17-01-2025 Tentative programme

As discussed with: Hendrik (University of Münster), @Hongyang Cheng @Deepak Tunuguntla

Training

Hackathon (Thursday afternoon Session 1 and 2)

• Goal is to lay the foundation so that we can build on this the whole year

• Split into two groups:

  • HPC

    • Led by Nicolin

  • Alternative Methods (AI/ML, MoR)

    • Led by Deepak/Hongyang

19-12-2024 Discuss tentative plan

After discussing with @Thomas Weinhart @Anthony Thornton

Training/Workshop (1 day)

• Introduction to virtual prototyping framework for modelling granular materials and see where ML/surrogate models can help

• Current ML/surrogate modelling techniques utilised for granular materials

  • PINNS, GNS, Model Order Reduction, et cetera

• ML methodologies/workflow

  • CRISP-DM and tools

• Exercises

  • Setting up a data science / ML environment

    • virtual environment and jupyter notebooks

  • Introduce renown frameworks

    • Keras, TensorFlow, PyTorch

  • Going from notebooks to scripts

  • Getting hands-on with the some of the methods presented

    • Will provide rotating drum data

    • Some notebooks with code 

• Uncertainty Quantification

  • GrainLearning

  • Includes tutorials and exercises

Hackathon (1/2 day)

Vision: Collection of surrogate model implementations with special focus on rotating drum dynamics

Goal: Setup the surrogate models codebase to build upon in the upcoming year

As a deliverable at the end of the hackathon, would be great if we end up with a GitHub repo with a certain folder structure acting as a placeholder for all the different ML/surrogate models we would like to implement.

Eventually, this repo will be converted as a python package called GranML, which should be integrable into other DEM open-source code bases.

16-12-2024 Original meeting discussion points

@Hongyang Cheng @Deepak Tunuguntla

• Start something high-level in granular material modeling, using existing examples implemented with Keras, TensorFlow, PyTorch, etc.

• Wednesday afternoon training, Thursday full-day hackathon (probably with one or two hours on advanced topics or more practical stuff)

• What topics should be covered during the hackathon?

  • model order reduction, encoder/decoder

  • local, global (spatial) time series

  • optional: uncertainty quantification and optimization

  • a GranML package which contains a wide range of available ML models

• Prepare rotating dataset before the hackathon: Roxana and Balazs have scripts ready to use

  • For the hackathon probably 2D or 3D with walls

  • Think about common codebases for future integration

• Whom to invite as more experienced participants?

  • UT: Dongwei Ye and his colleagues

  • JKU?

  • Anyone from Saxion?

• Communication with the working groups about our tasks (@Deepak Tunuguntla )