A preliminary large scale validation of the Stochastic Approach for Discontinuity Shear Strength

The shear strength of large discontinuities is an important parameter in stability assessment and design of rock slopes, especially where infrastructure, operations and/or human life may be at risk. However, estimating the shear strength of large in situ discontinuities is not trivial. Researchers and practitioners either test laboratory sized samples and/or employ empirical methods for estimating shear strength of large discontinuities. This in turn introduces another challenge, accounting for change of scale commonly referred to as the “scale effect”. The effects of scale on discontinuity properties have been continually studied since the 1960s. However, there is currently no consensus or a satisfactory method to account for surface variability and manage the ‘scale effect’ phenomenon. A few approaches have been established for estimating shear strength of large in situ discontinuities, among which the well-known Barton-Bandis model, but these approaches or models carry application and prediction reliability limitations and/or concerns. In recent years, a new stochastic approach was established and offers a way to estimate the shear strength of in situ discontinuities. Applicable at problem scale, it can bypass potential scale effect issues. The key concept of the stochastic approach is that a daylighting discontinuity profile possesses roughness information, which is representative of the whole in situ surface. Employing random field modelling and a shear strength model capable of processing 3D surfaces within a Monte Carlo framework, the stochastic approach produces a distribution of peak and residual shear strength from which stochastic design parameters can be obtained. The stochastic approach was successfully validated at small and a preliminarily application to a 4 m2 discontinuity surface returned promising results but lacked experimental data for validation. The foundation research initiated a continued line of research with an emphasis on large-scale application development and validation. The first part of the thesis addresses issues encountered with generating large scale synthetic surfaces using random field modelling, notably the mismatch of the asperity gradient distributions between the seed trace and generated surfaces. This mismatch translated to inaccurate stochastic predictions of shear strength. Acknowledging the multiscale nature of roughness, a new rigorous multiscale approach using random field modelling was developed and validated using 25 seed traces. The thesis goes on to presents the process undertaken to create 2 m by 2 m mortar discontinuity replicas, the design and characteristics of the large-scale shear device and the experimental results of the first series of direct shear tests. The large scale experimental direct shear data is needed to validate the employed semi-analytical discontinuity shear strength model and the stochastic approach’s prediction capabilities. The last part of this thesis deals with the first rigorous application of the employed semi-analytical discontinuity shear strength model and the stochastic approach on the 2 m by 2 m discontinuity replica surface, with comparison of stochastic predictions to experimental shear strength data. The shear strength model was observed to produce satisfactory deterministic peak and residual shear strength predictions. The predictive capability of the stochastic approach was explored using 16 seed traces. Using 100 synthetic surfaces for the stochastic analysis, it was observed that, for seed traces with gradient statistics equivalent to that of the surface, the predictions closely resemble the experimental results. Predicting shear strength from different seed traces (ones with different gradient statistics) resulted in more variability of predictions. However, the stochastic approach was found to better predict the shear strength of the experimental surface compared to the Barton-Bandis model, demonstrating its potential.