Abstract
When the artificial ground freezing technique is applied near existing underground structures, adfreezing behavior, characterized by ice bonding between the frozen soil and the existing structures, becomes a critical factor in assessing the stability of these structures. In this study, punch shear test data were employed to evaluate adfreezing behaviors at the frozen soil-structure interface under zero confinement conditions, representing critical states. Since machine learning (ML) algorithms have offered a powerful data-driven predictive modeling in geotechnical engineering, this study discussed the application of ML approaches to broaden the feasibility of the punch shear test for assessing the adfreezing behavior. Four ML algorithms, i.e., support vector regression (SVR), feedforward neural network (FNN), random forest (RF), and extreme gradient boosting (XGB), were adopted to develop predictive models based on the punch shear test results. To ensure optimal model performance, Bayesian optimization and five-fold cross-validation methods were employed to effectively train the ML models and identify the best hyperparameter combinations for each model. The predictive performance of these models was compared using three regression metrics: root-mean- square error (RMSE), mean absolute error (MAE), and determination coefficient (R2). The models were ranked based on their performance as follows: XGB > RF > FNN > SVR. Among them, the XGB model demonstrated the highest accuracy, with an RMSE of 0.0037, an MAE of 0.0015, and an R2 of 0.9999. The reliability and interpretability of the XGB model were further enhanced through post-hoc analysis estimating the prediction interval and SHAP values.
Key Words
adfreezing; artificial ground freezing; interfacial behavior; machine learning; post-hoc analysis; punch shear test
Address
Sangyeong Park: Departmentof Civil, Environmental and Architectural Engineering, Korea University, 145 Anam-ro, Seongbuk-gu, Seoul, 02841, Korea;
Department of Petroleum Engineering, Texas A&M University, 400 Bizzell Street, College Station, Texas, 77840, USA
Chaemin Hwang, Byeonghyun Hwang and Hangseok Choi: Departmentof Civil, Environmental and Architectural Engineering, Korea University, 145 Anam-ro, Seongbuk-gu, Seoul, 02841, Korea
Abstract
In the present study, a simplified non-associated joint model is employed to evaluate the influence of apparent friction coefficient and cohesion on earthquake response of concrete gravity dam-water-foundation rock systems. The formulation and its theoretical concepts are discussed initially. Then, a primary verification of the model is considered to control its implementation and to show its characteristics. Thereafter, the sliding behavior of Pine Flat gravity dam-water-foundation rock is studied for different ratios of foundation rock to dam elastic moduli by application of the described nonlinear joint-model. It is shown that joint opening/closing as well as sliding affects the seismic response of gravity dams significantly for rigid foundation assumption case. However, its influence reduces as foundation rock modulus decreases and radiation damping effects increases. This is observed by comparing displacement histories and envelope of principal stresses in dam body. In particular, locations and magnitudes of maximum tensile and compressive stresses occurring in the dam body are also changed when joint modeling is considered in comparison with corresponding linear cases.
Key Words
dam-water-foundation rock interaction; gravity dams; interface elements; non-associated joint model; radiation damping; sliding
Address
Vahid Lotfi and Amin Shiehnezhad: Department of Civil and Environmental Engineering, Amirkabir University of Technology, Tehran, Iran
Abstract
In this paper, the nonlinear bending of graphene platelet reinforced metal foams (GPLRMF) doubly curved shells is analyzed. Based on the Reddy' higher-order shear deformation theory, the nonlinear equations of motion are obtained by using Hamilton's principle, which are solved by utilizing assuming modal method and Newton-Raphson approach. In the end, the parameters are systematically studied to illustrate the effects of porosity coefficient, geometric imperfections, graphene platelets (GPLs) mass fraction, GPLs and pore distribution types on the nonlinear bending characteristics of GPLRMF doubly curved shells. It can be found that these parameters have a significant impact on this issue.
