gmane.comp.python.fipy

Re: bool source term

Jonathan Guyer — 2008-08-20T20:24:33

On Aug 20, 2008, at 4:10 PM, franck kalala wrote: I don't understand what you're asking. Those have nothing to do with fipy. It's just Python and NumPy syntax. '&' is correct. 'and' is not.

Re: bool source term

franck kalala — 2008-08-20T20:10:15

Thanks that right I have not subversion how can I deal with "&" or "and" in fipy? Regards 2008/8/20 Jonathan Guyer public.gmane.org>

Re: bool source term

Jonathan Guyer — 2008-08-20T19:42:29

On Aug 20, 2008, at 2:57 PM, franck kalala wrote: In an equation for phi? PowerLawConvectionTerm(coeff=2 * p.getFaceGrad()) You've also got a p.getFaceGrad().getDivergence() term that isn't in your .pdf and the sign of your ImlicitDiffusionTerm is wrong in eqp.

Re: bool source term

franck kalala — 2008-08-20T18:57:55

Thanks for this hint. Did you not forgot a term in the second equation? eqp=ImplicitDiffusionTerm(coeff=1)==_ExplicitSourceTerm(q1s+q2s) eqphi=TransientTerm(coeff=2)==_ExplicitSourceTerm(q1s-q2s) after subtracting the two equations of the problem I sent in the attached file, there a term which looks like 2\nabla(phi\nabla p), then how to represent it? Franck Regards 2008/8/20 Jonathan Guyer public.gmane.org>

Re: bool source term

Jonathan Guyer — 2008-08-20T17:06:13

On Aug 20, 2008, at 12:10 PM, franck kalala wrote: I understood what you're trying to do, but this appears to be a bug in version 1.2. What you've written (using '&') works on trunk/, but not on version 1.2. As a workaround, you can do from fipy.terms.explicitSourceTerm import _ExplicitSourceTerm eqp=ImplicitDiffusionTerm(coeff=1)==_ExplicitSourceTerm(q1s+q2s) eqphi=TransientTerm(coeff=2)==_ExplicitSourceTerm(q1s-q2s)

Re: bool source term

franck kalala — 2008-08-20T16:10:54

Re: bool source term

Jonathan Guyer — 2008-08-20T13:08:39

On Aug 20, 2008, at 6:19 AM, franck kalala wrote: "&" not "and"

bool source term

franck kalala — 2008-08-20T10:32:51

Thanks to show me how to use characteristic function as a source. But I still have a problem. I am solving these equations 1)eqp=ImplicitDiffusionTerm(coeff=1)==source1, where source1=q1s+q2s(pressure equation) 2)eqphi=TransientTerm(coeff=2)==source2 where source2=q1s-q2s(saturation eqaution) where: qin=40*((phi>=.1)and(phi<=.3))+20*((phi>=.7)and(phi<=.9)) qout=60*((phi>=.4)and(phi<=.6)) sin=.8 q1s=sin*qin+s*qout q2s=(1-sin)*qin+(1-phi)*qout when I run the I program I have the following error: Evaluating diphabis.py Traceback (most recent call last): File "", line 1, in AttributeError: 'bool' object has no attribute 'solve' I don't understand this error, could any one have an idea? see my code below. Regards, Franck ********************************************************************************************************************************************************************************************* nx=400 dx=L/nx timeStepDuration=0.012 from fipy.meshes.grid1D import Grid1D mesh

bool source term

franck kalala — 2008-08-20T10:19:03

Thanks to show me how to use characteristic function as a source. But I still have a problem. I am solving these equations 1)eqp=ImplicitDiffusionTerm(coeff=1)==source1, where source1=q1s+q2s 2)eqphi=TransientTerm(coeff=2)==source2 where source2=q1s-q2s where: qin=40*((phi>=.1)and(phi<=.3))+20*((phi>=.7)and(phi<=.9)) qout=60*((phi>=.4)and(phi<=.6)) sin=.8 q1s=sin*qin+s*qout q2s=(1-sin)*qin+(1-phi)*qout when I run the I program I have the following error: Evaluating diphabis.py Traceback (most recent call last): File "", line 1, in AttributeError: 'bool' object has no attribute 'solve' I don't understand this error, could any one have an idea? see my code below. Regards, Franck ********************************************************************************************************************************************************************************************* nx=400 dx=L/nx timeStepDuration=0.012 from fipy.meshes.grid1D import Grid1D mesh=Grid1D(dx=dx, nx=nx) from fipy.variable

