No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Evaluating blast-induced backbreak in open pit mines using the LSSVM optimized by the GWO algorithm
Abstract
Backbreak, a recurring issue in blasting operations, causes mine wall instability, equipment failure, inappropriate disintegration, lower drilling efficiency, and increased cost of mining operations. This study aims to address these issues by developing a hybrid LSSVM-GWO model for predicting blast-induced backbreak in open pit mines. To evaluate the effectiveness of the proposed model, its predictive performance was compared with three convolutional models, such as the support vector machine, K-nearest neighbor, and the least square support vector machine. Results demonstrated that the LSSVM-GWO model outperformed the other three models, achieving coefficient of determination values of 0.998 and 0.997, mean absolute error values of 0.0068 and 0.1209, root mean squared error values of 0.0825 and 0.1936, and a20-index values of 0.99 and 1.01 for training and testing datasets, respectively. Furthermore, the SHAP machine learning technique was applied to evaluate the feature importance, revealing that the powder factor had the highest influence, while the burden exhibited the least impact on backbreak. Sensitivity analysis confirmed these findings, highlighting the robustness of the hybrid model. The study concludes that the LSSVM-GWO model significantly enhances the prediction and evaluation of backbreak in open pit mines, providing critical insights to improve blasting operations, reduce costs, and ensure mine safety.
... Moreover, Shahani et al. [65,66,69] developed four gradient boosting machine learning algorithms, namely, gradient boosted regression (GBR), Catboost, light gradient boosting machine (LightGBM), and extreme gradient boosting (XGBoost), to predict the UCS of soft sedimentary rocks of the Block-IX at Thar Coalfield, Pakistan. Wet density, moisture, dry density, and Brazilian tensile strength have been used as input variables and 106-point dataset was allocated identically for each algorithm with a ratio of 70/30 for the training and testing phases respectively. ...
... Predictions of static and dynamic Young's moduli of intact sedimentary rocks, based on four inputs such as uniaxial compressive strength; density; p-wave velocity; and s-wave velocity have been carried out by [65,66,69]. The authors developed a novel machine learning regression model, namely an extreme gradient boosting (XGBoost) and results have been compared with those obtained from multiple linear regression (MLR) and artificial neural network (ANN) approaches. ...
... Related RMSE and MAE obtained from both methods are relatively low, with values of 0.0794 m/h and 0.254 m/h for RMSE, and 0.0123 m/h and 0.041 m/h for MAE respectively. In the paper of [65,66,69], six novel machine learning (ML)based intelligent regression models were developed to predict the impacts of wet density, moisture, dry density, and Brazilian tensile strength on elastic modulus. A ratio of 70/30 has been employed for training and testing on each model while fundamental statistical investigation tools have been used to categorize the most dominant and important input parameters. ...
The city’s population growth in developing countries over the past two decades led to a sharp increase in water supply needs. This generates proliferation of drilling companies and a highly competitive environment, in Cameroon in particular. A regular optimization of drilling operations in geological formations appears to be crucial and urgent, so as to reduce drilling costs since the drilling equipment used is too expensive and sometimes scarce. The present paper investigated an accurate machine learning model for the penetration rates (ROP) prediction in lateritic soil covers layers. The present study investigates various machine learning techniques including the linear regression, K-Nearest Neighbors, ridge regression and Random Forest, to predict the penetration rate. Data from four (04) defined parameters: the percussion pressure, Pp (MPa); the blowing pressure, Ps (MPa); the pressure of compressor, Pc (MPa); the rotation speed, Vr (tr/min) have been used to build the dataset, for training and validation tests, with an 70/30 ratio. The drilling time and drilling depth ranges from 0.3 to 2.8 h, and from 0.85 to 4.6 m respectively, for a constant rotation speed of 1350 tr/min while values of pressures range between 2.5 and 294.3 MPa. The key performance metrics including correlation coefficient (R2), mean absolute error (MAE) and root mean square error (RMSE) were calculated for each method to evaluate the accuracy of the predictions. Results show that the Random Forest model exhibited the best accuracy, with a R2 value of 0.999, with RMSE and MAE values of 0.0225 and 0.0121 respectively. A relatively high accuracy has been obtained for the K-nearest neighbors method with R2, RMSE and MAE values of 0.933, 0.1547 and 0.0802 respectively. Relatively low to values of metrics have also been obtained for linear regression and ridge regression methods. Related R2, RMSE, MAE values obtained are respectively 0.8612, 0.2232 and 0.1724 for the linear regression, and 0.8447, 0.2361 and 0.1894 for the ridge regression method. The discussion of the obtained results shows that, the Rain Forest model predicts the ROP value with a highly good accuracy, and can thus greatly contribute in reducing the costs and time related to drilling operations in lateritic soil covers context.
... This study builds upon the work of (Shahani et al. 2024a) on the LSSVM model and GWO-LSSVM by incorporating GA-LSSVM and PSO-LSSVM for more comprehensive analysis of the dataset. In addition, it aims to further optimize the LSSVM model to achieve improved accuracy for future applications. ...
... Several researchers (Ghasemi et al. 2016;Sharma et al. 2021) have predicted backbreak by applying different machine learning models, i.e., NLMR, GP, RT, ANFIS, and MRVA, using the same dataset, which can be found in the online literature. Building upon Shahani et al. (2024a) work, this study incorporates the GA-LSSVM and PSO-LSSVM models and performs a more comprehensive analysis. ...
Backbreak is an undesirable outcome in blasting operations caused by factors such as equipment failure, improper fragmentation, unstable mine walls, reduced drilling efficiency, and other issues that contribute to increased mining operation costs. To overcome these problems effectively, this study developed a least square support vector machine (LSSVM) model and optimized it using metaheuristic algorithms, including genetic algorithm (GA)-LSSVM, particle swarm optimization (PSO)-LSSVM, and grey wolf optimization (GWO)-LSSVM, to predict the efficiency and accuracy of backbreak due to blasting in surface mines using burden (m), spacing (m), stemming (m), powder factor (kg/ms), and stiffness ratio (m/m) as input parameters. Among the models evaluated, the GWO-LSSVM model demonstrated superior performance compared to the LSSVM, GA-LSSVM, and PSO-LSSVM models, achieving a coefficient of determination of 0.998 and 0.997, mean absolute error of 0.0068 and 0.1209, root mean squared error of 0.0825 and 0.1936, and SI of 0.021 and 0.044 on the training and testing datasets, respectively. Sensitivity analysis of the GWO-LSSVM model revealed that the powder factor exerted the most significant influence, while the burden had the least impact on backbreak. This developed method has proven to significantly enhance the performance evaluation of backbreak in surface mines and offers valuable engineering applications in mining and other related fields.
