The implementation of a radial basis function network to reconstruct the distorted spectrum reflected by a FBG sensor is reported.The non-ideal system performance caused by debonding of part of the sensor and the quantization errors might also cause a distortion in the reflected spectrum. In this case

if using conventional peak detection methods directly

there will be a big wavelength detection error.The error can be reduced

the demodulation precision can be improved and the multiple spectrum can be reconstructed using the RBFN

compared with using other wavelength detection methods

such as conventional peak detection

Gaussian fit.