Data fusion schemes in damage identification: Evaluation and comparison

A data fusion process plays a pivotal role to achieve reasonable accuracy and precision in damage identification. An appropriate data fusion scheme can increase the robustness and reliability by reducing imprecision, uncertainties and incompleteness. The current work is to evaluate three major data fusions schemes, conjunctive, disjunctive and compromised, in terms of their efficiency for predicting structural damage. For this purpose, (1) signal characteristics, time-of-flight (ToF), were extracted from signals rendered by an active sensor network, which were embedded in the composite laminates containing single- or multi-damage, in order to form a locus of possible damage location (i.e. the prior perception as to damage from individual sensors); and (2) the entire structures under inspection were evenly meshed and each mesh node established its value pertaining to the occurrence probability of damage at the node position with regard to loci obtained from (1) (i.e. the prior perception as to damage probability at each mesh node from individual sensors). ToF-based perceptions from (1) and belief-based perceptions from (2) were then fused using three schemes, respectively, to establish the posterior consensuses of the overall health status of the composite laminates under interrogation. The comparison of efficiency by using different data fusion schemes is beneficial to the selection of the better, if not optimal, data fusion process.