Non Linear Analysis - Hinge example
Moderatoren: ccad, mz15, auroraIco, Lehrstuhl
Non Linear Analysis - Hinge example
Good day Aurora Team
I tried the Hinge example in order to learn how to manage Non Linear Analysis. I believe I correctly followed the instructions and applied the indicated model, mesh, BCs. etc. Finally I think I got the same results of the guide. But...
The guide points that the elapsed time for the Pardiso calcutation should be about 45' (for a 4 CPU / 16 GB RAM system). In my system (WXP 32 Bits, 3 CPU, 4 GB RAM), this time has been of 8' 53" for Pardiso, and 14' 39" for SORCG. This differences cause me doubts about the results I got and/or if I made any mistake. (I'm running AURORA V2A)
Attached some images for you to evaluate this subject. Thanks a lot in advance
I tried the Hinge example in order to learn how to manage Non Linear Analysis. I believe I correctly followed the instructions and applied the indicated model, mesh, BCs. etc. Finally I think I got the same results of the guide. But...
The guide points that the elapsed time for the Pardiso calcutation should be about 45' (for a 4 CPU / 16 GB RAM system). In my system (WXP 32 Bits, 3 CPU, 4 GB RAM), this time has been of 8' 53" for Pardiso, and 14' 39" for SORCG. This differences cause me doubts about the results I got and/or if I made any mistake. (I'm running AURORA V2A)
Attached some images for you to evaluate this subject. Thanks a lot in advance
- Dateianhänge
-
- SORCG
- HINGE SORCG.jpeg (114.01 KiB) 8560 mal betrachtet
-
- PARDISO
- HINGE PARDISO.jpeg (120.18 KiB) 8560 mal betrachtet
-
- RESULTS AT PATH 25
- HINGE NO LINEAR.jpeg (86.83 KiB) 8560 mal betrachtet
Re: Non Linear Analysis - Hinge example
Hello selopez,
please note that the runtimes specified in the guide are mere guidelines and vary depending on the system. Also we introduced an accelerated (RS) version of our solvers with the release of V2a, which might account for the decrease.
So, judging by your results and the console readouts, I would not worry about the decreased runtime.
Have fun using our software!
Best regards
Felix Viebahn
please note that the runtimes specified in the guide are mere guidelines and vary depending on the system. Also we introduced an accelerated (RS) version of our solvers with the release of V2a, which might account for the decrease.
So, judging by your results and the console readouts, I would not worry about the decreased runtime.
Have fun using our software!
Best regards
Felix Viebahn
Re: Non Linear Analysis - Hinge example
Felix, thank you very much for your answer.
I'm not really "worried" about the runtime, but I'm still confused about some other subjects related to the outputs of this case.
I observed that if a Linear Analysis is applied to the same model and BCs, the total displacement results are the same than that of the 25th path at the Non Linear analysis. At the NLA I also observed that the progressions of loads and displacements along the 25 paths, are clearly linears.
So, is the difference between LA and NLA just a matter of graphical output? I know that this last is not true and I would like to have your opinion.
As usual, thanks a lot.
I'm not really "worried" about the runtime, but I'm still confused about some other subjects related to the outputs of this case.
I observed that if a Linear Analysis is applied to the same model and BCs, the total displacement results are the same than that of the 25th path at the Non Linear analysis. At the NLA I also observed that the progressions of loads and displacements along the 25 paths, are clearly linears.
So, is the difference between LA and NLA just a matter of graphical output? I know that this last is not true and I would like to have your opinion.
As usual, thanks a lot.
Re: Non Linear Analysis - Hinge example
Hello selopez,
it matters on the amount of load. If you put a small strain on it, you will have the same results as a linear simulation. The module for the non linear analysis is only for large deformations of the FE-mesh. That occurs only, if you have high loads. The linear analysis is only for small deformation of the FE-mesh. Therefore, it could occur, that you have the same results. Whether you simulate linear or nonlinear.
Best regards
Michael Frisch
it matters on the amount of load. If you put a small strain on it, you will have the same results as a linear simulation. The module for the non linear analysis is only for large deformations of the FE-mesh. That occurs only, if you have high loads. The linear analysis is only for small deformation of the FE-mesh. Therefore, it could occur, that you have the same results. Whether you simulate linear or nonlinear.
