Modeling Of Physical Processes Research Articles

Determination of the Curie temperature of structural alloys by analyzing the electrical characteristics of wires

The Curie temperature for many structural multicomponent alloys can only be estimated approximately, for example, from state diagrams. The necessity of an operational assessment of this parameter arises in the design of technologies for processing and controlling structural materials, in mathematical modeling of physical processes. The paper describes an experimental method for determining the temperature of magnetic transformation by measuring the temperature dependences of inductance and electrical resistance of wire samples. Verification of the procedure was carried out by determining the Curie temperature of nickel and comparing it with data obtained by other methods. We established the dependence of the Curie temperature of nickel on the carbon content and determined the Curie temperatures for alloys of Fe-Cr and Fe-Ni systems. Using numerical modeling and experimental data, temperature dependences of the relative magnetic permeability of materials were obtained. The results show the presence of the Hopkinson effect in the materials under study and the presence of inflection points on the electrical resistance curves.

THERMAL FIELD IN A HORIZONTAL WELL APPLIED TO THE DETERMINATION OF FLUID FLOW BY ACTIVE THERMOMETRY

Today, in the context of falling oil production and increasing water cut, the importance of geophysical control over field development is increasing. A promising method of production logging is active thermometry, it is based on the creation of a thermal disturbance area in the wellbore, for example, due to induction heating of a section of a metal pipe (liner or casing, depending on the design of the well). As a result of heat exchange with a locally heated pipe section, heat marks are formed in the liquid, moving along the wellbore. The article presents the results of physical modeling of thermal processes during local induction heating of a section of a horizontal well. It has been experimentally established that with an induction heater power of about 1 kW and flow rates in the pipe up to 50 m3/day, the heating of the pipe wall reaches 12 K. As part of the experimental studies, the movement of heat marks was monitored using two modules of temperature sensors distributed over the pipe cross section. It is shown that the nature of temperature change in time is determined by the position of the temperature sensor in the pipe section. The greatest thermal anomalies are noted for sensors pressed against the pipe in its upper part, the smallest - for sensors located in the lower part of the pipe. The fluid flow rate in the pipe was estimated from the rate of passage of the heat mark between the temperature modules. It is shown that the calculated flow rate is overestimated relative to the actual one due to the influence of natural thermal convection, which causes the occurrence of a natural convective fluid flow in addition to the forced movement with a given flow rate. A new analytical solution has been obtained for calculating the dynamics of the pipe temperature in the induction heating section, a comparison of the results of calculations using the analytical model with the experimental results showed good agreement. The results of experimental studies can be used in interpreting the results of field logging wells using active thermometry, in particular, the sensitivity of the pipe temperature in the induction heating section to the flow rate can be used to solve inverse problems of estimating the flow rate in a pipe by measuring its surface temperature using the obtained analytical solution.

Применение научной визуализации в моделировании низкотемпературной газоразрядной плазмы

This article discusses the importance of developing software and hardware complexes for data visualization and improving methods of presenting them in the modeling of low-temperature gas discharge plasma phenomena and related physical processes. In particular, issues related to creating a specialized distributed environment for solving gas discharge modeling tasks are considered. The use of computer graphics for visualization has allowed for a visual analysis of the obtained data and making necessary adjustments to the modeling process interactively with the data. The paper presents the results of visualizing a model of mercury-argon plasma in a coaxial microwave discharge. The application of local cross-sections of the computational domain for visualization has revealed the internal structure and peculiarities of the distribution of the electric field and electron concentration in the discharge tube. The obtained data served as the basis for creating an experimental prototype of a coaxial gas discharge optical radiation source.

Self-Similar Solution to the Problem of Unsteady Film Condensation on a Vertical Semi-Infinite Plate

Changes in the regimes of condensate film flow along the underlying surface in the presence of a body force are considered within the framework of a simplified physical and mathematical model of hydrodynamic and thermal processes occurring under unsteady film condensation on a vertical semi-infinite plate. The model is used to determine the applicability conditions of the zero-gravity approximation when considering film condensation. It is found that the validity of using the zero-gravity approximation to calculate the film thickness growth rate on any plate section depends not only on the magnitude of the body force, but also on time and distance between this section and the leading edge of the plate.