Re: matrix diffusive coefficient and characteristic source function

Jonathan Guyer — 2008-08-18T19:14:21

On Aug 18, 2008, at 2:57 PM, franck kalala wrote: (phi >= 0.7) & (phi <= 0.9)

Re: matrix diffusive coefficient and characteristic source function

franck kalala — 2008-08-18T18:57:57

Thanks for this answer 1) \nabla(\nabla phi)=source=q(phi) where source=q(phi)=charactristic function of the intervall [0.3,.7] how do I set a characteristic function in fipy ie my given source. Define the variable relationship and just add it as a source to the equation with "+" or "-". my source term for the above problem looks like source=q(phi)=40*chi[0.1,.03](phi)+20*chi[0.7,0.9](phi) where chi is the characteristic function of [0.1,0.3] which is 1 if phi belongs to [0.1,0.3] and 0 otherwise. how do I set such a function in fipy? Regards 2008/8/18 Daniel Wheeler public.gmane.org>

Re: Plotting flux

Jonathan Guyer — 2008-08-18T18:58:00

On Aug 18, 2008, at 10:16 AM, Rashmi Ranjan Mohanty wrote: You're right. Our vector variables don't support `getCellVolumeAverage()` (actually, on trunk/ the call doesn't fail, but the answer is meaningless). It looks like you can calculate the average vector components like this, though: >>> xave = vec.dot((1,0)).getCellVolumeAverage() >>> yave = vec.dot((0,1)).getCellVolumeAverage() You can the average magnitude (which is not the magnitude of the average vector) with: >>> magave = vec.getMag().getCellVolumeAverage() You'll want to normalize the values to the masked area, like we discussed in http://thread.gmane.org/gmane.comp.python.fipy/854

Re: matrix diffusive coefficient and characteristic source function

Jonathan Guyer — 2008-08-18T18:43:30

On Aug 18, 2008, at 2:24 PM, franck kalala wrote: Our svn installation instructions are at: http://www.ctcms.nist.gov/fipy/svn.html We are not allowed to run Vista here, so we can't offer any specific help with that platform. Perhaps some of our other users have experience?

Re: matrix diffusive coefficient and characteristic source function

franck kalala — 2008-08-18T18:24:56

Many thanks for your message ans explanation. However I still have one question. I had many problems to fix my fipy on vista and honestly I don't know exactly how to update the trunk you are talking about. May you give me simple steps on how to update easly? Franck, Regards 2008/8/18 Daniel Wheeler public.gmane.org>

Re: matrix diffusive coefficient and characteristic source function

Daniel Wheeler — 2008-08-18T15:36:38

On Mon, Aug 18, 2008 at 8:56 AM, franck kalala public.gmane.org> wrote: As I said, you will need to update to trunk and place the matrix as the first element and only element in a higher order iterable. Define the variable relationship and just add it as a source to the equation with "+" or "-". Cheers

Re: the contaminant transport problem

Daniel Wheeler — 2008-08-18T15:22:33

On Wed, Aug 13, 2008 at 8:08 PM, franck kalala public.gmane.org> wrote: Which version of fipy are you using? You should be aware of two things. 1) If you write the coefficient as "((0,1),(2,3))". This will be interpreted as a fourth order diffusion term. The first index refers to the order of the term. 2) Released versions of fipy (1.2 and earlier) do not have anisotropic diffusion (the diffusion term allows one to submit a matrix). You need to update to trunk in order to use this. To get what you want, you need to update to trunk (using subversion) and change your diffusion coefficient to be "(((0,1),(2,3)),)" for example. This will give you a second order diffusion term. If you miss out the final "," you will have a fourth order diffusion term with weird coefficients. Cheers

