BMS Forecasting of Bridge Health Condition Degradation Using AI Machine Learning

Abstract Bridges are the vital spine of any efficient transportation network. Thereby, nations exert great effort to maintain modern bridge networks include the preservation of existing bridges. This can be achieved by effective Bridge Management System (BMS) implementation which includes maintenance and rehabilitation planning over all the network levels. One BMS vital component is the bridge structural condition degradation forecasting. This important component usually is a module tool in the BMS forecasts a bridge future structural condition, health, based on a given set of bridge present and past conditions and variables. This article presents a part of research for enhanced implementation of a MENA region particularly built BMS. The method aims to provide a simplified and robust bridge structural degradation forecasting tool as part of the national BMS to reach effective bridge maintenance planning and prioritization. The method uses datasets created by the authors’ teams for the actual exiting bridges in MENA region and applies the AI concepts of Decision Tree (DT), Bagging, Extreme Gradient Boosting (XGBoost) and tree-based ensemble machine learning algorithms. These different algorithms are performed on the subject bridge network datasets, then tested and evaluated and provided efficient forecasting. The results indicated a mean absolute percentage error (MAPE) of DT, Bagging, XGBoost models of 3.097%, 2.917%, and 2.6%, respectively. Eventually, the proposed XGBoost model is proven as a reliable model for a BMS utilisation.