[peteeng]

Hi Spot - I would not use TIG on such a job either. MIG is the faster, less heat process. Speak to a welding supply company for parameters they will be up to speed with their equipment. Peter

[joeavaerage]

Hi,
you are talking MIG territory.

MIG is a constant voltage process. Even with 1.2mm solid wire typical voltages would be 20V-22V. The current is largely determined by how fast you feed the wire in there.

TIG is a constant current process. Say 150A may maintain a stable weld pool depending on the thermal properties of the material. The arc length then largely determines
the voltage. If you hold a short arc you might get a voltage as low as 16V but with a longer arc maybe more like 22V. The rate of wire feed will be largely determined by the rate of advance of the weld pool.

As I say using wire is really a MIG process and I suspect you'll have a lot more success if you hire a welder for the job rather than try to use a TIG.

Craig

[G-Spot]

Well it looks like static machining will soon be gone.

While getting my laser stuff ordered I see that you can purchase for $1000 a laser that can project a plane at under 750um thick out to 2m

You would not even bother to have to use the astronomy code to find the center. Could just be a simple yes or no, can you see the light, OK stop machining.

But still the lasers I am going to use are only $10 each and my Google cloud subscription to run the machine learning code is only $15 a month. I think I could get a lot better resolution than that using multiple cheap lasers and machine learning.

What would be the typical figures for misalignment of linear bearings and screws on a CNC machine?

This grinding machine build is perfect for dynamic machining where you just direct the grinding wheel to the location where grinding is needed based on instructions from a camera that is watching.

No need for expensive linear bearings or ground screws etc, your machining is referenced from a plane of light and not a metal surface.

I priced up a set of XYZ linear guides and screws for $700 that I can grind any size I beam up to 3m including the stepper motors and controllers. Going to delay the purchase until I have the laser guidance system working because I know there could be something about the laser projection and camera setup that might change the design of the grinder.

[peteeng]

Hi Spot - depends on what accuracy you want? 50um? sub 1um? all possible. In regard to the pipe dream of moving things accurately without rails etc just because you may "know" where something is does not mean its actually there or does not mean it can follow the commanded path. You are walking down the path of satellite and rocket control and they still have mishaps. Control of non constrained dynamic motion is pretty difficult..... Peter

The lasers used to set up good machines are accurate to >1um run at >0.01um. 750um is 2B pencil stuff...

[G-Spot]

Following Moore's law with my designs. Had a look at Space Frames came up with this metal crushing monster at 185 N/um with 2 tons of steel.

Definitely going to go this route, love the complexity, this is actually made up with 500 straight sections of 400mm long 85mm OD steam pipe.

Best thing is I used to make turbo manifolds and made a machine just to cut these multiple pipe merges, could cut them better than any CNC machines because that is all the machine was designed to do, my machine had to cut the pipe sections to any angle, this is the same angle throughout so this is easy peasy. I also can weld these up so you would swear they were machined from a solid billet.

Lots of experience working with 75mm steam pipe versus little experience working with huge steel beam sections of 300kg each. Each section of this is 20Kg.

Note that in the work area of this pyramid the deflections are in the 250N/um range. And the work area is more than 2m long.

[G-Spot]

The lasers used to set up good machines are accurate to >1um run at >0.01um. 750um is 2B pencil stuff...

Managed to get 2 Chinese lasers good for 60um line width instead of the 40um of the greedy Americans for a 10th of the price. I am sure with good cameras instead of that 2B pencil stuff industrial cameras I can get better accuracy than a lot of commercial machines.

Made a lot of discoveries these last few weeks. Industrial CNC is a scam.....lol

And these collector welds are TIG territory because most of the time just merging the existing metal is all that is needed without any filler material.

[G-Spot]

You are walking down the path of satellite and rocket control and they still have mishaps. Control of non constrained dynamic motion is pretty difficult

Had a great career in aerospace, I am sure it will help me immensely

[G-Spot]

Anyone thinking of following this route, let me save you a lot of time. You need to decide your base structure then know all its dimensions before you start off or you will never get it to pass FEA.