Geo-engineering problems are known for their complexity and high uncertainty levels, requiring precise definitions, past experiences, logical reasoning, mathematical analysis, and practical insight to address them effectively. Soft Computing (SC) methods have gained popularity in engineering disciplines such as mining and civil engineering due to computer hardware and machine learning advancements. Unlike traditional hard computing approaches, SC models use soft values and fuzzy sets to navigate uncertain environments. This study focuses on the application of SC methods to predict backbreak, a common issue in blasting operations within mining and civil projects. Backbreak, which refers to the unintended fracturing of rock beyond the desired blast perimeter, can significantly impact project timelines and costs. This study aims to explore how SC methods can be effectively employed to anticipate and mitigate the undesirable consequences of blasting operations, specifically focusing on backbreak prediction. The research explores the complexities of backbreak prediction and highlights the potential benefits of utilizing SC methods to address this challenging issue in geo-engineering projects.
Geopolymer concrete is an eco-efficient and environmentally friendly construction material. Various ashes were used as the binder in geopolymer concrete, such as fly ash, ground granulated blast furnace slag, rice husk ash, metakaolin ash, and Palm oil fuel ash. Fly ash was commonly consumed to prepare geopolymer concrete composites. It is essential to have 28 days resting period of the concrete to attain compressive strength in the structural design. In the present investigation, several soft computing models were employed to form the predictive models for forecasting the compressive strength of ground granulated blast furnace slag (GGBFS) concrete. A complete dataset of 268 samples was extracted from published research articles and analyzed to establish models. The modeling process incorporated seven effective parameters such as water content (W), temperature (T), water to binder ratio (w/b), ground granulated blast furnace slag to binder ratio (GGBFS/b), fine aggregate (FA) content, coarse aggregate (CA) content, and the superplasticizer dosage (SP) that were examined and measured on the compressive strength of GGBFS concrete by utilizing various modeling techniques, viz., Linear Regression (LR), Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Support Vector Regression (SVR), Grey Wolf Optimization (GWO), Differential Evolution (DE), and Mantra Rays Foraging Optimization (MRFO). The compressive strength of the training datasets was predicted using the SVR-PSO and SVR-GWO models, with a reliable coefficient of correlation of 0.9765 and 0.9522, respectively.
Elastic modulus (E) is a key parameter in predicting the ability of a material to withstand pressure and plays a critical role in the design of rock engineering projects. E has broad applications in the stability of structures in mining, petroleum, geotechnical engineering, etc. E can be determined directly by conducting laboratory tests, which are time consuming, and require high-quality core samples and costly modern instruments. Thus, devising an indirect estimation method of E has promising prospects. In this study, six novel machine learning (ML)-based intelligent regression models, namely, light gradient boosting machine (LightGBM), support vector machine (SVM), Catboost, gradient boosted tree regressor (GBRT), random forest (RF), and extreme gradient boosting (XGBoost), were developed to predict the impacts of four input parameters, namely, wet density (ρwet) in gm/cm³, moisture (%), dry density (ρd) in gm/cm³, and Brazilian tensile strength (BTS) in MPa on output E (GPa). The associated strengths of every input and output were systematically measured employing a series of fundamental statistical investigation tools to categorize the most dominant and important input parameters. The actual dataset of E was split as 70% for the training and 30% for the testing for each model. In order to enhance the performance of each developed model, an iterative 5-fold cross-validation method was used. Therefore, based on the results of the study, the XGBoost model outperformed the other developed models with a higher accuracy, coefficient of determination (R² = 0.999), mean absolute error (MAE = 0.0015), mean square error (MSE = 0.0008), root mean square error (RMSE = 0.0089), and a20-index = 0.996 of the test data. In addition, GBRT and RF have also shown high accuracy in predicting E with R² values of 0.988 and 0.989, respectively, but they can be used conditionally. Based on sensitivity analysis, all parameters were positively correlated, while BTS was the most influential parameter in predicting E. Using an ML-based intelligent approach, this study was able to provide alternative elucidations for predicting E with appropriate accuracy and run time at Thar coalfield, Pakistan.
Drilling rate index (DRI) is a fundamental parameter in the investigation of rock drillability, as drillability is considered one of the
main problems in rock engineering. Several researchers have continuously tried to analyze and correlate rock DRI, but the
problem remains unchanged. This study elucidates the machine
learning approaches, namely long short term memory (LSTM),
simple recurrent neural network (RNN) and random forest algorithm
(RFA) to predict DRI of rocks using multivariate inputs, that is, uniaxial compressive strength in MPa; Brazilian tensile strength (BTS) in MPa; brittleness value (S20); Sievers’ J value (Sj); modulus ratio (MR); shore hardness (SH), porosity (n) in %; shimazeks F abrasitivity in N/mm; and equivalent quartz content in %. For all proposed methods, the original dataset was divided into 70% for training and the remaining 30% for testing. Next, the performance indices, such as correlation coefficient (R2), root mean square error (RMSE), variance accounts for (VAF) and a-20 index of each proposed method were determined to examine the accuracy of the predicted data. In this study, according
to the results of LSTM, simple RNN and RFA methods, the LSTM revealed the best prediction output for DRI with the strongest R2, the lowest RMSE, the largest VAF and an appropriate a-20 index values as 0.999, 0.13416, 0.997, and 0.999 in the training stage and 0.998, 0.19479, 0.996, and 0.997 in the testing stage, respectively. Therefore, LSTM is an applicable machine learning approach that can be applied to accurately predict the DRI.
Young’s modulus (E) is essential for predicting the behavior of materials under stress and plays an important role in the stability of surface and subsurface structures. E has a wide range of applications in mining, geology, civil engineering, etc.; for example, coal and metal mines, tunnels, foundations, slopes, bridges, buildings, drilling, etc. This study developed a novel machine learning regression model, namely an extreme gradient boosting (XGBoost) to predict the influences of four inputs such as uniaxial compressive strength in MPa; density in g/cm³; p-wave velocity (Vp) in m/s; and s-wave velocity in m/s on two outputs, namely static Young’s modulus (Es) in GPa; and dynamic Young’s modulus (Ed) in GPa. Using a series of basic statistical analysis tools, the accompanying strengths of each input and each output were systematically examined to classify the most prevailing and significant input parameters. Then, two other models i.e., multiple linear regression (MLR) and artificial neural network (ANN) were employed to predict Es and Ed. Next, multiple linear regression and ANN were compared with XGBoost. The original dataset was allocated as 70% for the training stage and 30% for the testing stage for each model. To improve the performance of the developed models, an iterative 10-fold cross-validation method was used. Therefore, based on the results XGBoost model has revealed the best performance with high accuracy (Es: correlation coefficient (R²) = 0.998; Ed: R² = 0.999 in the training stage; Es: R² = 0.997; Ed: R² = 0.999 in the testing stage), root mean square error (RMSE) (Es: RMSE = 0.0652; Ed: RMSE = 0.0062 in the training stage; Es: RMSE = 0.071; Ed: RMSE = 0.027 in the testing stage), RMSE-standard deviation ratio (RSR) index value (Es: RSR = 0.00238; Ed: RSR = 0.00023 in the training stage; Es: RSR = 0.00304; Ed: RSR = 0.001 in the testing stage) and variance accounts for (VAF) (Es: VAF = 99.71; Ed: VAF = 99.99 in the training stage; Es: VAF = 99.83; Ed: VAF = 99.94 in the testing stage) compared to the other developed models in this study. Using a novel machine learning approach, this study was able to deliver substitute elucidations for predicting Es and Ed parameters with suitable accuracy and runtime.
