Application of data clustering algorithm in analyzing structural features of rock mass

When dealing with problems of geotechnology, it is nec essary to select a method for processing structural data in conformity with such criteria as unambiguous interpretation of the result, minimum subjectivity inflicted by a specialist, and data processing speed. The article presents a cluster analysis algorithm for structural data of geotechnical oriented borehole drilling. Clustering allows structuring and interpreting complex and manydimensional data, which is particularly important when studyi ng nonuniform natural systems. These systems have a complex spatial geometry describable by azimuth and dip angle among other things. The analysis of such data needs methods capable to reveal hidden laws and to generate uniform groups of data for the further examination. Clustering involves two stages. The first stage is preliminary processing of data, including a dimensional scaling procedure and standardization. Reduction of values of attributes to a uniform scale is a necessary condition of the correction calculation of distances in a manydimensional attribute space. The second stage is clustering using k-means. The latter is selected because of its efficiency in handling a large bulk of data. The k-means minimizes intracluster dispersion, and ensures compactedness and discriminability of distinguished groups. Furthermore, k-means can point at the natural groups of objects. Statistical processing of structural data, in particular, the clustering analysis, enables sound decision-making on clustering of oriented structures or systems of cracks, makes it possible to decrease the degree of the human subjectivity in detection of systems of cracks, and provides repeatability of the process at different stages of such data analysis, which is important in geotechnical research.