Hi!

I am trying to develop a cfx model to calculate the sediment transport around structures(scour around bridge piers) in a river. But I'm having some difficulties to find the right approach. I have tried to make a multiphase model in a box with just sand and water, but then you have to define sand particles as a fluid? I couldnt make it work because there obviously isn't any viscousity for a solid. Any sugestions on a way to approach the problem will be most appreciated.

Roger

Hi,

Particle deposition and picking-up (there are probably high-tech names for these processes) cannot be modelled in CFX currently. You would need to consider coupling CFX with a discrete element model such as EDEM. Also there may be some basic functionality in this area in CFX12. But you will have to wait until next year for that.

Glenn Horrocks

Hi thx for the reply... hurry up Ansys!

Hi,

I recommend you look into discrete element models anyway as they will allow you much more physics. Look into coupling CFX with EDEM or some other DEM software.

Glenn Horrocks

Hi Thanks for the advice on the edem modelse. I have read a little about them, but I think it will be too computing intensive to have any practical use. Anyway I got the eulerian-eulerian approach to work, but the chalenge is still the change in state from when the particles are part of the river bed to becoming dispersed solids. What is bothering me is that I have seen something like this done in fluid, but I think it could be done in cfx aswell.

kind regards

CFX can be used for modelling sediment transport *within* the flow. It doesn't (as Glenn mentions) solve for the dynamics of every particle exactly, but suitable approximations to the transport of particles *within* the fluid can be made, and the Eulerian-Eulerian approach you mention can certainly be used to do this.

However, as you note, this is not the real problem.... what CFX (or for that matter any Navier-Stokes solver) *cannot* do is to tell you the conditions under which particles leave the river bed and enter the flow, or the conditions under which particles leave the flow and adhere to the river bed. The reason is that this is governed by much more detailed physics (such as electrostatics, capillary forces, flow forces on tiny individual particles, particle roughness/shape) that CFX just cannot resolve.

So, in order to proceed, you will need an empirical correlation which relates the rate of particle resuspension to the flow conditions and particle properties. This relation then needs to be implemented in CFX as a mass source term for the sand phase at the river bed boundary, probably using some CEL.

The big problem is it can be very hard to identify an appropriate model - really you need to conduct an experimental study based on the particular mud which forms your river bed. Whilst some correlations exist in the literature, they vary greatly and it is hard to identify the correct one.

I got to know this the hard way, as I was asked to finish off a study someone else had proposed, and I was missing this crucial piece of information.... It was very hard work to make convincing arguments in the absence of firm data.

So, finally I can't offer you much help, except to emphasise that this is not a pure CFD problem. You need extra data about the resuspension of sand to complete your model which CFD cannot provide. So your next, and most important, step is to identify a suitable model in the literature (backed up by experimental tests if possible). Once you've found a suitable model, then you can return to the CFD....

Good luck. Andy

Thx very much for the reply Andy. Your idea with the source term will surely be valuable. But at first I will try to make the model fit to a scour experiment with even distributed grain sizes whitch make it a little easier.

I have defined the riverbed as porous domain and the water phase as fluid. Did you use a similar approach? At the moment i'm a little sceptical to it...

best regards

my work was for thin layers of adhesive sediment in a metal floored tank - so i didn't need a porous medium model. my work certainly has differences from yours, but the need for extra information about particle resuspension to complete the model is similar.

i remember coming across many publications about river beds which were not appropriate for my study, but might help your work - so i hope you are able to find the necessary information more easily than me.

good luck andy

[mdallali]

Hi everyone,
I m workinf in modeling sediment transport in open channel , I work in fortran code, I solve the diffusion -convection EQ for sediment ,in passive mode , I need 2 Boundary condition for solving the eq,no flux on the surface and net flux in bottom, but I can't take the expected result(log profile for concentration), I put this condition at bottom:

-nu*(dC/dy)=D(E-Db)
nu is diffusivity
D is channel height
E Is erosion
and Db is deposition as ws*C0( settling velocity by refrence concentration)

I have problem in implementing the Computational part for Boundary condition.
I read about the refrence height , shoulD I have a initial value just for the cells befor refrence cell?

Regard

[ghorrocks]

Why are you writing your own model? Why are the built in models inadequate?

I do not understand your question. What are you asking?

[mdallali]

it s not my own model , just I want to add it as a part of a code. But I can t set the Boundary condition , how could i discreitze this Boundary condition with finte-Diff? is thi sBoundary condition is correct? I found it in articles.
what do you suggest me?

REgard

[ghorrocks]

It looks just a like a specified flux boundary. So just use a specified flux boundary condition on the volume fraction or additional variable which contains the sediment phase.

[mdallali]

I have already done it , but it does not give me the expected profile . I have a jumb from the second cell adjcent to buttom to value zero .

it s what I wrote for Bottum :nu*(c(1)-c(0))/dy=D(E-Db(1))

I chose my refrence point at cell 1 and give it initialy also a value for sediment concentration.

[ghorrocks]

I do not know the details of how to implement boundary conditions in CFX fortran because I have never had to - the boundary condition you describe does not sound like it needs fortran as the boundary conditions CFX have built in already do it. So I do not see why you would not just use the boundary condition already in CFX.

[mdallali]

Could you please tell me what is the B.c in CFX for this CASE?

[ghorrocks]

Specified flux.

[mdallali]

I tried to add settling velocity to the simulation of suspention particle in teh code.I used the fulmula suggested by Van Rijn for fine particle as :
w=10*nu/d*{[1+(.01*(s-1)*g*d^3/nu^2)]^0.5-1}

it s in order of 0.03 if the particle size used as 230 Micro meter, But this velocity deposit all teh suspension rapidly , Should i used another formulation?
Does any one knows a refeternce experimental and computational data to compare with ?

[ghorrocks]

Why are you specifying the settling velocity? The settling velocity is a natural result of the simulation when you have the particle drag, size and density correct.

[mdallali]

In convection diffusion eq. I should specify it , :
dc/dt+ d/dxi[(ui -ws)*C -nu*dC/dxi]=0

[ghorrocks]

Oh yes, you are using a eularian approach for this - my comment was relevant to lagrangian particle tracking. Sorry for loosing the track there.

But your question is about what settling velocity equation should you use and this has little to do with CFX. You are going to have to do your own literature review to answer that question.