Time-Delay Neural Networks Reveal Pressure-Independent Fault Rupture Processes in Laboratory Acoustic Emission (Data & Code)

This archive contains code and data for analysing acoustic-emission waveforms and event catalogues from triaxial compression experiments on Alzo granite (5–40 MPa). It includes: (i) processed, per-experiment training datasets for model inputs and targets; (ii) MATLAB scripts for time-delay neural network optimisation with genetic algorithms, leave-one-experiment-out training and evaluation for stress/strain, and time-dependent feature-importance analysis; and (iii) documentation to reproduce figures and key results. MATLAB R2021b+ with Deep Learning and Global Optimization toolboxes is required. A README explains folder layout, paths, and exact commands to regenerate the main results. The companion article DOI will be added under “Related Identifiers” when available.