Nederlands
Developing a parallel solver for mechanical problems

Konrad Kaliszka
Daily supervisor: Martin van Gijzen

Site of the project:
Plaxis
Delftechpark 53
2628 XJ Delft

start of the project: October 2009

In February 2010 the Interim Thesis has been appeared and a presentation has been given.

The Master project has been finished in September 2010 by the completion of the Masters Thesis and a final presentation has been given. For working address etc. we refer to our alumnipage.



Summary of the master project:
Plaxis is using and developing a range of finite element packages intended for 2D and 3D analysis of deformation, stability and groundwater flow in geotechnical engineering. Geotechnical applications require advanced constitutive models for the simulation of the non-linear and time-dependent behavior of soils. In addition, since soil is a multi-phase material, special procedures are required to deal with hydrostatic and non-hydrostatic pore pressures in the soil.

Within the finite element formulation large linear systems have to be solved. At this moment fast and robust iterative solvers are available for sequential computing. Since more and more present day computers consists of more cores, Plaxis is working on parallelization of these solvers. It is not so easy to parallelize the current solver with the same amount of iterations. In this project other techniques are investigated in order to develop robust and efficient parallel solvers.

As a first approach domain decomposition methods have to be studied. After a good DD method has been selected the properties of this method is investigated for the problems originating from geotechnical applications. A combination of this method with a second level preconditioner is the next step in this investigation. Finally, instead of an arbitrary data partitioning, the decomposition of the computational domain should be based on the physical properties of the domain. This decomposition can be used to have special preconditioners for certain subdomains and can also be used to develop a good second level preconditioner.

If time permits another research question is: how to accelerate the solution of a non-linear problem by using information from the solution process of previous linear systems.




Plaxis 3D Tunnel



PlaxFlow

Contact information: Kees Vuik

Back to the home page or the Master students page of Kees Vuik