Probabilistic bridge deterioration prediction models based on Markov matrices using real and simulated data from deterministic models

Abstract

Abstract This study uses real and simulated information from 885 bridges in Brazil. A total of 2,655 available inspection data were collected from the database, and 37,170 additional data were simulated from deterministic deterioration prediction models developed in previous studies. The probabilistic Markov matrices-based models obtained include one covering all the bridges, specific models for non-aggressive and aggressive environments, and models for Average Daily Traffic (ADT) of less than and more than 4,000. Validation showed good metrics, with a coefficient of determination of 0.6268, a mean absolute error and mean squared error below 0.5, and an accuracy of 66.25%. Finally, these tools enable more accurate forecasting, and a better understanding of the risks associated with the deterioration of structures for safe and cost-effective bridge management.

Keywords

• bridge
• deterioration prediction model
• bridge management system
• deterministic model
• probabilistic model