Quantitative classification of adhesive bondline dimensions usingLamb waves and artificial neural networks

摘要

Adhesive bonding of metal assemblies is gaining acceptance for usenwith safety critical structures, and there is a need for effectiveninspection for both quality assurance (QA) and the assessment ofncondition in service. One aspect of QA is the need for the dimensions ofnadhesive bondlines to be within tolerance and measurable. This paperndescribes the application of ultrasonic Lamb waves in the determinationnof the principal dimensions of two forms of adhered joints (Lap andnT-form) between metal plates. Low order Lamb wave modes (s0 and a1) arenpropagated across adhered bond-lines, and the received signals arentransformed to the modulus frequency domain (FD). The FD data are usednas input to artificial neural networks (ANNs), which are trained tonassociate features in the input data with principal bondline dimensions.nThe performance of different network structures and simplified forms ofnthese is examined, and the technique gives reliable estimates of thenrequired dimensions in bondlines not included in network training. Theninterconnected weights of simplified networks provide evidence of thenfeatures in Lamb wave signals that underlie the successful operation ofnthe method