The use of predictive likelihood to estimate the distribution of extreme bridge traffic load effect

To assess the safety of an existing bridge, the loads to which it may be subject in its
lifetime are required. Statistical analysis is used to extrapolate a sample of load effect values from the simulation period to the required design period. Complex statistical methods are often used and the end result is usually a single value of characteristic load effect. Such a deterministic result is at odds with the underlying stochastic nature of the problem. In this paper, predictive likelihood is shown to be a method by which the distribution of the lifetime extreme load effect may be determined. An estimate of the distributions of lifetime maximum load effect facilitates the reliability approach to bridge assessment. Results are presented for some cases of bridge loading, compared to a return period approach and significant differences identified. The implications for the assessment of existing bridges are discussed.