Documents published in Scipedia

• Parametric Physics-Informed Neural Networks for Material Model Calibration from Full-Field Displacement Data

  D. Anton, H. Wessels

  particles2023.

  
  

  Abstract

  The identification of material parameters occurring in material models is essential for structural health monitoring. Due to chemical and physical processes, building structures and materials age during their service life. This, in turn, leads to a deterioration in both the reliability and quality of the structures. The material parameters indicate possible damage and material degradation, as they directly reflect the resistance of the structure to external impacts. We further developed physics-informed neural networks (PINNs) [1] for the calibration of the linear-elastic material model from full-field displacement data and global force data in a realistic regime [2]. For a realistic data regime, the optimization problem had to be conditioned. The advantage of this method is a straightforward inclusion of observation data. Unlike grid-based methods, such as the least square finite element method approach, no computational grid and no interpolation of the data are required. However, directly solving inverse problems using PINNs is computationally expensive and prone to realistic noise levels in the measurement data. Moreover, the PINN must be trained completely from scratch for each new full-field displacement measurement, even if the geometry and material of the structure remain unchanged. In our ongoing work, we are therefore focusing on learning parameterized solutions of parametric partial differential equations using PINNs, such as [3]. We further investigate the ability of parametric PINNs to act as a surrogate for the identification of material parameters from full-field displacement data. By learning parameterized solutions, the PINN does not need to be completely re-trained for each full-field displacement measurement. The calibration of the material model can thus be drastically accelerated, and information about the material condition can be provided near real-time. Furthermore, we also plan to apply the parametric PINN to more complex material models, such as those for hyper-elastic and elasto-plastic materials.

  Abstract
  The identification of material parameters occurring in material models is essential for structural health monitoring. Due to chemical and physical processes, building structures [...]

  • 0
  •  read

• DEM-BPM modelling of coated particles to investigate their crash absorbing behaviour

  W. Safdar, S. Heinrich, A. Düster

  particles2023.

  
  

  Abstract

  The collision of ships remains a significant cause of accidents, resulting in severe environmental consequences such as oil spillage from oil tankers. Enhancing the crashworthiness of ship structural design is crucial, and one approach being explored is the filling of the double hull structure with granular material [1]. Coating these particles with environmentally friendly materials can optimize their ability to absorb kinetic energy and transfer loads from the outer to the inner hull [2]. However, the mechanical behavior of the coated particles depends on the type of coating material, posing challenges in developing numerical simulation models. To address this, an open-source Discrete Element Method (DEM) code called MUSEN [3] is utilized to numerically model the behavior of coated particles. Furthermore, MUSEN can be extended using the Bonded Particle Method (BPM) to simulate particle breakage through solid bridges. However, the inclusion of coating material in the model increases the number of parameters, as well as computational time and cost. This requires a robust methodology to characterize the mechanical behaviour regardless of the type of coating material with reduced computational cost, keeping in mind the actual application inside the ship double hull. Sensitivity analyses and parametric studies are conducted to understand the effect of input parameters and identify the influential ones. Subsequently, algorithms such as the Particle Swarm Algorithm are employed for parameter optimization. Finally, models of different fidelity are used to compare the results from multi particle compression tests. The findings from these simulations will be presented in this contribution

  Abstract
  The collision of ships remains a significant cause of accidents, resulting in severe environmental consequences such as oil spillage from oil tankers. Enhancing the crashworthiness [...]

  • 0
  •  read

• A Novel Parallelisation Scheme for DEM on Distributed Memory Computers

  E. Staercke, C. Nougiuer-Lehon, F. Frolio

  particles2023.

  
  

  Abstract

  Several strategies have been proposed for high performance computing with DEM. Yet, quite few works have been published on this topic (e.g., [1-3]), compared to the extensive use of DE simulations as a modelling tool in research and industry. This is however not surprising, since granular media at the grain scale appear as highly evolutive and disordered systems, which makes effective parallelisation a challenging task. We present an approach which exploits an inherent limitation of DE simulations, as the opportunity for an efficient and flexible parallel implementation on distributed memory computers. Namely, the “numerical sound speed” set by the representative particle diameter and the time step defines an upper bound for the celerity at which perturbations can propagate across the discrete medium. We show how this numerical artefact (of negligible importance for most applications) can be turned into an original criterion of spatial domain decomposition, which leads to a DEM-suited parallelisation scheme. We present our approach through its actual implementation into the DEM’ocritus code [4,5]. We analyse its performance through benchmarks and parametric analyses of biaxial tests, on assemblies of up to 15 million circular particles.

