The reliable determination of geomechanical parameters of rocks such as Unconfined Compressive Strength (UCS) using laboratory methods is problematic and time-consuming. In this regard, the construction of reliable predictive models for assessing the UCS is of advantage.
The main purpose of this work is to propose the use of a reliable PSO-based ANN approach for predicting the UCS of sandstones.
For this purpose, laboratory tests were performed on 60 sandstone specimens. The laboratory tests comprise P-wave velocity, dry density, Schmidt hardness and UCS. Apart from the latter, the other laboratory tests were set as model inputs. Prediction performance of the constructed model was assessed according to the criteria including coefficient of determination (R^{2}), Root Mean Squared Error (RMSE) and Variance Account For (VAF).
Results (R^{2}= 0.974 and RMSE = 0.086 and VAF = 97.5) showed the reliability of the constructed PSO-based ANN model to predict UCS of sandstones.
Hence, this study recommends utilizing PSO-based ANN as a feasible tool for assessing UCS of sandstones. Nevertheless, further research is suggested for model generalization purposes.
Open Peer Review Details | |||
---|---|---|---|
Manuscript submitted on 26-12-2019 |
Original Manuscript | A Reliable PSO-based ANN Approach for Predicting Unconfined Compressive Strength of Sandstones |
Rock characterization plays an important role in design of geotechnical structures. There are several methods for determining the engineering properties of rocks. These methods are often divided into direct and indirect methods. Unconfined Compressive Strength (UCS) of rock is one of the direct methods for evaluating the compressive behavior of rock samples. The test is standardized by ISRM [1ISRM, "The Complete ISRM Suggested Methods For Rock Characterization, Testing And Monitoring: 1974–2006", Suggested Methods Prepared by the Commission on Testing Methods, International Society for Rock Mechanics. ISRM Turkish National Group: Ankara, Turkey, .]. However, determining UCS is relatively costly and destructive. Additionally, sometimes providing high-quality rock specimens is a difficult task to be accomplished more especially in the case of porous, thinly bedded, foliated, weak and weathered rocks. These impeding factors encourage laboratory technicians to utilize easier methods (indirect methods) for assessing the compressing strength of rocks. Indirect methods including point load index text, Is_{(50)}, P-wave velocity test, Schmidt hammer test, to name a few, are relatively easier and quick. These index tests can be related to the UCS test. Many researchers recommend conventional regression-based equations for relating rock index tests to UCS. In this regard, Momeni et al. [2E. Momeni, R. Nazir, D. Jahed Armaghani, M.F. Mohd Amin, and E. Tonnizam Mohamad, "Prediction of unconfined compressive strength of rocks: A review paper", J. Teknol., vol. 77, no. 11, pp. 43-50.
[http://dx.doi.org/10.11113/jt.v77.6393] ], in their review paper, provided a comprehensive list of these equations. In another study, Nazir et al. [3R. Nazir, E. Momeni, D. Jahed Armaghani, and M.F. Mohd Amin, "Correlation between unconfined compressive strength and indirect tensile strength of limestone rock samples", Electron. J. Geotech. Eng., vol. 18, no. I, pp. 1737-1746.] proposed a relatively reliable correlation between UCS and Brazilian tensile strength of rocks. Nazir et al. [4R. Nazir, E. Momeni, D. Jahed Armaghani, and M.F. Mohd Amin, "Prediction of unconfined compressive strength of limestone rock samples using L-Type schmidt hammer", Electron. J. Geotech. Eng., vol. 18, no. I, pp. 1767-1775.] also recommended a correlation between UCS and Schmidt hammer rebound number (SRn). Sari [5M. Sari, "Investigating relationships between engineering properties of various rock types", Global J. Earth Sci. Eng., vol. 5, no. 1, pp. 1-25.
[http://dx.doi.org/10.15377/2409-5710.2018.05.1] ] proposed two correlations between UCS and P-wave velocity test as well as UCS and Schmidt hammer rebound number. In the recent past, the application of artificial intelligence in geotechnical engineering is underlined in many studies [6H. Rezaei, R. Nazir, and E. Momeni, "Bearing capacity of thin-walled shallow foundations: An experimental and artificial intelligence based study", J. Zhejiang Univ., vol. 17, no. 4, pp. 273-285. [Springer.].
[http://dx.doi.org/10.1631/jzus.A1500033] -20H. Xu, J. Zhou, P.G. Asteris, D. Jahed Armaghani, and M. Md. Tahir, "Supervised machine learning techniques to the prediction of tunnel boring machine penetration rate", Appl. Sci. (Basel), vol. 9, p. 3715.
[http://dx.doi.org/10.3390/app9183715] ]. Other researchers highlighted the feasibility of soft computing in solving civil Engineering problems [21P.G. Asteris, and V.G. Mokos, "Concrete compressive strength using artificial neural networks", Neural Comput. Appl., .
[http://dx.doi.org/10.1007/s00521-019-04663-2] -37E. Momeni, "Feasibility of soft computing in civil engineering- A delusion", Third International Conference on Applied Researches in Structural Engineering and Construction Management (secm2019), Sharif University of Technology: Tehran, Iran, .]. Sometimes, to have a better evaluation, it is suggested to investigate the effect of a number of parameters on the parameter of interest. For example, although UCS can be estimated from BTS using conventional correlations, some studies suggest the prediction of UCS from multivariate equations. Nevertheless, in general, multivariate equations are not working well enough for non-linear problems when the contact nature between input and output parameters is unknown. Among different artificial intelligence techniques, many researchers suggested the feasibility of artificial neural networks in predicting the UCS of rocks [38Y. Abdi, A.T. Garavand, and R.Z. Sahamieh, "Prediction of strength parameters of sedimentary rocks using artificial neural networks and regression analysis", Arab. J. Geosci., vol. 11, p. 587.
[http://dx.doi.org/10.1007/s12517-018-3929-0] ]. However, as stated in the literature, ANN suffers from two major drawbacks [39E. Tonnizam Mohamad, D. Jahed Armaghani, E. Momeni, and S.V Alavi Nehzhad Khalil Abad, "Prediction on unconfined compressive strength of soft rocks: A PSO-based ann approach", Bull. Eng. Geol. Environ., vol. 74, no. 3, pp. 745-757.
[http://dx.doi.org/10.1007/s10064-014-0638-0] ]: slow rate of learning and getting trapped in local minima. Several studies reported the improvement of ANN by implementing the optimization algorithms including particle swarm optimization algorithm [40E. Momeni, "Prediction of uniaxial compressive strength of rock samples using hybrid particle swarm optimization-based artificial neural networks", Measurement, vol. 60, pp. 50-63.
[http://dx.doi.org/10.1016/j.measurement.2014.09.075] ], genetic algorithm [41E. Momeni, R. Nazir, D.J. Armaghani, and H. Maizir, "Prediction of pile bearing capacity using a hybrid genetic algorithm-based ANN", Measurement, vol. 57, pp. 122-131.
