Reaction force measurement

[aba24]

Hi, I'm trying to model a force test. I want to extimate the reaction force resulting from a surface of 1mm width (made of rubber material) which is under pressure on a side and on the other one is encastred. What I want to model is a 1 second test. in this time, the pressure has to increase from 0bar to 1 bar and the force exerted from the opposite face (encastred) has to be measured. Anyone can help me setting this problem?
I have already performed a simulation, I have my geometry and I set as time of the step 1s, a load (pressure) of 0.1MPa on a face(which is 1 bar since I have all dimensions in mm scale), an amplitude, tabular, on the total step time 0--->0(first row)
1--->1 (second row). Anyway this simulation take long times and values are much higher than experimental data.
In the picture there is my geometry, a side view, just to help. i simulated the displacement of the top place, now i want fix the top face and get the force exerted by pressure
thank you

 

[rb1957]

force = p*A, no?

are you saying that the FEA is predicting very large displacements ? typical with plate theory.

why apply the load in steps ?

another day in paradise, or is paradise one day closer ?

 

[desertfox]

Make your file attachment a PDF I cannot read tif's on Ipad

 

[Tmoose]

"why apply the load in steps ?"

I think that may be Abaqus' standard.

 

[btrueblood]

What is your material (constituent) model, and does it accurately replicate tensile, shear, and compression tests of the rubber material? Are you using an element which is recommended for that material model?

Was the material model done after several load/unload cycles (typically this is done to try and reduce the Mullin's effect which is the softening of as-molded rubber due to those same load/unload cycles)? Do you test your sample after a similar number of load/unload cycles to similar strain ranges?

"What I want to model is a 1 second test"

It is likely that the rubber material model testing was done at much slower strain rates, with the result that the model predicts a much wider contact patch (larger deflections) and higher resultant forces than the test you are performing.

 

[aba24]

thank you all for the replies,

btrueblood, the material is Ecoflex0050 and I meshed it with hexaedrical elements. I acquired a load/unload cycle (10mm/min) data, then performed a material test on abaqus, the result is that Ogden model (N=3) is optimal. I te experimental test, I have the same structure I posted in the pic, I pull air at pressure of 1bar with compressor in it. The structure is in contact with a load cell to acquire force data (there is no gap between the top surface of my structure and the sensitive surface of the load cell, so there are no displacemets). with this set-up I have 6.5N force @1bar. I want to model the same (or similar) in Abaqus. you're right, maybe 1s of simulation is too small, and it is not my experimental case. I'll try to augment the simulation time. Anyway, if you'll have to simulate a pressure that starts from 0bar and reach 1bar in, let's say 10s for example, how do you will set the step, amplitude, and load? this could help me in understanding how to reason.
Thank you

 

[btrueblood]

Not sure the time simulation of the problem is appropriate, as you are looking for the static equilibrium value of the reaction force? As Tmoose says, though, perhaps the stepped load case is appropriate for modelling nonlinear materials in Abaqus (I have no experience with that solver). My point was more that: if your part never sees the deflection/strain that was developed in the material test, then it will be much stiffer than predicted by material testing at larger strains, due to Mullin's effect. Similarly, if you test a brand new, freshly molded part, without cycling it to the strain levels and number of cycles that were used in the material test, then it will also be stiffer than the model predicts. In the end, the equilibrium case you model will give you results for the material stiffness you picked (your constituent model), so pick the stress/strain data for your model from the load case you believe is closest to what you are actually testing. It's usually a bit of trial and error to get there.

How certain are you that the full reaction force is sensed by the load cell? Have you played with the contact boundary condition, with a "no slip" versus a frictionless contact boundary, and compared the resultant load from both cases?