2. Explicit Dynamics Workflow
2.14. Postprocessing
2.14.6. User Defined Results for Explicit Dynamics Analyses
For general information about User Defined Results, see User Defined Results in the ANSYS Mechanical User's Guide
Shown here are the User Defined Results that are specific to an explicit dynamics analysis using the Autodyn solver. The number of nodes bonded to the faces on an element
during the analysis. A value of -1 is shown where all the bonds for the face have broken.
BOND_STATUS
Element Nodal Beam cross section area
C_S_AREA Beam cross section number
CROSS_SECTION Effective Geometric Strain of a cell
EFF_STN
Element Nodal Effective Plastic Strain. Note: This is calculated
incrementally, unlike the equivalent plastic strain
(EPPLEQV), which is calculated as an instantaneous value.
EFF_PL_STN
Element Nodal Energy resulting from fracture for the Johnson-Holmquist
brittle strength model
>0 - eroded. (will not be displayed)
Type Description
Variable
Element Nodal Effective Plastic Strain Rate
EPS_RATE Internal energy of the material
INT_ENERGY
Element Nodal Mass of material in an element
MASS
Elemental Material index. The material index as defined in the
Explicit solver. There is not always a direct one-to-one MATERIAL
correlation with materials defined in Engineering Data and those used in the Explicit solver.
For layered section shells, the MATERIAL for individual layers can be shown by using the Layer property in the results details view.
Element Dilation pressure for the Johnson-Holmquist brittle
strength model
2 – undergoing plastic flow
3 – failed due to effective criteria (with healing) 4 – failed due to effective criteria
5 – failed due to stress/strain in principal direction 1 6 – failed due to stress/strain in principal direction 2 7 – failed due to stress/strain in principal direction 3 8 – failed due to shear stress/strain in principal direction 12
9 – failed due to shear stress/strain in principal direction 23
Type Description
Variable
10 – failed due to shear stress/strain in principal direction 31
For layered section shells, the STATUS for individual layers can be shown by selecting the Layer number in the results details view.
Elemental Stochastic factor applied when the stochastic property as
defined in the material failure model STOCH_FACT Total strain XY. These are tensor shear strains, and not
engineering shear strains.
STRAIN_XY
Element Nodal Total strain YZ. These are tensor shear strains, and not
engineering shear strains.
STRAIN_YZ
Element Nodal Total strain ZX. These are tensor shear strains, and not
engineering shear strains.
STRAIN_ZX
Element Nodal Shell total strain XX, sub-layer #. These are tensor shear
strains, and not engineering shear strains.
SUB_STN_X_SHELL_LAYER__#
Element Nodal Shell total strain YY, sub-layer #. These are tensor shear
strains, and not engineering shear strains.
SUB_STN_Y_SHELL_LAYER__#
Element Nodal Shell total strain ZZ, sub-layer #. These are tensor shear
strains, and not engineering shear strains.
SUB_STN_Z_SHELL_LAYER__#
Element Nodal Shell total strain XY, sub-layer #. These are tensor shear
strains, and not engineering shear strains.
SUB_STN_XY_SHELL_LAYER__#
Element Nodal Shell total strain YZ, sub-layer #. These are tensor shear
strains, and not engineering shear strains.
SUB_STN_YZ_SHELL_LAYER__#
Element Nodal Shell total strain ZX, sub-layer #. These are tensor shear
strains, and not engineering shear strains.
SUB_STN_ZX_SHELL_LAYER__#
Element Nodal Effective plastic strain, sub-layer #
SUBL_EPS_SHELL_LAYER_# Element category (element number returned)
TYPE
HEX: 100-101 PENTA: 102 TET: 103-104,106
Type Viscous pressure due to artificial viscosity. No results will
display for an Eulerian part.
VISC_PRES
For Euler (Virtual) Analyses
The following results are multi-material variables in the Autodyn solver.
• EFF_PL_STN
For each Eulerian (Virtual) body in the analysis, a separate component will be available, which will allow the user to plot the result for the particular material associated with that body. The component name will be derived from the body name. There will also be an “ALL” component, which will displays results for all materials. Results for Lagrangian bodies can be viewed by selecting this “ALL” component. For a purely Lagrangian analysis, only the “ALL” component will be available to the user.
