DETECTING FAILURE SURFACE DEVELOPMENT OF LABORATORY-SCALE LANDSLIDES USING PARTICLE IMAGE VELOCIMETRY

Understanding the state of stress within steep hillslopes is essential for studying shallow landsliding mechanisms. Failure planes develop where shear strength of the material is exceeded following the Mohr-Coulomb criterion. Detecting incipient motion of the failure plane would improve understanding of failure surface propagation for given stress conditions, but is not well-documented. In this study, we simulated an unsaturated hillslope in the lab and controlled the stress conditions until failure to observe particle motion throughout the experiment. Dry, medium-coarse sand was compacted in an acrylic box 62 cm long, 40 cm high, and 20 cm wide, to a uniform bulk density (1.49 g/cm³). The volume was brought to a slope of 35^o by removing a sliding panel on the box. A pedestal 4 cm long was fixed to the slope crest, and sand was added to a receptacle on the pedestal at the constant rate of 6.25 g/s until visible failure of the slope occurred. Digital photographs of the slope were taken at constant time intervals during the experiment through the front of the box. A recently-developed particle image velocimetry (PIV) tool designed to detect soil deformation divided each digital image into a mesh of 30- by 30-pixel patches, and cross-correlated the locations of patches between two images to calculate displacement vectors in a 2-D cross section of slope. Results of the PIV analysis yielded incipient particle motion immediately below the crest and inward from the slope surface well before visible failure occurred. Deformation propagated downward and toward the toe of slope as the experiment progressed with a sharp interface between relatively high-magnitude vectors, and areas showing zero displacement. Post-failure location of the calculated interface in the images coincided with that of the failure plane in the experimental setup. The geometry of the slope with different normal stress values at the crest was input into a new finite-element slope stability model. Areas of the slope where motion was first detected from the PIV analysis were calculated to have local factors of safety of less than 1.0 for the same stress conditions. Our results show that the PIV technique can be used to detect incipient motion of a failure surface in a simulated slope failure event.