Abstract:
Partial safety factors present in codes are defined for the design of broad populations of structures. As a consequence, they do not always reflect correctly the uncertainties for specific existing structures. The possibility of adjusting them has therefore great interest when assessing an existing structure. The concept of design point connected to FORM offers a simple method to calibrate partial safety factors for individual
variables. This paper discusses the overall methodology and develops a set of analytical expressions for the probabilistic models more common in structural reliability.
As it will be seen the partial factor for a particular variable can be adjusted once defined: (1) probabilistic model used to describe the uncertainty in that variable; (2) coefficient of variation of the variable; (3) fractile
implicit on the characteristic value used to quantify the design value; (4) importance of the variable in the limit state under consideration (measured by the respective sensitivity factor), and (5) target reliability
index.
The choice of the probabilistic model influences significantly the partial factors and this influence rises as the coefficient of variation increases. As a consequence, more attention must be paid when choosing a probabilistic model for a variable with high coefficient of variation.
