Simulating industrial scenarios: with the open-source software MercuryDPM

  • Thornton, Anthony (University of Twente)
  • Nguyen, Quang-Hung (University of Twente)
  • Polman, Harmen (University of Twente)
  • Bisschop, Jan-Willem (MercuryLab)
  • Weinhart-Mejia, Raquel (MercuryLab)
  • Post, Mitchel (University of Twente)
  • Fitzsimmons, Donna (MercuryLab)
  • Vesal, Mohammad Reza (MercuryLab)
  • Ostanin, Igor (University of Twente)
  • Weinhart, Thomas (University of Twente)

Please login to view abstract download link

Creating accurate predictive computer simulations, i.e. virtual prototypes, of complex granular industrial processes has many challenges. In this presentation I will review, with examples from industry, recent advances in creating such virtual prototypes. I will introduce the open-source code MercuryDPM [1], which is often applied to complex industrial applications via the spin-off company MercuryLab. I will briefly discuss how to accurately import industrial geometries and deal with large numbers of particles and wide size distributions, see figures below. Then I will focus on how to create a computer representation of an actual granular material, the so-called model calibration. For calibration, I will start by reviewing what parameters need to be measured and what experimental characterisation machines are available. I will present an industrially practical calibration method, where certain parameters are directly measured and others are indirectly calibrated, using a variety of machine-learning techniques including the open-source code GrainLearning [2] and two other popular supervised learning algorithms: Neural Network (NN) and Random Forest (RF) regression. With GrainLearning, one can find local optima in only two to three iterations, even for complex contact models with many microscopic parameters. On the other hand, after a training period consisting of hundreds of DEM simulations, the NN and RF are capable of providing a database which can be used to find the micro-parameters that correspond to the experimental static angle of repose. Finally, I will demonstrate the whole workflow in AWS deployed cloud platform: MercuryCloud; highlighting our custom characterisation machines and calibration tools. [1] T Weinhart, L Orefice, M Post, et al, Fast, flexible particle simulations – An introduction to MercuryDPM, Computer Physics Communications, 249, 107129 (2020). [2] H Cheng, T Shuku, K Thoeni, P Tempone, S Luding, V Magnanimo. An iterative Bayesian filtering framework for fast and automated calibration of DEM models, Computer methods in applied mechanics and engineering, 350, 268-294 (2019).