Can a CNC program malfunction?

[Nargo]

Greetings! The company where I work received a valve with its driving nut, where an actuator engages to open and close it, too large. The specified width across flats of this hex nut is 2.236 +/- 0.005 inch, but it was delivered with this dimension at 2.282 inch. All the other features of this nut were correctly machined. The manufacturer of this valve has not submitted yet the Root Cause Analysis for this defect, I have requested this Analysis, so meanwhile I am gathering information so I can be prepared to review it. I don’t have experience with CNC machining so, if CNC was used to machine this nut (I don’t know yet if it was done in a CNC milling machine, because if it was manually machined, then this would be simply human error by the machinist or wrong information in a shop drawing or in another manufacturing document) is it possible that a CNC program loses the reference, skips a program step, or experiences another type of malfunction, and produces such a large deviation in the size of this nut? I imagine that CNC programs are always executed by the machine in the same way. I would appreciate very much any opinion on this. Thanks.

 

[3DDave]

For CNC the most likely cause was a cutting tool of the wrong diameter that was used solely for these surfaces.

The second most likely cause was the programmer did a roughing step and failed to program the final material removal.

Way down on the list is a failure of the CNC machine to follow the correct tool path.

 

[Nargo]

Thank you very much, 3DDave, for the quick reply and for the very useful explanation. But, while reading it, I realized that I didn't mention in my post, that we received five valves, the nuts of the other four were fine at 2.237 inch between flats. However, I don't know if the four SAT nuts were done in the same batch as the UNSAT nut, and this is a crucial clue that I don't have yet, I expect this will be indicated when the the valve manufacturer sends me the Root Cause Analysis. So, to cover for the information that I don't have at this moment, let me make my original question more specific with the most complex situation: Can a CNC program malfunction only once in a batch? So, a single defect in a batch would leave out Cause # 1 you indicated (wrong tool). However, your Cause # 2 looks promising. So, from my very little knowledge of CNC, I understand that CNC programs can be made continuous, so for example they will perform the roughing pass followed automatically by the finishing pass without machinist intervention, and even that CNC machines can replace by themselves tools between machining passes. However, from your answer I imagine that it is possible that these two passes were programmed separately, maybe to replace the roughing end mill with the finishing end mill, so the machinist could have ran the roughing code on that nut, and then s/he forgot to run the program of the finishing pass. But, I am more curious about your Cause # 3, even that you deemed it unlikely. Does it ever happen that a CNC deviates for a few seconds from the path that it has been following for the entire shift, and then go back to normal and continues in the expected path? The possibility of this erratic behavior on a machine is an interesting topic to study, I think.

 

[EdStainless]

They swapped a tool that was the wrong size.
These machines swap tools frequently during the machining process.
The other possibility is that there was debris that interfered with the zeroing of the tool position.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed

 

[mfgenggear]

OP
Most likely it was operator error.
Tool proofing requires a first article and it must be inspected complete.
The finishing pass missed or the operator added a new tool and did not mdi the tool correctly. When changing tools the z zero of the tool must be reset. The part inspected and the program adjusted. Any time a tool change has occured. Or
Line of code was accidently skipped.
Ether way operator has to verify his or her part.

 

[3DDave]

CNC deviates for a few seconds from the path that it has been following for the entire shift, and then go back to normal and continues in the expected path?

The Universe is infinite. Most deviations simply snap the tool off and destroy the work piece.

 

[IRstuff]

Are all 3 widths off by the same amount?

If so, then it sounds like wrong tool, worn tool, or programming error, which might be the same, or possibly calibration error or incorrect calibration data for the tool.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[MintJulep]

How does the surface finish compare to the "good" nuts?

I agree with the possible causes suggested by others above.

The probability of a validated CNC program momentarily having a hick-up that affects only one feature is very small.

 

[Nargo]

Excellent replies, each of you contributed very useful points, so now I have several things to review when I receive the Root Cause Analysis from the valve manufacturer, and also got two actions for me to investigate. I summarized both types of points below.

And, thank you 3DDave, you have replied twice. But, from mfgenggear post, after first-article inspection, what is the usual inspection practice for batches of, say, less than 20 pieces: either to verify the critical dimensions of each one, or to follow a sampling plan based on a predefined AQL? For a machining system that should be very consistent such as CNC, I imagine that checking a small sample should suffice.

Here is my summary:

Most-likely causes (not necessarily root causes): (1) Tool swapped to another one of different size. (2) Wrong zeroing due to: (2a) debris interference or (2b) operator missing reset step after tool change.

