Reliability-based design of concrete-filled steel tubular (CFST) truss systems by advanced analysis

Abstract

Concrete-filled steel tubular (CFST) truss is a type of composite structure with CFST chords and hollow tubular braces. CFST trusses have been increasingly used in large-scale structures such as towers, bridge girders, piers and arch ribs. The compression and flexural behaviour of CFST trusses are greatly improved compared to hollow tubular trusses due to the concrete infill in chords. For such complex composite structural systems, traditional structural analysis approaches are readily applied for the safety checks of individual members and connections, whilst system-level inelastic analysis and reliability calibration are very limited. In the past, relevant studies on the reliability of CFST structures mostly focused on structural components such as individual columns and beams, and there is no mature system-based design regulations nor reliability evaluations on this complex composite system. This thesis aims to address this gap by adopting stochastic finite element analysis (FEA) modellings of CFST truss systems through numerical approach considering both the structural nonlinearities and random uncertainties. With the complex configuration, nonlinear material interaction and sophisticated construction process, initial imperfections may largely affect the strength and stability of a CFST truss structure. The deterministic and probabilistic studies of CFST trusses with random imperfections based on the statistics of imperfections obtained through experiments and on-site measurements are carried out first, which lays a foundation for further study on the comprehensive reliability analysis of composite truss structures with uncertainties in material properties, structural configurations, initial imperfections and model uncertainties. The verified FEA models of CFST truss systems (two-chord, three-chord and four-chord) are then further developed to implement random uncertainties, including the elastic modulus of steel (Es), the yield strength of steel (fy), the thickness of steel (ts), the cylinder strength of concrete (fc’), the initial steel imperfection (χs) and the initial concrete imperfection (χc1, χc2), to facilitate a comprehensive reliability analysis. To consider the random variation of these uncertainties based on statistics, the Monte Carlo (MC) simulation and the Latin Hypercube Sampling (LHS) technique are incorporated on the basis of the above-described Abaqus modelling. A large number of stochastic FEA models of CFST truss samples are randomly analysed using Abaqus-Python technique with random variables considered to estimate the statistics of the flexural strengths of CFST trusses. Using the obtained statistics of system resistance, reliability analysis is then undertaken to calculate the system reliability indices (β) of the three typical CFST truss systems. Random dead loads and live loads are implemented to the FEA models. The structure reliability of the CFST trusses is evaluated in MATLAB using the First-Order Reliability Method (FORM). The relationship between reliability indices (β) and system resistance factors (ϕs) under various load cases is obtained. Finally, a comprehensive reliability-based design guideline of CFST truss systems by advanced analysis is suggested. The proposed novel computational approach and the reliability analysis contribute to both the practical design and the standard drafting for CFST truss systems.