Radial-basis function network for the approximation of FBG sensor spectra with distorted peaks

The implementation of a radial-basis function network to approximate spectra of the signal reflected by a fibre Bragg grating sensor is reported. This algorithm helps the interpretation of the data acquired with equipment as an optical spectrum analyser. Results using a double-peaked spectrum from a uniform fibre Bragg grating sensor show that the common practice of fitting the spectrum with different interpolation methods and finding its peak, or directly finding the maximum intensity position of the raw spectrum, would cause a larger error when compared to searching for the peak of an approximated spectrum using the proposed neural network. An example is demonstrated through two experiments measuring the volumetric shrinkage of polymeric resin using a uniform FBG and a HiBi FBG embedded in the material. The obtained accuracy is higher than that obtained with the simple non-processed peak detection.