Actions for me (the valves are in a plant, several miles away from the office building where I work): (1) Measure three widths across flats. (2) Check the finish of the six nut surfaces.

Thanks once again!

 

[CWB1]

They're not the most likely cause but CNC malfunctions do happen regularly in any large shop/plant, and in small shops with older equipment. Many older stepper drives are open-loop so a bit of wear or dirt will cause a skipped step. Likewise, newer controls are often Windows-based and have internet connections so are susceptible to the usual plethora of PC bugs, software updates, and employees watching Netflix or ordering lunch on their machine. Given a split-second of control lag and the part is off.

 

[Nargo]

Thank you, CWB1, for this very interesting post. As I explained in my original question, I don't have experience with CNC (to give a little more details, I have worked for 32 years, 29 of those as mechanical engineer, but never in a CNC dept.). So, due to my ignorance, I wrote yesterday that CNC machining is "consistent", but your post has been an eye-opener to me. So, now I imagine that a good practice would require to inspect a higher percentage of parts, but also that it is difficult to find the right balance between cost of inspection and cost of rejections, so this has to be decided on a case-by-case basis and even with adjustments, increasing or decreasing the frequency to inspect machined parts in accordance to changing conditions.

Thanks!

 

[mfgenggear]

OP
I had to refrain from answering to fast.
Most companies have quality systems setup on their customers requirements.
While inspection of products 100% is not economical and causes delays and raises cost.
How ever depending on the AQL specifications and AQL level small lots can be inspected 100%. Generally very close tolerances are inspected 100%.
For example .ooo5 tolerance .
Some companies have setup lead programers/cnc machinist and have operators running productions. Some companies have very skilled cnc machinist running complicated parts.
Because it is not a perfect world defective parts do slup through.
And checks and balances failed.
For what ever reason , operator or machinist failed to find the part and correct it.
Inspection failed to detect the defect.
When a shop produces thousands of parts it happens. Now it appears to me this was a very small lot and should have been inspected 100%. View the required specification and AQL level. It will give the Inspection criteria or sampling requirement.

 

[Nargo]

Thanks mfgenggear, for another very interesting post (your first one was very useful too). I am with you on all the points you mentioned on this one. I feel that now I am well prepared to review the explanation of Root Cause and description of Preventive Actions that I requested to the valve manufacturer (I work at a Supplier Quality Dept.), and to have a conversation with them if I need it. However, I will be happy to read and acknowledge any other post that comes in the next days. Thanks again, have a good weekend to everybody if we don't post anything else later today.

 

[EdStainless]

It is common to have the tools touch a reference before machining to assure positioning.
We usually programed it to do so after machining as well.
If the tool/machine combination didn't hold zero, then we know that the part has issues.
This let us pull a fairly loose AQL and still get excellent quality.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed

 

[ShataCastings]

Cause 3# is very rare in CNC program. Machines are more trustable than human beings, 99.999% CNC defects are caused by human factors.

In factory, if there are a complex CNC processe involve multipe clampings, especially for large quantity orders, factory will use more machines, each machine just takes one process.

For this nut, maybe the factory has two CNC machines-A+B, A for rough machining, and B for precision machining. Worker confused this semi-finished parts with precision-machined parts, then you received four good nuts and one out of tolerance.

Thank you very much, 3DDave, for the quick reply and for the very useful explanation. But, while reading it, I realized that I didn't mention in my post, that we received five valves, the nuts of the other four were fine at 2.237 inch between flats. However, I don't know if the four SAT nuts were done in the same batch as the UNSAT nut, and this is a crucial clue that I don't have yet, I expect this will be indicated when the the valve manufacturer sends me the Root Cause Analysis. So, to cover for the information that I don't have at this moment, let me make my original question more specific with the most complex situation: Can a CNC program malfunction only once in a batch? So, a single defect in a batch would leave out Cause # 1 you indicated (wrong tool). However, your Cause # 2 looks promising. So, from my very little knowledge of CNC, I understand that CNC programs can be made continuous, so for example they will perform the roughing pass followed automatically by the finishing pass without machinist intervention, and even that CNC machines can replace by themselves tools between machining passes. However, from your answer I imagine that it is possible that these two passes were programmed separately, maybe to replace the roughing end mill with the finishing end mill, so the machinist could have ran the roughing code on that nut, and then s/he forgot to run the program of the finishing pass. But, I am more curious about your Cause # 3, even that you deemed it unlikely. Does it ever happen that a CNC deviates for a few seconds from the path that it has been following for the entire shift, and then go back to normal and continues in the expected path? The possibility of this erratic behavior on a machine is an interesting topic to study, I think.