Unsupervised Classification of Acoustic Emission Sources from Aerial Man Lift Devices

摘要

Where complicated Acoustic Emission (AE) signatures are present, (e.g. inrncases where high background noise exists, or in composite structures wherernseveral failure mechanisms have to be discriminated), conventional graphicalrnand statistical analysis may not provide the necessary resources for sourcerndiscrimination. In such cases, Unsupervised Pattern Recognition (UPR)rntechniques extend the AE user's capabilities in identifying the hiddenrnstructure and correlation of data categories in a multidimensional space. Inrnthis work, Unsupervised Pattern Recognition techniques are applied for thernanalysis and evaluation of AE data recorded during testing of five InsulatedrnAerial Man Lift devices.rnVarious types of AE sources are expected during testing of Aerial Man LiftrnDevises, arising from the fibreglass components, the metal parts of the arm,rnthe high strength pins, the welds, as well as the hydraulic systems and thernlift mechanisms. The use of pattern recognition analysis, as applied in thernpresent work, aims to identify noise sources from the mechanisms used tornmanipulate the arm movements and to discriminate signals from variousrnfailure mechanisms arising from the different materials.rnResults from different unsupervised classification schemes, applied eitherrnon the AE feature set, or to its principal component projection arernpresented. Discussion is focused on the validity of the resulting partitionsrnby using numerical optimisation criteria and common Acoustic Emissionrnpractices such as cumulative plots and emissions during load hold.rnThe proposed methodology proved efficient for the discrimination of AErnsources recorded during proof testing of Aerial Man Lift Devices and can bernused as a basis for automating the evaluation of Acoustic Emission data fromrnfuture tests of similar devices.