In bench blasting, backbreak is the unwanted result that causes instability to the highwall and can lead to safety hazards. Hence, it is utmost necessary to minimize the generation of backbreak to improve mine’s safety. It has always been difficult to predict the backbreak because of various parameters involved, i.e. blast design, explosive properties, rock mass, etc. In this study, multivariate regression analysis (MVRA) and genetic programming (GP) techniques were performed on 70 blast data sets of previously published papers. Both the models have been developed and tested with the same mine data set. For validation of the models, a total of 14 trial blasts have been conducted in Indian coal mines with different geological strata and other parameters. The values of R2, RMSE, MAPE and prediction level at 25% and 90% were computed for GP and MVRA techniques. Also, the GP model is compared with the other state-of-the-art techniques. It has been found that the level of prediction for validation data set at 25% using GP is 78.57% and for MVRA is 21.42%. The mean magnitude of relative error (MMRE) value for GP and MVRA is 0.18 and 0.42, respectively. The results show that the GP is a more efficient tool for prediction of backbreak in comparison with MVRA. On performing sensitivity analysis, it has been found that stemming length and powder factor are the most influencing parameters to backbreak.
Liquefaction has caused many catastrophes during earthquakes in the past. When an earthquake is occurring, saturated granular soils may be subjected to the liquefaction phenomenon that can result in significant hazards. Therefore, a valid and reliable prediction of soil liquefaction potential is of high importance, especially when designing civil engineering projects. This study developed the least squares support vector machine (LSSVM) and radial basis function neural network (RBFNN) in combination with the optimization algorithms, i.e., the grey wolves optimization (GWO), differential evolution (DE), and genetic algorithm (GA) to predict the soil liquefaction potential. Afterwards, statistical scores such as root mean square error were applied to evaluate the developed models. The computational results showed that the proposed RBFNN-GWO and LSSVM-GWO, with Coefficient of Determination (R²) = 1 and Root Mean Square Error (RMSE) = 0, produced better results than other models proposed previously in the literature for the prediction of the soil liquefaction potential. It is an efficient and effective alternative for the soil liquefaction potential prediction. Furthermore, the results of this study confirmed the effectiveness of the GWO algorithm in training the RBFNN and LSSVM models. According to sensitivity analysis results, the cyclic stress ratio was also found as the most effective parameter on the soil liquefaction in the studied case.
Circular failure can be seen in weak rocks, the slope of soil, mine dump, and highly jointed rock mass. The challenging issue is to accurately predict the safety factor (SF) and the behavior of slopes. The aim of this study is to offer advanced and accurate models to predict the SF of slopes through machine learning methods improved by optimization algorithms. To this view, three different methods, i.e., trial and error (TE) method, gravitational search algorithm (GSA), and whale optimization algorithm (WOA) were used to investigate the proper control parameters of least squares support vector machine (LSSVM) method. In the constructed LSSVM-TE, LSSVM-GSA and LSSVM-WOA methods, six effective parameters on the SF, such as pore pressure ratio and angle of internal friction, were used as the input parameters. The results of the error criteria indicated that both GSA and WOA can improve the performance prediction of the LSSVM method in predicting the SF. However, the LSSVM-WOA method, with root mean square error of 0.141, performed better than the LSSVM-GSA with root mean square error of 0.170.
This study proposes a new uncertain rule-based fuzzy approach for the evaluation of blast-induced backbreak. The proposed approach is based on rock engineering systems (RES) updated by the fuzzy system. Additionally, a genetic algorithm (GA) and imperialist competitive algorithm (ICA) were employed for the prediction aim. The most key step in modeling of fuzzy RES is the coding of the interaction matrix. This matrix is responsible for analyzing the interrelationships among the parameters influencing the rock engineering activities. The codes of the interaction matrix are not unique; thus, probabilistic coding can be done non-deterministically, which allows the uncertainties to be considered in the RES analysis. To achieve the objective of this research, 62 blasts in Shur River dam region, located in south of Iran, were investigated and the required datasets were measured. The performance of the proposed models was then evaluated in accordance with the statistical criteria such as coefficient of determination (R2). The results signify the effectiveness of the proposed GA- and ICA-based models in the simulating process. R2 of 0.963 and 0.934 obtained from ICA- and GA-based models, respectively, revealed that both models were capable of predicting the backbreak. Further, the fuzzy RES was introduced as a powerful uncertain approach to evaluate and predict the backbreak.
Back break is an unsolicited phenomenon caused due to rock condition, blast geometry, explosive and initiation system in mines. It does not help in creating a smooth high wall and free face for next blasting due to cracks, overhang and under-hang. It can cause rockfall during drilling due to the cracks present in the in situ rock mass at the perimeter. Due to improper free face created from the previous blast and the presence of loose strata in the face increases the overall cost of production. Therefore, predicting and subsequently optimising back break shall reduce their problems to some extent. In this paper, an attempt is made to predict back break using the random forest method. The variables used for the study was such as burden to spacing ratio, stemming to hole-depth ratio, p-wave velocity and the density of explosive. For the random forest model, R2 0.9791 and RMSE 0.87899 and for linear regression was R2 was 0.824 and root mean square error (RMSE) 0.72, respectively. From the field trials, it was evident that the use of low-density emulsion can help in reducing the back break and optimise the overall cost of the blasting process. The same results were validated using Random forest method wherein the model R2 was 0.9791 and RMSE was 0.8799.
The geomechanical properties of rock, including shear strength (SS) and uniaxial compressive strength (UCS), are very important parameters in designing rock structures. To improve the accuracy of SS and UCS prediction, this study presented an evolving support vector regression (SVR) using Grey Wolf optimization (GWO). To examine the feasibility and applicability of the SVR-GWO model, the differential evolution (DE) and artificial bee colony (ABC) algorithms were also used. In other words, the SVR hyperparameters were tuned using the GWO, DE, and ABC algorithms. To implement the proposed models, a comprehensive database accessible in an open-source was used in this study. Finally, the comparative experiments such as root mean square error (RMSE) were conducted to show the superiority of the proposed models. The SVR-GWO model predicted the SS and UCS with RMSE of 0.460 and 3.208, respectively, while, the SVR-DE model predicted the SS and UCS with RMSE of 0.542 and 5.4, respectively. Furthermore, the SVR-ABC model predicted the SS and UCS with RMSE of 0.855 and 5.033, respectively. The aforementioned results clearly exhibited the applicability as well as the usability of the proposed SVR-GWO model in the prediction of both SS and UCS parameters. Accordingly, the SVR-GWO model can be also applied to solving various complex systems, especially in geotechnical and mining fields.
