Hello guys. I am currently trying to model unloading (springback) after bending. I tried making the stamp go back in the second step and modeling the change to remove the stamp, but they are not working. With the stamp going back, it calculates for one increment and then breaks. With the model change, I remove the stamp and contact pair, and then the workpiece starts moving somewhere:
You don’t have to remove the stamp, just the contact pair so that the workpiece and the stamp don’t “see” each other.
You should provide some minimum constraints to the workpiece so that it doesn’t fly away. Ideally symmetry plus whatever else is needed to fix the rigid body motion (well-defined symmetric model may need fixing just one node in one direction then). The other boundary conditions can be removed with OP=NEW.
Thank you for the reply. I added a rigid body to the stamp and matrix, and also added a constraint on the workpiece on the Y axis. It didn’t help. Can you explain more on how that’s done? Or maybe you have an example of that?
You should constrain the workpiece in all directions (to prevent all RBMs), but without overconstraining it. You could use the 3-2-1 method, but the easiest way is to utilize symmetry.
There are some examples in Abaqus documentation. They use import technique (transfer between implicit and explicit solver), but you could check how they constrain the model utilizing symmetry (their models are typically axisymmetric, though).
Thank you for the suggestions. I tried this 3-2-1 method on a simpler model in Prepomax. As I understood, I fixed 3 points on 1 plane in the X axis, 2 on the second plane in the Y axis, and 1 on the third in the Z axis. On the second step, the workpiece just explodes. With symmetry, the same thing happens. I fixed one surface on the X axis, the second on the Y, and one point on the Z. Same thing, the workpiece explodes.
I tried automatic incrementation instead of a direct one, and it worked. Although it is a bit strange (the left side is in the air and didn’t touch the matrix), it springs:
… and use a couple of elements across the thickness like @Calc_em 's model. Also, I would recommend just having rigid bodies for the punch and the holder, sometimes that makes it easier to converge.
Especially since I used C3D8R elements, so too few layers = hourglassing in bending.
Right, I have that in the example I shared as well. I colored them gray because the default yellow blended with the contact pair definitions. But their lines and ref points are also shown.
Thanks for the advice. Can you share how you made that symmetric model? Did you use PrePoMax built-in functions? I made a quarter of the original model in KOMPAS-3D and used it in a more complex model with the test subject being split into grains, but it is slower than it was. I also tried using gravity on the test subject, but one side is still floating in the air:
Can you share a picture/simple sketch of the shape you want to obtain ? What you got in the screenshots on the right seems rather reasonable for the assumed process conditions (stamp and die shape, no blank holder and so on). Keep in mind that correct friction coefficient is very important in such metal forming processes.
Basically, I want to model simple 3-point bending and see how different texture compositions would affect springing. The shape at the end should look like a V.
Is that set up possible?. There is a RB definition beside a master surface definition. That is not allowed in Calculix. Do that model work in prepomax?
It seems to be a V shape already, just rounded due to the shape of the stamp. Unless you mean the asymmetry, but I covered it above.
Yes, it works. There could be some issues due to how CalculiX handles rigid bodies, but I don’t recall the documentation explicitly prohibiting this approach or otherwise mentioning such a limitation. So I use the standard way of defining rigid-deformable contact in Abaqus where the rigid surface has to be assigned as the main (master) surface.
Edge effects could do that. You can eliminate such 3D effects by using a 2D model or constraining displacements in the plane of the bending motion for all nodes. Whether that is the situation you may judge from the the size and distribution of the little bit.
Edit to add: And should this be the kind of saddeling you see, it actually indicates a realistic behaviour as in a real life bending you would also see the cross section in the bottom of the V would be trapezoid rather than rectangular, with the inside width (where it contacts the tool) will be wider than the width on the outside.
. Thank you very much
