open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

we are modelling reservoir flushing using Sisyphe coupled internally with Telemac.
The simulation is running well by using
- one grain size and non erodable bed or
- multiple grain sizes without non erodable bed
but
when we combine multiple grain sizes and non erodable bed, then we get the following error message:

ITERATION 0 TIME: 0.0000 S

SISYPHE COUPLED WITH: TELEMAC2D
THE LIQUID BOUNDARIES FILE CONTAINS
6 LINES WITH:
Q(1)
ERROR ON FRACTIONS
LAYER 1 CLASS 1 POINT 23225
AVAIL-1= 2.41280481479977382E-004
ZFCL= 8.88862241943416836E-007
EVOL= 8.88862241943416836E-007 ELAY= 8.88647832653077785E-007
PROBLEM IN LAYER: J,TEST1 23225 1.0002412804814800
IN LAYER 1 SUM OF FRACTIONS NOT 1
STOP AFTER AN ERROR IN LAYER

In this context we don't understand the role of the fortran files init_avail.f and init_compo.f., because we have a uniform bed (no spatial variation) and only one inactive layer. Do we have to change something despite this simple configuration?

Two other question:
What are the differences between the keywords:
SEDIMENT DIAMETERS and MEAN DIAMETER OF THE SEDIMENT and MEAN GRAIN DIAMETER?

We want to feed the model at the upstream liquid boundary with bed load. How can we do that?

Many thanks,

Gabi & Clemens

I have not been able to find a solution for your first question, but I am still looking. Generally, note that a new version of SISYPHE is soon to be avalaible, for which a good cleaning of keywords was made. It highly probable that your problem might not be one once you get the new version.

Regarding your second question, both SEDIMENT DIAMETERS and MEAN DIAMETER OF THE SEDIMENT have the same MNEMO in the dictionnary and would therefore be the same (again, the new version has been cleaned) - often, we keep old keywords we remain backward compatible for a few years. MEAN GRAIN DIAMETER does not exist as a keyword anymore in v6p1.

Regarding your third question, the value of the bedload flux would need to be specified at the upstream boundary actually in the code (I think - I don't believe there's a way to do it as there is an easy way to do it in the cas file). The place to look, is in bedload_soliddischarge.f Section IV BOUNDARY NODES WITH IMPOSED FLOW ... unless someone else has a better way ...

Hope this helps,

Sébastien.

Hello Sébastien,
thank you very much for the answer!

Regarding the bed load input at the boundary, the subroutine conlit.f can be used. Sorry for the question, it is actually described in the user manual.

According to the suggestion in the manual, p. 31:
"In order to avoid numerical problems such as negative sediment fractions, it is advised to use finite
elements combined with the positive depth algorithm – as explained in § 4.5 for the treatment of rigid
bed - when dealing with graded sediment and a thin-sized active layer."

we tried the finite element method and treatment of erodable bed = 3 in the mean time and now we have stable simulations. However, option 3 for the treatment of erodable bed is not described in the manual anywhere. Is it generally considered obsolete?
And is it recommended to use generally the finite element methods when dealing with non-erodable bed and graded sediments?

Another problem appeared: when we use a zonal distribution in the init_compo.f, that means, one grain size in every defined zone, Sisyphe computes the bed evolution for every size class only in the correspondig zone which is not phyiscal in our case. At least some fractions should be transported downstream.

Please find attached our test case.
Maybe someone has any suggestions?

Best regards,
Gabi & Clemens

Hello,

Here is a first partial answer:

Now with finite elements, the treatment of non erodable beds is naturally ensured by the monotonicity of the algorithm solving the Exner equation. The sediment layer above non erodable bed will always remain positive. Hence key-word "treatment of erodable bed" is obsolete in this case as no specific treatment is done.

With berst regards,

Jean-Michel Hervouet

Hi,

thank you for the answer.

Unfortunately the setup with the finite element method has not solved our problem. After about 72,000 – 100,000 time steps the same `ERROR ON FRACTIONS´ occurs.
By adding grain feeding by using the subroutine conlit.f the problem gets worse. If we feed more than one grain size, the simulation stops after the first iteration with an ´ERROR ON FRACTIONS` in a lot of nodes (a sequence is listed here):
LAYER 1 CLASS 7 POINT 23103 comment: first node after inflow boundary
AVAIL= -1.53861939898187309E+303
ZFCL= 0.0000000000000000
EVOL= 1.53861939898187311E+307 ELAY= 10000.000000000000
ERROR ON FRACTIONS
ERROR ON FRACTIONS
LAYER 1 CLASS 2 POINT 23381 comment: non erodable bed
AVAIL-1= 1.18199465981066410E+303
ZFCL= 1.18199465981066417E+307
EVOL= 1.18199465981066417E+307 ELAY= 10000.000000000000
ERROR ON FRACTIONS
LAYER 1 CLASS 1 POINT 23436 comment: near to outflow boundary
AVAIL= -1.78744957770961721E+302
ZFCL= 0.0000000000000000
EVOL= 1.90442784105100088E+306 ELAY= 10000.000000000000
ERROR ON FRACTIONS

The nodes with the error on fractions are located near the inflow (second node), near the outflow and on the non erodable bed. I do not understand why this error occurs after the first iteration.
If there is just one grain size added by the conlit.f the simulation is starting and running until approx. 72.000 time steps. Then our old ERROR ON FRACTIONS stops the simulation.

What can I try to change? Maybe I make some big mistakes....

I have uploaded our princi-file with the conlit-subroutine. The other project files are in our last post (Clemens & Gabi)

Many thanks,
Gabriele

Hello,

In case it would bring an improvement, I enclose layer.f of version 6.1, which is compatible with version 6.0 and has been improved. I hope this will improve the case that shows problem only after 72000-100000 time steps.

I do not known about grain feeding, but I am not sure that the information on fractions is then compatible, it should probably be updated, which is probably tricky to implement. I'll ask my other colleagues in charge of Sisyphe.

With best regards,

Jean-Michel Hervouet

Hi,

thank you very much for the very quick answer.

The new layer.f subroutine has solver our `ERROR ON FRACTIONS´problem. In case of feeding just one grain size it is still working, but by feeding more than one grain size I get the error message 'ERROR IN LAYER CASE 5'.

The sediment transport seems to be calculated well in the first time steps (aprox. 100,000), but in longer runs the sediment seems to be “catched” at the outflow boundary. We derive sediment depositions up to 3m at the outflow (in our case the very short before the weir), which is not physically correct.

Do I have forgotten to add something? We used a S-D-C to define the water level at the outflow.

Many thanks
Gabi

is this problem solved?
now I'm on struggle with it.

These are my condition.
1. multiple grain size into 'initial fraction for particular size class'
2. NOER.f is used.
3. NOER level is 2m only near inlet, and the level of others is 10m
(I select this value because -100m evolution is appeared when I compute sysiphe without NOER.f)

My result is..
1. when i use just 10m for all, I can get right result. that means NOER.f delete negative value under 10m
2. but when i choice NOER level as 2m near inlet and 10m for others,
'ERROR ON FRACTIONS' is appeared.

My question is..
1. Should I put same value for NOER level?
2. Is there any other solution for this problem without NOER.f?

Could anyone help me?
Please give me hint to solve this problem.
I'm crying...............
Thank you.