A Reproducibility Audit of Deep Learning Papers in Structural Health Monitoring

1. Introduction

We attempted to reproduce the headline result of every deep-learning paper published in four structural health monitoring venues between 2020 and 2024 (n = 64). Eleven provided a runnable artefact; of those, four reproduced within one standard deviation of the reported metric. We report the taxonomy of failures — missing preprocessing, undisclosed hyperparameters, proprietary sensor data — and argue that the field's reported accuracy gains are not currently distinguishable from evaluation drift. We publish our audit harness and per-paper notes.

This is a demonstration full-text galley generated by DemoContentSeeder. In a live journal this HTML would be produced from the copyedited manuscript during production, or uploaded as a JATS XML rendition.

2. Method

The method section would appear here, followed by results, discussion and the reference list that is already attached to this article as structured citation records.

3. Conclusion

Open access publishing works best when the full text is genuinely readable in a browser, not only downloadable as a PDF. Both renditions are galleys in this system.