# Why loads in serviceability limit states usually don't include factors of safety?

For example, characteristic combination found here:

It can be seen that the dead loads $$G_k$$ are taken as the characteristic values instead of multiplying them by some factors of safety as in the case of ultimate limit states. Why are loads serviceability limit states taken as characteristic values instead of applying partial safety factors into their values?

## There's a reason why they're called "factors of safety"

When designing a structure, we need to make sure of two things:

• that the structure will remain a structure and not end up as a pile of rubble (i.e. that the structure is safe)
• that the structure can actually be used as intended.

The first is easy enough to understand: you either have a structure or a pile of rubble; do the math to ensure it's the former even under extraordinary circumstances (as described by factors of safety)

The second is more subjective... what does it even mean, really? Well, basically, that human beings will be able to perform whatever tasks they wish while using the structure.

Would you be able to sleep at night if you look up and the concrete beams are all cracked? If your bed bounces with every step your neighbor takes during their late-night walk to the kitchen for a cup of milk? Would you feel safe if you get to your office and the beam at the reception is drooping like Atlas' shoulders?

Of course not. So that's the purpose of serviceability limit state checks; they aren't there to make sure the structure is safe, that's the job of ultimate limit state checks.

Serviceability limit state is merely about whether the beam will suffer large deflections, and/or vibrations (and other checks, depending on the situation, such as cracking in concrete). And you only care about these under "normal" circumstances, which are described by the characteristic load values (if the structure undergoes an extreme, once-a-generation loading, do you really care if it deflects a bit too much for comfort?).

• To me it seems you've defined serviceability limit states, rather than answer my question. What I'm wondering about is this: factors of safety are there to take into account inaccuracies inherent in design, originating from inaccuracies in theory, and statistical uncertainty in things like material properties (for example, elastic modulus of concrete). So why aren't they present in serviceability calculations? If we are interested in deflection of an office roof, why do we take the loads as exact now? Why is the load deflecting the roof now exact, when in UTS calculations there is uncertainty? Jan 20 '21 at 21:48
• @S.Rotos: Keep in mind that the characteristic loads applied are already obscenely high. EN 1991-1 defines the applied dynamic load in a residential building as 1.5-2kN/m2. That's two people per square meter over the entire floor. It's nuts, when you consider that much of the real use of a residential area is actually unloaded. And even if it is loaded (with furniture or a person standing), it'll likely be under 2kN/m2. So just the characteristic values are already quite conservative. There's no need to add even more conservatism when you're not talking about safety.
– Wasabi
Jan 20 '21 at 22:39

Because the service load system, which is the old system being replaced by the LFRD method, limits the strength of the structural members like steel or concrete to a factor of usually around 60%. Roughly a safety factor of 150%.