Ask the Stamping Expert: Are we taking the best approach to meeting statistical capability on a die set?

Q: We are attempting to meet statistical capability with a minimum process capability index (CPK) of 1.33 on six stamping dies as one large data set. All six dies produce parts that meet capability individually, but not as a group. For this to occur as a group, all six stamping tools would need to act the same all the time for the rest of the program’s life, which is almost impossible to do, especially on difficult dimensions. For this reason, we sometimes keep data separate on programs that have multiple stamping dies or dies that have two or more stampings per stroke, like a back and front or a left and right. Thoughts?

A: You are correct that it is not practical to have multiple tools/fixtures/machines meet capability if they make the same product and have combined data for any one feature, especially on very tight-tolerance dimensions. You are almost forced to chart each one individually.

Most of our customers—medical, automotive, aerospace, electronics—have critical dimensions on their prints that require us to record their minimum statistical capability. In fact, most require this data to be sent with every shipment, because many of our products go into assembly machines, and assembly machines want every part exactly the same. If a minimum CPK is not met, the product is not accepted.

Even if you are shipping two lots of product that are made by identical machines or stamping tools, it is possible that a critical dimension can meet statistical requirements but be significantly different. This can easily happen if the range of the data is very tight but not centered on the same nominal number.

With that said, let’s look at a typical example. Assume you have two stamping dies to meet the volume required, or possibly one stamping tool that puts out two parts per stroke. Let’s say you have a critical dimension on the part with the spec of 1.0 +/-0.002, capability of CPK 1.33 required. The first capability study of that dimension from one of the parts reveals that the data runs at a mean of -0.001 with a standard deviation of 0.0002. The second capability study of the same dimension from the other part reveals that the data runs at a mean of +0.001 with a standard deviation of 0.0002. Both individually meet capability (with an awesome standard deviation!), but if you put the data together, they would fail capability. And if you have 10 stamping dies, this problem gets magnified and you’re forever chasing your tail on which die to adjust. But the customer wants everything the same, every time.

To meet customer volume requirements, we often have run multiple-part-out or even two identical, duplicate stamping tools. The customer feeds the parts we produce into assembly machines that bring together six to eight other components and drop off one assembly. These machines are tweaked to run fine with parts that all meet capability, but when the customer dumps in a batch of the “same product” from a different tool, the machines jam up. Even though the parts meet capability, they are different.

The problem lies in the stackup dimensions of all the components. Imagine if all eight to 10 components in the assembly ran on the very low end of the tolerance. This would be very different than if all the components ran on the high end of the tolerance. Most likely, the assembly machines would then have to be retweaked.

Theoretically, this should never happen, by design. If the product engineers on the customer end have done a thorough job, the assembly equipment should be able to take the full range of product variation as long as it meets capability. Unfortunately, that is most often not the case.

The lesson here is that when you make the same components from different tools, be sure never to mix them. You should be able to determine which parts you deliver to the customer were made off which tools, at any point in time. That way, the worst-case scenario is that the customer runs the parts in batches and might have to tweak the assembly machines when going from lot to lot of parts. That’s better than the customer having to babysit the assembly machines because the components have significant variation. Of course, balancing cost and inventory levels and increasing the batch sizes also help minimize this issue.