Development of a Long-Term Bridge Element Performance Model Using Elman Neural Networks

Abstract

A reliable deterioration model is essential in bridge asset management. Most deterioration modeling requires a large amount of well-distributed condition rating data along with all bridge ages to calculate the probability of condition rating deterioration. This means that the model can only function properly when a full set of data is available. To overcome this shortcoming, an improved artificial intelligence (AI)-based model is presented in this study to effectively predict long-term deterioration of bridge elements. The model has four major components: (1) categorizing bridge element condition ratings; (2) using the neural network-based backward prediction model (BPM) to generate unavailable historical condition ratings for applicable bridge elements; (3) training by an Elman neural network (ENN) for identifying historical deterioration patterns; and (4) using the ENN to predict long-term performance. The model has been tested using bridge inspection records that demonstrate satisfactory results. This study primarily focuses on the establishment of a new methodology to address the research problems identified. A series of case studies, hence, need to follow to ensure the method is appropriately developed and validated.