A manufacturer can typically process 176-264 devices a given period of concern. In this situation the processing event is either pass or fail. There is apriori info that on any given set of ~20 devices approx 1device is likely to fail( i.e. binomial w/ .95 prob of success). It is desired that the process should have a process reliability of .995 (which allows approx 3 out of 176 to fail.) If we want to have a confidence level of 90% for a .995 process reliability what is the necessary sample size to achieve this.. given the .95 prob of success for a typical batch of 20 devices? What is the preferred method of approach?