Hi,
a customer wants to reduce the effort testing plastic devices. To be more concrete:
They get 3 different plastic devices from suppliers (Top, Cover, Bottom), In the future 4 cavities will be delivered in one cast. The devices will be put together and then they test if it fits in the cavities. To be sure they will fit they would put each part in each cavity, which is much effort.They just check if it fits or not, so a binary Response.
They think using DOE could help them to reduce the testing if the parts will fit together.
My idea would be to start earlier and check if the devices follow the specs and test this first, before doing any final test on the cavities. If there is something suspicous the suppliers have to do their homework. But in case that's all fine, how can they check if there is more than just a random error that the devices will not fit with the cavities?
Theoretically I could create categorical factors for the cavities (Location in the one cast) with 4 Levels, and another with 3 Levels for the different devices. That could provide me a reduced number of runs compared to the full factorial, that is testing each device with each form of the cavity.
But I wonder if this makes statistically sense to try to reduce the number of tests using this DoE Approach.
Thanks for advice, hope I described it clear enough.
Martin