Test failure Root cause not found

Hi Everyone,

Dealing with a return of a unit that failed at the customer testing level. They sent it back and we can't figure out what caused the failure. We have a few potential causes but not a clear cause. Right now its just at the engineering level hasn't been risen to the quality level yet. We've delivered 600 of these units in the past 10 years and this is the first occurrence like this. How do I explain to the customer that no root cause was found? Do I list out the potential causes? Just don't want to open a can a worms where they get an idea to send back 100's of units for us to check when we are confident those are fine. Just having a hard time coming up with a sucient statement.

We are an AS9100 facility.
Thanks!

Are you unable to replicate the failure? Or was no fault/failure found when you tested the unit?

There isn't really instances of "no root cause" and, from a customer standpoint, that's the last thing they want to hear. There are instances in which failures cannot be replicated and you can provide proof of that. But if there was a true failure and the unit is not working within typical parameters, then something is causing an issue.

Try to repeat the whole engineering tests on this unit, may help this.

even at a purely "engineering level" or stage...correlation is key...replicate exactly the test setup is its electrical, if mechanical/dimensonal an agreed to QAF or Quality Acceptance Fixture is urged. The QAF is generally from or used in the automotive fields, but is key when solving issues.

mgoodman08 said:

- We've delivered 600 of these units in the past 10 years and this is the first occurrence like this.
- Just don't want to open a can a worms where they get an idea to send back 100's of units for us to check when we are confident those are fine

Click to expand...

Is your confidence based on anything besides the prior 600 units? That is, do you have some evidence that would show that this one defective unit could have escaped whereas no other defective units could have escaped? If not, then the customer's concern is valid.

mgoodman08 said:

Dealing with a return of a unit that failed at the customer testing level.

Click to expand...

Do you test the units? If so, how does your test method compare to the customer's?

mgoodman08 said:

They sent it back and we can't figure out what caused the failure. We have a few potential causes but not a clear cause.

Click to expand...

What prevents you from testing for the potential causes?

mgoodman08 said:

Right now its just at the engineering level hasn't been risen to the quality level yet.

Click to expand...

What does this mean?

mgoodman08 said:

We've delivered 600 of these units in the past 10 years and this is the first occurrence like this. How do I explain to the customer that no root cause was found? Do I list out the potential causes?

Click to expand...

You're assuming here that no root cause can be found. Is this the case? Have you exhausted all possibilities through testing?

mgoodman08 said:

Just don't want to open a can a worms where they get an idea to send back 100's of units for us to check when we are confident those are fine. Just having a hard time coming up with a sucient statement.

Click to expand...

You were apparently confident that the failed unit was OK too. I really don't think that "We can't figure it out" is going to instill confidence in the customer.

I have on at least one occasion been unable to determine the root cause of a failure beyond isolating the component in our assembly that failed. It was not for lack of effort, and we had the customer participating so they knew we were not just blowing smoke or trying to take the easy way out.

We (the customer, the component supplier, and my company) exhausted everything we could think of trying to determine how/why the component failed but couldn't. I think it was about a 1% failure rate IIRC. Because of the application, we ended up developing a rigorous screening test that would detect the rare failures, but never were able to identify and prevent the root cause.

One possibility is that the customer's test is different from yours. Or maybe something has changed so their test and yours are no longer equivalent. Annual calibration, by itself, is no guarantee that things can't change in the weeks and months after the calibration was done. You could suggest a round-robin test where the same sample(s) are tested by both your measurement and the customer's, and results compared.

Jim Wynne said:

What prevents you from testing for the potential causes?

Click to expand...

Another possibility is what I call a disappearing problem. We had one hard-to-solve problem with leaking assemblies. We suspected a pinched o-ring in the joint. When the leakers were disassembled, the o-ring snapped back into position and it was not possible to recreate the leak by assembling the same parts which previously leaked. There was no visible damage to the o-ring. There were only 4 leakers discovered out of 40000 assemblies shipped, so what sample size would you test to solve a 100 ppm problem?

We hand-made assemblies with intentionally pinched o-rings until we could pass end-of-line leak-test, and the leak became manifest later in the sequence, which proved the validity of this black swan. Then our investigation figured out how the pinched o-ring was occurring in the manufacturing process, and fixed that. No more occurrence of leaks.

John, your o-ring leak story caused the temporary recurrence of a tic I developed over a similar problem once. If I never see another hydrogen leak tester again, it'll be too soon.....

@Mike S., there was more to the story. The supplier was quite proud of their Helium leak tester, they told me it could measure a leak as small as 10cc/40 years. That was their assurance that they weren't shipping leakers. Their smugness disappeared when I showed them their $500K machine would pass parts with the o-ring completely missing from the joint if you put enough oil on the threads. I thought I would have to pick them up off the floor. I was able to document the pinched o-ring in the disappearing problem investigation by freezing the joint in liquid nitrogen, and then I could unscrew the joint and photograph the pinched o-ring before the rubber thawed. Total time on the problem, 4 days.