Decoding rate of penetration of tunnel boring machine in Deccan Traps under varied geological and machine variables using response surface analysis

Performance of TBM is significantly influenced by the ground conditions and machine variables. To achieve an optimum rate of penetration (ROP) during TBM excavation, it is important to assess the interaction between rock mass properties and machine operational/performance variables. This paper presents a systematic analysis of TBM performance based on the data collected from the MetroLine-3 UGC-01 project in Mumbai, India, and proposes a few performance prediction models for the Deccan Traps. The current work attempted to bring out the combined effect of RQD x Js as a single predictor variable while suggesting a reliable RSA modeling technique which considers the simultaneous interaction of variables. The database consisted of engineering-geological and machine variables; the selected variables were analyzed using artificial neural networks (ANN) for identifying the significant variables. Subsequently, multivariate regression (MVRA) and response surface analysis (RSA) were utilized to develop a model for predicting TBM ROP. The first model developed using MVRA as a function of rock mass variables yielded a coefficient of determination (R-2) of 0.80, whereas the second composite model developed as a function of geological and machine variables yielded an R-2 of 0.85. The third model was developed utilizing RSA which resulted in 2FI (two-factor interaction) model with improved R-2 of 0.88. Further, the best-performing RSA model accuracy is compared with the existing models and subsequently validated using new datasets and yielded an R-2 of 0.79. The developed model equation indicates that UCS, RQD x Js, and thrust variables show significant influence on the TBM ROP.

Automated summary

This paper presents a systematic analysis of TBM performance based on the data collected from the MetroLine-3 UGC-01 project in Mumbai, India, and proposes performance prediction models for the Deccan Traps. The models indicate that UCS, RQD x Js, and thrust variables have significant influence on TBM ROP.