ANN-Based Bridge Condition Rating Models Using Limited Structural Inspection Records

Abstract

The total expenditure for bridge maintenance, repair and rehabilitation (MR&R) and bridge asset extension increases rapidly every year in Australia. Computer-aided Bridge Management Systems (BMSs) are used to establish the best possible bridge MR&R strategies which ensure an adequate level of safety at the lowest possible bridge life-cycle cost. To achieve this, keeping up-to-date bridge information is crucial for a BMS software package. Although most bridge agencies in the past have conducted inspections and maintenance, the format of such bridge inspection records is dissimilar to those required by BMSs. These data inconsistencies inhibit correct BMS implementations. This paper presents an Artificial Neural Network (ANN) based prediction model, called the Backward Prediction Model (BPM), for generating historical bridge condition ratings using very limited existing inspection records. The BPM employed historical non-bridge datasets such as traffic volumes, populations and climates, to establish correlations with the existing bridge condition ratings from the very limited bridge inspection records. Such correlations can help fill the condition rating gaps required for an effective and accurate BMS implementation. The outcome of this study can contribute to minimising BMS operational problems due to limited inspection records.