Inspecting Total Runout

Q

qcchick

I am I silly for requiring the Total Runout be checked 100% on a new build prototype of a welded assembly? The assembly has many variables that contols cannot be determined in a single build cycle. Due to the complexity of the welded assembly and the fact that the runout tolerance is critical I feel it should be checked on each part. There are only 36 pcs. I am having a hard time conveying this point to one of our new engineers who thinks that if we build one part that "in theory" the next one will be exactly the same. (booksmart with very little practical experience). I feel if we can ensure that in the first build we are meeting the runout requirement that then we can establish process contols and decide on a sampling size for future builds. This assembly involves welding a rectangluar part that consists of 2 seperate parts with 2 welded inserts that hold a machined shaft. The shaft mounts to a bearing and the whole assembly will spin. The total runout tolerance is only .030" which isn't much granted the variables that could occur in the manufacutring process. Just would like opinions on how I can convey this properly with tact. This person's opion is that we should only have to check it one time and if it is good then the rest will also be good. (If it was that easy we wouldn't need process controls right?)
 
T

The Specialist

I am I silly for requiring the Total Runout be checked 100% on a new build prototype of a welded assembly? The assembly has many variables that contols cannot be determined in a single build cycle. Due to the complexity of the welded assembly and the fact that the runout tolerance is critical I feel it should be checked on each part. There are only 36 pcs. I am having a hard time conveying this point to one of our new engineers who thinks that if we build one part that "in theory" the next one will be exactly the same. (booksmart with very little practical experience). I feel if we can ensure that in the first build we are meeting the runout requirement that then we can establish process contols and decide on a sampling size for future builds. This assembly involves welding a rectangluar part that consists of 2 seperate parts with 2 welded inserts that hold a machined shaft. The shaft mounts to a bearing and the whole assembly will spin. The total runout tolerance is only .030" which isn't much granted the variables that could occur in the manufacutring process. Just would like opinions on how I can convey this properly with tact. This person's opion is that we should only have to check it one time and if it is good then the rest will also be good. (If it was that easy we wouldn't need process controls right?)


You are correct that you cannot rely on 'in theory' methodology for quality control.

It is possible that a small % (in process checks) could be used instead of 100% inspection, however, you will first have had to demonstrate through documented testing that you have a stable process (validation excersise).

IPC (% checking) can then be used mearly to demonstrate that the process is maintained (stable) during a specific process run.

If you intend to do a lot of these process runs, it may be worth taking this route.

If this is a one-off or rare process, it may be easier/less time-consuming to carry on with your current 100% inspection method.
 
M

Markaich

My understanding is that this would relate to the introduction of a new product to manufacturing.

As part of this introduction, you would need to collect sufficient data regarding the processes' capability (Cp & Cpk) of consistently achieving the 0.030" tolerance. While 100% may seem a large sample, on only 36 parts the sample size would need to be of this order to ensure that the statistical analysis was relevant.

If the process can be shown to be capable, then the number of parts measured can be reduced, with occaional samples taken using SPC to 1) show the continued capability of the process and 2) to warn if the process was going out of control.

In my experience, 'engineers' (particuarly those of the design rather than manufacturing variety) do not understand the concept of natural variation, irrespective of book learning or experience.

One way of reducing variation would be to machine the bearing holes after welding so that the variation intruduced through that rather variable process is eliminated.

Hope this helps
M
 

Jim Wynne

Leader
Admin
I am I silly for requiring the Total Runout be checked 100% on a new build prototype of a welded assembly? The assembly has many variables that contols cannot be determined in a single build cycle. Due to the complexity of the welded assembly and the fact that the runout tolerance is critical I feel it should be checked on each part. There are only 36 pcs. I am having a hard time conveying this point to one of our new engineers who thinks that if we build one part that "in theory" the next one will be exactly the same. (booksmart with very little practical experience). I feel if we can ensure that in the first build we are meeting the runout requirement that then we can establish process contols and decide on a sampling size for future builds. This assembly involves welding a rectangluar part that consists of 2 seperate parts with 2 welded inserts that hold a machined shaft. The shaft mounts to a bearing and the whole assembly will spin. The total runout tolerance is only .030" which isn't much granted the variables that could occur in the manufacutring process. Just would like opinions on how I can convey this properly with tact. This person's opion is that we should only have to check it one time and if it is good then the rest will also be good. (If it was that easy we wouldn't need process controls right?)

