Mesh convergence on Explicit dynamics
ive been testing trying to fine a suitable mesh size for the structure that i'm simulating however when i change the mesh size, the equivalent maximum stress tend to vary a lot, instead of decreasing linearly when mesh becomes coarse and increasing up till a limit when mesh becomes fine
What ive been doing so far is using parametize set to parametize the mesh while obtaining maximum equivalent stress within the structure
meanwhile my simulation is an impact simulation of an Arrowhead auxetic
Why is this so? is there a limit to how fine a mesh can be before it gives invalid results? what is the explanation behind it ? is there a way to fix this and how do i find the perfect mesh size in such a scenario?
Best Answer

peteroznewman Member
Data from Explicit Dynamics solutions are well known to be noisy. Stress, which is a derivative of displacements, is even more noisy. This is because the formulation of the equations is based on tracking pressure waves that travel at the speed of sound in the material and bounce around inside the structure.
I suggest you pick a fixed point on the model and plot the data at that point. This is better than the overall maximum, which can jump around to different locations as the mesh size changes
Copy the tabulated stress data at that point and paste it into a program that can low pass filter the data to remove the high frequency noise. Take the maximum value of the filtered data as the measure of stress at that point. Repeat this process for different element sizes.
Here is an example the unfiltered data on force and displacement from three simulations with a parameter change between them, solved in Explicit Dynamics.
Here is that data lowpass filtered and now the three curves are easily seen.
You should always filter data from Explicit Dynamics if it is noisy. Some output is inherently smoothed because it averages many elements, such as the system energy plots.
Answers
@djtan
Data from Explicit Dynamics solutions are well known to be noisy. Stress, which is a derivative of displacements, is even more noisy. This is because the formulation of the equations is based on tracking pressure waves that travel at the speed of sound in the material and bounce around inside the structure.
I suggest you pick a fixed point on the model and plot the data at that point. This is better than the overall maximum, which can jump around to different locations as the mesh size changes
Copy the tabulated stress data at that point and paste it into a program that can low pass filter the data to remove the high frequency noise. Take the maximum value of the filtered data as the measure of stress at that point. Repeat this process for different element sizes.
Here is an example the unfiltered data on force and displacement from three simulations with a parameter change between them, solved in Explicit Dynamics.
Here is that data lowpass filtered and now the three curves are easily seen.
You should always filter data from Explicit Dynamics if it is noisy. Some output is inherently smoothed because it averages many elements, such as the system energy plots.
thankyou ! Peter!!!!😁😁😁
@djtan
Here is a helpful paper on filtering Explicit Dynamics data to get smooth curves. Look in particular at section 2.3
https://bodietech.com/pdfs/DSP_for_Explicit_FEA_AUC1999.pdf
Thankyou @peteroznewman !