The "negative rho error" in case of FSI / ALE simulation with Radioss

Radioss simulations which include Fluid-Structure Interaction with ALE formulation may stop with an error message indicating a negative or zero density issue, such as the following :


MESSAGE ID :        167

 ** STOP: NEGATIVE RHO=-1.4673475693963E-07 FOR 3D-ELEMENT ID=2882209



This may be due to multiple causes and here is a list to help you resolve this issue :

1- Check boundary conditions. Are they correctly defined? Not too close from high gradients (pressure, vel etc..) ?

• Material can escape the domain through boundaries, check BCs, etc…
• Extend the ALE domains to move inlets and outlets away from zones with high gradients.

2- Check EoS parameters: is EoS well defined (physical)?

• Inconsistent EoS can lead to erratic pressure change (very high values) and then unphysical convection.
• Check initial pressure (and all other state variables) in h3d contour.

3- Check CFL condition: did you input scale factor greater than 0,5?

• If scheme is unstable, it can convect more mass than available in from one cell to the adjacent one.
• A decrease of the Scale Factor of the CFL can be considered to solve the issue (try with 0.3 for instance).

4- Check CEL interface (/INTER/TYPE18) gap and stiffness.

• The gap may be too small and stiffness too high. Start from advised values in the manual and then iterate to adjust values.

5- Check the mesh.

• Distorted or collapsed elements can affect the stability of the simulation.
• A coarse mesh in zones where the flow is accelerated or highly disturbed can produce negative density issues.

6- Release MUSCL gradient reconstruction - BETA coefficient (for multi-materials laws only) - /ALE/MUSCL.

• Over-compressive gradient limiter (activated by default) is the most accurate, but it can be unstable with a coarse mesh and the law 151 (rarely with law 51). If the mesh cannot be refined, you can reduce the compression coefficient for gradient reconstruction (BETA) from 2.0 to 1.0 (exact second order gradient reconstruction).

7- Check very low pressure in liquids - cavitation cases.

• When the flow or expansion waves may lead to a very low pressure in liquids or even cavitation, it is recommended to initialize the liquid volume with a small added volume fraction of gas at the same pressure (for instance: 1e-4 of air in 0,9999 of water) or to use the Pmin parameter to bound the minimum pressure.

8- Do not use a single precision version for FSI / ALE simulations.

• Single precision versions are not recommended for FSI / ALE simulations, except for some low speed cases.