Developing an Explainable Artificial Intelligent (XAI) Model for Predicting Pile Driving Vibrations in Bangkok's Subsoil

Explainability matters for engineering decisions: The researchers used SHAP analysis to make their AI model's predictions transparent and interpretable, revealing that distance from the pile, hammer weight, and pile size are the key factors affecting vibrations—insights that practicing engineers can actually understand and act upon, rather than treating the model as a "black box."

AI can outperform traditional methods while being more transparent: The explainable deep learning model achieved better accuracy (MAE of 0.276) than both conventional empirical methods and other machine learning approaches, demonstrating that transparency and performance don't have to be trade-offs in responsible AI development.

Real-world data and practical deployment are essential: The model was trained on 1,018 actual pile driving measurements and deployed through a web-based application for engineers, showing that responsible AI requires bridging the gap between academic research and real-world usability rather than remaining confined to research papers.

Responsible AI includes openness and reproducibility: By making the model and source code publicly available, the researchers enabled independent verification and broader adoption, which are critical practices for building trust in AI systems used for infrastructure and environmental decisions.

AI can reveal complex patterns that improve practices: The model uncovered non-linear relationships and threshold effects in how vibrations propagate through soft clay, providing new scientific insights that could help optimize construction practices and reduce environmental impacts in urban areas like Bangkok.