[peteeng]

Hi - I was checking through the model to see what was moving and found the wall to base connection was only at the bolt heads. So I fixed that and the stiffness became very good. 18/16/38 vs 18/17/79 for the fully bonded model. This means the bolted connections are quite efficient. I'll look further into the Z delta... So this gives me more confidence in the clamshell idea... Peter

[StrawberryBoi]

I guess I'm not following what you mean by clamshell. Are you hinging this? Like a briefcase?

[peteeng]

Hi Strawb - No hinges just bolted together like a gearbox or pump housing... Not sure if it has a proper descriptor..., in boat building we call it shoebox construction but there are no bolts in the hull flanges.... Peter

https://en.wikipedia.org/wiki/Gear_housing

maybe a split housing, but it does not house anything. So a split base??

[StrawberryBoi]

GOTCHA!

Yeah, it would be called a split housing in a gearbox, so a split base does check the box for my automotive engineering brain.

So you're thinking to do a split iso base? Putting the iso faces together?

[peteeng]

Yes Strawb - Here is an image with one of the parts removed showing the internal grid. But the gearbox article describes a "bathtub" design that may suit better ie a deep isogrid for one side and a closing flat piece at the top or bottom. This means only one major set up for the deeper part vs two small parts. I'll have to cost each way to decide but it also means I don't have lots of exposed bolts on the top side of the base. In the z axis image the FE is exaggerating the crush of the bolt preload. I'm having trouble parametrically modelling isogrids and diagrids. They are a pain to do manually... If I use a quad grid thats easy as the CAD does ortho patterning... Peter

[StrawberryBoi]

Grids are really difficult to do parametrically unless the dimensional variation is small because the corner fillets can eat features with relatively small changes.

I'm not familiar with the CAD package you're using, but I would think to make either a single triangle and mirror it, then circular pattern, then linear grid pattern, OR make a 2 triangle corner square and then do the circular pattern and then grid pattern. Either way, it will get nasty exponentially fast due to the patterning.

You'd need to calculate everything parametrically pre-sketch dimension, then reference the calc results for the sketch and pattern numbers. I know I could do it reliably with Fusion 360 or Inventor, but I guarantee our old frenemy Solidwork would raise a stink pretty quickly (ie it would just crash like half the time).

[peteeng]

Hi Strawb - I have managed to semi auto the isogrid. I use alibre FYI. Using a thin extrusion module then a pattern vs a full feature. It patterns out well. But if I include a fillet in the pattern it does not want to play, which means adding the fillets at the end which is tedious... But for purposes of FE and development I can live without the fillets... May find a solution in the meantime. I found using "thin extrusion" you can have duplicate lines which is verbotten in a full feature sketch. This means the sketch can be patterned and the overlapping lines don't matter. So diagrids, hexgrids and isogrids are now possible... I do like hex-grids for aesthetics but they are not good structurally Onward to Milli HR-13 plus Peter

edit see hexgrid - a couple of global parameters to keep the 6th digits correct and I think I've cracked the grillage. Now I have to include holes and fillets in the module to make life easier. Peter

[peteeng]

Evening All - I have solved the parametric grillage problem for now, can do any grid I like but putting the holes in parametrically is a deal... The outer holes are on a different module to the inner holes.... perseverance usually works. What's the emoji for sweat? Peter

[peteeng]

Evening all - since I had a couple of grids figured I compared them for stiffness. Racking and transversely. I thinned the circular one till it weighed the same as the triangular one 14.5kgin aluminium. They both deflected very similarly in two digits. The 3rd and forth are a bit different. So overall same. Its the skins that matter in bending and shear (racking so will get that on the go. I'll work on a bathtub split design... Seems the holes will be put in last manually, trying to do it parametrically has done me in for the night...Peter

[StrawberryBoi]

Hmmm this isogrid thing has got me thinking a lot. I've used parametric hexagon structures open faced for lightweighting a motor end plate on a large axial gap motor made with PEEK for all structural components. It was ridiculously effective, but now I wonder if the isogrid would have been better.