I made a text document with the following info so that my faces would align correctly and my FEA would not grind to a halt. If when the FEA does the analysis the results of comparing two points in space do not equate to zero then that is another thousand calculations to be done before it can continue.

Pyramid dimensions.

width and length = 400

Height of pyramid and distance to center from base nodes = 282.84271247461900976033774484194
x2 = 565.68542494923801952067548968388
x3 = 848.52813742385702928101323452582
x4 = 1,131.3708498984760390413509793678

Base interal Face angles = 60
Top internal Face angle = 60
Length of face = 346.41016151377545870548926830117
Distance of face to centre = 200
Angle of face from base = 54.735610317245345684622999669981

Angle of edge from base across centre = 45

[G-Spot]

Took my FEA game up another level, upgraded my memory to 64G and my CPU has 8 cores running at 5Ghz with 2 GPUs.

Looks like a frame that is 100N/um stiff with less than 1000kg of steel is possible.

This space frame optimized down to 40mm NB pipe with a 5mm wall thickness.

I know that the 35mm linear bearings I intend to use are only good for 50N/um so I might be able to get it down to less than 750Kg.

[peteeng]

Hi Spot - Make the same external shape in a suitable wall thickness plate that results in the same weight as the space frame and see if its the same stiffness... I was at a machinery exhibit today and was looking at 1.5kW laser welders for $14,000 AUD, seemed a good buy to me if I was going to do a lot of welding of any metal <3mm thick...Peter

also when doing a such a structure its sometimes better to boolean the truss into one solid vs having it have so many analytical connections (I see in some of your images the mesh is not contiguous). The mesh is cleaner and will solve faster and more accurately if its one body... Plus you mention autogenous welding, you have to be careful with that as when you just "meld" with no filler you can get hot and cold cracking in the weld.. We use filler for some good reasons....

https://australianmanufacturingweek.com.au/

[G-Spot]

Spent 2 weeks on software issues, switched over to beam modelling. OMG! if the topic itself wasn't such a mindf$$k then the bugs in the software will surely fry your brain.

But never the less I managed to sort out the multitude of issues and can do the simulations that were taking 1hr and 30 minutes in less than 90 seconds now. So now I can return to optimising my design.

This is my base design, a four sided equilateral pyramid which is the strongest most rigid shape you can use, especially if its suspending a tool that cuts at a point in space, the 3 sided pyramid is stronger but does not tessellate.

The attached images show that it's possible to build a 20 N/um machine using exhaust pipe...lol....this is 50mm OD pipe with a 2.5mm wall thickness.

I upped the size of my tube lengths from 400mm to 500mm to get me 1.5m x 1.5m x 500mm on my axis, and it only weighs 750 Kg.

The second image shows the axial loads on each beam so you can see which ones are in tension and which are in compression.

Should be able to make rapid progress now in getting to a buildable design.

[G-Spot]

Hi Spot - Make the same external shape in a suitable wall thickness plate that results in the same weight as the space frame and see if its the same stiffness..

Not too bothered about the actual shape, just looking for the stiffest design in the space available, I'm maxed out at a 3m x 3m base with a 2.5m height
Now that I have the software and memory issues sorted out and can do the simulations in seconds again instead of the hours it had started to take since switching to a Space Frame I am going to find out. But it's looking real bad for the monocoque on what I have seen so far.

That is still too much money for a laser, especially when EDM is a 1/3 of the price. Anyway whatever I use to machine with, it has to be directed and moved in space. So onwards with my frame design.

The way I weld up pipe is not really autogeneous but I hardly use any filler just enough to keep the puddle flowing, and turbo manifolds crack if your welding is suspect, so I wont be worried about cracks.

