From Fatigue-to-Fracture: Probabilistic Total-Life of Steel Bridge Structures

Abstract

The remaining fatigue-life of existing steel bridge structures is an unresolved thorny issue. Current code-based fatigue rules for new structure design were established on conservative principles that are intended to deliver dependable designs. Conversely, using the design basis in an inverse form shows the predicted fatigue-life to be unrealistically short– only a few decades, whereas many such bridges exceed this fatigue-life by a wide margin. For example, for an older bridge an assessment may indicate that the structure should have reached its (theoretical) fatigue-life some 10-years ago. Moreover, the first observed fatigue crack is not necessarily fatal; there always remains some post-crack life during the propagation stage. Instead of being panic-stricken, this additional life may be used wisely to carry out the remedial works. Knowing this, how should the asset owner's engineer approach this apparent dichotomy. The present research uses first non-parametric (ARMA) and then parametric fatigue damage models to re-calibrate historic fatigue test data in a probabilistic framework. Fracture mechanics is then used to extend the model from first fatigue crack (initiation) through propagation to final fracture. Most cycle counting methods are based on full reversals; however, steel bridges are subjected to elevated tensile mean stresses that are known to significantly affect the fatigue-life. Under typical train loading conditions, results show that for short span bridges each freight car contributes to stress reversals while for longer spans the overall train effect along with small fluctuations due to cars contribute to stress cycles. An integrated four-step model which accounts for all major sources of aleatory and epistemic uncertainties is used to assess the remaining fatigue-life as well as the post-crack initiation-life. Results show an expected fatigue-life of 129 years with 19 years of propagation-life for a short span riveted railway bridge while an expected fatigue-life of 3556 years with 511 years of propagation-life for a longer span bridge. Fatigue-to-fracture-life fragility curves indicate some 15% additional life until unstable crack propagation (fracture) occurs for riveted steel bridges. This offers the asset owners sufficient time to inspect, evaluate, propose, and execute the remedial works to restore the bridge back to full unrestricted service.