Use of monitored daily extreme stress data for performance prediction of steel bridges: Dynamic linear models and Gaussian mixed particle filter.

Highlights • A procedure for effective incorporation of daily data for reliability prediction. • A modelling approach of dynamic linear models based on historical data. • An effective use of Gaussian mixed particle filter with data and dynamic models. Abstract Sensors of modern bridge monitoring systems provide a huge amount of data used for reliability prediction. The proper handling of these data is one of the main difficulties in the field of structural health monitoring. To reasonably predict structural time-variant reliability based on the monitored daily extreme stress data, the objectives of this paper are to present: (a) a procedure for the effective incorporation of monitored daily extreme stress data for dynamic reliability prediction of bridge components, (b) a modelling approach of dynamic linear models based on historical daily extreme stress data in structural reliability prediction, and (c) an effective use of Gaussian mixed particle filter combining the monitored daily extreme stress data and dynamic linear models for dynamically predicting structural reliability. The monitored data obtained from an existing bridge is provided to illustrate the feasibility and application of the procedures and models proposed by this paper. [ABSTRACT FROM AUTHOR]