Force and Sound Data Fusion for Enhanced Tap Testing Scanning of Composites

Coin or hammer tap testing is one of the oldest methods of NDT inspection. Inspection techniques originated with human operators observing the tonal change of the audible sound produced by striking were later automated with use of mechanical instruments that measured surface stiffness from the contact time of the striker. Numerous applications have evolved from this technique, from predicting rock falls to defect detection and identification in composite structures. Since operator applied tap testing requires extensive training and knowledge of the part’s structure to accurately locate defects, it is widely regarded as a subjective method and does not allow for digitised recording of results or validation against reference standards. In addition, this type of tap testing is generally confined to simple structural materials such as thin skin composites with foam or honeycomb cores, where defects can easily be identified. More complex structures with varying thicknesses present a much greater challenge for this method, as defects may have a similar response signal to thinner, non-defective regions, so neither force nor sonic data can differentiate between the two. This article seeks to introduce a novel analysis technique, applying the principle of resonant membranes to global and local frequency perspectives to generate two functions. The first sharpens tap testing images from the sonic and force responses returning greater clarity when observing the underlying structure, the second creating a local ranking of defect positions allowing an automatic highlighting of regions of high depth flux. The outcome is a process that enables operators to identify disbonds within an unknown composite structure with greater precision than either force or sound approaches on their own in lieu of prior information of defect, surface, or global resonance modes. The developed technique is suitable for application with a robotic platform to unknown curved composites surfaces since future developments will aim to achieve robotised TT deployment. The algorithm is validated within a laboratory environment on a physical reference sample, representative of an RNLI Severn class lifeboat hull, imitating a dry-dock inspection scenario.