Acoustic Emission Features from Laboratory Earthquake

This dataset supports the findings of paper  "Machine learning reveals early preseismic signals of laboratory earthquakes." Earth and Planetary Science Letters. Ying, Huang, et al. (2026)

ContentsThe .zip archive contains two folders:

acoustic_emissions/ — Example waveform samples of acoustic emissions (AE) recorded during laboratory earthquake. The original dataset is on the order of terabytes (sampling rate: 3 MHz); this folder contains a representative excerpt (load 3, sensor 23) for reference.

features/ — Tabular feature data extracted from the AE waveforms using a moving-window method. Each row corresponds to one window; column names and their definitions follow the conventions described in the associated paper.

Experimental setupLaboratory earthquake experiments were conducted on a granite sample using a biaxial loading system under controlled normal stress conditions. A dense array of 32 piezoelectric sensors recorded continuous acoustic emission at a sampling rate of 3 MHz. Multiple loading stages produced a total of 122 stick–slip events. Full details of the loading system, loading protocol, and sensor configuration are provided in the associated publication.

Feature extractionFeatures were computed from the raw AE time series using a sliding-window approach. The window parameters and detailed definitions of each feature are described in the Methods and Supporting Information of the associated paper.

CitationIf you use this dataset, please cite both this Zenodo record and the associated publication.