We've often heard of busy managers telling their underlings, "Don't come to me with problems--come to me with solutions." My experience has been that when it comes to engineers in general, and design engineers especially, the reverse is true. Even if you know the answer, it's usually best to let the engineer think it was his idea. While it might be too late in this instance, what you should do is go to the engineer with a dilemma--you have this runout requirement and there are multiple variables to deal with--and let her (with a little directing help from you) develop the solution.

As it stands presently, the best you can do is explain the variables and how before you can settle on a verification strategy you have to understand--through measurement--how they all should be controlled.
 
A

Aliveguy

I agree with you qcchick !! It is always wise to add inspections in an effort to show process stability and efficacy.

I suggest being easy on your work associates, as booksmart can help in some areas where field experience falls short. You may need their help later.
I like the fact that you posted your thoughts first instead of taking immediate actions. That is a lesson in of itself....

Good Luck...
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
I am I silly for requiring the Total Runout be checked 100% on a new build prototype of a welded assembly? The assembly has many variables that controls cannot be determined in a single build cycle. Due to the complexity of the welded assembly and the fact that the runout tolerance is critical I feel it should be checked on each part. There are only 36 pcs. I am having a hard time conveying this point to one of our new engineers who thinks that if we build one part that "in theory" the next one will be exactly the same. (booksmart with very little practical experience).

Wow. 1 pc. capability study. Thank you engineering schools. I feel a Charlie Sheen quote coming on here.

I feel if we can ensure that in the first build we are meeting the runout requirement that then we can establish process controls and decide on a sampling size for future builds. This assembly involves welding a rectangular part that consists of 2 separate parts with 2 welded inserts that hold a machined shaft. The shaft mounts to a bearing and the whole assembly will spin. The total runout tolerance is only .030" which isn't much granted the variables that could occur in the manufacturing process. Just would like opinions on how I can convey this properly with tact. This person's opinion is that we should only have to check it one time and if it is good then the rest will also be good. (If it was that easy we wouldn't need process controls right?)

Even a 36 pc build is not sufficient to determine capability on that process - unless that is going to be the typical lot size. Not only that, runout is a unilateral tolerance with a hard limit (zero), so capability is tricky. It is non-normal, and usually falls into a Weibull or beta distribution. Bottom line: not intuitively obvious.

I guess the natural question I would ask - and any customer would ask, also - how do you know ever other part is going to be the same? Has the engineer developed a total variance equation to prove that point? Has the engineer considered process controls?
 
Q

qcchick

I managed to convince everyone involved that we needed to check total runout on all 36 pcs to establish capability. It was a difficult process but thanks to all for good ideas to help. It was much appreciated!
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
If you want to post the data and get some opinions on how to analyze it...we can help you with that.
 
K

Ka Pilo

I am I silly for requiring the Total Runout be checked 100% on a new build prototype of a welded assembly? The assembly has many variables that contols cannot be determined in a single build cycle. Due to the complexity of the welded assembly and the fact that the runout tolerance is critical I feel it should be checked on each part. There are only 36 pcs. I am having a hard time conveying this point to one of our new engineers who thinks that if we build one part that "in theory" the next one will be exactly the same. (booksmart with very little practical experience). I feel if we can ensure that in the first build we are meeting the runout requirement that then we can establish process contols and decide on a sampling size for future builds. This assembly involves welding a rectangluar part that consists of 2 seperate parts with 2 welded inserts that hold a machined shaft. The shaft mounts to a bearing and the whole assembly will spin. The total runout tolerance is only .030" which isn't much granted the variables that could occur in the manufacutring process. Just would like opinions on how I can convey this properly with tact. This person's opion is that we should only have to check it one time and if it is good then the rest will also be good. (If it was that easy we wouldn't need process controls right?)
Try The Square Root of N Plus One Sampling Rule
 
Top Bottom