Mechanical properties of model ice ridge keels

Rubble ice presents pressure dependent yield strength and its behaviour can be described by mathematical models based on several mechanical parameters. They are investigated for HSVA model rubble ice through the analysis of three different tests: the oedometer test, the pile test and the punch test. This last test is analysed with the non-linear Eulerian finite element method. The tests were performed on 4 ice ridges with two different submersion times. A 0.5 to 1.2 kPa model scale Mohr–Coulomb cohesion (0.6 to 1.5 kPa Drucker–Prager cohesion), depending on the ridge history, was used in the simulations of the model scale punch tests. The friction angle is estimated between 30 and 45° (40 and 50° Drucker–Prager friction angle). The upper value was used in the punch test simulations. A 0.9 MPa Young modulus was derived and the hydrostatic compressive yield curve was determined. The numerical model is able to estimate the rubble action during the entire penetration of the punch test in the keel and it is shown that a cohesive softening occurs in the rubble. In order to reproduce the experimental load time series for the short submersion time ridges it was necessary to use a vertical distribution of the cohesion representing the vertical distribution of the freeze-bond strength. A sensitivity analysis of the punch test shows that the keel depth and the ice density are the main parameters governing the keel frictional resistance. A precise determination of these parameters is therefore crucial for a correct determination of the rubble mechanical properties from the numerical simulation of experimental punch tests. The punch test is not appropriate for the determination of the friction angle due to the low confinement pressure at the failure plane. The numerical analysis of the punch test allows the estimation of different assumptions used in analytical models for the rubble failure: the cohesion averaging is an under-conservative approximation, and the non-simultaneity of the cohesive and shear resistance maximum values can be considered in the peak load estimation by the computation of their quadratic mean. The comparison with full scale values shows a reasonably good scaling of the cohesion for the model ice ridges with a long submersion time.