Finite element modelling of a shear box experiment on ice rubble

Analytical and numerical analysis of laboratory-scale shear box experiments on ice rubble have been performed. The experiments were done at NTNU in 2010 (Serré et al., 2011 and Repetto and Høyland, 2011). Three phases in the force-time history of the tests were identified (similar to what Serré et al., 2011 did). The first phase lasted about 2 seconds (compared to 6 seconds in Serré et al., 2011) and ended with a shift in the force-time gradient corresponding to plastic behaviour. Phase one was studied numerically while phase two and three were analysed conceptually. The experiments were done with different vertical confinement and different submersion times, altogether 14 individual tests were carried out. Each combination of parameters (initial and boundary conditions) was repeated three times. There was a large difference within the three tests with the same initial and boundary conditions. Mostly 2 tests were similar, while one test was different. The use of Hellmann’s (1984) definition of 3 phases in shear box testing cannot directly be used to analysis these tests, as the boundary conditions were different. Hellmann had zero displacement perpendicular to the moving piston, while Serré et al. (2011) had constant pressure. The friction angle was derived from phase three assuming that all freeze-bonds were broken and that the rubble was essentially without cohesion. The test with 20 hours (long) submersion times had lower friction angle (9? to 31?) than the ones with 10 minutes (short) submersion times (33? to 74?). This derivation is only indicative, but suggests that the submersion time govern the material properties. Phase 1 was studied numerically with the use of a continuum finite element model with Lagrange method (ABAQUS/Standard 6.10-2). An elastic-perfect plastic model with Drucker-Prager material model was applied, and the three most important material parameters, namely Young modulus, cohesion and friction angle were the main focus. These three were retrieved by trial and error method so as to match the experimental results. Young modulus give the initial slope of the force-time curve, while DP model cohesion and friction angle determine when the yielding points occurred and the force at that moment. So by carrying out numerical simulation, the values of Young modulus for different situations were given and a series pairs of cohesion and friction angle in DP model were given as well so as to match the experiment results. The long submersion times had lower Young modulus (0.9 MPa) than the short submersion times gave (2-4.5 MPa). But which pair of cohesion and friction angle that was the most suitable for the ice rubble material in nature could not be derived from the present study. The numerical simulation showed that the crack patterns predicted by numerical simulation were more or less the same as the ones observed. The question on whether or not a cap hardening model should be used to simulate the primary phase was also discussed and only in the Extra_High_Short test, a cap model is needed. Scaling problems were discussed by assuming that Froude scaling law could be used and it is found that the cohesion after scaling up to full-scale was larger than two recent studies: one is model-scale punch tests done by Serré (2010), another is full-scale punch tests done by Heinonen (2004). The results of this thesis could provide a better understanding for the material properties of freeze-bonds and be beneficial for future study in this field. The conceptual analysis of phase two and three showed that there were some shortcomings in the experiment setup and based on this, some improvement of the experiment equipment and suggestions were given.CoMEM - Coastal and Marine Engineering and ManagementSection Hydraulic EngineeringCivil Engineering and Geoscience