عنوان مقاله [English]
Stability of open stopes in hard rock underground mining is typically evaluated using the stability graph method. Due to the empirical nature of this method, lack of an unique graph for various conditions and considering the same conditions for the various surfaces of the stope, the risk of wrongly interpreting the results is high. In this paper new models to evaluate stability state of open stopes based upon input parameters of stability graph method were developed separately for roof and walls using logistic regression and support vector machine (SVM). For this purpose, a database was established containing conditions of roof and walls from open stope mines in Canada and Ghana. The results indicated that the accuracy of stability graph method, LG model and SVM model in prediction of roof stability were 29%, 86% and 95%, respectively. In addition, the accuracy of wall stability condition prediction using stability graph method, LG model and SVM model were 71%, 81% and 90%, respectively. These results confirm that the performance of developed model is better than conventional stability graph method. Besides, it was concluded that SVM models possesses a higher performance in stability evaluation when compared to the LG models. Findings of this paper show that separation of different open stopes surfaces as well as incorporating statistical and intelligent methods in stability evaluation increase the reliability of predictions in comparison with conventional stability graph method.