As Benjamin Franklin noted:
An ounce of prevention is worth a pound of cure.
You have a Sweepy, use it.
2 Likes
DO NOT USE AI FOR TOOLPATHS – I too learned this the hard way.
Carbide Create Pro has some learning curve, but not too bad at all. Pro upgrade is a good value, so you can output toolpath NC files and get stuff you can understand and fix. For me this was a relatively painless way to get up to speed.
Also recommend the writeup at Feeds & speeds | Shapeoko CNC A to Z – very helpful.
2 Likes
This talk about maintaining your Shapeoko mill had me look at the link on Feeds & speeds. Having had over 30 years’ experience in the machine tool industry working for a company that manufactured gear cutting machines. I started out with the companies two apprentice program. As I advanced through the shop I got to run the first iteration of CNC, known as NC machines. These Numerical Controlled machines used an 8bit code on a reel to reel reader. Think of the old reel to reel tape recorders. And before I retired from the company we were into CNC machines using Fanuc, Allen Bradely and Siemens controllers.
This statement to me is a little misleading.
“climb milling used to have a bad reputation for being dangerous to use on machines with a lot of backlash. While this was perfectly true on older manual mills, the point is moot on CNCs in general and the Shapeoko in particular.”
If you’ve ever experienced a CNC machine climb cutting and it grabs the part due to too much backlash in the lead screw or worn-out ball screw and hadn’t the need to change your underwear, then you haven’t experienced one of the more catastrophic crashes on a CNC machine.
The only difference between a manual machine and a CNC is on the manual the operator gets to turn the handwheel as apposed to a motor on a CNC machine. Granted the CNC will be steadier than cranking it by hand. Whether manual to CNC how much backlash is in the axis plays a big part of getting through the cut without a major issue.
Bernd
1 Like
William, I understand your position. You’ve made it quite clear. No, I will not be using the Sweepy right now, for all of the reasons I already explained. I am new to CNC, I am testing unfamiliar and sometimes AI generated toolpaths, and right now I want completely unobstructed visibility of the cutter and every movement of the machine. That is a conscious decision, not something I need another Benjamin Franklin or Norm Abram quotation to correct.
More importantly, Sweepy is an optional accessory. Unless Carbide 3D’s position is that a Shapeoko cannot be safely operated without purchasing and continuously using an optional dust collection accessory, ordinary sawdust exposure should not cause multiple ball bearing assemblies to develop serious problems within approximately two weeks of operation. If that amount of ordinary use can contaminate the motion components badly enough to produce the failures I am experiencing, then that is useful information that should be clearly communicated to customers. From my perspective, however, a machine developing these kinds of mechanical problems this quickly points to something beyond simply telling the customer, “You have a Sweepy, use it.”
I wasn’t aware of the potential for this level of contamination, and I have now learned how to disassemble, clean, and repack the bearings myself. I’m taking responsibility for maintaining the machine going forward. What I don’t need is another proverb, woodworking television quotation, or safety lecture. I contacted Carbide 3D because a nearly new machine is experiencing mechanical failures and I’m trying to get it running reliably. Fleming is helping me with that, so I’m happy to continue working with him from here.
1 Like
I appreciate the advice, Timothy, but I absolutely will continue using AI for toolpaths. For me, that is part of learning where this technology is going, including learning its limitations and when its output needs to be corrected. I certainly don’t blindly trust what AI produces, which is precisely why I watch the machine so closely and verify what it is doing.
I’ve used Carbide Create Pro, and while it certainly has value, it doesn’t come remotely close to solving the problem AI is solving for me. AI allows me to describe what I’m trying to accomplish, iterate rapidly, work through geometry and machining strategies, troubleshoot problems, and produce things that would otherwise require me to become proficient in several different pieces of software before I could even begin experimenting. That capability is only going to improve.
So for me the lesson isn’t “don’t use AI.” The lesson is learn how to use AI intelligently, understand that it can make mistakes, verify its work, and never confuse generated output with guaranteed correct output. I’d much rather learn how to work with the technology now, including occasionally learning something the hard way, than avoid it because it isn’t perfect yet. I do appreciate the feeds and speeds resource, though. That kind of information is exactly what helps me better evaluate what the AI is producing.
2 Likes
I use AI for lots of things, and really appreciate claude code. But I use it for things I already know how to do, and could do just fine without. For me, it’s a big productivity boost and I’ve had really good results that way.
