Predicting Load-Deflection of Composite Concrete Bridges Using Machine Learning Models

Abstract

The main objective of this study is to predict accurately the load-deflection of composite concrete bridges using two popular machine learning (ML) models namely Random Tree (RT) and Artificial Neural Network (ANN). Data from 83 track loading tests conducted on various bridges in Vietnam were collected and analyzed. Various input parameters namely bridge's cross-sectional shape, length of concrete beam, number of years in use, height of the main girder, distance between the main girders were selected for the modelling. Validation indicators like R, RMSE, and MAE, and Taylor diagram were used for validation and comparison of the models. Results of this study showed that both RT and ANN are good for prediction of the load-deflection of composite concrete bridges, but RT outperforms ANN. Thus, the developed ML models can facilitate efficient bridge health monitoring and management by predicting the load-deflection of simple-span concrete bridges.