How to measure "1 in 20 parts"

[therustygear]

I work in a foundry, and a customer of ours required 1 in 20 pieces be tested for mechanical properties. We can only pour 8 pieces at a time, and have no serialiation to identify "this is the 20th piece, this is the 40th piece"

How the heck can we know when 20 pieces have been made when anyone only ever sees 4 or 8 pieces at a time, interspersed between dozens (sometimes hundreds) of other parts?

I know serialization may be the answer, but I've been told that "is not a viable option at this point" . . .

 

[CoryPad]

Where do your pieces go after they have been poured, solidified and removed from the mold/die? Conveyor, bin, floor, ??? Are they large parts, and it takes a long time to make more than eight?

How do you provide the parts to your customer? I bet the shipping department knows how many parts go out the door.

Worst case scenario, you estimate how long it takes to make/send 20 pieces and take one from that time period for your testing.

Regards,

Cory

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips Fora.

 

[MadMango]

I would contact the customer and see if their requirement can be changed to 1 part per your lot of 8 pieces.

"Art without engineering is dreaming; Engineering without art is calculating."

Have you read faq731-376 to make the best use of Eng-Tips Forums?

 

[dingy2]

Most foundries check the mechanical properties per heat number and I am pretty sure you have those.

One piece in 20 is overkill for sure.

Dave D.

http://www.qmsi.ca

 

[therustygear]

CoryPad -- after shakeout, the gates/risers removed and the part is heat treated. After heat tratment, it is moved to finishing/grinding where flash etc are removed and fit is checked. This is likely the point at which we would send out the keel block (sample) for testing (after heat treatment)- The shipping department may know how many are shipped, but we need to know when we are pouring if we need testing because we need to cast a representative test block at the same time as the part.

These are very large castings (450 lbs) and 8 is the most we have the capacity to pour at once ie one heat = 8 pieces MAX

MadMango -- I considered your option, but that is alot of added cost, testing more than twice what the customer requires. If anything I would ask if we could test one per order, which is likely less than 1 in 20.

dingy2 -- the only mechanicals we check per heat is hardness. The #1 thing we produce is wear parts for mining applications, so most of our customers are more concerned with wear performance then tensile strength or anything else which is why these types of requirements are new to us.

 

[dingy2]

OK, got the picture but I will have to say that moving a 450 lb casting around for brinnelling may be a pain.

I remember a long time ago, we would surface grind with a hand grinder the brinnelled area and then had a portable machine and it worked out pretty well.

I can understand the customer requirement now.

Dave D.

http://www.qmsi.ca

 

[CoryPad]

What is the time delay from order, to pouring, to heat treatment, to shipping? Why can't you have the shipping department keep a running tally of shipped parts? When this reaches 12, you know you need to put a test block into the next pour of 8 parts.

Regards,

Cory

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips Fora.

 

[KENAT]

Just to throw a spanner in the works. Measuring the 20th, 40th, 60th etc part is statistically not always the same as "1 in 20'.

I once got lectured by a customers QA on how the '1 in 20' should be picked at random.

Your customers requirement may better define this.

KENAT, probably the least qualified checker you'll ever meet...

 

[pwtomlin]

Some sort of log where you track how many are poured? Or possibly better, a log based on orders.

 

[PSE]

You could try randomly or periodically, depending upon the customer requirement, inspecting 4 samples out of every 10 pours. My personal preference is random sampling as operator's tend to pay closer attention to running a part that they know will be tested which can skew results.

Since you are providing data to the customer, I also trust/hope you are utilizing the data yourself for feedback into process improvement and control. Make the expense of the testing work for you and add value if possible. If you are consistently able to show the customer that your are meeting or exceeding their requirments, perhaps they would be willing to lower the testing rate.

Regards,

 

[jackdaniels]

I fully agree with PSE. My experience, although not on large items such as those indicated, is that it is always preferable to understand the process capability than isolated samples of the process.
I also agree with the 'golden' job - this is watched by all and sundry, so everyone works exactly to procedure and takes great care. If this is representative of 'normal' production at your place, you are very successful (or lucky!).

My advice is for you to want to understand your job and process, this will give you the knowledge to establish a sampling regime that your customer will hopefully buy into due to your wealth of confidence in the product you make.

By the way, my company insist on mechanical tests for each piece we order. It annoys the supplier but we have little confidence (demonstrated information) from them.

"Help yourself, to help others"