Time-Dependent Deep Learning Manufacturing Process Model for Battery Electrode Microstructure Prediction

In simple shear testing, the specimen boundaries play a pivotal role in the transmission of shear forces. Waffle-style porous stones, plates with ribs or similar types of projections are used in experiments to reduce slippage at the boundary and transmit shear throughout the specimen. Conventionally, Discrete Element Method (DEM) simulations often model the top and bottom caps as flat boundaries with artificially enlarged friction coefficients, or use geometrical configurations that are computationally efficient. This research compares flat high-friction boundaries with ribbed and more novel-designed boundaries incorporating large pyramid, and small pyramid projections, aiming to improve shear transmission capability while ensuring computational efficiency. DEM simulations were conducted on specimens of steel bearings, with experimental validation using identical setups and particle properties. Specimens featuring different boundaries showed different void ratios in the layers close to the boundaries, with boundary effects diminishing towards the central zones. DEM simulations with projection boundaries demonstrated good agreement with experiments in terms of the macroscopic response, validating the effectiveness of the projection boundaries. The conventional flat boundaries exhibited limited shear transmission capability, resulting in insufficient development of shear stress and inadequate particle engagement. Conversely, projections on boundaries significantly improved shear stress transmission, ensuring the simple shear condition throughout the entire specimens. Projection boundaries introduced manageable increases in computational cost despite increased mesh complexity. This study highlights the importance of boundary design and recommends the adoption of projection-based boundaries in both experimental and numerical simple shear tests to ensure effective shear transmission and reliable results.

Modelling and simulation of fruit drop tests by discrete element method

Biomass has potential as a carbon-neutral alternative to petroleum for chemical and energy products. However, complete replacement of fossil fuel is contingent upon efficient processes to eliminate undesirable characteristics of biomass, e.g., low bulk density, variability, and storage-induced quality problems. Mechanical size reduction via comminution is a processing operation to engineer favorable biomass flowability in handling. Crumbler rotary shear mill has been empirically demonstrated to produce more uniformly shaped particles with higher flowability than hammermilled biomass. This study combines modeling and experimentation to unveil fundamental understandings of the relation between granular particle characteristics and biomass flow behavior, which elucidate underlying mechanisms and guide selection of critical processing parameters. For this purpose, the impact of critical material attributes, including particle size (2–6 mm), particle shape (briquette, chip, clumped-sphere, cube, etc.), and surface roughness, on the angle of repose (AOR) of milled pine chips were investigated using discrete element method (DEM) simulations. Forest Concepts Crumbler rotary shear system is used to produce milled pine particles within the same size range considered in DEM simulations. AOR of different sets of these particles were measured experimentally to benchmark DEM results against experimental data. Specific energy consumption for the comminution of biomass with different particle size and moisture content are measured for technoeconomic analysis. Our results show that the smaller size (2 mm) of pine particle achieves better followability (i.e., smaller AOR) while the energy cost of comminution is significantly higher and bulk density is almost the same as the 6-mm pine particles. For the 2-mm particle size, Crumbles from veneer have better flow properties than Crumbles from chips. Contrarily, no significant difference was observed between the AOR of the two materials for the 6-mm particle size. Furthermore, from DEM simulations, mechanical interlocking between particles was found as a dominant factor in determining AOR of complex-shaped particles such as milled pine, which cannot be accurately captured by using simple particle shapes (e.g., mono-sphere) with a rolling resistance model. Conversely, clumped-sphere model alleviates this limitation without increasing computational cost significantly and can be used for accurate representation of biomass granular particles when simulating free-flow behavior.

Influence of calendering process on the structural mechanics and heat transfer characteristics of lithium-ion battery electrodes via DEM simulations

Improvement of a pharmaceutical powder mixing process in a tote blender via DEM simulations

The presence of overlarge gravel particles poses significant challenges for laboratory testing on prototype gravelly soils due to sample size limitations. To address this issue, replacement techniques, such as substituting overlarge particles with finer materials, offer practical solutions. However, the impact of these techniques on the mechanical behavior of gravelly soils, particularly shear strength and stiffness, remains poorly understood. This study aims to bridge this knowledge gap by investigating the particle size effect on the shear behaviors of binary mixtures using a series of Discrete Element Method (DEM) simulations. Updated scaling relations, based on Iai's generalized scaling relations, were proposed to correct for particle size effects. DEM simulations, including drained triaxial tests and shear modulus measurements, were performed to validate the proposed law. The results indicate that the gravel replacement technique has a minor effect on peak shear strength but significantly reduces soil stiffness, especially at high gravel contents. The scaling relations effectively correct for the particle size effect, enabling the accurate prediction of shear behaviors of the prototype gravelly soils from those of the model gravelly soils. These validations demonstrate that for addressing the soil deformation problem instead of the stability problem in ultimate state, the developed scaling relations are highly effective for correcting the particle size effect. Based on the developed scaling relations, engineers can predict prototype-scale shear behaviors of gravelly soils with overlarge particles using scaled laboratory models, reducing reliance on costly large-scale equipment. Additionally, future studies, through both DEM simulations and laboratory experiments, are recommended to further validate and refine the proposed method across diverse soil conditions and loading scenarios, such as cyclic loadings.

