A process capability analysis checks whether the entire production process is capable.
We remember the How To for machine capability: All "variables" should be constant. Same batch of raw material, same worker, same operator... The variants now come into play in the process analysis.
But what exactly is the difference to machine capability?
In the case of process capability, individual subgroups are drawn. The sample size specified on the characteristics mask is decisive for the calculation. A classic sample size is "5".
This specification "5" therefore defines how many components are removed per time unit. E.g. 5 components every hour.
Each 5 component, the individual subgroup, now again follows the rules that apply in a machine capability analysis: There must be no changes to the "variables" within a single subgroup.
If the values mask is switched to the individual display of a single characteristic, this sample number can be displayed here:
And using the example of the additional data field "Machine number" (K0010), you can see here how a subgroup has the same additional data
The change of such additional data can be displayed in the value chart:
A mix of different additional data could be seen in the QCC, on the differently coloured separator bar:
But why is subgroup compliance so important?
The values of the individual subgroup are used to calculate the "estimators of location and variation" of the sample.
And these estimators of location and variation of the individual samples are used in the evaluation strategy to perform the first check of stability.
The correct adherence to the values within a subgroup therefore has a significant influence on the calculation, the defined distribution shape and the subsequent quality control chart, and defines in new strategies whether it is a C-value (stable QCC) or a P-value (unstable).
Now the next buzzwords have already come up. Distribution time models; QCC; distributions...
The process analysis defines much more. It is the beginning of the big topic of "SPC". Statistical Process Control.
But what is "the process"? Is the process as shown in the example above to mix the measured values of productions on different production machines the process, or can in the later course anyway only each machine be controlled for itself? With the aforementioned control charts?
Continuing the fictitious example from above: this "process" is considered unstable. Because the measured values from machine 5 always deviate slightly.
However, it is just as easy to define that each combination of part and production machine is a separate process. In this fictitious example, all "individual processes" would be stable in themselves.
These basic definitions (what is actually the process) are part of the K-field workshop.
Link to K-field workshops
What is the process? What can be regulated? And only then can we start with process capabilities.
The software can of course "fold" each data set, separate it according to additional data, summarise it and much more. However, quality control charts can only be saved per effectively saved characteristic.