Skip to content

Latest commit

 

History

History
59 lines (33 loc) · 2.29 KB

README.md

File metadata and controls

59 lines (33 loc) · 2.29 KB

Running the code

Run the code by typing python main.py. You need numpy and scipy for executing the code. Output folders with experimental results are created. The folders are named in the following way: seriesX_SHORTNAME_DATETIME. See below for details on the different experimental series.

You can plot the results by running python postproc.py <foldername>. You need matplotlib for plotting.

Convergence Study

This code can be used to perform convergence studies in a partitioned multi-physics multi-scale setup. As a benchmark scenario the 1D heat transport equations is solved. For details see [1]. The convergence studies performed by this code are identical to the ones described in [1].

Parameter set from [1]

If you clone this repository, you can directly perform the experiments described in [1] with the respective parameters:

  • maximum simulation time T = 1
  • largest timestep size tau_0 = 1/4
  • considered timestep sizes tau_i = [1/4, 1/8, 1/16, ..., 1/1024, 1/2048]
  • coarse mesh resolution h = 0.2
  • fine mesh resolution h/4 = 0.05

Parameter set with lower runtime

For a lower runtime, you can also use a modified parameter set:

  • maximum simulation time T = 1/32
  • largest timestep tau_0 = 1/128
  • considered timestep sizes tau_i = [1/128, 1/256, 1/512, 1/1024, 1/2048]

The plots below are created from the modified parameter set.

Experimental Series 1

Order degradation to first order if classical implicit coupling is used with higher order schemes a regular domain

Experimental Series 2

Second order if customized coupling schemes are used with second order schemes a regular domain

Experimental Series 3

Second order if Strang splitting coupling is used with higher order schemes on a non-regular domain

Experimental Series 4

High order if waveform relaxation coupling is used with higher order schemes on a non-regular domain

Reference

If you want to refer to this code, please use the following reference:

[1] Rüth, B., Uekermann, B., Mehl, M., & Bungartz, H.-J. (2018). Time Stepping Algorithms for Partitioned Multi-Scale Multi-Physics. Proceedings of ECCM VI / ECFD VII, (June).