[http://dx.doi.org/10.1016/j.measurement.2014.08.007] ], and imperialist competitive algorithm [42D. Jahed Armaghani, E. Tonnizam Mohamad, E. Momeni, M. Monjezi, and M. Sundaram, "Prediction of the strength and elasticity modulus of granite through an expert artificial neural network", Arab. J. Geosci., vol. 9, no. 48, .] to name a few. This paper is aimed to predict UCS of sandstones using the PSO-based ANN model. Additionally, the Multiple Linear Regression (MLR) was used for estimating UCS. Although UCS prediction has drawn considerable attention in the literature, presenting new real datasets from different parts of world is always of interest as rock behavior is site specific varies from a place to another place.
Numerous correlations have been proposed for estimating UCS in the literature. In Table 1, the regression-based correlations are tabulated and Table 2 presents the proposed artificial intelligence-based predictive models of UCS. Among others, Meulenkamp and Grimma [43F. Meulenkamp, and M.A. Grima, "Application of neural networks for the prediction of the Unconfined Compressive Strength (UCS) from equotip hardness", Int. J. Rock Mech. Min. Sci., vol. 36, pp. 29-39.
[http://dx.doi.org/10.1016/S0148-9062(98)00173-9] ] developed a predictive model of UCS using ANN. The size of the dataset in their study was 194. They used Equotip value, porosity, density, and grain size as their input parameters. According to their conclusion, the coefficient of determination of their model was high enough (R^{2} = 0.94). In another study, Singh et al. [44V.K. Singh, D. Singh, and T.N. Singh, "Prediction of strength properties of some schistose rocks from petrographic properties using artificial neural networks", Int. J. Rock Mech. Min. Sci., vol. 38, pp. 269-284.
[http://dx.doi.org/10.1016/S1365-1609(00)00078-2] ] proposed an ANN-based predictive model of UCS based on 112 datasets. They used petrography data to train and test their model. Also, Dehghan et al. [45S. Dehghan, G.H. Sattari, C.S. Chehreh, and M.A. Aliabadi, "Prediction of unconfined compressive strength and modulus of elasticity for travertine samples using regression and artificial neural network", Min. Sci. Technol, vol. 20, pp. 0041-0046.] recommended the feasibility of ANN in predicting UCS. They developed the model using 30 datasets. Their input data included Vp, point load index values, SRn, and porosity. They reported R^{2} = 0.86 and concluded that their predictive model is reliable enough. Rezaei et al. [46M. Rezaei, A. Majdi, and M. Monjezi, "An intelligent approach to predict unconfined compressive strength of rock surrounding access tunnels in longwall coal mining", Neural Comput. Appl., vol. 24, pp. 233-241.
[http://dx.doi.org/10.1007/s00521-012-1221-x] ] developed a Fuzzy Inference System (FIS) to predict UCS of rocks based on 93 datasets . Their model input consisted of porosity, SRn and density. Monjezi et al. [47M. Monjezi, H.A. Khoshalan, and M. Razifard, "A neuro-genetic network for predicting uniaxial compressive strength of rocks", Geotech. Geol. Eng, vol. 30, pp. 1053-1062.
[http://dx.doi.org/10.1007/s10706-012-9510-9] ] used the same input parameters. However, Monjezi et al. utilized a GA-based ANN model for developing their model. The reliability of the predictive model of UCS proposed by Monjezi et al. in terms of R^{2} was 0.96. Using 72 datasets, Beiki et al. [48M. Beiki, A. Majdi, and A.D. Givshad, "Application of genetic programming to predict the uniaxial compressive strength and elastic modulus of carbonate rocks", Int. J. Rock Mech. Min. Sci., vol. 63, pp. 159-169.
[http://dx.doi.org/10.1016/j.ijrmms.2013.08.004] ] developed a predictive model of UCS using genetic programming. The input parameters used in their study were density, porosity and Vp. They reported R^{2} equal to 0.94 as the performance index of their model. Yagiz et al. [49S. Yagiz, E.A. Sezer, and C. Gokceoglu, "Artificial neural networks and nonlinear regression techniques to assess the influence of slake durability cycles on the prediction of uniaxial compressive strength and modulus of elasticity for carbonate rocks. Int", J. Numer. Anal. Met., vol. 36, pp. 1636-1650.
[http://dx.doi.org/10.1002/nag.1066] ] utilized ANN for developing a predictive model of UCS using 54 datasets . Their input parameters were porosity, point load index value, dry unit weight, Vp, and SRn. However, according to their conclusion, their proposed model was not reliable enough (R^{2} = 0.50). Using 105 sets of data, Torabi-kaveh et al. [50M. Torabi-Kaveh, F. Naseri, S. Saneie, and B. Sarshari, "Application of artificial neural networks and multivariate statistics to predict UCS And E using physical properties of asmarilimestones", Arab. J. Geosci., .
[http://dx.doi.org/10.1007/s12517-014-1331-0] ] recommended an ANN-based predictive model of UCS. Their proposed model with R^{2} = 0.95 was highly reliable. Similar to the aforementioned works, they used density, porosity and Vp as the model inputs. Yesiloglu-Gultekin et al. [51N. Yesiloglu-Gultekin, C. Gokceoglu, and E.A. Sezer, "Prediction of uniaxial compressive strength of granitic rocks by various nonlinear tools and comparison of their performances", Int. J. Rock Mech. Min. Sci., vol. 62, pp. 113-122.
[http://dx.doi.org/10.1016/j.ijrmms.2013.05.005] ] implemented an Adaptive Neuro-fuzzy Inference System (ANFIS) for UCS prediction. They used Brazilian tensile strength and Vp as the model inputs. The size of the dataset in their study was 75. The reported R^{2} in their study was 0.60. In another study, Abdi et al. [38Y. Abdi, A.T. Garavand, and R.Z. Sahamieh, "Prediction of strength parameters of sedimentary rocks using artificial neural networks and regression analysis", Arab. J. Geosci., vol. 11, p. 587.
[http://dx.doi.org/10.1007/s12517-018-3929-0] ] predicted the UCS of sedimentary rocks using ANN. In order to train their model, they collected 196 different types of rock specimens including limestone, conglomerate, sandstone, and marl. The input parameters of their suggested model were dry unit weight, Vp, porosity, and water absorption. The correlation coefficient, R, of their proposed model for testing data was 0.93. Overall, the aforementioned studies, as well as studies highlighted in Table 2, indicate the workability of artificial intelligence more especially artificial neural network in predicting UCS of rock using different rock engineering properties as well as rock index tests.
Artificial Neural Network (ANN) can be seen as a black box which is utilized when there is a highly nonlinear relationship between input and output of a model. The term black box sometimes is given to the ANN because unlike regression-based techniques, there is no specific formula for estimating the parameter of interest. Multilayer Perceptron (MLP) is the most common type of ANN_{S} for both estimation and classification problems [91G. Dreyfus, Neural Networks: Methodology And Application., Springer Berlin Heidelberg: Germany, .-95M. Hajihassani, D. Jahed Armaghani, H. Sohaei, E. Tonnizam Mohamad, and A. Marto, "Prediction of airblast-overpressure induced by blasting using a hybrid artificial neural network and particle swarm optimization", Appl. Acoust., vol. 80, pp. 57-67.
