Hi, I was wondering if you can provide insight as to what shell element formulations are recommended for different situations. There are so many element formulations the LS-PrePost provides, but I have no clue on how to differentiate between them. I have tried reading the theory manual, but it is too complex and I have found no insight as to why someone might choose one element formulation over another. Do you have any recommendations as to where to read about them? Or if you know about them, would you mind explaining them? Thank you.
Great question! So first off, using the defaults in LS-DYNA is usually an okay place to start, in this case, it is element formulations (ELFORM) 2. This can be used for something structural like a tube or 2D wall ect. It's also one of the fastest formulations. However, it can hourglass, which is a complicated topic in itself. The simple explanation is that the nodal position of the elements deform without stress being registered, causing the elements to literally look like a bunch of hourglasses. Like turning a square into a trapezoid. This can cause instabilities during simulations. A way to counteract this in LS-Dyna is to give the Part an *Hourglass card. This is a completely different topic so perhaps we can start a new topic on hourglassing to get fully into it, but just use an hourglass with IHQ=3 for soft materials and IHQ=5 for stiff materials for now and leave the defaults.
A way to avoid hourglassing is to use fully integrated elements ELFORM 16. This is usually my choice if I have a shell part that requires a certain degree of accuracy. Fully integrated elements do not hourglass (no hourglass is needed for the part) but take longer to simulate.
Another situation would be if you were to model something without bending stiffness like a blanket, you wouldn't want to use either of the previously mentioned ELFORMS as the part would not fold like you would expect a blanket to do. For this, you can use membrane elements: ELFORM 5 for the membrane version of 2, or ELFORM 9 which is fully integrated membrane elements.
Also, keep in mind that different types of shell elements will require different ELFORMS. But this shouldn't be an issue as LS-DYNA will automatically change the hourglass of things like Triangular (Tria) elements to an appropriate ELFORM if given something like ELFORM 2.
Also note that tria and tetrahedral elements cannot hourglass.
I hope this helps.
There is another thread on 3D element formulations here.
Hi, this was definitely very helpful. I have been constantly struggling to find a source where I can find an explanation and purpose to the shell ELFORMs, and I am glad to have a reliable source of information! I am in the process of learning how to use LS-DYNA, and I will definitely be visiting this site frequently.
I was actually hoping you can make a video on your YouTube channel going in depth with the things you know about how to apply different shell ELFORMs and maybe some examples.
Thanks!
Glad I could help. Yeah I’m actually making a few videos right now and I’ll add one on ELFORMS too. Feel free to ask any other questions, just make a new topic if it’s unrelated.
The fully integrated element isn't always the best and most accurate option. In explicit analysis elements with reduced integration (with hourglassing control) will be more accurate because of the so-called shear locking phenomenon. The deformation in outcomes could be significantly smaller than in reality. Less integration points better results 🙂
It's true that fully integrated isn't always the best option and can add stiffness, however less integration points does not always mean "better" results. Reduced integration is simpler and almost always preferred in solid elements, but fully integrated shells have many benefits. The biggest for me being that you don't need to mess with hourglass controls, which can have an enormous effect on material stiffness. The stiffness of a soft material can literally double depending on the hourglass formulation/qm value. There's a lot of "tuning" with hourglass coefficients.
Really good points though! I over-simplified when I said that fully integrated is always more accurate.
You mentioned how ESORT works for 3D elements, but how to do work for shell elements?