0

I work in the aerospace industry and would like to utilize simple, mass market COTS assemblies as critical components.

Let's assume it's a simple solenoid valve.

Rather than create a contract or determine all of the criteria the seller qualifies their components against, let's assume that we are buying a quantity of them blindly. I need to purchase X quantity for the next set of products we are creating, but I know nothing regarding traceability and where (what lot/batch) each subcomponent in the valve comes from.

Now let's assume that I want to determine, with a Y% certainty, that the X amount of components will survive our qualification testing. How many, Z, components to I need to test?

Is it possible to create a testing scenario for this? Something that allows me to vary X, Y, and Z in order to find a combination that satisfies whatever our needs are? Is it possible to create a binomial distribution with this information? How about utilizing an attribute sampling test or some form of AQLs (acceptance quality level)?

I'm completely new to quality engineering, so even topics to explore would be helpful. Thanks!

Jee
  • 101
  • 1
  • You should try to get away from the idea of random and instead try to consider the concept of removing bias- that the criteria for selection be as far removed as possible from the pass/fail criteria. One way to completely remove selection related bias is to select everything. – Abel Jan 10 '24 at 12:28
  • To answer your question, you'd need to make some assumptions about the distribution of defects. It's good to be wary when doing this. Defects have a tendency to cluster (not always but often), and this adds the challenge of controlling batch-to-batch vs within-batch defect rates - especially as the source may be an upstream supplier the supplier you buy from (e.g. if you're buying a cable, maybe it's the manufacturer of the inner wire, and once in a while there's a bad spool). Ofc you still need a test plan so you have no choice but to make some assumptions. – Pete W Jan 15 '24 at 14:58
  • Also, if literally using non-aerospace COTS items in an application that aerospace industry calls "critical", then 100% testing may be called for. IME it's fairly easy to defend this choice to management. It adds mfg cost (# of cycles of the test, someone else's department) but not necessarily engineering effort (cost to your boss) since you have to design/spec/explain the test either way. – Pete W Jan 15 '24 at 15:03

1 Answers1

2

Standard 9138 will define your testing requirements

Of course there is math available to answer your question. I suggest that because you have a very demanding environment you should only use that math if you are willing to define the math yourself.

In non-aviation environments, this process is known as Acceptance Quality Limit (AQL). As near as I can tell, this isn't used in aviation because the correct number of failures is zero. This takes the math and puts it into charts to tell you the sampling required.

Standard 9138 (from various organizations like SAE or IAQG, you should determine which organization is yours, there are various versions across the globe, but all are essentially similar). They take the math and help you apply it to the aerospace industry. A knowledge of statistics is necessary to navigate this standard. I would recommend some type of training/certification in this standard.

Tiger Guy
  • 7,376
  • 10
  • 22