System Effects in Identifying Risk-Optimal Data Requirements for Digital Twins of Structures

TL;DR

This paper evaluates how different data collection methods like inspection and SHM influence risk management decisions for structures, emphasizing the importance of system-level models over simple data sum approaches.

Contribution

It introduces a framework for quantifying the combined value of inspection and SHM data using VoI analysis in structural risk management.

Findings

System-level analysis differs from simple sum of marginal VoI estimates.

System-level decision making requires integrated models.

Julia code for probabilistic models and influence diagrams is provided.

Abstract

Structural Health Monitoring (SHM) technologies offer much promise to the risk management of the built environment, and they are therefore an active area of research. However, information regarding material properties, such as toughness and strength is instead measured in destructive lab tests. Similarly, the presence of geometrical anomalies is more commonly detected and sized by inspection. Therefore, a risk-optimal combination should be sought, acknowledging that different scenarios will be associated with different data requirements. Value of Information (VoI) analysis is an established statistical framework for quantifying the expected benefit of a prospective data collection activity. In this paper the expected value of various combinations of inspection, SHM and testing are quantified, in the context of supporting risk management of a location of stress concentration in a railway bridge. The Julia code for this analysis (probabilistic models and influence diagrams) is made available. The system-level results differ from a simple linear sum of marginal VoI estimates, i.e. the expected value of collecting data from SHM and inspection together is not equal to the expected value of SHM data plus the expected value of inspection data. In summary, system-level decision making, requires system-level models.