For example, an analysis has two Eulerian (Virtual) bodies (Solid, Solid) and a Lagrangian Body (Surface Body), as shown in the image of the Outline View below.
In the User Defined Result Expression Worksheet, there are three components available for the multi-material results, named SOLID, SOLID_2, and ALL.
Note
It may be necessary to delete and reinsert multi-material results in order to view result for databases created prior to Release 13.0
For NBS Tetrahedral Elements
The element variables listed below can be used to visualize the variable values at the nodes. The variable values presented in the element are a volume weighted average of those at the nodes.
• TEMPERATURE
• SOUNDSPEED
• DENSITY
• COMPRESS
• STRAINS (NORMAL AND SHEAR)
• EFF_PL_STN
• TIMESTEP
• INT_ENERGY
The following variables are available as calculated directly from the solver in the element:
• EFF_STN
Dynamics
3.1. When Implicit Models Can be Run in Explicit
Implicit and Explicit finite element solvers use different methods to evaluate the underlying equations.
A simple high level overview is given in the figure below. There is an overlap in the "Quasi-Static" ap-plication area, where both Implicit and Explicit methods can be used to solve a model. Implicit methods are typically bounded by the amount of deformation and contact nonlinearity that is taking place, where Explicit methods are typically bounded by the problem's time scale, which would lead to excessive run times.
Figure 3.1: Different applications of the two solvers with respect to velocity
Problems that are in this "Quasi-Static" range have a good chance of being solved by either method until the limitations of a particular solver are reached. At that point, it can be beneficial to consider the use of the alternative solver.
This chapter describes the steps necessary to transform a model that was initially set up for simulation in the Implicit solver to a model setup for simulation in the Explicit solver. Typically, you would want to consider doing this when the degree of nonlinearity in the model is starting to pose problems for Implicit methods. Because of the nature of the two methods, the explicit solver is more suitable for nonlinear problems, working with less computationally heavy but a much larger number of iterations that can follow the physical parameter changes at a much higher frequency. The implicit solver work with much more complex calculations for each iteration but has a lot fewer of them.
3.2. When to Consider an Explicit Analysis
There are numerous reasons why an implicit methods fails to solve properly. This section tries to give realistic guidelines for when to switch to an Explicit method.
3.2.1. Incorrect Model Setup 3.2.2. Large Deformations 3.2.3. Large Contact Models 3.2.4. Rigid Body Deformations
3.2.1. Incorrect Model Setup
A model may fail when the Implicit method is used simply due to poor model setup, in which case Ex-plicit methods will fail also. However, an incorrect model setup may be easier to detect with an ExEx-plicit analysis because the solution progresses with very small timesteps, and results can be visualized during the solution (by using result trackers (p. 105), or using the Autodyn component system). Once the problem is identified using the explicit dynamics analysis, it can be corrected and solved using implicit methods.
The explicit dynamics solver is very useful when working with complex interacting mechanisms and geometries. The solver can be used to quickly check for fit and how the parts are positioned with respect to each other at the end of the simulation.
The example model shown in Figure 3.2: Example model run with explicit dynamics showing problem area (right) (p. 89) does not converge when run with the Static Structural (implicit) solver. The output messages recommend checking for an 'insufficiently constrained model'. The geometry has multiple angles and edge lengths so the problematic area is not obvious. This is a good example of where explicit dynamics methods can be used to quickly identify model problems. The same geometry with the same setup is run with the default mesh using the explicit solver. In cases like this, where we only need to observe the general fit and alignment and part interaction in the setup, an arbitrary end time value is set (in this case 1 millisecond). Starting from 1 second, the end time is reduced to give a satisfactory quick 'Est. Clock Time Remaining' value in the solution information, interrupting the solve before each change. In all other cases the user should make sure that the end time and boundary conditions are set properly (for further discussion into analysis and setup settings see Analysis Settings (p. 98) and Boundary Conditions (p. 96)).
The endtime is chosen to obtain a fast solution in order to observe the relative movement of the parts (for further discussion into time scaling see Adjusting Load Cases for Reasonable Run Times (p. 96)).
