Dynamic analysis and liquefaction

Question

Hi there, 
 I am in the process of conducting a dynamic analysis to asses the stresses and strains in a secant wall during an earthquake event. The parkade is about 28m deep. The water table is close to the surface and the excavation wall consists of secant wall and anchors. There is a liquefiable layer close to the surface (~2m thick) that I am modeling with PM4Sand and a relative density of 65%. I am facing issues with the model converging during the shaking. It slows down dramatically at around 3.7-4 seconds, when the stronger shaking starts. Yesterday it ran for 12 hours and didn't progress past 4 seconds. 
 I tried a few modeling approaches to set up the initial conditions, but none of these made it past 4 seconds during the shaking: 
 1) Excavate the building in stages, pre-stressing the anchors and prescribing a GW boundary at the base of the excavation to bring down the water table inside the building area. All soils were modeled with undrained A conditions. The water table is based on a phreatic case. The secant wall was modeled initially with linear elastic material and switched before the seismic phase to a HSSmall model, calibrated based on slurry wall triaxial data. The liquefiable layer is switched during the seismic phase as recommended in the PM4Manual. 
 2) Excavate the building in stages, pre-stressing the anchors and prescribing a GW boundary at the base of the excavation to bring down the water table inside the building area. All soils were modeled with Drained conditions. The water table is based on a steady-state flow. this approach did not reach convergence. 
 3) Simplified geometry, where the building is excavated and installed in one phase (shown below). All soils were modeled with undrained A conditions. The water table is based on a phreatic case. 
 For all cases I included a 10m thick elastic layer of soil to the side boundaries. The elastic soils have equivalent properties to the corresponding layers and are modeled in the drained condition. 
 Are there any recommendations on how to speed up the model performance and get past the 4 second mark of the earthquake? Is the liquefaction what is causing this to slow down dramatically? I am only running a crustal motion (~38 seconds) and will have to run a strong subduction later, which will be stronger with longer duration. (~150 seconds). Any help would be appreciated. thank you!

Stefanos Papavasileiou · Answer

Dear Boris, 
 Very interesting project! Thank you for all the details. However, as this relates to a very specific project with given material properties and boundary conditions, it is almost impossible to give any accurate advice without checking the project details. 
 In order to be able to assist you with your project and identify the reason for the problem you experience, please submit a service request. 
 Then, one of our support engineers can help you in detail: https://apps.bentley.com/srmanager/ProductSupport

Stefanos Papavasileiou · Answer

Dear all, 
 I just wanted to add our few comments on the subject. 
 Many aspects can influence the solution (model parameters, mesh, time step, boundary conditions, etc.), and all questions are project-specific, which makes it very difficult to give an answer without checking the project. 
 For instance, some questions can arise: I can see a very wide model; is it necessary? Can it be shortened? What is happening at the bottom of the excavation? Were there any localised heave zones? What is the state of the materials before starting the dynamic analysis? 
 The distortion of the elements might be related to the quality of the elements, which is particularly important in liquefaction analysis, where very low stiffness and, consequently, very large deformations occur. Of course, all implicit mesh-dependent methods, such as the finite element method, have limitations when simulating very large deformations. 
 Nevertheless, as mentioned before, appropriate element size and quality, time step, BCs, etc., may help. 
 
 In all cases, feel free to reach out to our support services, and one of our engineers can give it a look: https://apps.bentley.com/srmanager/ProductSupport