How to validate hundreds or even thousands of parts?

[mfgengineer]

Hi Forum,

I am well-versed in process validations, ISO13485, 21 CFR 820.75, etc. However, I am still scratching my head when it comes to efficiently/effectively validation a process that involves 1000+ parts.

A practical example may be a machine shop or an injection molding/extruding process. Technically, thousands of parts can be produced from this process (with different parameters and tooling, of course). I was wondering if anyone on this forum has any experience in tackling a situation like this?

The best approach I could come up with is categorize major part 'features' and validate the process that way, but even still, that might take me from validating 1000 'parts' to about '100' which is still unrealistic.

Any advice would be greatly appreciated. Thanks.

 

[ncwalker]

Take a step back. Don't validate the individual processes, validate the controls - the steps you take to NOT make a bad part.

Those won't be so many.

 

[mfgengineer]

Let me try to clarify the situation:

And I may be inaccurate because I don't work with injection molding, but for simplification:

You have 1 injection molding machine
You have X number of fixturing/templates
You have 1000 different parts that your company manufactures
Obviously for each of the 1000 parts, depending on its design and size, the parameters such as resin mix, temperature, etc all vary.

In a perfect world, you would validate the manufacturing process for each of the 1000 parts. But in a realistic world, that is not cost-effective, nor exactly feasible. Has anyone had to deal with this situation? Any input would be appreciated.

 

[Ajit Basrur]

Let me try to clarify the situation:

And I may be inaccurate because I don't work with injection molding, but for simplification:

You have 1 injection molding machine
You have X number of fixturing/templates
You have 1000 different parts that your company manufactures
Obviously for each of the 1000 parts, depending on its design and size, the parameters such as resin mix, temperature, etc all vary.

In a perfect world, you would validate the manufacturing process for each of the 1000 parts. But in a realistic world, that is not cost-effective, nor exactly feasible. Has anyone had to deal with this situation? Any input would be appreciated.

Here comes the risk management approach. You can identify the process and product that could encompass the requirements across your product mix, define the validation approach and then execute accordingly.

 

[Bev D]

I have lots of experience in injection molding and machining etc.

first, these are not "special processes" and so the validation one would do is easier.

Second, each unique part number is in fact validated for most industries. certainly we require it as does such industries as aerospace and defense and medical devices where the parts have critical functions.

the extent of the validation is most certainly risk based and can range from a simple first article to full characterization of the process input space (referred to as "Operating Qualification in the medical device world) and full functional testing in the end system under real world use conditions...

It may seem large to you but suppliers of these parts know this is a requirement (although many don't like it as the validation must typically be approved by the Customer prior to implementation of the process and any subsequent changes) and those who buy these parts are willing to pay for it.

Speaking from my own experience in aerospace, defense and medical device industries, the cost of failures that aren't caught in validation far outweighs the cost of the validation itself.

 

[Stoic]

the extent of the validation is most certainly risk based and can range from a simple first article to full characterization of the process input space (referred to as "Operating Qualification in the medical device world) and full functional testing in the end system under real world use conditions...

@Bev D, I'm curious about this interpretation of how risk can determine the extent of process validation, when in the context of FDA regulations. To date, I've not read anything from the FDA, GHTF, or anyone else indicating that any part of IQ/OQ/PQ can be waived based on risk, when it's been determined that the process requires validation. Certainly there are implicit or explicit risk decisions involved in simply identifying the Critical-to-Quality characteristics which are evaluated in OQ and PQ, in determining which products in a given size/configuration matrix need to be tested, and in determination of sample sizes and acceptance criteria. But limiting "validation" to just a first article inspection? Could you please elaborate on scenarios in which this would be appropriate and the logic which supports the approach?

 

[Bev D]

we would have to read the entirety of my post for proper context. even in the portion you quoted I don't say that IQ/OQ/PQ can be 'waived' when it is required by the Customer or when there is a 'special process" or when the final characteristic is a critical characteristic. my response aslo focused on part validation not so much on process validation and it encompassed many industries as the OP was discussing injection molding which can supply many industries.
So called special processes must always be validated even in most non-FDA industries.
Ditto fro

There are some component parts that are engineered and fully characterized for how they effect a product. in the original design these are of course validated per IQ/OQ/PQ type of studies - we must validate that the characterization is correct. However for these parts when a change occurs it is often sufficient to simply perform a first article inspection (for parts across the input tolerance range - which some my refer to as OQ at least from the suppliers perspective) since if the changed part meets the drawing requirements we know it will function correctly. This is common in aerospace and automotive but not as common for medical devices. The key here is that the design space - which is multi-layered - is engineered, well characterized and validated in the original design. It also will depend on the severity of the effect of a failure of the part in the final system. If the effect of failure is trivial (again well known and validated in the original design - not just a guess), there is little need for more than a first article inspection.

Hope this helps?

 

[Stoic]

Yes, thank you for the clarification! This aligns with my understanding of the FDA QSR.