A Simplified Approach to Recognize Vortex-Induced Vibration Response Using Machine Learning

Abstract

The vortex-induced vibration (VIV) problem has been of critical concern for the wind-resistance of long-span bridges. Usually there are four types of approach for VIV studies: wind tunnel tests, field monitoring, computational fluid dynamics and mathematical models. However, traditional approaches have shown some limitations, such as high cost and low efficiency. In order to improve the efficiency and accuracy of VIV studies, this article has taken the VIV problem of a split three-box girder in a cable-stayed and cooperative suspension system bridge as an instance, and conducted a series of VIV wind tunnel tests. An approach based on machine learning is described that is able to serve as a complement to the wind tunnel tests. The proposed approach involves two steps: firstly, based on the dataset produced by wind tunnel tests, a clustering algorithm is introduced to separate the VIV signals automatically from other vibrations. Then, an artificial neural network is utilized to recognize the VIV response and aerodynamic force in the lock-in region directly. It is shown that the clustering algorithm can be a good tool for the recognition of VIV signals. Moreover, the proposed artificial neural network models show good ability for recognizing VIV amplitude and aerodynamic lift force.