Issue:
The time to run the Explicit Dynamics analysis (in Inventor Nastran) or Event Simulation (in Fusion 360) is long or predicted to be long.Solution:
The predicted runtime = (time to solve 1 time step)*(Event Duration)/(time step size). To reduce the runtime, one or more of those three factors must be adjusted.Time to solve 1 time step
The time to solve 1 time step is related to the number of nodes and elements in the analysis. Reduce the number of nodes and elements to reduce the runtime.- Increase the mesh size. Use "Mesh > Mesh Controls" (Inventor Nastran) or "Local Mesh Control" (Fusion 360) to create smaller elements only in areas where smaller elements are necessary.
- Use symmetry instead of the full model if the model and results are symmetric. (Vibration conditions do not produce symmetric results since some mode shapes are not symmetric.)
- Reduce the number of contacts defined if possible.
Event Duration
The predicted runtime is directly proportional to the event duration. If a 0.001 duration simulation takes 1 minute to solve, then a 0.01 duration simulation will require 10 minutes to solve, and a 1 second duration simulation would take 1000 minutes (16 hours) to solve. Edit the "Dynamic Setup" in the model tree (Inventor Nastran) or "Manage > Settings > General" (Fusion 360) to change the duration.
Be sure to update the Transient Table Data for the loads to reflect the new duration.
In some analyses, the event duration can be used to your advantage. For example,
- Use a very short event duration to minimize the runtime to check if the setup is reasonable; that is, the loads, constraints, contact, and so on are defined properly. Keep in mind that the dynamic effects will be higher than reality due to the shorter duration.
- After the setup is confirmed, change the event duration to the desired value. The ratio of durations will provide a good estimate for the final runtime.
- The difference between a dynamic and static analysis is the duration and the inertial effects that occur. To approximate a static analysis, create a "quasi-static" load that starts slowly, ramps up quickly, then ends slowly. This will minimize the dynamic effects at the beginning and end of the analysis. Create a half-sine curve for the transient table data.
- Create 20 or more points in Excel, then copy and paste the data into Nastran.
- Multiplier = A*sine(2*pi*x/T-pi/2)+C where A=0.5 to get a maximum multiplier of 1, T = twice the duration (so that the max occurs at the duration), C=0.5 to begin with a value of 0, and x are values of time from 0 to the event duration.
Time Step Size
The solver determines the time step from the smallest elements with the highest natural frequency. Therefore, the following factors affect the time step size:
- The mesh size. Increase the mesh size when possible. Remove any small features that force the mesh to be small if the features are not critical to the analysis. For a model with a mostly coarse mesh and a portion of a very fine mesh, the time step size will be determined by the smallest element.
- Distorted elements. Distorted elements can cause the time step size to be reduced.
- Avoid long skinny triangular elements. These are often created when the geometry includes sliver surfaces, extremely short edges, or sharp corners.
- If using a parabolic mesh with "Project midside nodes" included, uncheck "Project midside nodes".
- Material properties.
- Some nonlinear materials can cause the time step to be reduced. Use nonlinear materials only when necessary.
- Parts that are much more rigid can be set to "Rigid" instead of artificially increasing the modulus of elasticity in the material properties. A "rigid" material eliminates all of those elements from affecting the time step size.
- Damping. Greater Rayleigh damping values α and β in the event simulation settings significantly extend the run time for the simulation. A stiffness-proportional damping factor β greater than 0.002 will likely cause the run time to be exceeded. If possible, set the stiffness-proportional damping factor β to 0 and/or reducing the mass-proportional damping factor α. See Structural damping has no effect in an Inventor Nastran explicit dynamics analysis.
Review Output
Review the output file (.out) for details about the initial time step and estimated runtime. The output file can be thought of as having two sections:- The first section is a summary of the input.
- The second section gives the progress of the analysis, showing the time, energy calculations, and estimated remaining runtime. (This output is shown “on the fly” in Inventor Nastran. Fusion 360 does not show this output until after the analysis finishes.) Here is an example of the second section:
Time Procedure Total Kinetic Internal Viscous External Total Elapsed Remaining
Inc Step Time Time Energy Energy Energy Work Energy Wall Time Wall Time
0 9.07e-09 0.00e+00 0.00e+00 0.00e+00 0.00e+00 0.00e+00 0.00e+00 0.00e+00 0 00:00:00 n/a
50 9.07e-09 4.53e-07 4.53e-07 1.56e-04 3.62e-05 2.33e-07 1.92e-04 2.23e-07 0 00:00:02 n/a
100 9.07e-09 9.07e-07 9.07e-07 3.79e-04 2.66e-04 2.70e-07 6.45e-04 2.59e-07 0 00:00:05 n/a
150 9.07e-09 1.36e-06 1.36e-06 6.05e-04 5.39e-04 2.23e-07 1.14e-03 2.13e-07 0 00:00:07 0 03:04:54
200 9.07e-09 1.81e-06 1.81e-06 8.65e-04 7.42e-04 1.99e-07 1.61e-03 1.88e-07 0 00:00:10 0 03:04:29
250 9.07e-09 2.27e-06 2.27e-06 1.17e-03 9.18e-04 1.52e-07 2.08e-03 1.42e-07 0 00:00:12 0 03:04:32
Note: Columns of data have been removed from the remaining output examples to make them readable for this article. These tables are in the “first” section of the output and are mostly shown in reverse order in this article compared to the actual output file. That is, the “Runtime Estimates” section is the last output in the first section, and the “unnamed block” is next to last, and so on.
