A new clustering method of rock discontinuity sets based on modified K-means algorithm

Identification of rock discontinuities caused by orientation is an important basis for rock mechanics analysis and slope stability assessment. In order to obtain more objective and accurate clustering results, a modified K-means clustering algorithm based on genetic algorithm and simulated annealing algorithm is proposed, which overcomes the shortcomings of traditional K-means algorithm and realizes global optimization. A similarity measurement based on the negative sine-squared value of the acute angle between discontinuity unit normal vector was used to replace the original Euclidean distance measurement. In addition, the Xie-Beni validity index and the Davies-Bouldin validity index were introduced to determine the optimal clustering number. The validity of the method was analyzed on artificial and literature data where the data are synthetic and the clustering results can be clearly grasped; the results show that the maximum clustering error of the artificial data is 5.87% and the minimum is 0.24%. Meanwhile, the proposed method was used to clustering the discontinuities of rocky slopes on National Highway 317 in Sichuan Province, Southwest China. The results clearly demonstrate that the method achieved good and realistic clustering results with stronger robustness than other methods.