## Question 316:

1## Answer:

No answer provided yet.If I understand the situation, if you want to estimate the fatigue point of a part, you want to know how many you need to sample. You also would like to keep it simple by using the normal distribution. Let me see what I can do.

First, I notice that your measure is number of cycles to failure, which is discrete, but in large enough numbers we can treat it like continuous data, just keep in mind that this will produce things like .5 cycles, which is impossible. My recommendation is to compute a confidence interval around your sample and use the end of the interval as your conservative estimate of the minimum fatigue level. You can adjust the confidence level to from 99, to 95, 90 or even 80 based on your need for precision. I'll provide an example:

I can provide an example but I need some idea of what the mean, standard deviation of these samples typically look like.