Fiber-optic strain sensing and neural networks for health monitoring of composite plates.

摘要

This work demonstrates the feasibility and application of extrinsic Fabry-Perot interferometric (EFPI) fiber optic strain sensors in impact-induced dynamic strain sensing and impact damage assessment of advanced composite structures. EFPI fiber optic strain sensors were surface-mounted on composite plates and their transient impact responses were determined. First, the fiber optic sensor responses for nondamaging impacts were compared to those from electrical resistance strain gages and polyvinylidene fluoride (PVDF) piezoelectric film sensors. An in-house finite element program that incorporates shear deformations was used to establish the validity of the first-peak strain response for the EFPI fiber optic sensor. Graphite/epoxy composite plates were used. Experimental and theoretical impact-induced first peak strain responses were in close agreement. Second, impact contact forces of graphite/epoxy composite plates were determined using a neural network. The in-plane strain responses from EFPI fiber optic sensors were directly used as inputs to a neural network to obtain contact force history. An instrumented impact tower was used for impact tests. These tests did not produce visible damage. The contact force history obtained from the neural network is closely matched to the experimental load cell results and to finite element model calculations. Third, damage-inducing low-velocity impact tests were conducted on glass/epoxy composite plates. Four PVDF strain sensors were mounted at various positions on each plate. The impact contact force history indicated damage initiation and propagation. There is an abrupt change in the leading edge of the main contact force peak. The impact-induced strain profiles show generally similar trends and differ from those for the non-damage-inducing impact tests.