Research paperCombining CFD and artificial neural network techniques to predict vortex-induced vibration mechanism for wind turbine tower hoisting
Introduction
When incoming air flows through a wind turbine tower structure, the boundary layer may be expected to separate due to pressure changes. This flow produces vortices, which then continue to diverge, resulting in periodic vortices. The generation and shedding of such vortices exerts cyclic forces along the flow direction around the tower and perpendicular to the direction of flow, respectively, which is known to cause vibration of the tower structure; this is referred to as vortex-induced vibration (VIV). VIV is a major cause of structural fatigue damage, especially in the process of tower hoisting [1]. When the vortex shedding frequency approximates the inherent frequency of the structure, a large “locking” motion known as vortex synchronization can be observed [2], which may result in abnormal operation of the wind turbine. For a tower without a nacelle, necessary measures should be taken to prevent vortex-induced vibration. Therefore, the development of methods to reduce or suppress tower vibration is of great significance.
Because wind turbine generator system (WTGS) towers are necessarily thin-walled hollow structures with a smooth external surface, the problem can be simplified as that of vortex-induced vibration caused by airflow around the cylinder. At present, two major methods have been explored in the relevant literature, including scale-model experiments and computational fluid dynamics (CFD). Irani [3] et al. tested the vortex-induced vibration of a turbine model at a low Reynolds number (Re), compared the test results with VIV responses measured in the field, and found no significant difference between the two datasets. Blackburn [4] et al. applied the direct numerical simulation (DNS) method by using 2D and 3D CFD models separately when Re = 556, and concluded that 3D simulation results were closer to the experimental data than 2D cases. Horcas [5] proposed an innovative aerodynamic coupling method to obtain the VIV shedding frequency and predict the aeroelastic response of a WTGS tower. Concerning the suppression of vortex-induced vibration, most studies have focused on surface characteristics. Kiu et al. [6] experimentally studied the inhibitory effect of surface roughness on VIV. Law and Jaiman [7] investigated the inhibitory effect of grooves in the spanwise direction, showing that the amplitude was suppressed to 37%, and the drag coefficient was reduced by approximately 25%. Sui [8] et al. explored VIV suppression in the case of a cylinder with spiral threads and found that the maximum suppression rate could reach up to 98%. Gao [9] et al. studied the impact of surface roughness on vortex-induced vibration and concluded that the lateral flow amplitude would be reduced with an increase in surface roughness.
At present, the digital implementation of the entire life cycle of wind turbines is a key direction in the industry. As a core part of the entire life cycle, the development of methods to simulate hoisting operations and maintenance are an important part of establishing a digital twin (DT) system of the entire life cycle. DT makes full use of physical models, sensors, and other data, and integrates multi-disciplinary, multi-physics, multi-scale, and multi-probability simulation processes in a virtual space to realize the dynamic characteristics of physical space [10]. Digital research on wind turbines requires a large amount of computational resources to model fluid dynamics. On the one hand, data are collected from experimental or field sensor measurement systems, and on the other, they are primarily generated from CFD simulation data. Therefore, combining digital model and CFD in simulation prediction research can address not only the problem of digitization, but also that of long computational time with a single model.
In this study, we calculated the vortex-induced vibration frequency of two types of towers (with and without a flow-disturbing device of triangular cross-section) under different wind conditions by using the CFD method to investigate corresponding cyclic oscillation rules, aiming to provide a theoretical basis for future engineering applications. Furthermore, considering the high computational cost of CFD approaches, particularly of sophisticated 3D CFD models, the development of methods to predict the aerodynamic performance of wind turbine towers with relatively high accuracy and efficiency is critical. Based on sensor data provided by the enterprise, we propose a hybrid metamodel prediction method. The CFD model is integrated into a digital model of a wind turbine and subsequently incorporated into an enterprise-scale DT model.