Hybrid residual deep learning models with physical knowledge for improving plant transpiration estimation

This research article proposes a deep learning framework that addresses two major hurdles in adopting deep learning techniques for solving physics-based problems. One is the requirement of a large data set for training the deep learning (DL) model and another is the consistency of a DL model with the physics of a phenomenon. The framework is generic that can be applied to model a phenomenon in physics if its behavior is known. A semi-supervised physics guided neural network (SPGNN) has been developed based on our framework to demonstrate the concept. SPGNN models the I–V characteristics of gallium nitride based high electron mobility transistors (GaN HEMTs). A two-stage method has been proposed to train a DL model. In the first stage, the DL model is trained via an unsupervised learning method using the analytical physics-based model of a field-effect transistor (FET) as a loss function of the DL model that incorporates the physics of the FET in the DL model. Later, the DL model is fine-tuned with a small set of experimental data in the second stage. Performance of SPGNN has been assessed on various sizes of the data set with 100, 500, 1000, 1500, and 2000 samples. SPGNN significantly reduces the training data requirement by more than 80% and provides better performance than a traditionally trained neural network (TTNN), even for the unseen test data set. SPGNN predicts 32.4% of the unseen test data with less than 1% of error and only 0.4% of the unseen test data with more than 10% of error.

Continuum modeling of granular segregation during hopper discharge

This paper describes an entire procedure to realize an image-based DEM (Discrete Element Method) simulation with lunar soil. Since the detailed grain shape information of the returned sample of the lunar soil is not available, one of its simulant, FJS-1, is used in this study. High resolution micro X-ray CT system at SPring-8, a synchrotron radiation facility in Japan, makes it possible to visualize 3-D image of FJS-1 granular assembly in details. Then a newly developed image analysis procedure is adopted to identify each individual grain from the others. Using the obtained grain shape data, a sufficient number of FJS-1 grains are directly modeled for DEM simulation. A preliminary simulation is performed using these FJS-1 DEM grains and its performance is discussed.

Seeded granulation process in twin-screw granulators: A DEM approach

Unwanted grain commingling impedes new quality-based grain handling systems and has proven to be an expensive and time-consuming issue to study experimentally. Experimentally validated models may reduce the time and expense of studying grain commingling while providing additional insight into details of the particle flow. In this study, grain commingling in a pilot-scale bucket elevator boot was first modeled in three-dimensional (3-D) discrete element method (DEM) simulations. Experiments on the pilot-scale boot were performed using red-dyed and clear (uncolored) soybeans to validate the 3-D DEM model. Predicted results from the 3-D boot model generally followed the experimental data but tended to underpredict commingling early in the process. To reduce computational time, quasi-two-dimensional (quasi-2-D) DEM simulations were also evaluated. Comparison of predicted average commingling of five quasi-2-D boot models with reduced control volumes (i.e., with boot widths from four to seven times the mean particle diameter) led to the selection of the quasi-2-D model with a boot width of 5.6 times the mean particle diameter (i.e., 5.6d) to reduce computational time. In addition, the 3-D and quasi-2-D (5.6d) models were refined by accounting for the initial surge of particles at the beginning of each test and correcting for the effective dynamic gap between the bucket cups and the boot wall. The quasi-2-D (5.6d) models reduced simulation run time by approximately 70% compared to the 3-D model of the pilot-scale boot. Results of this study can be used to accurately predict commingling levels and improve grain handling, which can help farmers and grain handlers reduce costs and maintain grain purity during transport and export of grain.

In this paper, a new procedure for the prediction of soil-tool abrasive wear is presented which drastically reduces the duration and, therefore, the cost of simulations in comparison to conventional 3D wear modeling. The goal is to extend the experimental data from a single scratch test to the wear of mixtures by means of equations obtained from discrete element method (DEM) simulations and geometric relations. We are predicting abrasive wear with a combination of numerical and experimental approaches taking two shapes of particles into account. Single wear is quantified by measuring the width of scratch induced by a single quartz particle. Geometrical relations together with the particle’s microscopic picture are used to find the depth of scratch. DEM mixture simulations result in equations for the number of contacts and normal contact forces. Finally, the wear rate is calculated for a specific soil sample as an example to clarify the developed prediction procedure. The DEM simulations are performed using PFC $$^{3{\text {D}}}$$ code for both a homogeneous soil sample and a mixture of two different soils. We are specially investigating a relation to predict the abrasive wear caused by a mixture of particles.

SUMMARYIn discrete element method (DEM) simulations of real scale, the spherical particles are commonly employed for increasing the computation speed, and the complex boundary models are represented by triangle meshes with controllable accuracy. A new contact detection algorithm has been developed to resolve the contacts between the spheres and the triangle mesh boundaries. The application of the barycentric coordinates makes this algorithm more efficient to identify contacts in the intersection test. As a particle probably collides with several triangles at the same time, the multiple contacts would be reported as face contacts, edge contacts, or vertex contacts. Moreover, the particle embedding in a triangle can be also contact with the edges or vertices of the next triangles. These contacts should be considered as invalid for updating contact forces in the DEM. To exclude invalid records from the multiple contacts, the algorithm gives attention to the mesh structure nearby contacts and analyzes all possible collision situations. Numerical experiments have been conducted to verify this algorithm by using the algorithm in the DEM simulation framework. The numerical results suggest that the algorithm can resolve all contacts precisely and stably when the spherical particles collide on the complex boundary circumstances. Copyright © 2013 John Wiley & Sons, Ltd.

An impact energy erosion model with an energy allocation rule for the discrete element method