  Abstract
  Several strategies have been proposed for high performance computing with DEM. Yet, quite few works have been published on this topic (e.g., [1-3]), compared to the extensive [...]

  • 0
  •  read

• Modeling and Simulation of Soil and Soil-Tool Interaction for Industrial Applications

  M. Harutyunyan, S. Emmerich, S. Steidel, M. Burger, K. Jareteg, J. Quist

  particles2023.

  
  

  Abstract

  Modeling of soil-tool interaction for industrial applications involves the coupling of soil models, mostly based on the Discrete Element Method (DEM), with multibody systems, representing construction machinery such as excavators or wheel loaders. To obtain accurate predictions of reaction forces on tools like wheel loader buckets, it is essential to have an appropriate parametrization procedure, which makes use of data obtained from laboratory tests such as the triaxial compression test or direct shear test for different types of soil. Simulations with suitable DEM-softwares can then be validated against the experimental data to assess the applicability and performance of the numerical methods [2, 3] The DEM software GRAnular Physics Engine (GRAPE) developed at Fraunhofer ITWM in Germany has been successfully used to simulate compression and shear tests and has been proven to yield good predictions of the soil-tool interaction and draft forces [4] for spherical particles such as coarse sand. A year-long collaboration with the Fraunhofer-Chalmers Centre (FCC) in Sweden has successfully resulted in the development of a soil simulation toolbox for FCC’s general purpose DEM solver Demify® incorporating and enhancing the simulation techniques used in GRAPE [1]. Co-simulation is enabled through an FMI interface coupling Demify® with multibody systems modeled e.g., in Simulink to evaluate relevant variables such as forces at critical linkage joints of the construction machines. To model real-life application scenarios, the entire workflow must be considered, from the soil parametrization process, to setting the particulate soil model and performing numerical simulations for the specific problem, to the final post-processing to analyze the load data. We will characterize and discuss the different steps of the workflow and present simulation results obtained with the toolbox Demify® for Heavy Machinery. 

  Abstract
  Modeling of soil-tool interaction for industrial applications involves the coupling of soil models, mostly based on the Discrete Element Method (DEM), with multibody systems, [...]

  • 1
  •  read

• Continuum Granular Flow Modeling and Beyond: Exploiting Meshless Methods

  K. Kamrin

  particles2023.

  
  

  • 15
  •  read

• Particle-based fiber models of woven materials for earth entry thermal protection

  A. Santos, L. Abbot, J. Haskins

  particles2023.

  
  

  Abstract

  Applications requiring materials with layer-to-layer strength, from basketry to thermal protection systems, use interlaced, three-dimensional woven materials. NASA is developing and deploying woven material heat shields for missions, including Artemis I (3D-MAT for compression pads) and potentially Mars Sample Return - Earth Entry System. These materials are complex, hierarchical and must protect from extreme environments and phenomena, such as deformation, impact and high-enthalpy heating. Woven material performance depends on microstructure, damage and weave geometry. Therefore, fiber-specific models are needed to simulate fiber contacts within the weave hierarchical geometry (fiber to tow to yarn to weave) and the inherent directionality of fibers. Explicit fiber models can simulate how weave microstructure evolution affects thermal and mechanical properties. We parameterize a discrete element bonded particle models (DEM-BPM) of fibers to capture thermal and mechanical behavior within and between fibers, with bonded and contact forces, respectively. We study the proportion of heat transfer and stress via the contact network, fiber bonds and the weave geometry, for example, with respect to yarn warp-weft identity (whether it interlaces weave layers). Our results demonstrate the importance of explicit fiber modeling for connecting microstructure with thermal and mechanical properties.

  Abstract
  Applications requiring materials with layer-to-layer strength, from basketry to thermal protection systems, use interlaced, three-dimensional woven materials. NASA is developing [...]

  • 0
  •  read

• GPUs Based Material Point Method for Compressible Flows

  P. Baioni, T. Benacchio, L. Capone, C. de Falco

  particles2023.

