Conference Agenda

As wind turbines operate under highly stochastic environmental and operational conditions (EOCs), robust Condition Monitoring (CM) strategies are essential for minimizing unscheduled downtime. Recent research suggests that integrating Trend Monitoring (TM) with Change Point Detection (CPD) can enhance diagnostic sensitivity by capturing both monotonic degradation and discrete structural shifts. This study provides the first experimental validation of an integrated TM–CPD framework using vibration data from wind turbine blades. Two laboratory case studies were conducted: leading-edge erosion (representing quasi-static surface degradation) and twist misalignment (representing a dynamic aerodynamic fault). For erosion, TM via Discrete Wavelet Transform (DWT) energy coefficients in the mid-frequency bands (3rd to 5th levels) showed a consistent increase (+34–45%), effectively characterizing the degradation trend. Bayesian Online Change Point Detection (BOCPD) provided no additional gain, confirming that TM is sufficient for steady-state faults. Conversely, for twist misalignment, Short-Time Fast Fourier Transform (STFT) revealed gradual spectral drifting, while BOCPD applied to the spectral centroid yielded a sharp, localized probability peak corresponding to the onset of aerodynamic imbalance. This integrated approach significantly improves the Probability of Detection (PoD) for dynamic faults. From an industrial perspective, this hybrid methodology mitigates the risk of catastrophic failure and optimizes the Levelized Cost of Energy (LCOE) by transitioning from reactive to predictive maintenance. Environmentally, the early detection of aerodynamic faults ensures optimal power curve performance, maximizing carbon-free energy yield and extending the structural fatigue life of the turbine assembly. This work establishes an evidence-based guideline: CPD is critical for faults inducing spectral instability, while TM suffices for smooth, monotonic degradation.

The continuous increase in the number of wind turbines is one of the major innovations of the 21st century worldwide for generating renewable energy and simul taneously reducing greenhouse gases. Since maintenance by mechanics only takes place at intervals without continuous monitoring and is associated with high profit losses due to shutdowns, intelligent sensors and algorithms are used to monitor the entire structure, in particular wind turbine blades (WTBs). The field study by Maelzer et al. [1] and the fatigue test in the laboratory by Simon et al. [2] have shown that a radar-based approach in the millimeter-wave frequency range for structural health monitoring of WTBs is promising.

This work describes the overall system design of a radar-based SHM system and its field deployment in a wind turbine. For this purpose, four sensor boxes are mounted on the main web of two rotor blades of a wind turbine. Each sensor box contains a frequency-modulated continuous wave (FMCW) radar at 60GHz and an acceleration sensor. Furthermore, one control unit per blade is mounted in the hub. Eachsensor box is connected to the corresponding control unit using the Power over Dataline standard for independent communication [3]. One of the three control units is connected to the mobile network for data transfer. A reference damage model developed by Rao et al. [4] and previously implemented on a WTB segment is applied here to simulate an artificial delamination. Initial measurement results are presented taking the changing environmental and operational conditions of the wind turbine into account.

References

[1] MORITZ MÄLZER, SEBASTIAN BECK, SERCAN ALIPEK, ELIAS REICHART, JOCHEN MOLL, VIKTOR KROZER, CHRISTOS OIKONOMOPOULOS, JÜRGEN KASSNER, MANFREDHÄGELEN,THOMASHEINECKE,etal. Radar-based struc tural monitoring of wind turbines blades: Field results from two operational wind tur bines. STRUCTURAL HEALTH MONITORING 2023, 2023.

[2] Jonas Simon, Thomas Kurin, Jochen Moll, Oliver Bagemiel, Raphael Wedel, Stefan Krause, Fabian Lurz, Andreas Nuber, Vadim Issakov, and Viktor Krozer. Embedded radar networks for damage detection in wind turbine blades: Validation in a full-scale fatigue test. Structural Health Monitoring, 22(6):4252–4263, 2023.

[3] Tobias Huemmer, Thomas Kurin, Moritz Maelzer, Katharina Fiedler, Sebastian Beck, Jochen Moll, and Fabian Lurz. System architecture of a radar-based structural health monitoring system for wind turbine blades using power over dataline. In IEEE Sensors 2025. IEEE, 2025.

[4] Manuel E Rao, Jochen Moll, Peter Kraemer, and Viktor Krozer. Experimental application of a reversible reference damage model for radar-based shm of glass fiber reinforced polymer structures. In 2025 IEEE 12th International Workshop on Metrology for AeroSpace (MetroAeroSpace), pages 780–784. IEEE, 2025.

Maintaining the operational efficiency and extending the service life of wind turbines are critical objectives in the wind energy sector, as structural degradation can lead to costly unscheduled downtime and increased maintenance costs. Structural Health Monitoring (SHM) solutions are thus fundamental for timely damage detection and optimized maintenance interventions. However, conventional SHM methods, often reliant on manual data analysis, are inherently time-consuming and prone to delays.

This research introduces a fully automated SHM workflow, leveraging Simcenter Testlab Workflow Automation (TWA) to streamline data management and monitoring tasks. Tasks include continuous acquisition of vibration data via both electrical and optical sensors, automated pre-processing for noise mitigation, and advanced Operational Modal Analysis (OMA) algorithms for extracting and tracking modal parameters. The methodology employs Simcenter Operational Polymax for modal parameter estimation, coupled with an Automated Modal Parameter Selection (AMPS) technique utilizing the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, and an Automated Modal Tracking (AMT) procedure applying statistical thresholds on modal parameters. By eliminating manual intervention, this automated pipeline efficiently tracks resonance modes and detects subtle structural changes across varying turbine operational states.

The workflow efficacy is established through comprehensive benchmarking on a laboratory-scale WT prototype, which allows controlled replication of diverse operating conditions and robust evaluation of the automated methodology. Building upon successful laboratory validation, the framework will be deployed on a full-scale operational wind turbine to demonstrate real-world applicability.

This study addresses the challenge of high-dimensional vibration data generated by dense accelerometer networks used for structural health monitoring (SHM) of offshore wind turbine jacket-type structures. The experimental dataset was obtained from a scaled jacket foundation instrumented with eight triaxial accelerometers, producing a high-dimensional feature space that increases computational cost and may degrade classification performance due to redundancy and noise. To overcome this limitation, a Parametric Uniform Manifold Approximation and Projection (Parametric UMAP) model is employed to perform nonlinear dimensionality reduction and extract compact latent representations of the vibration responses. The reduced feature set is then used to train a Light Gradient Boosting Machine (LightGBM) classifier for the automatic identification of four bolted joint conditions: healthy state (12 N·m), two intermediate loosening levels (9 N·m and 6 N·m), and complete bolt loss. The proposed Parametric UMAP–LightGBM framework achieves high classification accuracy while significantly reducing the dimensionality of the original dataset and improving computational efficiency. The results demonstrate that the combination of nonlinear dimensionality reduction and gradient boosting classification constitutes an effective and robust strategy for vibration-based damage assessment in bolted joints. This approach provides a scalable solution for real-time SHM and establishes a strong foundation for integration into digital twin platforms and predictive maintenance systems for offshore wind energy infrastructure.