Artificial neural network based fiber bragg grating sensor interrogation

Abstract

This thesis proposes novel artificial neural network (ANN)-based Bragg wavelength detection algorithms for a given fiber Bragg grating (FBG) sensor. In the introductory part of this thesis, the FBG sensor, its construction, working principle, applications, and interrogation schemes are discussed. Two popular interrogation techniques are described: one utilizing a CCD array spectrometer and the other employing a tunable filter. Then, two separate albeit related ANN-based algorithms are proposed for each of these two interrogation methods. The proposed algorithms address the actual non-ideal behaviors of the components, such as optical intensity variation over time, as well as the poor repeatability and poor linearity of the tunable filter. These algorithms also take into account the fact that the nominal Bragg wavelength of a given FBG sensor often changes during field deployment due to variations in pre-strain or ambient conditions, ensuring that the trained model remains valid and usable even with these variations. Furthermore, the proposed ANN algorithms are compared with traditional FBG Bragg wavelength determination methods, such as curve fitting and centroid finding methods. Finally, the scope of further work in this direction is also discussed toward the end of this thesis.