Key Words
doubly curved shell; metal foams; nonlinear bending; snap-buckling
Address
Gui-Lin She, Lei-Lei Gan and Jia-Qin Xu: College of Mechanical and Vehicle Engineering, Chongqing University, Chongqing 400044, China
Abstract
Earthquake-induced fault ruptures present a considerable risk to structures, especially underground systems like pile foundations. Batter pile foundations, among the various foundation types, are commonly employed for their effectiveness in withstanding inclined forces. Therefore, it is crucial to comprehensively understand how batter pile groups respond to fault ruptures under diverse geotechnical conditions to enhance geoengineering practices. In this study, 3D numerical modeling was used to investigate the internal force and damage distribution mechanisms of different batter pile groups subjected to various normal fault ruptures. Additionally, five novel machine learning regression models (i.e. Light Gradient Boosting Machine (LightGBM), CatBoost, Extreme Gradient Boosting (XGBoost), ExtraTrees, and Random Forest (RF)) were developed to learn and predict the impact of four input parameters related to batter piles and normal fault ruptures. A database comprising 375 datasets was extracted from numerical modeling results to build the learning and testing framework. The comprehensive results indicate that LightGBM has the highest potential for estimating the internal force and concrete damage distribution along batter pile foundations due to normal faults. The coefficient of determination (R2) exceeded 0.90 across all models, with reasonable values for mean square error (MSE), root mean square error (RMSE), and mean absolute error (MAE). This study provides an effective method for estimating the response of batter pile foundations to normal fault ruptures. The findings can assist engineers in designing batter pile foundations and evaluating the damage conditions of structures subjected to fault ruptures prior to detailed inspections.
Key Words
battered pile foundation; earthquake; fault rupture; machine learning
Address
Mukhtiar Ali Soomro, Zhu Ziqing, Sharafat Ali Darban and Zhen-Dong Cui: School of Mechanics and Civil Engineering, China University of Mining and Technology, Xuzhou, Jiangsu, 221116, P.R. China
Abstract
Accurately predicting rock mechanical properties, such as uniaxial compressive strength (UCS) and internal friction angle (o), is crucial for various subsurface engineering applications. Traditional laboratory testing methods for determining these parameters are often expensive and time-consuming. This research presents a novel methodology that integrates two key techniques, machine learning (ML) and geostatistics, to more efficiently and accurately estimate UCS and o from routinely measured P- and S-wave velocities (VP and VS) based on well-logging operations. The methodology involves training three machine learning models, including multivariate adaptive regression splines (MARS), least absolute shrinkage and selection operator (Lasso), and Ridge regression, on 70% of the data to predict UCS and o. Predictions were validated through cross-validation on the remaining 30% of the data. Next, the Ordinary Kriging (OK) method was employed to evaluate the accuracy and robustness of the applied methods. Finally, all the results were assessed using various metrics, including mean biased prediction error (MBPE), mean absolute prediction error (MAPE), mean squared prediction error (MSPE), and R-squared (R2). The results indicate that the Ridge model delivers the best performance for predicting o, with the lowest MSPE of 2.29 and the highest R2 of 0.98. Additionally, the value of MAPE is the lowest at 0.91, and MBPE has the lowest distance to zero. For UCS, the MARS demonstrates the lowest MSPE and MAPE values, as well as the highest R2, indicating superior performance compared to the other models.
Address
Fataneh Fakhri and Hossein Baghishani: Department of Statistics, Shahrood University of Technology, Shahrood, Iran
Danial Mansourian: Mewbourne College of Earth and Energy, Oklahoma University, USA
Ayub Elyasi: Department of Petroleum Engineering, College of Engineering, Knowledge University, Erbil 44001, Iraq
Esmael Makarian: Department of Mining Engineering, Sahand University of Technology, Tabriz 94173-71946, Iran
Fatemeh Saberi: Harold Hamm School of Geology & Geological Engineering, University of North Dakota, USA
Abstract
Rock mass classification of TBM tunnel is an important reference index to analyze the TBM performance and determine the support mode. It is of great significance for the safe and efficient construction of tunnel to quickly recognize the rock mass classes of tunnel face. In this paper, the data preprocessing is carried out by statistical method, and the input eigenvalues of model are selected through correlation analysis. A LIBSVM rock mass classification model is established, with the penetration rate, thrust and cutterhead torque as the model inputs, and the rock mass classes obtained by the Hydropower Classification (HC) method as the model output. The test results show that the average precision of LIBSVM model is 0.919, and the recall rate of Class II rock mass is as high as 0.980. The data overlap of rock mass of Class III, IV and V is an important factor affecting the precision of the model. Analyzing the importance of the 6 input eigenvalues of model in each two-classifier, the results show that the mean value of the penetration rate, the thrust and the cutterhead torque play a major role, and the standard deviations of their variation play a supplementary role.
Key Words
driving parameters; LIBSVM; Rock mass classification; Tunnel Boring Machine (TBM)
Address
Yimin Xia and Laikuang Lin: College of Mechanical and electrical Engineering, Central South University, 932 Lushan South Road, Changsha, China;
State Key Laboratory of Precision Manufacturing for Extreme Service Performance, Central South University,
932 Lushan South Road, Changsha, China
Dun Wu and Jie Ke: College of Mechanical and electrical Engineering, Central South University, 932 Lushan South Road, Changsha, China