[http://dx.doi.org/10.1016/j.apacoust.2014.01.005] ]. The structure of the MLP network includes input, hidden and output layers. There is no specific approach for determining the number of hidden layers as well as the number of hidden nodes . However, almost in most civil engineering problems, the use of one hidden layer is good enough. It is due to the fact that in civil engineering problems, collecting experimental data for training the ANN-based predictive models is difficult. In other words, when the size of dataset is small, implementation of more than one hidden layer can increase the likelihood of the model overtraining and overfitting. Nevertheless, the input, hidden and output layers are connected to each other using hidden nodes or neurons. That is the input parameters are connected to hidden nodes in the hidden layer and the aforementioned hidden nodes are connected to the model output. Each connection has a weight. Depending on the training algorithm, the weight of each connection is determined. One of the most famous training algorithms is backpropagation technique. In this method, the network starts to feed forward. In the hidden layer, to have a hidden node output value, a transfer function is applied to the input values of each hidden node. The input value of each hidden nodes is simply determined by the summation of connection weights of that hidden node as well as a threshold value known as bias. This process is repeated between the hidden layer and the output layer and the value of the output parameter is predicted. The predicted output is compared with the target (already known) and the error is estimated. The network has to backpropagate and update its weights (if the error is not desirable). In the hybrid ANNs, instead of conventional methods, the weights are optimized using optimization algorithms. Detail of ANNs is out of the scope of this paper and readers are referred to other studies to get more detail [9E.T. Mohamad, D.J. Armaghani, and E. Momeni, "Rock strength estimation: A PSO-based BP approach", Neural Computing and Application., Springer, .
[http://dx.doi.org/10.1007/s00521-016-2728-3] ].
Particle Swarm Optimization was originally proposed by Eberhart and Kennedy [96R. Eberhart, and J. Kennedy, "A new optimizer using particle swarm theory", Proc 6^{th} International Symposium on Micro Machine and Human Science, pp. 39-43.Nagoya, Japan4–6 Oct 1995
[http://dx.doi.org/10.1109/MHS.1995.494215] ]. Several studies have been conducted on the efficiency of PSO in achieving optimal solutions in large search spaces, e.g., the study by Mendes et al. [97R. Mendes, P. Cortes, M. Rocha, and J. Neves, "Particle swarms for feed forward neural net training", Proc IEEE Int Joint Conf on Neural Networks, pp. 1895-1899.Honolulu, HI, USA]. In this regard, a primary advantage of PSO is the simple evolutionary process that distinguishes it from other optimization algorithms. Victoire and Jeyakumar [98T. Victoire, and A.E. Jeyakumar, "Hybrid PSO–SQP for economic dispatch with valve-point effect", Electr. Power Syst. Res., vol. 71, pp. 51-59.
[http://dx.doi.org/10.1016/j.epsr.2003.12.017] ] showed that PSO is an effective computational tool that cuts the required amount of memory in comparison to other similar algorithms. PSO works with two simple equations.
(1) |
(2) |
In Eqs. 1 and 2, v_{new}, v, p_{new}, and p denote the new velocity, current velocity, new position, and current position, respectively, for a given particle among a set of particles. Also, C_{1} and C_{2} show acceleration constants; p_{best} indicates the best position of a particle and g_{best} stands for the best position of the particles in the set; r_{1} and r_{2} are arbitrary values in the range of 0 to 1. PSO optimization is initiated by randomly selecting some particles. In the current context, it starts with initializing the ANN weights. A random velocity and position are assigned to each particle (i.e., ANN weight). Next, an iterative process is employed to seek the optimum solution by recording the P_{best} and g_{best} values of each particle during each interaction. Afterward, using Eqs. 1 and 2, the changes in the positions of the particles with respect to their experiences and the position of the other particles are determined [99R. Poli, J. Kennedy, and T. Blackwell, "Particle swarm optimization", Swarm Intell., vol. 1, no. 1, pp. 33-57.
[http://dx.doi.org/10.1007/s11721-007-0002-0] ]. These positions are updated until reaching a predefined “termination criterion” or the optimum solution [100Y. Shi, and R.C. Eberhart, "Empirical study of particle swarm optimization", Proceedings of the 1999 IEEE Congress on Evolutionary ComputationIEEE: New York, . 1999, pp. 1945-1950.
[http://dx.doi.org/10.1109/CEC.1999.785511] ].
To enhance the efficiency of ANNs, often optimization algorithms like PSO are implemented. This is generally attributed to the shortcomings of ANNs in getting trapped in local minima. The problem can be solved if ANN is coupled with global search algorithms like PSO. The PSO component of such a hybrid system enables it to find a global minimum while searching. Accordingly, a hybrid PSO-based ANN model enjoys the benefits of both PSO and ANN; PSO searches for the entire minima in the search space and ANN then employs them to obtain the best solution. It is worth mentioning that in the hybrid system, the ANN weights are optimized using PSO. Each particle of PSO consists of the ANN weights. When the number of iterations is reached in PSO, the algorithm returns the best solution (G_{best}) which contains the whole ANN weights. The obtained weights are applied to train the network.
One of the key steps in network designing is data collection. For this purpose, 10 sandstone blocks were selected from different locations in Central Iran and Sanandaj-Sirjan zones (Fig. 1). These sandstones belong to the upper red formation in southwest Qom and northeast Hamedan, and Jurassic sandstones in the east of Hamedan. Then, the collected blocks were cored in the laboratory to prepare core samples with NX size (54.1 mm diameter) based on the ISRM [1ISRM, "The Complete ISRM Suggested Methods For Rock Characterization, Testing And Monitoring: 1974–2006", Suggested Methods Prepared by the Commission on Testing Methods, International Society for Rock Mechanics. ISRM Turkish National Group: Ankara, Turkey, .]. To develop the ANN model, different engineering properties (physical or mechanical) of 60 sandstone samples such as UCS, Schmidt hardness, P-wave velocity, porosity, and dry density were determined according to the ISRM [1ISRM, "The Complete ISRM Suggested Methods For Rock Characterization, Testing And Monitoring: 1974–2006", Suggested Methods Prepared by the Commission on Testing Methods, International Society for Rock Mechanics. ISRM Turkish National Group: Ankara, Turkey, .]. The study by Momeni et al. [9E.T. Mohamad, D.J. Armaghani, and E. Momeni, "Rock strength estimation: A PSO-based BP approach", Neural Computing and Application., Springer, .
[http://dx.doi.org/10.1007/s00521-016-2728-3] ] shows that there should be a good correlation between the input parameters and the model output. Hence, in this study, apart from UCS, other properties (which are related to UCS) were set as model inputs based on an extensive literature review (Tables 1 and 2). The results of the laboratory tests of selected samples are listed in Table 3. As presented in this table, the porosity values of studied sandstones range from 2.19 to 14.51%. The P-wave velocity of samples varies between 1.09 and 5.38 km/sec. The results for dry density show that γ_{d} range from 2.04 to 2.94 gr/cm^{3}. Also, the Schmidt hardness number for selected sandstones varies between 23 and 46. The results of this work indicate that the UCS values range from 29.94 to 143 MPa.