[peteeng]

Hi Spot - Its your connections that are driving the solution time up not the mesh. If the mesh was contiguous it would solve in seconds. Plus with a space frame such as you have there is no value in using pipe. The ligaments require axial stiffness not bending stiffness so your 40NBx5m is Dia45x5mm so has a cross area of 628mm2. This is the same as a Dia28mm rod so use that. Easier to deal with in building... Run the same geometry as a beam rod 28mm and you will see....Peter

in your image attached I can see your mesh connections don't match and you are using polynomial elements. Use linear elements (more of them to capture the curvature) and correctly grouped solids and the mesher will make the nodes at the connections agree. Then it solves in seconds...

[G-Spot]

Hi Spot - Its your connections that are driving the solution time up not the mesh. ].

No kidding...lol.....the problem is a development one, 517 pieces of pipe to mesh takes more time to set up as a single contiguous mesh than the hour and a half to solve using multiple connections. I probably would have settled with the 90 minute wait compared to the error inducing nightmare of reconnecting a large mesh.

The mission is to be able to make quick changes to the design, that is impossible to do using the method you suggested, I tried, try removing a single section and replacing with a larger diameter section or different profile and you have a days worth of work ahead of you. I thought the previous designs with 30 or so elements was a bit of a mind bender to realign the mesh after changes but its an almost impossible task on hundreds of elements.

The only reason I didn't give up was because I knew that all those SpaceFrames out there could not have been designed from changing the mesh each time. Did a stupid amount of research on it and from the makers of the software themselves, there are basically 2 approaches, messing about with meshes on a small numbers of elements, or using a connection based approach with a large amount of elements.

I will try seeing what the result is of replacing with solid bar will be, but I think it will make it a lot worse as your weld connection area will be smaller and then your angles and places of attachment will allow for more rotation than with a larger bore pipe. If this was the case, why don't they make racing cars out of solid steel bar and not tube ? If it is a space frame connected with pieces with small golf ball type joints it will be true, if it is a space frame welded together with large profile pieces, where they will be welded far from the centre of the geometric joint it wont be true.

Then you have to consider the elements themselves, this is not like a motor car or airplane with hundreds of unique components, they are all exact copies of each other the only difference is their placement and angles etc, it is an absolute nightmare to edit anything, if you look away from the screen to check a measurement, when you look back you have lost your frame of reference.

Then there are some crazy bugs which cause multiple connections, the level of detail you have to maintain to avoid errors is stupid, stuff like working out how many possible connections each cell must have with the next, then going and checking about 100 different directories with a different calculation to see whether the correct number of elements is found.

And then the "automatic tools" don't work in a way that helps you, if I manually check for errors, when I find an error I can correct it, with the automatic tool it gives you a poorly described list that takes much longer to sort through than checking yourself.

[peteeng]

Hi Spot - What is your tool chain? CAD? - FE? - all of the issues you describe are sorted by the good CAD and Fe suppliers. If your using beam elements then its very easy. The difficulty is only in the CAD side. The way I would do the geometry is make a pyramid of required module size, pattern it out to the size needed, then delete the ligaments that you don't want. Then boolean the entire thing together and the mesher will do the rest. Peter

You can make a solid block that represents the middle void and subtract that from the spaceframe, then delete the little bits left sticking out... if you use something like grasshopper you can write a script that will produce the shapes you want. Thats how they do complex buildings.

There is a japanese company that makes large buildings from space frames. They use a system that has a steel spherical node so that all the tubes are cut square. So the fab is simple. The node to node distance is always known. Car roll cages are under high torsion loads because the spaceframe is not a pyramidal assembly. Although its a spaceframe its not like yours... Zaha Hadid architects are wizards at doing extremely complex very large spaceframe buildings...

[G-Spot]

]all of the issues you describe are sorted by the good CAD and Fe suppliers.]

The cost of the cheapest Fusion license is 150% more than I pay for unlimited internet at 500mbs per month, so that is never going to happen for a hobby software tool. If I wanted to get commercial with this I would use open source FEA code and write my own GUI.

