Optimization And Design Of Fixed-Structure Controllers For Processes With Time Delays Under Constant Disturbaces

In recent years, as the extension of the concept of disturbance, researchers pay more attention on study of the disturbance rejection performance of the control system. In addition, in chemical industry, the disturbance rejection is the primary concern. Therefore, the study of disturbance rejection is a hot research topic and is significant for chemical industry. This thesis focuses on the optimization and design of the widely used fixed-structure controller for disturbance rejection in process industry, and some model synthesis and data-driven controller design methods are proposed for both single-input-single-out (SISO) and multiple-input-multiple-output (MIMO) processes with time delays.1. The design of the model synthesis-based PID and disturbance observer (DOB)-based control system. Among the model synthesis-based controller design methods, internal model control (IMC) is a typical one (other methods can be converted into or be equal to IMC). This thesis deeply studies the controller design based on IMC and proposes some new IMC filters satisfying multiple control objectives for different types of processes. Some optimization techniques are used to obtain the optimal filter constants. In a similar way, the DOB-based control system is designed.2. The optimization of PID controller for disturbance rejection. Generally speaking, for the advanced PID design methods, the PID parameters are obtained by solving a constrained optimization problem, where the output performance indices (defined in the time domain) are formulated as the objective function and the robustness performance indices (defined in the frequency domain) are formulated as the constraints. This kind of constrained optimization problem is a non-convex optimization problem, a special optimization problem. In order to solve this special optimization problem, this thesis firstly studies the stability of the closed-loop system and obtains the stabilizing parameters space. Then, the parameters space satisfying the constraints are constructed. As a result, the optimal solution can be found in the above intersection. By this means, the resulting optimal solution is the globally optimal solution. At the same time, the proposed method only depends on the frequency response of the process model. Hence, it can be applied to the higher order processes. The results are used to design the controller for processes with uncertain delays.3. The design of the data-driven DOB-based control system. The data-driven controller design is a performance index-based control design. For such methods, the controller is obtained by optimizing a given performance index without the process model. It is relatively easy to optimize the performance for set-point tracking since we can easily perform a set-point change. However, if the disturbance rejection performance is optimized, a significant disturbance signal is needed to be introduced into the system, which sometimes is unsafe for the control system. Hence, its application on disturbance rejection is limited. Motivated by this fact, this thesis proposes an indirect data-driven method. Specifically, we do not optimize the controller directly but optimize the parameters of the process model, and the controller is determined by a specific function in terms of the process model. This specific function is determined for disturbance rejection. As a result, the optimal model parameters are obtained, and the goal of disturbance rejection is achieved.4. Multivariate control design for disturbance rejection. Firstly, this thesis considers the design of the decentralized control sturcture for disturbance rejection. The effect of the disturbance on the output is analyzed, and a method for disturbance rejection based on the proposed concept of equivalent non-parametric model is proposed. Besides, another typical control strategy for multivariate system is the full decoupling control, which is used to remove the interactions among loops. However, from the perspective of disturbance rejection, the interactions are helpful for disturbance rejection in some cases. Motivated by this fact, this thesis proposed a partial decoupling control contructure, where the interactions that are against the disturbance rejection of the decoupled loops are removed and the interactions that are helpful for the disturbance rejection are retained.