Re: Detalization of the mesh

Jonathan Guyer — 2008-08-18T13:24:28

On Aug 17, 2008, at 7:22 PM, Igor Telezhinsky wrote: Enough to get the accuracy you need. I'm not aware of any simple formula to evaluate this (Daniel might know). Some problems are very tolerant of coarse meshes. Some are not. If the critical length scales of your problem are small, then you'll need an appropriately fine mesh. In this case, one might well wonder why you're even solving this problem on a domain of L=1000. Everything interesting happens in 0 < x < 1. >>> mesh = Grid1D(nx=7) + (-3.5,) will give you a mesh with cell centers ranging from -3. to +3. One option might be to only have a fine grid at small x. You should be able to do: >>> mesh = Grid1D(nx=1000, dx=0.001) + (Grid1D(nx=999, dx=1.) + (1.,)) but this is both extremely slow (Tony Yu has provided us a patch that should fix that) and it throws an error for some reason. Instead, you can do: >>> Grid1D(dx=[0.001] * 1000 + [1.] * 999) Alternatively, to make a logarithmic grid: >>> mesh=Grid1D(dx=10**numeri

Re: Mesh Problem

Jonathan Guyer — 2008-08-18T13:10:37

On Aug 15, 2008, at 3:17 PM, Tony S Yu wrote: Thanks very much for offering the solution, Tony. Daniel and I were both out last week, so it's great when FiPy users can help each other out. The array() shouldn't be necessary. `[[0.],[2.]]` or `((0.,),(2.,))` should do it. We'll take a look at it. We presently don't do any work to determine what the offset "means". In FiPy 1.2 and earlier, the coordinates were specified as `(0,2)` the way Zhiwen was doing. One consequence of the array re-ordering we did since then was that the "correct" shape of the offset is now `((0,),(2,))`. I agree, though, that it's clumsy and that supporting its transpose doesn't seem likely to be ambiguous.

matrix diffusive coefficient and characteristic source function

franck kalala — 2008-08-18T12:56:42

Dear all, How can solve this problem 1) \nabla(S.\nabla phi)=source where S is a 2 by 2 matrice? 1) \nabla(\nabla phi)=source=q(phi) where source=q(phi)=charactristic function of the intervall [0.3,.7] how do I set a characteristic function in fipy ie my given source. Franck Regards.

Detalization of the mesh

Igor Telezhinsky — 2008-08-17T23:22:03

Dear all, Thank you for previous answers! Now I have a few tricky questions: Lets say I have a mesh mesh = Grid1D(dx = dx, nx = nsteps) x = mesh.getCellCenters()[...,0] x is now a space variable of my problem. How many steps in space is it essential to use for the convection problem? (see PDE solution problem thread for details). In what way the number of steps in space changes the solution precision? I would like to have small steps in space at least near 0 (i.e. 0, 0.001, 0.002 ... ). This is needed for my further research. I experimented a bit and found that fipy slows down dramatically when number of steps in space increases (as N^2)! That's why I thought of logarithmic scale so that all space dependent functions would now depend on 10^x. I met a problem, however: ARE THERE NEGATIVE NUMBERS ON THE MESH? For example, I would like to have a range of x (-3,3) what would respond to the range (0.001,1000) with 1000000 steps if had to do this detalization in linear scale

Re: Mesh Problem

Zhiwen Liang — 2008-08-15T20:16:07

Hi Tony, Thank you very much! Your solution works for me. You are a great help! Regards, Zhiwen On Fri, Aug 15, 2008 at 3:17 PM, Tony S Yu public.gmane.org> wrote: