We are a first and second tier supplier of Automotive Components, with the majority of of parts being produced by injection moulding. Historically, we have carried out capability studies on each cavity of multi- cavity tools. Due to some tools having in excess of 24 cavities, this results in a time consuming exercise, and 24 seperate cpk values. We are led to believe that it is more relevant to conduct a study taking random samples from all cavities, resulting in one cpk value. Does this represent an acceptable method of achieving process capability for multi-cavity tools?
Yes, a random sample should be a good method of determining process capability. However, theoretically you have 24 little distributions inside that random sample distribution.
Basically two items are of concern - what does your customer want and what is important to you. I know most automotive customers want individual by-cavity Cpk calculations and an overall calculation of all cavities combined. If during PPAP you determine that the variation between cavities is not significant then your SPC or other data collection could be from random samples.
Most of what I have seen in multi-cavity tooling is that there is always some cavities that are statistically different. Sometimes it matters and sometimes it does not.
Multi-Cavity Capability Studies
Linear Mode
