reliable results

[morpheus87jr]

I hope I don't make doubles. . .

I wanted to propose as a discussion the reliability of the results of a fem analysis.

Taking some details of which I carried out analytical calculation, I tried to perform also the fem analysis that returned me comparable values only in some areas.
I know that this depends both on the simplifications made by analytical theories and also on the difficulties in managing constraints and loads.
(I recall the remarkable difference between the axially binding the entire shouldering surface of a tree and the binding one point).


in complex cases how do you proceed to see if the numbers that the computer returns to me are reliable?

(I know that it is often a matter of experience and eye on the part of the designer, but are there procedures that help in this sense? )

 

[cacciatorino]

in complex cases how do you proceed to see if the numbers that the computer returns to me are reliable?

(I know that it is often a matter of experience and eye on the part of the designer, but are there procedures that help in this sense? )

I think your way of proceeding is sensible. I also usually develop the problem on a simple problem of which I can predict the trend by analytical way. once the "simple" model solution is reliable, step to the complex one using the same assumptions.

If the solutions of the complex problem were easily known by analytical means, numerical analysis would become useless, right? then, as you say, it is very important to experience this kind of problem, otherwise the risk of producing simple colored images, instead of analysis of stress/deformation/etc, is concrete.

p.s.

There are some methods to verify that the analyses you are doing follow some rules of "good practice", such as the continuity of the constant tension lines or the absence of sudden inversions of the sign of the quantities. you can find some considerations in this regard, expressed in a simple way, on the first book of ingenuine.

 

[morpheus87jr]

you can find some considerations in this regard, expressed in a simple way, on the first book of ingenuine.

the library will soon receive visits (post examination of thermodynamics -.- ).
thanks for the reference!


sewing here and I found it also who performs "in series" simulations starting from very wide mesh by inflating away in points of greater interest and verifying the convergence of the various solutions obtained. . .

It seems to me a little indirect way to verify also because there should be rather gross errors for a non-convergence of solutions gradually more dense....


question from student "sbarbatello" (without great experiences)... when a relationship arises and results fem, it is necessary to return a range of confidence or in any case a reliability of the same or just to indicate the parameters used?

 

[meccanicamg]

Bye to all,
I would say that fem analysis is a bad beast if you do not impose it correctly. I've also done my unmatched models to understand how simulators behave. I would say that the bearing is perhaps among the worst. you should really consider what bearing you use....if you allow or not orientation. I also did my tests on this and as soon as I have a moment of time I find myself posting the clam differences and the sense of the fem in this regard.

 

[bisogna_rimanere_umili]

...
you can find some considerations in this regard, expressed in a simple way, on the first book of ingenuine.

whose title is?

 

[morpheus87jr]

I think it refers to this:


"structural design with computer. guidelines in the automatic calculation of structures"

 

[stef_design]

I think it refers to this:


"structural design with computer. guidelines in the automatic calculation of structures"

Does anyone have this book? Are you referring to the use of a particular software or is it a generic book?
examples faced with what fem software are executed?

 

[cacciatorino]

Does anyone have this book? Are you referring to the use of a particular software or is it a generic book?
examples faced with what fem software are executed?

I bought it, I even got a second, but I still have to go pick it up in the library.

no, it does not refer to specific software, although I think I understand that the examples are taken from ansys, but I repeat the level of abstraction from the used software is remarkable. It's a book for beginner "finti" in the sense that if you already have a minimum of infarination you understand better the concepts expressed. It is however to be considered as a "divulgative" book on fem technique, in fact the examples mentioned are always treated so as to extract general teachings, but you will not find examples of command files for a certain fem or another. I found it a useful reading, I think if one has some money to spend (even many at the end) it sure doesn't hurt to read it. If you're already very expert on the fem, maybe it's better to focus on more specific texts.

 

[cacciatorino]

I think it refers to this:


"structural design with computer. guidelines in the automatic calculation of structures"

That's right. the first book however treats only linear analysis.

 

[stef_design]

thanks for info hunting:finger:
the second volume is:advanced structural design techniques. calculation of composite materials, instability, dynamics and non-linearity ?

