That Day, the 1.08 Figure on the CPK Report Made the Production Line Engineer's Face Turn Pale
I still remember two years ago, it was a Friday afternoon, and we were just about to finish work when Ah-Zhe from the production line rushed over, his face pale. He exclaimed, "Brother, look! The CPK report for this new batch of products just came out at only 1.08! The client requires at least 1.67; what do we do now?" I took a look, and my heart sank. This new product batch had very tight tolerances and was measured using the latest non-contact laser scanning machine, yet the data was so poor that the entire shipment was effectively halted. Everyone started pointing fingers at the machine, claiming the laser scanner was inaccurate, and some even wondered if the program was written incorrectly. But honestly, this new laser machine cost a fortune; how could it perform like this?
If CPK Is Insufficient, Is It Poor Product Quality, or Inaccurate Measurement?
In reality, the problem Ah-Zhe encountered, when stripped down, was simply inadequate "Measurement System Analysis" (MSA). When you see a poor CPK, what's your first thought? Product defects? Poor process capability? But sometimes, the real reason might be that your "measurement tools" are not up to standard. Especially now, with the increasing use of non-contact measurement methods like laser scanning and 3D scanning, while they are convenient and fast, how do we ensure they truly yield accurate data? This is the key to MSA for non-contact measurement.
In other words, if your measurement system itself has significant variation, then no matter how well your product is made, the measured data will fluctuate, and the CPK will naturally not improve. It's like using an inaccurate ruler to measure something; the results will, of course, be unreliable.
How to Conduct MSA for Non-Contact Measurement?
Our problem that time was later found to be due to poor "Repeatability" and "Reproducibility" of the laser scanning machine.
- Repeatability: Simply put, it's whether the data remains consistent when the same person uses the same machine to measure the same object multiple times. At that time, our laser scanner's data fluctuated significantly just by repeatedly measuring the same point. For example, if the same point was measured 10 times, theoretically it should be 100±1um, but it gave us readings like 100um, 102um, 98um, 105um, showing large variation.
- Reproducibility: This refers to whether the data remains consistent when different people use the same machine to measure the same object. Non-contact measurement less commonly encounters operator variation, but if 3D scanning requires manual placement of the workpiece, different operators' placement postures could lead to reproducibility issues.
So, here's the key: for MSA of non-contact measurement, you need to conduct Gage R&R. Yes, just like with traditional contact measurement, but there are some nuances in operation.
First, you need to prepare a sufficient number of "master parts." The dimensions of these master parts must be "golden samples" measured with very high-precision instruments. Next, you need to select multiple measurement points. For instance, if you're laser scanning a phone case, you can't just measure one point; you need to measure different features like edges, corners, and flat surfaces.
Then, have your machine repeatedly scan these points for multiple rounds. If your 3D scanner requires manual part placement, then have multiple engineers take turns placing and scanning the parts. Finally, input the data into Gage R&R software for analysis. If your GRR (Gauge Repeatability and Reproducibility) percentage is too high, exceeding 10% or even 30%, then your measurement system is compromised. For instance, in our case, the GRR percentage directly soared to 35%, no wonder the CPK was only 1.08! The estimated DPMO also surged to 6210.
The Most Common Pitfalls: Software Settings and Environmental Interference
Frankly, the most common pitfalls in non-contact measurement are often not hardware issues with the machine, but rather "software settings" and "environmental interference."
In the case of our laser scanner that time, it was later discovered that the "point density" setting for scanning was too low, leading to insufficient sampling and naturally unstable data. Additionally, the laser's "power" and "scanning speed" also affect accuracy; these parameters must be optimized through experimentation.
Furthermore, environmental issues are often overlooked. For instance, laser scanning is highly sensitive to "temperature" and "vibration." One time, when an adjacent machine vibrated a bit more, the measurement data started to drift. At that time, I even joked, "Should we ask the production line aunties to walk more lightly?" It was later discovered that the machine lacked proper vibration damping. The same applies to 3D scanning; changes in lighting and background noise can affect scanning results. Therefore, don't assume that everything will be fine once the machine is purchased; subsequent parameter tuning and environmental control are the real skills.
One Thing You Can Do Today
Quickly go back and check your non-contact measurement machine's Gage R&R report; it might be long overdue for an update.