AI Geotechnical Analysis
Machine learning models that predict soil bearing capacity, settlement, liquefaction risk, and classification from borehole logs and lab test data—improving geotechnical assessment accuracy from R² of 0.68 to 0.99.
Definition
Geotechnical analysis is being transformed by machine learning. Traditional empirical methods for predicting bearing capacity achieve R² values of 0.68-0.82 with MAPE exceeding 19.6%. Multiple 2025 studies demonstrate ensemble ML models dramatically outperforming these methods: XGBoost, Gradient Boosting, Random Forest, and CatBoost achieve R² values above 0.988 and MAPE below 5.07% for shallow foundation bearing capacity prediction. SHAP analysis consistently identifies foundation depth and soil friction angle as the most influential parameters, validating that ML models learn geotechnically meaningful relationships. AI-enhanced soil classification using Random Forest within the AASHTO framework achieved 100% accuracy on 349,015 samples, maintaining 93.1%+ accuracy even when individual input variables are missing. Applications include: automated boring log interpretation, real-time ground improvement monitoring, pile capacity prediction from CPT data, and landslide and slope stability assessment.
Examples
XGBoost model predicting allowable bearing capacity for 200 foundation locations from CPT data in minutes
AI soil classification processing 3,000 borehole samples to produce a site characterization map in 4 hours instead of 2 weeks
Settlement prediction model identifying a high-risk zone beneath a planned storage tank where conventional analysis showed no concern
Nomic Use Cases
See how Nomic applies this in production AEC workflows:
Frequently Asked Questions
Geotechnical analysis is being transformed by machine learning. Traditional empirical methods for predicting bearing capacity achieve R² values of 0.68-0.82 with MAPE exceeding 19.6%. Multiple 2025 studies demonstrate ensemble ML models dramatically outperforming these methods: XGBoost, Gradient Boosting, Random Forest, and CatBoost achieve R² values above 0.988 and MAPE below 5.07% for shallow foundation bearing capacity prediction. SHAP analysis consistently identifies foundation depth and soil friction angle as the most influential parameters, validating that ML models learn geotechnically meaningful relationships. AI-enhanced soil classification using Random Forest within the AASHTO framework achieved 100% accuracy on 349,015 samples, maintaining 93.1%+ accuracy even when individual input variables are missing. Applications include: automated boring log interpretation, real-time ground improvement monitoring, pile capacity prediction from CPT data, and landslide and slope stability assessment.
XGBoost model predicting allowable bearing capacity for 200 foundation locations from CPT data in minutes. AI soil classification processing 3,000 borehole samples to produce a site characterization map in 4 hours instead of 2 weeks. Settlement prediction model identifying a high-risk zone beneath a planned storage tank where conventional analysis showed no concern.
Project Research: Instantly access all project-critical information from a single search interface.