Controle com aprendizado iterativo para processos em batelada

• Autor(a) principal: Granzotto, Matheus Henrique
• Data de Publicação: 2010
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/15149
• topic: Controle de processos em batelada; Controle com aprendizado iterativo; Controle de processos químicos; Batch process control; Iterative learning control; CNPQ::ENGENHARIAS::ENGENHARIA QUIMICA

Batch processes are usually systems for which defined raw materials quantities are combined in specific recipe during a designated time to produce intermediate or final high added value products and generating the smallest waste of raw material. Most of the batch processes with economic interest are inherently nonlinear, making the control task a challange. The iterative learning control is a methodology that involves the use of learning strategies in order to achieve the better reference trajectory tracking and disturbance rejection. Thus, the addressedmethodology uses important features for the control of industrial batch plants, as the disturbances rejection, the inclusion of restrictions on the control variables making control more realistic, added to the ability to track reference trajectories over the batch, among others. In this context, this work presents the iterative learning control methodology successfully applied to various plants with distinct features, ranging from minimum and nonminimal phase SISO and MIMO linear systems and nonlinear systems. Comparative studies about the proposed control quality against classical control methods were also performed to show the degree of increasing performance due to the iteractive learning strategy. The results of the use of iterative learning control strategies looked promising since they provide versatility and tuning with easy, making the controller intuitive. For LTI minimum phase processes, the learning was almost immediate. The iterative learning controller was able to compensate for nonminimum phase behavior resulting from delay or nonminimum phase zeros, leading to a perfect tracking of the reference trajectory. The use of the methodology for nonlinear plants control was evaluated for reaction systems in which are tried two different ways to get the learning matrix, such as FIR models and locally linearized models. The iterative learning control methodology brought improvement on the setpoint tracking in all studied cases.