Delamination Monitoring of Composite Plates using Vibration-based Surrogate Assisted Optimisation

With the increasing applications of laminated composites in all kinds of industries, it becomes imperative to detect delamination, a frequently occurring damage which causes substantial stiffness loss in composite structures. This paper presents an approach to assess delaminations (in terms of their size, in-plane location and interface) in fibre reinforced laminated composite plates using natural frequencies as indicative parameters and surrogate assisted optimization as the inverse prediction tool. A 3D finite element model has been used to model eight-layer ([0/45/-45/90]s) graphite/epoxy plates having embedded delaminations at different interfaces. Contact elements are employed between the sub-laminates to prevent interpenetration between them. Modal analysis was performed using the finite element simulation to generate a database of natural frequencies for up to 20 modes for various combinations of interface, X and Y locations and sizes of simulated delaminations in the fibre reinforced composite plate. This data was used to train a back propagation neural network to an acceptable level of accuracy. The trained neural network is then used as a surrogate model in the optimization routine for solving the inverse problem to predict the X, Y and Z locations and the dimensions of the delamination. The surrogate assisted optimisation is tested with numerically simulated data and validated experimentally with frequency shifts measured from graphite/epoxy and glass/epoxy plates with artificially induced delaminations. The results show that, the surrogate assisted optimization method, can determine the locations and sizes of delaminations in real plates from measured frequency shifts.