Weird error message Log Out | Topics | Search
Moderators | Register | Edit Profile

FlexPDE User's Forum » User Postings » Weird error message « Previous Next »

Author Message
Top of pagePrevious messageNext messageBottom of page Link to this message

Pascal Mosnier (p_mosnier)
Junior Member
Username: p_mosnier

Post Number: 3
Registered: 06-2005
Posted on Friday, June 01, 2007 - 09:34 am:   

For a problem of thermal stresses and moisture diffusion in a composite using FlexPDE version 4, I got an error that I do not understand :

Floating-point Divide by Zero
---Called from timsolve::get_wates
---Called from timsolve::onestep
---Called from timsolve::gearstep
---Called from timsolve::evolve
---Called from timsolve::do_runjob

This does not occur always during the same timestep. Moreover I used about the same set of equation for others problems and boundaries conditions which did not gave me that message.

Any suggestions or solutions would be greatly helpful !

Thanks in advance
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 864
Registered: 06-2003
Posted on Friday, June 01, 2007 - 02:37 pm:   

The diagnostic means that you have divided by zero. This is not legal.

Somewhere in your script (which you have not included) there is a divide, the divisor of which can go to zero in some circumstance. This circumstance is occurring, and causing a divide by zero.

You should be aware that FlexPDE constructs trial solutions, some components of which may oscillate beyond what you expected to be the limit values. For example, if you divide by a variable that in some places is near zero, then you may get a divide by zero as the trial solutions oscillate to or beyond zero.
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
New member
Username: yenrab54

Post Number: 1
Registered: 07-2007
Posted on Monday, July 02, 2007 - 12:44 pm:   

I am getting a similar msg; only mine reads:

Floating-point Divide by Zero
---Called from domainspec::vedge_scale
---Called from domainspec::grid_layer
---Called from mesher3::newmesh3
---Called from mesher3::grid3d
---Called from control::do_gridgen

at this point i am merely trying to generate the geometry of my system. my goal is to, eventually, run a heat transfer simulation. my first assumption was that there were to high of a cell density, but i did some experimenting with the mesh generation commands and found the aspect ratio command to be the most effective in reducing cell density. nonetheless, the domain run gets to a certain point and either gives this error, or totally crashes flex.

i am definately attaching the file. its of the latest flex version.

Thanks for being available.

Barney
application/octet-streamProblem
BN Configuration.pde (2.0 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 888
Registered: 06-2003
Posted on Monday, July 02, 2007 - 08:44 pm:   

The diagnostic implies there is a vertical edge of zero length.

However, I am concerned that the washers on top and bottom of layer 4 do not define a body in layer 4 (see attached graphic). Check and see that these rings are included in layer 4.

There may also be trouble in layer 3 because the washer exists on the top and not on the bottom, and the layer is thin. You might need to include the washer on layer 3 as well, with the appropriate material properties.
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
New member
Username: yenrab54

Post Number: 2
Registered: 07-2007
Posted on Tuesday, July 03, 2007 - 12:44 pm:   

Ok, my 4th layer is where the bottom washer exists. The 5th layer is where the washers are on the top and bottom. I figured out the zero length vertical edge issue; I was carrying the decimal too far. I changed that, but know I have a different error which I am even more oblivious to its solution. Here is the modified version of the last descriptor. I've also changed how the layers and regions are labled to, hopefully, make more sense.

Thank you for all the help.

Barney
application/octet-stream
BN Configuration.pde (2.1 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Junior Member
Username: yenrab54

Post Number: 3
Registered: 07-2007
Posted on Friday, July 06, 2007 - 12:30 pm:   

To clarify, the new error reads:

Cell 963 Nabor 1004 Nonreciprocal!
Mesh is tangled near (6.65291, 2.98252)!

It lets me continue, but I immediately get a series of NO COMMON CELL errors.
I understand there is mesh tangling. I'm not sure which selector to use to fix it, or whether a selector is even necessary. I'm afraid the geometry is too complicated. This descriptor is present in the previous msg.

In a different simulation I am running, I am getting a different, possibly less serious, error:

Variable has unknown range.
Provide threshhold clause ()
to quantify error results.

I tried to put a threshhold clause in directly after the variable entry, but it still didn't like it.
application/octet-streamThreshhold clause error
flat resistor.pde (2.1 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 892
Registered: 06-2003
Posted on Friday, July 06, 2007 - 03:07 pm:   

The flat_resistor problem runs without complaint. Since there are no boundary conditions or sources, the solution is all zero.
Did you post the wrong file?

Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 4
Registered: 07-2007
Posted on Friday, July 06, 2007 - 04:28 pm:   

yes, i did post the wrong file, and when I went to access it, flex didn't save it. I tried to put the same changes as I did before, and this time I ran it, I got a different error:

equation is empty in region 38

I don't even have a region 38.

anyway, here it is. concerning the other simulation, do i need to define the material properties before we can resolve the "no common cell" issue?
application/octet-streamflat resistor
flat resistor.pde (3.1 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 893
Registered: 06-2003
Posted on Friday, July 06, 2007 - 04:40 pm:   

You don't need to define material properties (except that every name must have some kind of definition) to resolve the meshing problem.

