Creating dam break case with water and sediment phase

[ChrisCivEng]

I want to create a two-phase model with water and sediment. The sediment-phase will be simulated as a NN-fluid and should represent sand. Therefore I need to set a value for the kinematic viscosity and the specific yield stress. How can I derive the right values for the kinematic viscosity and the specific yield stress for the sediment phase?

So far I modified the case example 01_WetDambreak of the mphase_nnewtonian folder. Now I need to validate the simulation parameters of the sand layer.

`

    <special>
        <nnphases>%Defines non-newtonian phases parameters
            <phase mkfluid="0">
                <rhop value="1000" comment="Density of the phase" />
                <visco value="0.000001" comment="Kinematic viscosity (or consistency index) value for phase (m2/s)" />
                <tau_yield value="0" comment="Specific yield stress of phase (Pa m3/kg)" />
                <HBP_m value="0" comment="Use 0 to reduce Newtonian liquid, order of 10 for power law and order of 100 for Bingham (sec) " />
                <HBP_n value="1" comment="Use 1 to reduce to Newtonian, <1 for shear thinning >1 for shear thickenning " />
                <phasetype value="0" comment="Non-Newtonian=0 only option in v5.0" />
            </phase>
            <phase mkfluid="1">
                <rhop value="2630" comment="Density of the phase" />
                <visco value="0.057034" comment="Kinematic viscosity (or consistency index) value for phase (m2/s)" />
                <tau_yield value="0.038023" comment="Specific yield stress of phase (Pa m3/kg)" />
                <HBP_m value="100" comment="Use 0 to reduce Newtonian liquid, order of 10 for power law and order of 100 for Bingham (sec) " />
                <HBP_n value="1.8" comment="Use 1 to reduce to Newtonian, HBP_n<1 for shear thinning HBP_n>1 for shear thickenning " />
                <phasetype value="0" comment="Non-Newtonian=0 only option in v5.0" />
            </phase>
        </nnphases>
    </special>
     [...]

`

Thanks for your help!

[alexbexe]

@gfourtakas can help here

[MaryamBC]

Hello dear user,

I have the same question and unfortunately I have not received response from developer. I am wondering if you could find the answer?
Best regards,
Maryam

[gfourtakas]

Hi,

Firstly let me be clear that here you will not be simulating "sand" as the NN model will only simulate HBP fluids.
All parameters related to the NN fluids are "specific" values, i.e. you need to divide by "rho" - I see you are using kinematic viscosity already but I am not sure about your Tau_yield.

Please let me know if you need anything else.

George

[ChrisCivEng]

Hi George,

Thank you for your reply!

I divided both the dynamic viscosity and Tau_yield by the density "rho" to get the kinematic viscosity and the specific yield stress respectively. My initial values for dynamic viscosity and Tau_yield are 150 Pa and 100 Pa respectively (see screenshot). I took these values from your paper "Modelling multi-phase liquid-sediment scour and resuspension induced by rapid flows using Smoothed Particle Hydrodynamics (SPH) accelerated with a Graphics Processing Unit (GPU)" (2016) here: https://www.sciencedirect.com/science/article/pii/S0309170816300926?via%3Dihub

I attended your SPH lecture in Parma in November 2024 where it was mentioned that sediments could be simulated as non-newtonian fluids with shear-thickenning characteristic. This is why I thought, that I could "approach" the behaviour of "sand" as a NN fluid.

Should I use an older version of DualSPHysics? In version 4.4.12 under examples\twophases\DSPH_v3.4_LiquidSediment I found a case for a dambreak with water and sediment. In this case are the values viscosity and yield strenght also "specific" values that are divided by "rho"?

How would you advise me to proceed?
Best regards
Chris

[gfourtakas]

Hi,

Yes you can use the NN solver to model such flows, I am just pointing out that this is not a soil specific model such as u-p-w formulation from Biot, etc.

I do not remember what we used in the journal paper, i.e. if the values are /rho or not to be honest but on the v5.0 version they need to be normalised.

Did you run the simulation or is it now working? Trying to understand what the problem is!

BW,
George

[gfourtakas]

PS: the units are on the xml -m2/s and Pa m3/kg :)

[ChrisCivEng]

Hi,

great to hear that I can use the solver of version 5. I will make a note in my Master Thesis that the NN solver is an approach to model a sediment-like phase. Firstly I wanted to confirm, that the model values and units are correctly understood.

The simulation is working. I used Paraview for visualization. Now I want to validate the model with experimental data. In the following there is a figure of a physical dambreak experiment with sand and water at and my simulation result both at time step 1 sec. between -1 m and 3 m. The initial water height was 0.35 m. (Note that the vertical scale of the physical experiment is stretched by factor 1.5):

As you can see in the simulation the water is not travelling as far as in the physical experiment. Plus, the sediment phase is more unstable compared to the experiment. It seems that the water flow looses too much energy due to friction, why the flow doesn't travel as far as in the experiment. In other words: the flow velocity of the simulation is slower than the real velocity of the experiment.
I tried to change the values for viscosity and yield stress of the sediment phase but without any visible changes. The travel distance stayed the same. I don't know whether it was an issue of Paraview or the model.

That is why I changed dp. Next I want to try different values of the artificial viscoity alpha.

The simulation at time step 1 sec. better matches the pyhsical experiment at 0.75 sec. in terms of flow distance:

If you need further information, please let me know. I appreciate your help.

Best Wishes
Chris

[lijunhao0218]

CaseWetDambreak2DNN_Def 1.txt
Hi, I have just used this software for a short while and this is my xml file. May I ask why the sand of my simulation result is not quite the same as yours, is something not quite right?

[ChrisCivEng]

There are some differences in the parameter section for the simulation shown above. I set "ViscoTreatment" to value 1 for artificial viscosity. In your file it is set to value = 3 (Constitutive eq.). This could be an explanation for different simulation results. How exactly is your result different?

I conducted further analysis of the model and found out that leaving "ViscoTreatment" to value 2 (Laminar+SPS) worked better for me. In this case the sediment phase is more stable, still the water level is overestimated. But I haven't tried out value 3 (Constitutive eq.).
Other parameters that are different in my simulation:
"StepAlgorithm" value = 1 (Verlet, default)
"ShiftTFS" value= 2.75 (for 3D)
"ShiftCoef" value= -10 (value that was preset in example case)

I can't say how StepAlgorithm or ShiftCoef influence the simulation. But I think it is important to change ShiftFS to 2.75, since it looks like you are simulating in 3D as well.

[sharanyansudarshan]

Hi ChrisCivEng...

I am an absolute beginner to this tool and I too am trying to run a similar case.
Could you kindly let me know how you calculated/obtained the yield stress value for the sediment phase. Did you use any of the empirical models such as Shield's criteria or Drucker-Prager Criteria to find the value?.If so could you maybe post the calculation here if possible?

Thanks a lot
Regards

Sharanyan