In general, trapezoidal channels are used in irrigation and drainage networks. When installing a side weir on the side wall of a trapezoidal channel, as excess water reaches the side weir plane, additional flow from the crest of the side weir is driven into the side channel. The main aim of this study is to predict the discharge coefficient of rectangular side weirs located on trapezoidal channels using support vector machines (SVMs). Based on the effective parameters on the discharge coefficient of side weirs in trapezoidal channels, six different models (SVM 1–SVM 6) are introduced. According to the analysis results of SVM 1–SVM 6 models, the superior model is introduced as a function of the Froude number (Fr), ratio of side weir length to the bottom width of a trapezoidal channel (L/b), ratio of side weir length to the flow depth upstream of the weir (L/y1), side slope of the trapezoidal channel (m) and ratio of flow depth upstream of the weir to the trapezoidal channel bottom width (y1/b). Based on the simulation results, the superior model has a reasonable accuracy. For example, the root mean square error, mean absolute relative error and correlation coefficient (R²) values calculated for the superior training model are 0.0156, 0.0327 and 0.884, respectively. Furthermore, the ratio of side weir length to trapezoidal channel bottom width (L/b) is identified as the most effective input parameter for modeling discharge coefficient. Additionally, a matrix is presented for superior model to estimate discharge coefficient of the side weirs.
Accurate short-term traffic flow prediction plays an indispensable role for solving traffic congestion. However, the structure of traffic data is nonlinear and complicated. It is a challenge to get high precision. The least square support vector machine (LSSVM) has powerful capabilities for time series and nonlinear regression prediction problems if it can select appropriate parameters. To search the optimal parameters of LSSVM, this paper proposes a hybrid optimization algorithm which combines particle swarm optimization (PSO) with genetic algorithm. The main contributions are twofold: (1) A hybrid optimization method is proposed, which can skip the local optimal pitfall with less learning time by introducing a selection strategy, crossover and mutation operators into PSO; (2) the crossover and mutation operators are controlled by adaptive probability functions. The crossover and mutation probabilities increase when the population fitness is concentrated, and decrease when the fitness is dispersed. It can effectively improve the precision and speed of convergence. The proposed model is verified based on the measured data. The experimental results show that our new model yields better prediction ability and relatively high computational efficiency compared with other related models.
This work proposes a new evolutionary multilayer perceptron neural networks using the recently proposed Bird Swarm Algorithm. The problem of finding the optimal connection weights and neuron biases is first formulated as a minimization problem with mean square error as the objective function. The BSA is then used to estimate the global optimum for this problem. A comprehensive comparative study is conducted using 13 classification datasets, three function approximation datasets, and one real-world case study (Tennessee Eastman chemical reactor problem) to benchmark the performance of the proposed evolutionary neural network. The results are compared with well-regarded conventional and evolutionary trainers and show that the proposed method provides very competitive results. The paper also considers a deep analysis of the results, revealing the flexibility, robustness, and reliability of the proposed trainer when applied to different datasets.
Due to the high degree of strong coupling and nonlinearity of marine lysozyme fermentation process, it is difficult to accurately model the mechanism. In order to achieve real‐time online measurement and effective control of bacterial concentration during fermentation, a generalized predictive control method based on least squares support vector machines is proposed. The particle swarm optimization least squares support vector machine (PSO‐LS‐SVM) model of lysozyme concentration is established by optimizing the regularization parameters and the kernel parameters of the least squares support vector machine by particle swarm optimization. To avoid the nonlinear problems in predictive control, the model is linearized at each sampling point and the generalized predictive algorithm is used to predict the bacteria concentration of lysozyme. The experimental simulation shows that the least squares support vector machine model with particle swarm optimization can achieve good prediction effect. The linearized model performs generalized predictive control, which makes the total activity of the enzyme increased from 60% to 80% and the yield improved by 30%. In order to achieve real‐time online measurement and effective control of bacteria concentration during fermentation, a generalized predictive control method based on least squares support vector machines is proposed. The particle swarm optimization least squares support vector machine (PSO‐LS‐SVM) model of bacteria concentration is established by optimizing the regularization parameters and the nuclear parameters of the least squares support vector machine by particle swarm optimization.
Unconfined compressive strength (UCS) of rock is a key parameter in drilling and stimulation of oil and gas wells such as wellbore stability and fracturing operations. Better estimates of UCS could increase the efficiency of drilling and stimulation operations. Current techniques for UCS determination either rely on laboratory measurements or empirical relationships using well logs. The laboratory measurements, although more reliable, are expensive and time-consuming. On the other hand, the empirical relations derived from core data and well logs are very unique because they are developed for a specific rock type; hence has limited applications. This paper presents a data-driven model for UCS prediction from petrophysical properties and elemental spectroscopy using artificial intelligent technique, namely support-vector regression (SVR).
Elemental spectroscopy, density, porosity, and UCS data presented in this paper are based on various geological formations. UCS is determined by the uniaxial compression test in the laboratory, while elemental spectroscopy was obtained from X-ray fluorescence (XRF) analysis. We first use SVR to establish a correlation between elemental spectroscopy, density, and porosity with UCS. We separate these data into two categories: training and testing data. Training data is used to train SVR and establishes the UCS prediction model. The model will generate UCS prediction using testing data and compared with the laboratory-measured UCS.
In total, 21 cases were run with different combination of input parameters. Good agreement was observed between the SVR-predicted UCS and the laboratory measurement. Two quantitative measures for estimation accuracy are calculated and examined including the coefficient of determination and the mean absolute percentage error. Considering limited number of available data used in this study, the SVR-predicted UCS produces very good coefficient of determination and small error. The results also demonstrate the significant influence of elemental spectroscopy on the UCS prediction because elements determine grain density, which contributes to the rock strength. This emphasizes the advantage of incorporating elemental spectroscopy, together with other petrophysical properties, for UCS prediction. The favorable results in this study demonstrate the promising capability of SVR to build a UCS-prediction model based on elemental spectroscopy and petrophysical properties. Further application of SVR can be adapted to predict UCS directly from mineralogy logs and conventional well logs.
Drilling and blasting is the predominant rock excavation method in mining and tunneling projects. Back-break (BB) is one of the most undesirable by-products of blasting and causing rock mine wall instability, increasing blasting cost as well as decreasing performance of the blasting. In this research work, a practical new hybrid model to predict the blast-induced BB is proposed. The model is based on particle swarm optimization (PSO) in combination with adaptive neuro-fuzzy inference system (ANFIS). In this regard, a database including 80 datasets was collected from blasting operations of the Shur river dam region, Iran, and the values of four effective parameters on BB, i.e., burden, spacing, stemming and powder factor were precisely measured. In addition, the values of the BB for the whole 80 blasting events were measured. The accuracy of the proposed PSO-ANFIS model was also compared with the multiple linear regression (MLR). Median absolute error, coefficient of determination (R²) and root mean squared error, as three statistical functions, were used to evaluate the performance of the predictive models. The results achieved indicate that the PSO-ANFIS model has strong potential to indirect prediction of BB with high degree of accuracy. The R² equal to 0.922 suggests the superiority of the PSO-ANFIS model in predicting BB, while this value was obtained as 0.857 for MLR model.
