Prediction of wind-induced vibrations of twin circular cylinders based on machine learning. (1st November 2021)

• Record Type: Journal Article
• Authors: Gu, Shanghao; Wang, Junlei; Hu, Gang; Lin, Pengfei; Zhang, Chengyun; Tang, Lihua; Xu, Feng
• Is Part Of: Ocean engineering. Volume 239(2021)
• Publication Date: 2021-11-01
• Subjects: Twin cylinders -- Circular cylinder -- Amplitude prediction -- Machine learning -- Gradient boosting regression trees; Ocean engineering -- Periodicals; Ocean engineering; Periodicals; 620.4162
• DOI: 10.1016/j.oceaneng.2021.109868
• Languages: English
• ISSNs: 0029-8018

Abstract

Abstract: This paper investigates vortex-induced vibrations (VIV) of a pair of circular cylinders with identical diameters at tandem and staggered configurations. The lattice Boltzmann method was used for two-dimensional computational fluid dynamics (CFD) simulations. Both cylinders were installed elastically and vibrate transversely and the Reynolds number Re = 150. The computation results reveal that mass ratio M ∗, the angle between the centerline of two cylinders and the fluid flow direction α, the reduced velocity U ∗ and the ratio of the distance between the center lines of the two cylinders L ∗ play key roles in the amplitude of the upstream cylinder and downstream cylinder. Subsequently, this study selected the above four parameters as input feature and trained two machine learning models to predict the amplitude of the upstream cylinder and the downstream cylinder, respectively. Three machine learning algorithms, namely decision tree regressor (DTR), random forest (RF), and gradient boosting regression trees (GBRT), were tested. Among them, the GBRT model performed optimally in predicting the amplitude of both the upstream and downstream cylinders. The GBRT model is capable of predicting the amplitude of the upstream and downstream cylinders precisely within the test range of M ∗, α, U ∗ and L ∗ . Highlights: This paper studies the amplitudes of twin cylinders with different configurations. Machine learning (ML) algorithms are employed to predict the amplitudes. Mass ratio, the angle and the distance ratio and the reduced velocity are studied. The performance of three algorithms (DTR, RF, GBRT) were tested and verified. This paper shows the application potential of ML in fluid-structure interaction.