When you’re using AI to do things while you’re learning along the way, and you’re not doing proper dust collection, then please – don’t complain when the learning curve is steep and expensive, and don’t criticize the product or company!
Also, dust collection is for your health just as much as to reduce cnc maintenance costs, so hopefully you’ll be able to hook it up soon.
2 Likes
Appreciate your comments on climb milling, sounds like an experience I don’t want to have! I’ll definitely pay closer attention to it than I’ve been doing – thanks!
1 Like
How do you verify the AI output is “good” before running it on multi-thousand dollar hardware that can cause significant injury?
From the initial problem report of bad noises and HM009 errors – if I run the Nomad such that it is unable to move through the material, I see similar symptoms (bad noises, errors, invalid position). I generally take that as a cue to seriously re-consider my approach, and have been able to proceed after re-initializing the machine (sometimes a few times) and changing the cutting approach (toolpath, feeds, speeds).
On the topic of dust collection – I recently spent a bunch of time making a dust boot. The choice is simply : remove as much material from the environment as you are able to… or don’t. If material cuttings are not collected, there are consequences: specifically it migrates throughout the machine, the environment, and the bodies in the environment.
2 Likes
Had a thought about using AI for your project – how about having the AI output stl files, then use proven software to turn the models into toolpaths? That avoids the serious bugs in AI-generated toolpaths, and you can view the stl files to check that they’re what you want.
4 Likes
Hi. Not trying to take any side here, but I was reading this page and was just curious if the above meant that you’re not using the Sweepy, or that you’re intentionally not using any dust collection at all?
The Sweepy itself is just a shroud and I doubt C3D cares whether you use their shroud, or you download and print something on your own. But I was just wondering if you’re using any kind of dust collection at all.
I am very unsure if this is just trolling
I understand wanting to tinker and learn new approaches, but it doesn’t make any sense to refuse to use dust collection
5 Likes
Esp since a clear sweepy base is included so you can observe the endmill at work.
“Two clear polycarbonate bases are included, one with a brush for maximum dust removal, and one without a brush for maximum visibility.”
I guess if you want to waste time doing clean up and unscheduled maintenance…
2 Likes
This thread gets me thinking about one of my favorite quotes. "In the beginner’s mind there are many possibilities, but in the expert’s there are few” hehe
6 Likes
I need to remember that one- it’s almost a corollary to Chesterton’s Fence, which is my favorite.
6 Likes
Good one Rob. I’ll have to remember that one as well
1 Like
Nice. I had to look that one up
2 Likes
If you are going to be watching it anyway. Have your shop vac going and with one of the nozzles (to keep your hands away from things) and suck up the chips as they come off. You can see what is going on, and get rid of the chips. Some of the CNC folks doing youtube tutorials do this so you can see the cut clearly.
John
Chesterton’s Fence
“Why keep this fence?” asks Ned at the mill.
“It blocks my CNC. Remove it I will.”
The door swings ajar; the spindle says, “No!”
Chips style his eyebrows in one rapid go.
The guard keeps sawdust and cutters in line;
The lock stops the spindle at opening time.
Ask what that barrier was built to defend,
Before you remove it, discover its end.
6 Likes
As someone who works with bearings on a daily basis, I’ll simply chime in to say cigarette smoke or anything larger counts as a red flag grade A emergency to a properly operating bearing. Of course, the dose makes the poison.
Lubricants should be as close to sterile as possible. Greases should not be mixed without careful testing. Chassis grease and bearing grease are different and behave differently, and a motor grease, spindle grease, and crusher grease are also different bearing greases. Often incompatibly so.
I strongly recommend anyone who isn’t an expert in bearings and lubrication defer to the OEM. I see maintenance professionals get it wrong, often.
Great that you can get it back together, but a can of brake kleen and a fresh bottle of Vactra will prevent further, long term damage.
Sorry - I’m a bearing expert. It’s a touchy subject for me. I don’t want to see surface- initiated fatigue spalling in 4 months and have you then think it’s a quality issue.
10 Likes
How are you creating toolpaths, are you literally just telling claude code to make gcode given some file (stl or whatever), giving it a toolpath type and some parameters (mill type, diameter), or are you doing something else?
A bit over a year ago I had Claude generate a fusion plugin for v-carveing on a curved surface. It was a fair amount of work on my part then, I bet it could just about one-shot it now.