Application of Artificial Neural Networks for Predicting the Stability of Rectangular Tunnels in Hoek–Brown Rock Masses

Abstract

An artificial neural network (ANN) model for predicting the stability of rectangular tunnels in rock masses based on the Hoek–Brown (HB) failure criterion is presented in this study. Since the safety assessment of the tunnel stability is one critical issue for civil engineers during the construction, it is very important to develop a reliable and accurate stability analysis of such problems. The finite element limit analysis (FELA) with the HB failure criterion is used to develop the numerical upper and lower bound solutions of the problem of rectangular tunnels in rock masses. A novel machine learning-aided prediction of this problem is then developed based on the datasets of the numerical bound solutions obtained from the FELA. The inputs consist of six dimensionless parameters including the cover-depth ratio of tunnels, the width ratio of tunnels, the normalized uniaxial compressive strength, the geological strength index, the mi parameter, and the degree of disturbance of rock masses. The results show that the optimal ANN models provide very great accuracy in predicting the stability of the rectangular tunnels based on the HB failure criterion. The solutions will provide a prompt assessment of tunnel stability in rock masses for geotechnical engineers during the construction of rock tunnels.

A machine learning regression approach for predicting the bearing capacity of a strip footing on rock mass under inclined and eccentric load

“…Besides, a three-noded triangular element with three undetermined stresses is employed in the lower bound (LB) analysis. The lower bound assessment aims to optimize the limit force of foundations by Yodsomjai et al, 2021a;Yodsomjai et al, 2021b;Chauhan, 2021;Keawsawasvong and Lai, 2021;Keawsawasvong et al, 2022b;Chauhan et al, 2022;Keawsawasvong et al, 2022c;Eskandarinejad, 2022).…”

“…Fortunately, together with the development of artificial intelligence, machine learning is applied in data analysis in many scientific fields, including geotechnical problems, which have significant issues in calculating data (e.g., Fernández-Cabán et al, 2018;Ghahramani et al, 2020;Bamer et al, 2021;Wu and Snaiki, 2022). Some machine learning methods, which can be considered as the successful models in geotechnical problems, are artificial neural networks ~ANN, extreme learning machines ~ELM, support vector regression ~SVR, Gaussian process regression ~GPR, and stochastic gradient boosting trees ~SGBT (e.g., Yuan et al, 2021;Keawsawasvong et al, 2022c). Nevertheless, Multivariate Adaptive Regression Splines (MARS), a curve-based machine learning method, is quite an efficient method compared to other methods (e.g., Wu and Fan, 2019;Raja and Shukla, 2021;Shiau et al, 2022).…”

Undrained stability of dual tunnels in layered soils with different strength

“…As an accurate and efficient method, FELA was used to address geotechnical stability issues 1 – 3 , 7 , 9 – 22 . FELA method can acquire accurate ultimate loads by a combination of the limit theorems of plasticity and finite elements 23 .…”

A comparative study of prediction of compressive strength of ultra‐high performance concrete using soft computing technique

“…are very effectively been used for modeling various complex engineering properties of composite materials. [22][23][24][25][26][27][28][29][30][31] An extensive review of the use of such ML techniques for forecasting the mechanical properties of concrete was carried out by Ben Chaabene et al 32 A number of studies have been conducted with the goal of predicting the mechanical properties of different types of recently developed modern concretes, including self-healing concrete, 33 recycled aggregate concrete (RCA), 34,35 high-performance concrete (HPC), 36 and ultra-HPC (UHPC). 37,38 To improve the model prediction accuracy, they used the RS algorithms optimization technique.…”

“…Various powerful ML models including tree‐based ensembles, support vector machines (SVM), artificial neural networks (ANNs), multivariate adaptive regression splines (MARS), and so forth. are very effectively been used for modeling various complex engineering properties of composite materials 22–31 . An extensive review of the use of such ML techniques for forecasting the mechanical properties of concrete was carried out by Ben Chaabene et al 32 A number of studies have been conducted with the goal of predicting the mechanical properties of different types of recently developed modern concretes, including self‐healing concrete, 33 recycled aggregate concrete (RCA), 34,35 high‐performance concrete (HPC), 36 and ultra‐HPC (UHPC) 37,38 .…”