Vibration-based cable condition assessment: A novel application of neural networks. (15th December 2018)

• Record Type: Journal Article
• Authors: Haji Agha Mohammad Zarbaf, Seyed Ehsan; Norouzi, Mehdi; Allemang, Randall; Hunt, Victor; Helmicki, Arthur; Venkatesh, Chandrasekar
• Is Part Of: Engineering structures. Volume 177(2018)
• Page Start: 291
• Page End: 305
• Publication Date: 2018-12-15
• Subjects: Cable-stayed bridges -- Vibration-based cable tension estimation -- Cable condition assessment -- Cable health monitoring -- Artificial Neural Networks (ANNs); Structural engineering -- Periodicals; Structural analysis (Engineering) -- Periodicals; Construction, Technique de la -- Périodiques; Génie parasismique -- Périodiques; Pression du vent -- Périodiques; Earthquake engineering; Structural engineering; Wind-pressure; Periodicals; 624.105
• DOI: 10.1016/j.engstruct.2018.09.060
• Languages: English
• ISSNs: 0141-0296

Abstract

Highlights: Proposing ANNs as a tool to estimate the cable tension in cable structures. Using L, m, EA, and cable natural frequencies to train the ANNs. Validation of proposed methodology using experimental data. Abstract: Vibration-based cable tension estimation methods demand complex computations especially when usage of comprehensive cable models is required. Avoiding mathematical calculations, this paper proposes a simple novel framework to estimate the cable tension based on Artificial Neural Networks (ANNs). Employing a comprehensive cable model, a set of data including cable length, cable mass per unit length, cable axial stiffness, cable bending stiffness, cable tension and the corresponding cable natural frequencies is generated for training, validation, and testing of the ANNs. The acquired ANNs are then used to estimate the cable tensions in new Ironton-Russell Bridge and the results are compared against the cable tensions directly measured by lift-off test. It will be shown that for new Ironton-Russell Bridge, using cable length, cable mass per unit length, cable axial stiffness, and first two cable natural frequencies as input features to ANNs, the cable tensions can be accurately estimated.