Tools for diagnosis and prediction of deterioration of timber bridges in Australia

Australia has a large number of aging timber bridges currently in need of new and effective management practices. Timber bridges have a high maintenance cost and are affected significantly by a number of deterioration mechanisms which require a systematic approach for diagnosis and treatment. <br><br>Early diagnosis of possible deterioration scenarios can lead to effective management strategies. Furthermore, infrastructure managers need predictive models for condition assessment to optimise the repair and maintenance management process over the life cycle of a given timber bridge. No published research work to date has been comprehensively addressing a deterioration diagnosis system and a forecasting model specifically validated for timber bridges. The present work addresses this gap in knowledge.<br><br>The research presented in this thesis is intended to facilitate understanding of the deterioration mechanisms affecting timber bridges, the development of an expert system capturing the practical expertise and the development of a practical deterioration prediction model based on the Markov process for timber bridge elements in Australia. The expert system can diagnose the deterioration mechanisms affecting timber bridges based on early signs of distress. The aims of the timber bridge diagnostic system are not only to help engineers to make appropriate decisions on the repair of existing deficiencies, but also to take corrective measures to prevent or reduce future deterioration by identifying the distress mechanisms. This required capturing the complexity of relating different types of defects, their signs and possible causes into a simple algorithm. The deterioration prediction model required understanding of the deterioration of major bridge elements with the age of the element so that infrastructure managers could make appropriate decisions on repair strategies and program maintenance schedules by accurately predicting the future condition of timber bridge elements.<br><br>The development of the expert system which aids the non-expert to diagnose the cause of distress in timber bridges utilised cause and effect diagrams and a fuzzy logic approach. The inputs to the system are linguistic variables such as the type of element, the visual symptoms, the environmental conditions, method of construction and the location of the bridge. The expert system executes fuzzy inference to evaluate the cause of the distress using these input data and built-in rules. The rules were initially developed based on data available on timber bridges and were then validated with the assistance of an expert engineer.<br><br>Markov models have been used extensively in modelling the deterioration of infrastructure facilities. These models can predict the conditions of bridge elements as a probabilistic estimate and the method is often used when the condition data have a high scatter and a deterministic estimate of deterioration prediction therefore becomes unrealistic. Condition data obtained from VicRoads and two local councils, Strathbogie Shire Council and the Corangamite Shire Council, have been used to develop transition probabilities of a Markov model. The percentage prediction method, regression-based optimisation method and non-linear optimisation technique have been used to predict transition matrices and transient probabilities from condition data. The most suitable deterioration model for timber bridge elements has been selected by evaluating the model performance using the goodness-of-fit test. The application of the model for decision-making is demonstrated using costing information for one bridge element.<br>