Research Projects

Tunneling in the Indian Himalayas has posed significant challenges over the years due to many hidden geological complexities within the mountains. Numerous tunneling projects in this region have encountered substantial delays and financial setbacks due to geohazards like water ingress, rock squeezing, spalling, rock bursts, roof collapses, and wedge failures, all triggered by unexpected geological conditions. To mechanize tunnel construction in the Indian Himalayas and achieve faster advance rates, some projects introduced tunnel boring machines (TBMs) as an alternative to the traditional drill-and-blast method (DBM). However, the use of TBMs in these conditions has generally been discouraging (with the Kishanganga hydroelectric project as a notable exception) due to several unexpected geological challenges. Although TBMs provide several advantages, they lack flexibility when unanticipated geological conditions arise along the tunnel alignment. Additionally, there remains a gap in understanding the stress-induced geohazards outlined above. Given the complexity of these issues, this research proposes a comprehensive machine learning (ML) approach using experimental and available in-situ TBM data (as documented in the literature) to reduce geological surprises during tunneling in the Indian Himalayas. The experimental setup will involve a true-triaxial loading facility and a miniature TBM to simulate various combinations of in-situ stress and geological conditions. Over the past five years, several ML models have been developed to predict geological surprises ahead of the TBM face, incorporating methods such as fuzzy logic, regression models, gene expression programming, and extreme gradient boosting (XGBoost). However, each model has been trained and validated using data from only a single project. This research addresses that limitation using physical model test data and in-situ TBM parameters to validate existing models and identify the most effective approach for the Indian Himalayas. Given the challenges associated with TBM tunneling in this region, a customized ML model will be crucial to prepare for future excavation challenges. The proposed ML model will also be instrumental in selecting the TBM type and support system for the tunnel construction in the Indian Himalayas.