Probabilistic Life-Cycle Connectivity Assessment of Transportation Networks Using Deep Learning

Abstract

Bridges and pavements are two major infrastructure components of a transportation network providing mobility of freight and commodities for economic vitality and access to a range of users as social benefits. However, the lack of a comprehensive infrastructure management system incorporating bridges and pavements inhibits accurate performance prediction, optimal maintenance actions, and the associated use of shrinking budgets. This paper presents an integrated probabilistic life-cycle connectivity framework for the performance analysis of transportation networks containing bridges and asphalt pavements by considering flexural and shear failure modes for prestressed concrete and steel bridges and four failure modes, including international roughness index, rut depth, alligator cracking, and transverse cracking, for asphalt pavements. In this framework, neural network–based deep learning models are used to assess the probabilistic performance of transportation networks and to provide guidance for the associated maintenance strategies. An existing transportation network consisting of bridges and asphalt pavement segments is selected to investigate its life-cycle connectivity reliability and component importance using the matrix-based system reliability method. Results show that the consideration of asphalt pavement failure probability has a significant effect on the probability of transportation network connectivity.