Figure 3.2: Example model run with explicit dynamics showing problem area (right)
After a run time of 1.2 minutes, the model solves with all options being left as default. The problematic area is obvious as can be seen above: the left notch of the upper part does not follow the bottom part geometry. This quickly points to where a change of the geometry is necessary.
3.2.2. Large Deformations
Many models require the simulation of rubber-like highly deformable materials. This is associated with the use of hyperelastic material models in the setup. The implicit solver makes a strong effort to solve these models with options like 'Large Deflection' and 'Nonlinear adaptivity', which are recommended when such materials are used. Nevertheless these solutions may not converge.
This would be a suitable situation to use explicit dynamics. It does require to specify all the input for the hyperelastic materials as opposed to the implicit solver, where the density and the incompressibility parameter can have zero values (see Materials (p. 93)), but it will provide a solution in most cases where the implicit cannot. Important things to look out for in the explicit solver when using hyperelastic ma-terial models are the energy error/hourglassing and excessive mesh element distortion requiring the use of erosion (p. 102). Models with high nonlinear deformations are also a good candidate for mass scaling (p. 101). The following example demonstrates how the same setup works with the two different solvers.Figure 3.3: Comparison between the implicit (left) and the explicit (right) solvers for maximum deformation values (p. 90) shows the highest displacement achieved of the disc relative to a hypere-lastic material complex part.
Figure 3.3: Comparison between the implicit (left) and the explicit (right) solvers for maximum deformation values
The implicit solver has troubles converging at around half of the total displacement and the explicit solver manages to go through the run simulating the high deformations.
3.2.3. Large Contact Models
Handling a large number of contacts can be problematic for the implicit solver. This is especially the case when the contact is not bonded but is sliding and moving. The explicit dynamics solver has standard out-of-the-box automatic contact options (trajectory contact) which work very well. Contact will be detected in the model automatically at any point without requiring the user to define specific contact regions. On top of that the user can specify contacts manually (or generate them automatically) separate from the trajectory contact which is done similarly to the Contacts feature in the implicit solver.
Figure 3.4: Model setup showing contact (left) and boundary conditions (right)
The model Figure 3.4: Model setup showing contact (left) and boundary conditions (right) (p. 90)
demonstrates this contact issue. The implicit setup has a manually defined frictionless contact consisting of 40 contact and 38 target faces between the two parts. The explicit dynamics model simply has the default frictionless trajectory contact enabled. All other boundary conditions are the same for both analyses: a fixed support and a displacement boundary condition. Both models have the same mesh type and mesh density (the implicit setup does not make use of midside nodes in order to achieve maximum similarity in comparison, since the explicit solver cannot use midside nodes). The implicit model has problems converging while the explicit solve completes without issues. This model exemplifies
the possible problematic contact handling in the implicit solver. The model will converge when using a denser mesh; however the differences are clear for comparable mesh size (the implicit solver using midside nodes with the standard mesh size also does not converge). The final stress results can be seen in Figure 3.5: Final stress values comparison between the explicit (left, 3.4E10 Pa) and implicit (right, 3.7E10 Pa) solvers (p. 91).
Figure 3.5: Final stress values comparison between the explicit (left, 3.4E10 Pa) and implicit (right, 3.7E10 Pa) solvers
3.2.4. Rigid Body Deformations
A common analysis in the quasi-static range is the simulation of physical mechanisms. This means that rigid body movements are included in the analysis. A common simulation is a rigid or much stiffer body that snaps over a soft and flexible one. Some examples include: rubber seals for waterproofing, snapping of softer metal elements to ensure a tight fit or snapping through a notch to prevent backward move-ment. In these situations the implicit solver possibly encounters problems modelling the high deform-ations right before the snap or the release of the high deformdeform-ations after the snap. These problems are inherently unstable for the nature of the implicit solver and can be a challenge to solve successfully.
Figure 3.6: The clip model setup in the implicit solver with final deformation values (right) (p. 91) is an example of a clip snap through model where a metal clip has to pass over a rubber step.