Perspectives of computational fluid dynamics methods in the study of membrane processes

In the mathematical modeling of membrane processes, the study of the structural features of the membranes used and the significant limitation of information in the formal description of their separation properties led to the development of physical models that take into account not only the structural features of real membranes, but also their functioning features. When compiling physical models of membrane processes, it is impossible to objectively carry out a quantitative accounting of most factors due to their great diversity and variability, which distances the mathematical model far from the real process. This is why CFD methods reliably and efficiently perform calculations for all physical models and types, including steady or transient flow, incompressible or compressible flow (from small subsonic to hypersonic), laminar or turbulent flow simulations, Newtonian or non-Newtonian fluids, ideal or real gas . The authors analyzed the possibility of using CFD to calculate the hydrodynamics of flows in a membrane bioreactor based on hollow fibers. Safarov R.R. etc., an electronic geometric model of the installation was built, mesh geometry with different densities was selected to optimize the calculation time and solution accuracy for a specific case, the kinetic dependence of cell growth was calculated, the flow rates of the nutrient medium into the intrafiber and interfiber spaces of the bioreactor were determined, and the hydrodynamic conditions were analyzed. All of the above confirms the prospects for using CFD methods for modeling membrane processes complicated by culturing cells on the surface of membranes.

AI‐enhanced iterative solvers for accelerating the solution of large‐scale parametrized systems

SummaryRecent advances in the field of machine learning open a new era in high performance computing for challenging computational science and engineering applications. In this framework, the use of advanced machine learning algorithms for the development of accurate and cost‐efficient surrogate models of complex physical processes has already attracted major attention from scientists. However, despite their powerful approximation capabilities, surrogate model predictions are still far from being near to the ‘exact’ solution of the problem. To address this issue, the present work proposes the use of up‐to‐date machine learning tools in order to equip a new generation of iterative solvers of linear equation systems, capable of very efficiently solving large‐scale parametrized problems at any desired level of accuracy. The proposed approach consists of the following two steps. At first, a reduced set of model evaluations is performed using a standard finite element methodology and the corresponding solutions are used to establish an approximate mapping from the problem's parametric space to its solution space using a combination of deep feedforward neural networks and convolutional autoencoders. This mapping serves as a means of obtaining very accurate initial predictions of the system's response to new query points at negligible computational cost. Subsequently, an iterative solver inspired by the Algebraic Multigrid method in combination with Proper Orthogonal Decomposition, termed POD‐2G, is developed that successively refines the initial predictions of the surrogate model towards the exact solution. The application of POD‐2G as a standalone solver or as preconditioner in the context of preconditioned conjugate gradient methods is demonstrated on several numerical examples of large scale systems, with the results indicating its strong superiority over conventional iterative solution schemes.

Mathematical Simulation in Nuclear Medicine for Optimization Diagnostic Accuracy of SPECT/CT

Purpose: To develop the method of mathematical simulation in the field of nuclear medicine and its
 practical application in research aimed at improving the diagnostic accuracy of the SPECT/CT method.
 The basic principles of the method of mathematical simulation in nuclear medicine and the main modules of the software package “Virtual Platform for Simulation Tests of the SPECT/CT Method”, created
 at the laboratory for simulation in nuclear medicine of Novosibirsk State University and the Institute of
 Theoretical and Applied Mechanics.
 Material and methods: The main principles of the developed software package are the realism of the ‘virtual patient’ mathematical model, the accuracy in modeling of physical processes in visualization, as
 well as simulation conducting close to clinical practice. All studies were carried out in collaboration
 with clinicians. The software complex was verified by comparison with clinical data. Numerical experiments have shown a close correspondence between the results of clinical and virtual studies. In simulation tests, the same errors were obtained on images that were observed on clinical images. Simulation
 computer tests were carried out using the software package “Virtual Platform...” in the field of nuclear
 oncology, cardiology and neurology, aimed at studying the accuracy of images of pathological foci to assess the possibility of switching to quantitative SPECT, as well as optimizing the protocol for examining
 patients.
 Results: Simulation tests have shown that in order to move to quantitative SPECT, it is not enough to be
 limited to the development of a methodology for calibrating SPECT systems for a particular radionuclide. There are problems without understanding and solving which it is impossible to talk about the diagnostic accuracy of the quantitative SPECT method. Such problems are edge artifacts that appear on
 the images of pathological foci, and the rule of stopping the iterative algorithm when the values of activity in the area of interest (pathological focus) that are closest to the true values are reached. A stopping
 criterion based on Pearson's modified chi-square test of goodness of fit is proposed. There is a significantly good correlation between the proposed stop criterion and the minimum RMS error of the image
 reconstruction. The results of simulation modeling with the aim of optimizing SPECT examinations of
 brain perfusion demonstrated the possibility of reducing the data recording time compared to the standard protocol by at least two times.
 Conclusion: The computer simulation method presented in this paper is a practical technology that
 contributes to the optimization and development of a quantitative SPECT method to achieve the best
 possible results in the imaging of pathological foci. Further development of the software package is focused on applications in radionuclide therapy and theranostics.

