July 2023
In this Smart Infrastructure Blog, Wei Bi, PhD candidate in Climate Resilience of Infrastructure Systems at the Centre for Sustainable Development & Centre for Smart Infrastructure and Construction, argues that the current preference for simplicity in network modelling has the potential to obstruct real advances in the application of network modeling for accurately measuring infrastructure resilience.
The impacts of climate change are real and have become increasingly evident over the past decade. Events such as extreme flooding have posed significant threats to existing infrastructure assets, thereby causing disruptions to the normal functioning of society and the economy. In this context, traditional prevention and protection-oriented engineering approaches are insufficient to address extreme risks, because investment in eliminating such high levels of risk is unlikely to be affordable or practical. In response, resilience has gained popularity as a mechanism to plan for ‘safe-to-fail’: it acknowledges that extreme and uncertain events could disrupt the operation of an infrastructure system, but the consequences can be minimised through proactive response, prompt recovery, and effective adaptation.
An overly simplified network model is an inadequate representation of a realistic complex socio-technical infrastructure system that would, consequently, diminish the credibility of the obtained results. The current trend of favoring simplicity in modeling has the potential to impede the tangible advancements in applying network modelling for effectively assessing infrastructure resilience Wei Bi PhD candidate in Climate Resilience of Infrastructure Systems, Centre for Sustainable Development & Centre for Smart Infrastructure and Construction(CSIC).
As engineers, we appreciate the value of quantifying things and simulating plausible scenarios to understand where we are and what we can do to achieve better outcomes. Developing methodologies to measure infrastructure resilience is therefore critical both to predicting how well a system will perform in the event of future disruptions and to testing the effectiveness of feasible interventions. The numerical results help to shape informed decision-making on prioritising resources for emergency response and climate change adaptation. Modelling infrastructure resilience is nevertheless a challenging task, as the infrastructure itself is a complex social-technical system and its resilience can be interpreted from varying dimensions.
Taking transport infrastructure as an example, the traditional engineering dimension focuses on the physical infrastructure itself – how robust the road/railway/metro network is to local disturbances and how much redundancy is embedded in the system to provide alternative pathways to service delivery. This ‘hard’ technical dimension of resilience is indeed essential, but we cannot disregard the role of the ‘soft’ human dimension. From the organisational perspective, the soft dimension relates to how the asset operators act in response to emergencies and prioritise decision-making based on the limited resources; from the social perspective, it concerns the extent to which the disruption affects passengers’ journeys and business activities. All these dimensions are indispensable, but the task of designing an approach that combines them into a comprehensive measure of infrastructure resilience is still under development.
Our recent paper entitled “Old wine in new bottles? Understanding infrastructure resilience: foundations, assessment, and limitations”, provides a review of approaches to infrastructure resilience assessment with a focus on applying complex network modelling to measure the resilience of transport infrastructure, where there is an emerging body of research. As modern infrastructure systems are becoming increasingly interdependent, network modelling becomes a trendy tool for capturing the physical connections and functional interactions between system elements and making the entire complex of relationships understandable from the system level. For instance, a metro system can be abstracted into a network with nodes representing metro stations and edges representing tracks. Disruption scenarios can be simulated by removing nodes and edges from the network, but there must be a basis for deciding which elements to remove and in what sequence. The recovery process can also be simulated by bringing the failed elements back to the network, but there is a need to consider realistic influential factors such as recovery priority, practical recovery time, and limited recovery resources. On this basis, resilience can be assessed by quantifying the change in system performance during the disruption duration.
This process of infrastructure resilience assessment seems straightforward. By incorporating programming techniques, the utilisation of complex network modelling should be a powerful tool to customise disruption scenarios, recovery processes, and resilience indicators in a practically sensible manner. Surprisingly, however, current network models for infrastructure resilience measurement (as it has typically been applied to date) are often overly simplified; they tend to only consider the resilience of the physical structure of the infrastructure network (i.e. analysing the technical dimension by calculating topological attributes based on mathematical theory, such as betweenness centrality and network efficiency) but neglect many practical features of operational performance (e.g. travel distance/time, passenger flows, and heterogeneity of stations in terms of geographical location and engineering design) that play critical roles as well. Moreover, the way that nodes or edges are removed (i.e. either random or targeted) and recovered is over-abstract for simulating geographical disasters such as floods; this requires consideration of how the flooding actually affects system operation, and how asset operators respond, but there is very limited evidence to suggest that interviews have been conducted with asset operators to distil such valuable information.
An overly simplified network model is an inadequate representation of a realistic complex socio-technical infrastructure system that would, consequently, diminish the credibility of the obtained results. The current trend of favoring simplicity in modeling has the potential to impede the tangible advancements in applying network modelling for effectively assessing infrastructure resilience. This reflects a common sentiment among researchers who work with complex network modelling: while complex network modelling can provide valuable insights and help us to understand the fundamental properties of complex infrastructure systems, there are challenges in accurately describing and predicting real-world systems using these models. To depict real-world systems with high resolution, a paramount prerequisite is engaging in conversations with industry professionals to uncover critical information such as system architecture, data flow, historical events, and operational processes.
Reference:
Bi, W., MacAskill, K. and Schooling, J., 2023. 'Old wine in new bottles? Understanding infrastructure resilience: Foundations, assessment, and limitations'. Transportation Research Part D: Transport and Environment, 120, p.103793.