Limitations to geotechnical datasets for machine learning: the case of P-S logging
3rd Workshop on the Future of Machine Learning in Geotechnics (3FOMLIG), Florence, Italy. Paper PS01-ID22
From the introduction
In this contribution, the effects of working with a low-quality dataset are illustrated. The importance of properly accounting for the uncertainties on the data is highlighted. The dataset of choice consists of measurements of compression wave velocity Vp and shear wave velocity Vs on sediments from the Southern North Sea with P-S logging. CPT-based correlations for Vp are not widely available but have applications in seismic inversion (Dalgaard et al., 2024). A simplified analytic model is compared to a purely data-driven method to highlight the differences between both approaches.
Related project
Cite this
Stuyts, B. & Dalgaard, E. (2025). Limitations to geotechnical datasets for machine learning: the case of P-S logging. 3rd Workshop on the Future of Machine Learning in Geotechnics (3FOMLIG), Florence, Italy. Paper PS01-ID22. https://solidground.xyz/papers/2025-StuytsDalgaard-3FOMLIG-ML-PS-logging.pdf
@misc{stuyts-dalgaard-2025-ml-ps-logging,
author = {Stuyts, B. and Dalgaard, E.},
title = {Limitations to geotechnical datasets for machine learning: the case of P-S logging},
year = {2025},
howpublished = {3rd Workshop on the Future of Machine Learning in Geotechnics (3FOMLIG), Florence, Italy. Paper PS01-ID22},
url = {https://solidground.xyz/papers/2025-StuytsDalgaard-3FOMLIG-ML-PS-logging.pdf}
}