Derivation of material factors

[LRJ]

I understand how to calculate the probability of failure using FOSM, FORM, Monte Carlo, etc. methods, and that different material factors will change the probability of failure, but what I want to know is how one might actually derive a material factor using a probabilistic method. Currently in design we appear to just apply the factors based on a codified approach which apparently has/is supposed to have a database behind it. Would it be possible to derive a material factor for design without such a 'failure database'?

 

[BigInch]

Already being done that way, but not as subject to probability as you might like. The preferred method is to be 99.975& safe.
Safe allowable stresses are typically set by statistical methods. If you test 100 specimens of pipe steel and the lowest yield stress encountered is Y, a safe allowable stress might be set at a certain percwentage, say 80% Y. If the material is one which is subject to brittle fracture, then a limit of 50% of the yield stress might be used. I say 99.975% because theoretically there is always some chance of failure.

 

[LRJ]

Hi BigInch. What do you mean by 'already being done that way'? What way?

What I want to know is if I could use a probabilistic method to justify an alternative selection of material factor for design. FCan I do with without empirical evidence? If I can, how do I demonstrate this? What is the method?