Fig. (1) Map of sampling points. |
Fig. (2) The effect of the number of iterations on the prediction performance of the PSO-based predictive model of UCS. |
The modelling procedure of the PSO-based ANN predictive model starts with data normalization and sensitivity analysis. In fact, one may perform different sensitivity analyses in order to obtain the proper PSO parameters. In this study, the number of iterations was set to be the termination criterion for the hybrid model. For this reason, using a sensitivity analysis, firstly, the number of iterations was set to 400 and then the prediction performance of the model was evaluated using mean square error. Fig. (2) shows the effect of the number of iterations on the model response. As shown in this model, after nearly 50 iterations there is no significant change on the MSE; therefore, the maximum number of iterations was set to be 50. The values of C_{1} and C_{2} were set to be 2 according to the literature. The study by Tonnizam Mohamad et al. [39E. Tonnizam Mohamad, D. Jahed Armaghani, E. Momeni, and S.V Alavi Nehzhad Khalil Abad, "Prediction on unconfined compressive strength of soft rocks: A PSO-based ann approach", Bull. Eng. Geol. Environ., vol. 74, no. 3, pp. 745-757.
[http://dx.doi.org/10.1007/s10064-014-0638-0] ] recommends that the swarm size does not have a remarkable effect on the model performance; hence the number of particles in this study was set to 200.
After proper determination of the PSO parameters, the structure of the main ANN model (i.e. number of hidden nodes) has to be configured. In this regard, two methods are recommended in the literature. Some studies recommend using a trial and error procedure for determining the proper number of hidden nodes. On the other hand, some studies suggest the implementation of a number of equations for hidden node determination. These equations are related to the number of inputs and output(s). In this study, based on the authors' experience, the trial-and-error method was implemented. In other words, the models were trained and tested with three, four and five hidden nodes, respectively and the prediction performances were evaluated based on R^{2} as well as Root Mean Square Error (RMSE).
It is worth mentioning that each model was trained and tested five times. Since increasing the number of hidden nodes can lead to model complexity, in this study, a maximum of five hidden nodes in a hidden layer were used for training the model. It should also be mentioned that 80 percent of the data was used for training the model and the remaining 20% of the data was set for testing the model performance.
Table 4 shows the results of the implemented trial and error procedure for determining the best network architecture. It is worth mentioning that the prediction performance of the testing data was considered as the main criterion for the selection of the best model. Table 4 shows that the prediction performances of all models are close to each other meaning that the likelihood of accidental good results is negligible. When results for different hidden nodes are close to each other, to have a less complex model, one may select the model with the lowest hidden node more especially when the size of the dataset is not large. Hence, in this study, the 5^{th} PSO-based ANN predictive model of UCS with three hidden nodes is selected as the best model.
Figs. (3-5) indicate the reliability of the constructed PSO-based ANN model. Fig. (3) plots the predicted UCS values (in the training step) versus the measured UCS values in the laboratory. As shown in this figure, the coefficient of determination of the model for the training step is 0.941 which suggests the reliability of the model. As demonstrated in Fig. (4), the estimated UCS using the PSO-based ANN model for testing data is close to the measured UCS (R^{2}=0.974), which indicates the relatively high reliability of the constructed network. As mentioned before, the coefficient of determination of more than 0.9 indicates that the predicted and measured values are in good agreement. Hence, based on the aforementioned results the predictive model of UCS is a feasible tool in predicting the UCS. It should be highlighted that implementations of such low- cost predictive models are simple and quick compared to direct or indirect methods of UCS estimation. However, the model should not be generalized when the estimated range of geomechanical properties are beyond the range of the presented data in Table 3. The proper prediction performance of the PSO-based ANN technique to estimate the UCS during training and testing steps is also shown in Fig. (5) for both training and testing data. As displayed in this figure, the predicted UCS values using the constructed network are close to the measured values, which indicate the reliability of the proposed PSO-based ANN model.
Fig. (3) Correlation between predicted and measured UCS values (training stage). |
Fig. (4) Correlation between predicted and measured UCS values (testing stage). |
Fig. (5) Comparison between predicted and measured UCS using the PSO-based ANN model a) Training Data b) Testing Data. |
Fig. (6) Correlation between predicted and measured UCS values from MLR technique a) training step; b) testing step. |
In this study, for comparison purposes, Multiple Linear Regression (MLR) method is also implemented for the prediction of UCS. Same dataset and input parameters were used in the MLR-based predictive model. To produce a multivariate equation, the statistical software package SPSS23 was used. Lastly, the best obtained multivariate regression equation is suggested for UCS estimation (Eq. 3).
(3) |
Fig. (6) shows the correlations between the estimated values of the UCS using MLR method (Eq. 3) and the observed values in the laboratory for both training and testing data. On the basis of the results, the R^{2} between the predicted and observed UCS for training and testing steps are 0.940 and 0.966, respectively. Hence, it can be concluded that the MLR is also capable of forecasting UCS with acceptable precision. However, the use of the PSO-based ANN predictive model of UCS is more preferable.
As mentioned earlier, the prediction performances of the developed models were assessed using different standard statistical indices including the R^{2}, and Root Mean Square Error (RMSE). At the final stage, Variance Accounts For (VAF) was also utilized for comparing both predictive models. The aforementioned performance indices were calculated using the following equations:
(4) |
(5) |
(6) |
Where X_{i} and Y_{i} are the observed and predicted data, respectively; and are mean of the measured and predicted data, and n is the number of data points. As previously mentioned, in theory, an estimation model is excellent when R^{2} is 1, RMSE is equal to 0 and VAF is 100. The values of the aforementioned statistical indices are tabulated in Table 5. The performance indices values shown in this table suggest that the PSO-based ANN model outperforms the MLR-based predictive model of UCS; therefore, this study recommends the implementation of the PSO-based predictive model as a feasible tool for estimation of UCS. However, further research is suggested for model generalization.
In this study, using 60 datasets, two predictive models of UCS were proposed. Porosity, P-wave velocity, dry density and Schmidt hardness number were set to be model inputs. Based on the results, it was concluded that the PSO-based ANN model with three hidden nodes performs good enough in assessing the UCS of sandstone. A comparison between the prediction performance of the intelligent-based predictive model and MLR-based predictive model showed that the PSO-based ANN model outperforms the MLR-based predictive model. However, it should be highlighted that the proposed predictive models work well enough when the future data are in the range of the data presented in Table 3. Further research with the larger dataset is recommended for overcoming the aforementioned limitation.
Not applicable.
Not applicable.
None.
The authors declare no conflict of interest, financial or otherwise.
Declared none.
