Abstract
This paper presents a system-based model for the fatigue assessment of old, deteriorating riveted railway bridge connections, which are built up of a number of basic components. A finite element model of a typical, short-span railway bridge is used to convert train loading into probabilistic fatigue load spectra via Monte Carlo train simulations. Uncertainties in terms of loading, resistance and modelling are taken into account. Fatigue damage calculations are based on the Theory of Critical Distances, which is a recently developed theory that considers the entire stress distribution ahead of any given stress concentration. The paper focuses on a typical riveted stringer-to-cross-girder connection which is treated through generic sub-systems that capture potential damage in identifiable hot-spots, such as rivets, holes and angle fillets. By treating these hot- spots as the elements of a structural system susceptible to fatigue failure, its reliability over time is evaluated using system reliability methods. The results show that the probability of failure of the connection depends significantly on the form of the system adopted for the analysis. Low remaining lives on specific hotspots on different connection components show that fatigue cracking may be imminent or may have already initiated in a number of similar existing bridge connections.