Uniaxial compressive strength (UCS) is one of the most widely used and important rock mechanical parameters in rock engineering. The main objective of the present study was to evaluate the ability of artificial intelligence models including Multi-layer Perceptron (MLP), Sugeno Fuzzy Logic (SFL), Mamdani Fuzzy Logic (MFL), Adaptive Neuro-Fuzzy Inference System (ANFIS) and Support Vector Machine (SVM) to predict the UCS of travertine rocks in the Azarshahr area (NW Iran). To attempt this objective, 85 core samples of travertine rock were collected from the study area and the laboratory tests were performed to determine the P-wave velocity [Vp (km/s)], porosity (n %), Schmidt rebound hardness (Rn) and UCS of the rocks at the Rock Mechanics Laboratory in the Tarbiat Modares University. The data set including Vp (km/s), n % and Rn as the inputs and UCS as the output were divided into training (80% of dataset) and testing (20% of dataset) subsets to construct the models. The coefficient of determination (R2), root mean square error (RMSE) and mean absolute error (MAE) were used to evaluate the models performance. The models accuracy followed the order SVM > ANFIS > SFL > MLP > MFL. The SVM model with RBF kernel function yielded the highest R2 (0.9516), and the lowest RMSE (2.14 MPa) and MAE (1.351 MPa) in the testing step. Accuracy results indicated that SVM model predictions were better than MLP, SFL, MFL and ANFIS models for prediction of UCS of travertine rocks.
This paper employs the recently proposed nature-inspired algorithm called Multi-Verse Optimizer (MVO) for training the Multi-layer Perceptron (MLP) neural network. The new training approach is benchmarked and evaluated using nine different bio-medical datasets selected from the UCI machine learning repository. The results are compared to five classical and recent evolutionary metaheuristic algorithms: Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Differential Evolution (DE), FireFly (FF) Algorithm and Cuckoo Search (CS). In addition, the results are compared with two well-regarded conventional gradient-based training methods: the conventional Back-Propagation (BP) and the Levenberg-Marquardt (LM) algorithms. The comparative study demonstrates that MVO is very competitive and outperforms other training algorithms in the majority of datasets in terms of improved local optima avoidance and convergence speed.
One of the most challenging safety problems in open pit mines is backbreak during blasting operation, and its prediction is very important for a technically and economically successful mining operation. This paper presents application of particle swarm optimization (PSO) technique to estimate the backbreak induced by bench blasting, based on major controllable blasting parameters. Two forms of PSO models, linear and quadratic, are developed based on blasting data from Sungun copper mine, Iran. According to obtained results, both models can be used to predict the backbreak, but the comparison of two models, in terms of statistical performance indices, shows that the quadratic form provides better results than the linear form.
Blasting is the process of use of explosives to excavate or remove the rock mass. The main objective of blasting operation is to provide proper rock fragmentation and to avoid undesirable environmental impacts such as ground vibration, flyrock and back-break. Therefore, proper predicting and subsequently optimizing these impacts may reduce damage on facilities and equipment. In this study, an artificial neural network (ANN) was developed to predict flyrock and back-break resulting from blasting. To do this, 97 blasting works in Delkan iron mine, Iran, were investigated and required blasting parameters were collected. The most influential parameters on flyrock and back-break, i.e. burden, spacing, hole length, stemming, and powder factor were considered as model inputs. Results of absolute error (Ea) and root mean square error (RMSE) (0.0137 and 0.063 for Ea and RMSE, respectively) reveal that ANN as a powerful tool can predict flyrock and back-break with high degree of accuracy. In addition, this paper presents a new metaheuristic approximation approach based on the ant colony optimization (ACO) for solving the problem of flyrock and back-break in Delkan iron mine. Considering changeable parameters of the ACO algorithm, blasting pattern parameters were optimized to minimize results of flyrock and back-break. Eventually, implementing ACO algorithm, reductions of 61 and 58 % were observed in flyrock and back-break results, respectively.
The prediction of blasting vibration characteristic parameters is very important to evaluate the situations of blasting vibration damage. Blasting vibration of rock mass is affected by lots of characteristics, such as charging parameter, rock type and geological topography. The characteristics should be comprehensively considered in order to accurately predict the blasting vibration. Based on training and testing 93 sets of measured data in an open-pit mine, Support Vector Machine (SVM) and Random Forest (RF) methods are applied to predict the peak particle velocity (PPV), first dominant frequency and duration time of first dominant frequency of blasting vibration. The other 15 groups of measured data are tested as forecast samples, of which the predicted results are consistent with the measured ones. Results show that the prediction accuracies of SVM and RF models were acceptable. The average error rate of SVM is lower than results using RF, and the weight of factors is determined using RF. It is a new approach to predict destructive effect on housing under blasting vibration using SVM and RF, which can be applied to practical engineering.
In blasting works, the aim is to provide proper rock fragmentation and to avoid undesirable environmental impacts such as back-break. Therefore, predicting fragmentation and back-break is a significant step in achieving a technically and economically successful outcome. In this paper, considering the robustness of artificial intelligence methods utilized in engineering problems, an artificial neural network (ANN) was applied to predict rock fragmentation and back-break; an artificial bee colony (ABC) algorithm was also utilized to optimize the blasting pattern parameters. In this regard, blasting parameters, including burden, spacing, stemming length, hole length and powder factor, as well as back-break and fragmentation were collected at the Anguran mine in Iran. Root mean square error (RMSE) values equal to 2.76 and 0.53 for rock fragmentation and back-break, respectively, reveal the high reliability of the ANN model. In addition, ABC algorithm results suggest values of 29 cm and 3.25 m for fragmentation and back-break, respectively. For comparison purposes, an empirical model (Kuz-Ram) was performed to predict the mean fragment size in the Anguran mine. A mean fragment size of 33.5 cm shows the ABC algorithm can optimize rock fragmentation with a high degree of accuracy.
In blasting operation, the aim is to achieve proper fragmentation and to avoid undesirable events such as backbreak. Therefore, predicting rock fragmentation and backbreak is very important to arrive at a technically and economically successful outcome. Since many parameters affect the blasting results in a complicated mechanism, employment of robust methods such as artificial neural network may be very useful. In this regard, this paper attends to simultaneous prediction of rock fragmentation and backbreak in the blasting operation of Tehran Cement Company limestone mines in Iran. Back propagation neural network (BPNN) and radial basis function neural network (RBFNN) are adopted for the simulation. Also, regression analysis is performed between independent and dependent variables. For the BPNN modeling, a network with architecture 6-10-2 is found to be optimum whereas for the RBFNN, architecture 6-36-2 with spread factor of 0.79 provides maximum prediction aptitude. Performance comparison of the developed models is fulfilled using value account for (VAF), root mean square error (RMSE), determination coefficient (R2) and maximum relative error (MRE). As such, it is observed that the BPNN model is the most preferable model providing maximum accuracy and minimum error. Also, sensitivity analysis shows that inputs burden and stemming are the most effective parameters on the outputs fragmentation and backbreak, respectively. On the other hand, for both of the outputs, specific charge is the least effective parameter.
Backbreak is one of the undesirable effects of blasting operations causing instability in mine walls, falling down the machinery, improper fragmentation and reduction in efficiency of drilling. Backbreak can be affected by various parameters such as the rock mass properties, blasting geometry and explosive properties. In this study, the application of the artificial neural network (ANN), an adaptive neuro-fuzzy inference system (ANFIS) for prediction of backbreak, was described and compared with the traditional statistical model of multiple regression. The performance of these models was assessed through the root mean square error, correlation coefficient (R
2) and mean absolute percentage error. As a result, it was found that the constructed ANFIS exhibited a higher performance than the ANN and multiple regression for backbreak prediction.