[1] | ISRM, "The Complete ISRM Suggested Methods For Rock Characterization, Testing And Monitoring: 1974–2006", Suggested Methods Prepared by the Commission on Testing Methods, International Society for Rock Mechanics. ISRM Turkish National Group: Ankara, Turkey, . |
[2] | E. Momeni, R. Nazir, D. Jahed Armaghani, M.F. Mohd Amin, and E. Tonnizam Mohamad, "Prediction of unconfined compressive strength of rocks: A review paper", J. Teknol., vol. 77, no. 11, pp. 43-50. [http://dx.doi.org/10.11113/jt.v77.6393] |
[3] | R. Nazir, E. Momeni, D. Jahed Armaghani, and M.F. Mohd Amin, "Correlation between unconfined compressive strength and indirect tensile strength of limestone rock samples", Electron. J. Geotech. Eng., vol. 18, no. I, pp. 1737-1746. |
[4] | R. Nazir, E. Momeni, D. Jahed Armaghani, and M.F. Mohd Amin, "Prediction of unconfined compressive strength of limestone rock samples using L-Type schmidt hammer", Electron. J. Geotech. Eng., vol. 18, no. I, pp. 1767-1775. |
[5] | M. Sari, "Investigating relationships between engineering properties of various rock types", Global J. Earth Sci. Eng., vol. 5, no. 1, pp. 1-25. [http://dx.doi.org/10.15377/2409-5710.2018.05.1] |
[6] | H. Rezaei, R. Nazir, and E. Momeni, "Bearing capacity of thin-walled shallow foundations: An experimental and artificial intelligence based study", J. Zhejiang Univ., vol. 17, no. 4, pp. 273-285. [Springer.]. [http://dx.doi.org/10.1631/jzus.A1500033] |
[7] | D. Jahed Armaghani, E. Momeni, S.V.A.N.K. Abad, and M. Khandelwal, "Feasibility of ANFIS model for prediction of ground vibrations resulting from quarry blasting", Environ. Earth Sci., pp. 1-16. [Springer]. [http://dx.doi.org/10.1007/s12665-015-4305-y] |
[8] | D. Jahed Armaghani, R.S. Faradonbeh, and E. Momeni, "Performance prediction of tunnel boring machine through developing a gene expression programming equation", Engineering with Computers, Springer, . [http://dx.doi.org/10.1007/s00366-017-0526-x] |
[9] | E.T. Mohamad, D.J. Armaghani, and E. Momeni, "Rock strength estimation: A PSO-based BP approach", Neural Computing and Application., Springer, . [http://dx.doi.org/10.1007/s00521-016-2728-3] |
[10] | E. Momeni, D.J. Jahed Armaghani, S.A. Fatemi, and R. Nazir, "Prediction of bearing capacity of thin-walled foundation: A simulation approach", Eng. Comput., . [http://dx.doi.org/10.1007/s00366-017-0542- xSpringer] |
[11] | A.R. Bunawan, "Experimental and intelligent techniques to estimate bearing capacity of cohesive soft soils reinforced with soil-cement columns", Measurement, vol. 124, pp. 529-538. [http://dx.doi.org/10.1016/j.measurement.2018.04.057] |
[12] | M. Duc Nguyen, and B. Thai Pham, "Development of an artificial intelligence approach for prediction of consolidation coefficient of soft soil: A sensitivity analysis", Open Construct. Build. Technol. J., vol. 13, pp. 178-188. [http://dx.doi.org/10.2174/1874836801913010178] |
[13] | G.R. Khanlari, M. Heydari, A.A. Momeni, and Y. Abdi, "Prediction of shear strength parameters of soils using artificial neural networks and multivariate regression methods", Eng. Geol., vol. 131–132, pp. 11-18. [http://dx.doi.org/10.1016/j.enggeo.2011.12.006] |
[14] | D. Jahed Armaghani, E. Tonnizam Mohamad, M. Sundaram Narayanasamy, N. Narita, and S. Yagiz, "Development of hybrid intelligent models for predicting TBM penetration rate in hard rock condition", Tunn. Undergr. Space Technol., vol. 63, pp. 29-43. [http://dx.doi.org/10.1016/j.tust.2016.12.009] |
[15] | E. Momeni, R. Nazir, D. Jahed Armaghani, and H. Maizir, "Application of artificial neural network for predicting shaft and tip resistances of concrete piles", Earth Sci. Res. J., vol. 19, no. 1, pp. 85-93. [http://dx.doi.org/10.15446/esrj.v19n1.38712] |
[16] | D. Jahed Armaghani, R. Shirani Faradonbeh, H. Rezaei, A. Safuan, A. Rashid, and H. Bakhshandeh Amnieh, "Settlement prediction of the rock-socketed piles through a new technique based on gene expression programming", Neural Comput. Appl., vol. 29, no. 11, pp. 1115-1125. [http://dx.doi.org/10.1007/s00521-016-2618-8] |
[17] | E. Tonnizam Mohamad, M. Koopialipoor, B. Ramesh Murlidhar, A. Rashiddel, A. Hedayat, and D. Jahed Armaghani, "A new hybrid method for predicting ripping production in different weathering zones through in situ tests", Measurement, vol. 147, .106826 [http://dx.doi.org/10.1016/j.measurement.2019.07.054] |
[18] | M. Hajihassani, S. Shah Abdullah, P.G. Asteris, and D. Jahed Armaghani, "A gene expression programming model for predicting tunnel convergence", Appl. Sci. (Basel), vol. 9, p. 4650. [http://dx.doi.org/10.3390/app9214650] |
[19] | L. Huang, P.G. Asteris, M. Koopialipoor, D. Jahed Armaghani, and M.M. Tahir, "Invasive weed optimization technique-based ann to the prediction of rock tensile strength", Appl. Sci. (Basel), vol. 9, p. 5372. [http://dx.doi.org/10.3390/app9245372] |
[20] | H. Xu, J. Zhou, P.G. Asteris, D. Jahed Armaghani, and M. Md. Tahir, "Supervised machine learning techniques to the prediction of tunnel boring machine penetration rate", Appl. Sci. (Basel), vol. 9, p. 3715. [http://dx.doi.org/10.3390/app9183715] |
[21] | P.G. Asteris, and V.G. Mokos, "Concrete compressive strength using artificial neural networks", Neural Comput. Appl., . [http://dx.doi.org/10.1007/s00521-019-04663-2] |
[22] | P.G. Asteris, I. Argyropoulos, L. Cavaleri, and H. Rodrigues, Masonry compressive strength prediction using artificial neural networks. International Conference on Transdisciplinary Multispectral Modeling and Cooperation for the Preservation of Cultural Heritage, Springer: Cham, Switzerland, pp. 200-224. |
[23] | M. Apostolopoulou, M.G. Douvika, I.N. Kanellopoulos, A. Moropoulou, and P.G. Asteris, "Prediction of compressive strength of mortars using artificial neural networks", 1^{st} international conference TMM_CH, transdisciplinary multispectral modelling and cooperation for the preservation of cultural heritage Athens, Greece,. 2018 Oct (pp. 10-13). |
[24] | P.G. Asteris, A. Ashrafian, and M. Rezaie-Balf, "Prediction of the compressive strength of self-compacting concrete using surrogate models", Comput. Concr., vol. 24, no. 2, pp. 137-150. |
