Modeling soft contact with parabolic distribution in Abaqus

[Tinni1]

Hello,

I am simulating a compression test on a cold-formed steel-lipped channel column (stud). The column is connected with a small length of a U section (track) at the top and bottom.

There is a gap between the stud and the track, and I have modeled the gap.

Initially, I defined the normal contact condition between the stud and the track as hard contact, but the analytical curve's slope was much stiffer than the test curve. I have tried refining the mesh, making the boundary condition more accurate, using an accurate material model derived from coupon test data. But the slope is not reducing.

Now, I am trying to model a soft contact formulation.

First, I tried the linear relationship by defining the contact stiffness k as a percentage of the base metals' Youngs's modulus value. Thus the slope of the analytical curve reduced a little.

Now, I want to try the parabolic relation of the soft contact formulation.

For this, we need to define two parameters:
1. Distance at the zero pressure.
2. Contact pressure at zero distance.

My queries are:


1. From my understanding, the distance at zero pressure could be the gap between the stud and the track. Is it correct?
2. How much contact pressure value should I put on the pressure at zero distance? Is there any thumb rule to guess the initial magnitude?

Could you please provide any guidance on this?

Many thanks

 

[FEA way]

The description and figure in the documentation chapter "Contact Pressure-Overclosure Relationships" (paragraph "Softened Contact Defined with an Exponential Law") can be helpful here.

Basically, clearance at zero pressure c_0 is when the surfaces start transmitting contact pressure between them. This way you can make the surfaces transmit pressure even before they actually touch. Then you also have to specify the pressure that will occur at zero clearance (when the surfaces are touching with no separation or penetration). It's hard to advise any general values, it's highly case-dependent and just needs testing in a given scenario.

 

[Tinni1]

Thank you so much for your response!