Back break is an undesirable phenomenon in blasting operations. It can cause instability of mine walls, falling down of machinery, improper fragmentation, reduced efficiency of drilling, etc. To solve this problem, parameters such as the physico-mechanical properties of rock mass, explosives specifications and geometrical particulars of blast design should be considered to obtain optimum design. Due to multiplicity of effective parameters and complexity of interactions among these parameters, empirical methods may not be fully appropriate for blasting pattern design. In this paper, the artificial neural network (ANN) technique was used to determine the near-optimum blasting pattern so that back break is reduced. The Gol-E-Gohar iron mine in Iran was considered as a case study. A four-layer ANN was found to be optimum with architecture of seven neurons in input layer, 15 and 25 neurons in first and second hidden layer, respectively, and one neuron in output layer. Applying the results obtained from this study, back break was reduced from 20 to 4 m.
In this letter we discuss a least squares version for support vector machine (SVM) classifiers. Due to equality type constraints in the formulation, the solution follows from solving a set of linear equations, instead of quadratic programming for classical SVM''s. The approach is illustrated on a two-spiral benchmark classification problem.
In small-sample settings, bolstered error estimation has been shown to perform better than cross-validation and competitively with bootstrap with regard to various criteria. The key issue for bolstering performance is the variance setting for the bolstering kernel. Heretofore, this variance has been determined in a non-parametric manner from the data. Although bolstering based on this variance setting works well for small feature sets, results can deteriorate for high-dimensional feature spaces.
This article computes an optimal kernel variance depending on the classification rule, sample size, model and feature space, both the original number and the number remaining after feature selection. A key point is that the optimal variance is robust relative to the model. This allows us to develop a method for selecting a suitable variance to use in real-world applications where the model is not known, but the other factors in determining the optimal kernel are known.
Companion website at http://compbio.tgen.org/paper_supp/high_dim_bolstering.
edward@mail.ece.tamu.edu.
Water ingress poses a common and intricate geological hazard with profound implications for tunnel construction's speed and safety. The project's success hinges significantly on the precision of estimating water inflow during excavation, a critical factor in early-stage decision-making during conception and design. This article introduces an optimized model employing the gene expression programming (GEP) approach to forecast tunnel water inflow. The GEP model was refined by developing an equation that best aligns with predictive outcomes. The equation's outputs were compared with measured data and assessed against practical scenarios to validate its potential applicability in calculating tunnel water input. The optimized GEP model excelled in forecasting tunnel water inflow, outperforming alternative machine learning algorithms like SVR, GPR, DT, and KNN. This positions the GEP model as a leading choice for accurate and superior predictions. A state-of-the-art machine learning-based graphical user interface (GUI) was innovatively crafted for predicting and visualizing tunnel water inflow. This cutting-edge tool leverages ML algorithms, marking a substantial advancement in tunneling prediction technologies, providing accuracy and accessibility in water inflow projections.
Coal pillar assessment is of broad importance to underground engineering structure, as the pillar failure can lead to enormous disasters. Because of the highly non-linear correlation between the pillar failure and its influential attributes, conventional forecasting techniques cannot generate accurate outcomes. To approximate the complex behavior of coal pillar, this paper elucidates a new idea to forecast the underground coal pillar stability using combined unsupervised-supervised learning. In order to build a database of the study, a total of 90 patterns of pillar cases were collected from authentic engineering structures. A state-of-the art feature depletion method, t-distributed stochastic neighbor embedding (t-SNE) has been employed to reduce significance of actual data features. Consequently, an unsupervised machine learning technique K-mean clustering was followed to reassign the t-SNE dimensionality reduced data in order to compute the relative class of coal pillar cases. Following that, the reassign dataset was divided into two parts: 70 percent for training dataset and 30 percent for testing dataset, respectively. The accuracy of the predicted data was then examined using support vector classifier (SVC) model performance measures such as precision, recall, and f1-score. As a result, the proposed model can be employed for properly predicting the pillar failure class in a variety of underground rock engineering projects.
Backbreak is an adverse phenomenon in blasting operation, which can cause, among others, mine walls instability, falling down of machinery, drilling efficiency reduction and stripping ratio enhancement. Therefore, this research aimed to develop two-hybrid RF (Random Forest) prediction models of random forest, which are optimized by Harris hawks optimizer (HHO) and sine cosine algorithm (SCA), for estimation of the backbreak distance. The HHO and SCA algorithms were adopted to determine two hyper-parameters (mtry and ntree) in the RF models, in which root mean square error (RMSE) was utilized as a fitness function. A database with 234 samples was established, in which six variables [i.e., hole length (L), burden (B), spacing (S), stemming (T), special drilling (SD) and powder factor (PF)] were used as input variables, and backbreak was defined as output variable. Additionally, three classical regression models (i.e., extreme learning machine, radial basis function network and general regression neural network) were adopted to verify the superiority of the hybrid RF prediction models. The predictive reliability of the proposed models was assessed by the combination of mean absolute error (MAE), RMSE, variance accounted for (VAF) and Pearson correlation coefficient (R2). The results revealed that the SCA-RF model outperformed all the other prediction models with MAE of (0.0444 and 0.0470), RMSE of (0.0816 and 0.0996), VAF of (96.82 and 95.88) and R2 of (0.9876 and 0.9829) in training and testing stages, respectively. A Gini index generated internally in the RF model showed that backbreak was significantly more sensitive to L and T than to SD.
The control of blast-induced ground vibration and generation of back break are main concerns of mine management. Because it damages structures near to the blasting face and makes highwall instable. It was observed during prediction of blast-induced ground vibration through scaled distance technique that it does not consider the superimposition effect of waves. Therefore, the predicted values of ground vibration differ from actual value. This study was conducted to see the superimposition effect using single hole blast concept and optimizing the blast design parameters. By this method, it is possible to nullify the chances of superimposition of waves and their influence on generation of blast-induced ground vibration and backbreak. From the study, it was found that at 92 ms delay between rows generate less backbreak compared to 42 ms and 67 ms delay. It has been found that a significant reduction in vibration can be achieved using 92 ms delay between rows, i.e., reduction up to 31% to 61% compare to 42 ms delay between row and 7% to 42% compare to 67 ms delay between rows, when blast intensity generated by single hole was 10 mm/sec.