[25] | P.G. Asteris, S. Nozhati, M. Nikoo, L. Cavaleri, and M. Nikoo, "Krill herd algorithm-based neural network in structural seismic reliability evaluation", Mech. Adv. Mater. Structures, vol. 26, no. 13, pp. 1146-1153. [http://dx.doi.org/10.1080/15376494.2018.1430874] |
[26] | P.G. Asteris, P.C. Roussis, and M.G. Douvika, "Feed-forward neural network prediction of the mechanical properties of sandcrete materials", Sensors (Basel), vol. 17, no. 6, p. 1344. [http://dx.doi.org/10.3390/s17061344] [PMID: 28598400] |
[27] | PG. Asteris, DJ Armaghani, and CG Hatzigeorgiou, "Predicting the shear strength of reinforced concrete beams using artificial neural networks", Comput. Concr., vol. 24, no. 5, pp. 469-488. |
[28] | M. Apostolopoulou, D.J. Armaghani, A. Bakolas, M.G. Douvika, A. Moropoulou, and P.G. Asteris, "Compressive strength of natural hydraulic lime mortars using soft computing techniques", Proc. Struct. Integr., vol. 17, pp. 914-923. [http://dx.doi.org/10.1016/j.prostr.2019.08.122] |
[29] | D. Jahed Armaghani, G.D. Hatzigeorgiou, Ch. Karamani, A. Skentou, I. Zoumpoulaki, and PG. Asteris, "Soft computing-based techniques for concrete beams shear strength", Procedia Structural Integrity, vol. 17, pp. 924-933. [http://dx.doi.org/10.1016/j.prostr.2019.08.123] |
[30] | P.G. Asteris, K.G. Kolovos, M.G. Douvika, and K. Roinos, "Prediction of self-compacting concrete strength using artificial neural networks", Eur. J. Environ. Civ. Eng., vol. 20, pp. s102-s122. [http://dx.doi.org/10.1080/19648189.2016.1246693] |
[31] | P.G. Asteris, A.K. Tsaris, L. Cavaleri, C.C. Repapis, A. Papalou, F. Di Trapani, and D.F. Karypidis, "Prediction of the fundamental period of infilled RC frame structures using artificial neural networks", Comput. Intell. Neurosci., vol. 2016, .5104907 [http://dx.doi.org/10.1155/2016/5104907] [PMID: 27066069] |
[32] | P.G. Asteris, and K.G. Kolovos, "Self-compacting concrete strength prediction using surrogate models", Neural Comput. Appl., vol. 31, no. 1, pp. 409-424. [http://dx.doi.org/10.1007/s00521-017-3007-7] |
[33] | A. Manouchehrian, M. Sharifzadeh, and R.H. Moghadam, "Application of artificial neural networks and multivariate statistics to estimate UCS using textural characteristics", Int. J. Min. Sci. Technol., vol. 22, no. 2, pp. 229-236. [http://dx.doi.org/10.1016/j.ijmst.2011.08.013] |
[34] | P.G. Asteris, and M. Nikoo, "Artificial bee colony-based neural network for the prediction of the fundamental period of in filled frame structures", Neural Comput. Appl., vol. 31, no. 9, pp. 4837-4847. [http://dx.doi.org/10.1007/s00521-018-03965-1] |
[35] | P. Sarir, J. Chen, P.G. Asteris, D.J. Armaghani, and M.M. Tahir, "Developing GEP tree-based, neuro-swarm, and whale optimization models for evaluation of bearing capacity of concrete-filled steel tube columns", Eng. Comput., . [http://dx.doi.org/10.1007/s00366-019-00808-y] |
[36] | H. Chen, P.G. Asteris, D.J. Armaghani, B. Gordan, and B.T. Pham, "Assessing dynamic conditions of the retaining wall using two hybrid intelligent models", Appl. Sci. (Basel), vol. 2019, no. 9, p. 1042. [http://dx.doi.org/10.3390/app9061042] |
[37] | E. Momeni, "Feasibility of soft computing in civil engineering- A delusion", Third International Conference on Applied Researches in Structural Engineering and Construction Management (secm2019), Sharif University of Technology: Tehran, Iran, . |
[38] | Y. Abdi, A.T. Garavand, and R.Z. Sahamieh, "Prediction of strength parameters of sedimentary rocks using artificial neural networks and regression analysis", Arab. J. Geosci., vol. 11, p. 587. [http://dx.doi.org/10.1007/s12517-018-3929-0] |
[39] | E. Tonnizam Mohamad, D. Jahed Armaghani, E. Momeni, and S.V Alavi Nehzhad Khalil Abad, "Prediction on unconfined compressive strength of soft rocks: A PSO-based ann approach", Bull. Eng. Geol. Environ., vol. 74, no. 3, pp. 745-757. [http://dx.doi.org/10.1007/s10064-014-0638-0] |
[40] | E. Momeni, "Prediction of uniaxial compressive strength of rock samples using hybrid particle swarm optimization-based artificial neural networks", Measurement, vol. 60, pp. 50-63. [http://dx.doi.org/10.1016/j.measurement.2014.09.075] |
[41] | E. Momeni, R. Nazir, D.J. Armaghani, and H. Maizir, "Prediction of pile bearing capacity using a hybrid genetic algorithm-based ANN", Measurement, vol. 57, pp. 122-131. [http://dx.doi.org/10.1016/j.measurement.2014.08.007] |
[42] | D. Jahed Armaghani, E. Tonnizam Mohamad, E. Momeni, M. Monjezi, and M. Sundaram, "Prediction of the strength and elasticity modulus of granite through an expert artificial neural network", Arab. J. Geosci., vol. 9, no. 48, . |
[43] | F. Meulenkamp, and M.A. Grima, "Application of neural networks for the prediction of the Unconfined Compressive Strength (UCS) from equotip hardness", Int. J. Rock Mech. Min. Sci., vol. 36, pp. 29-39. [http://dx.doi.org/10.1016/S0148-9062(98)00173-9] |
[44] | V.K. Singh, D. Singh, and T.N. Singh, "Prediction of strength properties of some schistose rocks from petrographic properties using artificial neural networks", Int. J. Rock Mech. Min. Sci., vol. 38, pp. 269-284. [http://dx.doi.org/10.1016/S1365-1609(00)00078-2] |
[45] | S. Dehghan, G.H. Sattari, C.S. Chehreh, and M.A. Aliabadi, "Prediction of unconfined compressive strength and modulus of elasticity for travertine samples using regression and artificial neural network", Min. Sci. Technol, vol. 20, pp. 0041-0046. |
[46] | M. Rezaei, A. Majdi, and M. Monjezi, "An intelligent approach to predict unconfined compressive strength of rock surrounding access tunnels in longwall coal mining", Neural Comput. Appl., vol. 24, pp. 233-241. [http://dx.doi.org/10.1007/s00521-012-1221-x] |
[47] | M. Monjezi, H.A. Khoshalan, and M. Razifard, "A neuro-genetic network for predicting uniaxial compressive strength of rocks", Geotech. Geol. Eng, vol. 30, pp. 1053-1062. [http://dx.doi.org/10.1007/s10706-012-9510-9] |
[48] | M. Beiki, A. Majdi, and A.D. Givshad, "Application of genetic programming to predict the uniaxial compressive strength and elastic modulus of carbonate rocks", Int. J. Rock Mech. Min. Sci., vol. 63, pp. 159-169. [http://dx.doi.org/10.1016/j.ijrmms.2013.08.004] |
