Statistical Analysis of Traffic Loads and Traffic Load Effects on Bridges

It is vital to ensure highway structures to cater for increasing demand in transport capacity. This involves the precise estimation of structural load-carrying capacity to withstand potential maximum load effects in residual lifetime. The scope of this research thus addresses estimation of traffic load effect. Many different methods have been used to model extreme traffic load effects on bridges for estimating characteristic value or distribution, including fitting a Normal or Gumbel distribution to the upper tail of load effects, the use of Rice formula based level crossing method, the block maxima method and the peaks over threshold (POT) method. A review of the mathematic and engineering background of these methods is presented with qualitative analysis. In addition, a quantitative comparison is carried out to evaluate the performances of these methods on estimating characteristic value and probability of failure. Results show that the POT method provides relatively good estimation. A thorough study on POT method is thus carried out in this thesis. Estimation of generalized Pareto distribution parameters is a crucial step in the application of POT method. Available methods include the method of moments (MM), the probability weighted moments (PWM), the maximum likelihood (ML), the penalized maximum likelihood (PML), the minimum density power divergence (MDPD), the empirical percentile method (EPM), the maximum goodness-of-fit statistic and the likelihood moment (LM). In order to provide guidance on selecting appropriate method when applying POT to bridge traffic load effects, we evaluate their performance by estimating characteristic values through three studies with: numerical simulation data, Monte Carlo simulated and measured traffic load effect measurements. From the simulated traffic load effect, the ML and PML provide more accurate estimates for large size sample, while the MM and PWM are better than others for small size sample. Moreover, results from monitored traffic load effects indicate that outliers or large observations have significant influence on the parameter estimators. A common cause of outliers or larger observation is observed when the traffic is a mixture of two distributions, which may be two distinct sub-populations. Traffic load effects induced by different numbers of vehicles are not identically distributed, and this violates basic assumption in the use of the classic extreme value theory to estimate the distribution of extreme bridge traffic load effects. A new method named mixture peaks over threshold (MPOT) is proposed which enables to estimate the distribution of extreme traffic load effects from mixture loading events. Its performance is evaluated by theoretical and traffic load effect examples, and results iii te l-0 08 62 40 8, v er si on 1 16 S ep 2 01 3