Made available in DSpace on 2015-05-08T14:59:39Z (GMT). No. of bitstreams: 1 parte3.pdf: 17677016 bytes, checksum: 0da8175d15bd36a40467f7f50abd9987 (MD5) Previous issue date: 2012-03-28 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / This work presents a study on the application of an ADALINE neural network acting as a notch filter applied to the estimation of the stator flux, in order to obtain the resulting electromagnetic torque of three-phase induction motors (MIT). The estimation of the stator flux was performed by means of the voltage model of the induction machine, in which a integrator is directly applied over the stator counter electromotive force. The ADALINE neural network adaptive filter is employed in this research with the purpose of eliminating existing CC levels, which are present due to the problem of the initial values of the integrator and in the voltage and current measurements. Simulated and experimental results are presented to validate the proposed strategy. The algorithm used in the ADALINE adaptive neural filter simulations was created on the MATLABTM language, and the algorithm used for both the simulations and the laboratory experiments to estimate the flux and the torque was created in the C/C++ language. The hardware used to confirm the effectiveness of the proposed method is based on the Texas Instruments DSP TMS320F28335 platform, along with and induction motor manufactured by WEG, model W21 High Efficiency / Apresenta-se neste trabalho um estudo do emprego de rede neural ADALINE funcionando como notch filter , aplicada na estimação do fluxo estatórico para consequente obtenção do conjugado eletromagnético de Motores de Indução Trifásicos (MIT). A estimação do fluxo do estator foi feita por meio do modelo de tensão da máquina de indução, em que é aplicado um integrador diretamente sobre a força contra eletromotriz do estator. O filtro neural adaptativo ADALINE é empregado nesta pesquisa com o objetivo de eliminar os níveis de cc presentes devido ao problema de valores iniciais do integrador nas medições de tensão e corrente. São apresentados resultados simulados e experimentais para validação da estratégia proposta. O algoritmo utilizado nas simulações do filtro neural adaptativo ADALINE foi elaborado na linguagem computacional MATLABTM, e o algoritmo utilizado tanto nas simulações, como nas experiências em laborátorio para estimação do fluxo e do conjugado foram elaboradas na linguagem computacional C/C++. O hardware utilizado para comprovar a eficácia do método proposto neste trabalho tem como base a plataforma DSP TMS320F28335 da Texas Instruments , juntamente com o um motor de indução fabricado pela WEG modelo W21 Alto Rendimento.

This thesis deals with the analysis of a three-phase winding of an induction engine. After the theoretical introduction concerning different types of windings, two types of their analysis and the importance of the winding factor, two methods of analysis and design tools are explained: Görges diagram and Tingley scheme. This scheme is then used for the design of all possible variants of winding for a certain number of stator slots and the theoretical shape of magnetic field is analysed. The next step is a deeper software analysis of the engines with variants of windings with different coil pitches and number of conductors per slot. Especially the finite element method is used in this part. The obtained values and their characteristics of the simulated engines are compared numerically and graphically. Then the changes of important values for different windings are described. The optimal winding according to chosen requirements is chosen.