Review and Evaluation of the Separation Factor Approach for Structural Reliability

Studies have indicated that resistance factors calculated using a first-order second moment reliability method that uncouples the load effect from the resistance, termed the “separation factor approach” (SFA), differ considerably from those calculated using more accurate methods that also consider statistical variations in both the load effects and resistance. This can be attributed, in part, to the SFA implementing a separation factor, α, equal to 0.55, which was determined from tentative loading criteria and statistics in the 1970s. This paper amalgamates the disparate literature/background on the SFA and investigates its sources of error to illustrate its inherent assumptions and limitations. Three studies are conducted whose results are used to recommend appropriate separation factors (with associated bounds) for use in the SFA when determining resistance factors for steel components.It is found that α = 0.55 was calibrated to an atypical range of live-to-dead load ratios and small values of VR, which undermines its applicability when used in conjunction with modern-day statistics. Despite this, α = 0.55 is found to perform well when the reliability index, β, is equal to 3.0. For β = 3.5 and β = 4.0, α = 0.70 and α = 0.80, respectively, give results that agree with more accurate reliability methods at a live-to-dead load ratio of 3.0.