Analysis of the Partial Safety Factor Method Using Probabilistic Techniques

Abstract

Partial Safety Factors (PSFs) are scaling factors which are used to modify the input parameters to a deterministic fracture mechanics assessment in order to consider the effects of variability or uncertainty in the values of the input parameters. BS7910 and SINTAP have adopted the technique, both of which use the First Order Reliability Method (FORM) to derive values for PSFs. The PSFs are tabulated, varying with the target probability of failure, p(F), and the Coefficient of Variance (COV) of the variable. An accurate assessment of p(F) requires a probabilistic method with enough simulations. This has previously been found to be time consuming, due to the large number of simulations required. The PSF method has been seen as a quick way of calculating an approximate, conservative value of p(F). This paper contains a review of the PSF method, conducted using an efficient probabilistic method called the Hybrid probabilistic method. The Hybrid probabilistic method is used to find p(F) at a large number of assessment points, for a range of different PSFs. These p(F) values are compared to those obtained using the PSF method. It is found that the PSF method was usually, and often extremely, conservative. However there are also cases where the PSF method was non-conservative. This result is verified by a hand calculation. Modifications to the PSF method are suggested, including the establishment of a minimum PSF on each variable to reduce non-conservatisms. In light of the existence of efficient probabilistic techniques, the non-conservatisms that have been found in the PSF method, coupled with the impracticality of completely removing these non-conservatisms, it is recommended that a full probabilistic assessment should generally be performed.