Reliability-Based Load and Resistance Factor Rating Using In-Service Data

Traditional bridge evaluation techniques are based on design-based deterministic equations that use limited site-specific data. They do not necessarily conform to a quantifiable standard of safety and are often quite conservative. The newly emerging load and resistance factor rating (LRFR) method addresses some of these shortcomings and allows bridge rating in a manner consistent with load and resistance factor design (LRFD) but is not based on site-specific information. This paper presents a probability-based methodology for load-rating bridges by using site-specific in-service structural response data in an LRFR format. The use of a site-specific structural response allows the elimination of a substantial portion of modeling uncertainty in live load characterization (involving dynamic impact and girder distribution), which leads to more accurate bridge ratings. Rating at two different limit states, yield and plastic collapse, is proposed for specified service lives and target reliabilities. We consider a conditional Poisson occurrence of identically distributed and statistically independent (i.i.d.) loads, uncertainties in field measurement, modeling uncertainties, and Bayesian updating of the empirical distribution function to obtain an extreme-value distribution of the time-dependent maximum live load. An illustrative example uses in-service peak-strain data from ambient traffic collected on a high-volume bridge. Serial independence of the collected peak strains and of the counting process, as well as the asymptotic behavior of the extreme peak-strain values, are investigated. A set of in-service load and resistance factor rating (ISLRFR) equations optimized for a suite of bridges is developed. Results from the proposed methodology are compared with ratings derived from more traditional methods.