DES Y3 cosmic shear down to small scales: Constraints on cosmology and baryons

We present the first analysis of cosmic shear measured in DES Y3 that employs the entire range of angular scales in the data. To achieve this, we built upon recent advances in the theoretical modelling of weak lensing provided by a combination of N-body simulations, physical models of baryonic processes, and neural networks. Specifically, we used BACCOemu to model the linear and non-linear matter power spectrum including baryonic physics, allowing us to robustly exploit scales smaller than those used by the DES Collaboration. We show that the additional data produce cosmological parameters that are tighter but consistent with those obtained from larger scales, while also constraining the distribution of baryons. In particular, we measured the mass scale at which haloes have lost half of their gas, log Mc = 14.38−0.56+0.60 log(h−1 M⊙), and a parameter that quantifies the weighted amplitudes of the present-day matter inhomogeneities, S8 = 0.799−0.015+0.023. Our constraint on S8 is statistically compatible with that inferred from the Planck satellite data at the 0.9σ level. We find instead a 1.4σ shift in comparison to that from the official DES Y3 cosmic shear, because of different choices in the modelling of intrinsic alignment, non-linearities, baryons, and lensing shear ratios. We conclude that small scales in cosmic shear data contain valuable astrophysical and cosmological information and thus should be included in standard analyses.

Comparison of damping techniques in non-linear spring-mass systems for power transformers under short-circuit condition

PurposeThe simplified modeling of many physical processes results in a second-order ordinary differential equation (ODE) system. Often the damping of these resonating systems cannot be defined in the same simplified way as the other parameters due to the complexity of the physical effects. The purpose of this paper is to develop a mathematically stable approach for damping resonances in nonlinear ODE systems.Design/methodology/approachModifying the original ODE using the eigenvalues and eigenvectors of a linearized state leads to satisfying results.FindingsAn iterative approach is presented, how to modify the original ODE, to achieve a well-damped solution.Practical implicationsThe method can be applied for every physical resonating system, where the model complexity prevents the determination of the damping.Originality/valueThe iterative algorithm to modify the original ODE is novel. It can be used on different fields of the physics, where a second-order ODE is describing the problem, which has only measured or empirical damping.

METHODS FOR CALCULATING THE THERMAL AND STRESS-STRAIN CONDITIONS OF THE MAIN COMBUSTION CHAMBERS FLAME TUBES

Large-scale russian aggression has created a number of new and critical challenges in the field of security and defense. The growing number of emerging tasks, and at the same time the need for greater efficiency in solving them, determine the need to update and implement rational ways to maintain the serviceability of the required number of aircraft, taking into account their own and international experience.
 Thus, the development of a scientific and methodological apparatus for modeling physical processes that occur in resource-limiting parts and assembly units of aircraft engines, regarding their impact on the durability of the material in the process of operating factors, is an urgent scientific task. At the same time, it is taken into account that absolutely every engine loading cycle during flight or ground testing corresponds to a certain proportion of damage to its components, which determines a certain quantitative measure of their exhaustion of the resource.
 The article proposes a methodology for calculating the thermal stress-strain states of the flame tubes of the main combustion chambers. The objectivity and accuracy of the calculations is determined by the reliability of the boundary conditions required to solve the non-stationary thermal conductivity equation. The proposed technique allows to take into account the distribution of temperature fields obtained due to numerical modeling of the working process in the main modes of engine operation.

