[G-Spot]

Hi Spot - They are the same order of magnitude and I'd have to understand how you restrained the model and loaded the model, meshed and connected the models to figure out the difference. 2.631/3.11= 0.85 There are many ways to validate an FE model and this is done at commercial level for many projects. I have no problem chasing bilbies down many rabbit holes for the fun of it... spent a lot of time doing that... Mental gymnastics are sometimes and mostly skewed and they need validation as well. Peter

I think it could be that the OD of a 40mmNB pipe with 5mm thick wall thickness is not 45mm OD as you used in your calculations.

[peteeng]

Hi Spot - to illustrate the point. I modelled three steel tubes in fusion. The tube is OD45 and ID 35. The rod has the same area so is 28.3mm dia. One tube consists of three pieces so in the simulation it has bonded connections at two places. They are 999mm long. Manual calc says they stretch 7.56um at 1000N. The FE says the rod stretches 5.63um the tube 6.48um and the bonded tube 7.13um. I used the default mesh settings. They should all be the same but are not. So models have to be understood to correctly interpret the results. So I'd have to dig into the mesh and connections to establish why they are different. Peter

[peteeng]

Morning All - This morning I refined the mesh, used "offset bonds" vs bonded and the result is much better. I also made the cross areas much closer. So now I would be happy to progress to a more detailed model if this was a subset of the actual model. Peter

[peteeng]

Morning All - over coffeee I had a further play with the tube model. I made the mesh finer and used an adaptive mesh setting. Plus I set the connections at bonded. This took more time to solve (due to adaptive setting) but each member stretches the same and its 7.56um which is close to the manually calculated stretch of 7.57um. Looking at the adapted mesh it has placed 3 or 4 elements across the thickness, this is what I would have aimed at if manually specing the mesh. Having one element across a thickness is poor practice and the stresses will be inaccurate. In this case stress is not important but it clearly influenced the stretch somehow. All good. Next would be to check a bonded connection in bending to establish that's its OK as well, then we could move onto the real model knowing outcomes are correct... Peter

so next step to speed up the solution is to spec a finer mesh but not use adaptive.

I have not played enough with Fusion that's why I have done this. I mainly use a system called simsolid as most of my current work involves large structures with lots of bolts and SS does bolted connections and extracts bolt loads very well. Fusion has bolted connections but does not extract the bolt load (maybe it does haven't dug that deep yet) I am confident that a good model in Fusion is accurate as this study has shown.... Since Fusion has a ansys and a simsolid port I think its running ansys by default...

[G-Spot]

I get the following results with my default mesh.

Tried a lot of different custom meshes the results always came out the same, got me thinking about what steel you used?

[G-Spot]

Here's one where one of the tubes has 3 divisions through its thickness.

I just manually calculated the stretch to be 7.94095 um using the structural steel with a modulus of 2e11Pa and a diameter of 28.3mm with a length of 999mm

[G-Spot]

So I'd have to dig into the mesh and connections to establish why they are different. Peter

Or you could just upgrade to ANSYS :P

[peteeng]

Hi Spot - My steel uses E=210Gpa that's the only difference. I agree that the mesh density for the purposes of deflections should not be critical. But once I understand what Fusion is doing then I'm happy. I use three FE systems Strand7, simsolid and fusion... each has its strengths and weaknesses. For confident modelling a building block process is used. Make small simple models that can be checked as you /we have done then build up to complex models knowing the blocks are good. Peter

[G-Spot]

This one is for the MESH POLICE....lol

This is giving 50N/um on the X-axis with bang on 1000kg of steel but 100N/um plus on the other axis.

Finally can make quick changes to whole sections of pipe and see the results within 90 seconds.

This was with mesh settings that take about 3 minutes, that's too slow for me, so I use the lower settings for investigation and the higher settings for conformation.

[G-Spot]

I used 5mm thick pipe 50mm OD in the highlighted sections and 2.5mm thick everywhere else.

[G-Spot]

Without these animations I wouldn't be able to optimise as easily.
They show you where the weakness really is.


OMG the 2000's called they want their video posting software back.

[G-Spot]

This is with adaptive turned off and with capture curvature and proximity turned on, which increases the mesh where 2 elements connect.

I've noticed the better your meshing gets the less the deflection in your results, which is a good thing, you are more likely to end up with a stiffer than expected frame.

I've also noticed that I get no error messages at all in my results, that happened after I managed to get rid of all the orphaned connections.

There's 517 individual pieces of pipe in this and over 3000 connections between the beams, if you get one connection wrong it gives errors.

If you use the automatic fix feature it uses a midpoint between the bad connection so your whole frame ends up crooked, even though the mesh is fixed.

[G-Spot]

Changed 2 pipes with the rails on for 100mm OD with 10mm wall thickness and almost doubled the stiffness.

This frame shape performs best, if I remove any of the outer pipes then there is a noticeable drop in performance.

The inner pipes don't really change the stiffness by much if you just change one or two of them, it's better to just size all of them up.

Did a simulation on the whole frame with 1mm mesh spacing.

