SPH Fluid Damping — How to Set Artificial Viscosity and Bulk Coefficients for Water vs PDMS in Abaqus/Explicit?

[stevekoja]

Hi everyone,

I’m working on my master’s thesis and would really appreciate some help understanding how to properly define artificial viscosity and bulk damping for different fluids in Abaqus/Explicit SPH.

Model setup:
I am modeling a Wind Turbine (WT) system as a single-degree-of-freedom (SDOF) spring-dashpot system. It is coupled to a chamber filled with fluid, which is modeled using SPH (PC3D) elements. I extract the response at a reference point (RP) on the WT to evaluate how each fluid damps the motion.

Fluids modeled (using EoS – USUP):

1. Water

• Density: 1000 kg/m³
• Dynamic Viscosity: 0.001 Pa·s (1 cSt)
• Speed of sound: 1450 m/s
• S = 2.0, gamma = 0.0347

1. PDMS-50 cSt

• Density: 960 kg/m³
• Dynamic Viscosity: 0.048 Pa·s (50 cSt)
• Speed of sound: 1004 m/s
• S = 1.68, gamma = 0.21

1. PDMS-1000 cSt

• Density: 971 kg/m³
• Dynamic Viscosity: 0.971 Pa·s (1000 cSt)
• Speed of sound: 1004 m/s
• S = 1.54, gamma = 0.21

The PDMS properties are taken directly from published sources.

The issue:

I am trying to calibrate damping behavior for the three fluids, using the Abaqus SPH formulation. However, the results don’t reflect the expected behavior:
PDMS-1000 cSt should damp significantly more than water, but in many cases it damps less, or water appears overdamped.

I have adjusted both:

• Artificial viscosity (α and β) → defined via *SECTION CONTROLS
• Linear and quadratic bulk viscosity coefficients → these affect time-step damping, and are separate from α and β

I understand that real viscosity defined via *Viscosity has no effect in SPH, so the damping must be controlled through α, β, and timestep bulk viscosity.

Despite trying many combinations, I can’t achieve physically consistent damping behavior across the fluids. I’m aware that low-viscosity fluids (like water) may need more bulk damping for stability, and high-viscosity fluids (like PDMS) can become unstable if bulk damping is too high.



Question:

Could someone please recommend realistic and physically meaningful values for:

• Artificial viscosity (α, β) in *SECTION CONTROLS
• Bulk viscosity damping coefficients (linear, quadratic) for timestep control

for each of the following fluids:

• Water (1 cSt)
• PDMS-50 cSt
• PDMS-1000 cSt

This would help me a lot in finalizing the setup for my thesis. The goal is to properly compare how each fluid damps the WT motion, and right now the artificial damping behavior is not making physical sense.

Thanks very much in advance for your help!

Best regards,
Steve

 

[FEA way]

Did you check the documentation (chapter "Smoothed Particle Hydrodynamics", paragraph "Artificial Viscosity") carefully ? Artificial viscosity is, as the name suggests, not meant to model real viscosity. It serves a similar role as *BULK VISCOSITY in Abaqus/Explicit and it’s definition is also equivalent. Those two features are only meant to reduce numerical oscillations in results but they sometimes affect the results significantly (often visible in energy output) and should be reduced from the default values. If "real" viscosity (*VISCOSITY) has no visible effect in your model, maybe those artificial effects are too large or there’s another issue with the model.

If you check the documentation examples, there is one called "Splashing of a figurehead" where *VISCOSITY is used with SPH. You could do some tests on it or just simplify your model to the absolute minimum.

 

[stevekoja]

Did you check the documentation (chapter "Smoothed Particle Hydrodynamics", paragraph "Artificial Viscosity") carefully ? Artificial viscosity is, as the name suggests, not meant to model real viscosity. It serve a similar role as *BULK VISCOSITY in Abaqus/Explicit and it’s definition is also equivalent. Those two features are only meant to reduce numerical oscillations in results but they sometimes affect the results significantly (often visible in energy output) and should be reduced from yhe default values. If "real" viscosity (*VISCOSITY) has no visible effect in your model, maybe those artificial effects are too large or there’s another issue with the model.

If you check the documentation examples, there is one called "Splashing of a figurehead" where *VISCOSITY is used with SPH. You could do some tests on it or just simplify your model to the absolute minimum.

Thank you — yes, I’ve checked the documentation and I understand the purpose of artificial viscosity vs real viscosity. I also looked into the "Splashing of a Figurehead" example, but in the .inp file I found, the <AlphaFactor> and <BetaFactor> tags are there but not assigned any values. Also, the step time doesn’t seem to include any bulk viscosity settings either, so it's hard to make a proper comparison.


Would it be okay if I upload my .inp file here for review? Maybe there’s something subtle I’m missing.