Faults with a crush zone can strongly affect the mechanical, geochemical, and hydrological properties of a rock mass. Because of this, fault zones are treated as essential elements for evaluating the underground geological environment and the engineering performance of rocks. Because of the limitations to borehole investigations, it is not always possible to obtain sufficient, high-quality geological data. In addition, the evaluation of results may differ depending on various factors such as geological conditions and skill of the engineer or geologist. Such uncertainty can lead to difficulty in evaluation and understanding of the geological environment at depth and in the decision-making and planning of underground construction, which, as a result, may increase potential risks during construction. To reduce the uncertainty, this study proposes a data selection method using multivariate analysis composed of principal component analysis and a clustering method using data from a deep borehole investigation in the Mizunami Underground Research Laboratory (Mizunami City, Gifu, Central Japan). Utilizing this method and the analyses, the rocks could be accurately classified depending upon their geological characteristics. It was also possible to discriminate subtle differences in the rockmass. Furthermore, the location and width of fault zones were determined objectively. Accordingly, multivariate analysis, in considering the variety of data from different sources, proved to be more effective than traditional methods that analyze items individually. Moreover, a method to rank the variables used in the principal component analysis was developed using a logical and quantitative index that can arrange the variables in their order of importance. The proposed method developed in this study can provide useful geological and engineering information for 3-D geological modeling, construction of underground structures and groundwater flow analysis.
View full abstract