matching time history output with experimental curve
Hi,
I want to carry out material (law 19 and 58) parameter fitting with Hyperstudy to match results of experimental four-point bend test.
The testing protocol consists of two stages:
-initial inflation to prescribed pressure and positioning under gravity
-four-point bend test (rigid body impactor measuring reaction force which is plotted as a function of midspan deflection).
Behaviors of inflatable fabric structures often involve coupled effects from inflation pressure such as fluid-structure interactions (FSI’s), thermomechanical coupling and nonlinear constitutive responses of the materials. Because of this it is nearly impossible to predict the position after initial stage: the time and deflection are unknown when the impactor initially comes into contact with the inflatable. I will use a similar approach as in MV-3000 & 3010, where time history output curve is optimized to match the targeted experimental curve which starts form 0 force @ 0 deflection.
The question is how to match the time history output with the experimental curve so the former also begins at 0 deflection?
<?xml version="1.0" encoding="UTF-8"?>
Answers
-
hello, @Ivan
could you explain simple what do you want?
my english isn't well, but maybe i can help you.
0 -
Do you want to compare test curve and computation curve in Hyperview?
0 -
is it what you want?
i did it manually.
0 -
Hi Mrt,
thanks for your interest.
Actually it is rather simple to compare curves in Hyperview; just load the data files from simulation & experiment and then offset/scale x or y components to match the curves.
What I am trying to achieve is DOE and optimization of material parameters to minimize the difference (minimize the sum of squared error) between the simulation and experimental curve using Hyperstudy. Please refer to tutorials MV-3000 and MV-3010 for more details. In order to do this the solver would have to automatically adjust the simulation curve so it starts at 0 deflection at 0 force. Attached bellow is the sample experimental curve that does not match the origin of simulation curve.
One possible solution would be to somehow utilize the interface sensor (which is already defined in the shared model) which triggers on initial impact the recording of time history output.
<?xml version="1.0" encoding="UTF-8"?>
0 -
ok.
but in my opinion, 6.5 s is so long for explct solver...
0 -
You are correct- in fact the experimental testing duration was 6 minutes and the simulation duration can be reduced to 1 s for computational efficiency while keeping the kinetic energy bellow 5-10% of internal energy to ensure a quasistatic simulation.
I have also tried to simulate inflatables in Radioss implicit and Optistruct, however explicit runs faster with AMS- it could also be I have not properly defined the implicit solution.
0 -
Hello @Ivan
i have tried to do it in Optistrucl, radioss implicit. in radioss implicit i have got nonrealistic forces.
if i have time i will finish it in optistruct. i can't do it realistic now.
0 -
if you can solve issue first, please, write it here.)
0
Altair Employee