Research On Signal Processing And Analysis For Disbond Detection In CFRP T-joint

The typical damage (disbonding) including generation and extensions of CFRP T-joint, understatic extension testing, were investigated by piezoelectric sensors and active Lamb wave monitoringtechnology. To achieve the damage alarming and determine the degree of the T-joint interfacialdisbonding, the characteristic parameters from sensor signal is abstracted, combined with the BPneural network system optimized by the dynamic adaptive particle swarm optimization (DAPSO).Optimization and experiments, based on the frequency dispersion of the Lamb waves spreadingin T-joint, were carried out to get kernel parameters, such as excitation signal center frequency andarrangement of sensors. After the mode mixing sensor signal was filtered, characteristic sensitiveparameters in time-frequency were identified using HHT method. On the application of improved BPneural network system, disbonding damages of T-joint were finally identified efficiently, preventingthe training from being partial optimum.Static tension test was carried out in two groups of CFRP T-joints one of which were healthysamples and the others with pre-setting damages. The experimental results showed that interfacialdisbonding appeared firstly in T-joint triangle filling area and then extended to flanges on both sides,signal energy and least square peak factor, both linearly decrease with time before failure, coulddescribe disbond extension of T-joint. The network training data improved by adaptive particle swarmoptimization algorithm were corresponded to experimental results with error range of3.8%~4.7%.The study, being sponsored by preliminary research funding and accomplished in state keylaboratory of mechanics and control of mechanical structures, provided helpful reference inparameters extraction and identification of damages for health monitoring of integral compositestructures.