/DT/AMS

Engine Keyword Time step control for Advanced Mass Scaling. Advanced Mass Scaling is an elementary time step method that increases the time step to a higher value than the usual elementary or nodal time step.

Format

/DT/AMS/Iflag

$Δ T_{s c a}$ $Δ T_{\min}$

If Iflag=1 or 2, insert the next line

Tol_AMS

If Iflag=2, insert the next line

Niter Nprint

Definitions

Data	Description	SI Unit Example
`Iflag`	Number of additional (optional) lines. = 0 or blank Only one card for $Δ T_{s c a}$ and $Δ T_{\min}$ is read. = 1 or 2 `Tol_AMS` is read from input. = 2 `Tol_AMS`, `Niter`, and `Nprint` are read from input.
$Δ T_{s c a}$	Scale factor on critical time step. It is recommended to set $Δ T_{s c a}$ to 0.67. Default = 0.9
$Δ T_{\min}$	Minimum time step. 2
`Tol_AMS`	Tolerance for AMS convergence. Default = 10^-3
`Niter`	Maximum number of iterations in conjugate gradient. Default = 1000
`Nprint`	Frequency (number of cycles) for writing additional output about the number of iterations before convergence in the conjugate gradient. Default = 0

Comments

To apply Advanced Mass Scaling, the Engine keyword /DT/AMS and Starter keyword /AMS must be defined.
The AMS time step method is applied on an element when the time step of an element becomes less than:
$\frac{Δ T_{\min}}{Δ T_{s c a}}$
If Iflag=0, the default values for Tol_AMS, Niter, and Nprint are used.
At each cycle, AMS solves iteratively the nodal accelerations $[M] γ = F$ using a conjugate gradient algorithm (Mass matrix is not diagonal anymore when using AMS).
Convergence of the conjugate gradient is assumed when:(1)
$‖ F - [M] γ ‖ \leq Tol_AMS \cdot ‖ F ‖$
If Iflag = 2, the maximum number of iterations in conjugate gradient Niter, and the frequency for additional output Nprint, must be entered. Otherwise, the maximum number of iterations in conjugate gradient is set to its default value (1000), and no additional output is provided.
If more than Niter iterations have been performed before convergence of the conjugate gradient, the following error message is output and the computation stops:
```
"** ERROR: AMS IS LIKELY DIVERGING"
```
If Nprint is specified, then at each Nprint cycle an additional output is provided including: the number of iterations before convergence of the conjugate gradient at this cycle, the final residual norm, and the force vector norm.

It is possible to apply Advanced Mass Scaling to a group of parts. The part group must then be specified in Radioss Starter Input deck (/AMS). For optimized computational performance, it is recommended to apply a classical mass scaling, /DT/NODA/CST to the parts not belonging to /AMS part group.

Example:

/DT/AMS/1
 $Δ T_{s c a} Δ T_{\min}$ 

Tol_AMS
/DT/NODA/CST
 $Δ T_{s c a} Δ T_{\min}$

If /DT/AMS and /DT/NODA/CST are used together, /DT/NODA/CST is only applied to parts not defined with AMS using the part group in /AMS. It is also recommended to use the same $Δ T_{\min}$ , because the simulation will always use the minimum time step. If no part group is specified on /AMS, then AMS applied to the whole model even if /DT/NODA/CST.

Advanced Mass Scaling (AMS) does not modify the global mass, so that the global momentum of the related nodes is conserved. It is, therefore, more accurate than /DT/NODA/CST.
/DT/Eltyp/Keyword3 is compatible with /DT/AMS (except /DT/INTER/CST, that only applies to the nodes where /DT/AMS does not apply.
When using /AMS with nonlinear stiffness contact interfaces TYPE7, TYPE11 or TYPE19, it may be necessary to use /DT/INTER/DEL in the Radioss Engine input deck; otherwise, AMS may converge slowly, or may even diverge.
For information about limitations related to /DT/AMS, refer to Capabilities and Limitations in Advanced Mass Scaling (AMS) in the User Guide.