Approximation Methods Of Optical Spectra: An Analysis Of The Application Of Constructive RBF Networks In Fiber Optic Sensing

Título: Approximation Methods Of Optical Spectra: An Analysis Of The Application Of Constructive RBF Networks In Fiber Optic Sensing

Autores: Paterno, A. S.; Arruda, L. V. R.; Kalinowski, H. J.

Resumo: This paper reports an analysis of the methods used to interrogate fiber Bragg grating sensors that use the whole reflection spectrum of the sensor. Standard methods like the identification of the peak of the spectrum in the Gaussian or polynomial fit, as well as the direct peak detection in the non-processed data and the determination of the peak by calculation of the centroid of the spectrum are also analyzed. A method recently proposed that uses a constructive Radial-basis function network is analyzed and compared to the other methods, demonstrating that this method produces the lowest uncertainty in the peak determination. A new methodology to find the optimum radius of the radial-basis function for fiber Bragg grating sensor interrogation is proposed. An analysis with respect to characteristics of the acquired data and processing time is also described.

Palavras-chave: Radial-basis function networks; fiber Bragg grating sensors; fiber optic sensors; fiber Bragg grating interrogation

Páginas: 10

Código DOI: 10.21528/lmln-vol3-no2-art2

Artigo em PDF: vol3-no2-art2.pdf

Arquivo BibTex: vol3-no2-art2.bib