From characterization to calibration in the cloud using the open-source software MercuryDPM

Many industrial granular processes remain poorly understood. The key reason for this poor understanding is opaque machinery in processes where it is hard to experimentally see inside. Virtual prototyping (VP) offers the solution by giving detailed insights into these processes. Our VP utilizes the particle simulations to create a computational representation of the system which allows industry to optimise their processes, lowering the cost and duration of product development and improving process quality for more sustainable products. However, accurate predictions need a carefully calibrated particle model. Furthermore, discrete element simulations require a high computational effort for large systems and timescales, and thus should be run on a highly scalable, on-demand computer architecture. This study shows the whole workflow from characterization to calibration of a discrete particle simulation in the cloud using the open-source software MercuryDPM [1].
We discuss how to choose appropriate characterization tests for a given powder considering the application’s process conditions. The transition into a computational representation of a granular material mixture – the calibration – will be the main subject for the remainder of this talk.
For calibration, we demonstrate the new MercuryCloud AWS deployed calibration toolbox. This has several key advantages: (i) A no-code interface which is easy to use, (ii) simulations are run in the Amazon EC2 cloud requiring no local hardware, (iii) a user can pick between several characterization tests in a single interface. This interface, its architecture and the calibration procedure based on the open-source code GrainLearning [2] is demonstrated for a Lactose/MCC powder mixture frequently processed in e.g., pharmaceutical wet granulation.
As the application example, we compare a twin-screw wet granulation particle simulation with a calibrated powder mixture to experimental data from a publicly available dataset [3].
[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). https://github.com/chyalexcheng/grainLearning
[3] Plath, Timo; Korte, Carolin; Sivanesapillai, Rakulan; Weinhart, T. (Thomas) (2021): Dataset as a basis for process modeling of twin-screw wet granulation: A parametric study of residence time distributions and granulation kinetics. 4TU.ResearchData. Dataset. https://doi.org/10.4121/14248433.v1