Forecasting daily and monthly streamflows is necessary for short- and long-term water resources management, particularly in extreme cases, e.g., flood and drought. Accurate models are needed to plan and manage water resources in watersheds. Recently, support vector regression (SVR) shows the ability to handle the hydrological forecasting issues, but the accuracy of SVR depends on the appropriate choice of parameters and selection of proper inputs. A new meta-heuristic algorithm called grey wolf optimizer (GWO) is used in the current research in order to improve SVR accuracy in forecasting monthly streamflow. The proposed approach is compared with other evolutionary methods, particle swarm optimization, shuffled complex evolution, and multi-verse optimization, that are employed in tuning SVR parameters. Furthermore, the proposed methods were also combined with wavelet transform and tested using monthly streamflow data from two gauging stations, Ain Bedra and Fermatou, in Algeria. To assess the performance of the developed models, Nash–Sutcliffe efficiency (NSE), correlation coefficient, root mean squared error (RMSE), and mean absolute error (MAE) were used. The obtained results indicate that multi-linear regression provides appropriate input variables to SVR models. According to all of the performance measures used, hybrid models exhibit better performances, in monthly streamflow prediction, compared with single versions. For example, for the Ain Bedra station, the NSEcriterion and the correlation coefficient values increased considerably from 27.36% and 0.5405, for the single models, to 95.72% and 0.9786, for the hybrid models. A great decrease has also been obtained for the RMSE and MAE values, which decreased from 0.1562 m³/s and 0.1244 m³/s to 0.6433m³/s and 0.3047 m³/s. In addition, the new GWO algorithm outperformed the other algorithms in terms of both forecasting accuracy and convergence.
Accurate estimation of soil moisture content is necessary for optimal management of water and soil resources. Soil moisture is an important variable in the hydrologic cycle, which plays an important role in the global water and energy balance due to its impact on hydrological, ecological, and meteorological processes. The purpose of the present study was to explore a newly developed hybrid intelligent model for simulating soil moisture content. A hybrid adaptive neuro fuzzy inference system (ANFIS) model-grey wolf optimization (GWO) algorithm was designed here and validated against the neural network (ANN), support vector regression (SVR) and standalone ANFIS models. The models input parameters were di-electric constant, soil bulk density, clay content and organic matter of 1155 soil samples. Various statistical indices were employed to evaluate the performances of the applied models. For a reliable ranking of models, the Global Performance Indicator (GPI) was utilized, which is a 5-agent index. The results evidenced the feasibility of the developed hybrid ANFIS-GWO model with superior simulation results. At the testing stage, the MAE and SI values for the ANFIS-GWO were 1.468% and 0.098, respectively, which indicated the superiority of the ANFIS-GWO compared to the ANFIS-Fuzzy C mean (MAE=6.427%, SI=0.354), and ANFIS-sub clustering (MAE=2.137%, SI=0.141) models. Based on the GPI, the ANFIS-GWO model was ranked as the best model, followed by the standalone ANFIS and SVR models, while the worst accuracy was attained through ANN model. The ANFIS-GWO model improved the simulation accuracy by 48% and 50%, respectively, compared to the standalone ANFIS and SVR models. In addition, based on the GPI, the ANFIS-GWO model presented an enhancement of about 77 percent compared to the ANN model. The high accuracy of the ANFIS-GWO model compared to the standalone ANFIS model represents the performance of the GWO algorithm for escaping local optima, which makes the ANFIS-GWO as a powerful tool for estimating soil moisture. Overall, the explored hybrid intelligent models demonstrated a reliable pedotransfer function of soil moisture where it can contribute to several geo-sciences engineering principles.
Pervious concrete is a widely used construction material thanks to its good drainage characteristics. Before application, its most important properties, i.e. the permeability coefficient (PC) and 28-day unconfined compressive strength (UCS) are required to be tested. However, conducting PC and UCS tests with multiple influencing variables is time-consuming and costly. To address this issue, this paper proposed, for the first time, an evolved support vector regression (ESVR) tuned by beetle antennae search (BAS) to accurately and effectively predict the PC and UCS of pervious concrete. To prepare the dataset of the ESVR model, 270 specimens in total were prepared and casted in a controlled environment in the laboratory. The water-to-cement (w/c) ratio, aggregate-to-cement (a/c) ratio, and aggregate size were selected as the crucial influencing variables for the inputs, while PC and UCS were the outputs of this model. The results indicate that both the PC and UCS firstly increased and then decreased with increasing w/c ratio. As the a/c ratio increased, PC increased, while UCS decreased. Moreover, BAS is more reliable and efficient than random hyper-parameter selection for hyper-parameter tuning. A low root-mean-square error (RMSE) and high correlation coefficient (R) indicate a relatively high predictive capability of the proposed ESVR model. The sensitivity analysis (SA) suggests the a/c ratio and aggregate size were the most sensitive variables for UCS and PC, respectively. This pioneering work provides a simple and convenient method for evaluating PC and UCS of pervious concrete.
Most of the mine operators use Non-electric and Detonating cord with cord relay initiation system for blasting in surface mines. The ground vibration predictor equation proposed by United States Bureau of Mines (USBM) using Non-electric or Detonating cord with cord relay initiation system gives erroneous results while predicting the Peak particle velocity (PPV) at point of interest. The predicted values of PPV using the scaled distance equation and actual values shows significant errors in case of blasting with pyrotechnic based delay initiation system. The errors between predicted and actual PPV may be attributed to the fact that the cap scatter in pyrotechnic based delay initiation system varies from ±10% to ±20%. Variation of actual firing time from the designated time delay may result in altered firing order, overlapping of holes and failure of the blasting sequence, which can cause altered maximum charge per delay resulting in high vibration levels. Therefore, it becomes very important to understand the impact of cap scattering on maximum charge per delay and influence of altered maximum charge per delay on blast-induced ground vibration.
In this paper, a new concept of modified charge per delay has been developed to predict the blast-induced peak particle velocity for blasts using different initiating systems having varying cap scatter while using the same predictor equation. It is seen that a significant reduction in errors has been obtained between actual and predicted PPV with values of root mean square errors (RMSE) up to 1.361 mm/s.
Thesupport-vector network is a new learning machine for two-group classification problems. The machine conceptually implements the following idea: input vectors are non-linearly mapped to a very high-dimension feature space. In this feature space a linear decision surface is constructed. Special properties of the decision surface ensures high generalization ability of the learning machine. The idea behind the support-vector network was previously implemented for the restricted case where the training data can be separated without errors. We here extend this result to non-separable training data.High generalization ability of support-vector networks utilizing polynomial input transformations is demonstrated. We also compare the performance of the support-vector network to various classical learning algorithms that all took part in a benchmark study of Optical Character Recognition.
A generalized form of the cross‐validation criterion is applied to the choice and assessment of prediction using the data‐analytic concept of a prescription. The examples used to illustrate the application are drawn from the problem areas of univariate estimation, linear regression and analysis of variance.
There is no straightforward and systematic method for adequately measuring blast damage in on-going mining operations. This paper considers various techniques available for the measurement of blast damage, making particular reference to recent underground hard-rock mine studies. This paper concludes by indicating the importance of understanding and accounting for geology and emphasizes the need for a robust damage-audit procedure.
