Eigenfrequency Z88 solver not supported by FC

About the development of the FEM module/workbench.

Moderator: bernd

User avatar
uwestoehr
Posts: 627
Joined: Sun Jan 27, 2019 3:21 am

Re: Eigenfrequency Z88 solver not supported by FC

Postby uwestoehr » Sat Feb 02, 2019 12:56 pm

thschrader wrote:
Fri Feb 01, 2019 8:18 pm
Running a FEM-analysis on FC is 10-times faster then using Z88.
It seems you misunderstood me. To clarify, I opened this thread because in real life it is important to use at least 2 different solvers. Of course I tried out CalculiX with FC and it works fine. Nevertheless, using the same mesh in Z88 gives me other Eigenfrequencies than CalculiX and this is what I expected. Therefore it would be very valuable for real life application if FC would offer to perform simulation with different solvers.

It might be that Z88 is not bleeding edge or doesn't have much user activity, but it seems that one can rely on its results. Therefore Z88 support should stay and it should be possible to use all analysis modes of Z88.

Concerning Z88 Aurion, I tried it of course already and that V1 of this solver did not have any output was kind of ridiculous. V2 offers now at least STL export. It is clear that this is not sufficient for integration with FC.
It seems to me that Z88 works together with companies and therefore they don't offer that many possibilities. This would also explain the low activity on the Z88 forum because as employee you are normally not allowed to post in forums (on what project would you book this time etc.).
User avatar
bernd
Posts: 8238
Joined: Sun Sep 08, 2013 8:07 pm
Location: Zürich, Switzerland

Re: Eigenfrequency Z88 solver not supported by FC

Postby bernd » Sat Feb 02, 2019 4:26 pm

no worries, z88 will stay as solver in FreeCAD.
User avatar
bernd
Posts: 8238
Joined: Sun Sep 08, 2013 8:07 pm
Location: Zürich, Switzerland

Re: Eigenfrequency Z88 solver not supported by FC

Postby bernd » Sat Feb 02, 2019 4:40 pm

@uwe

if you know Z88 input file format and Python, extending this in FreeCAd is really not difficault at all.
User avatar
bernd
Posts: 8238
Joined: Sun Sep 08, 2013 8:07 pm
Location: Zürich, Switzerland

Re: Eigenfrequency Z88 solver not supported by FC

Postby bernd » Sun Feb 03, 2019 8:31 pm

bernd wrote:
Sat Feb 02, 2019 4:40 pm
@uwe

if you know Z88 input file format and Python, extending this in FreeCAd is really not difficault at all.
may be not. Two years ago on GSoC a new system for choosing the analysis type was introduced. The equation object. ATM this only works for ElmerSolver. If more analysis types should be used in Z88 I would recommend to use the Elmer Equation system. We would need to switch for Calculix too. I would take care of CalculiX if you decide to give it a try.

bernd
User avatar
uwestoehr
Posts: 627
Joined: Sun Jan 27, 2019 3:21 am

Re: Eigenfrequency Z88 solver not supported by FC

Postby uwestoehr » Mon Feb 04, 2019 11:24 am

bernd wrote:
Sun Feb 03, 2019 8:31 pm
If more analysis types should be used in Z88 I would recommend to use the Elmer Equation system. We would need to switch for Calculix too. I would take care of CalculiX if you decide to give it a try.
Hi Bernd,
I have no experience yet with Elmer, started to try it yesterday. However, I have now a fixed real life project where I will have to perform simulations and I will have to use Elmer.
I make now my first steps and when I am into it I want of course give something back. But since I am too new in FC FEM and need to learn first, I can currently only offer to test out things and give feedback.

As I wrote, it is important for real life projects to use at least 2 different solvers and I am amazed that FC offers this quite nicely. If for all 3 supported solvers, CalculiX, Z88 and Elmer FC would use the same equation system setup, it would be a big advantage to perform quickly multi-solver analyses.

Uwe
User avatar
bernd
Posts: 8238
Joined: Sun Sep 08, 2013 8:07 pm
Location: Zürich, Switzerland

Re: Eigenfrequency Z88 solver not supported by FC

Postby bernd » Mon Feb 04, 2019 4:34 pm

uwestoehr wrote:
Mon Feb 04, 2019 11:24 am
I make now my first steps and when I am into it I want of course give something back. But since I am too new in FC FEM and need to learn first, I can currently only offer to test out things and give feedback.
You gave lot of value feedback already. Keep going :D


uwestoehr wrote:
Mon Feb 04, 2019 11:24 am
As I wrote, it is important for real life projects to use at least 2 different solvers and I am amazed that FC offers this quite nicely.
Exact this is the plan including the possibility to add even more solvers.


uwestoehr wrote:
Mon Feb 04, 2019 11:24 am
If for all 3 supported solvers, CalculiX, Z88 and Elmer FC would use the same equation system setup, it would be a big advantage to perform quickly multi-solver analyses.
This is just a matter of time.
thschrader
Posts: 1407
Joined: Sat May 20, 2017 12:06 pm
Location: Germany

