When to define contact ?

[Student BE]

Hi,

I would like to ask when contact should be defined in structural mechanics analyses ? Could you explain it as for a beginner ?

I've heard that every assembly model needs contacts applied but I think that it is possible to avoid them in many cases. I understand that when contact between 2 faces appears during simulation (initially faces are separated) then it should be defined properly because otherwise there could be wrong penetration/collision result (let's say we are modelling hardness testing or ball being dropped on something) and we don't know the area of contact.

But what about situation when faces are already touching each other with their whole area at the beginning ? For example we can imagine tower consisting of a few blocks put on top of each other and we simulate compression or tension treating it as fully welded/glued construction with no chance of separation between blocks. Can contacts be avoided in such case ?

I've also heard that contacts should be used mostly when there is motion in our model (for example gearbox). Is it true ?


Thanks in advance for your reply

 

[goeasyon]

Generally, we classify the contact into the bonded, frictionless, and frictional. If the contacted bodies are initially attached and won't change position to each other, you could apply the bonded contact, for example using MPC algorithm. If the contact can separate or slip as the motion goes on, use frictionless or frictional.

The computational speed order (from fast to slow): bonded > frictionless > frictional

https://welsim.com

 

[klaus.]

well you as an engineer have to decide what informations ( results) you expect from a FE calculations
According to this you need to do the modeling and input so the output is reasonable.

Basically the better output you want the more input ( modelling e.g. contact) you need to do

There is no general rule for this........

 

[Student BE]

Thank you very much for fast reply.

Could you tell me which contact should be applied when bodies are initially separated and one hits another (bullet and plate or ball and floor etc.) ?

What about threaded connection ? Should contact be applied to all touching surfaces here ? For example here it's automatic bonded contact everywhere:

https://www.youtube.com/watch?v=p1GcedxmnlI

Is it enough ?

And one more example: if we want to simulate spring being subjected to compression and we know that coils will touch each other st some point of simulation then contact should be applied ? If yes then which one ?

 

[NecroSPEL]

Hi Student BE,

For sure you shouldn't define bonded or tie contact definition between parts which move between each other. You can use friction or frictionless contact (it depends on your simulation). Algorithm check it in all iteration if parts are in contact then contact will be work (Let say in a very simplified form ) Regarding that thread connection, I'm not really sure what mean "automatic bonded contact" here. Maybe is it something like "general contact" (all parts in model are in contact) If yes, it's enough Same situation in spring case.

BR,
NecroSPEL

 

[Student BE]

In that YouTube example of threaded connection modeling, which I am referring to, there is default setting that all parts that have their surfaces closer than 0.004in from each other are automatically modelled with bonded contact.

So in case of spring with coils separated at the beginning of simulation bonded contact is not a good idea and friction/frictionless should be used ?
And in case of threaded connection bonded contact should be enough when we assume that thread is stuck in place ?

And last question - do you know anything about Augmented Lagrangian and Penalty methods ? These types of contact are available in COMSOL from what I’ve seen on my university.

 

[NecroSPEL]

1) If you decide to use bonded contact between spring coils, then after you compress the spring it's can't go back to beginning state (coils just "stuck" together).
2) Yes
3) Lagrangian and Penalty method it's very width topic, generally penalty method is always use in symmetric or self contact when you use a lot of parts in the same contact zone. Lagrangian is used in 2D models and when obtain big material non-linearity. But of course it's always DEPENDS WHAT YOU SIMULATE!

BR,
NecroSPEL

 

[Student BE]

Do you know which type (penalty or Lagrangian) should be used in case of FSI model of closing heart valve where leaflets touch each other due to blood flow ?
And which type (penalty or Lagrangian) to use when simulating ball hitting or pressing against flat surface (second one is like in hardness testing) ?

 

[Dave442]

whichever one works best for your specific model/scenario.

 

[Student BE]

Hi,

I would like to ask you one more question concering contacts. I've seen some YouTube tutorials on how to perform impact analysis (explicit dynamics) of car or bullet hitting obstacle. When car was already modelled as touching (connected with) obstacle then bonded contact was used. But when car was separated from obstacle and applied velocity was causing him to hit the wall during simulation then no contact was defined. Like here:

https://youtu.be/-cu__bX3pg8?t=8m26s

So is contact definitione unneecessary in such case ? I thought that contact must be defined whenever 2 parts touch each other, even if they are separated from the beginning. Otherwise I thnik that objects should pass each other like ghost would walk through wall. My theory is that if objects deform so much (like in case of car or bullet impact)in explicit dynamics simulation then it's not treated like regular static contact case in which faces interact (generate stress/pressure) but don't penetrate each other physically.

Thanks in advance for your help

 

[Mustaine3]

Explicit codes (and also some implicit codes like Abaqus/Standard) have a General Contact algorithm. That means that the solver recognized when and where regions penetrate each other and prevents this. You can't define contact pairs in a real crash analysis. It would need thousands of definitions.
This also includes the contact between car and obstacle. Using bonded here is wrong in a real crash analysis.

 

[Student BE]

Now everything's clear, thanks