Aplicações de técnicas computacionais à interrogação de sensores ópticos baseados em grades de período longo

Abstract

The application of computational techniques, signal processing and feature extraction to the interrogation of optical sensors based on long-period gratings is the object of intensive development of recent research and still requires advancement. Considering the increasingly wide application of optical fiber sensors, this work aims to present alternative forms of interrogation which do not rely on spectral analysis or high-cost devices. Two proposals are presented. The first one is based on the separation of the sensor signal in spectral sub-bands and the ensuing analysis of the optical power in each sub-band by an artificial neural network. The second proposal is based on feature extraction applied on reflectometry traces measured from fibers containing long-period gratings and on the use of feature descriptors as explanatory variables in a multiple linear regression algorithm. Both proposals rely on computational methods in order to achieve the goal of recovering the information collected by the sensor and making it available in a suitable format. The obtained results indicate that the proposed methods, with simple experimental setups, present better performance than similar methods, or present comparable performance while adding the inherent advantages of in-fiber sensing.