My opinion on this is fairly untapped in the industry. Also, Validation should help guide engineering studies to come up with documented and defended settings.
I think packaging engineering should follow a three step approach.
1 - Using a Design of Experiment (DoE), come up with a performance profile.
2 - Pick a result that you want (seal strength, speed, etc.).
3 - Run qualification at a few settings, and confirm that the results align with your performance profile (step 1).
Your DoE supports the limits of operation. Your qualification confirms the DoE. You have also tested “ranges” more than just high/low of each setting, you have also looked for attribute interaction, curvature, etc using WAY more robust statistical analysis than traditional. Overall you have shown that you understand your process and that it is in control, which is what validation is intended to do.
Alternatively. You could do the old school approach (I don’t love this method, but it also works).
1 - List all your parameters
2 - Lock down some parameters which WILL NOT change - so these are fixed at one set point, and don’t need to be challenged during qualification.
3 - Determine (thought engineering studies) appropriate ranges for all the remaining settings (not fixed in step 2)
4 - Determine which settings are “Critical” and will be qualified and which settings are “suggested” and can be changed based on equipment efficiency.
4a - for example rail position might be an efficiency setting and might be adjusted after some wear.
4b - for example temperature and speed (dwell time) might be a critical setting, and which needs to be controlled tightly.
NOTE: Packaging qualification can change due to differences in stock material. Somehow make allowances for that. A qualified setting for one lot of materials might not be optimal for another lot of materials.
Overall, your qualification DOES need to test the ranges of operation. But which ranges? That is up to you. Temperature and speed are some of the settings which require qualification of ranges. You asked “should EACH range be tested?” The answer is, how is that possible? There are so many ranges, you could not possible test all the combinations of each range - thus using statistical DoE is the better method to get the most information possible.
Do you guys have a statistician who can help set this up?