A modelling approach for evaluating the effects of design variables on bridge condition ratings
Abstract
AbstractWhile routine inspections are commonly used to assess the structural integrity, safety and maintenance needs of individual highway bridges, data from these inspections can also be used to study performance of bridges at the inventory level. This paper presents a novel method by which inspection data can be used to evaluate design variables and inform future designs. In particular, inspection data from prestressed concrete bridges in south-eastern United States were used to develop artificial neural networks models for estimating the condition rating of bridge decks and superstructures as a function of skew angle and span length, as well as, bridge age, width and traffic level. Once developed and validated, the neural network models were used for an array of simulations that were designed using a full factorial approach. The objective of the simulations was to identify skew angles and span lengths that correlate with the highest inspection ratings. It was determined that deck ratings are highest for smaller skew angles and shorter span lengths, whereas superstructure ratings are minimally impacted by larger skews and unrelated to span length. The conclusions of this study will be helpful in understanding the implications of bridge design variables on the long-term performance of bridge decks and superstructures. Though the trends and conclusions noted in this study are to be seen within the scope of the data considered, the approach demonstrated in this paper can be applied to address other questions of bridge performance.