[49] | S. Yagiz, E.A. Sezer, and C. Gokceoglu, "Artificial neural networks and nonlinear regression techniques to assess the influence of slake durability cycles on the prediction of uniaxial compressive strength and modulus of elasticity for carbonate rocks. Int", J. Numer. Anal. Met., vol. 36, pp. 1636-1650. [http://dx.doi.org/10.1002/nag.1066] |
[50] | M. Torabi-Kaveh, F. Naseri, S. Saneie, and B. Sarshari, "Application of artificial neural networks and multivariate statistics to predict UCS And E using physical properties of asmarilimestones", Arab. J. Geosci., . [http://dx.doi.org/10.1007/s12517-014-1331-0] |
[51] | N. Yesiloglu-Gultekin, C. Gokceoglu, and E.A. Sezer, "Prediction of uniaxial compressive strength of granitic rocks by various nonlinear tools and comparison of their performances", Int. J. Rock Mech. Min. Sci., vol. 62, pp. 113-122. [http://dx.doi.org/10.1016/j.ijrmms.2013.05.005] |
[52] | S. Kahraman, M. Fener, and E. Kozman, "Predicting the compressive and tensile strength of rocks from indentation hardness index", J. South. Afr. Inst. Min. Metall., vol. 112, no. 5, pp. 331-339. |
[53] | R. Altindag, and A. Guney, "Predicting the relationships between brittleness and mechanical properties (UCS, TS and SH) of rocks", Sci. Res. Essays, vol. 5, no. 16, pp. 2107-2118. |
[54] | C. Gokceoglu, and K. Zorlu, "A fuzzy model to predict the unconfined compressive strength and modulus of elasticity of a problematic rock", Eng. Appl. Artif. Intell., vol. 17, pp. 61-72. [http://dx.doi.org/10.1016/j.engappai.2003.11.006] |
[55] | K. Karaman, F. Cihangir, B. Ercikdi, A. Kesimal, and S. Demirel, "Utilization of the brazilian test for estimating the uniaxial compressive strength and shear strength parameters", J. South. Afr. Inst. Min. Metall., vol. 115, no. 3, pp. 185-192. [http://dx.doi.org/10.17159/2411-9717/2015/v115n3a3] |
[56] | E. Broch, and J.A. Franklin, "Point-load strength test", Int. J. Rock Mech. Min. Sci., vol. 9, pp. 241-246. [http://dx.doi.org/10.1016/0148-9062(72)90030-7] |
[57] | Z.T. Bieniawski, "Point load test in geotechnical practice", Eng. Geol., vol. 9, no. 1, pp. 1-11. [http://dx.doi.org/10.1016/0013-7952(75)90024-1] |
[58] | S. Kahraman, O. Gunaydin, and M. Fener, "The effect of porosity on the relation between uniaxial compressive strength and point load index", Int. J. Rock Mech. Min. Sci., vol. 42, pp. 584-589. [http://dx.doi.org/10.1016/j.ijrmms.2005.02.004] |
[59] | A. Basu, and A. Aydin, "Predicting uniaxial compressive strength by point load test: Significance of cone penetration", Rock Mech. Rock Eng., vol. 39, pp. 483-490. [http://dx.doi.org/10.1007/s00603-006-0082-y] |
[60] | I. Yilmaz, and A.G. Yuksek, "An example of Artificial Neural Network (ANN) application for indirect estimation of rock parameters", Rock Mech. Rock Eng., vol. 41, no. 5, pp. 781-795. [http://dx.doi.org/10.1007/s00603-007-0138-7] |
[61] | K. Diamantis, E. Gartzos, and G. Migiros, "Study on uniaxial compressive strength, point load strength index, dynamic and physical properties of serpentinites from central greece: Test results and empirical relations", Eng. Geol., vol. 108, pp. 199-207. [http://dx.doi.org/10.1016/j.enggeo.2009.07.002] |
[62] | S. Kahraman, "The determination of uniaxial compressive strength from point load strength for pyroclastic rocks", Eng. Geol., vol. 170, pp. 33-42. [http://dx.doi.org/10.1016/j.enggeo.2013.12.009] |
[63] | K. Sarkar, V. Vishal, and T.N. Singh, "An empirical correlation of index geomechanical parameters with the compressional wave velocity", Geotech. Geol. Eng., . [http://dx.doi.org/10.1007/s10706-011-9481-2] |
[64] | A. Kilic, and A. Teymen, "Determination of mechanical properties of rocks using simple methods", Bull. Eng. Geol. Environ., vol. 67, pp. 237-244. [http://dx.doi.org/10.1007/s10064-008-0128-3] |
[65] | I. Cobanoglu, and S. Celik, "Estimation of uniaxial compressive strength from point load strength, schmidt hardness and p-wave velocity", Bull. Eng. Geol. Environ., vol. 67, pp. 491-498. [http://dx.doi.org/10.1007/s10064-008-0158-x] |
[66] | A. Tuğrul, and I.H. Zarif, "Correlation of mineralogical and textural characteristics with engineering properties of selected granitic rocks from Turkey", Eng. Geol., vol. 51, no. 4, pp. 303-317. [http://dx.doi.org/10.1016/S0013-7952(98)00071-4] |
[67] | A. Aydin, and A. Basu, "The schmidt hammer in rock material characterization", Eng. Geol., vol. 81, pp. 1-14. [http://dx.doi.org/10.1016/j.enggeo.2005.06.006] |
[68] | V. Gupta, "Non-destructive testing of some higher himalayan rocks in The Satluj Valley", Bull. Eng. Geol. Environ., vol. 68, pp. 409-416. [http://dx.doi.org/10.1007/s10064-009-0211-4] |
[69] | P.K. Sharma, and T.N. Singh, "A correlation between P-Wave velocity, impact strength index, slake durability index and uniaxial compressive strength", Bull. Eng. Geol. Environ., vol. 67, pp. 17-22. [http://dx.doi.org/10.1007/s10064-007-0109-y] |
[70] | Z.A. Moradian, and M. Behnia, "Predicting the uniaxial compressive strength and static young’s modulus of intact sedimentary rocks using the ultrasonic test", Int. J. Geomech., vol. 9, pp. 1-14. [http://dx.doi.org/10.1061/(ASCE)1532-3641(2009)9:1(14)] |
[71] | M. Khandelwal, "Correlating P-Wave velocity with the physicomechanical properties of different rocks", Pure Appl. Geophys., vol. 170, pp. 507-514. [http://dx.doi.org/10.1007/s00024-012-0556-7] |
[72] | R. Altindag, "Correlation between P-Wave velocity and some mechanical properties for sedimentary rocks", J. South. Afr. Inst. Min. Metall., vol. 112, no. 3, pp. 229-237. |
[73] | S. Xu, P. Grasso, and A. Mahtab, "Use of schmidt hammer for estimating mechanical properties of weak rock", Proc. 6th International IAEG Congress, vol. 1, Balkema: Rotterdam, pp. 511-519. |
[74] | A. Jamshidi, H. Zamanian, and R.Z. Sahamieh, "The effect of density and porosity on the correlation between uniaxial compressive strength and P-wave velocity", Rock Mech. Rock Eng., vol. 51, no. 4, pp. 1279-1286. [http://dx.doi.org/10.1007/s00603-017-1379-8] |
