Hi, i have some troble for getting coherent results in a keyseat simulation. I think this is because of the mesh.
let me show you what i mean:
standard mesh with control of 0.5mm, relation 1.2, the mesh is not even(for me)
based on curvature mesh:( i think it is a better. the thing is that i only want to focus o the end fillet, not the whole part and i can´t do it with this tipe of mesh)
minimun size of element: 0.5mm
number of circles: 36
relation: 1.2
What else can i do?
I am going to assume here that you want get a better mesh than what you are getting with this being defined as lots of elements where you want them and not so many where you don't and a nice transition between the two areas. Which mesher you use is up to you. A key thing to remember is that when you want a smoothly transitioning mesh and you use use a low number of say 1.1 or 1.2 as the growth ratio you need to select the surfaces that get the mesh control wisely. Wisely usually means that you pick a set of surface that may extend beyond what you are exactly after so that the transition area of the mesh happens on surfaces that have the room to accomodate the growth before you move to another surface patch. Otherwise the mesh has to grow much faster since you don't have the room to execute the growth rate. Some of your previously displayed meshes had behavior that looked like they were not wisely (as mentioned above) choosen.
Actually those meshes were on the other post. You can use split lines to delinate the surfaces where you want to have a mesh control with more specificity.
thank you guys for the answers.
Bill, you are right. I didn´t had in mind the concepts you just said. i didnt define the control mesh wisely.
with coherent i meant this: i made a design study on a based on curvature mesh varying the minimun size of the elements from
5
4
2
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
all in mm. i got this:
is like the results never converge and the 0.2mm based on curvature solution is not coherent for me.
with all of those hypothesis, which mesh should i consider the most apropiate?
converge on displacement, not stresses or other parameters
don't look at just the model max and min, look at a couple of locations
and don't choose any until you figure out why the spike is happening
i forgot to tell you that the values of the simulation were defined by a sensor located at the end of the fillet
Your mesh converged at 70 - whatever those units are kgf/cm^2 if the bottom axis is the inverse of the element size - on your schedule. The bump at the end is from an "over constrained mesh" - yeah it happens. When doing these mesh convergence studies you don't need to get more than 2 that are close. When I do these I exactly half the element size for mesh convergence studies - 4 mm then 2 mm then 1mm.
the issue is that the smallest radius of the fillet is 0.5mm and i thought that a mesh of 1mm wont be appropiate.
maybe a size of 0.2mm or lees would be satisfactory. excluding that extrange value, it looked like it was converging.
ok, thank you jared. i didnt know that either
should i use the equivalent displacement or in one defined direction?
as noted, multiple locations gives you better confidence. multiple values will give you even more.
i located tree sensors for defining the equivalent displacement.
and i ran the analisys. the results were almost cosntant.
i am really dizzy since i need to know the value of the stresses...
see bills comment and also do a search for singularities. disaplcements converge, stresses won't always. or you may have a spike due to some other issue. This is where looking at multiple parameters helps.
i'd recommend really sitting down wth an expert to go over this one-on-one. if you'd like we can help you with that here at Hawk Ridge Systems. jared@hawkridgesys.com if you want to discuss some of the options.
thanks guys, i really appreciate your help.
the only experts who are helping me out are you.
based on your previous post, i suspeect "coherent" results is likely more an aspect of your setup vs the mesh.
curvature vs standard mesher = equivalent. curvature mesher is generally better at handling bad geometry and is multithreaded for faster meshing. but otherwise pretty equivalent from an accuracy perspective.
regarding this actual problem, i don't know what the problem defintiion is or how you think it is affecting the results.
i think your question is "why can't i create a mesh control with curvature based mesher"?
and "why isn't the standard mesher more uniform"?