Conclusion and Limitations – Virtual City Model

Conclusion

This study applied a multi-objective optimization framework to evaluate maintenance strategies for a heterogeneous infrastructure system, integrating life-cycle environmental, economic, and operational performance indicators within a unified 25-year analysis horizon. Rather than identifying a single optimal solution, the analysis focused on the structure of trade-offs emerging from the interaction between maintenance duration, environmental impact, and cost. The results demonstrate that Pareto-based optimization provides a robust means of exploring these trade-offs, even under stochastic sampling, while also revealing intrinsic limitations imposed by system structure and boundary definitions.

Limitation

A key limitation of the current analysis arises from the pronounced heterogeneity of the integrated system. Subsystems with short operational lifespans and high maintenance intensity—most notably offshore wind turbines—tend to dominate the optimization outcome within the fixed 25-year horizon. In contrast, long-lived components such as buildings and foundations contribute relatively little decision flexibility within the same time window, as only a small subset of their maintenance cycles is represented. As a result, maintenance strategies for long-lifespan subsystems are effectively constrained, leading to an imbalance in which optimization primarily reflects the behavior of short-lived, high-impact components rather than the integrated system as a whole.

This structural imbalance suggests that the choice of system boundaries and temporal scope plays a decisive role in shaping the decision space, often more strongly than the optimization algorithm itself. One potential avenue for addressing this limitation is the adoption of modular or hierarchical system definitions. By grouping subsystems with comparable lifespans and maintenance characteristics—such as offshore wind turbines and associated electrical infrastructure—optimization could be conducted on more homogeneous subsystems before being integrated at a higher system level. Such an approach may reduce extreme lifespan mismatches and improve the interpretability of trade-offs without abandoning system-level coordination.

Further limitations are associated with the temporal discretization of maintenance decisions. In the present model, maintenance timing is defined at an annual resolution, which restricts the range of feasible intervention schedules, particularly for short-lived components with narrow decision windows. Increasing temporal resolution to semi-annual, quarterly, or monthly intervals could recover additional decision flexibility and enable finer differentiation between maintenance strategies. However, this improvement would come at the cost of increased computational complexity and higher data requirements, underscoring the trade-off between model expressiveness and practical tractability.

Finally, the fixed operational horizon excludes the possibility of full component replacement, most notably the renewal of offshore wind turbines. Allowing turbine replacement would fundamentally alter the system boundary by extending the effective lifetime of the integrated system and shifting the role of long-lived structures from auxiliary components to primary decision drivers. Such an extension would require explicit modeling of decommissioning processes, manufacturing impacts of replacement components, and transitional system states. While beyond the scope of the present study, this perspective highlights the importance of aligning system boundaries with the research question, particularly when optimization spans multiple infrastructure generations.

Future Research

Overall, this work illustrates that in highly heterogeneous infrastructure systems, optimization outcomes are strongly conditioned by structural assumptions embedded in system definition, temporal scope, and decision granularity. Pareto-based multi-objective optimization remains a valuable exploratory tool under these conditions, but its results must be interpreted as reflections of the modeled decision space rather than definitive prescriptions. Future research should therefore treat system design, boundary selection, and temporal resolution not as secondary modeling choices, but as central components of the optimization problem itself.