Site Specific Modelling of Traffic Loading on Highway Bridges

Accurate traffic loading models based on measured weigh-in-motion (WIM) data are
essential for the accurate assessment of existing bridges. Much work has been
published on the Monte Carlo simulation of single lanes of heavy vehicle traffic, and
this can easily be extended to model the loading on bridges with two independent
streams of traffic in opposing directions. However, a typical highway bridge will have
multiple lanes in the same direction, and various types of correlation are evident in
measured traffic, such as groups of very heavy vehicles travelling together and heavy
vehicles being overtaken by lighter ones. These traffic patterns affect the probability
and magnitude of “multiple presence” loading events on bridges, and are significant
for the maximum lifetime loading on the bridge.
This paper analyses traffic patterns using multi-lane WIM data collected at four
European sites. It describes an approach to the Monte Carlo simulation of this traffic
which seeks to replicate the observed patterns of vehicle weights, same-lane and interlane
gaps, and vehicle speeds by applying variable bandwidth kernel density
estimators to empirical traffic patterns. This allows the observed correlation structure
to be accurately simulated but also allows for unobserved patterns to be simulated.
The process has been optimised so as to make it possible to simulate traffic loading on
bridges over periods of 1,000 years or more, and this removes much of the variability
associated with estimating characteristic maximum load effects from shorter periods
of either measured or simulated data. The results of this analysis show that the patterns
of correlation in the observed traffic have a small but significant effect on bridge
loading.