Nonlinear adaptive control of industrial process in cascade configuration and its application to paper vacuum dewatering process.

摘要

Many industrial processes are in a cascade configuration in which the material being processed goes through a sequence of processing units. In this dissertation, control of such type of industrial process will be discussed with an emphasis on the application of paper vacuum dewatering process. A pre-emptive control algorithm that makes use of in-process measurement of upstream processing units to prevent errors from propagating to downstream processing units has previously been proposed. The limitation of this control algorithm is discussed and an improved design is proposed to realize the output control objective. Moreover, in many cases, in-process measurements of the relevant variable other than the output of the last processing unit is not feasible, except by the use of surrogate or model based soft measurements. For the vacuum dewatering application, the use of airflow through wet sheet can be used as a surrogate measurement for moisture content. Such a surrogate measurement model is experimentally validated. Because these models are subject to great variabilities, an adaptive calibration algorithm is therefore proposed in this dissertation to adaptively calibrate the surrogate measurement model online. The robustness issue of the proposed adaptive calibration algorithm to the actuator dynamics is also investigated. Both known and uncertain actuator dynamics are considered. Three different control algorithms are designed to deal with three different types of actuator dynamics. Simulation results on the vacuum dewatering process application show the effectiveness of the proposed control algorithms.;To experimentally verify the proposed control algorithms, an experimental setup is built up in our lab because an actual paper manufacturing process is not available for us to do the experiments. The setup mimics the vacuum dewatering process in a reverse way. Because of the discrete nature of the experimental setup, the proposed control algorithms are modified into discrete versions. The possibility of multiple box feedforward is also investigated for the situations where the actuator dynamics are significant. The experimental results show the effectiveness of the proposed the control algorithms.