An explainable AI-based framework for predicting and optimizing blast-induced ground vibrations in surface mining
Blast induced ground vibrations (BIGV) pose critical challenges in surface mining, threatening structural integrity, worker safety, and environmental compliance. This study proposes a novel hybrid artificial intelligence (AI) framework that integrates physics informed neural networks (PINNs) with co...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | Results in Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590123025021188 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Blast induced ground vibrations (BIGV) pose critical challenges in surface mining, threatening structural integrity, worker safety, and environmental compliance. This study proposes a novel hybrid artificial intelligence (AI) framework that integrates physics informed neural networks (PINNs) with conventional machine learning (ML) algorithms for the accurate prediction and optimization of BIGV. Unlike empirical equations that lack generalizability or black box ML models with limited transparency, the proposed approach embeds domain specific physical laws while leveraging data driven learning to improve both predictive accuracy and interpretability. A multiobjective optimization scheme is employed to balance competing goals: minimizing peak particle velocity (PPV), maximizing fragmentation efficiency, and reducing operational costs. Crucially, the framework incorporates Explainable AI (XAI) techniques such as Shapley Additive Explanations (SHAP) and Local Interpretable Model Agnostic Explanations (LIME) and uncertainty quantification (UQ) methods based on Bayesian Neural Networks to provide insight into model decisions and confidence in predictions. Validation across five operational mines in the Godavari Valley Coalfields (India) demonstrates strong generalizability, achieving up to a 20% reduction in RMSE compared to empirical baselines. The improvement is statistically significant (p<0.01) as confirmed through a paired t-test across cross-validation folds. These findings highlight that a physics informed, explainable, and uncertainty aware AI framework can substantially improve vibration prediction, ensure regulatory compliance, and support safer, more sustainable blasting operations in modern surface mining. |
|---|---|
| ISSN: | 2590-1230 |