A new soil classification system for the quadrilátero ferrífero province using multivariate statistical analysis

Abstract

The Unified Soil Classification System (USCS) is widely used in geotechnical engineering, but it has limitations in classifying tropical soils due to characteristics such as the presence of saprolites or laterites. This study proposes a regional soil classification system for the mineral province of the Quadrilátero Ferrífero, in Brazil, aiming to enhance the understanding of soil behavior in the region. A total of 101 soil samples, both natural and compacted, were analyzed, including variables such as effective friction angle, cohesion, plasticity index, specific gravity of particles, and fines content. Principal Component Analysis (PCA) and k-means clustering were used to develop the proposed classification system. This categorizes the soils into three distinct classes (A, B, and C) based on their geotechnical parameters. The analysis showed that the proposed system outperforms the USCS in differentiating soil behavior in the Quadrilátero Ferrífero province, establishing a classification chart that explains 81.68% of the variability of the analyzed parameters. Compared to the USCS, the new system provides a more accurate tool for predicting soil behavior, being useful in foundation engineering, excavation projects, and other geotechnical applications in the region.