Online neurofuzzy controller: aplicação, análise de parâmetros e contribuições

• Ano de defesa: 2019
• Autor(a) principal: Hugo Rafael Nogueira Gomes
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Taxa de aprendizagem dinâmica; Controlador NeuroFuzzy; ONFC; Controle de processo; Engenharia elétrica; Inteligência computacional
• Link de acesso: http://hdl.handle.net/1843/BUBD-BANMR9

The usage of computational intelligence in process controllers has significantly increased. This work studies the OnLine Neurofuzzy Controller (ONFC), an adaptive fuzzy controller with low computational cost and few parameters. A review of all ONFC versions is made, with computational simulations for the studied controllers. The ONFC learningrate () and the error range (EM) are studied. The Dynamic Learning Rate is discussed and a new approach for the learning rate calculation is proposed, based on the process model. A dynamic adjustment for the EM is presented, called Context Adaptation, and a new approach for the derivative action for the ONFC is proposed, the ONFCDwDe controller.The controllers are applied in a didactic control system, with industrial equipments, to control water flow, in tracking setpoints and disturbance rejection experiments. All controller versions have their performance compared with the PI controller used in the plant. The results highlight how the Dynamic Learning Rate and the Context Adaptation improved the controller performance.