ExplainabilityTransparencySafetycs.LG
Investigating the effect of CPT in lateral spreading prediction using Explainable AI
Cheng-Hsi Hsiao, Ellen Rathje, Krishna Kumar
March 17, 2025
Geotechnical Frontiers 2025
3 citations
Key Takeaways
-
AI can improve earthquake hazard prediction: The study demonstrates that machine learning models using compressed CPT (cone penetration test) data achieved over 83% accuracy in predicting lateral spreading during the 2011 Christchurch earthquake, outperforming traditional approaches.
-
Explainability reveals actionable insights: By using Explainable AI techniques, researchers identified that soil conditions between 1-3 meter depths are most critical for assessing liquefaction risk—a finding that could guide where engineers focus their investigations and resources.
-
Automation can make complex data more useful: The autoencoder approach compressed 200 detailed soil profiles into just 10 key features while preserving important information, demonstrating how AI can make large datasets more manageable without losing predictive power.
-
Responsible AI requires transparency in model decisions: The study prioritized explaining why the model made predictions (which soil depths mattered most) rather than just accepting accurate predictions as a "black box," supporting better decision-making by engineers and planners.
Abstract
This study proposes an autoencoder approach to extract latent features from cone penetration test profiles to evaluate the potential of incorporating CPT data in an AI model. We employ autoencoders to compress 200 CPT profiles of soil behavior type index (Ic) and normalized cone resistance (qc1Ncs) into ten latent features while preserving critical information. We then utilize the extracted latent features with site parameters to train XGBoost models for predicting lateral spreading occurrences in the 2011 Christchurch earthquake. Models using the latent CPT features outperformed models with conventional CPT metrics or no CPT data, achieving over 83% accuracy. Explainable AI revealed the most crucial latent feature corresponding to soil behavior between 1-3 meter depths, highlighting this depth range's criticality for liquefaction evaluation. The autoencoder approach provides an automated technique for condensing CPT profiles into informative latent features for machine-learning liquefaction models.