Hydraulic fracturing through the core of earth dams is a known trigger mechanism for internal erosion and can lead to catastrophic failure of the structures. Many of the previous laboratory studies on hydraulic fracturing in soils have used materials with high fines contents (>50%). Here in New Zealand, there are large earth dams constructed with glacial till cores. These till materials are usually composed of widely graded sands and gravels, with much lower fines contents (10% - 40%}. This study aims to fill this gap in the literature by investigating how this wider particle size distribution influences hydraulic fracture behaviour.

To achieve this aim, a laboratory testing programme was conducted using a purpose-built Hydraulic Fracture Cell (HFC}. The HFC is a modified true triaxial apparatus capable of testing samples containing particles in the gravel size range (300 mm cubical sample size). Bulk material recovered from an active hydropower dam was screened to three different particle size distributions for testing: Material A (36% fines, Dmax = 2.36 mm, Cc= 1.1), Material B (25% fines, Dmax = 9.5 mm, Cc= 3.5), and Material C (16% fines, Dmax = 19 mm, Cc= 4.9}.

The outputs from the hydraulic fracture testing were fluid injection pressure plotted against injected volume (at a constant flow rate). The pressure curves for Materials A and B were similar - both in terms of the magnitudes of fracture initiation and the overall shape. These results also matched the general shape of previously published curves for fine-grained materials (>50% fines). Material C, however, behaved differently. No discernible peak was observed in the fluid injection pressure vs. injected volume curves. Instead, after breaking from the near linear rate of rise, the fluid pressure remained constant with increased injected volume.

The similar results for Material A, B and previous studies using fine-grained materials suggest these materials failed by hydraulic fracturing. Material C appeared to be susceptible to a different failure mechanism approaching seepage flow. The differences in particle size distribution from Material B to Material C changed the way the material dissipated fluid pressure at the tested injection flow rate. It is thought that this was caused by a) an increase of hydraulic conductivity, and b) a change in the matrix structure causing a lower effective stress transfer factor (a> 1). For the widely graded materials tested at an injection rate of 10 ml/min, the transition point between hydraulic fracture and seepage behaviour types is between Material B (26% fines content) and Material C (16% fines content).