<title>Fiber optic vibration sensing and neural networks methods for prediction of composite beam delamination</title>

Abstract

Extrinsic Fabry-Perot interferometric (EFPI) fiber optic sensors were used to characterize delamination size and location in laminated composite bemas. Six eight-ply glass/epoxy composite beams, each 26.04 cm long and 2.33 cm wide, were fabricated with five midplane delamination sizes ranging from 1.27 cm to 6.35 cm long. The five delaminated beams as well as undamaged beams were tested for their first five modal frequencies. The modal frequencies shifted with changes of delamination size and location. The EFPI fiber optic sensors measured identical model frequencies as piezoelectric ceramic sensors. However, EFPI fiber optic sensors showed more sensitivity and better signal-to-noise ratios. Analytical classical beam theory and finite element methods validated the EFPI modal frequency measurements. A feedforward backpropagation neural network predicted the size and location of a prescribed mid-plane delamination in the composite beam using the EFPI fiber optic sensor modal frequency measurements. Modal frequency data sets from classical beam theory were used for training and testing the network. The delamination size and location predictions from the neural network had an average error of 5.9% and 4.7% respectively.