Exploitation
Soufi Amine; Zerradi Youssef; Soussi Mohamed; Ouadif Latifa; Bahi Anas
Abstract
The aim of this study is to thoroughly analyze the relaxation zone developing around sublevel stopes in underground mines and identify the main parameters controlling its extent. A numerical approach based on the finite element method, combined with the Hoek-Brown failure criterion, was implemented to ...
Read More
The aim of this study is to thoroughly analyze the relaxation zone developing around sublevel stopes in underground mines and identify the main parameters controlling its extent. A numerical approach based on the finite element method, combined with the Hoek-Brown failure criterion, was implemented to simulate various geometric configurations, geological conditions, and in-situ stress states. A total of 425 simulations were carried out by varying depth, horizontal-to-vertical stress ratio (k), rock mass quality (RMR), foliation orientation and spacing, as well as the height, width, and inclination of the sublevels. The results enabled the development of robust predictive models using regression analysis techniques and artificial neural networks (ANNs) to estimate the extent of the relaxation zone as a function of the different input parameters. It was demonstrated that depth and the k ratio significantly influence the extent of the relaxation zone. Additionally, a decrease in rock mass quality leads to a substantial increase in this zone. Structural characteristics, such as foliation orientation and spacing, also play a decisive role. Finally, the geometric parameters of the excavations, notably the height, width, and inclination of the sublevels, directly impact stress redistribution and the extent of the relaxation zone. The overall ANN model, taking into account all these key parameters, exhibited high accuracy with a correlation coefficient of 0.97. These predictive models offer valuable tools for optimizing the design of underground mining operations, improving operational safety, and increasing productivity.
Rock Mechanics
Anant Saini; Jitendra Singh Yadav
Abstract
The goal of this research work was to use an Artificial Neural Network (ANN) model to predict the ultimate bearing capacity of circular footing resting on recycled construction waste over loose sand. A series of plate load tests were conducted by varying the thickness of two sizes of recycled construction ...
Read More
The goal of this research work was to use an Artificial Neural Network (ANN) model to predict the ultimate bearing capacity of circular footing resting on recycled construction waste over loose sand. A series of plate load tests were conducted by varying the thickness of two sizes of recycled construction waste (5 mm and 10.6 mm) layer (0.4d, 0.6d, 0.8d, 1d, and 1.2d, d: diameter of footing) prepared at different relative densities (30%, 50%, and 70%) overlaying. The ultimate bearing capacity obtained for various combinations was used to develop the ANN model. The input parameters of the ANN model were thickness of recycled construction waste layer to diameter of circular footing ratio, angle of internal friction of sand, unit weight of sand, angle of internal friction of recycled construction waste and unit weight of recycled construction waste, and the model's output parameter was ultimate bearing capacity. The FANN-SIGMOD_SYMMETRIC model with topology 3-2-1 provided a higher estimate of the ultimate bearing capacity of circular footing, according to the ANN findings. The sensitivity analysis also revealed that the unit weight of sand and angle of internal friction of sand had insignificant effects on ultimate bearing capacity. The estimated ultimate bearing capacity was most affected by the angle of internal friction of recycled construction waste. The result of multiple linear regression analysis was not as good as the ANN model at predicting the ultimate bearing capacity.
Exploration
Abdelrahem Khalefa Embaby; Sayed Gomaa; Yehia Darwish; Samir Selim
Abstract
This study aims to develop an empirical correlation model for estimating the uranium content of the G-V in the Gabal Gattar area, northeastern desert of Egypt, as a function of the thorium content and the total gamma rays. Using the recent MATLAB software, the effect of selecting tan-sigmoid as a transfer ...
Read More
This study aims to develop an empirical correlation model for estimating the uranium content of the G-V in the Gabal Gattar area, northeastern desert of Egypt, as a function of the thorium content and the total gamma rays. Using the recent MATLAB software, the effect of selecting tan-sigmoid as a transfer function at various numbers of hidden neurons was investigated to arrive at the optimum Artificial Neural Network (ANN) model. The pure-linear function was investigated as the output function, and the Levenberg-Marquardt approach was chosen as the optimization technique. Based on 1221 datasets, a novel ANN-based empirical correlation was developed to calculate the amounts of uranium (U). The results show a wide range of uranium content, with a determination coefficient (R2) of about 0.999, a Root Mean Square Error (RMSE) equal to 0.115%, a Mean Relative Error (MRE) of -0.05%, and a Mean Absolute Relative Error (MARE) of 0.76%. Comparing the obtained results with the field investigation shows that the suggested ANN model performed well.
Areeba Qazi; Kanwarpreet Singh
Abstract
The rock mass classification system is utilized to categorize rocks, and has been used in engineering projects and stability investigations. It focuses on the parameters of rock mass and engineering applications, which include tunnels, slopes, foundations, etc. Rock mass classification is valuable in ...
Read More
The rock mass classification system is utilized to categorize rocks, and has been used in engineering projects and stability investigations. It focuses on the parameters of rock mass and engineering applications, which include tunnels, slopes, foundations, etc. Rock mass classification is valuable in the areas where the collection of samples and yielding of observation is difficult. With the advancement in technology, various machine-based model algorithms have been used, i.e., ANN and MLR in rock mass classification from prior few years. In the present work, the rock mass classification has been discussed, i.e., rock load, stand up time, RQD, RMR, Q, GSI, SMR, and RMi along with their applications. Considering all the parameters, it is concluded that for slope stability in a poor rock condition, the applicability of GSI is sufficient when compared with RMR. GSI also provides a highly accurate valuation of geo-mechanical properties, making it a valuable tool for the engineers and geologists. Also, the RMR values obtained from the ANN model provide better results for tunnels when compared with MLR and the conventional method. The ARMR classification of Slate, Shale, Quartz Schist, Gneiss, and Calcschist at 5 different locations of the world were 51-54, 66-70, 57-60, 35, 65-70, respectively. The range for slate and shale was found to be moderately anisotropic, while quartz schist, gneiss, and calcschist were found to be slightly anisotropic and highly anisotropic.
Z. Bayatzadeh Fard; F. Ghadimi; H. Fattahi
Abstract
Determining the distribution of heavy metals in groundwater is important in developing appropriate management strategies at mine sites. In this paper, the application of artificial intelligence (AI) methods to data analysis,namely artificial neural network (ANN), hybrid ANN with biogeography-based optimization ...
Read More
Determining the distribution of heavy metals in groundwater is important in developing appropriate management strategies at mine sites. In this paper, the application of artificial intelligence (AI) methods to data analysis,namely artificial neural network (ANN), hybrid ANN with biogeography-based optimization (ANN-BBO), and multi-output adaptive neural fuzzy inference system (MANFIS) to estimate the distribution of heavy metals in groundwater of Lakan lead-zinc mine is demonstrated.For this purpose, the contamination groundwater resources were determined using the existing groundwater quality monitoring data, and several models were trained and tested using the collected data to determine the optimum model that used three inputs and four outputs. A comparison between the predicted and measured data indicated that the MANFIS model had the mostpotential to estimate the distribution of heavy metals in groundwater with a high degree of accuracy and robustness.