Markov modelling of the fatigue damage in welded structures under in-service inspection

Abstract

This paper describes a probabilistic cumulative fatigue damage model based on a simple Markov chain approach. The objective is to develop a tool for a reliability assessment and a strategy for periodic inspection of welded joints in marine steel structures. The model is a unit-jump discrete-state/discrete-time Markov chain where damage states are related to the fatigue crack depths. The model parameters are estimated from the experimental statistics of the fatigue crack initiation and growth in fillet-welded joints tested under uniaxial constant-amplitude (CA) loading. By the use of an automatic crack monitoring system the statistics of the lifetime to reach given crack depths are obtained. Based on a fracture mechanics approach and the statistics of S-N data the model is extended to deal with variable-amplitude loading. The possibilities of taking into account crack initiation and crack threshold are discussed. The practical engineering utility of the model is emphasized and a case study for an articulated column in the North Sea is included.