I can only find use case of hexagons in a sandwich arrangement and yet the isogrid research and proponents only seem to off-hand say isogrid is better while hexagon sandwiches are more efficient. No comparison to open face hex, no apples to apples or even numbers for sections of identical mass.

Something I may play with on my own (don't spend time on my curiosity). If you do happen to have a reference that actually discusses strength density of an open face hex I'd love to see it though.

[ardenum]

Found a really cool leaflet with frame pictures and fea etc.
https://www.hasmak-t.com/cms-uploads...ginal-ver0.pdf

[peteeng]

Hi Ard - Interesting have a wide "gantry" on a narrow column.

Hi Strawb - There are two applications to consider 1) bending 2) in-plane loads. If a plate is in bending then it is a candidate for sandwich construction. The analogy is an I beam is more efficient in bending vs a solid beam in terms of specific stiffness. The flanges resist bending, the web keeps the flanges apart and transfers shear between the two. If the load is tensile (in-plane) the shape of the section does not matter its the cross sectional area that matters. This is the in-plane loading condition ie a rope will work or a column in compression.

If you get a piece of honeycomb (plastic or paper) and pull it in plane you will realise it has zero in-plane stiffness. If you have an iso-grid sample it has in plane stiffness as it has contiguous longitudinal elements that resist in-plane stretching/compression. Plus the triangle is inherently stable vs a hex which is geometrically unstable. Hex, quad, pent structures depend on the connection stiffness for global stiffness.

The hex or any other shape works as a sandwich core because its main function is transferring shear from face to face and holding the faces in-plane parallel. The shape of the core does not matter. In an isogrid the longitudinal elements are tuned to resist the in-plane or axial loads and the diagrid elements are tuned to resist the torsion loads. Up until lately the maths involved to optimise at this level has been tedious so the grillage has been averaged to solve these issues. There used to be a technique called smearing which averaged various properties to establish the grid sizing. But now the whole thing can be scripted in FE or isogeometric modelling and optimisation can be achieved relatively easily. I got into domes when I left school and did some work on aluminium extruded sections for domes. I was speaking to a friend last night and he wants top go to arizona next year to compete in the hand launched glider champs there. He's a world champion and builds gliders that weigh 2g! The dome is an isogrid timber structure made from laminated timber. A beautiful thing... There are domes that use hexs as their main shape but they use internal wires to stabilise the hex usually.

So to make a statement - hexs are not efficient structurally as they are geometrically unstable. If there is a plate both sides the plate is doing the work, if its one side be careful in what the hex is doing. If min weight is the target tris are probably the solution... Peter

This also bears out in non linear FE. If you use a triangular mesh (plate or tet solid) you will get mesh locking in non linear materials problems. If you use quads or hexs they have some room to give and allow the solver to increment properly....

[peteeng]

Morning all - By the way the hex shape comes from the method of manufacture of paper honey comb. Its made in strips with glue lines and when its expanded it forms hexs. Paper honeycomb was first made to use as air flow straighteners in wind tunnels I believe. Cheers Peter

B-CORE SLAB interesting sandwich construction for buildings ships etc

https://fb.watch/7LIXcBkadb/

NASA work on isogrids https://femci.gsfc.nasa.gov/isogrid/...rid_Design.pdf

[StrawberryBoi]

Thanks, excellent explanation and now I totally see the plethora of zero force members that make up an unsupported face in a hex. Really flipped that switch for me pete!

[peteeng]

Morning All - I am in a dilemma. I want to buy a 100mm slab of aluminium and make lots of swarf so I get a light efficient structure. But I know this will be many $$$ and this is not an aerospace product where swarf ratios are huge and costs are acceptable. The option is a plate build, minimum swarf but no isogrid... So I looked at a wall with a grid and just webs to see if it made a big difference. I loaded it in bending and in racking. I added an extra web to the LHS one so it was a similiar weight as the quad grid one. 54 vs 59kg The quad grid is slightly stiffer in racking but not in bending... The webbed version is a bit lighter I could put in another web.. My head says plate is the cheaper path... just more parts to organise and assemble.. Thats why I wanted to cast in the first place.. so onward to the next version... Peter

This is the same issue with intercostal beams in ships. Should they be connected to the main beams? or pass through them? Maybe I can egg crate the isogrid in plate...