Physics-Informed Neural Networks for the Heat Equation with Source Term under Various Boundary Conditions

Modeling of physical processes as partial differential equations (PDEs) is often carried out with computationally expensive numerical solvers. A common, and important, process to model is that of laser interaction with biological tissues. Physics-informed neural networks (PINNs) have been used to model many physical processes, though none have demonstrated an approximation involving a source term in a PDE, which modeling laser-tissue interactions requires. In this work, a numerical solver for simulating tissue interactions with lasers was surrogated using PINNs while testing various boundary conditions, one with a radiative source term involved. Models were tested using differing activation function combinations in their architectures for comparison. The best combinations of activation functions were different for cases with and without a source term, and R2 scores and average relative errors for the predictions of the best PINN models indicate that it is an accurate surrogate model for corresponding solvers. PINNs appear to be valid replacements for numerical solvers for one-dimensional tissue interactions with electromagnetic radiation.

Identifying Probable Submarine Hydrothermal Spots in North Santorini Caldera Using the Generalized Moments Method

The presence of active hydrothermal vent fields near residential areas and their possible link to volcanic activity poses a potential hazard to the environment, society, and the economy. By capitalizing on Autonomous Underwater Vehicle sampling methodologies and applying the Generalized Moments Method model for geological and physical processes in these environments, we shed light on the underlying dynamics shaping the physicochemical characteristics of the vents. In this study, we focus on the Northern Caldera of Santorini and, more specifically, on the recorded CTD data (Conductivity, Temperature, Depth). The data sets were collected in 2017 in Santorini using an Autonomous Underwater Vehicle during the GEOMAR POS510 mission. Our research shows that the active vent field within the caldera probably follows a multifractal behavior and exhibits a weak memory effect. Depth Profiles and Time Series show similar behavior among conductivity and temperature. The variance and moments of both parameters underline the existence of two different mechanisms governing the behavior of the vent field. Finally, the structure function shows that changes in the time series are described by a Cauchy–Lorentz distribution.

Lithium Ion Battery Electrode Calendering Study By Combining X-Ray Computed Tomography and 3D-Resolved Physical Process Models

Significant amount of research has been done recently on how to increase energy densities while maintaining or lowering costs in response to the growing demand for lithium-ion batteries (LIBs). In order to achieve optimization of the manufacturing process of batteries, it is essential to understand the influence of each process stage on the architectures of the electrodes, which affects the energy, power, lifetime and safety of the LIB cells. Our ERC-funded ARTISTIC project1 develops a digital twin that enables predicting the electrode architectures and their electrochemical performances from parameters at each stage of the LIBs manufacturing process.In this work, we present our most recent research, which studies the manufacturing process of LiNixMnyCo1-x-yO2 (NMC)-based cathodes and especially focuses on the electrode calendering stage2. Our methodology encompasses coarse-grained molecular dynamics and discrete element method to simulate the calendering process of electrodes that are virtually produced after the drying and slurry simulations. In this new work, we account in our model for non-spherical NMC secondary particles with realistic 3D shapes derived by X-ray computed tomography (XCT), together with spherical particles representing the carbon binder domain. An in situ compressing micro-XCT experiment is carried out on the NMC electrode to monitor the change of the electrode morphology upon the calendering process. The findings from the simulation are validated by comparing our predicted electrode architectures with the experimental ones. Our approach captures the alterations in orientation of the secondary particles throughout the calendering processes and the relaxation process, as well as their possible deformation and spring back effect brought on by the pressure. Additionally, we discuss the comparison of variations in porosity, tortuosity factor, conductivity, different phases distribution and pore size distribution during the calendering process. ERC Artistic : Home Artistic. http://www.erc-artistic.eu/.Xu, J., Ngandjong, A.C., Liu, C., Zanotto, F.M., Arcelus, O., Demortière, A., and Franco, A.A. (2023). Lithium ion battery electrode manufacturing model accounting for 3D realistic shapes of active material particles. Journal of Power Sources 554, 232294. 10.1016/j.jpowsour.2022.232294.

Modelling and simulation of fast pyrolysis of pomace from three-phase olive mill targeting optimal yields of pyrolysis products

Biomass can be transformed into useful bioproducts (biofuels, biomaterials) through its thermochemical conversion. This study concerns modeling and simulation of fast pyrolysis of olive pomace from three-phase olive mill using Super Pro Designer (SPD) software, which was selected because of its vast databank of specific chemical compounds as well as specific unit operations specifically designed for modeling and simulation of biological, physical and chemical processes. The simulation was carried out at pyrolysis temperatures ranging from 400 to 675 °C and residence times varying between 0.1 and 15s. Simulation results indicate that fast pyrolysis yielded maximum bio-oil yield of 23.72% at 650 °C and a residence time of 0.1s, and maximum syngas yield of 41.17% at 675 °C and a residence time of 15s. Predicted product yields were in accordance with experimental data collected from the literature, with relative errors in the range of 9%, which may be due to variable feedstock properties. The developed model provides very useful information on olive pomace fast pyrolysis conditions.

