Assessment of bolted connections for supporting structures of offshore wind turbine towers

Abstract

In the past two decades, offshore wind has emerged as a new source of renewable energy. This highlights the requirement for the utilisation of larger and more efficient offshore wind turbines (OWTs). The connections used in support structures of OWTs are critical to ensure the excellent structural performance of OWFs. An alternative option is the C1 wedge connection (C1-WC) to join virtually all the wind turbine generator (WTG) towers to their foundations. This connection shows promising potential in reducing construction, installation, and maintenance costs by eliminating the ring flange and using smaller diameter bolts.
Till now, C1-WC has undergone three generations of development. A more comprehensive research program is required to explore its implementation in a wind farm. The load transfer mechanism and critical component of C1 wedge connections are different to the conventional bolted ring flange (RF) connections. It is important to understand the mechanical behaviour of this connection. Meanwhile, support structures are exposed to the harsh environment during the service life of the OWTs. Material degradation and local cracks in the connection are inevitable to affect the serviceability of OWTs. A need for a reliable and rigorous structural health monitoring (SHM) system for the connection is evident. As one of the non-destructive techniques (NDT), Acoustic emission (AE) has been extensively used in early damage detection and real-time assessment of steel structures. Despite its successful applications, challenges still exist in using AE technique for monitoring applications, especially in analysing the recorded data. Therefore, the research aims to assist in understanding mechanical behaviour and evaluating the health status of the innovative connection.
An extensive experimental program was conducted to evaluate the static and cyclic behaviour of the C1-WCs. Additionally, a detailed 3D non-linear finite element (FE) model of the C1-WCs has been developed. The incorporation of material non-linearity and ductile damage allows the FE model to model the post-necking and final fracture of the connection. The FE model replicates with a good agreement the experimental tensile static and cyclic tests up to the final damage and reproduces the joint behaviour correctly. Parametric studies investigate the influence of bolt grade, the friction coefficient between contact surfaces, and the preloading force level on mechanical behaviour. Moreover, a quantitative comparison between C1-WC and two types of connections (RF connection and RF connection with defined contacts) is performed to provide practical insights into the selection and application of such connections and further optimization. FE-assisted analyses were performed to examine the effect of applied boundary conditions, bolt pretension level, and steel grade on the behaviour of the connections.
In addition to mechanical behaviour analysis. this research is also focused on developing data processing methods to address the challenges of AE monitoring for the C1-WCs. A hybrid model is proposed to identify the deformation stage of metal material. This method combines a self-adaptive denoising technique and an Artificial neural network (ANN). To reduce noise in the AE signals, a decomposition-based denoising method is proposed based on singular spectral analysis (SSA) and variable mode decomposition (VMD), referred to as SSA-VMD. After denoising, an ANN is constructed to identify the deformation stage of steel materials using features extracted from the filtered AE signals as input.
Fatigue damage of the C1-WCs could result in catastrophic failure of OWTs. Due to space constraints, it can be challenging to detect surface cracks in the lower segment holes of the C1-WC using commercial sensors. Thin PZT sensors are lightweight and small, making them suitable for use in restricted-access areas. However, their poor signal-to-noise ratio can limit their effectiveness in AE monitoring. A criterion for selecting the optimal thin PZT sensors is proposed and a configuration is designed for multiple sensors. Two signal processing methods are then proposed in terms of this issue. Firstly, a data fusion-based method is proposed to enhance the functionality of thin PZT sensors in AE applications. Convolutional neural networks (CNNs) combined with principal component analysis (PCA) are employed for signal processing and data fusion. Secondly, a baseline-based method is proposed to provide early warning of the fatigue damage of C1-WCs using thin PZT sensors. A benchmark model correlating to the damage state is created by breaking pencil leads. Multi-variate feature vectors are extracted and then mapped to the Mahalanois distance for identification.
Based on this research work, an efficient FE method has been developed to further improve the design of C1-WC. By providing an in-depth guideline for evaluating the mechanical performance of connections used in OWTs, this research has the potential to contribute to the development of more robust and reliable wind turbine structures. Moreover, the proposed signal processing methods for identifying the deformation stage and early fatigue damage can be further explored in structures with similar damage mechanisms. This can lead to the development of more accurate and effective methods for monitoring and assessing the health of offshore wind turbines, ultimately contributing to improved safety and reliability in the renewable energy industry.