  
  

  Abstract

  Particle-In-Cell (PIC) methods such as the Material Point Method (MPM) can be cast in formulations suitable to the requirements of data locality and fine-grained parallelism of modern hardware accelerators such as Graphics Processing Units (GPUs). While continuum mechanics simulations have already shown the capabilities of MPM on a wide range of phenomena, the use of the method in compressible gas dynamics is less frequent. This contribution aims to show the potential of a GPU-based MPM parallel implementation for compressible fluid dynamics, as well as to assess the reliability of this approach in reproducing supersonic gas flows against solid obstacles. The results in the paper represent a stepping stone towards a highly parallel, Multi-GPU, MPM-base solver for M ach > 1 Fluid-Structure Interaction problems.

  Abstract
  Particle-In-Cell (PIC) methods such as the Material Point Method (MPM) can be cast in formulations suitable to the requirements of data locality and fine-grained parallelism [...]

  • 16
  •  read

• Computational Error Estimation for the Material Point Method in 1D and 2D

  M. Berzins

  particles2023.

  
  

  Abstract

  The Material Point Method (MPM) is widely used for challenging applications in engineering, and animation but lags behind some other methods in terms of error analysis and computable error estimates. The complexity and nonlinearity of the equations solved by the method and its reliance both on a mesh and on moving particles makes error estimation challenging. Some preliminary error analysis of a simple MPM method has shown the global error to be first order in space and time for a widely-used variant of the Material Point Method. The overall time dependent nature of MPM also complicates matters as both space and time errors and their evolution must be considered thus leading to the use of explicit error transport equations. The preliminary use of an error estimator based on this transport approach has yielded promising results in the 1D case. One other source of error in MPM is the grid-crossing error that can be problematic for large deformations leading to large errors that are identified by the error estimator used. The extension of the error estimation approach to two space higher dimensions is considered and together with additional algorithmic and theoretical results, shown to give promising results in preliminary computational experiments.

  Abstract
  The Material Point Method (MPM) is widely used for challenging applications in engineering, and animation but lags behind some other methods in terms of error analysis and [...]

  • 0
  •  read

• Determining Coke Moisture Content Through Images Analysis Methods and Machine Learning Models

  M. Li

  particles2023.

  
  

  Abstract

  Coke moisture content plays a crucial role as a quality indicator in ironmaking. Fast and accurate measurement of coke moisture content can effectively ensure operational stability, thereby guaranteeing the quality of pig iron. Coke with different moisture levels exhibits variations in light reflection, refraction, and powder adhesion characteristics. Drawing from this premise, we employed an image analysis approach to analyse the colour and texture features of coke images with varying moisture content. The results indicated that certain specific image features are highly sensitive to changes in coke moisture content. This study also conducted testing and comparison of performances of three common machine learning models in predicting coke moisture content based on image analysis. The Support Vector Machine (SVM) predictive model for coke moisture content based on image analysis showed optimal performance. This demonstrated a close connection between coke image features and moisture content.

  Abstract
  Coke moisture content plays a crucial role as a quality indicator in ironmaking. Fast and accurate measurement of coke moisture content can effectively ensure operational [...]

  • 1
  •  read

• Simulating industrial scenarios: with the open-source software MercuryDPM

  A. Thornton, Q. Nguyen, H. Polman, M. Post, J. Bisschop, R. Weinhart-Mejia, D. Fitzsimmons, M. Vesal, T. Plath, I. Ostanin, T. Weinhart

  particles2023.

  
  

  Abstract

  Creating predictive computer simulations, i.e. virtual prototypes, of complex granular industrial processes has many challenges. In this paper we review recent advances in creating such virtual prototypes. We introduce the open-source code MercuryDPM [1], which is often applied to complex industrial applications via the spin-off company MercuryLab. We briefly discuss how to import complex industrial geometries and how to deal with large numbers of particles and wide size-distributions. Then we focus on how to create a computer representation of an actual granular material, the so-called model calibration. For calibration, we start by reviewing what parameters need to be measured and what experimental characterisation machines are available. We present an industrially practical calibration method, where certain parameters are directly measured and others are indirectly calibrated, using a variety of machine-learning techniques, implemented in the open-source codes GrainLearning [2], TensorFlow [3] and scikit-learn [4]. 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, TensorFlow and scikit-learn use two popular supervised learning algorithms, Neural Network (NN) and Random Forest (RF) regression, respectivly. After a training period consisting of hundreds of particle simulations, NN and RF are capable of providing a mapping between the micro-parameters and the bulk behaviour, which can be used to find the optimal micro-parameters that correspond to the experimentally observed behaviour.

  Abstract
  Creating predictive computer simulations, i.e. virtual prototypes, of complex granular industrial processes has many challenges. In this paper we review recent advances in [...]

  • 14
  •  read