Practical experience validating CNC Mills


Trusted Information Resource
If you are being forced into an IMDRF (I always think of GHTF!) approach, and are committed to it (in principle), based on this OP comment, I want to believe that it won't be too hard to satisfy an external reviewer (auditor or customer) or internal customers (operators, shop leads, maintenance techs).

I have seen a few posts on the forum regarding validation of CNC mills. Most discuss the general aspects of process validation, which is not particularly helpful. I think that most likely the critical process parameters are tool feed and speed. These parameters are different for every cutting tool and come from the tool manufacturer's recommendations. There might be 50 different tools, so this alone is no small task. Possibly, I can use retrospective data to satisfy the PQ. We should have data from parts that were cut with each tool that may show good process capability. However an OQ will be more challenging. Establishing process range for feed and speed on every cutting tool and for every mill will take the mills down for a signficant time. It's not likely that this is feasible. Possibly, I can establish a machine equivalence so that full validation is only required on one mill and an abbreviated validation on the other 20 mills. Still, the job seems daunting. Anyone else have any practical experience actually performing a CNC mill validation in a regulated medical device environment?

If you've been collecting data from each part, I want to believe that you are in a good position to do some retrospective analysis and make an assessment of the job you have been doing, along with an assessment in the "control chart" space (imagine me doing some hand-waving on the last bit).

In the IMDRF model, a natural approach would have you perform a process qualification for each new part coming out of the CNC. Since you already have ~50 different parts (and historical data), I would
  1. Identify the (operator selectable) machine/process parameters which have allowable ranges such that you can get the largest variance in the outputs - might require some head-scratching analysis -
  2. Identify the specific (key) elements of specific parts that are the most sensitive to those variances (it could be different specifications on different parts)
  3. Run the OQ against (only) those parts and inspect those key characteristics... I'd do a 100% verification anyway as a sanity check, but I need to see the key characteristic data to assess the OQ.
The individual tooling for each part could factor into the decision, especially because you don't want parts of the "OQ phase space" to be ignored, but I would not sweat specific angles of attack etc. unless your historical data is showing some peculiar variances.

I'm also assuming the same material for each part in this approach, different (intended) materials could impact outputs... I'm not trying to be an expert I'm just thinking about sources of known/controlled variability.

This approach is roughly what I did a lifetime ago to validate a wave solder process that was being used to assemble approximately two dozen different PCBA. I did not have to run the OQ on each PCBA, but first I did have to do a deep dive into each assembly to understand which features of which parts would be most sensitive to the process controls. This may not come as a complete surprise, but the best outcome was that we got rid of the old, worn-out carriers as they were introducing 100% of all process flaws. I want to believe most CNC operators are better at taking care of their worn tools!


Quite Involved in Discussions
I have learned a little more since my initial post. Luckily, we only machine in one material, which makes this somewhat easier. It seems the process parameters are spindle speed, cutting depth, cutting width and feed speed. The tool manufacturers are able to calculate the tool surface speed and feed per tooth on any tool based upon these parameters. It seems like it might be possible to perform an OQ for the process range of surface speed and feed per tooth. This would dramatically reduce the extent of validation over trying to validate every parameter for every cutting tool. Frankly, validating all the inputs seemed close to impossible and very difficult to maintain. New tools are introduced all the time and every new part program has different cutting depths, widths and feeds. I still need to figure out how to ensure mill programs don't exceed the validated tool surface speed and feed per tooth. Anyone have any experience validating in this manner? I fear that I might be reinventing the wheel here, but I haven't found much specific information on how others have validated a CNC mill. Most of the information I find is very general and applicable to any process validation.
Top Bottom