I am looking at it now, and haven't yet figured out what it is complaining about.
The manifestation of the problem is that the bottom surface of PCB has a hole in it, and the top surface does not. Clearly, the top surface of PCB is confused about region membership. But I don't yet know why.

Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 894
Registered: 06-2003
Posted on Friday, July 06, 2007 - 06:41 pm:   

In Flat_Resistor:

Granted, the diagnostic is not helpful. "Regions" are tabulated internally layer by layer. "Region" 38 is region 5 layer 4. Here 5 is the sequence number, not your tag number.

However...
You have given a default value of 0 for "cond", which guarantees that if it is not redefined in some layer/region, the first equation will collapse to 0=0.

If you had simply omitted the value on the initial declaration of cond, then FLexPDE would have checked definitions for you and reported
"COND is undefined in region 5 layer 4", and put the editor cursor on the offending region 6 (the fifth declared region). See attached.

You should not use 0 as a default if you expect to always override it. Leave the value out so FlexPDE can check definitions.

I will fix the "collapsed equation" diagnostic to use the same report as the "Undefined parameter" diagnostic.

application/octet-stream
flat_resistor.pde (3.1 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 5
Registered: 07-2007
Posted on Saturday, July 07, 2007 - 06:16 pm:   

ok, that worked and i also figured out how to implement the threshhold clause. I started the simulation and left it running. There is a notice saying "step failure retry," or something like that.

How long should flex take to give results for this kind of system? I let it go for 15 min, so far. I understand this could take hours, but I have no idea.
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 895
Registered: 06-2003
Posted on Sunday, July 08, 2007 - 12:36 am:   

Send me the final script and I'll run it and see if there is anything impairing the performance.
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 6
Registered: 07-2007
Posted on Sunday, July 08, 2007 - 02:49 pm:   

Great, here it is. Thanks for all your help; I hope you aren't doing all this for fun.
application/octet-streamflat resistor
flat resistor01.pde (3.1 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 897
Registered: 06-2003
Posted on Monday, July 09, 2007 - 05:04 pm:   

You appear to have misunderstood the way FlexPDE constructs 3D figures.

The layout in the BOUNDARIES section is a map of the projection of all material divisions into the XY plane.

The REGION layout in the BOUNDARIES section follows the rule that REGIONs defined later overlay and hide any REGIONS defined earlier that they overlap.

Above each REGION in the BOUNDARIES section, there is a stack of layers formed by cutting the REGIONS with the EXTRUSION SURFACES. Each LAYER in each REGION may have a different set of material properties.

Your region 13 overlays and hides all of your region 2.
Your region 14 overlays and hides all of your region 3.
Your regions 11 and 12 overlay and hide part of your region 5.

A material property (e.g. K=25) stated in a region without LAYER qualification, applies to the entire stack of layers above the region.

So I don't think you are getting what you want with this layout. The physical configuration is probably what you expected, but the assignment of material properties is probably not.

You can examine this question by adding plots of material properties on various cut planes to see if they are what you want.

And yes, we do this for fun. And money. Fun and money.
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 7
Registered: 07-2007
Posted on Tuesday, July 10, 2007 - 12:26 pm:   

yeah...i did misunderstand the protocol. do you think the same misunderstanding fudged the BN configuration layout?
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 8
Registered: 07-2007
Posted on Thursday, July 12, 2007 - 06:32 pm:   

so, here's a new one for you:

timestep has fallen below 1e-009 of starting value!
you may have temporal discontinuity in parameters (see ...) or you may simply need a smaller BY clause or HAULT clause in the TIME section.
Run halted:
---Called from timsolve::gearstep
...::evolve
...::do_runjob
...::runphase
...::control_loop

I included the elipses b/c I failed to record the actual wording. If it's totally necessary I can run it again. The issue with that is it happened after two hours of runtime. It seems an accuracy manipulation would fix this, but I'm not exactly sure. And this is the latest descriptor.

application/octet-streamlatest
flat resistor01.pde (2.6 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 900
Registered: 06-2003
Posted on Friday, July 13, 2007 - 03:00 pm:   

1.
The way you have posed the problem, all voltages are initially at zero, then the boundary voltage jumps instantaneously from zero to 3 at time=0.
The voltage equation is time independent, so all the internal voltages jump from zero to their equilibrium value over the first timestep.

FlexPDE tries to find a timestep to start the simulation by cutting the timestep until the time behavior is polynomial over the first two steps.

But since the voltage always jumps from zero to equilbrium in one step, no matter how small the step, there is no termination to the timestep cutting operation.

Hence the "Timestep has fallen...." diagnostic.

There are several ways to address this inconsistency (see Help->Tech Notes->Avoid Discontinuities):
a) Ramp the applied voltage from zero to 3 over a realistic time.
b) start out with initial values of U that are a self-consistent solution to the potential equation. You can do this by applying values that approximate the equilibrium values, or by running an auxilliary problem first to establish the potentials, then TRANSFERring these into the temperature equation as initial values of U.

