Probabilistic Translational Stability of Multilayered Cover System Based on Limited Experimental Data

Abstract

Multilayered cover systems (MLCSs) of waste containment facilities are generally prone to translational instabilities, largely governed by their interfacial shear strength parameters. For reliability assessment of a MLCS, it is necessary to characterize the interfacial shear strength parameters probabilistically. Owing to the uncertainties involved with the sample preparation and subsequent laboratory experimentation, estimation of shear strength parameters based on deterministic linear regression becomes an oversimplifying paradigm. Conducting a very large number of repeated experimental trials to estimate shear strength parameters is practically infeasible; thus, geotechnical engineers often resort to best possible inferences from limited data. This necessitates a Bayesian regression–based approach catering to limited data for layerwise estimates of interfacial shear strength parameters of the MLCS. Informative (normal) and noninformative (uniform) priors are utilized to obtain updated probability distributions of shear strength parameters of the interfaces. These estimates are subsequently adopted in reliability assessment of a MLCS. A novel conceptual paradigm of a probabilistic vulnerable interface diagram (PVID) is introduced to identify the most probable vulnerable interfaces of the MLCS when subjected to several factors leading to instabilities.