Figure 3.6: The clip model setup in the implicit solver with final deformation values (right)
The solution does not converge unless the mesh is much coarser - this means the initial clip to rubber step contact is missed (without any special settings). Also there is a problem of missed contact between the clip and the hinge, which you can solve in the implicit solver by applying a cylindrical support. This same model with the same setup for boundary conditions and less constraints (no cylindrical support or equivalent), can be successfully solved by the explicit solver (as seen in the following figure). The setup uses mostly default settings apart from a Static Damping value which is added because of the hyperelastic material (see Damping (p. 103)). The model will run successfully without damping, but due to the nature of the materials, strong oscillations will be introduced. This means the maximum stress on the clip will spike at a much larger value than in the damped solution and then gradually converge on a similar final value when the vibrations decrease.
Figure 3.7: The clip model setup in explicit dynamics with final deformation values (right)
3.3. Setting up the Explicit Dynamics Analysis
This section will take you through the steps necessary to perform the implicit analysis using the explicit solver.
3.3.1. Attaching an Explicit Dynamics System to an Existing Static Structural System
In general, you should use the implicit analysis to set up the explicit dynamics analysis. When you identify the need the use the explicit dynamics solver, you must attach an explicit dynamics system to the existing implicit one. You do this in the same way that you would attach systems in any other Workbench project schematic operation - by drag and drop. You have four choices of what to include in the component system information transfer (see the following figure) .
Figure 3.8: Choices for information sharing between cells of implicit and explicit systems
If you drop the Explicit Dynamics system on the Engineering Data cell, only the material data would be transferred. This is not what you want to do. Dragging and dropping on the cell Geometry or Model cell should be used when you want to transfer the model from implicit to explicit. Dropping the system on the Solution cell transfers all of the end results, deformation and stress from the implicit solution so it is used only in prestressing cases.
If you drop the system on the Geometry cell, all of the implicit setup has to be recreated manually for the explicit solver. This is the better choice when dealing with very simple models with very few options
for the mesh, virtual topology, contacts and boundary conditions. This connects the two systems, but the model is launched in two separate Mechanical instances, only sharing the material and geometry data.
If you drop the system on the Model cell, the models are much more connected. They share a single Mechanical instance and the same meshing and contact options. Due to the large variation in boundary conditions, they are not transferred automatically.
Note
Changing some options for meshing, materials, or others to make the explicit analysis work might interfere with the implicit solver and make the model not solve properly. These options are discussed in the next sections. If you want to create an explicit simulation using the Model cell transfer, it is recommended that this is done in a duplicate project file.
3.3.2. Materials
There are some material models that are not available for both solvers. Whenever a question mark symbol is observed next to the Engineering Data cell, it must be properly addressed. By inspecting the materials, it should be clear where the problem is. For example, it might be a missing density value or a parameter which has not been set; something which might be required for the explicit solver but not for the implicit one. This is the case with hyperelastic materials using the Mooney-Rivlin material model.
To get a value for the incompressibility parameter the user must either have the experimental data and use curve fitting, use a value from another material specification, or just use the rubber model in the explicit material database.
Another issue you might encounter is where parameter required for the explicit simulation can interfere with the implicit solution and make it unable to solve. This often occurs since both systems share the same material data, and can be fixed by using different material assignments (if you are using the Geometry cell data transfer and have separate Mechanical instances). A problem with unsupported material model types is usually seen as an error message in the solver.log file or the Solution In-formation when a solve is attempted.
Another common example of a problem is having tabular data input for a material property in implicit with, for example, 12 stress strain pairs. This would trigger an error in the explicit solver, which only supports 10 or less stress strain pairs. An easy workaround for this would be to take the curve formed by the 12 points and delete two points and relocating the others so that the curve shape remains the same.
3.3.3. Meshing
Before running the simulation the meshing has to be thoroughly checked to ensure all requirements are met. The explicit and implicit solvers require different types of meshes. The simplest way to differ-entiate is to switch the Physics Preference option between Explicit and Mechanical. However, if the Model cell connection is used, the models are going to make use of the same mesh; this might mean that when the mesh is made to work with the explicit solver it might not solve anymore with the implicit solver. Generally, with a complex geometry we do not want to use the same mesh for both solvers.
Figure 3.9: Meshing options menu - physics preference
3.3.3.1. Uniform Mesh Works Best
The implicit solver works well when the areas of interest have much denser mesh. This is not the case
The implicit solver works well when the areas of interest have much denser mesh. This is not the case