In the history of research of the learning problem one can extract four periods that can be characterized by four bright events: (i) Constructing the first learning machines, (ii) constructing the fundamentals of the theory, (iii) constructing neural networks, (iv) constructing the alternatives to neural networks.
a b s t r a c t The application of artificial neural network (ANN) in predicting pile bearing capacity is underlined in several studies. However, ANN deficiencies in finding global minima as well as its slow rate of convergence are the major drawbacks of implementing this technique. The current study aimed at developing an ANN-based predictive model enhanced with genetic algorithm (GA) optimization technique to predict the bearing capacity of piles. To provide necessary dataset required for establishing the model, 50 dynamic load tests were conducted on precast concrete piles in Pekanbaru, Indonesia. The pile geometrical properties, pile set, hammer weight and drop height were set to be the network inputs and the pile ultimate bearing capacity was set to be the output of the GA-based ANN model. The best predictive model was selected after conducting a sensitivity analysis for determining the optimum GA parameters coupled with a trial-and-error method for finding the optimum network architecture i.e. number of hidden nodes. Results indicate that the pile bearing capacities predicted by GA-based ANN are in close agreement with measured bearing capacities. Coefficient of determination as well as mean square error equal to 0.990 and 0.002 for testing datasets respectively, suggest that implementation of GA-based ANN models as a highly-reliable, efficient and practical tool in predicting the pile bearing capacity is of advantage.
Backbreak is one of the destructive side effects of the blasting operation. Reducing of this event is very important for economic of a mining project. Involvement of various parameters has made the backbreak analyzing difficult. Currently there is no any specific method to predict or control the phenomenon considering all the effective parameters. In this paper, artificial neural network (ANN) as a powerful tool for solving such complicated problems is used to predict backbreak in blasting operation of the Sangan iron mine, Iran. Network training was fulfilled using a collected database of the practiced operation including blast design details and rock condition. Trying various types of the networks, a network with two hidden layers was found to be optimum. Performance of the ANN model was compared with statistical analysis using datasets which were kept apart from the original database. According to the obtained results, for the ANN model there existed a higher correlation (R2 = 0.868) and lesser error (RMSE = 0.495) between the predicted and measured backbreak as compared to the regression model. Also, sensitivity analysis revealed that the inputs rock factor and number of rows are the most and the least sensitive parameters on the output backbreak, respectively.
Backbreak is an undesirable phenomenon in blasting operations. It can cause instability of mine walls, falling down of machinery, improper fragmentation, reduced efficiency of drilling, etc. The existence of various effective parameters and their unknown relationships are the main reasons for inaccuracy of the empirical models. Presently, the application of new approaches such as artificial intelligence is highly recommended. In this paper, an attempt has been made to predict backbreak in blasting operations of Soungun iron mine, Iran, incorporating rock properties and blast design parameters using the support vector machine (SVM) method. To investigate the suitability of this approach, the predictions by SVM have been compared with multivariate regression analysis (MVRA). The coefficient of determination (CoD) and the mean absolute error (MAE) were taken as performance measures. It was found that the CoD between measured and predicted backbreak was 0.987 and 0.89 by SVM and MVRA, respectively, whereas the MAE was 0.29 and 1.07 by SVM and MVRA, respectively.
Backbreak is one of the unfavorable blasting results, which can be defined as the unwanted rock breakage behind the last row of blast holes. Blast pattern parameters, like stemming, burden, delay timing, stiffness ratio (bench height/burden) and rock mass conditions (e.g., geo-mechanical properties and joints), are effective in backbreak intensity. Till date, with the exception of some qualitative guidelines, no specific method has been developed for predicting the phenomenon. In this paper, an effort has been made to apply artificial neural networks (ANNs) for predicting backbreak in the blasting operation of the Chadormalu iron mine (Iran). Number of ANN models with different hidden layers and neurons were tried, and it was found that a network with architecture 10-7-7-1 is the optimum model. A comparative study also approved the superiority of the ANN modeling over the conventional regression analysis. Mean square error (MSE), variance account for (VAF) and coefficient of determination (R
2) between measured and predicted backbreak for the ANN model were calculated and found 89.46 %, 0.714 and 90.02 %, respectively. Also, for the regression model, MSE, VAF and R
2 were computed and found 66.93 %, 1.46 and 68.10 %, respectively. Sensitivity analysis was also carried out to find out the influence of each input parameter on backbreak results, and it was revealed that burden is the most influencing parameter on the backbreak, whereas water content is the least effective parameter in this regard.
An ideally performed blasting operation enormously influences the mining overall cost. This aim can be achieved by proper
prediction and attenuation of flyrock and backbreak. Poor performance of the empirical models has urged the application of new approaches. In this paper, an attempt has been
made to develop a new neuro-genetic model for predicting flyrock and backbreak in Sungun copper mine, Iran. Recognition of
the optimum model with this method as compared with the classic neural networks is faster and convenient. Genetic algorithm
was utilized to optimize neural network parameters. Parameters such as number of neurons in hidden layer, learning rate, and
momentum were considered in the model construction. The performance of the model was examined by statistical method in which
absolutely higher efficiency of neuro-genetic modeling was proved. Sensitivity analysis showed that the most influential parameters
on flyrock are stemming and powder factor, whereas for backbreak, stemming and charge per delay are the most effective parameters.
تنفيذ عملية التفجير يؤثر بشكل كبير من الناحية المثالية تكاليف التعدين عموما. ويمكن تحقيق هذا الهدف عن طريق التنبؤ السليم وتخفيف
flyrock وbackbreak. وحثت ضعف الأداء من نماذج تجريبية لتطبيق النهج الجديد. في هذه الورقة ، وقد بذلت محاولة لوضع نموذج جديد
الاعصاب الوراثية للتنبؤ flyrock وbackbreak في منجم للنحاس Sungun وإيران. الاعتراف النموذج الأمثل مع هذا الأسلوب بالمقارنة
مع الشبكة العصبية الكلاسيكي هو أسرع ومريحة. واستخدم الخوارزمية الجينية لتحسين المعلمات الشبكة العصبية. واعتبرت معلمات مثل
عدد الخلايا العصبية في طبقة خفية ، معدل التعلم والزخم في بناء النموذج. تم فحص أداء نموذج من الأسلوب الإحصائي الذي ثبت كفاءة
أعلى على الاطلاق من وضع نماذج الاعصاب الوراثية. حساسية التحليل أظهرت أن المعلمات الأكثر تأثيرا على flyrock هي النابعة ومسحوق
للعامل في حين backbreak الناشئة والمسؤول عن تأخير في والمعلمات الأكثر فعالية.
KeywordsFlyrock-Backbreak-Neuro-genetic approach-Sungun copper mine
Although blasting is the most principal method of fragmentation in hard rock mining, the significance of the costs of blast induced rockmass damage in terms of mining efficiency and safety is becoming increasingly recognized. Backbreak is one of the adverse phenomena in blasting operations that causes the instability of mine walls, falling down of equipments, improper fragmentation, reduced efficiency of drilling, etc., and consequently increases the total cost of a mining operation. In this paper, predictive models based on fuzzy set theory and multivariable regression have been developed for predicting backbreak in Gol-E-Gohar iron mine of Iran. To evaluate performance of the employed models, the coefficient of correlation (R2) and the root mean square error (RMSE) indices were calculated. It was concluded that performance of the fuzzy model is considerably better than regression model. For the fuzzy and regression models, R2 and RMSE were equal to 95.43% and 0.44 and 34.08% and 1.63, respectively. The fuzzy model sensitivity analysis shows that the most effective parameters on backbreak phenomenon are stemming length, hole depth, burden and hole spacing. Application of this model in the Gol-E-Gohar iron mine considerably minimized backbreak and improved blasting efficiency.
