OpenNHM

A short month with a lot of progress! It went over so quickly, even this post missed the end of the month…

• The big focus this month was the implementation of our dense flow kernel (com1DFAPy) in python/cython. The aim is to replicate the results of our current com1DFA module to be able to swap them. This meant diving deep into every numerical/model detail, things like

  • artificial viscosity
  • friction forces
  • particle initialisation
  • and many more.

  Alongside this development a lot of supporting work was needed, i.e. making things easier for us, as one example: being able to read and export particle positions.

Welcome to the first instalment of our monthly update of 2021. This month we started getting into the nitty-gritty of our dense flow module’s numerical implementation:

• To facilitate and improve our development we added new test cases. This time they are more geared towards testing specific numerical issues and more targeted issues, concentrating for now on the dense flow module com1DFAPy.

  • flat plane: a pile of sand (i.e. particles with coulomb friction) on a flat plane. The pile should or should not move depending on the steepness of the sides of the pile

Winter has arrived here, so this is basically the first winter update:

This month focused on getting our code base ready for the upcoming release of version 0.1.

• The documentation got a lot of attention with updates and improvements basically everywhere. Most notably we started to include installation and quick start sections for you to follow.

• Testing took a big step forward. We expanded out standard tests, generated more benchmarks and created scripts to run these tests on demand as well as within our continuous integration.

Winter is slowly creeping up on us, so here is the next AvaFrame update:

• We heavily worked on the dense flow avalanche module com1DFA, bringing functionality from its native code over to python functions. Now there is the option to do parameter variations on friction parameters and release thickness automatically. Besides this new option, more and more simulation parameters can now be set in python.

• Regarding the necessary executable for the com1DFA (dense flow) module: we discovered some discrepancy which we currently work on to resolve. The differences to our reference results are minuscule, but we would like to understand the source of it. Once this is resolved, we will release the executable as a first step to bringing it fully open source.