Large-scale closed and generalized networks of ribosome flow model with different site sizes

Dynamical systems play a key role in understanding the different biophysical aspects of many complex transport phenomena. The ribosome flow model with different site sizes (RFMD) is a dynamical system relevant to model various biological and physical processes such as phosphorelay, motion of vehicles along an isolated road, and more. However, many transport processes take place concurrently which give rise to networks comprising multiple tracks e.g., vehicular motion, cargo transport along the cytoskeletal highway, etc. Here, we present two large-scale network models: RFMDs network with a pool (RFMDNP), and the generalized network of RFMDs (RFMDN). The RFMDNP is a closed system that can model the simultaneous movement of particles along tracks having sites of different capacities in a resource-limited environment. We prove that the RFMDNP admits a continuum of linearly ordered steady-state points and it always phase-locks with the periodic excitations in the parameters. Furthermore, we also show that increasing any of the parameters in an RFMD leads to an increase in the output rate of this RFMD, and the output rate of other RFMDs all increase or decrease. Next, we analyze a generalized network of RFMDs that models static connections between the RFMDs and prove the important mathematical properties: (a) the network admits a unique steady-state, and (b) the problem of determining the connecting weights to maximize the steady-state network output rate is a convex optimization problem. We believe that these networks can have wide applicability to model various non-linear phenomena since their behavior is predictable and ordered.

ON INFLUENCE OF PHYSICAL AND CHEMICAL MODIFICATION OF WETTABILITY OF FILTRATION SURFACE ON INTERCHANGEABILITY OF IMMISCIBLE FLUIDS IN RESERVOIR FORMATIONS

The article presents and discusses the results of experimental studies of the processes of mutual substitution (displacement) of immiscible fluids, proceeding by the mechanism of impregnation and drainage in layered-heterogeneous reservoirs with hydrodynamically connected and separated interlayers.
 Physical modeling of filtration processes, taking into account the similarity criteria, made it possible to establish that during the mutual substitution of immiscible fluids at the phase boundary and lithological differences, nonequilibrium capillary phenomena develop, which, through the action of capillary forces, significantly affect fluid filtration and saturation distribution in an inhomogeneous porous medium. In order to eliminate the influence of the viscous instability factor on the motion of the displacement front, the processes of displacement of a less viscous fluid by a more viscous one were studied.
 It has been established that non-equilibrium capillary phenomena manifest themselves most intensively in reservoirs with a heterophilic surface, where capillary forces can initiate or block fluid flows between differently permeable interlayers. It has been experimentally shown that in layered-heterogeneous reservoirs with a heterophilic filtration surface in the interlayers, capillary phenomena largely determine the nature of the movement of the fluid displacement front (from the almost complete suppression of the “run-up” of the displacement fronts in the interlayers to the formation of an advanced front in the low-permeability interlayer). In order to exclude the influence of errors of a systematic or random nature on the results of the experiments, their reproducibility and the significance of differences were assessed using the statistical criteria of Student and Cochran, which showed the reliability of the above conclusions.
 The results obtained indicate the dominant effect of crossflows between interlayers on the formation of the displacement front of immiscible fluids in layered-heterogeneous reservoirs, especially with a heterophilic wettability characteristic, as well as the possibility of controlling this process by means of physicochemical modification of the wettability conditions of the filtration surface, for example, using surface active substances (surfactants) of various nature and structure.

