Forecasting long-term bridge deterioration conditions using artificial intelligence techniques

摘要: About a quarter of 600,000 bridges in the USA are deficient. The objective this study is to make accurate predictions future bridge deterioration condition using artificial neural networks, and ensure developed models applicable for practical use. Using inspection data provided by Connecticut Department Transportation ConnDOT, networks-based model modelling complex relationships between input output identify patterns within data. net variables included geometry, construction service, while were ratings deck, superstructure, substructure. nets used research demonstrated an ability produce results down root mean square error 10.05% on best trial. This shows potential develop tool predict assist agencies program planning.