Fea bearing Models

[fnm]

I want to analyse stress distribution in a surface when I apply loads through a ball but I don't know how to handle this type of problem. How do I stablish in my drawing the type of elements at the initial point of contact between the ball and the surface? The problem I want to solve is to predict axial displacement of a bearing when I apply axial loads. Anybody can help me?

 

[GBor]

Can you post an image of what you are trying to do? Also, are you using MES?

 

[fnm]

Attached is a section of a Hiwin model AGH25 linear guideways. This bearing is part of a router moving table and I want to predict displacements.
What do you mean by MES?

 

[GBor]

MES = Mechanical Event Simulation

This looks like a 2-dimensional outline of a cross-section. Is your intention to analyze it as a 2-D, linear, static model? Can you describe your loads and boundary conditions?

Garland E. Borowski, PE
Engineering Manager
Star Aviation

 

[fnm]

Of course drawing is an outline of a cross section and I'm interested in 3D only, not 2D, static stress analysis. My difficulties begin when I have to chose the element type and its size and adequate meshing at the interface between the ball and the surface of the rail and linearguide. Every ball has only 4 contact points against its outside surfaces... but my difficulty is located at the contact point between the ball and the rail surface.

 

[4Pipes]

I would ask Algor support first.

I would guess:
a) Put balls in separate part. Deactivate balls(or deactivate in orginal cad model).
b)Insert construction points at each ball to raceway contact.
d)Draw a rigid element between each pair of contact points to create pseudo balls. Each would be a separate part.
e)Add boundary conditions to define location of pseudo balls.
f)Add surface contact between pseudo ball parts and raceways. Add temperature to pseudo balls if preloaded.
g)Run Algor to calculate forces and deflections at pseudo balls.
h) create a second model of a half ball (in brick) and part of the raceway. Apply forces. Check deflections to make sure they are comparable to the pseudo "ball". Maybe a hand calc of local ball loads might be better.
Let us know how you got on.

 

[GBor]

Not sure I agree with 4pipes on his methodology, but for this type of problem, I agree with him that you should contact Algor. If you aren't familiar with the acronym, "MES", then you are probably out of your depth for this level of analysis. Hertzian contact is very non-trivial.

If you are in Spain, there are a couple of really good power users in Great Britain that may be able to help you out. There are about 5 people in the US that I think are probably capable of this level of sophistication with Algor including Algor's tech support. Maybe one in South Africa. Beyond that small group, this is probably something that will frustrate you more than give you confidence in the results. Convergence is going to be VERY difficult.

 

[4Pipes]

Gbor,
.... and I do not agree because...?

My first to comments were key - "Contact Algor" and "I would guess".

My suggestion was to aimed at generating discussion.

Why MES? The problem is a moving router table. I would have expected the loads causing overall deflection would be more or less static.

My guess was based on the logic that inclusion of the balls in the full model would be far too complicated for the reason you have stated. (My wording for step "d" is not very clear - 4 rigid element would be required to build a ball with 4 contact points).

fnm

Is the orginal problem to determine the deflection of the overall table (or bearing assembly) when subject to reaction loads applied to the router cutter plus dead weight of wood and table? i.e. What cutting tolerances will be achieved for a particular cut on a particular size and type of wood?

 

[GBor]

4pipes,

My first statements were also key "not sure...". I snapped off a quick response, but wanted time to think about your proposed methodology. I think answers to your last questions are very important.

The rigid elements, if I'm envisioning them correctly per your methodology, may artificially stiffen an area of interest and could lead to underestimating displacement and overestimating stress at the tips of the rigid members. They are also bi-directional where the contact would be uni-directional.

I'm also not sure I understand their purpose. It looks like your intent is to construct a rigid frame representing the ball, but then I'm not sure I follow what boundary conditions you would supply for this.

I think the logic is reasonable, I'm just not sure I believe the methodology would produce accurate results.

An option may be to model the bearing and check to see, based on a unit load applied at each of the contact points, what type of stiffness you can expect from the bearing. Instead of using rigid element, use springs of the appropriate stiffness.

What type of analysis are you doing, fnm?

 

[fnm]

Thank you 4pipes for your hints ...but before I explain what I pretend to do I want to tell a little bit about my background. I'm an electronic engineer and I have been working with Algor software in the past during more than 10 years for measuring microstrains applied to load cell designs. Period 1983- 1996 During those times element types were only bricks and meshing was by hand drawing. I never used gap elements and no MES during those times¡¡¡. I stopped this activity and also upgrading my Algor soft... so I lost contact with FEA analysis until now.
Today I'm designing a CNC Router as a hobby and I want to predict the displacement of its gantry as a function of applied forces at the tool and thickness of supporting plates. As a first step I did and analysed a model and in order to simplify it I substituted the linear bearings by single rigid elements to transmit forces and moments and evaluate displacements. Now the second step is to check the accuracy of my model, so I want to find a simple model equivalent to the linear bearings. As you can see my question is very simple: how I implement a model of a ball against a groove surface.

 

[4Pipes]

I would suspect that the local deflection of a ball and raceway will be negligible in comparison to the other deflections. It might be worth getting the info from the bearing supplier. Modelling the bearing elements as blocks might be the simplest and most accurate solution. Modify the E value / size of the the block until it is as stiff as specified by the manufacturer. Trying to account for individual balls requires details of the bearing that the manufacturer is very unlikely to provide.

I assume from your dxf import file that you are creating your model in Fempro. It might be possible but drawing in Fempro is fairly basic. I would suggest using a solid modelling application. Download a copy of Alibre Xpress. This is the free version and has a few limiting features but should be good enough.
Modelling hemisphere and then cutting both symmetrical planes to produce a 1/8 ball, a section of raceway and putting both in an assembly for export into Algor should then be pretty simple.
I have the full version of Alibre which has Algor listed as an add-on. If not, create a step file and open from Algor. Somebody else might advise on interfacing Alibre Xpress.

 

[Stringmaker]

Do you really want to use a FE code to do this? If you want to get meaningful results you'll need to have an extremely fine mesh because contact area is so small for rolling element bearings.

We do pretty high tech bearings and I know bearing guys at our company typically use an analytical approach for this analysis.