Probabilistic slope stability assessment of laterite borrow pit using artificial neural network

Abstract

ABSTRACT Assessment of slope stability of abandoned laterite borrow pits in residential areas is highly desirable as theconsequence of its failure could be fatal. This study applied artificial neuralnetwork (ANN) to conduct probabilistic slope stability assessments of theborrow pits. To determine the corresponding factor of safety (FOS), randomshear strength parameters, slope geometry, structure load on the slope andstructure distance from the slope crest were used as inputs infinite-difference numerical simulations. The FOS was combined with ANNtechniques to derive a mathematical model for predicting the failureprobability. The effects of variability of soil shear strength parameters andcross-correlation between the parameters on the probability of slope failurewere examined. Results showed that the performance level of the pit slopes washazardous. Variability in shear strength parameters significantly influencedthe slope stability, while negative correlation coefficients between theparameters reduced the probability of the slope failure.