Location of impacts on composite panels by embedded fiber optic sensors and neural network processing

Publisher Website
Google Scholar
Abstract
Location of impacts on an anisotropic polymer matrix composite panel was demonstrated by using a neural network to process the outputs of embedded fiber optic strain sensors. Three extrinsic Fabry-Perot interferometer sensors were embedded in a graphite/bismaleimide composite with a unidirectional lay-up. The location of an impact can be calculated directly by triangulation from the difference in arrival times of the impact-generated stress waves at the embedded sensors. A data set of 132 experimental results was generated by impacting the panel at evenly spaced locations, and measuring the time differences. A back-propagation neural network was simulated using commercial software, and the data set was used to train the network. Following training, the network was capable of determining the location of randomly located impacts with an accuracy of a few centimeters. These results were comparable to the accuracy achieved for impact location on an isotropic aluminum plate, indicating that the neural network performance is independent of material anisotropy.

This publication has 0 references indexed in Scilit: