yade-dev team mailing list archive

[Christian Jakob <jakob@xxxxxxxxxxxxxxxxxxx>]

first of all:

thank you very very much for your time spending on answering my question(s)!


Zitat von Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>:

> On 20/02/14 10:52, Christian Jakob wrote:
> > Hi,
> >
> > I want to do some updates/improvements of PFV method.
>
> Don't do anything yet!

oops, too late :(

> There is some code cleaning going on locally
> (different variable names, removing useless functions, and so on). It
> will mess up everything when I will commit if you have local diffs. Make
> you a favor and wait one more week, or be ready for tedious merge.
>
>
> > First thing I wanna do is a simulation with an increasing pore
> > pressure in all cells to investigate liquefaction of a loose sand at
> > different stress conditions. I tested imposePressure(), but for now it
> > is not possible to use it within a loop over PFV cells. Actually it
> > needs position of the cells (or at least a point within the cell), but
> > I want it to do something like:
> >
> > for i in range(flow.ncells()):
> >   flow.imposePressure(i,p)
>
> This loop is strange. Do I miss something?

of course you missed at least one if condition ... sorry i should have  
been more clear on this point ...

> 1/ If you impose pressure everywhere, there is no point using a fluid
> solver (which aim is to compute pressure, at the end of the day)
> 2/ If pressure is the same everywhere, it applies no force on particles
> (forces only come whith pressure gradients), again no point to simulate
> a fluid.
> 3/ In liquefaction, pressure increase comes from the hydromechanical
> coupling (and the model gives that) not from an externally applied pore
> pressure

that is main question of my study: which processes leads to increased  
pore pressure? and how much excess pressure is needed to trigger  
liquefaction ?
for this purpose i have to impose pressure at a larger part of the  
model. this is why i need a to loop over cells and use imposePressure  
heavily!

> 4/ The element number "i" is not something you can rely on, this is why
> such function does not exist. The mesh is changing all the time as
> particles move, thus element number "i" is meaningless. Applying
> pressure at a specific location in space is more appropriate.

from my point of view it should be possible to do this via "i". I  
already implemented it (not tested yet). Also I am planning to do  
similar update to other methods, that use "pos" as input (imposeFlux,  
getPorePressure, ...)

> Besides, I agree that there is a need for writing loops on all elements
> in other cases. We (you, Thomas, me, ...) can add it after the code
> cleaning step.

I will do it (see above) ...

> > I had a short look at the source code, it is quite confusing. I
> > searched code for imposePressure. It finds 6 lines in
> > pkg/dem/FlowEngine.hpp, but where is corresponding code in .cpp file?
> > I noticed there is a macro called TPL. How can edit the code of this
> > macro?
> imposePressure is defined in the hpp [1].
> The TPL macro as well [2].
>
> [1] https://github.com/yade/trunk/blob/master/pkg/dem/FlowEngine.hpp#L330
> [2] https://github.com/yade/trunk/blob/master/pkg/dem/FlowEngine.hpp#L24

yes, i noticed already ... sorry for this silly question, it was is not my day

> The template aspect makes things a bit more difficult to read, but it is
> the only way I found to not duplicate the whole PFV code each time there
> is something new. Periodic BC, drainage model, solute transport, etc.
> will normally be only different instances of FlowEngine with different
> template parameters (and maybe additional functions). It is the case for
> periodic BC already, which do not duplicate anything.

aah, i see

> I'm open to suggestions if somebody has a better idea of design. The
> root of the problem is the data type in each element/node, which is the
> template parameter of CGAL::Triangulation and is different for each new
> model.

Sorry, I am really not the guy to ask for C++ design stuff!
But I see the problem ...

> Cheers.
>
> Bruno

Again, thanks a lot!

> _______________________________________________
> Mailing list: https://launchpad.net/~yade-dev
> Post to     : yade-dev@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~yade-dev
> More help   : https://help.launchpad.net/ListHelp