Aerodynamic and aeroelastic responses of short gap twin-box decks: Box geometry and gap distance dependent surrogate based design

• Published in: Journal of wind engineering and industrial aerodynamics 2020-06, Vol.201, p.104147, Article 104147
• Main Authors: Nieto, Felix, Cid Montoya, Miguel, Hernández, Santiago, Kusano, Ibuki, Casteleiro, Alejandro, Álvarez, Antonio J., Jurado, José Á., Fontán, Arturo
• Format: Article
• Language: English
• Subjects: CFD; Deck shape; Flutter derivatives; Flutter speed; Gap distance; Quasi-steady load model; Surrogate modeling; Twin-box
• ISSN: 0167-6105; 1872-8197
• DOI: 10.1016/j.jweia.2020.104147

The aerodynamic and aeroelastic performances of twin-box deck bridges are difficult to anticipate when changes in the deck geometry are introduced. Considering moderate changes in the gap distance and in the box geometry, a surrogate model may provide the full picture of the aerodynamic properties, and even the critical flutter speed. In this work the boxes’ width and depth, and the slot between girders, are selected as input of a surrogate model. A set of 25 samples is defined using the Latin Hypercube Sampling method for training the Kriging surrogate, and for these samples, the force coefficients and their slopes are obtained by means of 2D URANS simulations. Force coefficients and flutter derivatives have been obtained experimentally for a subset of three samples. These experimental data have been used to validate the CFD results, finding a good agreement. Furthermore, based upon these experimental results, the values adopted for the aerodynamic centers in the quasi-steady approximation of the flutter derivatives have been corrected, improving their accuracy. Using the quasi-steady formulation, the critical flutter velocity is obtained over the whole design domain, finding that the gap distance and the boxes’ width are very influential in the aeroelastic performance of cable-supported bridges. •A surrogate model is defined that provides the aerodynamic response of twin-box decks.•The gap distance and the width and depth of the boxes are the input of the surrogate.•The CFD simulations of 25 randomly defined samples are used for training the surrogate.•Based on the quasi-steady formulation and the surrogate output, the flutter response is obtained over the design domain.