[75] | D.C. Entwisle, R.N. Hobbs, L.D. Jones, D. Gunn, and M.G. Raines, "The relationship between effective porosity, uniaxial compressive strength and sonic velocity of intact borrowdale volcanic group core samples from sellafield", Geotech. Geol. Eng., vol. 23, pp. 793-809. [http://dx.doi.org/10.1007/s10706-004-2143-x] |
[76] | A.K. Verma, and T.N. Singth, "Aneuro-fuzzy approach for prediction of longitudinal wave velocity", Neural Comput. Appl., vol. 22, no. 7–9, pp. 1685-1693. [http://dx.doi.org/10.1007/s00521-012-0817-5] |
[77] | R. Singh, V. Vishal, T.N. Singh, and P.G. Ranjith, "A comparative study of generalized Regression neural network approach and adaptive neuro-fuzzy inference Systems for prediction of unconfined compressive strength of rocks", Neural Comput. Appl., vol. 23, pp. 499-506. [http://dx.doi.org/10.1007/s00521-012-0944-z] |
[78] | M. Rezaei, and M. Asadizadeh, "Predicting unconfined compressive strength of intact rock using new hybrid intelligent models", J. Min. Environ., vol. 11, no. 1, pp. 231-246. |
[79] | B. Saedi, S.D. Mohammadi, and H. Shahbazi, "Prediction of uniaxial compressive strength and elastic modulus of migmatites using various modeling techniques", Arab. J. Geosci., vol. 11, p. 574. [http://dx.doi.org/10.1007/s12517-018-3912-9] |
[80] | E. Rabbani, and F. Sharif, "Application of neural network technique for prediction of uniaxial compressive strength using reservoir formation properties", Int. J. Rock Mech. Min. Sci., vol. 56, pp. 100-111. [http://dx.doi.org/10.1016/j.ijrmms.2012.07.033] |
[81] | L.K. Sharma, V. Vishal, and T.N. Singh, "Developing novel models using neural networks and fuzzy systems for the prediction of strength of rocks from key geomechanical properties", Measurement, vol. 102, pp. 158-169. [http://dx.doi.org/10.1016/j.measurement.2017.01.043] |
[82] | N. Ceryan, U. Okkan, and A. Kesimal, "Prediction of unconfined compressive strength of carbonate rocks using artificial neural networks", Environ. Earth Sci., vol. 68, pp. 807-819. [http://dx.doi.org/10.1007/s12665-012-1783-z] |
[83] | K. Zorlu, C. Gokceoglu, F. Ocakoglu, H.A. Nefeslioglu, and S. Acikalin, "Prediction of uniaxial compressive strength of sandstones using petrography-based models", Eng. Geol., vol. 96, pp. 141-158. [http://dx.doi.org/10.1016/j.enggeo.2007.10.009] |
[84] | R. Jahanbakhshi, R. Keshavarzi, and M.J. Azinfar, "Intelligent prediction of uniaxial compressive strength for sandstone", 45^{th} US Rock Mechanics/Geomechanics Symposium, American Rock Mechanics Association, . |
[85] | A. Majidi, and M. Rezaei, "Prediction of unconfined compressive strength of rock surrounding a roadway using artificial neural network", Neural Comput. Appl., vol. 23, pp. 381-389. [http://dx.doi.org/10.1007/s00521-012-0925-2] |
[86] | K. Sarkar, A. Tiwary, and T.N. Singh, "Estimation of strength parameters of rock using artificial neural networks", Bull. Eng. Geol. Environ., vol. 69, pp. 599-606. [http://dx.doi.org/10.1007/s10064-010-0301-3] |
[87] | Q. Fang, B.Y. Bejarbaneh, and M. Vatandoust, "Strength evaluation of granite block samples with different predictive models", Eng. Comput., pp. 1-18. [http://dx.doi.org/10.1007/s00366-019-00872-4] |
[88] | D.A. Mishra, and A. Basu, "Estimation of uniaxial compressive strength of rock materials by index tests using regression analysis and fuzzy inference system", Eng. Geol., vol. 160, pp. 54-68. [http://dx.doi.org/10.1016/j.enggeo.2013.04.004] |
[89] | D. Jahed Armaghani, E. Tonnizam Mohamad, E. Momeni, M. Sundaram, and M. A. Mohd For, "An adaptive neuro-fuzzy inference system for predicting unconfined compressive strength and young’s modulus: A study on main range granite", Bull. Eng. Geol. Environ., vol. 74, pp. 1301-1319. [http://dx.doi.org/10.1007/s10064-014-0687-4] |
[90] | E.T. Mohamad, and D.J. Armaghani, "Rock strength estimation: A PSO-based BP approach", Neural Comput. Appl., vol. 30, pp. 1635-1646. [http://dx.doi.org/10.1007/s00521-016-2728-3] |
[91] | G. Dreyfus, Neural Networks: Methodology And Application., Springer Berlin Heidelberg: Germany, . |
[92] | B. Tiryaki, "Predicting intact rock strength for mechanical excavation using multivariate statistics, artificial neural networks and regression trees", Eng. Geol., vol. 99, pp. 51-60. [http://dx.doi.org/10.1016/j.enggeo.2008.02.003] |
[93] | M.T. Mohamed, "Artificial neural network for prediction and control of blasting vibrations in Assiut (Egypt) limestone quarry", Int. J. Rock Mech. Min. Sci., vol. 46, pp. 426-431. [http://dx.doi.org/10.1016/j.ijrmms.2008.06.004] |
[94] | A. Cevik, E.A. Sezer, A.F. Cabalar, and C. Gokceoglu, "Modeling of the uniaxial compressive strength of some clay-bearing rocks using neural network", Appl. Soft Comput., vol. 11, pp. 2587-2594. [http://dx.doi.org/10.1016/j.asoc.2010.10.008] |
[95] | M. Hajihassani, D. Jahed Armaghani, H. Sohaei, E. Tonnizam Mohamad, and A. Marto, "Prediction of airblast-overpressure induced by blasting using a hybrid artificial neural network and particle swarm optimization", Appl. Acoust., vol. 80, pp. 57-67. [http://dx.doi.org/10.1016/j.apacoust.2014.01.005] |
[96] | R. Eberhart, and J. Kennedy, "A new optimizer using particle swarm theory", Proc 6^{th} International Symposium on Micro Machine and Human Science, pp. 39-43.Nagoya, Japan4–6 Oct 1995 [http://dx.doi.org/10.1109/MHS.1995.494215] |
[97] | R. Mendes, P. Cortes, M. Rocha, and J. Neves, "Particle swarms for feed forward neural net training", Proc IEEE Int Joint Conf on Neural Networks, pp. 1895-1899.Honolulu, HI, USA |
[98] | T. Victoire, and A.E. Jeyakumar, "Hybrid PSO–SQP for economic dispatch with valve-point effect", Electr. Power Syst. Res., vol. 71, pp. 51-59. [http://dx.doi.org/10.1016/j.epsr.2003.12.017] |
[99] | R. Poli, J. Kennedy, and T. Blackwell, "Particle swarm optimization", Swarm Intell., vol. 1, no. 1, pp. 33-57. [http://dx.doi.org/10.1007/s11721-007-0002-0] |
[100] | Y. Shi, and R.C. Eberhart, "Empirical study of particle swarm optimization", Proceedings of the 1999 IEEE Congress on Evolutionary ComputationIEEE: New York, . 1999, pp. 1945-1950. [http://dx.doi.org/10.1109/CEC.1999.785511] |