Advanced Machine Learning PCPT Interpretation in Offshore Geotechnical Investigations

1. Home
2. Publications
3. Online Library
4. Advanced Machine Learning PCPT Interpretation in Offshore Geotechnical Investigations

Piezocone Penetration Testing (PCPT) is the preferred technique for evaluating ground conditions in offshore geotechnical engineering. However, the large and complex datasets generated by PCPT tests pose challenges for traditional analysis methods, often relying on substantial engineering judgment. This study presents an integrated methodology leveraging both supervised and unsupervised machine learning (ML) algorithms to analyse PCPT data to rapidly define stratigraphic boundaries, characterise soil properties, and develop predictive models which can be readily utilised by engineers to develop reliable engineering design parameters. The methodology consists of pre-processing PCPT datasets including normalisation and noise reduction, followed by the identification of distinct soil layers using unsupervised clustering algorithms such as K-Means and agglomerative clustering. Dimensionality reduction techniques, notably Principal Component Analysis (PCA), are then applied to enhance data visualisation and interpretation. Supervised learning algorithms are used to characterise soil properties by clustering the datasets. Decision Trees (DT) and Random Forests (RF) are implemented for classification tasks, identifying key features that differentiate soil types. For developing characteristic equations describing soil behaviour, Linear Regression and Support Vector Regression (SVR) are applied establishing relationships between tip resistance (q_c), sleeve friction (f_s), and pore water pressure (u₂). To ensure robustness, anomaly detection techniques are integrated into the methodology. Autoencoders are then used to detect and isolate anomalies in the data, and Isolation Forests further assist in identifying outliers, ensuring data integrity for model training. Through development of this workflow, the authors demonstrate the potential of the integration of ML to aid geotechnical interpretation workflows aimed at PCPT data interpretation, thus offering a systematic approach that can significantly benefit offshore geotechnics.

5th International Symposium on Frontiers in Offshore Geotechnics (ISFOG2025)

5 - Data Analytics and Machine Learning