ABSTRACT
Bridge management data can be used to create predictive models for prioritizing repairs of bridges, which may lead to better budget allocation. Thus, this study proposes a framework to derive a predictive model for the state of the bridges in Ontario, Canada, based on the available data from the Ontario bridge inventory. 1691 bridges were selected and categorized from this inventory into 27 bridge groups based on existing bridge management data, such as bridge age, maintenance history, traffic volume, and region. Then, the probability distribution of the bridge condition index (BCI) for these 27 groups was determined using the Anderson-Darling test, which revealed that the lognormal distribution is the best for describing the BCI across the bridge groups. Finally, multivariate linear regression analysis was applied to relate the distribution parameters to the bridge group characteristics, resulting in a suggested model for predicting the bridge condition in Ontario. The suggested model can aid safe and economic decisions on bridge maintenance.
