Control of industrial processes using predictors-based control structures

Abstract

This work discusses the analysis and design of predictors-based controllers applied to single-input single-output (SISO) and multiple-input multiple-output (MIMO) stable, unstable and integrative dead-time processes. Dead-time is a characteristic behavior of several industrial processes, capable of leading the system to undesired behaviors and even instability. The greater the delay, the more difficult it becomes to design efficient controllers, and an effective way to address the challenges presented is by using dead-time compensator (DTC) structures. Thus, this thesis proposes new structures for controllers based to the Simplified filtered Smith predictor (SFSP) in order to extend its advantages to multivariable processes and to linear parameter-varying (LPV) dead-time processes. First, it is proposed a delay compensator series cascade control structure for two first-order processes plus dead-time (FOPDT). The industrial environment has some systems with these characteristics, so designing a controller that improves the performance and robustness of these types of systems is quite relevant. The controller incorporates a predictor for each process to manage unstable processes in the discrete-time domain, with the detail that each robustness filter is adjusted related to the perturbation applied to its respective loop. Simulation results show that, when compared with another controller present in the literature, the proposed controller presented better performance results, robustness, better attenuation to disturbances (mainly in the internal loop, due to the cascade structure) and noise, both in the nominal case and in the presence of uncertainties. Second, it is proposed a DTC structure for parallel cascade control of systems with dead time. For the proposed structure there is no need for integrators in the primary controller. Also, a robustness filter is used to reject disturbances and guarantee zero error at a steady state. Simulation results show equivalent performance compared to other recently published work, and better rejection to noise. This thesis also presents a robust dead-time compensator for two-input two-output (TITO) processes with multiple dead time based on the generalized predictive control (GPC). The proposed strategy focus mainly in disturbance rejection by means of a predictor structure proposal. Simulation results show better disturbance rejection performance compared to other DTC in literature. Finally, it is proposed a method to design a LPV controller for dead-time systems based on the SFSP. The advantage of this structure is that there are fewer parameters to tune, as there is no explicit integrator in the primary controller, which only consists of an LPV gain. For this work, the dead time is considered fix and uncertain, so it is treated as uncertainty and an LPV robustness filter is designed in order to deal with disturbances. The main contribution of the proposed SFSP-LPV is the possibility of dealing with nonlinear systems with dead time in an LPV framework. Simulations performed for stable and unstable systems show that the SFSP-LPV provides better performance when compared to other LPV controllers based on the Smith predictor published recently.