General Recommendations

Every time step control method has advantages and limitations.

Good engineering judgement should be used to evaluate if the method used applies to a simulation. When in doubt, a simulation without time step control can be compared to a simulation with time step control to make sure the results are acceptable.

Structural Dynamic Simulations

Nodal Time Step and Mass Scaling

For non-uniform meshes, the nodal time step method will lead to a slightly higher time step then the elementary method. This can be activated with the default options using:

/DT/NODA
0.9 0

Most simulations can benefit from a small amount of mass being added to the nodes with the lowest time step. The following option will activate the nodal time step with mass added to meet the minimum time step entered:

/DT/NODA/CST
0.9  $Δ T_{\min}$

The process of finding a specific time step that adds a reasonable amount of mass can be done by running a model a short amount a time without any mass scaling to find the time step of the simulation. This is also a good time to see how long Radioss estimates the simulation will take to run. Next, increase the entered minimum time step and start running the model again to note the amount of mass added. Continue this process until a reasonable amount is added at the beginning of the simulation. Depending on the simulation, more mass may be added later, which would require another modification of the entered minimum time step.

Good engineering judgement must be used to determine how much mass is an acceptable amount to be added to a model to reduce the runtime. Adding too much mass can affect the physics of a drop or impact simulation. This is because the object being simulated weighs more than the real part. In general, it is recommended to keep the amount of mass added to be less than 5%, but more may be acceptable depending on a particular simulation.

Contact Interface Controlling the Simulation

When a contact has the minimum time step in a simulation, the Engine output lists INTER and the contact interface number that has the minimum time step. This indicates the time step was reduced to prevent a secondary node from passing through a main segment in one cycle. A contact can occasionally have the minimum time step during a simulation.

However, if the contact controls the time step for an extended amount of time, then the following possible causes should be investigated.

Initial intersections in the model
Incorrect penetrations in the model
Incorrect material definition, which leads to too soft of a contact stiffness
Small contact thickness or gap

Another possible solution is to use /DT/INTER/DEL to remove from the contact the secondary nodes that are causing the time step to drop. This is typically done by setting /DT/INTER/DEL $Δ T_{\min}$ to 10 to 100 times less than the $Δ T_{\min}$ used in /DT/NODA/CST.

Switch to Small Strain Formulation

If the default large strain formulation is used, large solid element deformation can give a cause a decrease in the time step. To prevent this from happening, the /DT/BRICK/CST can be used to switch to a small strain formulation at the entered minimum time step for elements whose property uses I_smstr= 2 or 12.

Example: /DT Engine Input

For a typical simulation, using a combination of the following time step control commands can help maintain a reasonable time step in the simulation.

/DT/NODA/CST
0.9  $Δ T_{\min 1}$ 
      
/DT/BRICK/CST
0.9  $Δ T_{m i n 2}$ 
/DT/INTER/DEL
0.9  $Δ T_{m i n 3}$

With $Δ T_{m i n 2}$ being two to four times smaller than $Δ T_{\min 1}$ and $Δ T_{m i n 3}$ being ten to a hundred times smaller than $Δ T_{\min 1}$ .

Quasi-static Simulations

For quasi-static simulations either traditional nodal mass scaling, /DT/NODA/CST, or Advanced Mass Scaling, /DT/AMS could be used. If the event is slow, then using larger amounts of nodal mass scaling often will not affect the results. Or using /DT/AMS per recommendations works well.

¹ Courant, Richard, Kurt Friedrichs, and Hans Lewy. "On the partial difference equations of mathematical physics." Mathematische Annalen 100 (1928): 32-74