Neural network demodulator for optical Bragg strain sensors

The reflection spectrum of a Bragg optical grating has a center wavelength defined by grating parameters. As strain sensors, Bragg gratings provide an absolute measure of strain and have excellent multiplexing capability. The change in the output of a Bragg grating for applied strain is characterized by a linear shift in the Bragg grating center wavelength. Other parameters can be associated with shift in Bragg wavelengths such as temperature. The complexity of conventional demodulation systems is much greater for optical fiber sensors based on spectral changes rather than irradiance changes. Hence, commercially-available demodulators for Bragg sensors can be expensive over those for irradiance-based sensors such as Fabry-Perot Interferometric sensors. This study looks into the development of an Artificial Neural Network (ANN) based demodulation system for Bragg strain sensors that can bring down the costs of its instrumentation significantly. Standard optical detectors and filter components provide parallel inputs to a neural network demodulator which determines the absolute strain or shift in the Bragg center wavelength. Efficient temperature compensation techniques using either simple temperature sensors or dummy Bragg gratings are proposed and software tested. A comprehensive foundation, through software simulation is established for further development. ANN models for Bragg grating strain sensors are developed for changing field parameters, such as noise, filter non-linearity, and amplitude variation, to account for practical issues regarding source, filters, and multiplexing.