A hierarchical data fusion scheme for identifying multi-damage in composite structures with a built-in sensor network

Multi-damage causes much more complex scattering phenomena in captured signals than mono-damage does. Examination of an individual signal may fail to provide sufficient information to identify all instances of damage. Upon comparative evaluation of the performance of forward and inverse inferences for damage identification, a data fusion scheme was developed for predicting multi-damage in a structure with the aid of a sensor network. The approach, conducted hierarchically by activating different sensors in a sensor network, fused an extracted signal feature, time-of-flight (ToF), at different levels, to provide an overall consensus as to all possible instances of damage. This consensus was presented in an intuitional contour map indicating the probability of damage occurrence. Benefiting from the sensor network, the dependence of identification processes on a specific sensor was minimized, and the need for interpreting complex signal scattering by multi-damage was avoided as much as possible. To facilitate the extraction of ToF from raw signals, a signal processing approach, scale-averaged wavelet power (SAP) analysis, was introduced. As validation, the proposed identification scheme was employed to gauge dual delamination in a CF/EP woven laminate with a built-in active piezoelectric sensor network. The results have demonstrated the excellent capability of the approach in evaluating multiple structural damage sites.