Evaluation of rock slope stability conditions through discriminant analysis.

• Autor(a) principal: Santos, Allan Erlikhman Medeiros
• Data de Publicação: 2018
• Outros Autores: Lana, Milene Sabino, Cabral, Ivo Eyer, Pereira, Tiago Martins, Naghadehi, Masoud Zare, Silva, Denise de Fátima Santos da, Santos, Tatiana Barreto dos
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UFOP
• Texto Completo: http://www.repositorio.ufop.br/handle/123456789/11109
• topic: Multivariate statistics; Principal component analysis; Boosting technique

A methodology to predict the stability status of mine rock slopes is proposed. Two techniques of multivariate statistics are used: principal component analysis and discriminant analysis. Firstly, principal component analysis was applied in order to change the original qualitative variables into quantitative ones, as well as to reduce data dimensionality. Then, a boosting procedure was used to optimize the resulting function by the application of discriminant analysis in the principal components. In this research two analyses were performed. In the first analysis two conditions of slope stability were considered: stable and unstable. In the second analysis three conditions of slope stability were considered: stable, overall failure and failure in set of benches. A comprehensive geotechnical database consisting of 18 variables measured in 84 pit-walls all over the world was used to validate the methodology. The discriminant function was validated by two different procedures, internal and external validations. Internal validation presented an overall probability of success of 94.73% in the first analysis and 68.42% in the second analysis. In the second analysis the main source of errors was due to failure in set of benches. In external validation, the discriminant function was able to classify all slopes correctly, in analysis with two conditions of slope stability. In the external validation in the analysis with three conditions of slope stability, the discriminant function was able to classify six slopes correctly of a total of nine slopes. The proposed methodology provides a powerful tool for rock slope hazard assessment in open-pit mines.