Prediction-Based Maintenance of Existing Bridges Using Neural Network and Sensitivity Analysis

Bridge deterioration is affected by various factors. However, neither the relationships between these factors and are explicitly determined, nor relative effect of each factor on well understood. This study proposed a methodology to resolve issues integrating an artificial neural network (ANN) sensitivity analysis method. The ANN was used predict deterioration, method applied evaluate influence deterioration. Testing with 3,368 bridge inspection data pieces indicates that (1) developed obtained accuracy about 65%; (2) seven were identified affecting established model has equivalent performance for three grades four types bridges. Two (the Shapley value Sobol indices) methods compared, they same five most important Consequently, can effectively avoid uncertainty providing importance list methodology’s predictive ability identification make it suitable decision-makers understand situations schedule further corresponding maintenance strategies.