System reliability of structural steel frames: Component- and system-based design methods

This study investigates the system-level reliabilities achieved by two component-based design methods, the Direct Analysis Method and the Advanced Elastic Analysis Method, and two system-based design methods, the Advanced Inelastic Analysis Method and the Direct Design Method. A series of benchmark structural steel frames were first designed using a structural design optimization framework and system reliability analyses, which included uncertainties in geometric properties, material properties, and applied loads, were then performed on the resulting designs using the Importance Sampling technique. The component-based design methods consistently produced system-level reliabilities exceeding target levels, with mean reliability indices of 3.65–3.75 and high variability (coefficient of variation of 18.7–20.6%), but yielded designs significantly heavier than those produced by the system-based design methods. In contrast, the system-based design methods produced significantly lighter designs – on average 13–17% lighter than the Direct Analysis Method – with more consistent reliability closer to target levels, with mean reliability indices of 2.8–3.1 and lower variability (coefficient of variation of 6.1–9.9%). A sensitivity study further showed that recalibrating the material reduction factor of the Advanced Inelastic Analysis Method from 0.90 to 0.85 raises the mean reliability index from approximately 2.85 to 3.15 at the cost of an average weight increase of only 3.4%. Based on these findings, recommendations are provided to improve system-level reliability calibration procedures and to support the implementation of system-based design methods in future design codes.