[peteeng]

Hi All - I have upgraded Fusion to the advanced simulation extension so can do generative design. Had a play this morning. Need to learn a few things. Started with this told it to halve the weight and ended up with Mr Blooby. Peter

[peteeng]

Hi All - Fusion reduced the 400kg model to 50kg at same stiffness. Interesting shape. Will keep learning. Peter

[peteeng]

Hi All - Fusions GD module has an interesting feature. You can put in a load and a target deflection. So I put in 100N and 0.001mm and it hit the 0.001mm in 37 iterations. Once I figure out protected areas and clearance areas and a few things perhaps Fusion can get a design in one go! Artificial intelligence taking over my job. Peter

edit - reduced 415kg to 48kg

[suspension]

Hi All - Fusions GD module has an interesting feature. You can put in a load and a target deflection. So I put in 100N and 0.001mm and it hit the 0.001mm in 37 iterations. Once I figure out protected areas and clearance areas and a few things perhaps Fusion can get a design in one go! Artificial intelligence taking over my job. Peter

edit - reduced 415kg to 48kg

I was wondering if I can use the generative design for my CNC frame. I was never able to get a decent stiffness with steel plates.

[peteeng]

Hi Sus - I'm sure you can. Its a learning curve. But its the way to go into the future. Machine stiffness is all about maximising part geometry or investigating exo-skeleton type designs. A traditional C column design is difficult to get very stiff unless its a big machine ie its geometry is already big. Peter

[peteeng]

Evening all & Sundry - Its been a bit over 2 years since I started this thread. Been a decent journey including many sidetracks and rabbit holes. Looking at #1 entry I think I should recap. I intended creating a cold casting material of E70 equivalent to aluminium. Failed except for carbon fibre which I can make E60 thru to E90 using std modulus CF. But that's very labour intensive but doable. The idea of casting is to remove all that labour, just pour something into a mould...

The Germans have an E80 concrete but I have not been able to replicate that. Ideally the material can be routed so the finishing can be done in one of my routers. Again this wish failed. But I can go with laminated aluminium which achieves E70 and is machinable in my routers. (Look up GLARE composite) I dicked around with lots of configurations and floundered at getting my target stiffness of 50N/um, I think was the target.... The envelope of 500x500x500 was ambitious for a benchtop machine. But I have come to the conclusion my preferred configs are a) like the Mori M1 or b) a high wall design (see images attached). The BT-30 head is a bit big and so an ISO20 is better. Or a high speed spindle with good bearings and adopt very HS machining techniques. Even for steel this is doable but would be on the bleeding edge of surface speeds for various coated tools. But I think the technology is there to play with and its certainly being used by various companies and hobbyists. Feeds and Speeds have consistently climbed over the last few years in metals and its into the high speed spindle zone now.

What's changed?
1) The exercise has given me a deeper understanding of various relationships within the machine design space. Geometry is king and the material is a bonus at the end. Start designing at the spindle and work outward, forget about material until the end. Don't start at the machine base as you take three steps fwd and 2 steps back as each next part does not work out. If you start at the spindle you can progress to the base in a relatively smooth workflow. The machine may be bigger then you expect but then you adjust it as needed.

2) Fusion - I have been using Fusion for 1.5 years now and I think I have the hang of it except some of the moving joints stuff. The support has been very good, they answer my questions quickly and they notify me when my thoughts get escalated to their to do list... or its already on the list (which must be from here to the Moon). Fusion has just clicked over 10 years old... In terms of designing a machine its faster then traditional solid modellers as its a top down modeller vs a bottom up modeller. Plus the models seem to be more robust. With Alibre and others you change things and change things in development and eventual the wheels fall off the relationships, the model freezes, stalls, complains and just becomes angry on both sides. So far Fusion has not failed unless its operator error. I'm impressed. Plus its nice having FEA integrated with the modeling. It also allows things like file names to be renamed, parts to be moved around in the file structure and the model does not fail. In other programs you need a vault or specialist file management software to do that sort of thing...

I do not understand why the timeline and browser are separate. Other CAD browsers make it much easier to find relationships and features as these are combined. The only thought I have is that people are generally using direct modelling techniques these days vs history based workflow. Maybe I should try more of that.

Recently I have upgraded the simulation side to allow generative design and that's what I'm about to do with Milli. Design a generic structure then let AI do the work. In generative design you can specify the stiffness required and it will do the crunching. My job maybe about to be superceded; probably it already has....

3) Material dampness - Many of the cast materials such as concrete and EG are held up as wonderful because they are relatively damp. But once you dig in you find out that material dampness cannot combat typical machine vibrations very well. They only take small steps towards helping. I've attached an article on this. You have to read between the lines a little because the author likes constrained layer structures (such as laminated metals). I now take the view that the structure has to be maxed out in static stiffness, made as light as possible, then use tap testing to identify the dynamic stiffness of the systems and then use those results to optimise the operation of the mill/router. Tap testing provides huge performance gains in removing vibration (by knowing where not to operate and where to operate) and then maxing MRR if that's what you need to do.

