Results Checking

How is the Energy Error in the output file (Runname_0001.out) computed?

(1)

% E r r o r = 100 (\frac{E_{k} + E_{k}^{r} + E_{i}}{E_{k, 1} + E_{_{k, 1}}^{r} + E_{i, 1} + E_{w k} - E_{w k, 1}} - 1)

Where,

$E_{k}$: Translational kinetic energy at current time
$E_{k}^{r}$: Rotational kinetic energy at current time
$E_{i}$: Internal energy
$E_{w k}$: Work of external forces (energy brought to the system)
$E_{, 1}$: Energy at beginning of the RUN (not at time t=0)

The Hourglass energy is not counted in this energy balance, so that a negative energy error generally occurs (except if using I_shell=24 (QEPH) or I_shell=12 (BATOZ) shells, and fully integrated solids, for which there is no Hourglass).

$% E r r o r$ is bounded to ± 99%.

The energy error is reset after each RESTART.

It is possible to stop a job and rerun it by using a control file containing /CHKPT. In this case, error and energy values will restart from their last value.

How to understand the Energy Error and what values are considered reasonable.

The Energy Error computed by Radioss is a percentage.

If the error is negative, it means that some energy has been dissipated.
In case of under integrated elements (Belytschko shells, solids with 1 integration point), the Hourglass energy can also explain a negative Energy Error since it is not counted in the energy balance. The normal amount of Hourglass energy is about -10% to -15%.
If the error is positive, there is an energy creation.
In case of using QEPH shell formulation (I_shell=24) or fully integrated elements, the Energy Error can be slightly positive since there is no Hourglass energy and the computation is much more accurate. An error of +1% or +2% is acceptable

If the positive energy error is greater than 2%, the source of this energy has to be identified. Incompatible kinematic conditions can lead to such a situation.
An increasing Energy Error that reaches ±99% can indicate the simulation has diverged. However, in certain situations high energy error can be acceptable:
- If the initial energy in the system is low, then it is possible to have large Energy Errors early in a simulation that reduce as energy is added to the system. This is because, small numerical differences in the energy causes large percentage Energy Errors.
- There are times when a large Energy Error at the end of a simulation can be caused by only one part diverging, but the rest of the model is giving correct results.
- Large contact energies relative to total energy can cause large negative Energy Errors because contact energy is not part of the Energy Error equation. If the simulation has friction and a lot of sliding contact, then the large contact energy and resulting energy error can be considered acceptable.

Is it necessary to take into account the added mass and how to check that the results are not modified too much?

The added mass can be due to Interface TYPE2, Spot_flag =1. In this case, the added mass is totally made at time t=0. It can also be due to options for constant time step (/DT/NODA/CST or /DT/INTER/CST).

In case of added mass in the model, it is necessary to check if it is not too important with respect to the total mass of the model (see the DM/M value in the last column in the Radioss Engine output file (Runname_nnnn.out)).

It is also important to post-process this added mass in order to check that it is not too large locally, since this could mean false results (for checking this, the Animations written with /ANIM/NODA/DMAS have to be visualized).

Also note that /DT/NODA/CST can lead to added inertia, which is much more difficult to appreciate from an engineering standpoint and cannot be post-processed prior to Radioss V9.0.

Added mass due to /DT/NODA/CST is often the cause of the divergence of computation.

If interfaces are badly defined (too small gaps, initial penetrations, and so on), /DT/NODA/CST cannot be the solution to increase time step. The model has to be improved. It is recommended not to use either /DT/NODA/CST or /DT/INTER/CST.

It is possible to first make a short run with added mass by using /DT/NODA/CST (one cycle is sufficient) and then to check the added mass by post-processing the Animation written at time t=0. The computation can then be continued without using /DT/NODA/CST, in order to limit the possible problems it can generate while getting a better time step.