CT scanning Accuracy/Resolution

[[Ni...]]

Looking into CT scanning technology to take our Inspection of the medical devices we make to the next level.   In the real world of manufacturing, I am curious if Zeiss offers a CT scanner capable of actually measuring our .001" profile callouts we see often on some of our parts.   Walking around IMTS last week hearing some of the claimed resolution created some concern about the accuracy this technology currently offers.   

Does anyone have real world experience trying to measure ±.0005" with CT scanning technology?

 

Thanks

[[Ti...]]

What were you hearing regarding the accuracy at IMTS? I'm curious what was stated.

I always scratch my head when an engineer gives you only 5 tenths to miss in one direction. Why not give us 7 or 8 tenths, what is so magical about the .001" profile number? At least the metric prints give you .03mm so you get an extra tenth 😉  When you start getting this tight your datum(s) become critical, so its not just the surface you need to be worried about it is how all the datums are stacking up together.

I'm still learning, but here's what I see. .0005 on both sides of a simple surface that isn't large is doable, but on a custom thread profile I don't see how to even machine it that good let alone measure it within that tol. The scans are going to create a mesh with radii corners when the print will show a sharp corner. So you are already out there, but you simply back off those areas. 

There are some variables you throw at the scan like kilavolts, watts, zoom/voxel size, milliseconds per position, positions per full rotation, etc. We will do a 10-25 minute scan. We had a trainer come in and he is more of like a job shop where a 45 minute scan isn't too long for him. He gets sharper edges than we do. I told a few people I'd like to let a couple parts bake in there for 2 hours and see how sharp we can get the meshes, just so I know the capability if needed. It's not a magic bullet, you don't just buy the CT scanner and throw a part in and open GOM and boom you measure .0002 profile for internal features. I would say be wary of expecting to hit .001" profile in general. .002" no problem imho.

 

 

[[Ja...]]

I'd think more holistically ,like Tim is alluding to.  To convert , you are talking about a tolerance band of 25 microns. the device in question/the process in question would have to be repeatable to single figure microns to get a reliable answer .  In reality this is barely acheieve able with any device let alone ct, many hours , controls and studies generally would need to be in place.  Tim rightly points out that when datums come into the game this is made harder still and most likely completely impractical.  

So first question always is , how is it measured today?  What is your current capability? 

You also have to consider the assessment of good enough and the information gained.

If say you assess a flatness , for ease of comparison, and you have a 3d mesh from a device.  If you took say 6 discrete points on that surface like a contact system you get a different value of flatness , and better repeatability than looking at the full surface. However you gain more information about your manufacturing process from the full surface.

I would approach by having some of the items captured by a device under test.   A. Across a range of parts do you see anything in this area? B. Get an impression of the repeat value of the machine and various methodologies for assessing the feature .

Then maybe move away from just a tolerance game and see if its good enough to get valuable information on this surface.

Beyond that it is a discussion with design on the options available and the limitations of reality .

I hope this helps in some way

[[Ni...]]

Currently for our bone screw thread and also head profiles we use a combination of one of our (5) Zeiss Contura's with a Vast XT probe head and/or an optical comparator with an overlay.  Some of our inspection methods are customer driven of course to achieve correlation (which is a pretty standard practice) so depending on the customer our inspection method does vary. 

Zeiss wasn't making huge promises in regards to accuracy.  They seemed to speak fairly realistic at approximately 10µm depending on the machine. Another manufacturer claimed 4.5µm which got my attention of course.   Before my time with this company we had sent parts to Zeiss for a CT study and of course we get our reports back with favorable results and the Cad model comparison showing a lot of interesting and seemingly useful data.  Having been alive for a little over an hour at this point in my life (sarcasm) I know reality can differ greatly from what a sales person would like us to believe. (some not all)  

I was having this discussion with one of our quality engineers and thinking how this CT tool could drive some print/tolerance changes with our customers but that is an entirely different conversation.  Like stated a .001 profile or true position tolerance is quite the ask from any Inspection tool currently offered. a tenth or two of error adds up quite a bit in those callouts.  

@Tim Jones    Can I ask what CT machine/model you are using?  How big are your parts you are scanning that take 25mins?  Does that run time out weight the other methods of inspection? mainly cost?   We have been eying up the Metrotom 6 Scout.  Over in the Werth booth the TomoScope XS was on display at IMTS and during my hour long demo I was pretty impressed.  That is of course if what they claim is achievable in reality. 

 

Thanks for the replies

[[Jo...]]

Hello all,

Really enjoying reading over the discourse here. Perhaps I spoke with you at IMTS, my name is John Moore. As it happens, I have quite the extensive application experience with the MT6 x Bone Screws.

At a basic level I like to broadcast that we can generally handle +-0.030um tolerances on bone screws. Of course, as you all have already alluded to, throwing datums into the mix can often lower this. This then becomes a case to case situation depending on feature size and datum design.

Please see the attached example for a demo Type 1 study I did on a bone plate with the MT6. Typically, titanium bone plates and bone screws perform similarly in the MT6 in terms of repeatability. Would love to speak with you further on the topic if these numbers grab your attention!

-John

[[Ni...]]

Thank you for the info John. Does Zeiss claim a minimum feature size it is able to capture?  (R.005"?) Also, what is the approximate scan time for the part you have shown above? From lets say closing the door to getting a dimensional report in your hands? 

 

Do you have any experience with the other models of CT that Zeiss offers?  Is there a "better" option which would allow us to achieve a higher level of accuracy? 

 

Thanks again, 

Nick

[[Ti...]]

Please sign in to view this username.

We only have 1 part that is over 20 min, I believe there is a way to reduce the time thru cropping. So some of the 18min scans are now at 14-15 min. Some parts are 10min. I'm not the best to explain the cost/time analysis, but I know there is a lot less parts cut in half and less wasted time checking dimensions. Maybe John can answer some questions you had and mine below...

Please sign in to view this username.

Do you have experience programming the scans? I'm curious what an avg scan time for you would be on a 15mm Height x 10mm wide part. Something like a full resolution 300 millisec setting is roughly 15min. Wondering if you are around that for some parts and what your times might be. Have you ever let a part bake in there for hours to see how sharp the corners would come out? I'm curious how/if the technology handles a fixture being out of dead center by .05mm which would introduce wobble. Do you ever see bad meshes produced when you are expecting to see two materials and it only sees one?  Thanks.

Edited September 28, 2022

[[Ja...]]

John , appreciate you have tried to qualify stating basic level and datum impact .

Please dont take this the wrong way but i think you have to be very careful making feature size and performance claims.

You have a nice result there no doubt , but you are pulling out the result of one dimension to make a judgement .

Youve taken a result of a diameter of a cylinder.   These usually perform well in r and r tests as the fitting of the feature inherantly helps smooth out noise in the process.

My point is that if you look at cylinder 3 cylindricity it is a sigma of 0.032 not 0.002 ..this is an entirely different Cg value.  This is of course caused by different calculation , number of points and sensitivity.   I.e. The same feature measured in a different way yields a different capability , just like a datum would impact.

This is entirely normal , but its easy for someone to go off with a number without fully understanding .

An endlessly tricky topic!