$FS = \dfrac{\sigma_u}{\sigma_{allow}}$ — F S equals sigma-u over sigma-allowable. $\sigma_u$ is the strength of the material, the stress at which it yields or breaks; $\sigma_{allow}$ is the working stress the design actually puts into it; $FS$ is the factor of safety, a bare ratio with no units at all, because it is a stress divided by a stress.

Read it as a declaration of ignorance rather than of strength. A factor of safety of 3 does not say the part is three times better than it needs to be; it says the engineer is keeping a threefold reserve against the load being heavier than assumed, the steel weaker than certified and the weld worse than drawn — all on the same bad morning. $FS$ below 1 means the working stress has already passed the strength: the part is failing on paper. $FS$ at 1.4 means it survives the load you wrote down, and nothing else.

Then a stress nobody applies. Heat a bar and it wants to grow by $\alpha \, \Delta T$ per unit of length; hold both ends and it cannot, so the stress arrives anyway: $\sigma = E \alpha \, \Delta T$ — sigma equals E alpha delta-T. $\alpha$ (alpha) is the coefficient of thermal expansion, about $12 \times 10^{-6}$ per °C for carbon steel; $\Delta T$ is the temperature change from the condition it was installed at; $E$ is the same Young's modulus as ever. Notice what is NOT in that formula: length. A restrained run builds the same stress at two metres as at two hundred — which is why expansion is designed around anchors, never around lengths.