Hi,
I was more thinking of the equipment qualification (installation qualification and operational qualification for the equipment part). All part and therefore process-specific things come later (like operational qualification for the process and process validation).
In my understanding this is a wrong interpretation of what process validation is and how it is structured. I'm not familiar with "equipment qualification". The first stage in process validation is installation qualification. It's true you look at the equipment in that stage, but it is in the context of a specific process and therefore also specific parts. I'm not aware of an "equipment validation" during which a blanket coverage for the equipment - no matter what parts it produces - is achieved. I'm also not aware of an operational qualification other than that of the a specific process and parts.
The CNC machine must be shown to fulfil acceptance criteria (e.g. accuracy along x1 x2 y .....axes, rotational axes, spindle speed, etc).
Yes, this is something that would normally be looked at during installation qualification, but typically through records from the machine manufacturer and provision of evidence that your specific installation has been carried out and completed according to all instructions from the machine manufacturer; such instructions are supposed to guarantee that the machine will perform accurately within its specification. It's the machine manufacturer's responsibility to establish such specification and instructions, and ascertain that they will produce the required results (part of the machine development). Good manufacturers will either install the machine for you or escort / support / supervise whoever else is installing, and will eventually sign-off that the installation is satisfactory. They might take necessary measurements of the machine and/or reference machined part to verify. All these should become records in your IQ.
I am somehow lacking the arguments why this cant be done in simply producing 30 parts and evaluating their characteristics according to their drawing.
Whatever measurements are taken, at whatever stage, the analysis must be statistically rigorous and there should be justification on record why those measurements are representative of what you're trying to establish. 30 parts are sometimes statistically enough, but this should be shown in every individual context.
- so many different parts will run on the machine, it would be too complex to derive why one part covers all need for the other parts
This is true; normally you have to validate every process (part) on the machine, unless you can justify bundling some in "families".
- if the machine is regarded as a black box with good parts coming out of the process, later on it cant be assured that a state of control is maintained if the equipment itself is not measured
A state of control is proven through proper analysis of the process parameters and noises and proper design of the validation runs; and through statistical rigor. To an extent the process is always treated as a black box, because when noise is concerned the influences are never fully understood (in the physical or causal sense) and of course never controlled. Regardless, processes can and should be shown to be under control (and capable). If not, the process has to be substantially modified.
- the machine axes are used only at special points, only a measurement of the whole traveling distance can make sure that it works correctly (and will later on with new or other parts being manufactured on the machine)
The process should be validated across the entire machining envelope that is intended to be used in production (typically, in the center and at the corners), and this should be captured in the analysis and report. You're not supposed to use the machine for a specific process/part outside the validated envelope.
- in case of a repair or replacement the whole part measuring must be repeated, when the machine itself is evaluated/measured only the isolated measurement covering the repair part can be repeated
If the repair is substantial, the entire validation might have to be repeated. If it's more focused, perhaps a subset of activities would suffice, backed by a strong documented justification. This is applicable for each and every validated process (part / family) running on this machine. Typically you'd be able to justify revalidating just a few processes and argue that given the nature of the repair they are representative of all current processes.
Cheers,
Ronen.