Best regards
Michael Frisch
Re: Non Linear Analysis - Hinge example
Estimated Mr. Fisch
As you point out, and as I understand it, for small displacements, the projected load variation (either by the angular variation of the application area or by its distance to the position constraints) is not enough to be relevant, so the displacement progression appears to be linear.
To see a clearer nonlinear progression I tried first to increase the load. Here I noticed that the 900,000 N of the example seem to be the maximum acceptable for the solver. Even with slightly higher charges occurred, at some of the paths, something like a non-resolvable loop from which the process could not continue. The higher the load, earlier occurred this loop. I extend this explanation because it would be very valuable for me to know the reason for this malfunction. If allow myself to fantasize, is as if the solver detects a kind of violation of the Poissons ratio of the material. Another possible explanation is that the angle of the load application area makes that the X value of the force approaches to 0 and this generates algebraic incongruity.
Finally I modified the geometry of the model in a way to do it "more flexible" (longer in Y - see picture). Here NLA results were clear. Considering the maximum displacements for each path, between 1/2 Δ = 8.01 MM, 12/13 Δ = 5.3 MM and 24/25 Δ = 2 MM. Result that is very consistent. In the NLA, the maximum displacement for step 25 was 136 MM, when for the LA it was 199 MM. In my opinion this difference is what highlights the importance of the NLA.
Thank you very much for your attention.
Regards
As you point out, and as I understand it, for small displacements, the projected load variation (either by the angular variation of the application area or by its distance to the position constraints) is not enough to be relevant, so the displacement progression appears to be linear.
To see a clearer nonlinear progression I tried first to increase the load. Here I noticed that the 900,000 N of the example seem to be the maximum acceptable for the solver. Even with slightly higher charges occurred, at some of the paths, something like a non-resolvable loop from which the process could not continue. The higher the load, earlier occurred this loop. I extend this explanation because it would be very valuable for me to know the reason for this malfunction. If allow myself to fantasize, is as if the solver detects a kind of violation of the Poissons ratio of the material. Another possible explanation is that the angle of the load application area makes that the X value of the force approaches to 0 and this generates algebraic incongruity.
Finally I modified the geometry of the model in a way to do it "more flexible" (longer in Y - see picture). Here NLA results were clear. Considering the maximum displacements for each path, between 1/2 Δ = 8.01 MM, 12/13 Δ = 5.3 MM and 24/25 Δ = 2 MM. Result that is very consistent. In the NLA, the maximum displacement for step 25 was 136 MM, when for the LA it was 199 MM. In my opinion this difference is what highlights the importance of the NLA.
Thank you very much for your attention.
Regards
- Dateianhänge
-
- HingeXNL.jpeg (84.09 KiB) 8490 mal betrachtet
Re: Non Linear Analysis - Hinge example
Friends of Aurora Team:
Please, excuse me for adding more questions to the subject of NLA.
In the previous email I pointed out the difference between the NLA final displacement (136 mm) and that of the LA (199 MM). In other words, the output the NLA was found to be 68.3% of the output of the LA.
Given that the NLA does not perform stress calculation, I proceeded to make a LA with a load of 615 000 N (900 000 x 0.683). This is, reducing the load in the same proportion in which the displacement is reduced in the NLA. In this LA the maximum displacement coincided with the final maximum displacement of the NLA (136 mm). Obviously, the calculated stresses showed the same proportionality: LA @ 900 000 N – maximum stress = 40 700 MPa stress / LA @ 615 000 N – maximum stress = 27 800 MPa stress.
I would like then your opinion about if this extrapolation of the displacement differences (NLA / LA) to the stress calculation, is reasonable and/or realistic.
Thank you very much. Best regards.
Please, excuse me for adding more questions to the subject of NLA.
In the previous email I pointed out the difference between the NLA final displacement (136 mm) and that of the LA (199 MM). In other words, the output the NLA was found to be 68.3% of the output of the LA.
Given that the NLA does not perform stress calculation, I proceeded to make a LA with a load of 615 000 N (900 000 x 0.683). This is, reducing the load in the same proportion in which the displacement is reduced in the NLA. In this LA the maximum displacement coincided with the final maximum displacement of the NLA (136 mm). Obviously, the calculated stresses showed the same proportionality: LA @ 900 000 N – maximum stress = 40 700 MPa stress / LA @ 615 000 N – maximum stress = 27 800 MPa stress.
I would like then your opinion about if this extrapolation of the displacement differences (NLA / LA) to the stress calculation, is reasonable and/or realistic.
Thank you very much. Best regards.
