Common node method of building from parts.

MikeMcMullen

I am starting to put rebar into my sometimes large concrete models. I tried entering the rebar as round truss elements, but the nodes did not seem to merge. Also larger networks of these truss elements were hard to build because the truss elements seemed to want to be manually placed some of the time rather than snapping to existing nodes. I manually mesh in Mecway as these models are very regular and suited to Hex 8 or Hex 20. Also difficulty with the node snapping into a different plane. Will there be a difficulty, if I get it it correctly assembled, running in CCX, will the conversion to solid elements mess up or increase the size of the mesh a lot? Will the long length to width ration of such elements cause a difficulty? Would there be a problem with square rebar connected to remaining materials by only one corner? There should be very little bending in these elements, but yielding behavior and concrete compression only behavior is needed.

For possible future inclusion, could the truss elements be also included in expanded solid element option form for easier switching back and forth between the internal solver and CCX?

Victor

I've heard of people modeling rebar in Mecway like this before, though with beams and possibly only linear.

If you were merging the nodes with Mesh tools -> Merge nearby nodes, it only acts on selected nodes (every member of the merge has to be selected) or the whole model if nothing is selected, so it could be that. I'm not sure why snapping to existing nodes didn't work. If you could give more details, it might turn out there's a bug. It does snap to whatever plane the nearest node (in screen space) is in which means if you have multiple nodes at the same screen location, it picks an arbitrary plane. Not great, sorry, but you could rotate the view slightly to separate them.

Expansion of beams to solids doesn't usually create too many nodes - just 4-6x per truss element and generally works fine connected to solids.

disla

Mecway also allows to extrude points into beams. That could be usefull to you.

MikeMcMullen

4 to 6 x added nodes does not sound bad. I may try again. My main issues were difficulties generating the truss members, the bigger problem being not snapping on the second point, but treating it as a manual point location. that when located in the meshed concrete solid the nodes would not merge even though they were in almost exactly the same location, and I provided a largish tolerance for the merge nodes. Note the truss members are a different material (steel than the concrete.)
I have not tried extruding into the solid, I have been working at refining my mesh to put matching holes for the rebars, I have seen a Mechway example for this a few years back. The mesh refinement needed generated about 10 times as many nodes.

Victor

If you're creating the truss elements using Mesh tools -> Create element or the toolbar icon for that, it should use the existing nodes you click on so there's no need to merge. Having the yellow circle appear around a node before you click it indicates that it'll be used instead of creating a new node. Is that not working correctly?

Merge nearby nodes doesn't consider the materials. I wonder if you could give more details on when it doesn't work so I can reproduce it?

MikeMcMullen

I will work on this a bit more. I think that issue may relate to the phrase "existing node". When I try to click on existing nodes, sometimes it selects an existing node, and sometimes creates a new one. I have no clue why one case and not the other. I had some luck creating a single truss element somewhere else with points created from numbers, then dividing it, then moving it into place, but where the nodes need to go is not always a uniform division. A uniform division may be good enough though ignoring some nodes.

MikeMcMullen

I think I thought out what the problem is. I think video card latency as the model gets larger causes the existing node to not be found quickly on the second pick for the line 2 element creation. I am running a Nvidia 1660 Ti. Used to use a Ryzen VII hoping to take advantage of the fast double precision speed, but so far for Calculix that is Cuda cards only so I gave that card to my grandson who can use it for AI using PyTorch. Currently trying your version 23 Beta. Was using version 22 when the problem showed up and had not used truss elements previously.

MikeMcMullen

Seems to be a difficulty keeping the common nodes common between Mecway and CCX with the forced rectangular element which has the nodes at the corners of the shape, and the Mecway truss members which have the nodes along the CL. I will need to change the rebars to two rectangles wide and use extruded shapes for my rebar.

Victor

Not sure what you mean there. The CCX truss elements do get expanded to rectangles with none of their nodes at the original end node locations, but CCX internally connects them with MPCs. You can also get the solution to show them as line elements rather than the expanded hex by requesting any of the section forces (axial force/etc.).

MikeMcMullen

It appears that the problem runs OK, though convergence is iffy. I got the stresses etc to show. Problem was that element and model edges obscured the color. The deflection of bars is still iffy not being very close to the concrete they are attached to or even adjacent bars, and somtimes significantly out of plane. CCX says there are 308600 nodes, Mecway says 12788. I think the conversion of the truss members in CCX is increasing the problem size too much.

Victor

Since the displacements of the truss elements aren't following the nodes they're attached to, that sounds like they're not really attached to those nodes. I'm not aware of a bug that causes them to not get connected by the solver. Check with View -> Open cracks in the modeler to see if they move away. Or rectangle-select a node to make sure it's really 1 node, not 2.

The numbers of nodes/etc. reported by CCX are called "estimated upper bounds" and it really means it. Those numbers are usually much higher than what actually ends up in the system matrix and the solution.

MikeMcMullen

I have been looking for cracks. Note stresses seem pretty regular, just deflections, and often unusual directions. Also having convergence issues, though sometimes reaches solution, though this may be expected with the compression only concrete model. None of the steel is getting into the non-linear stress range. I probably have other issues in the model I will correct. Convergence is somewhat sensitive to support stiffnesses. I will send you a copy of the input when I get things debugged till a single problem stands alone, and maybe if I get it fixed totally. It may end up being simply a need for quadratic elements.