Approximate Bayesian network formulation for the rapid loss assessment of real-world infrastructure systems

Gehl, Pierre; Cavalieri, Francesco; Franchin, Paolo

doi:10.1016/j.ress.2018.04.022

Cited by 29 publications

(13 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is worth noting that the construction of loss scenarios useful for crisis management purposes is not limited to the building damage assessment considered in this paper, but may include consideration about the potential for induced effects such as landslides or soil liquefaction [73], functional losses resulting from structural damage [74], or potential domino effects that could lead to NaTech events [75]. The assessment of building damage is therefore only one of the central steps in a broader process of qualifying the resilience of systems to earthquakes [76].…”

Section: Impact On the Use Of Damage Scenarios For Earthquake Crisis Managementmentioning

confidence: 99%

Sensitivity of Earthquake Damage Estimation to the Input Data (Soil Characterization Maps and Building Exposure): Case Study in the Luchon Valley, France

et al. 2021

Self Cite

View full text Add to dashboard Cite

This article studies the effects of the soil data and exposure data of residential building inventories, as well as their spatial resolution, on seismic damage and loss estimates for a given earthquake scenario. Our aim is to investigate how beneficial it would be to acquire higher resolution inventories at the cost of additional effort and resources. Seismic damage computations are used to evaluate the relative influence of varying spatial resolution on a given damage model, where other parameters were held constant. We use soil characterization maps and building exposure inventories, provided at different scales from different sources: the European database, a national dataset at the municipality scale, and local field investigations. Soil characteristics are used to evaluate site effects and to assign amplification factors to the strong motion applied to the exposed areas. Exposure datasets are used to assign vulnerability indices to sets of buildings, from which a damage distribution is produced (based on the applied seismic intensity). The different spatial resolutions are benchmarked in a case-study area which is subject to moderate-to-average seismicity levels (Luchon valley in the Pyrénées, France). It was found that the proportion of heavily damaged buildings is underestimated when using the European soil map and the European building database, while the more refined databases (national/regional vs. local maps) result in similar estimates for moderate earthquake scenarios. Finally, we highlight the importance of pooling open access data from different sources, but caution the challenges of combining different datasets, especially depending on the type of application that is pursued (e.g., for risk mitigation or rapid response tools).

show abstract

Section: Impact On the Use Of Damage Scenarios For Earthquake Crisis Managementmentioning

confidence: 99%

Sensitivity of Earthquake Damage Estimation to the Input Data (Soil Characterization Maps and Building Exposure): Case Study in the Luchon Valley, France

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…BBN is an extensive probabilistic model that is used to characterize the uncertainty that is associated with variables that constitute the model [49]. BBN is a graph-based model comprising nodes and edges, where nodes represent model variables and edges that represent the relationship between the nodes and also the conditional dependencies [50,51]. BBN can also be defined as a class of graphical models that presents a brief representation of the probabilistic dependencies between a given set of random variables [52].…”

Section: Bayesian Belief Networkmentioning

confidence: 99%

Flood Resilience of Housing Infrastructure Modeling and Quantification Using a Bayesian Belief Network

2021

View full text Add to dashboard Cite

Resilience is the capability of a system to resist any hazard and revive to a desirable performance. The consequences of such hazards require the development of resilient infrastructure to ensure community safety and sustainability. However, resilience-based housing infrastructure design is a challenging task due to a lack of appropriate post-disaster datasets and the non-availability of resilience models for housing infrastructure. Hence, it is necessary to build a resilience model for housing infrastructure based on a realistic dataset. In this work, a Bayesian belief network (BBN) model was developed for housing infrastructure resilience. The proposed model was tested in a real community in Northeast India and the reliability, recovery, and resilience of housing infrastructure against flood hazards for that community were quantified. The required data for resilience quantification were collected by conducting a field survey and from public reports and documents. Lastly, a sensitivity analysis was performed to observe the critical parameters of the proposed BBN model, which can be used to inform designers, policymakers, and stakeholders in making resilience-based decisions.

show abstract

“…The distribution of the terms η and ξ i is directly provided by the shake-map outputs; while the spatial correlation between the ξ i terms is approximated by a Dunnett-Sobel decomposition (Dunnett and Sobel 1955), as detailed by Bensi et al (2011) and Gehl et al (2018) in the case of spatially distributed systems. Given a spatial correlation matrix R between the N sites, and standard Gaussian variables w and v i , the Dunnett-Sobel decomposition provides the following:…”

Section: Bayesian Updating Framework For the Estimation Of New Paramementioning

confidence: 99%

Investigation of damage to the water network of Uki City from the 2016 Kumamoto earthquake: derivation of damage functions and construction of infrastructure loss scenarios

Gehl

Matsushima

Masuda

2020

Bull Earthquake Eng

Self Cite

View full text Add to dashboard Cite

The 2016 Kumamoto earthquake (main-shock: Mw 7.0, April 16th) has induced a widespread disruption of lifeline systems across the affected area. In Uki City (Kumamoto prefecture, Japan), water supply had been intermittently cut off for several days: as a result, around 50 repair operations on various locations of the underground pipeline system have been carried out by municipal services. In order to better constrain the vulnerability of such infrastructure with respect to future earthquakes, the collected empirical data is exploited to derive repair-rate functions for ductile pipelines (ductile iron and polyvinyl chloride) of small diameter (less than 100 mm). Due to the relatively low number of data points over a limited range of seismic intensity, the derivation of purely empirical damage function appears to be subject to significant statistical biases. Therefore, a Bayesian updating framework is adopted, where prior information on the parameters of the repair-rate function is estimated from existing damage functions from the literature. Moreover, the uncertainty related to the characterization of the Peak Ground Velocity at the location of the pipelines is taken into account by: (i) the generation of shake-maps with different assumptions on ground-motion prediction equations or fault models, and (ii) the inclusion of a spatially correlated field of the intra-event shake-map error term (i.e., modelling of the inherent variability of the seismic intensity). The results show that the derived repair-rate equation is consistent with some existing functions for ductile and earthquake-resistant pipeline segments. The effect of specific land conditions (e.g., topographic/geological factors) is also investigated, with the possibility to further parametrize the repair-rate function. Finally, the developed damage functions are applied to stochastic simulations of the seismic performance of the water network, while keeping track of various sources of uncertainties and quantifying their impact on the system’s loss distribution.

show abstract

Approximate Bayesian network formulation for the rapid loss assessment of real-world infrastructure systems

Cited by 29 publications

References 23 publications

Sensitivity of Earthquake Damage Estimation to the Input Data (Soil Characterization Maps and Building Exposure): Case Study in the Luchon Valley, France

Sensitivity of Earthquake Damage Estimation to the Input Data (Soil Characterization Maps and Building Exposure): Case Study in the Luchon Valley, France

Flood Resilience of Housing Infrastructure Modeling and Quantification Using a Bayesian Belief Network

Investigation of damage to the water network of Uki City from the 2016 Kumamoto earthquake: derivation of damage functions and construction of infrastructure loss scenarios

Contact Info

Product

Resources

About