Altair EDEM Professional 2022.0

EDEM GPU Solver (CUDA) Multi-GPU Solver Engine
The newly developed EDEM Multi-GPU Solver (CUDA) enables users to solve bigger simulations with more particles than ever before. By making full use of the latest Nvidia graphics cards, this cutting-edge EDEM Multi-GPU Solver automatically balances computation workload and memory usage on the GPU devices to maximize the utilization of the hardware. This doesn't require any additional user input and delivers good scalability across multiple GPU cards with larger particle numbers. The EDEM Multi-GPU Solver is compatible with Polyhedral, Sphero-Cylinder and MultiSphere particles as well as supporting most of EDEM API functionality. For full details of the GPU capabilities, see the table at the end of this document.
Database for Powder Material Models
A database of powder material models has been added to EDEM. This is a set of materials which can be sorted based on the static angle of repose and the steadystate shear stress responses. Once the appropriate result has been selected, scaling rules can be applied to the result to generate an appropriate material model for the application.
EDEMpy Custom Properties for post processing
EDEMpy now supports the option to Add, Remove and Edit custom properties in a simulation deck. This can be used in post-processing to add user defined timedependent attributes to particles, geometries, contacts, or the whole simulation. This information is then saved within the deck, where it can be visualized and exported using the standard EDEM Analyst tools.
Volume Packing improvements
The Volume packing functionality has been extended to support Bulk Materials which have a User-Defined Particle Size Distribution, support periodic boundaries and no longer require a material block to be generated. It is still possible to generate a material block from the generated material, however this is not done by default. This will improve performance and avoid the material database increasing in size. The Periodic boundaries can be used in conjunction with the block factory tool to generate large beds of materials or generate silos with a defined solid fraction. This allows simulations where there is a repetition in the material to be set up much quicker. The Volume Packing tool now supports the Bonded and Heat Transfer Physics models. This means that volumes can be filled using the packing tool and then bonded and volumes can be generated with particles which have specific temperatures and heat flux values.
Tavares UFRJ Breakage Model for CPU
The Tavares UFRJ Breakage model has been added to the internal physics models. This model captures the various body breakage mechanisms that occur during particle collisions. Specifically, it describes the adaptation of a detailed breakage mechanism of brittle materials. It accounts for the variability and sizedependency in breakage probability and weakening by repeated stressing, and delivers the final size distribution of the material.Analyst tools have also been added to better understand the breakage in the model. These tools include the ability to plot and export particle size distributions in Breakage simulations. This incorporates the distribution of "fines": particles that are below the minimum cut-off diameter of the simulation. Although fines are too small to participate in the simulation, their mass and size is still accounted for in this distribution.
Wet Mixing Model (Experimental)
A wet mixing model has been added to the experimental physics models. The model is primarily aimed at concrete mixing. Particles with different moisture contents are added to the system. The transfer of moisture between particles is modeled and the cohesive interaction between the particles is modified depending on the amount of moisture.
Spray Coating for Polyhedral, Multi-Sphere and SpheroCylinder Particles
A spray coating model has been added to the Physics models. This model captures a spray of coating material hitting particles and adding that mass to those particles. This model was previously available as an API model on the user forum, this is now available within EDEM and has been updated to be compatible with all particle types and is available from the GUI. The transfer of spray between coated particles is not modelled, however a post-processing Python model is available to analyze this.

Creator

Use CAD file name for Geometry Sections
An option has been added to use the CAD file name for geometry sections when importing them into EDEM. By default, EDEM will continue to use the part name located within the CAD file itself. This option will allow more control of geometry names if the name within the CAD file is not found. If the CAD file is composed of more than one geometry section a numerical suffix will be added to the name.
Particle Limits applied periodically
Particle limits can now be looped: they can be applied periodically throughout the simulation. This can be used to remove excess movement of particles from the simulation. This is particularly aimed at scenarios where large numbers of particles are placed quickly into a simulation and would normally result in a particle explosion. This acts as a calming function, where velocities are repeatedly reset to a low value until the overlap is small.
EDEM File Compression
The 'saving last timestep' message will prevent EDEM from closing until the data compression is complete, this ensures all time step data is compressed and that users benefit from the reduced file size data compression offers.
Meta-Particles for Sphero-Cylinder and Multi-Sphere Particles
The GPU CUDA solver can generate Meta-Particles composed of Multi-Sphere or Sphero-Cylinder particles. Using these with the Bonded V2 physics model will allow users to simulate flexible particles with complex shapes. The Meta-Particle interface has been added to the Sphero-Cylinder options. Multi-GPU factories will not generate Meta-Particles which cross the boundary between two GPUs. This will affect the random nature of particle generation.

