Proceedings of the

The 33rd European Safety and Reliability Conference (ESREL 2023)
3 – 8 September 2023, Southampton, UK

Digital Twins or Equivalent Infrastructure Models? The Role of Modeling Granularity in Regional Risk Analysis of Infrastructure

Fabrizio Nocera1 and Paolo Gardoni2

1Department of Civil, Environmental and Geomatic Engineering, University College London, UK.

2Department of Civil and Environmental Engineering, University of Illinois Urbana-Champaign, USA.


Accurate regional risk analysis requires representative mathematical models of infrastructure. One of the main challenges in developing mathematical models of infrastructure is defining their modeling granularity, i.e., the level of detail in the topology of the model. Different modeling granularities affect our ability to capture the spatial variability of the impact arising from the changes in the capacities of infrastructure and service demands. A recent trend in infrastructure modeling is to develop detailed digital twins to mimic all aspects of the real infrastructure. However, detailed digital twins might require data not readily available, and their analyses often have prohibitive computational costs, making digital twins not always the most suitable option to model the performance of infrastructure. The goal of selecting the optimal modeling granularity is to allocate computational resources to the model that best delivers the desired information with the desired accuracy level. This paper presents a mathematical formulation to systematically select the appropriate modeling granularity of infrastructure. The formulation adaptively increases the granularity starting from a low-granularity infrastructure model until we reach the desired tradeoff among accuracy, simplicity, and computational efficiency. To define the tradeoff, we introduce metrics that measure the level of agreement between estimates of the quantities of interest computed using different levels of granularity. Such metrics include global measures that assess if a model is insufficiently detailed to capture the quantities of interest and local measures that identify specific regions of the model that may require further refinement. As an example, we apply the illustrated formulation to select the granularity of the potable water infrastructure model in Seaside, Oregon, to quantify its performance following a seismic event.

Keywords: Digital twin, Infrastructure, Model selection, Network granularity, Resilience.

Download PDF