So watch this space occasionally to see what AI comes up with.... Onward and upward Chatbot, Suri, Hal etc - Peter

[peteeng]

Hi all - Here is the extract about material being inadequate for good damping. Peter

[peteeng]

Hi all - heres my block model and first GD runs. The solver wants to connect the spindle to the base when I run for minimum mass, ie it wants to create a closed loop structure. I'll keep playing, the red block is the exclusion zone. Peter

[peteeng]

Hi All- Here's another one. It can't be stiff? I'll have to do some reading and more playing. I'm also playing with shape optimisation in the simulation space... Peter

[peteeng]

Hi all - I think I like the shape optimiser better. It has a slider and shows the shape it can be at different weights...Peter

[peteeng]

Hi All - I'm tending to think the shape optimiser is the go vs generative or automated. We know the outside maximum dims of each part (bounding box of each part) and this is the starting point as this is maximum geometric inertia. Then the optimiser removes non stressed areas. I'll keep playing with generative as well. Peter

The nominal design is all steel, weighs 1486kg and has a static stiffness of 67N/um. The optimiser has removed around 50% of the material and has dropped the stiffness from 75 to 67N/um. The process has been quite smooth.

Its X = 660mm Y= 330mm and Z=300mm

So Fusion has automated modelling that can join "required" geometry - it has Generative design which can connect required geometry (both of these work on stress minimisation and may not give a max stiffness geometry) - it has the optimiser which works with loadpaths which is better for machine design... I now think my earlier thoughts of building a 50N/um benchtop machine was a pipe dream... Now to build one closer to reality including bolt access and drives that meets the original Milli intent. Only 4.5 years I must be a slow learner...

[peteeng]

Morning all - I'm going to call this last model Milli-24OPT1. 24OPT-1 does not meet the Milli objective in a couple of places. 1) its not a benchtop machine; at 1.5 tonnes of steel its not small 2) it has a high speed spindle and the configuration makes it difficult to achieve a slow speed one for steel 3) This also applies to my favoured high rail design. So this means I need to move to the Mori 200 type of config which has an outreach providing real estate for a motor and pulley/spindle.

In another thread we spent time on spindles for steel. It comes down (for me) to a 750 or 1000W AC servo with an upspeed pulley to give 0-6000rpm or 0-9000rpm or use a high speed spindle with downspeed pulley to give 6000-12000rpm say or even 4000-9000rpm....

The other aspect is that 24OPT1 was designed from the spindle backwards and arrived at 600x300mm envelope. This is driven by the 80mm dia clamp and 20mm flanged bearings. Each bearing patch sort of grows as it goes from Z axis to the column. So the next version needs to use 15mm square bearings to minimise the bearing patches...

I did more reading on shape optimisation in Fusion and it definitely is the better approach then generative for machine parts until I become better at generative. GD is a tool and all tools you need to get competent with. I do like the M1 config so will model the motor/pulley/spindle and see if I can package it into the 24OPT1 or 2 Z axis....

So onward to Milli-24OPT-2 Peter

[peteeng]

Hi all - Milli is also a materials thread. So if I built it from aluminium it would be 3x softer then 24OPT1 but a lot lighter. (2700/7800= 2.9 ~1/3 lighter) Since I removed material that did not help stiffness it would not be worthwhile putting some back to try to recover the stiffness. if I made it from concrete It would be less stiff then aluminium and then there's EG less again... See what OPT-2 brings Peter

[joeavaerage]

Hi peteeng,

In another thread we spent time on spindles for steel. It comes down (for me) to a 750 or 1000W AC servo with an upspeed pulley to give 0-6000rpm or 0-9000rpm or use a high speed spindle with downspeed pulley to give 6000-12000rpm say or even 4000-9000rpm....

I strongly agree. For a mill such as you envisage being able to cut steel is the goal, and in which case high torque at low(ish) speeds are required from your spindle.
An AC servo, with an upspeed pulley is viable, as is a high speed AC motor with a downspeed pulley. While both lack the top speed of the 24000rpm asynchronous spindles
so favoured by hobbyists, they are a better match for larger tools in steel....no matter the cost or inconvenience.

You know already I have an AC servo driven spindle for steel. It is direct coupled so the spindle is capped at 3500rpm of the servo, but is none the less very useful, and it requires
no more real estate than a 24000rpm asynchronous spindle. I swap back and forth between my highspeed (24000rpm) spindle and this low speed (3500rpm). Takes a few minutes
so is not too bad.

