedback identification, the key and difficult problem is how to accurately identify all of the model parameters by means of a single relay feedback test and without a priori information about the process dynamics, in other words, how to improve the relay identification accuracy and efficiency. As to controller tuning, the key point of research lies in how to develop an auto-tuning control method which is not only simple for analysis and design, but also convenient to achieve the trade-off between the robustness and nominal performance of the closed-loop system.Aimed at the aforementioned key problems, this paper researches the model identification methods based on relay feedback technique and optimal analytical design methods of PID controller based on frequency-domain control theory and robust optimal performance indexes for stable dead-time processes and integral dead-time processes with inverse response, respectively. The quantitative relationship between controller parameters and system performance indexes are further established. By combining the above identification methods with control algorithms, the PID quantitative auto-tuning strategy is developed. Besides, for the unstable processes with large time delay which are difficult to stabilize and control, a biased relay feedback test system, with an additional proportional-derivative (PD) inner loop, is improved to identify the model. The stabilization problem of PID controller is studied and two-degree-of-freedom control strategy is developed. The main contribution lies in:(1) For the stable processes with time delay, a systematic approach for auto-tuning of quantitative H_∞PID controller is presented based on the biased relay feedback identification technique.Firstly, the relationship between the ultimate frequency characteristic of the real process and the transfer function of First-Order-Plus-Dead-Time (FOPDT) model is researched. Then, a modified identification method using a biased relay with hysteresis is proposed for stable FOPDT model. It is a dominant merit that three parameters of the FOPDT model are obtained by a single run of biased relay feedback test without any a priori information about the time delay or the steady-state gain. According to the estimated model, an analytical design method for H_∞PID controller is developed based on H_∞optimal performance index, thus the system performance and robustness can be tuned as required. By incorporating the above identification and control methods, the H_∞PID quantitative auto-tuning strategy is developed. In comparison with the standard relay feedback auto-tuning method, the proposed method yields very accurate identification results and feasible controller tuning rules. This new method is not only applicable to FOPDT processes, but also suitable for high-order or large dead-time processes. Besides, considering that the control system usually has the characteristic of nonlinear saturation, which will make the real system performance far away from the expectation, an anti-windup scheme is presented to solve the industrial practical problems.(2) For the integrating processes with inverse response and time delay, which are typically encountered in the level control of boiler steam drum, a quantitative approach for auto-tuning of H_∞PI/PID controller is proposed.According to the proposed method, without any a priori information about the model, four parameters of the integrating model with inverse response and time delay are obtained effectively by means of only a single relay feedback test. Since the exact analytical expressions of limit cycle are derived in terms of biased relay feedback measurements, more accurate parameter estimations are achieved. Besides, the proposed method has capability of disturbance rejection and robustness for model mismatch. Once the process model is identified, the PI/PID controllers are designed analytically based on H_∞optimization and IMC theory. The quantitative relationships between the controller parameters and system performances are established so as to achieve the quantitative auto-tuning control strategy.(3) For the unstable processes with time delay, especially with large time delay, which are difficult to stabilize and control, an additional proportional-derivative (PD) controller is incorporated as an inner feedback loop in the biased relay feedback test system to identify the model.The advantage of the proposed method is that it can relax the constraint on the ratio of delay to the unstable time constant (θτ) and identify unstable processes withθτ< 2, while the conventional method can only identify unstable processes withθτ< ln 2. For the unknown unstable processes, the parameters of stabilizing PD controllers can be determined according to the output to relay feedback system. Three parameters of the unstable FOPDT model can be exactly identified without any a priori information.(4) On the basis of the identified unstable FOPDT model, a new two-degree-of-freedom control strategy is proposed.Firstly, the procedures for determining the complete sets of stabilizing PID controllers for unstable processes with large time delay are presented based on Hermite-Biehler Theorem. Secondly, the setpoint-tracking controller of the two-degree-of-freedom control structure is analytically derived based on H_∞optimization and IMC theory. Finally, the robust performance is analyzed and the practical tuning rule of the disturbance rejector is developed. Its dominant merits are that the controller is easy to be tuned and the load disturbance can be decoupled from the setpoint response. Besides, the unstable processes with large time delay can be successfully stabilized and effectively controlled.