This paper aims to seek to develop a digital twin (DT)–based framework for identifying systemic delivery risk in construction projects through latent performance–loss states. Conventional monitoring approaches rely primarily on observable indicators such as productivity or schedule deviation, which often reveal problems only after degradation has already propagated through interdependent project activities. The study therefore proposes a monitoring approach that infers underlying system conditions from operational data, enabling earlier detection of emerging risk and supporting more proactive decision-making in complex construction delivery systems.
The framework is demonstrated within a representative construction delivery case study using design science research methodology and requires further validation across different project types and live operational environments. The findings establish latent system risk as an inferable condition of interdependent construction delivery systems and provide a basis for further research into adaptive dependency modelling, project health, systemic resilience and intelligent construction control.
Across the five scenarios, the framework achieved mean absolute error (MAE) values of 0.06–0.11 and root mean square error values of 0.08–0.15, reducing estimation error by 21%–34% relative to key performance indicator (KPI)-based monitoring. Latent-state inference provided a median detection lead-time gain of 4.2-time units, equivalent to a 35% increase in the available early-warning window. The findings demonstrate that systemic degradation can be inferred before significant deterioration becomes observable through conventional KPIs.
The study demonstrates the proposed framework using a representative case study, which may limit the generalisability of findings across all project contexts. Further empirical validation across diverse project types and larger dependency networks would strengthen the robustness of the approach. Future research may also explore integration with automated control strategies, advanced sensing technologies and machine learning–based state estimation methods to enhance scalability and predictive capability in DT–enabled construction monitoring systems.
The framework enables stakeholders to move from reactive monitoring of realised deviations towards earlier identification of emerging system deterioration. By identifying where degradation is developing and how it may propagate through subsystem dependencies, DT-enabled risk intelligence can support project managers, contractors and clients in prioritising earlier and more targeted interventions, with potential to mitigate downstream schedule, productivity and cost consequences.
The study shifts construction monitoring from observing realised performance deviations towards inferring latent system risk as an evolving condition of interdependent delivery systems. The study proposes an integrated approach to performance loss -state inference, subsystem coupling and DT synchronisation to transform observable project data into systemic risk intelligence. This enables emerging degradation and its propagation to be identified before conventional KPI thresholds are reached, providing an earlier and more targeted basis for construction project control.