Re: Non Linear Analysis - Hinge example
Dear selopez,
such a stress extrapolation is not possible, it is very dangerous to use it. If you want to get the stresses in the nonlinear case, just use the results given by the nonlinear solver. Therefore, you have to switch to "stresses at the Gauss points". The reason is that stresses are related to the outer forces which do not differ between the linear and the nonlinear case. Stresses are not related (except if the material law is introduced) to the displacements which differ between linear and nonlinear analysis.
If you have further questions, do not hesitate to ask!
Best regards,
Christoph Wehmann
such a stress extrapolation is not possible, it is very dangerous to use it. If you want to get the stresses in the nonlinear case, just use the results given by the nonlinear solver. Therefore, you have to switch to "stresses at the Gauss points". The reason is that stresses are related to the outer forces which do not differ between the linear and the nonlinear case. Stresses are not related (except if the material law is introduced) to the displacements which differ between linear and nonlinear analysis.
If you have further questions, do not hesitate to ask!
Best regards,
Christoph Wehmann
Re: Non Linear Analysis - Hinge example
Dear Mr. Wehmann, thank you very much for your reply.
Although in this case the results for Stresses at Gauss Points of the LA @ 615 000 N are quite similar to the NLA @ 900 000 N, I share with you the warnings about the risks of a simple extrapolation. I don't fully understand when you refer to "introducing the material laws" in the NLA. I guess you are not referring to re-crystallization processes occurring at the plastic deformation, which modify the material properties.
A minor bug report. At the "Filter" of the "Post-processor", when a new maximum or minimum value is set, it’s not possible to return to the unfiltered view as it was in Aurora V1. The only way I found to restore the unfiltered view, is to close the project and reopen it.
Best regards.
Although in this case the results for Stresses at Gauss Points of the LA @ 615 000 N are quite similar to the NLA @ 900 000 N, I share with you the warnings about the risks of a simple extrapolation. I don't fully understand when you refer to "introducing the material laws" in the NLA. I guess you are not referring to re-crystallization processes occurring at the plastic deformation, which modify the material properties.
A minor bug report. At the "Filter" of the "Post-processor", when a new maximum or minimum value is set, it’s not possible to return to the unfiltered view as it was in Aurora V1. The only way I found to restore the unfiltered view, is to close the project and reopen it.
Best regards.
Re: Non Linear Analysis - Hinge example
A quicker way to reset the Min/Max values of the filter is to leave and re-enter the postprocessor. That way you do not have to close and reload the whole project.
Re: Non Linear Analysis - Hinge example
Finally I could discover the utility of the little brush. It works fine. Thanks a lot.
Re: Non Linear Analysis - Hinge example
Hello,
you're welcome. I just saw that this is not in the documentation, so you might not know it.
regards
mz15
you're welcome. I just saw that this is not in the documentation, so you might not know it.
regards
mz15
Re: Non Linear Analysis - Hinge example
Aurora Team:
Concerning similar issues, check please the Z Limit tool. Here it seems to be also impossible to restore the initial view (un-clipped) and the little brush does not help. But, IMHO, these are not so important matters as the fact that stresses in NLA are exposed just at the Gauss Points, what requires a more careful reading.
Regarding this last, in the Theory Manual, at 3.2.8 , pages 46-47, INTOS_TYP_1…24”, it can be read: “For isoparametric elements Nr.16 and 17…A good value is 7 for type 16. For type 17 1 gauss point may suffice." Is this correct?
As usual, thank you very much for your attention.
Concerning similar issues, check please the Z Limit tool. Here it seems to be also impossible to restore the initial view (un-clipped) and the little brush does not help. But, IMHO, these are not so important matters as the fact that stresses in NLA are exposed just at the Gauss Points, what requires a more careful reading.
Regarding this last, in the Theory Manual, at 3.2.8 , pages 46-47, INTOS_TYP_1…24”, it can be read: “For isoparametric elements Nr.16 and 17…A good value is 7 for type 16. For type 17 1 gauss point may suffice." Is this correct?
As usual, thank you very much for your attention.
Re: Non Linear Analysis - Hinge example
Aurora Team:
I apologize for pointing out a malfunction of the Z Limit tool. Actually the problem is in the way I set my graphics card to run Z88 AURORA (in order to see round picking points). Best regards.
I apologize for pointing out a malfunction of the Z Limit tool. Actually the problem is in the way I set my graphics card to run Z88 AURORA (in order to see round picking points). Best regards.