Tensor-Train-Based Multiuser Multivariate Multiorder Physical Markov Process Informed Multimodal Prediction for Industrial Trajectory Applications

By combining data driven approaches and physical laws for industrial trajectory prediction can improve the performance of industrial applications, such as path planning of robots, routing selection of vehicles in transportation system. In the era of industrial big data, the application of industrial trajectory prediction based on physical Markov process and tensor model has attracted much attention. To synchronously improve the prediction accuracy and computational efficiency, this paper proposes a tensor-train (TT) based multiuser multivariate multiorder (3M) physical Markov prediction approach for multimodal industrial trajectory pattern mining. First, we propose a TT-based unified product calculation rule with its scalable computation approach based on decomposed TT cores to speed up the execution efficiency. Then, a TT-based 3M (TT-3M) Markov transition approach is presented. Furthermore, we put forward a TT-based power method to calculate the stationary joint eigentensor (SJE) and an SJE-based multimodal prediction algorithm to mine concealed trajectory patterns. Experimental results based on real GPS trajectory dataset show that compared with the original tensor-based 3M approach, the TT-3M approach can improve the computational efficiency up to 3 times and reduce the memory space proportion to a minimum of 1‰ while ensuring basically consistent prediction accuracy.

Evaluating the Precipitation Biases over the Western Periphery of the Sichuan Basin with the ECMWF Operational Forecast Model

Abstract Based on the hourly merged precipitation product, the performance of the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecasting System (IFS) in simulating the diurnal variations of precipitation during warm season over the western periphery of the Sichuan basin (SCB) has been evaluated, and the underlying physical causes associated with the wet biases have also been investigated. The results show that the IFS well reproduces the spatial distributions of precipitation amount, frequency, and intensity over the SCB, as well as their diurnal variations, but the simulated precipitation peaks earlier than the observation with notable wet biases over the western periphery of the SCB. In addition, the strong wet biases exhibit notable regional differences over the western periphery of the SCB. The simulated wet biases over the southwestern periphery of the SCB expand westward to higher altitudes along the windward slope, with the maximum wet biases occurring at night. The westward expansion of the simulated stronger upward motions results in a westward shift of precipitation. However, the simulated precipitation over the northwestern periphery of the SCB has small difference in terms of the location; hence, the overestimated precipitation is associated with the stronger atmospheric instability, resulting from the higher potential temperature and the larger specific humidity near the surface. The findings revealed in this study indicate that the ECMWF forecast shows distinct uncertainties over the different complex terrain, and thus offers a promising way forward for improvements of model physical processes.

Three-dimensional physical modeling of the wet manufacturing process of solid-state battery electrodes

Poor design of composite electrodes in Solid-State Batteries (SSBs) is one of the main reasons for their low performance. Although modeling techniques offer great potential in investigating multiple conditions for microstructure optimization, SSB manufacturing remains almost unaddressed in terms of computational studies. We present here a three-dimensional physics-based modeling workflow to investigate the impact of wet manufacturing process parameters on the properties of SSB tape casted composite electrodes based on LiNi0.6Mn0.2Co0.2O2 as the active material and Li6PSCl5 sulfide solid electrolyte. Our proposed workflow models the entire process of wet manufacturing of SSB electrodes, starting with the slurry consisting of active material, carbon additive, binder, solid electrolyte and solvent, and continues through its drying process, and the calendering of the resulting electrodes. Our focus is in particular on the impact of the calendering degree on the microstructure of the electrode. We characterize the resulting microstructures in terms of electronic and ionic conduction properties. We believe that this first-of-its-kind wet manufacturing process model for SSB cathodes is an important step towards the development of systematic modeling approaches that can provide practical optimization of the interfaces between materials in electrodes to improve SSB performance and durability.

Boundary Graph Neural Networks for 3D Simulations

The abundance of data has given machine learning considerable momentum in natural sciences and engineering, though modeling of physical processes is often difficult. A particularly tough problem is the efficient representation of geometric boundaries. Triangularized geometric boundaries are well understood and ubiquitous in engineering applications. However, it is notoriously difficult to integrate them into machine learning approaches due to their heterogeneity with respect to size and orientation. In this work, we introduce an effective theory to model particle-boundary interactions, which leads to our new Boundary Graph Neural Networks (BGNNs) that dynamically modify graph structures to obey boundary conditions. The new BGNNs are tested on complex 3D granular flow processes of hoppers, rotating drums and mixers, which are all standard components of modern industrial machinery but still have complicated geometry. BGNNs are evaluated in terms of computational efficiency as well as prediction accuracy of particle flows and mixing entropies. BGNNs are able to accurately reproduce 3D granular flows within simulation uncertainties over hundreds of thousands of simulation timesteps. Most notably, in our experiments, particles stay within the geometric objects without using handcrafted conditions or restrictions.