How low should you go for allowable stress in steel?

​I posted this question as a poll on LinkedIn.

It's a very simple question but the answer is not so simple.

​We all have our reasons and I don’t think there is only one answer but here are my thoughts.

​Stress Below Yield

Basically, if the nominal tensile or compressive stress on a steel component is below yield, it is safe provided that the loading is a single static load (not fatigue).
 
By safe, I mean there will be no permanent deformation but there may be localised yielding.
 
No one cares if there is localised yielding in a ductile steel component with a static load because stress raisers (holes etc) just redistribute the plastic stress over a wider area and a short distance away from the hole, it's elastic.
 
Fantastic!
 
But is it really safe?
 
Not if it's shear stress it ain't but if it's tensile or compressive stress then yes. Shear stress has an allowable of about half that of tensile or compressive stress.
 
But is it safe enough? 

​​Stress Below 2/3 of the Yield

Where did the number 2/3 of yield come from?
 
The thing is, loading is rarely a one-time event. It's usually hundreds or thousands of load cycles.
 
With a nominal allowable tensile stress of 2/3 of yield, it provides a bit of insurance that the part won't fail if it is loaded more than once.
 
There’s also a bit of a safety factor to allow for uncertainty.
 
Let's take Eurocode 3 as an example, a variable load has a load factor of 1.5.
 
Assuming we are using elasticity checks, double the load and you double the stress.
 
And 1/1.5 = 2/3
 
So, increasing the load by a factor of 1.5 is the same as reducing the allowable tensile stress to 2/3 of the yield.
 
You don't need to find stress concentrations around a small hole in a structural beam of a building because most buildings simply won't rack up enough loading cycles to cause any problems with fatigue.
 
In fact, a lot of the design codes specify the allowable nominal tensile stress should not exceed 2/3 of yield strength.
 
According to many design codes, tensile stress that is less than 2/3 of the yield strength is safe.

​Stress Below UTS

It's not a great idea to push nominal tensile stress levels all the way to Ultimate Tensile Strength (UTS). That would be like a candle in the wind, will it fracture or not?
 
On the other hand, it could be argued that if the component reached UTS in places but didn’t actually fracture, and it held the load, it is therefore safe.
 
That said, not all types of stress are equal.
 
For example, the bearing stress of a pin in a hole is not as important as the shear stress in a pin.
 
The worst that can happen if there is too much bearing stress in a hole, is you can elongate the hole.
 
The worst that can happen if you have too much shear stress in a pin, is the pin will shear off.

It really depends on how you define safe.

Depends on Lots of Factors

It does depend on lots of factors.
 
For a lifting beam carrying a ladle of molten steel. I wouldn’t be happy to be told the nominal tensile stress is 2/3 of the yield.
 
On the other hand, if a computer monitor arm had nominal tensile stress of 2/3 of yield, that would be more than adequate.
 
The design codes generally provide methods to calculate the allowable stress for the given application.

​​A Brief History of Design Codes

Design codes have a lot of history behind them.
 
Let's take a quick look at building design for structural steel for example focusing on direct tension.
 
Back in the 1800's, building structures were usually made from timber, masonry or cast/wrought iron.
 
At the time the recommended safety factor for tension/ bending (followed by some) was UTS / 4. Others assumed UTS / 5 based on an average UTS.
 
This seems high these days but was necessary due to common casting flaws and
the brittle nature of cast iron, since brittle materials snap rather than bend.
 
In the latter part of the century around the 1880's, rolled steel started to be used for channels, angles and I beams.
 
At first, the same permissible stress of UTS /4 was generally applied but the UTS of rolled steel was greater than cast/wrought iron. Therefore structures became lighter and leaner because the permissible stress was greater.
 
The first edition of BS 449 was released in 1932. The safety factor at this time was 1.8 for direct tension.
 
After the war and in 1948, BS 449 modified the safety factor to about 1.7 for direct tension.
 
The first edition of BS 5950 was released in 1985 and this design code reduced safety factors by about 10%.
 
Eurocodes have now replaced BS 5950.
 
The permanent load (dead load) factor was reduced from 1.4 in BS 5950 to 1.35 in Eurocode 3.
 
The variable load (live load) factor was reduced from 1.6 in BS 5950 to 1.5 in Eurocode 3.
 
So, the trend over time has been to reduce the factors of safety every time a new design code replaces an old one.
 
The upside to this is reduced construction costs and less steel is required.
 
The downside of this, is the design codes have increased in complexity with more load case combinations, rules and calculations.
 
These days there are many design codes for all sorts of things and the more dangerous the consequences of failure are, the greater the safety factor is specified in the design code.

​Final Thoughts

There are no design codes that I’m aware of that allow nominal tensile stress to exceed the yield strength, although some plasticity is acceptable in localised regions.
 
Quite a few design codes specify that the nominal tensile stress must be below 2/3 of the yield.
 
Therefore, if there is no design code for your application, I would use 2/3 of the yield for nominal tensile stress in ductile steel for a low number of cycles.
 
If the loading is severe, I would make sure the nominal tensile stress is below yield for a low number of cycles.

​For a once in a lifetime fault condition, it may be acceptable for some components to completely fracture or act like a plastic hinge provided that the main structure does not collapse. Or in other words, demonstrate that the structure fails safe for very rare and extreme loading events. The technical authority would need to agree with this approach beforehand.
 
If there is a design code available then I would follow the guidance from the code.

These are all just my personal opinions, there will be many good engineers who have very different opinions.