ABSTRACT
Road bridges are life line structures essential to ensure community mobility. Management of an aging bridge network on limited resources can be a challenge for the state authorities. Prediction of future condition states of bridges at component level is essential to optimise maintenance actions and ensure the safety and serviceability of the network. This is the first step in developing an advanced management strategy. A research project funded by the Australian Research Council in partnership with the Roads Corporation of Victoria (VicRoads) and conducted at RMIT University in Melbourne examined the level 2 visual inspection data collected by over 4–5 inspection cycles with the objective of predicting deterioration of bridge components. The quality and quantity of the data were analyzed and observed to be adequate for developing component level deterioration prediction models. A range of deterioration prediction methods including the deterministic models and stochastic models were examined to understand the validity of the different methods in predicting the deterioration of bridge components using visual inspection data. A case study with a demonstration on concrete open girder is investigated for the well-known linear models and the commonly used Markov deterioration model. The results show that the Markov model can produce more reasonable prediction than the linear model when presented with short time inspection data. Furthermore, preparation of valid data is also important to ensuring accurate and sensible prediction from deterioration models. This study also shows that the issue of first inspection has significant impact on predictive outcomes and thus should be carefully considered in preparation of valid data.
