Abstract:The efficiency of tunnel boring machines (TBMs) in underground projects has great significance for the mining and tunneling industries, demanding a reliable estimation of the TBM’s performance in different geotechnical conditions. The current research work attempted to suggest an optimal predictor model of TBM performance as a reliable alternative to experimental and numerical techniques. To achieve this target, three data-mining techniques, namely neural network (NN), gene expression programming (GEP), and multivariate adaptive regression splines (MARS), were employed for modelling the TBM performance, and then the most robust predictive model was optimized via a metaheuristic search method known as whale optimization algorithm (WOA). For the modelling purpose, an experimental database was compiled by performing a field assessment program in a tunneling project in Malaysia and then conducting laboratory testing on the derived rock specimens. Based on the measured experimental data, the six most influential parameters were identified and served as model inputs to predict penetration rate (PR). In order to indicate the capability of the developed GEP and NN models, a stepwise linear regression model, i.e., MARS, was designed for PR prediction as well. The predictive capacity of the constructed models was quantified using a series of statistical indices, i.e., root mean squared error (RMSE), determination coefficient (R2) and variance account for (VAF). Based on the computed indices for testing records, both the proposed GEP and NN models (with RMSE values of 0.1882 and 0.2120 and R2 values of 0.9058 and 0.8735, respectively) yielded more accurate predictive results than the MARS model with RMSE of 0.2553 and R2 of 0.8346. Hence, by achieving the most robust performance compared to the rest, GEP-based model can provide a new practical equation with a high level of accuracy. In other part of this study, the six input parameters of the GEP model and its equation were, respectively, defined as decision variables and objective function for the WOA technique to find the optimum values of PR. As a consequence of optimizing the GEP equation, the maximum value of PR rose from 3.75 m/h to 4.022 m/h, equivalent to an increase of 7.25% in PR value. The findings of this study verified the applicability of the proposed hybrid GEP and WOA approach in the site investigation phase of tunneling projects constructed by TBMs.

Evaluation and Interpretation of Blasting-Induced Tunnel Overbreak: Using Heuristic-Based Ensemble Learning and Gene Expression Programming Techniques

Optimization of Blasting Parameters Considering Both Vibration Reduction and Profile Control: A Case Study in a Mountain Hard Rock Tunnel

Multivariate regression and genetic programming for prediction of backbreak in open-pit blasting

Study on influence of key blasthole parameters on tunnel overbreak

Enhanced Wear Prediction of Tunnel Boring Machine Disc Cutters for Accurate Remaining Useful Life Estimation Using a Hybrid Model

Towards lightweight excavation: Machine learning exploration of rock size distribution prediction after tunnel blasting

A hybrid GEP and WOA approach to estimate the optimal penetration rate of TBM in granitic rock mass

Performance of Hybrid SCA-RF and HHO-RF Models for Predicting Backbreak in Open-Pit Mine Blasting Operations

A Combination of Expert-Based System and Advanced Decision-Tree Algorithms to Predict Air-Overpressure Resulting from Quarry Blasting

Estimation of the TBM advance rate under hard rock conditions using XGBoost and Bayesian optimization

Machine-Learning-Aided Determination of Post-blast Ore Boundary for Controlling Ore Loss and Dilution

Applications of Two Neuro-Based Metaheuristic Techniques in Evaluating Ground Vibration Resulting from Tunnel Blasting

Developing hybrid XGBoost model integrated with entropy weight and Bayesian optimization for predicting tunnel squeezing intensity

Enhancing ground loss rate prediction in soft-soil shield tunneling: a synergistic approach of peck back analysis and eXtreme Gradient Boosting and bayesian optimization

Research on Prediction of EPB Shield Tunneling Parameters Based on LGBM

Utilizing heuristic strategies for predicting the backbreak occurrences in open-pit mines, Gol Gohar Mine, Iran

Performance evaluation of hybrid WOA-XGBoost, GWO-XGBoost and BO-XGBoost models to predict blast-induced ground vibration

Intelligent prediction of rockburst in tunnels based on back propagation neural network integrated beetle antennae search algorithm

Predictive modeling of shallow tunnel behavior: Leveraging machine learning for maximum convergence displacement estimation

Excavation-Induced Fault Instability: A Machine Learning Perspective

Optimization Methods of Blasting Parameters of Large Cross-Section Tunnel in Horizontal Layered Rock Mass