Fuzzy Clustering Analysis And Validity On Dominant Orientations Of Rock Mass Dicontinuity

The rock mass is a complex composed of the structural blocks of rock blocks and rock fractured or discontinuities.Generally,there are plenty of discontinuities existing in the rock mass.Because of the existence of discontinuities,the strength,stability and impermeability of rock mass are far less than that of complete rock.It is important to establish a three-dimensional network model of rock mass and to scientifically evaluate the engineering characteristics of rock mass by statistical analysis and partitioning of dominant discontinuities orientation.Therefore,it is very important to study the distribution of the discontinuity of the rock mass and extract the dominant orientation of discontinuities to carry on the rock engineering research and analysis.The traditional partition method of discontinuities dominant orientation have the rose diagram and pole iso-dense diagram,but these methods are subjectively different from the proficiency of engineers,can only get qualitative grouping results.Based on the mathematical model of discontinuities,this paper constructs the objective function based on the fuzzy C-means algorithm(FCM)to quantitatively group the dominant orientation of discontinuities.In this paper,a new method of discontinuities dominance group partitioning based on firefly fuzzy C means algorithm(FFCM)is proposed,and the corresponding program is programed in order to partition discontinuities automatically.Based on the fuzzy decision theory,the fuzzy weighting index is optimized.The noise data of discontinuities data set is studied,and a method of noise data processing is proposed.A heuristic intelligent algorithm—firefly algorithm(FA)is introduced,and the parameter of the FA algorithm are improved.The validity index of the clustering algorithm is studied,and the validity of the clustering is evaluated by the double validity index.In order to systematically vertify the performance of partitioning of proposed method,a simulated data using Monte-Carlo method is applied to the proposed method.And two field rock discontinuities orientation is grouped using the proposed method.By analyzing the final results,we can conclude that the proposed method can effectively solve defects of the traditional methods like it needs to specify the number of groups in advance,and the clustering result are extremely sensitive to the initial value and noise data.The results obtained using the proposed method are more accurate and consistent with the actual situation.