| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 9.02 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A levitação magnética consiste na suspensão de um objeto no ar, utilizando forças magnéticas de atracão ou repulsão, contrariando assim a força gravítica aplicada nesse objeto. Nos tempos correntes ´e cada vez mais utilizada em áreas de transporte de passageiros e carga pois reduz drasticamente as perdas mecânicas que os sistemas convencionais acarretam. No entanto, sistemas que empregam levitação magnética necessitam de um algoritmo de controlo pois são, normalmente, instáveis e não lineares. A utilização de algoritmos de controlo encontra-se presente na maior parte dos sistemas, existindo diversas formas de controlo. Atualmente ainda os controladores clássicos, como o controlo PID, são os mais utilizados, mas graças à evolução científica e tecnológica na área da inteligência artificial foram desenvolvidas técnicas de controlo inteligente, como o controlo neuronal. As redes neuronais são agora muito estudadas e utilizadas em áreas de processamento de imagem, problemas de otimização complexos, controlo de sistemas, entre outros. Sendo assim, nesta dissertação são desenvolvidos, para controlar um sistema de levitação magnética, controladores clássicos e controladores neuronais de modelo inverso, modelo interno e modelo de referência e são comparados os seus desempenhos. E assim averiguada a capacidade dos controladores neuronais fazerem o sistema seguir um sinal de referência de maior amplitude, ao invés dos controladores clássicos, devido à não linearidade do sistema. Por último, os controladores clássicos e neuronais desenvolvidos são implementados num micro-controlador, é desenvolvido um algoritmo na linguagem C que permite monitorizar o desempenho do sistema e ´e realizada uma comparação dos diferentes controladores testados.
Magnetic levitation consists of the suspension of an object in the air, using magnetic forces of attraction or repulsion, thus counteracting the gravitational force applied to that object. In current times it is increasingly used in areas of passenger and cargo transportation because it drastically reduces the mechanical losses that conventional systems entail. However, systems that employ magnetic levitation require a control algorithm as they are normally unstable and nonlinear. The use of control algorithms is present in most systems, with several forms of control. Currently, classical controllers, such as PID control, are still the most used, but thanks to scientific and technological developments in the field of artificial intelligence, intelligent control techniques have been developed, such as neural control. Neural networks are now widely studied and used in the areas of image processing, complex optimization problems, systems control, among others. Thus, in this dissertation, classical controllers and inverse model, internal model and reference model neural controllers are developed to control a magnetic levitation system and their performances are compared. Thus, the ability of neural controllers to make the system follow a reference signal of greater amplitude is verified, unlike the classic controllers, due to the nonlinearity of the system. Finally, the classical and neural controllers developed are implemented in a micro-controller, an algorithm is developed in C language that allows monitoring the system performance and a comparison of the different tested controllers is carried out.
Magnetic levitation consists of the suspension of an object in the air, using magnetic forces of attraction or repulsion, thus counteracting the gravitational force applied to that object. In current times it is increasingly used in areas of passenger and cargo transportation because it drastically reduces the mechanical losses that conventional systems entail. However, systems that employ magnetic levitation require a control algorithm as they are normally unstable and nonlinear. The use of control algorithms is present in most systems, with several forms of control. Currently, classical controllers, such as PID control, are still the most used, but thanks to scientific and technological developments in the field of artificial intelligence, intelligent control techniques have been developed, such as neural control. Neural networks are now widely studied and used in the areas of image processing, complex optimization problems, systems control, among others. Thus, in this dissertation, classical controllers and inverse model, internal model and reference model neural controllers are developed to control a magnetic levitation system and their performances are compared. Thus, the ability of neural controllers to make the system follow a reference signal of greater amplitude is verified, unlike the classic controllers, due to the nonlinearity of the system. Finally, the classical and neural controllers developed are implemented in a micro-controller, an algorithm is developed in C language that allows monitoring the system performance and a comparison of the different tested controllers is carried out.
Description
Keywords
Levitação Magnética Sistema não linear Controlador em avanço Controlador em avanço-atraso Identificação Neuronal Controlador neuronal Magnetic levitation Nonlinear system Lead compensator Leadlag compensator Neural identification Neural controller