Guided waves for autonomous online identification of structural defects under ambient temperature variations

Ultrasonic guided waves that are excited by piezoelectric transducers can be used for the autonomous online identification of structural defects in thin structures. The proposed technique in this paper continuously analyzes a damage metric which is defined as the maximum residual amplitude of the differential signal. A special focus is on the decision making to discriminate the undamaged from the damaged state of the structure where the appropriate detection thresholds are derived statistically from the inverse cumulative distribution function of the damage metric during an initial training phase. An integrated trend analysis by means of the moving average mitigates the impact of statistical outliers and reduces the probability of erroneous identifications.Long-term measurements under ambient temperature variations have been conducted on an aluminum and a composite plate to study the properties of the proposed novelty detection framework. In this process the temperature effect was compensated by the well-known combination of optimal baseline selection (OBS) and baseline signal stretch (BSS). In case of the aluminum structure two artificial cracks with different sizes have been identified reliably. Consistent results were found on the composite specimen where an impact damage was identified for different excitation frequencies.