STABILITY CONDITIONS EVALUATION OF SLOPE BY MULTIVARIATE ANALYSIS

Abstract

Technological advances have contributed to applications of nonparametric methodologies with the objective of predicting slope stability conditions. The objective of this paper was to determine a discriminant function capable of predicting the stability condition of the slopes of the database under study. It is important to note that the methodology does not replace the stability analysis, but it can work very well for a preliminary analysis by selecting the slopes that must be intervened. The database used is composed by 59 slopes with relevant parameters in slope stability analysis with circular failure. A combination of multivariate statistical techniques, specifically principal component analysis (PCA) and discriminant analysis, was used to determine the slope stability condition. The principal component analysis was used to reduce the dimensionality of the database. The discriminant analysis was used to determine the boundary between stability conditions. Two types of discriminant function validations were performed, cross validation and external validation. The cross validation presented a global probability of success of 89.83%, the errors obtained in the cross validation were in favor of safety, with 5 stable slopes classified as unstable and only 1 unstable slope classified as stable. In the external validation were used 12 new slopes, which 8 slopes were correctly classified correctly.

Keywords

• stability condition prediction, multivariate analysis, principal component analysis, discriminant analysis, non-parametric techniques.