Re: Eigenfrequency Z88 solver not supported by FC

Postby thschrader » Mon Feb 04, 2019 9:39 pm

uwestoehr wrote:
Sat Feb 02, 2019 12:56 pm
thschrader wrote:
Fri Feb 01, 2019 8:18 pm
Running a FEM-analysis on FC is 10-times faster then using Z88.
It seems you misunderstood me. To clarify, I opened this thread because in real life it is important to use at least 2 different solvers. Of course I tried out CalculiX with FC and it works fine. Nevertheless, using the same mesh in Z88 gives me other Eigenfrequencies than CalculiX and this is what I expected. Therefore it would be very valuable for real life application if FC would offer to perform simulation with different solvers.

It might be that Z88 is not bleeding edge or doesn't have much user activity, but it seems that one can rely on its results. Therefore Z88 support should stay and it should be possible to use all analysis modes of Z88.

Concerning Z88 Aurion, I tried it of course already and that V1 of this solver did not have any output was kind of ridiculous. V2 offers now at least STL export. It is clear that this is not sufficient for integration with FC.
It seems to me that Z88 works together with companies and therefore they don't offer that many possibilities. This would also explain the low activity on the Z88 forum because as employee you are normally not allowed to post in forums (on what project would you book this time etc.).
I agree to to your comment.
To 80%. The missing 20% are:
Why do you compare solver versus solver (means: calculix versus Z88)
Why dont you calibrate your model at an existing analytical solution?
Can you give a screenshot of your problem. Freecad CAD file?
Thomas
User avatar
HarryvL
Posts: 1042
Joined: Sat Jan 06, 2018 7:38 pm

Re: Eigenfrequency Z88 solver not supported by FC

Postby HarryvL » Mon Feb 04, 2019 10:17 pm

thschrader wrote:
Mon Feb 04, 2019 9:39 pm
Why dont you calibrate your model at an existing analytical solution?
Thomas
I agree with Thomas. I don't see why you would need 2 solvers just so you can check the validity of your solution. In my experience you are best off by making a simplified analysis first (either by hand or your favorite solver) and then refine where required.
User avatar
uwestoehr
Posts: 627
Joined: Sun Jan 27, 2019 3:21 am

Re: Eigenfrequency Z88 solver not supported by FC

Postby uwestoehr » Mon Feb 04, 2019 11:16 pm

thschrader wrote:
Mon Feb 04, 2019 9:39 pm
Why do you compare solver versus solver (means: calculix versus Z88)
That dates back some years when I worked at the university and we found out that Comsol gave other results that the simulation in Solidworks. The simulation guys checked that the mesh and geometry was the same. Since that I have this in mind. I don't know how the different solvers calculate internally exactly and I don't have the time at work to investigate this.
thschrader wrote:
Mon Feb 04, 2019 9:39 pm
Why dont you calibrate your model at an existing analytical solution?
In reality we have a complex structure with many different holes (different diameters) in different "levels" of a block-like structure. Not trivial. I could run different solvers on a simple block without holes to "calibrate" them but I don't know how the solvers work internally to judge if this "calibration" result can be transferred to complex meshes. I mean for my real geometry I already get different Eigenfrequencies with Z88 and CalculiX using the same mesh.
I'll now try to run both solvers on a simple block and see how they behave. However, I don't have the time to produce such a block and test if using the calculated Eigenfrequencies lead in maximal vibration amplitude for the produced part. And that is the point - I want to simulate as best as possible to modify the real block to shift the Eigenfrequencies to values above e.g. 300 Hz. Then I know there won't be problems in reality. Thus if e.g. Z88 say the first Eigenfrequency is above 300 Hz while CalculiX says no, I better change the block geometry so that both solvers give me frequencies above 300 Hz. It is much quicker and much cheaper to run some more simulations than to produce the real part.

I must admit, that I don't have that much expertise in simulations for real life so if you think I make a logical mistake please tell me that I can learn or correct myself.
thschrader wrote:
Mon Feb 04, 2019 9:39 pm
Can you give a screenshot of your problem. Freecad CAD file?
I am an industry guy and cannot make screenshots of real bodies.
User avatar
bernd
Posts: 8238
Joined: Sun Sep 08, 2013 8:07 pm
Location: Zürich, Switzerland

Re: Eigenfrequency Z88 solver not supported by FC

Postby bernd » Tue Feb 05, 2019 7:21 am

I see both points. I have to face them in daily business too.

May be all this depends on the part to analyses. I'm structural engineering guy. As long as I only calculate forces, there will be a possibility to make some simple calculations to get some result. The FEM should not be more than 20 ... 30 % away, may be even less 5 ... 10 %. This depends on the part. But I can imagine there is part geometry out there on which a calculation on paper is just not possible.