Neural networks improving robustness on fiber Bragg gratings interrogation systems under optical power variations

The herein article presents a signal processing methodology based on the process knowledge learned by Artificial Neural Networks - ANN aiming to compensate the undesirable power variations of the light in a Fiber Bragg Grating - FBG demodulation system. Conventional ratiometric signal conditioning requires the acquisition of the intensity light signal related to the power light source, minimizing only linear variations of the light imposed to the interrogation system. The proposed method brings better benefits, particularly when the measuring range of the system shall be extended, because of the redundant information generated by the addition of more fixed filters, improving the generalization capacity of the ANN. A temperature interrogation system is also presented and arranged on a typical and useful demodulation architecture, which adopts three filters to extend the measuring range. Preliminary results from a temperature experiment showed the ANN ability to make the FBG demodulation system robust to light variations, including some non-linear characteristic.