However, continuum models are more complex and require control of
several numerical model parameters to ensure accuracy of results.
Advanced Eulerian-mesh based continuum models usually use explicit finite element method due to its ability to handle very high deformations and complex contacts (ice-soil and soil-pipe). The explicit method utilizes a non-iterative (implicit) time integration procedure using very small time steps and does not check for solution convergence after every step. An appropriately selected time step (either calculated by the FE program or specified by the user) generally ensures stability and accuracy of solution, but may yield inaccurate results without any warning messages in some incorrectly set-up analyses. Careful model setup, including appropriate mesh refinement and solution control parameters, and post-analysis checks are necessary to ensure accuracy of results.
Validation of this approach using large scale field test data is necessary before its use in design due to the novel nature of this approach, numerical modeling complexities, and reduction in conservatism compared to the current state-of-practice. A recent study by Philips et al. (2010) compares continuum model predictions against centrifuge test data and PRISE equations. They conclude that numerical modeling can provide invaluable insight; and highlight the importance of achieving a steady state in gouging simulations and the need to use finer meshes and develop better constitutive models.
Finite Element Modeling Approach
Abaqus/Explicit Coupled Eulerian Lagrangian (CEL) technique is used in the present study to analyze the ice-gouging process. Abaqus CEL approach has previously been used by J ukes et al. (2008) and Phillips et al. (2010) for simulation of the ice-gouging process. This method uses the Eulerian mesh for components of the model that can undergo extreme deformations (plastic strain in the soil below and in front of the ice keel is of the order of 1000%) and the Lagrangian meshes for the other components. The Eulerian mesh is fixed in space and material flows through the mesh, unlike the conventional Lagrangian mesh that is attached to and deforms with the material.
Finite element modeling techniques used in the present study are briefly described in this section. Additional details are included in Lele et al. (2011). The soil is modeled using an Eulerian mesh with 8-noded linear reduced integration hexahedral elements (type EC3D8R). The pipe and ice keel are modeled with Lagrangian meshes. Ice keel is currently modeled as rigid (conservative assumption); however, appropriate ice mechanics based material models may also be used for defining ice keel behavior. The pipe is modeled using linear reduced integration shell elements (type S4R). Only half of the geometry is modeled by utilizing model-symmetry and applying appropriate symmetry boundary conditions. A schematic representation of the model is shown in Fig. 4.
Necessary contact definitions and model constraints are defined in an initial step and the simulation is conducted using two analysis steps as described below.
Initial Step: Define boundary conditions to prevent flow of soil material outside the Eulerian mesh region Define contact between ice keel and soil, and soil and pipe Define necessary boundary conditions to prevent rigid body rotation of ice-keel Specify initial geostatic stresses state due to soil overburden Step-1: Gravity loading Impose gravity loading on the soil at the beginning of the step. Dynamic effects due to this sudden loading are small due to initial geostatic stresses defined earlier. Continue simulation with no additional loading for few seconds (simulated time) to allow small numerical fluctuations due to application of gravity loading to dampen out. Step-2: Ice-gouging simulation Apply velocity boundary condition in the gouging direction to the ice-keel. All other degrees of freedom for the keel are constrained.
Fig. 4 Finite Element model
The pipe in the model is sufficiently separated from the initial position of the keel to ensure that the gouging process has reached a steady state when the keel passes over the pipe (of the order of ten times the gouge depth as a rule of thumb). At the end of the simulation the keel has traveled beyond the pipe, but must stop a sufficient distance away from the boundary of the soil included in the model to minimize the effect of boundary conditions.
The 8-noded linear hexahedral (brick) element with reduced integration formulation includes only one solution point at the center of the element. Hence, an appropriate level of mesh refinement is necessary to ensure accuracy of results in complex ice-gouging simulations. Fig. 5 shows the level of pipe-soil mesh refinement necessary to ensure accurate simulation of pipe-soil interaction. The mesh includes sufficient elements along the circumference of the pipe to model curvature, and the soil mesh also includes several elements along the pipe diameter. A sufficiently refined mesh is also necessary in the gouge region to accurately model ice-soil interaction.
Fig. 5 Mesh refinement in the pipe-soil contact region
Due to the level of mesh refinement needed and necessity of including a large soil region to minimize any effects of model boundary conditions, the FE model meshes may include a large number of elements (of the order of hundreds of thousands to a few million depending on pipe/keel geometry). Meshing techniques such as biased element size, with finer mesh in the gouge region and coarser mesh away from the gouge may be employed to reduce the total number of elements and hence the computational time. 617
Journal of Constructional Steel Research Volume 10 Issue None 1988 (Doi 10.1016 - 0143-974x (88) 90034-x) Egor P. Popov Michael D. Engelhardt - Seismic Eccentrically Braced Frames PDF