FEA Ansys question

CatFromHue

hi

i was wondering if anyone could help me with a problem im encountering using ansys.

firstly im a distance student (i dont attend classes or tutorials as i work full time) in an engineering program and this is with a project i have to do.

the problem is that when i model the part and then go to mesh it i get an error message saying that the licence of the program im using cant acomodate the amount of nodes. its a non commercial licence.

the part is a plate with 10 holes in it. its not symmetrical. i have divided up the part into smaller separate sections. when i use a smaller mesh size though i get error messages indicating a poor mesh.

i have tried using the compressed number function in the numbering controls section but this hasnt worked.

does anyone have any ideas on how i can get around this problem or how i could restructure the mesh to reduce the node number?

all help is much appreicated

FesterBeatty

Sounds like you need to spend a few quid and get a decent version of this programme. Either that or keep it simple - first principles hand calculations!

Liquidus

The academic version has a node limit. I presume you are using solid elements for the plate? You will just have to increase mesh size for node element or reduce model size.

CatFromHue

yep its the academic version and yeah im using solid elements. due to the gemotery of the plate other element types are suitable.

i have tried increasing the mesh size but i get shape function errors and poor mesh errors.

the plates dimensions are 200mm long, 4.5 high, and 15 wide. when you say reduce the model size, do you mean reduce the dimensions to say 100mm long, 2.25 high and 7.5 wide and then scale down the load accordingly?

thanks for the replies too

breadmonkey

You could try using the adaptive analysis function. Type "adapt, a, b" into the command window and ANSYS will create a mesh and refine it where necessary.

a = number of cycles you will permit the program to do (to limit computation time)

b = target error

so, "adapt, 5, 10" would allow 5 cycles and seek a target error of 10% or less.

Also, I would have thought shell elements would be more suited to a plate problem.

Turbulent Bill

CatFromHue wrote: »

the part is a plate with 10 holes in it. its not symmetrical. i have divided up the part into smaller separate sections. when i use a smaller mesh size though i get error messages indicating a poor mesh.

Have you glued several volumes together to make the complete beam? If so, ANSYS could be refining the mesh excessively at the volume boundaries. Keep the model as simple as you can.

Can you run a quick analysis of the plate without any holes, just to prove it meshes and solves something correctly? From memory even the academic version could handle 100k+ nodes, more than enough for a plate, so it's probably just a settings issue.

CatFromHue

what i have done is actually divide the plate into a series of 2 smaller volumes i.e. around each hole is one volume and then the bits between the holes would be another volume.

im using solid 95 elements.

all good ideas. thanks everyone. another problem i have now is i cant get out to the university!

the non comercial version they have there has a limit of 32000 nodes i think. i have a copy of the student version but that has a limit of 1000 nodes so ill try some of your ideas when i get home.

thanks again everyone

Liquidus

Yes 32k with academic introductory and 512k with academic advanced. With those dimensions it should work as the model isn't large. As Turbulent Bill said it's probably something with the volume boundaries. Solid elements should work with this model. Shell element are usually used with tubes etc.

RobHill

Have you tried using inflation layers? You can use these to create a dense mesh around the areas that you expect to see boundary layer separation or other areas of interest. This allows you to keep a course mesh for the majority of the model and not lose too of the detail you are interested in.

enda1

I don't know ansys, but I presume its not so different to other codes.

Shell elements definitely first of all. Also limit to first degree elements. So you will only have 4 nodes per element, really 1 and a bit unique nodes per element on average.

Next make sure you have no coincident nodes and respect connectivity.

Around your holes have 6/8 elements in a ring. You can double the ring concentrically if you are particularly interested in stress results around the ring.

32k node should be more than enough.

What is hole diameter and general plate dimensions? Is is tapered or planar?

Paul_Designer

Have you tried solving the problem using 2D? FRom your description you should be able to use a 2D element to solve the problem?

CatFromHue

thanks for all the input but in the end i had to do a 50%scaled model to get it to run.

the plate was too thick to use shell elements (4.5mm thick, 15mm wide, 190mm long)

there are holes in the plate which are important factors in how much it deflects under a load so i think that rules out 2D.

thanks for all your help but if a mod could close this thread it would be great.