Just had notification that NZ Customs are going to release my new 3.5kW HSK35 ATC spindle to me.....after quite some correspondence. Bottom line is do not BS Customs....
you'll regret it!!!!

Craig

[peteeng]

Hi Craig - What is the torque of your AC and what would the biggest dia tool you'd run in it? rpm?

1000W 3000rpm 4Nm with 2x up gives 0-6000rpm at 2Nm
2.2kW 24000rpm (0.8Nm) at 3x down gives 2000-6000rpm at 2.4Nm a 12mm tool at 6000rpm has a surface speed of 226m/min so probably don't want to go faster then that...

servo probably gives more options...

Peter

[joeavaerage]

Hi peteeng,

What is the torque of your AC and what would the biggest dia tool you'd run in it? rpm?

The specs say the rated torque of my new spindle is 3.4Nm. I would guess that I could run a 12mm four flute tool in steel at 2500rpm or 94m/min surface speed.
This is a guess based on what I know from my servo driven spindle. What I am looking forward to is that my new spindle is rated at 10,000rpm so I expect to easily get
down to 5000rpm, and possibly as low as 2500rpm with good spindle security.......but will have to wait and see.

Really this new spindle is about ATC, it has an HSK32 tool interface. While its rated at 10,000rpm its max is 40,000rpm. I'm thinking that I have a useful range of 25000rpm to
40000rpm.......and ATC.

It is a moderately expensive purchase for me, $4000NZD, so I'm hoping that the overall increase in capability justifies the cost. You can well imagine my sigh of relief when Customs
released the spindle, I was a hairs breadth from losing it altogether with zero compensation. That would have hurt!.

I still have to buy a VFD for it, and it requires a high frequency model, so that in itself is likely to set me back another $1000NZD.

servo probably gives more options...

I agree. Modern AC servos have so much overload capacity, you'd swear they are rather more powerful than the nameplate suggests.
Should you encounter a brief moment where the tool experiences a very great increase in torque demand (toolpath anomaly or a sudden unanticipated plunge) a servo will 'man up'
and deliver whereas AC asynchronous motors tend to stall.

I would urge you consider:

https://www.fasttobuy.com/delta-2kw-...nc_p33936.html

It is 2kw, 6.4Nm @3000 rpm cont. , 19.1Nm peak and 5000rpm max. You have a viable speed range of 0 rpm to 5000rpm, which I think you'll agree is pretty damned good. Cost $895USD plus shipping.
(At 5000rpm is in field weakening mode so torque is 6.4 x 3000/5000 =3.84Nm)

My Allen Bradley servo, previously pictured, is 1.8kW 6.1Nm @3500rpm cont. 18Nm peak. Personal experience is that 6.1Nm is highly useful.....you'll always find the need for more, especially for rigid tapping, but nonetheless 6.1Nm
does a power or work. I have run 16mm four flute carbide tools in med tensile steel quite happily, if cautiously, with it. 16mm is about as big as the ER25 toolholder can manage.

Is there a need to upspeed?. If anything I'd be tempted to downspeed so I could use 3 and 4 inch face mils and really start removing material! Be great for rigid tapping as well.

Craig

[peteeng]

Hi Craig - Its intended as a benchtop Makers grade machine. I need to start in the middle range of Makers but have the flexibility for them to up speed or down speed so they can tailor to their needs. I don't want too many options up front that's for the Maker to figure out... One person will never touch steel, someone else may never touch Alum.

But also I think a bolt on, bolt off spindle is a good thing. Can then use high speed and servo like you do.... I'll start modelling a suitable front end. What frame size is your servo? I'm thinking 100mm is too big... will mean the machine grows up too big again...Peter

[jackjr-123]

The specs say the rated torque of my new spindle is 3.4Nm. I would guess that I could run a 12mm four flute tool in steel at 2500rpm or 94m/min surface speed.
This is a guess based on what I know from my servo driven spindle. What I am looking forward to is that my new spindle is rated at 10,000rpm so I expect to easily get
down to 5000rpm, and possibly as low as 2500rpm with good spindle security.......but will have to wait and see.

Really this new spindle is about ATC, it has an HSK32 tool interface. While its rated at 10,000rpm its max is 40,000rpm. I'm thinking that I have a useful range of 25000rpm to
40000rpm.......and ATC.

I have one of these spindles. With the Chinese it's hard to get correct information. While asking for the torque/frequency curve of the motor I got 3 different replies, and their VFD settings were all wrong (for them rated speed = max speed).
In the end if the rated torque is 3.4 Nm (like mine) and it's a 4 pole motor the rated speed should be 16'350 rpm. This is what I set the VFD to. I fitted an encoder, so the spindle can reach full torque down to 0 rpm, but 3Nm is still not much for steel.