Simulator

Fixed velocity particles
Particles now have the option to preserve the velocity they are generated with. This means that particles generated will travel at a constant speed. This is applied on a per particle basis. This allows some particles to have a fixed velocity. A notable application using this feature is fiber modeling, where the 'root' of the stem can have a fixed velocity and the rest of the fiber is free to move. Plot Particle Size Distribution against the expected distributionThe solve report now includes graphs comparing the particle size distribution of each particle type against the distribution that was requested. This helps users identify cases of factories being too small to allow larger particles to be generated. This is applicable to all solvers.
Multi-Sphere Factory using Physical Radius for CUDA solver
The ability to use the physical radius option for factories has been added to the CUDA solver, this matches the behavior of the CPU and OpenCL solver.
Preserve contact across elements on CUDA
An option to preserve contacts across geometry elements has been added to the Multi-Sphere CUDA solver. This preserves the contact history between the particle and the geometry and eliminates the contact force spikes when a particle moves between elements.

Analyst

Copy particle ID into manual particle selection list through "Copy and Paste" style
The ability to paste particle IDs into the manual selection has been added to the Analyst. This functionality allows the study of specific individual particles or groups of particles which would have otherwise been difficult to select.
Polyhedral Particle contact attributes
Three additional attributes are available for Polyhedral particle contacts. These are the Penetration Depth, Overlap Volume and Contact Normal. These attributes can be used in the Analyst Queries, Data Export, Selection Contact data and contact coloring. Only the Contact Normal coloring is available when using legacy rendering.
Rendering improvement for Cones and Vectors
Rendering of particle cone, particle vector, and contact vector representations has been improved to be smoother.

GPU

Residence Time on EDEM GPU Solver (CUDA)
The particle data query 'residence time' tracks the amount of time each particle has spent in the simulation. Previously this was not implemented for the CUDA solver, but it has now been implemented and may be used in the same way as for the other solvers. Note that decks with pre-existing particles created using the CUDA solver and earlier versions of EDEM will not have correct data for this property. Only new simulation runs with the CUDA solver will have correct values.
EDEM GPU Solver (CUDA) Direct Force Adding
An option to 'Use Direct Force Adding' has been added to the CUDA solver. This reduces memory usage, particularly in simulations with wide particle size distribution. This functionality requires a Graphics Card which supports CUDA Compute Capability 6.0 (Pascal architecture or above). Older cards will continue to run using the CUDA solver, however they will not be able to use this feature.
Note: If a contact model is using a particle custom property value to calculate the same custom property value delta, then the contact calculation can get an updated particle custom property value instead of the original one. None of the internal models are impacted, as the same value is not read and modified at the same time. Using this option can lead to minor variations between repeated simulation runs, however this variation will not affect the overall bulk behavior.
EDEM GPU Solver (CUDA) Performance for Size Distributions
Alternate optimizations have been made to the EDEM GPU Solver (CUDA) to improve performance, these include contact sorting in global memory rather than shared memory and significant improvements in the contact detection filtering. These changes are particularly apparent where there is a large particle size distribution.

Add a warning for Block Factory usage with all CUDA solvers
When using Block factories on the CUDA solver, they will generate particles, but they currently don't have contact information. When a simulation is run on CUDA using block factories, a warning message will appear.
Default Solver Precision settings in the CUDA Engine Settings
Default Solver Precision settings moved to CUDA Engine Settings. OpenCL Alternative contact detection will use double by default in new decks.
Calibration Kit - Cone Penetrometer Test
The Cone Penetrometer method has been added to the Calibration Kit. This test is used for characterizing ground materials. The kit contains an input deck and EDEMpy post processing kit. This can be found in the EDEM installation folder and can be used in conjunction with EDEMCal to develop a ground material model which matches the experimental cone penetrometer results.
Updated Calibration Kits
Calibration Kits have been updated to the latest version. These updates include removing any specific IDE requirements and using the latest version of EDEMpy.

Deprecated Features

Linear Cohesion contact model removed
The Linear Cohesion model has been removed from EDEM. This has been replaced by the Linear Cohesion model V2 which addresses the physics issues in the Linear Cohesion model. If a simulation is opened which had Linear Cohesion selected in the physics section, the model will automatically be converted to Linear Cohesion V2. The results will differ between the two contact models, particularly when using a size distribution.
Sentinel (legacy) licensing removed
The legacy license system is no longer supported. EDEM is now available on the Altair Units licensing system and will not work with the legacy license system. Users still on legacy licenses who are paying maintenance should contact their account manager.