Section snippets
Background and problem description
At present, wind power enterprises combine a large amount of operating data from wind turbines with information on their operating status, and use technologies such as big data and artificial intelligence to detect and diagnose faults. Meanwhile, DT technology is gradually being introduced to optimize the design of wind turbine systems and perform and intelligent operation and maintenance. The key technologies of large-scale wind turbine systems include high-fidelity modeling and simulation
Computational methodology
For the low-speed incompressible flow, we used a pressure-based solver [13] to solve the governing equations, including the momentum, energy, and turbulent flow equations. The governing equations in computational fluid dynamics are in integral form and can be manipulated to obtain discrete equations based on the conversation of each variable.
The law of mass conservation can be expressed as follows [14], [15]: where, is the density of the fluid; is time; ,
Comparison of flow-disturbing device
The towers were modeled with no flow-disturbing device and with a spiral flow-disturbing device of triangular cross-section. The average wind speed ranged from 3 m/s to 10 m/s. Due to high computational cost, the wind velocity was set at 3 m/s, 7 m/s, and 10 m/s, respectively.
Prediction model and performance evaluation
From Section 3, it may be observed that the CFD simulation a considerable period of time. A 90-core workstation was used to simulate the 10-second vortex-induced vibration of the tower; the shortest calculation time was 10 h. Calculating CFD simulations with a large number of element nodes, requires increased time and equipment costs. In addition, a large amount of reanalysis is required for variations in wind conditions. To solve this problem, in this study, we use artificial neural networks
Conclusions
In this study, we have investigated the effects of a flow-disturbing device of triangular cross-section on vortex-induced vibration. Considering the high cost of CFD calculation, a hybrid metamodel has been proposed to analyze and predict tower vortex-induced vibration. Based on digital model, digital research on the hoisting, operation, and maintenance of wind turbines was conducted. Based on the results, we can draw the following conclusions.
(1) For the three wind velocities, the tower
CRediT authorship contribution statement
Yiming Chen: Conceptualization, Writing – original draft. Xin Jin: Methodology, Software. Peng Cheng: Calculation and writing of prediction part. Huali Han: Calculation, Investigation. Yang Li: Validation.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgment
This paper is sponsored by Natural Science Foundation of China (approval No.: 51975066, Pitch Bearing Crack Growth Evolution Mechanism and Risk Assessment of Offshore Wind Turbine Based on Meso-Macro Scale Correlation). The authors appreciate the anonymous reviewers for their valuable comments, which are helpful to improve the paper.
References (23)
- et al.
Numerical simulation of a freely vibrating circular cylinder with different natural frequencies
Ocean Eng
(2018) - et al.
A complementary numerical and physical investigation of vortex-induced vibration
J. Fluids Struct
(2001) - et al.
Effects of uniform surface roughness on vortex-induced vibration of towed vertical cylinders
J Sound Vib
(2011) - et al.
Passive control of vortex-induced vibration by spanwise grooves
J Fluid Struct
(2018) - et al.
VIV Suppression for a large massdamping cylinder attached with helical strakes
J Fluid Struct
(2016) - et al.
Numerical simulation of vortex- induced vibration of a circular cylinder with different surface rough nesses
Mar Struct
(2018) Cfd modeling of varying complexity for aerodynamic analysis of H-vertical axis wind turbines
Renew Energy
(2020)- et al.
Direct numerical simulations of fluid–solid systems using the arbitrary Lagrangian–Eulerian technique
J Comput Phys
(2001) - et al.
CFD Simulation of the atmospheric boundary layer: wall function problems
Atmos Environ
(2007) - et al.
Comparison of linear regression and artificial neural network model of a diesel engine fueled with biodiesel-alcohol mixtures
Alexandria Eng J
(2016)
Vortex-induced vibrations
Ann Rev Fluid Mech
Cited by (4)
Nonlinear vortex-induced vibration of wind turbine towers: Theory and experimental validation
2023, Mechanical Systems and Signal Processing