[joeavaerage]

Hi,

In the end if the rated torque is 3.4 Nm (like mine) and it's a 4 pole motor the rated speed should be 16'350 rpm. This is what I set the VFD to. I fitted an encoder, so the spindle can reach full torque down to 0 rpm, but 3Nm is still not much for steel.

I took that off the manufacturers data that accompanies the spindle. The Torque/Speed curve clearly indicates 10,000rpm as rated speed.
It is four pole, ie 333Hz is 10,000rpm, 700Hz is 21,000rpm and 1333Hz=40,000rpm.

You might be surprised what can be done with 3Nm. In another thread, corresponding with peteeng, recently I measured the torque required to cut mild steel using the current demand of my AC servo spindle. The correspondence of
current to torque is a very linear relationship and I found that even 12mm tools would absorb 2Nm to 3Nm even when making quite aggressive cuts, certainly up to the limit of my machine.
So while I agree 3.4Nm is hardly a surfeit of torque for steel neither is it too shabby.

My current high speed spindle (24000rpm, 800W) has a stall torque of 0.3Nm.and so cutting steel with it is an exercise in patience but I can and do use it for that purpose.
I did so particularly when my servo spindle was out of action for several months. My new spindle is ten times better and so my steel capacity (without spindle change) has
increased dramatically, and additionally this spindle is faster, up to 40,000rpm. I have better low speed torque and yet have a significantly higher top speed as well.

According to the wiring diagram the spindle is fitted with a SinCos encoder already, although as I have yet to receive it I cannot confirm that. While an encoder is nice its not really a priority
for me. I more interested in its free running performance in the range 5000rpm to 30,000 rpm. If I want to do rigid tapping then I would fit my servo spindle. It, being a servo, has position control
and is very useful for rigid taping, aside from having much higher torque.

Quite some years ago, maybe eight or more, I bought a re-manufactured servo ( Vickers) from circa 95. It is 3kW, up to 4000 rpm, 12Nm cont and 48Nm peak. It was always my intention to
make a spindle with it. It is fitted with a resolver rather than an encoder and thus drives are hard to find and expensive when you do. Electronics is my thing and so I have designed and built
my own Field Oriented Control drive for it. I had a prototype lashed up about six years ago, and it worked but it was still a lash up. Since then I've sold my home, shifted, got another job, bought a business...
and so the project has been in abeyance for six years...but should i require genuine high torque for face mils or rigid tapping, this is the motor for it.

Craig

[joeavaerage]

Hi peteeng,

But also I think a bolt on, bolt off spindle is a good thing. Can then use high speed and servo like you do.... I'll start modelling a suitable front end. What frame size is your servo? I'm thinking 100mm is too big... will mean the machine grows up too big again...Peter

I agree. Making the machine spindle agnostic makes huge sense, and is and always was my intention. It would be nice to think I could fit one spindle to do everything I ever wanted, and they are available
but they are totally beyond my means and they are huge and heavy. No, the best I could do is have two or as many as three different spindles that I could fit as need demands. I do anticipate that my new
spindle having both higher and lower rpm's will be less frequently swapped out for a genuine high torque spindle, but I still anticipate the need for a second spindle.

My servo spindle slips into the same holder that my 65mm diameter 24000rpm spindle slips into. I do not have to unbolt anything, just loosen two clamping screws, slide one
spindle out and slide another in. Pretty easy. That is the advantage of having the servo direct coupled, it is physically similar in size and shape to a high speed asynchronous spindle,
and a swap is easy to make.

My 1.8kW Allen Bradley spindle is a 100mm frame (measured across the flats), and I disagree that its too big. Remember this is 2.5hp, with a three fold overload. That make AC servos
perhaps the most power dense motors available, and a small fraction the size of a 2.5hp induction motor. The 2kW Delta servo to which I linked yesterday is also 100mm frame.
Note that Delta have another B2 model, also 2kW but in a 130mm frame. Its rated at 9.55Nm @ 2000rpm and 3000rpm max. Peak overload is 28.7Nm. So a modest increase in frame size
results in quite a jump in torque, sadly the price has crept up to $1100USD plus shipping. None the less it would make for a good motive source for a combined upspeed/downspeed arrangement.
Sadly having that flexibility comes at the cost of complexity and parts/materials costs.

The market for ATC spindles is very competitive. There seems to be a real flush of Chinese made ATC spindles with ISO20, ISO25, HSK25, BT30 tool interfaces in the range of 2.5kW to 5kW
ranging in price from below $1000USD to $2000USD. Very VERY hard to complain about the pricing. It was only a few years ago when such spindles when they were available were two and
three times those prices.

My spindle is a good example, $2000USD for a 3.5kW HSK32 ATC spindle, with ten HSK32/ER20 toolholders and free shipping to New Zealand, that's a damned good deal.

All-in-all with the price of servos coming down and the increased availability of very useful highspeed ATC spindles the market is moving to where hobbyists are......to our advantage.

Craig