The way you describe how you would do it, is exactly how I approached it, but it is a lot different in real life from the software companies sales blurb. For a start, if you are actually going to do it from a development perspective and not how quickly you can make a cool geometric shape, then you need to be able to describe different areas and sections in intricate detail, problem in node 3452 doesn't help you. I used to have my own company supplying hardware to the CAD industry many years ago and the way you describe how CAD software works and how it actually works...does not tally. Then from research I learned that nearly all the Space Frames out there are made to combat gravity or carry a load, so their jointing systems are designed for loads in a particular direction. If you look at it from a mathematical perspective and you want stiffness in all axis, then it's equilateral triangles and not pieces of different lengths. Kind of sucks and takes all the fun out of it but if you want stiffness in all directions then you are going to end up working in a sphere. Next best thing is a pyramid.

Had a close look at how to actually build a space frame for the purpose I want to use one for, and making it much more complicated and double the cost by introducing bolted connections to make it more simple to assemble doesn't help you if you are the one making the complicated connections and doing the assembly, far simpler to just notch the tube then weld together, from a lot of experience notching tube of this size I know if you make accurate cuts, which is simple to do if you make a good jig then it is easy to assemble, the pipes will only sit flush and snug at the correct angle and position, if you spend a lot of time making a good notcher then its easier to get right than a frame made from square tube.

Then your accuracy of your frame has nothing to do with the accuracy of your machine, how you set your rails on the machine determines how accurate the machine will be, the frame could be wonky by centimeters but if the frame is super stiff and you set your rails on it accurately then you will have a great machine.

I am not going to worry about distortion because I think I would have to weld up hundreds of joints up badly to have a situation were welding stress of a few bad joints would have the power to pull the frame out of shape, then I have 100% confidence in making a laser system were any movement of the rails though the distortion of the frame can be detected and the controller software can remap itself to still cut accurately. Going to spend my money on a CNC system that uses a laser on each axis(not attached to machine) to maintain accuracy and not the position and shape of a lump of steel.

[peteeng]

Well since you won't tell me what systems you are using I can't help. A triangle is a triangle whether its equilateral or not. There are various systems to optimise shapes. Its not the FE that's the issue, its generating complex geometry easily (prefer it to be parametric and function driven) and grasshopper is the go to for that. Well the way I described it is the way it works for me. Designed lots of spaceframe domes and shelters in my time plus did the fe... Good luck on your adventures. Peter

[G-Spot]

] A triangle is a triangle whether its equilateral or not. ]

So a truck stand is just as effective if the lengths are unequal? Of course not, if you want equal stability in all directions it has to be equilateral.

And after you designed your space frame in whatever software package, how do you go about checking it before you build it? In my playbook you need an independent method to verify the results, which is why I cam e up with checking the mathematical possibilities with the number of actual connections. From doing that I discovered there is a bug in the software where if you remove an object from the project that has a beam profile, then the beam's profile gets orphaned and then recreates the beam you deleted at mesh time.

So FEA where you build something is a lot more work than FEA where you just design something. And you cant build something using parametric functions at some point you will have that object in your hand and have to deal with it as a unique object not as a group.

From my experience that tells me it is better to come up with a strong descriptive system for each of your elements before you start or you will always be taking your results from the software in blind faith.

I have worked in nuclear power and aerospace for some of the biggest companies in the fields, none of them were innovative, they just went to the right software vendors, the right hardware vendors, everything was always done according to some expert's playbook. I find that type of engineering mind numbingly boring and spent as much time as I could in the field which was the opposite, using the vendors and tools you happen to have on that particular day to achieve your goals.

Did I mention that for me this is a DIY project where the more interesting option is the correct option and not the one which will be a commercial success.

[G-Spot]

40NBx5m is Dia45x5mm so has a cross area of 628mm2. This is the same as a Dia28mm rod so use that. Easier to deal with in building... Run the same geometry as a beam rod 28mm and you will see....Peter

I think you made an error in your calculations.

[peteeng]

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