boundary effects at sediment transport

sebastiana · 2016-07-05 15:40:23

Dear Developers

I am trying to model the river Saltina in 1D and 2D during a historic event of duration 140 hours and I have some very unsatisfying results concerning the sediment transport.

- in 1D: No sediment got transported at all when using my initial geometry. I then tried to change this by adding an additional spillway chute above the former upstream boundary to ensure sufficient flow velocity and sediment transport capacity into my river. Now I get some sediment transport between my first and second cross section, but not any further - not even until the end of the spillway, which would be the 3rd cross section. At the same time, the bed elevation difference diagram does not show any accumulation, nowhere. So basically my sediments seem to just get lost! I am using the HSR output script by the way.
- in 2D: I can model some erosion and depostion but the majority of sediments gets deposited in one single element, which is situated at the inflow boundary, causing the mesh geometry to change very drastically. I use paraview for illustration and I get one element that builds a "tower" so high that it vanishes out of the screen. (Unfortunately it seems I can not add the image for illustration in this forum, but I would be happy to send it to the person who is answering this post).

In both models I use an inflow hydrograph and an inflow sediment hydrograph as upstream boundary conditions. I use backwater as my initial condition for the 1D model and dry for the 2D model. I have a pre-defined experimentally derived grain size distribution which I broke down to classes. Generally, the simulations seem to perform slightly better if I do not include the classes with large diameters in the mixture (though this is not extremely satisfiying, because they do exist in the real river...) but I did not yet manage to perform one simulation that shows good outcomes with respect to sediment transport. Other outputs, like velocities, froude numbers and shear stress, seem to be within reasonable range.

I would be happy for any inputs on what to change to make the models perform better.

Thank you.

Matteo Facchini · 2016-07-05 16:35:45

Hi sebastiana,
a couple of thoughts:

a) you could try to do a simple calculation with the single grain meyer-peter and mueller formula for each of your grain classes to see if the discharge you are using is able to move all of them.
b) try to start from the restart files of purely hydraulic simulations

It is very difficult for us to help you if you do not provide us at least your setup (which you can copy paste here, or load it to a cloud and write us the link). Plus, it would be useful for us to know a little bit more about how many (and which) grain classes you are using and what does your geometry look like.

Thanks

Matteo

sebastiana · 2016-07-06 10:09:09

Hi Matteo

Please find the files here: https://polybox.ethz.ch/index.php/s/sr6ecrZuVr8kAcR

I did try some of the hint (a) already, I broke down the grain size distribution to 50% / 50% for the two lowest classes (which in reality make 40% of the sediments only) for the 2D model. I can pursue this analysis further. However, the grain size distribution was measured and the hydrograph did actually happen, I am modelling an event that existed (within the uncertainties of measurement of course), so it sort of cant be that it cannot happen at all... ;-)
For (b), I usually cause basement to collapse when I go for the "continue" option but I will retry.

I would be glad if you can look at the geometry and files.

Thank you,
sebastiana

Matteo Facchini · 2016-07-06 11:01:23

Some comments on your 1D setup:

1. under HYDRAULICS/PARAMETER try to use the roe riemann_solver instead of the upwind one
2. the discretization of your grain size distribution curve seems a little bit strange, try to follow what parker suggests here [1] in chapter 2
3. I get the following error:"Block-Error: 'diameters' in GRAIN_CLASS block has more than one grain class defined, which is not compatible with the single grain bedload formula 'smartjaeggi'. Please change bedload formula or grain classes!" therefore I suggest you to use smartjaeggi_multi, mpm_multi, wu or parker formula.
4. consider to lower a little bit your CFL, like to 0.8 or 0.9.
5. your bmc file uploads a bmg file (gemoetry) which is not included in the polybox folder you shared with me.

comments on 2D setup:

1. your sedigraph is quite steep, this might cause some problems
2. the discretization of your grain size distribution curve seems a little bit strange, look at comment nr. 2 in 1D comments
3. consider to lower a little bit your CFL, like to 0.8 or 0.9.

A general comment on your post: remember that you are modelling an event, not reproducing it with 100% of reliability. Try to start for example from a simplified case (e.g. 1 grain class, i.e. d50 of your distribution and a simplified hydrograph) and see if the behavior remains strange.

Bests

[1] http://hydrolab.illinois.edu/people/par … ctures.htm

sebastiana · 2016-07-07 12:11:46

Thanks again Matteo!
I uploaded the correct 1D geometry to the polybox folder as well, sorry for that. The topo_schussrinne geometry is with the mentioned spillway chute, converted2 is without - the latter would be my final goal to use but if a spillway chute really is necessary thats ok for me. However, so far it did not show any positive effects.
I checked the grain size distribution, I did as Parker. I have a stream from a very wild montainous catchment so I guess that caused your questioning. Also, I suspect that causes my problems in calculation...
I tried today a computational run 2D with lower CFL and it did not help.
Best

Matteo Facchini · 2016-07-11 10:57:01

Have you tried to simplify the hydrograph and the sedigraph and perform a simulation with one representative grain size (e.g. d50)?
If you do that and you get results that are close to what you expect than you can proceed and add complexities to your simulations, I would not start directly with the "real" configuration. Starting from simplified configurations help to better understand if the model is behaving properly
Cheers

User Forum of Software BASEMENT

#1 2016-07-05 15:40:23

boundary effects at sediment transport

#2 2016-07-05 16:35:45

Re: boundary effects at sediment transport

#3 2016-07-06 10:09:09

Re: boundary effects at sediment transport

#4 2016-07-06 11:01:23

Re: boundary effects at sediment transport

#5 2016-07-07 12:11:46

Re: boundary effects at sediment transport

#6 2016-07-11 10:57:01

Re: boundary effects at sediment transport

Board footer