Damage detection in composite structures using multi-sensing, actuation, and neural network systems

This dissertation presents the results of the research investigation on damage detection in composite structures using multi-sensing, actuation, and neural network systems. The effect of prescribed delamination on natural frequency of laminated beam specimen is examined both experimentally and theoretically. Modal testing of a perfect beam and beams with different delamination size is conducted. A back propagation neural network model is developed using the results from the beam theory and used to predict the delamination size. The neural network model successfully predicted delamination size. Broadband acoustic emission (AE) signatures are also used to detect delaminations in composite beams. The AE wave propagation along a perfect composite beam and composite beams with known delamination sizes are investigated. The results show that AE strength and AE energy decrease significantly over delamination area. This phenomenon is used successfully to determine delamination location and size in composite beams. Waveform-based acoustic emission technique is used to determine the location of low velocity impact in composite plates. The Gaussian cross-correlation method and the Hilbert transformation are used to obtain arrival time differences among three broadband AE sensors and to overcome dispersion problems. The locations of impact points are successfully determined. A method for determining the low velocity impact induced contact force on laminated composite plates is developed using finite element method. A backpropagation neural network is used to estimate the contact force on the composite plates using strain signals. The neural network approach for estimation of contact force proves to be a promising alternative to more tradition techniques, particularly for an on-line health monitoring system.