DEMIN - Departamento de Engenharia de Minas
URI permanente desta comunidadehttp://www.hml.repositorio.ufop.br/handle/123456789/510
Navegar
2 resultados
Resultados da Pesquisa
Item Evaluation of machine learning methods for rock mass classification.(2022) Santos, Allan Erlikhman Medeiros; Lana, Milene Sabino; Pereira, Tiago MartinsSolutions in geotechnics have been optimizing with the aid of machine learning methods. The aim of this paper is to apply different machine learning algorithms in order to achieve rock mass classification. It is demonstrated that RMR classifi- cation system can be obtained using only variables which are closely related to rock mass quality, instead of all RMR variables, without missing significant accuracy. The different machine learning algorithms used are the naı ̈ve Bayes, random forest, artificial neural networks and support vector machines. The variables to calculate RMR, selected by factor analysis, are: rock strength, rock weathering, spacing, persistence and aperture of discontinuities and presence of water. The machine learning models were trained and tested thirty times, with random subsampling, using two-thirds of the total database for training sample. The models presented average accuracy greater than 0.81, which was calculated from the confusion matrix, using the proportion of true positives and true negatives in the test sample. Significant values of efficiency, precision and reproducibility rates were achieved. The study shows the application of machine learning algorithms allows obtaining the RMR classes, even with a small number of variables. In addition, the results of the evaluation metrics of the developed algorithms show that the methodology can be applied to new database, working as a valuable way to achieve rock mass classification.Item Rock mass classification by multivariate statistical techniques and artificial intelligence.(2020) Santos, Allan Erlikhman Medeiros; Lana, Milene Sabino; Pereira, Tiago MartinsThis study aims to improve the quality and accuracy of RMR classification system for rock masses in open pit mines. A database of open pit mines comprising basic parameters for obtaining the RMR was used. Techniques applied in this research were multivariate statistics and artificial intelligence. In relation to multivariate statistics, factor analysis was capable of identifying underlying factors not observable in the original variables, using the variables of these factors in the classification system, instead of all RMR variables. The proposed classifier was obtained by training neural networks. The results of the factor analysis allowed the identification of three common factors. Factor 1 represents the strength and weathering of the rock mass. Factor 3 represents the fracturing degree of the rock mass. Finally Factor 2 represents water flow conditions. Thirty artificial neural networks were trained with randomly selected training samples. The trained networks proved to be effective and stable. Regarding the validation of the networks, the values obtained for the overall probability of success and apparent error rate showed normal distributions and a low dispersion rate, with average rates of 0.87 and 0.13, respectively. Regarding specific errors, error values were recorded only between contiguous RMR classes. The major contribution of the study is to present a new methodology for achieving rock mass classifications based on mathematical and statistical fundamentals, aiming at optimising the selection of variables and consequent reduction of subjectivity in the parameters and classification methods.