Flat disk with pressure. Results are too low.

[laminarflow]

I have a diaphragm disk, .0625" thick, 4.944" diameter, 17-4 PH H-900. I fixed the very edge around the diameter, the .0625" thick OD edge, and put 50 psi on one face. I end up with .004" max deflection in the very center. That seemed low so I had a friend run the same test on Pro-E and he came up with .042" which seems correct. Hmmm, off by a factor of 10???

I check units, pressure, etc. and everything seems OK.

Could someone run this piece and see what they get?

Thanks!

 

[GBor]

Were you both using the same mesh density? What do you get from the hand calculation out of Roark's or a textbook? I'll run it, but "gut feel" tells me your friend's answer is correct.

 

[GBor]

Quarter symmetry model using plate elements, fixed around the perimeter. Displacement: 0.044". see attached bitmap. I used Algor.

Roark's, 6th edition, Table 24, Case 10b with r0 = 0:

y = -q*a^4/(64D) = -0.04438"

Looks pretty good to me.

Garland E. Borowski, PE
Engineering Manager
Star Aviation

 

[MechEng2005]

The constraints were to completely fix the edge around the diameter? Was this consistant in both? Same edge (pressure side versus non-pressure if the constraint was tied to an edge and not the .0625" wide surface)? Same constraint (not that one of you fixed translation only and the other fixed translation and rotation)?

I could definately see a descrepency between fixing only translation versus with rotation too... Don't know about the factor of 10, but possibly.

-- MechEng2005

 

[laminarflow]

Before we continue, I have to admit I have little experience using FEA. Only what I have done with Algor and Cadre. So I don't quite know what you mean by "Quarter symmetry model using plate elements". The Roark's and your model confirm the max displacement. I was just modeling a disk in Alibre and sending it to Algor design check, which is a single part only license that comes packaged in Alibre, my CAD package. Any chance of a bit of education on what I am trying to do? Thanks!

 

[GBor]

Well, I have "played" with Alibre, but really have no idea what flexibility you have in "Design Check". If it is anything like Cosmos Express that comes with SolidWorks...DUMP IT NOW AND RUN!!!

By "Quarter Symmetry Model", I have set boundary conditions along the x-axis and y-axis that simulate the entire disk. Think about it for a minute: Any deflection that would pull the nodes on the x-axis in the positive y-direction has an equal force trying to pull it in the negative y-direction. Because of this, you can set "x-symmetry" meaning that the line of nodes on the x-axis can only displace along the x- or z-axes...not the y. Using that same logic, you can constrain the y-axis nodes to move only along the y- or z-axes. That means that the middle node can only move straight down.

As for your model, I'm guessing you are using brick elements, which are completely inappropriate for a thin disk. They are stiff in bending (reduces displacement).

Are you able to set element types in Design Check? If so, you should use plates in your case. This should result in a mid-plane mesh of your solid disk. As an alternative, see if you can use the surface modeling capabilities and just model a 2-D surface of the disk. Apply your boundary conditions and loads and try to run design check. If it recognizes the 2-D geometry, it should use plate theory to calculate displacements and your answer should be closer to correct.

Hopefully this will get you started.

 

[4Pipes]

Design Check is very good for what it is. However, it is a feature limited version of Algor. I suspect the same applies to most similar cut down versions of major fe apps. You are limited only to brick elements. Great if brick elements are appropriate but not much use otherwise.
Deflection will be very inaccurate. You need 3 brick elements through the thickness. This is probably not practicable with Design Check. You could try quarter symmetry as suggested by GBOR and reduce element size which will get you closer.
Design Check will only do symmetry on orthoganal axes. It will not do any other smaller angle or axisysmmetic.

 

[laminarflow]

Not having access to the full version of Algor, I apparently didn't realize how limited it is. Too bad they don't offer an intermediate version. Any suggestions?

Thanks for running this and making me realize the limitations of Design Check.

I apparently have a lot to learn.

 

[GBor]

Since AutoDesk took over Algor, things have changed, so I am no longer confident of anything, but you used to be able to get their linear stress package fairly reasonably and they were willing to work with you on payment plans. With AutoDesk at the helm...all bets are off.

Design Check sounds exactly like Cosmos Express, which I usually refer to as "dangerous". The package is too limited and you have too little control. If you don't know what you are looking at, it will get you in trouble. If you do know what you are looking at, you probably have a more capable package.

At least now you know a little bit of what you didn't know before. Keep searching...you will find something useful.

If cost is the issue, a low-cost, but very good package is Roshaz (

http://www.Roshaz.com).

Right now, they are running a special for $595 for the package and 1 month of maintenance. Additional maintenance is $600/year. Couple this with the free, open-source code Calculix (

http://www.bconverged.com)

and you have a VERY capable package for a VERY reasonable pricetag. The Calculix pre-processor used to be horrible, but Roshaz helps ease the pain.

 

[laminarflow]

The problem I have buying a package is that I only need to check something only a few times a year. Hard to even learn and stay current like that I know.

Oh well, I have friends with complete packages so we'll just trade work.