Since your conductivites do not seem to be temperature dependent, you don't really need to solve simultaneously for the potentials and temperatures. You can use the auxilliary problem to define potentials and merely import them as fixed data, solving only for temp.

2.
You have stated a threshold of 650 for the temperature. The THRESHOLD is the smallest variation you care about. So, in effect, you have said you don't care about variations smaller than +- 325. This effectively disables timestep control for the run, since your temperature variations are much smaller than this. Use a more realistic value like 1 for the threshold. When the actual variations become larger than this, the threshold will be ignored.


In the attached script I have done some of these things. It runs reasonably, except that the temperature in the substrate seems to be clamped at zero, making the distribution through the resistive layer hard to model with only a few mesh cells. If this is what is supposed to happen, you should increase the mesh density in the resistive layer.
application/octet-stream
flat_resistor02a.pde (2.8 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 9
Registered: 07-2007
Posted on Friday, July 13, 2007 - 04:21 pm:   

thanks a bunch. concerning the other simulation, i've been trying to figure out a better way of setting up the geometry, but i haven't really had any luck. i understand the hole through the PCB layer is causing the problem, but that part of the system is the important part. let me know what you think.

Barney
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 901
Registered: 06-2003
Posted on Tuesday, July 17, 2007 - 09:48 pm:   

I don't know what to think, because I don't really know what you want it to look like.
Is it really a hole? Or some other material?
Do you have a drawing of what it is supposed to be modeling?
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 10
Registered: 07-2007
Posted on Wednesday, July 18, 2007 - 11:20 am:   

no, i don't have a picture, but i can try to describe it to you. I drill a hole through a circuit board, then in the hole I place a copper cylinder. On both sides of the circuit board, I place a washer surrounding the cylinder. Lastly, I put a gold pin in the hole and through the cylinder.

I guess my real question is whether I can use a circle as a region, then another circle inside of it, then another circle inside of it, meanwhile giving them different properties after extrusion. Or, do I need to define two rings and a circle?
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 902
Registered: 06-2003
Posted on Wednesday, July 18, 2007 - 02:12 pm:   

My question was whether there was really a hole, and it sounds like there is not, because it should be filled by the gold pin.

You can build circular regions as long as you define the biggest one first, and descend in size. This way, the small ones overlay the big ones, and each can have separate material properties.
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 11
Registered: 07-2007
Posted on Wednesday, July 18, 2007 - 02:32 pm:   

thanks, I'll see what I can do.
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 12
Registered: 07-2007
Posted on Wednesday, July 18, 2007 - 03:40 pm:   

ok, I started to redo the whole geometry from the bottom up, excluding what I thought were the trouble regions(smaller circular parts. The errors begin during the meshing problem in the layer'lower copper'region 1. I tested the thickness with a few different values which all gave the same result:

Incomplete Bounding Mesh
An internal processing error has occured
Please contact PDE...
---Called from domainspec::grid_layer
... mesher3::newmesh3
...mesher3::grid3d
...control::do_gridgen
...ccontrol::runphase

Is this still confusion about region membership?

application/octet-streamBottom UP
new BN config.pde (1.3 k)
Top of pagePrevious messageNext messageBottom of page Link to this message

Robert G. Nelson (rgnelson)
Moderator
Username: rgnelson

Post Number: 904
Registered: 06-2003
Posted on Wednesday, July 18, 2007 - 07:09 pm:   

You didn't pay attention to my post of 7/9:

"The REGION layout in the BOUNDARIES section follows the rule that REGIONs defined later overlay and hide any REGIONS defined earlier that they overlap."

Your region PCB appears third, which means that in any surface in which it appears, it DELETES the outlines of the resistor and epoxy.

We always suggest that your maximum outline appear as "region 1", and that all smaller figures be laid on in order of decreasing size.

If you move your PCB layer to the head of the REGION list, it will no longer delete any of the smaller figures when it appears.


Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 13
Registered: 07-2007
Posted on Thursday, July 19, 2007 - 12:37 pm:   

Yeah, you're right; it works and makes sense, now.
Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 14
Registered: 07-2007
Posted on Thursday, July 19, 2007 - 12:56 pm:   

Now for a different, but similar configuration. This msg is preventing a full domain review:

Memory Protection Fault
---Called from domain::merge_edges
...domain::domain_analysis2
...domainspec::domain_surface
...mesher3::grid3d
...control::do_gridgen

Top of pagePrevious messageNext messageBottom of page Link to this message

Barnabas Kipapa (yenrab54)
Member
Username: yenrab54

Post Number: 15
Registered: 07-2007
Posted on Thursday, July 19, 2007 - 08:51 pm:   

here's the descriptor.


application/octet-streamnew config
BNless Configuration.pde (2.5 k)

Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action:

Topics | Last Day | Last Week | Tree View | Search | Help/Instructions | Program Credits Administration