Unsupervised Classification of Acoustic Emission Signal to Discriminate Composite Failure at Low Frequency

Rahman, Noor A’in A.; May, Zazilah; Mahmud, Muhamad Shazwan

doi:10.1007/978-981-16-2183-3_75

Cited by 1 publication

(2 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A challenge in AE monitoring is to establish a clear link between the recorded AE signals and the corresponding source. The possibility of identifying the signatures of damage mechanisms is a well-established field [ 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 ]. Most of the time, the analysis of AE data is established through empirical correlations between the damage mechanism and the recorded signal.…”

Section: Introductionmentioning

confidence: 99%

“…The descriptor selection helps focus on the most informative data, thus reducing complexity [ 6 ]. This is crucial for improving the classification accuracy, real-time performance, and stability in AE signal detection [ 7 ]. The methodologies based on machine learning are widely used to classify different AE sources [ 9 , 10 , 11 ].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Reduction in the Sensor Effect on Acoustic Emission Data to Create a Generalizable Library by Data Merging

Chen,

Godin,

Doitrand

et al. 2024

Sensors

View full text Add to dashboard Cite

The aim of this paper is to discuss the effect of the sensor on the acoustic emission (AE) signature and to develop a methodology to reduce the sensor effect. Pencil leads are broken on PMMA plates at different source–sensor distances, and the resulting waves are detected with different sensors. Several transducers, commonly used for acoustic emission measurements, are compared with regard to their ability to reproduce the characteristic shapes of plate waves. Their consequences for AE descriptors are discussed. Their different responses show why similar test specimens and test conditions can yield disparate results. This sensor effect will furthermore make the classification of different AE sources more difficult. In this context, a specific procedure is proposed to reduce the sensor effect and to propose an efficient selection of descriptors for data merging. Principal Component Analysis has demonstrated that using the Z-score normalized descriptor data in conjunction with the Krustal–Wallis test and identifying the outliers can help reduce the sensor effect. This procedure leads to the selection of a common descriptor set with the same distribution for all sensors. These descriptors can be merged to create a library. This result opens up new outlooks for the generalization of acoustic emission signature libraries. This aspect is a key point for the development of a database for machine learning.

show abstract