 

[cacciatorino]

thanks for info hunting:finger:
the second volume is:advanced structural design techniques. calculation of composite materials, instability, dynamics and non-linearity ?

yes!

vabbe', today allora:
http://www.claudiogianini.it/pubblicazioni_tecniche/pubblicazioni_tecniche.htmlI hope this message is not taken as spam! :biggrin:

 

[MBT]

yes!

vabbe', today allora:
http://www.claudiogianini.it/pubblicazioni_tecniche/pubblicazioni_tecniche.htmlI hope this message is not taken as spam! :biggrin:

:angry:

 

[morpheus87jr]

I also recommend the books of prof. gugliotta.
(extensively treat the mathematical part of finite elements)

 

[cacciatorino]

I also recommend the books of prof. gugliotta.
(extensively treat the mathematical part of finite elements)

It's not exactly an umbrella reading then, unless you put yourself in full and matrices on the sand, hoping that you don't get the baby pallonata near the last step! :biggrin:

 

[michele81]

I also recommend the books of prof. gugliotta.
(extensively treat the mathematical part of finite elements)

was my prof.
Somewhere on the site of the polytechnic of torino are downloadable the dispenses of his course, which resume his publications on the theory that lies behind the fem. for nothing a informative reading....:wink:

 

[paolino]

was my prof.
Somewhere on the site of the polytechnic of torino are downloadable the dispenses of his course, which resume his publications on the theory that lies behind the fem. for nothing a informative reading....:wink:

Hi.
It was also my professor... I would like to point out also the pantry on assisted design of mechanical structures of prof. Somà, always of the polytechnic of torino, also here not the ideal under the umbrella :biggrin: but in my opinion the treatment is much clearer and more linear than that of gugliotta... ravanando in the website of the pole should be found

 

[michele81]

Hi.
It was also my professor... I would like to point out also the pantry on assisted design of mechanical structures of prof. Somà, always of the polytechnic of torino, also here not the ideal under the umbrella :biggrin: but in my opinion the treatment is much clearer and more linear than that of gugliotta... ravanando in the website of the pole should be found

quoto! excellent dispenses of his course...
Among other things it was a nice course with the use of ansys both in line and in dynamic.
with him I have also done a good thesis.. .

 

[stef_design]

quoto! excellent dispenses of his course...
Among other things it was a nice course with the use of ansys both in line and in dynamic.
with him I have also done a good thesis.. .

Does qiualcuno have the link of its dispenses?

 

[michele81]

Does qiualcuno have the link of its dispenses?

Look here http://www.mondovi.polito.it/ebook/pubbl.htmlthe link is no longer that of a few years ago but I think it is still all freely downloadable.

 

[DeltaS4]

I have my personal and modest experience in the fem field. as mentioned earlier when you start a simulation you have to have in mind already more or less what you have to come out, what I mean is that you do not have to take for good everything that the fem gives us, you can do a parallel with the cad, who knows the design we are not the program. Another important point is to look for an existing analytical solution of the problem, although a little different but that at least gives us an order of magnitude of what must come from the calculations. great importance is the model implemented it must reflect in a satisfactory way the reality, this does not mean starting from the beginning with something complicated (it weighs only the calculation) but studying the first data pulled out and then honing it more and more. in this part I would also insert the mesh, it starts from a simple discretizzazzione and away it hones there where you expect gradients. in parallel it is useful to also study the convergence of the result (in the area of interest) to vary the number of elements of the mesh (just a simple graph) and you will see that it does not take much.
other thing to keep in mind is the type of element to use, because it is precisely the mathematical description of the same that rounds it more or less suitable for this purpose. for example for the beam of timoshenko or use enriched elements or try to appear with a sub-integrated one, to avoid the phenomenon of looking. It is useful for example to launch simulation with two different types of elements and see the most conservative solution. Another question can be the extrapolation of data, from gauss points to knots, generally happens in a linear way, after an average weighed between the elements afferent to that node, a procedure used for example and to make the gauss point happen right at the exact point where you want to have the result, at that point I have an exact and unextracted value.

as recommended first read the book of the professor gugliotta that despite everything is one of the simplest around, I studied it I loved, then the mathematics behind the fem is there and serves you to understand what they do.
Don't make you think of reading zienkiewicz, despite the fem's dad, it's unbeatable, it's a pure analysis treaty without explanations though :biggrin: , is an encyclopedia for those who are already really expert in the matter.

I hope I have contributed to knowledge :biggrin:

Hi.