EDEMpy

EDEMpy retrieve physics properties of EEPA model
A function has been added to return all user defined parameters for the EEPA model. The existing equivalent function for getting Hertz-Mindlin with JKR parameters has also been updated to return the whole data set instead of just the first entry.
Retrieve Kinematic properties using EDEMpy
A kinematics class has been added. This contains functions to get information on kinematic attributes, start point, end point and point of application from the EDEM simulation.
Implement "getPressure" method in EDEMpy
A getPressure() method has been added to the timestep geometry class which returns an array containing pressure per triangle.

EDEM 2022 contains fixes for the following issues:
- Using volume packing with the position optimization option selected would not fill the volume properly.
- Particle generation failed when generated particles intersected dynamic domain boundaries.
- The auto-grid size option defined in the GUI was not overwritten by the grid cell size written in the command line. Using the auto-grid and -g flags together in the command line interface will now give the user a warning. Using the -g flag in the command line interface will automatically set auto-grid to false.
- An issue could occur with the transfer of tangential overlap between timesteps, this could lead to the contact not acting exactly perpendicular to the contact normal. This issue was not common but may result in a change of trajectories to individual particles.
- Particles could be generated outside the volume when using the volume packing tool with volumes with sharp edges.
- The CUDA solver would produce particles with large ID numbers, they are now assigned ascending and continuous ID values.
- Simulations with large particle size distributions would fail when run on the GPU solver (CUDA).
- The graph settings would not be preserved in the interface. This meant that graph settings would need to be re-entered multiple times, these setting are now saved, so graphs can be modified without needing redefined.
- Static factories were never marked complete in the solve report when the factory was defined with 0 particles.
- Simulation data could become corrupted at the end of a GPU solver simulation if the coupling disconnected, or the simulation stopped unexpectedly.
- A crash could occur when kinematics were added to a simulation through the Coupling Interface.
- A crash could occur if multiple geometries were deleted when contacts existed with those geometries.
- Contacts were not hidden when using the 'Hide Out of Bounds' coloring option and contact transparency was not supported using standard or instanced rendering.
- Rendering Sphero-Cylinder particles in legacy mode or while simulating could cause EDEM to become unresponsive.
- A crash could occur when using the Hysteretic Spring model on the CUDA solver, due to invalid values for the tangential overlap.
- Simulator settings were not fully transferred to new versions of EDEM.
- A crash could occur when selecting a Meta-Particle before selecting a particle.
- Particles of interest would not be fully populated with all custom properties on the GPU CUDA solver.
- The 'Contacting Neighbours' custom property was removed from the EEPA contact model, this was an unused custom property, but will still be visible for old simulations in the Analyst.
- Big simulations run using the CUDA solver used a large amount of memory on the GPU.
- The geometry position could be affected by interacting with it in the Simulator and transparent geometries did not appear when switching to the simulator.
- The EDEM coupling failed to connect to RecurDyn VR4. Users are now required to open the 'EDEM1_1_0.xml' in a text editor and add the full path to the EDEMRecurDynCouplingClient.dll where it is indicated to do so.
- An issue could occur with simulations containing multiple factories when run with the CUDA solver. If there was more than 1 active factory generating particles, then the particles could be created in the wrong place.
- The Edinburgh Elasto-Plastic Adhesion model would report an error when using a User Defined Size Distribution sum was not exactly 100%.
- A crash could occur when using the Volume Packing with particle containing very small spheres.
- The API function isSphere() would return false where is should have returned true.
- The Total Over Time query in the Analyst was not displayed in the query name, leading to misleading query titles.
- Simulations could not be saved to a network drive while using data compression if hard drive was not mounted or mapped.
- A crash could occur when recording a video using older and compute GPU cards.
- An incorrect particle size distribution could be generated when scaling by radius was used in the user defined size distribution options.
- Factories on the CUDA solver would not behave correctly in certain situations.
- The API function getNodeIDs() would return false where is should have returned true.
- Export to STL did not account for partial time steps.
- If a material block contained bonded particles, only the first block would be created with bonds.
- The last time step could become corrupted if it's selective and compression were used.
- The default coloring for Stress Tensor, Axial Stress and Von Mises attributes would be a uniform color unless minimum and maximum values were updated.
- Memory usage was high when using Geometry Bin Groups.
- Expired coupling motions or deleted kinematics could leave a velocity on simulation elements when running on the CUDA solver. This may also cause some very minor results differences on some existing CUDA simulations, however these will not be significant.
- There was inconsistent behavior between Sphero-Cylinders and Polyhedral simulations when run on command line with the invalid flag to use the OpenCL solver.
- Particles could explode when using periodic boundaries with the CUDA solver.
- Meta-particle factories could corrupt a simulation deck if a particle size distribution was used, and the particles type was updated in the factory.