Tu-SML-02 - Dual-Descriptor VAE Characterisation of AE-Band Structural Signals in Reinforced Concrete

Event

EUROSENSORS 2026
2026-09-06 - 2026-09-09
Zurich

Band

Poster

Chapter

Signal Analysis, Machine Learning, Physics-Informed Machine Learning And Artificial Intelligence For Sensing

Author(s)

X. Zhang - HafenCity University Hamburg,Hamburg (Germany)

Pages

698 - 699

DOI

10.5162/eurosensors2026/Tu-SML-02

ISBN

978-3-910600-12-6

Price

free

Abstract

We report an offline baseline-relative VAE study for fixed-window characterisation of continuous highfrequency structural acoustic signals from AE-band piezoelectric sensors in reinforced concrete. RMSnormalised spectra represent spectral shape, while pre-normalisation log-RMS preserves amplituderelated energy. In a bending test, principal load cycles were dominated by energy increases, with weaker spectral-shape shifts, latent-space drift, and broader latent dispersion during final load-to-failure. Results support unsupervised descriptor-based continuous signal characterisation under varying load.