Reliability Index and Probability of Failure ASCE 7-16 Hurricane Winds 3

[MegaStructures]

See this paper published by NIST:

https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=917689

The author reports reliability indices for structures designed to ASCE risk categories II, III, and IV. Reliability indices can then be converted to a probability of failure. My question is how does ASCE use AISC and ACI "strength" values to construct these reliability values. I assume structures designed to LRFD and ASD will have different reliability values, as will concrete and steel structures, is the reliability index the worst possible case for all materials and design codes?

See also this table from ASCE 7-16

“The most successful people in life are the ones who ask questions. They’re always learning. They’re always growing. They’re always pushing.” Robert Kiyosaki

 

[ClearCalcs]

This article is a good summary of the reliability concept behind load & resistance factors for AISC:

https://www.aisc.org/globalassets/a...winners/load-and-resistance-factor-design.pdf

As you mentioned, reliability works both ways - load and resistance side. If you go through the math for say a single beam, you'll see that you're left with an equation at the end where for a target reliability index, your resistance factor is a function of your load factor. So technically, any combination of the two works as long as you're respecting that equation. The tough part is in generalizing to every beam, column, in every structure, out of every construction material. It takes lots of data and there is no one "right" solution. Thus, a bit of an arbitrary process was used, as far as I understand.

1. Existing buildings were surveyed and analyzed under ASD provisions to back-calculate a typical reliability index for existing structures
2. Factors on loads were determined somewhat arbitrarily. Something like: Load Factor = Bias Factor * (1 + k * COV) where k is a multiple of the coefficient of variation. I think it's typically been taken as 2, which corresponds roughly to a 5% probability of exceedance. That's why you see 1.2 on dead load vs 1.6 on live load - the COV is different for them. And it's also why you see 1.0 used for wind loads - instead of using this arbitrary formula, the wind loads are directly based on a mean recurrence interval which directly incorporates the 95% probability of exceedance.
3. Once the load factors were set, the typical reliability index for existing structures was set as the target reliability index for LRFD, and this was sent out to AISC, ACI, AWC, etc for their respective codes.
4. With load factors and a target reliability index, the equation I mentioned earlier is easy to solve to find the optimal resistance factor. So for instance the AISC would go over different failure modes & properties, find a bias factor and COV, and from there, back-calculate the optimal resistance factor for that failure mode. Failure modes with similar phi factors would be grouped together for simplicity.

Now with this all done, I think the general assumption is that most codes have been calibrated to maintain the same reliability index. So changing the "loads" side without changing the "resistance" side can only affect the reliability index, regardless of material.

I hope that helps - this was a nice excuse to back into my reliability analysis notes

http://www.ClearCalcs.com

 

[MegaStructures]

Great answer and very interesting! BTW clearcalcs.com is a beautiful website, well done

“The most successful people in life are the ones who ask questions. They’re always learning. They’re always growing. They’re always pushing.” Robert Kiyosaki