Validation of Processes 7.5.6 - Do cosmetic welds need process validation?

[GoSpeedRacer]

We are a small contract manufacturer and have been building a medical device system (class II) for one of our customers for more than 10 years. They have now decided that it would be good to evaluate our special processes, ensuring that we have validations for all processes where "the resulting output cannot be verified by subsequent monitoring or measurement. I understand this requirement in theory, however in practice we are working on things like the welds on the corners of the sheet metal control box that are purely cosmetic. The weld could fail completely and the control box would continue to function perfectly.
I looked throughout the standard trying to find an out. I don't want to go to our sheet metal fabricator and develop a validation for this process. Could a risk assessment show that there are no hazards associated with this weld and therefor give us relief from 7.5.6?
Any help appreciated.

 

[Sidney Vianna]

however in practice we are working on things like the welds on the corners of the sheet metal control box that are purely cosmetic. The weld could fail completely and the control box would continue to function perfectly.

This type of welding requires NO validation. I even used a similar example to make my point in the Welding - Is welding always a special process per ISO9001:2000 Clause 7.5.2 thread.

 

[Ronen E]

Process validation is required where verification of results is impossible or impracticable (it can also be used voluntarily, where process validation makes more economical sense than 100% verification).

What is verification? It's obtaining evidence that a specified requirement has been met.

What is the relevant requirement in this case? For discussion sake, I'll phrase it as "the weld must have a visual appearance of being complete, i.e. no voids, cracks etc. are allowed." Please feel free to adjust to meet the actual case. My wording is based on the statement that the weld is purely cosmetic and has no structural role.

Can this requirement be verified? Certainly - through 100% visual inspection. If 100% is unfeasible or not economical, you might be able to justify a sampling plan; however, it must be very robust statistically, so that it provides close to 100% confidence (I would say >99%, preferably >99.5% confidence). It might sound very burdensome but that's the deal with verification.

Where process outcomes can be, and are in fact, verified, process validation is not mandatory.

The key here is that you never validate a process "in general", but to a specified outcome. This is where formalised Design Input (AKA engineering product requirements) becomes very important.

Cheers,
Ronen.

 

[pkost]

I agree with both comments above, there is no need to validate, however I think Ronen_E is being a little strict on the sampling confidence level required, I suspect a risk based approach could justify a low threshold or a minimal inspection method