[G-Spot]

OK I removed all the ugly cross bracing I added and left it as built solely from equilateral pyramids of 500mm.

What matters is mounting your 2 sets of rails on 100mm OD 10mm thick pipe the other pipes can all be 50mm OD with only 2.5mm wall thickness.

This things looks amazing when you rotate it, check the 45 degree side view.

[G-Spot]

So I am going to use my pyramid space frame design as rebar surrounded by concrete, good old bridge cement.

Taking the advice of the experts, see attatched pics....lol

Also if anyone is thinking of putting any kind of fibers in their cement they need to watch this.

https://www.youtube.com/watch?v=qNej42HAuXk&t=624s

I have seen dozens of others videos with the same result, cement composites with fibers in them are a lot weaker than cement without them in.

You would think the Chopped Mat scandal back in the 90s or whenever it was would have made people wary of this fibers in cement story. The thinking then was exactly the same, the experts said you could just chop up the fibers and add them to the mix and just spray boats into existence. Turned out that just adding chopped fibers to the composite weakened it and did not add anything other than cosmetically appearing better.

Fibers in your cement, great for paving etc where you don't want surface cracking to make it look so worn.

Look at the test in the video with the fiber cement, even though it is broken it still looks OK....ish.....lol.

If anyone wants me to, I can go find the video where they test that new carbon fiber impregnated 3D printer filament where it fails before the standard ABS does, the guy doing the testing also looked as surprised as the Hydraulic press guys did.

I have been looking for a single video where these fiber cements have been shown to be stronger than regular cement, it's just another Chopped-mat engineering scandal.

And look at the strongest and most rigid concrete they tested, Bridge cement they called it, but it is the same cement I have been using for the last 10 years. Check how it exploded compared to how the fiber cement just failed like a damp sponge. The more violent the explosion the more rigid the concrete was, the more the press got stretched and not the concrete squashed.

Also see how adding chopped steel fibers made it weaker.

When I was still at school, we were warned never to allow fibers to get in the cement mix as it would weaken it, seems like they stopped teaching that at schools.

[G-Spot]

Going to get my hands on some of this Sika Intraplast-100 and add it to the cement I use and see how it works in some pipe. It's the same stuff that they used to make those concrete dampers in that science paper I posted. The prof says it is just aluminum powder, which when it oxidises gives of hydrogen which cause the concrete to expand and not shrink, I might just try removing the oxide layer from aluminum powder myself and try it. Depends on the cost.

The whole point to using concrete in a machine design is because of how cheap it is compared to steel, if you now make it expensive with expensive additives then there is no point. Anyway I wont be using any additives at all when I make my frame as the concrete will be on the outside.

But I am going to build my machine the traditional way, with the concrete outside, will use the space frame as my rebar plan of steel inside the concrete.

Also going to build my machine to the original dimensions, well a bit less, 12m instead of 15m. Wont be going up vertically will keep it at 12m x 6m x 2m.

I have little use for a machine that works on 1m x 1m x 500mm. Plenty of stuff and jobs that need at least 4m x 2m x 1.5m.

Also it seems obvious to me after all my FEA work that there is a minimum amount of steel you need to make a commercial grade machine, and its about 5000Kg for 1m x 1m x 500mm. Any less and you will be breaking the laws of physics.

Also it seems from my calculations that anyone thinking of doing the same would be better off spending the money needed for 5000Kg of steel on a laser instead, which is about the same price.

So I am going to build a machine outside with concrete and rebar, and I wont be using any of those extortionately priced rail or screw systems. Which would have been poor at maintaining overall accuracy at that scale anyway. Will purchase a milling machine and lathe for less than the bearings and screws would have cost me and make my own bearing system and linear motors for less.

The best linear motors are the most simplest to make, even on my worse day I can produce something that is more accurate over 10m than any of these ballscrews could ever be. Already made them and have all the wire and formers etc.

[peteeng]

Hi Spot - Bambergs work is very good as a general guide but IMO not good for some of the details. Being a thesis and ending in a machine being built it seems to me some decisions where made that went against the solutions found. I'd recommend not to use aluminum as a shrinkage compensator. The results lead to accelerated corrosion of the steel. Engineering grouts are already made double compensated so why experiment? Just buy a CSA based class C grout and all the research has been done for you. These do not use al as a compensator... that's a very old approach.

Machine elements are generally in bending and fibre reinforced concrete is clearly stronger than non-reinforced concrete in bending. But high strength is rarely required in machine frames due to the stiffness required. Compression tests are a poor test and have several issues to check before you can rely on them. Keep at it, the Grail is out there... Peter

[peteeng]

Hi Spot - What is the point of burying a steel spaceframe in a concrete casing? One of the structures will be dominate. The steel being 200Gpa will provide a preferred loadpath and the 30GPa concrete just adds mass. Or the concrete having the greater inertia (being outside) will be dominate and the steel will be doing naught. I doubt they will co-operate in a useful way. You have a philosophical issue to resolve here me thinks. Peter