Runtime Estimates: Total Number Nodes in Mesh .......................... 66764 Total Number Elements in Mesh ....................... 36968 Total Mass of the Mesh .............................. 1.456e-02 Minimum Mesh Characteristic Length .................. 5.913e-06 Maximum Mesh Characteristic Length .................. 9.140e-04 Results Time Interval ............................... 4.000e-05 Step Duration ....................................... 2.000e-03 Initial Courant Stability Limit ..................... 1.788e-09 Courant Stability Limit with AutoMass Scaling ....... 9.069e-09 Speedup from Automass Scaling ....................... 5.071e+00 Initial Time Step ................................... 9.069e-09 Estimated Number of Time Increments ................. 220542The above block of text includes the following values of interest:
- “Step Duration” is the total duration of the analysis entered by the user.
- “Initial Courant Stability Limit” is the smallest time step size from all parts based on the mesh size. See the section “Element Block Statistics”.
- “Initial Time Step” is the time step used by the analysis at time 0. This is determined by adding mass to individual elements to increase the Courant Stability Limit. See the section “{unnamed block}”.
- “Estimated Number of Time Increments” is the minimum number of time steps that need to be solved. This is calculated from “Step Duration” / “Initial Time Step”.
- “Estimated Wall Clock time” is how long before the analysis is completed. It is calculated by knowing how many time increments are required and having an estimate of how long it takes to solve each time step. The actual time could be shorter (each time step takes less time to solve than expected) or longer (the time step size is reduced because the elements distort).
{unnamed block of text just before the Runtime Estimates}
AMS Mass AMS Ratio Num Elem
Part Increase Stablity Minimum Max/Min with AMS
Name Part Mass (%) Limit Elem Length Elem Length Increase
---- ---------- ---------- --------- ----------- ---------------------
P1 1.410e-03 9.874e-01 9.069e-09 1.032e-05 1.569e+01 75
P2 1.411e-03 1.042e+00 9.069e-09 9.641e-06 1.775e+01 64
P3 1.246e-03 -4.245e-06 7.617e-08 2.050e-04 3.843e+00 0
P4 1.877e-03 4.202e-02 9.069e-09 5.913e-06 1.546e+02 10
P5 1.879e-03 -9.400e-06 2.736e-08 7.362e-05 1.242e+01 0
P6 2.389e-03 2.038e-03 9.069e-09 1.066e-05 7.273e+01 83
P7 2.388e-03 2.003e-03 9.069e-09 1.066e-05 7.132e+01 89
P8 9.928e-04 5.487e-06 2.259e-08 1.126e-05 4.289e+01 0
P9 9.944e-04 -1.051e-05 2.389e-08 1.191e-05 4.022e+01 0
The above block of text shows how increasing the mass of some elements allows the minimum time step size to be increased. AMS means Automatic Mass Scaling.
- “AMS Mass Increase” shows how much the mass of the part was increased. The target mass increase is no more than 1%. In this example, part 2 (P2) determines the time step for the analysis since it has reached the 1% AMS increase. (Part 1 is closely behind at 0.99% AMS increase.) There is no need to increase the mass of the other parts and get a larger time step size since the time step size for the analysis is limited by part 2.
- “AMS Stablity Limit” (yes, there is a typo) shows the new minimum time step size for each part. If a few parts have a time step much smaller than the other parts, those parts would be the ones to concentrate on as far as increasing the time step size. Compare this time step size with the “Minimum Courant Stability Limits” in the next section.
- “Ratio Max/Min Elem Length” shows the result of dividing the maximum element length (removed from this display of the output) by the minimum element length. When the ratio is large, such as 154.6 for part 4, it implies that there may be small feature in the part. If the feature were removed, if the smallest element were larger, if the larger time step size were not the minimum for all the parts, then removing and remeshing the model would be worthwhile. In this example, part 4 with the largest ratio is not limiting the minimum time step size.
- “Num Elem with AMS Increase” shows how many elements had a mass increase, just for curiosity.
- See the sections “Part Definitions”, “Section Definitions” and “Material Definitions” to correlate the part names to the components in the model.
