ABSTRACT
In the aftermath of a major seismic event, transportation networks are crucial to rescue victims and preserve access to hospitals and shelters, while later they are instrumental in supporting the economic recovery. Fragility curves are required for the probabilistic evaluation of performance of an earthquake-damaged transportation network. Recently, fragility curves of several hundreds of bridges have been obtained via 3D inelastic response history analysis, and then used to calibrate a Bayesian Network model of seismic fragility (Figure 1). This paper investigates the sensitivity of network-level results to the use of the developed BN-based fragility model. Figure 2 shows the case study network chosen to probe such sensitivity. Monte Carlo simulations were employed to take into account the relevant uncertainties, both in the seismic hazard and in the physical damageability of bridges. Two sets of simulations were carried out, the one employing the exact FEM-based fragility curves and the one employing the BN-based ones. Network flow analysis, establishing traffic flow and flow speed on all network edges via user equilibrium, was carried out on the damaged network for thousands of seismic events. Two global flow-based performance indicators have been computed: a) the average travel time increment, ATTI, that is the average value over the seven bridges of the travel time with respect to pre-earthquake conditions; b) the Drivers’ Delay, DD, defined as the difference between the total travel time in damaged and undamaged conditions. Results are encouraging showing that even though the BN-based surrogate fragility model does not perform in an equally good manner over all bridges, the simple fact that network performance does not in general depend on any single worst-performing bridge (from the point of view of approximating its fragility) leads to overall more accurate results at the network level paving the way for the replacement of generic bridge fragility functions in large scale network analyses. BN structure for the LD limit state assessment (input and output variables in white and black shading, respectively). https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315207681/cd556cd4-4dcf-4efe-8e29-56fc67b8bfbd/content/fig152_1.tif"/> Königsberg road network. Source-to-site distances do not follow the scale. https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315207681/cd556cd4-4dcf-4efe-8e29-56fc67b8bfbd/content/fig152_2.tif"/>
