Entity-based System Dynamics for Bridge Asset Management

Abstract

The Netherlands is on the verge of initiating its most extensive maintenance project in history. The Dutch government will have to invest an estimated minimum of 260 billion Euros for the renewal of its aging civil structures like bridges, viaducts, and tunnels. This large maintenance operation is labeled as the replacement and renovation task. Within the replacement and renovation task, so-called ‘Baby Boomer bridges’ require the most attention. These bridges, accounting for a substantial portion of the 85,000 bridges in the Dutch inventory, face challenges stemming from prolonged stress and corrosion issues. The Merwedebrug's narrowly averted catastrophe in 2016 served as a wake-up call, highlighting the critical need for proactive and strategic maintenance practices. A key challenge in the replacement task is the limited construction capacity, especially with an expected surge in maintenance between 2040 and 2080. Efficient and systemic asset management is pivotal in minimizing the required capacity increase and to help ‘flatten the curve’.

One of the core strategies identified by the Dutch government to facilitate the transition towards more systemic and efficient infrastructure asset management is to cluster bridge maintenance projects. Project clustering involves consolidating multiple maintenance projects with similar characteristics or geographical proximity into one portfolio. Existing research has highlighted the positive impacts of project clustering on project-specific performance indicators. However, there is a lack of studies examining the broader implications of different clustering strategies on the entire transport system. The current literature predominantly comprises retrospective studies analyzing historical project data, providing valuable insights into project clustering effectiveness but falling short of assessing its influence on future infrastructure system behavior. This research aims to bridge this gap by conducting an exploratory modeling analysis to explore the effects of different maintenance cluster strategies on the performance of the transport infrastructure network.

This thesis employs Entity-based System Dynamics (SD). Entity-based SD is a relatively new modeling methodology and can be seen as a combination of agent-level (ABM) modeling and macro-level (SD) modeling. This combination allows for the modeling of the (spatial) behavior and attributes of individual bridges, roads and regions, while retaining the capability of doing macro-level analyses. Furthermore, as the bridge maintenance problem is subject to deep uncertainty, Entity-based SD was paired with the Exploratory Modeling and Analysis methodology. This allows for the exploration of the repercussions of various combinations of assumptions about uncertain factors in the system. To allow for the combination of the two methodologies, a novel EMA Workbench-Ventity connector was constructed for this thesis.

Because Entity-based SD is still a relatively novel methodology, there is a lack of spatially explicit applications within the existing scientific literature. As such, this thesis pursues two objectives, (1) developing, and reflecting on the added value of a novel spatially explicit Entity-based SD modeling method when modeling the effect of bridge maintenance cluster policies on the wider infrastructure system, and (2) identifying maintenance cluster policies that are effective at facilitating a steady and predictable maintenance capacity demand. To pursue these two objectives, an abstract network was constructed using the Entity-based System Dynamics methodology.

The analysis of the model outcomes shows that the model was able to generate spatially explicit relationships between traffic flows and bridge degradation. The model was also able to capture the performance of maintenance cluster strategies and showed expected behavior. Six policies were tested with the model, three variations of geographical clustering (small, medium, and large), construction type clustering, construction year clustering, and a no clustering policy. The model results indicate that larger maintenance clusters bring about more fluctuating changes in capacity utilization, while smaller clusters lead to a higher total number of expected projects over a 100 year simulation period. Larger clusters also result in a higher average load capacity for the bridge set, mainly due to increased preventive maintenance. Despite larger clusters generally outperforming no cluster policies, their overall effectiveness is diminished, especially concerning the critical outcome of change in capacity utilization. A geographical cluster policy with small clusters, an average of 1.9 bridges per cluster, stands out for its more stable maintenance capacity utilization compared to a no cluster policy option and slightly better performance in other key outcomes.

As such, policymakers should implement policies that encourage the formation of small maintenance clusters. However, as the network specification plays a crucial part in the performance of cluster policies, policymakers should adopt a flexible approach, considering the specific characteristics of the infrastructure network when formulating maintenance clustering policies. Future applications for the model could add additional external effects to the model, introduce finite maintenance capacity and a finite maintenance project size, or include dynamics in the model that allow for the modeling of traffic jams.

At a methodological level, it can be concluded that Entity-based SD is a suitable approach to infrastructure modeling. The added value of the novel spatially explicit Entity-Based SD approach can be described in five points. First, the method holds a high degree of replicability. Because entity types can be independently defined, infrastructure components can be individually modeled and reused in other models. Additionally, the method makes use of externalized network initialization data, which separates the dynamics of infrastructure components and the network specification data, allowing for components to be altered individually without the need to alter the other. Second, the computational requirements of the approach are limited compared to other infrastructure modeling approaches. Third, similar to SD modeling, the model is made up of a clear model structure with stocks, flows, and causal links which enhances communicability and supports group model building with stakeholders. Fourth, as an extension of SD modeling, the method provides a holistic approach to infrastructure modeling, which means that it enables the modeling of not only individual infrastructure components but also the broader system in which these components operate. Lastly, the approach is capable of coping with high degrees of uncertainty due to the EMA Workbench connector that was constructed for this thesis.

Based on the arguments presented in this thesis, the novel spatially explicit Entity-based SD approach is considered to be a suitable new avenue for infrastructure modeling. However, this study should be considered as a first investigation into this approach and is therefore incomplete. Although the abovementioned added values have been identified during the course of the thesis, some limitations and future improvements still exist. Firstly, the Entity-based SD methodology offers limited documentation, as the maturity of the approach is low. Secondly, the approach is not meant to generate precise forecasts. As such, if this is the objective of the modeler or stakeholder, Entity-based SD should not be considered as a candidate approach. Future applications of this modeling approach could look at the performance of the approach when modeling larger networks. Therefore, spatial Entity-Based SD should not be seen as a replacement for current infrastructure modeling approaches. Rather, it should be viewed as a new addition to the scientific field of infrastructure modeling.