Projeto de uma nova topologia de motor esférico e controle em malha aberta utilizando inteligência artificial

Abstract

A new concept of a spherical actuator with multiple degrees of freedom (M-DOF) is proposed in this thesis. It is also presented, with details, the design of a prototype developed for the validation of the new concept. After a thorough review of the state of the art on spherical motors, it can be seen that the main challenges for the use of this motor are: 1) the position control system of the rotor; 2) the monitoring system for precise determination of the rotor location; and 3) the great number of coils present in the prototypes exhited in the literature. Unlike all the spherical motor models found, this design is innovative because it presents a minimalist concept in the quantity and in the arrangement of stator coils, taking into account that it proposes only three coils, which facilitates its construction and its control system. The stator coils are located in the lower hemisphere of the motor. The rotor is composed of a permanent magnet of ironboron-neodymium, which is attached to a system of axes that allows the typical movement of a gyroscope and can reach any position on the upper hemisphere. In order to correctly identify the positioning of the rotor, a rod was adapted to it, which, by means of image recognition, defines its position. The coil currents are individually controlled by a circuit that uses an ARM Cortex-M3 micro controller and a current drive with the LMD18200 integrated circuit. According to the currents that are applied to the coils, a resulting magnetic field causes the rotor shaft to move to a certain position in the actuation area. For the location of the movable rod, it was used a camera monitoring system, which captures the image of the rod on the surface of the spherical motor. With a pattern recognition algorithm, it was possible to identify and calculate the position to which the rotor axis points in spherical coordinates, generating a database that relates the currents in the coils and a rotor position. With a series of data (currents x position) collected, artificial intelligence based algorithms were used and trained, that learned, from the examples, which sets of currents must be applied to the stator coils in order to position the rotor in a desired coordinate, or determine a sequence of currents for the rotor follows a certain trajectory on the spherical surface.