Element Block Statistics: Block Courant Stability Limits Elem Characteristic Length Name Minimum Maximum Average Minimum Maximum Max/Min -------- --------- --------- --------- --------- --------- --------- P1:TET10 1.913e-09 3.002e-08 1.907e-08 1.032e-05 1.618e-04 1.569e+01 P2:TET10 1.788e-09 3.174e-08 1.963e-08 9.641e-06 1.711e-04 1.775e+01 P3:TET10 7.617e-08 2.927e-07 2.023e-07 2.050e-04 7.877e-04 3.843e+00 P4:TET10 2.197e-09 3.396e-07 1.324e-07 5.913e-06 9.140e-04 1.546e+02* P5:TET10 2.736e-08 3.397e-07 1.569e-07 7.362e-05 9.140e-04 1.242e+01 P6:TET10 3.962e-09 2.881e-07 6.589e-08 1.066e-05 7.754e-04 7.273e+01 P7:TET10 3.962e-09 2.826e-07 6.576e-08 1.066e-05 7.604e-04 7.132e+01 P8:TET10 2.259e-08 9.690e-07 2.442e-07 1.126e-05 4.831e-04 4.289e+01 P9:TET10 2.389e-08 9.607e-07 2.427e-07 1.191e-05 4.790e-04 4.022e+01 * A ratio of Max/Min characteristic element lengths of more than 2 orders of magnitude probably indicates the mesh has some very poor geometry.The above block of text includes the following:
- “Courant Stability Limits Minimum” is the original time step size for each part. The time step size is controlled by the mesh quality, mass density, and the stiffness of the material. Compare this original time step with the final time step determined after scaling the mass in the section “{unnamed block}”.
- “Elem Characteristic Length Max/Min” is the same as before. It shows the result of dividing the maximum element length by the minimum element length. When the ratio is large, such as 154.6 for part 4, it implies that there may be small feature in the part. If the feature were removed, if the smallest element were larger, if the larger time step size were not the minimum for all the parts, then removing and remeshing the model would be worthwhile. In this example, part 4 with the largest ratio is not limiting the minimum time step size.
- “Elem Characteristic Length Minimum” gives the smallest dimension in each part that sets the original time step size. Since the time step size is related to the minimum element length, increasing the minimum element length can help to increase the initial time step size for the part. (Whether that translates to a larger time step size for the analysis depends on if the part is limiting the time step size for the analysis.)
Part Definitions:
-----------------
Number Element
Part Name Section Name Elements Type
--------- ------------------------------ -------- ---------
P1 P1-3D Solid 713 TET10
P2 P2-3D Solid 648 TET10
P3 P3-3D Solid 380 TET10
P4 P4-3D Solid 1400 TET10
P5 P5-3D Solid 875 TET10
P6 P6-3D Solid 7223 TET10
P7 P7-3D Solid 7338 TET10
P8 P8-3D Solid 9138 TET10
P9 P9-3D Solid 9253 TET10 The above section gives the cross reference between the “Part name” and “Section name”. Both are used in the various sections. Unfortunately, Nastran does not care about real part names and does not have a mechanism to transfer the names from the CAD model to the analysis. For all practical purposes, the “part name” is generic.
Section Definitions:
--------------------
Number
Section Name Section Type Material Name Subsections
------------ ---------------------- --------------- -----------
P1-3D Solid 3D Solid MID1 0
P2-3D Solid 3D Solid MID2 0
P3-3D Solid 3D Solid MID3 0
P4-3D Solid 3D Solid MID4 0
P5-3D Solid 3D Solid MID5 0
P6-3D Solid 3D Solid MID6 0
P7-3D Solid 3D Solid MID7 0
P8-3D Solid 3D Solid MID8 0
P9-3D Solid 3D Solid MID9 0 The above section gives the cross reference between the “Section name” and the “Material Name”. Using the Material Definition section (see below) in combination with information about the parts (such as the total mass), it is usually possible to get a good idea of which part in the output corresponds to which part in the model.
Material Definitions:
---------------------
Material: MID1
Material Type ..................... Linear Elastic
Density ........................... 8940
Youngs Modulus .................... 1.175e+11
Poissons Ratio .................... 0.345
Shear Modulus ..................... 8.73606e+10
Bulk Modulus ...................... 1.26344e+11
Dilatational Modulus .............. 1.84584e+11
Dilatational Wave Speed ........... 4543.9
Material Deletion Enabled ......... FALSE
Isotropic Thermal Expansion Properties:
Isotropic Coefficient Thermal Expansion ... 1.67e-05
Material: MID2
Material Type ..................... Linear Elastic
Density ........................... 8940
Youngs Modulus .................... 1.175e+11
Poissons Ratio .................... 0.345
Shear Modulus ..................... 8.73606e+10
Bulk Modulus ...................... 1.26344e+11
Dilatational Modulus .............. 1.84584e+11
Dilatational Wave Speed ........... 4543.9
Material Deletion Enabled ......... FALSE
Isotropic Thermal Expansion Properties:
Isotropic Coefficient Thermal Expansion ... 1.67e-05