[peteeng]

Afternoon All & Sundry - With my thoughts returning to plate I spoke to the laser cutters today about plate thicknesses. They can cut 16mm fast on one machine and 20mm slower on another. About 1.75 years ago I started developing a plate mill for a contact in New Zealand. After getting some freight quotes that where absolutely ridiculous I decided that I would find a machinist near him and send them the files to make the bits vs making it here and freighting it. Plus aluminium seems to be a bit cheaper in NZ for some reason... Anyways that didn't come off for various reasons. The interesting design feature was a split saddle to decrease the stack width of the z axis. I was figuring out the drive details when the project stopped. Its sort of where Milli has ended up. So I'll make some walls from 16mm plate an see how that stacks up. Having the plate laser cut oversize then CNC finishing it is an economical workflow... Peter

[peteeng]

Afternoon again - I looked at some reciprocal structure solutions but they didn't work out. This would have allowed bolting pieces into a pattern. But hollow structures always are lighter so the bathtub/shoebox approach is nearly always better. So I added a hollow "wall" to the prior model and even with no internal stiffeners it did better especially in torsion. Even with it not connected at top and bottom. So I think I'm heading back to a plate design... Peter

edit -taking the sandwich concept further I placed a 6mm SS plate on the back vs a 16mm al one. I made the walls the same thickness and size. The SS sandwich beat them and was same weight. So as long as the bolted connections are efficient this is something to work with... Peter

[peteeng]

Evening all - Been playing with plates, sandwiches and a saveloy. Its all too far away from the original concept of minimal pieces hogged out of something. Too many bolts connections and things to get right. So its back to billet parts & isogrids. Keep making - Peter

[StrawberryBoi]

I think the billet parts and isogrids seems to be your best overall approach. Given that you were working with aluminum, the swarf incurs a lower cost than it would with steel.

Setup time and fixturing has it's own cost as well, piece pricing at low volumes might be worse with the medium/high volume 'economical' workflows.

[mactec54]

Supplier quality management is a dedicated position in even fairly small companies for a reason, for sure.

I completely empathize with you struggles in getting consistent results. I also agree that unless it's a robotic welder, you will never truly have the same result even with the best human welder and the same one every time.

I've dealt with similar issues with CNC processes that were multi-vendor as well. Center less grinding, of all things! We had a lamination stack for a large servo motor that needed to be +/- 0.0005" per Y14.5-2009, so a profile tolerance on the diameter and not just the average dimension.

The grinder had the proper facility and equipment, but they didn't really take seriously that we could check this tolerance with our CMM in our Q/A department. They themselves could check it, but didn't because the machine was "dead nuts", per them. Theoretically, sure, but not when you don't keep on top of wheel wear, or chucking it with sufficient precision... particularly because this was a lamination stack and the OD and ID needed to be concentric to a tight requirement as well. The same vendor would first center less grind the OD and then ID grind on a different machine.

Wr were getting only 1 in 5 to spec and they refused to acknowledge the failure. Eventually we did a visit and made them show us the entire setup process. They didn't even dial the part in on the second setup!!!

Bingo, issue found. Thst supplier caused almost $60k in losses over just a few dozen parts, not to mention our customer in the semiconductor industry was delaying a multi-million dollar machine over a tens of thousands of dollars motor.

Even in house processes require constant vigilance and process development, so yeah, going a single machine process with a single vendor makes a LOT of sense.

Center less grinding is not that accurate never has been, if they could hold .0005" that was quite good, .0005" is a big number for some, as most new cnc mills and lathes can hold that kind of tolerance

Laser welding is the same all the time no robot needed