Medição indireta ótima estocástica e neuronal em sistemas dinâmicos.

Abstract

Indirect measuring is an alternative wherever is difficult to perform a measuring by means of sensors directly associated with variations in the magnitude to be measured. This work aims at developing procedures for indirect measurement systems, whether this measuring is done in real time or not. Indirect Measurement Systems (IMS) are classified in this Thesis in such a way as to specify the sort of treatment to be used for the attainment of a given magnitude, also including the means to analyze those results. Each method must be used according to the application, whether isolated or in groups, as to obtain the magnitude of interest. In this thesis, two techniques used in case studies are approached: the Kalman Filter and the Artificial Neural Network. Considering the complexity of the applications performed in case studies, it is crucial to develop methods addressed to the application as well as simulation and prior analysis of the results until the system’s implementation. Four IMS are developed: indirect speed measuring of an space aircraft; indirect temperature measuring of an object inside an oven; indirect rotor speed measuring of an induction motor and, as supplement to this case study, the robust indirect speed measuring. The IMS implementation allows us to check the viability of using mathematical tools to conduct indirect optimal measurements, both, stochastic and neuronal in dynamic systems.