System reliability of semi-rigid steel frames designed with Direct Design Method

Abstract

Acknowledging the remarkable benefits and potential of advanced analysis methods, several steel structure design specifications, such as the American Standard AISC 360-16, have incorporated provisions to facilitate the application of these methods for system-based design. However, a key challenge in implementing this approach in practice lies in the determination of appropriate system resistance factors that fulfil the high-reliability demands of the structural system.

This thesis introduces a comprehensive framework for an advanced system-based analysis methodology applied to steel structures. The primary focus involves the determination of system resistance factors while accounting for inherent uncertainties in the ultimate strength of the frames. All essential parameters affecting frame strength are treated as stochastic variables, and Monte Carlo simulations are conducted to assess their impact. A significant contribution of this study extends the existing Generalised Component Method (GCM) to model the full-range behaviours of bolted angles connections, by introducing two new components, i.e., the slip component to model the bolts slip, and the gap component to model the contact of beam and column. Moreover, the GCM framework was employed to develop probabilistic models for various connection types. Furthermore, the semi-rigid properties of the connections were taken into consideration during the modelling of the steel structures.

The proposed methodology is applied to a diverse set of two-dimensional steel frames. Different connection types are selected to encompass various stiffness and strength properties. Through advanced analyses, the performance of the steel structure is assessed. Finally, the reliability of the steel structure is determined, and the relationship between system reliability and system resistance factors is established, facilitating the derivation of appropriate system resistance factors based on recommended target reliability levels.