Este trabalho pretende verificar o potencial e as restrições da abordagem adaptativa no projeto de um sistema de controle de velocidade de motores elétricos de corrente contínua na presença de atrito seco. O compensador foi obtido através do método de hiperestabilidade aplicado sobre um sistema adaptativo com modelo de referência e estrutura paralela, considerando um modelo simplificado do motor elétrico. O sistema controlado foi analisado através de simulações onde o compensador foi implementado de forma discreta e o modelo do motor era mais complexo e incluía não linearidades como saturação e atraso. As seguintes modificações mostraram-se necessárias: normalização do sinal do erro, interrupção das integrações quando ocorre saturação do sinal de controle, e interrupção da adaptação quando a norma do erro relativo é menor que um certo limite. O compensador modificado foi utilizado em ensaios experimentais cujos resultados apresentaram boa concordância com os resultados das simulações. Apesar dos bons resultados obtidos, pois o sistema se mostrou estável e o efeito do atrito seco foi atenuado, certas restrições à abordagem adaptativa foram percebidas: a teoria não fornece subsídios para a determinação de importantes parâmetros de projeto e as modificações adotadas exigem um certo conhecimento prévio do sistema.

Any mechanism is subjected to dry friction, in a bigger or smaller degree. This friction may limit the mechanism performance, mainly if the requirements are critical. In this case one must fall back upon non-linear compensations. The adaptive approach was chosen because it deals with Sistema whose parameters are not well known or change during the operation, as normally it occurs with dry friction. On the other hand, the applications of adaptive approach has been limited once adaptive compensator robustness problem is, in many ways, an open question. The existing applications are generally based on ad hoc solutions that normally depend on considered applications. This work intends to verify whether some of this suggestions apply on direct current motors in the presence of dry frictions. First two parallel direct reference adaptive system were designed through classic methods, the Lyapunov method and the Hyperstability method based on the nominal model of the system, where several simplifications were admitted. Since the obtained algorithms were quite similar, we chose the one which was got through the Hyperstability method in order to analise it with more details and to implement it. We verified that this adaptive system is nominally stable, and that the error between the system and reference model states is bounded if the son anticipated external inputs are also bounded. In the sequel the obtained algorithm is converted in order to allow its implementation through a digital computer, and suitably changed in order to increase its robustness.The discrete algorithm was obtained through Eulers method, and the adopted changes were the use of normalized error signal in the adaptive algorithm, adaptation process were frozen when relative error norm is smaller than a certain bound, and the integrations performed by the algorithm were frozen when saturation occurs. The simulations showed that such modifications are needed, in other words, the nominal compensator drives the system to the instability when more real conditions are considered. This modified compensator was implemented (with no others changes) in an actual system, the carried out experimental tests showed that the system has good performance, stability and robustness characteristics, and that supports the simulations results. This sort of approach has, however, some limitations: the theory does not give means to define important design parameters as the adaptive gains, the used modifications depend on a priori knowledge as the noise upper bound and saturation bounds. Furthermore the tests showed some situations where the adaptive process was not stopped, what might end up in instability if the adjustable compensator parameters drifted, which fortunately does not happen, at least in the tested conditions. Nevertheless we believe that use of adaptive approach is feasible in this specific control problem, by reason of good performance, stability and robustness characteristic showed by the system during the tests.