11 - Dynamic feature evaluation and information mining for AE data stream

Abstract

Two dynamic evaluation models are developed for AE features. The acoustic emission (AE) features of composite single-lap joints under tensile loads are experimentally investigated. The Laplace score and the multicluster feature selection method are introduced and modified into a dynamic feature evaluation model suitable for streaming data. The real-time evaluation results of 15 common AE features are analyzed and the effect of damage accumulation on AE features is explored. The reliability of damage mode identification for composites based on global AE data is verified and the information contained in each AE feature is further explored and explained. Results have shown that different AE features show a certain preference for damage characterization, such as damage mode identification, damage process description, or some purpose in between. However, these AE features still fail to quantitatively characterize the remaining load-bearing capacity or the remaining useful life of the composite specimen.