Abstract
The prediction of tunnel boring machine (TBM) performance from the rate of penetration (ROP) point of view has yet to draw a lot of attention since it is one of the main challenges for excavation with TBMs. This study examined six tunnels excavated with TBM to develop predictive models of the ROP estimation using five algorithms: Gradient Boosting (GB), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine, Adaptive Boosting (AdaBoost), and CatBoost (categorical-features-include-GB). A dataset has been developed, including uniaxial compressive strength (UCS), Rock Type, the distance between planes of weakness (DPW), and thrust force (TF), and contains more than 575 data points for each parameter. The developed models showed that the XGBoost model outperformed the other models, followed by the CatBoost, according to seven different evaluation metrics used to rank the models when the models were modeled with the default values of the algorithm parameters. After tuning the hyperparameters, the GB model outperformed the others, while the other models remained relatively unchanged. By using the overall ranking according to the metrics and considering the parameter tuning time, XGBoost and CatBoost were presented as the two best models. SHAP (Shapley additive explanations: an explainable artificial intelligence tool) values and dependency plots showed that the TF has the highest impact on the ROP, followed by UCS, Rock Type, and DPW. It is concluded that the XGBoost and CatBoost models, with coefficients of determination of 0.9878 and 0.9682, respectively, may be used for estimating the TBM penetration rate for similar rock types.
| Original language | English |
|---|---|
| Article number | 108985 |
| Journal | Engineering Applications of Artificial Intelligence |
| Volume | 136 |
| DOIs | |
| Publication status | Published - Oct 2024 |
Funding
The response to the demand has been increasing worldwide since computer-based and programming-based techniques, like artificial intelligence (AI), came to the attention of researchers and industries. That started in the early 21st century with using neural networks (NN) to predict the ROP when Grima Alvarez employed artificial neural networks (ANN) and adaptive neuro-fuzzy inference systems (ANFIS) to predict TBM performance (Grima et al., 2000). These techniques have drawn various attention where other researchers also used them to estimate TBM performance by employing different input parameters including rock type, uniaxial compressive rock strength (UCS), Brazilian tensile strength (BTS), Young's modulus, BI, DPW, α, RQD, percentage of quartz, RMR, TBM thrust and torque, joints spacing (Js) and conditions (Jc), punch slope index (PSI) refers to rock brittleness, cohesion, internal friction angle, Poisson's ratio, density, RPM, cutter torque (CT), thrust force (TF), and AR (Afradi et al., 2019; Benardos, 2008; Eftekhari et al., 2010; Gholami et al., 2012; Gholamnejad and Tayarani, 2010; Salimi and Esmaeili, 2013; Torabi et al., 2013; Yagiz et al., 2009; Zhu et al., 2021; Oraee et al., 2012); the prediction performance of such models varies from 0.69 to 0.939 from the R-squared point of view, or using other models derived based on the NN, like deep NN (DNN) (Koopialipoor et al., 2019), probabilistic NN (PNN) (Harandizadeh et al., 2021), back-propagation NN (BPNN) (Yan et al., 2023), convolutional NN (CNN) (Li et al., 2022b). Employing different algorithms by introducing evolutionary optimization methods is another use of AI in TBM performance prediction that has been utilized by various researchers. It started by using particle swarm optimization (PSO) employed by Yagiz and Karahan (2011), and went further by applying the Differential Evolution (DE) algorithm and Grey Wolf optimizer (GWO) to predict the ROP by using the parameters of DPW, α, UCS, and BI (Yagiz and Karahan, 2015). Other optimization techniques plus their combination with different algorithms have caught the researchers' attention, where Armaghani et al. used the combined methods of PSO-ANN and Imperialism Competitive Algorithm (ICA)-ANN to predict the ROP (Armaghani et al., 2017). Using optimization methods has gone further, so that various techniques, either individually or by combining with other algorithms, have been employed to develop models for estimating ROP; such methods include biogeography-based optimization (BBO), moth flame optimization (MFO), social spider optimization (SSO), and multiverse optimization (MVO) combined with extreme gradient boosting (XGBoost) (Zhou et al., 2021), and grasshopper optimization algorithm (GOA) (Akbarzadeh et al., 2022). Other AI techniques have been utilized to predict TBM performance, such as support vector machine (SVM)-based techniques such as support vector regression (SVR) (Mahdevari et al., 2014; Salimi et al., 2016; Yang et al., 2020), least-squares support vector machine (LS-SVM) (Ge et al., 2013), and finally the algorithm based on decision trees (DTs) algorithm including classification and regression tree (CART) (Salimi et al., 2019). There are some other techniques, particularly two or several algorithms called “hybrid methods,” employed to predict the ROP, which is a summary of previously developed models and their input parameters presented in Table 1.This work was supported by the Faculty Development Competitive Research Grant program of Nazarbayev University in Kazakhstan, Grant Number 021220FD5151. This work was supported by the Faculty Development Competitive Research Grant program of Nazarbayev University in Kazakhstan, Grant Number 021220FD5151.
| Funders | Funder number |
|---|---|
| BPNN | |
| CNN | |
| Coins for Alzheimer's Research Trust | |
| Office of Defense Nuclear Nonproliferation | |
| Universidade de Caxias do Sul | |
| Paul Scherrer Institut | |
| Nazarbayev University in Kazakhstan | 021220FD5151 |
UN SDGs
This output contributes to the following UN Sustainable Development Goals (SDGs)
-
SDG 4 Quality Education -
SDG 9 Industry, Innovation, and Infrastructure -
SDG 11 Sustainable Cities and Communities -
SDG 12 Responsible Consumption and Production
Keywords
- Artificial intelligence
- Gradient boosting
- Machine learning
- Rate of penetration
- Rock properties
- Tunnel boring machine
ASJC Scopus subject areas
- Control and Systems Engineering
- Artificial Intelligence
- Electrical and Electronic Engineering
Fingerprint
Dive into the research topics of 'Estimating the penetration rate of tunnel boring machines via gradient boosting algorithms'. Together they form a unique fingerprint.Cite this
- APA
- Standard
- Harvard
- Vancouver
- Author
- BIBTEX
- RIS