Probability based damage detection on a composite fuselage panel based on large data set of guided wave signals

This paper reports on the application and challenges of a guided wave-based structural health monitoring (GWSHM) methodology on an industrial 5 meter-long curved composite fuselage panel, for detecting barely visible impact damage (BVID), based on integrated sensors data. An extensive network consisting of 72 piezoelectric transducers installed on the panel via diagnostic film, a layer including sensors and inkjet printed flexible circuits, that provides uniform bonding quality and repairability. Threshold setting for damage detection is proposed based on statistical distribution of pristine signals, obtained throughout the span of one month. The threshold is determined through an Outlier Analysis and validated through an iterative cross-validation approach. The model validation yields a Gamma distribution with 1.5% false positive rate. A total of 20 impacts at different energies are conducted on the panel for BVID generation and these are respectively categorized based on their location. Methodologies were developed in order to detect, characterize, and localize the BVID generated from the impacts at various locations of the panel. The reliability of the SHM system is quantified by the Probability of Detection for each damage category by computing the damage area with 90% probability and 95% confidence level (a90 | 95 values). For damages in the skin and at the foot of the stringer a90 | 95 values of 233 mm2 and 365 mm2 are, respectively, obtained. Finally, the summary of the experimental work is presented through the detection of all imparted damages.