Quantifying the value of information in Bayesian inference using deflection measurements of prestressed concrete bridges

Excessive deflection in long-span prestressed concrete bridges (PSCBs) poses a significant engineering challenge due to the highly stochastic and time-dependent behaviour of these structures. Bayesian inference using static deflection measurements offers a promising approach to reducing uncertainties of key influencing factors, such as concrete creep and shrinkage. However, optimising the scheduling of inspection and maintenance activities to maximise the benefits of Bayesian inference remains an open issue. This study addresses this gap by evaluating the value of information (VOI) associated with Bayesian inference using static deflection measurements in PSCBs. A prior-decision optimisation framework is employed to assess the minimal expected costs of various inspection and maintenance strategies, accounting for uncertainties in concrete creep and shrinkage. The conditional VOI (CVOI) is defined as the difference in minimal expected costs before and after Bayesian updating. Furthermore, the VOIs of future inspections are quantified through the expected CVOI of predicted deflection values, generated by a Gaussian process trained on existing deflection data. The proposed method is applied to an actual PSCB, demonstrating that CVOI varies with specific deflection observations, while the VOI of future inspections increases nonlinearly with longer observation intervals. These findings provide valuable insights for optimising inspection and maintenance schemes.