Application of the statistics of extremes to the reliability assessment and performance prediction of monitored highway bridges

The present paper examines how statistics of extremes can be used to enhance the assessment and performance prediction of monitored highway bridges. This is achieved by proposing an approach to obtain a monitoring-based live-load for use in a reliability analysis. Time effects that correlate observed data to code required return periods are considered. Additionally, a method to identify, minimise and properly account for the epistemic uncertainty inherent to any monitoring record within a reliability analysis is presented. The approach can be utilised to plan data collection efforts or to maximise the utility of a limited amount of data. The extension of extreme value statistics to the monitoring of highway bridges is developed first using simulations and is then demonstrated on a case study using 90 days of in-service data collected from a bridge located in Pennsylvania, USA.