Search

Wind energy as a renewable energy source has gained in popularity in recent years as a viable means to replacing fossil fuels [1]. Wind turbines form extremely sophisticated systems, operating under extreme and time-varying loads of polymorphic nature (e.g., wind, waves) and under adverse environments (highly varying temperatures and icing conditions). This implies that beyond the operation of mechanical components such as the gearbox, the robustness and resilience of structural components are of the essence but certainly overlooked so far in terms of monitoring. The operation of wind turbines relies on the use of supervisory control and data acquisition (SCADA) systems for their monitoring and control [2]. These systems typically measure operational quantities in and around the nacelle [3], such as wind speed and direction, generator temperature, as well as the generated power [4], and are provided as ten-minute averages. Currently, no standard monitoring solution exists that can be easily integrated to assess the performance of critical structural components, such as the blades (e.g., its aerodynamics) [5]. However, the need for this is becoming increasingly important as wind turbine dimensions rapidly increase and blades become more flexible [1]. Such integrated monitoring systems would need to withstand harsh weather and operational conditions on a blade, which is a nontrivial task. This makes published measurements on operating rotor blades in real conditions extremely rare [6]. However, recent advances in electronics, wireless communications, and micro-electromechanical systems (MEMS) have enabled the acquisition of data directly on the blade, in a cost-effective and energy-efficient way [7]. ISSN:1094-6969 ISSN:1941-0123

As MW-scale wind turbines are becoming larger with increasingly flexible blades, the need for sophisticated blade health monitoring techniques is growing. Recent work has shown that Microelectromechanical systems (MEMS) sensor strips are well suited for the purpose of monitoring wind turbine blade aerodynamics [1]. While these systems could potentially cover an entire airfoil section of a small wind turbine blade, for larger blades with more extensive chord lengths only relatively small segments can feasibly be instrumented. In such cases, it becomes important to place the sensing strips in optimal locations, which maximize the collection of pertinent data required for the downstream monitoring tasks (e.g. damage detection, angle of attack estimation, etc.). While these downstream tasks can be of diverse nature, we choose to focus on a more general problem: flow reconstruction. Reconstructing the flow around a blade entails using the pressure measured on its surface in order to extrapolate the features (pressure and velocity fields) of the surrounding flow. Reconstructed solutions can then be utilized to enhance the performance of other monitoring tasks. In previous work, we have demonstrated how Graph Neural Networks (GNNs) could be employed to reconstruct the flow field surrounding arbitrary 2D airfoils, using only as an input the surface pressure [2]. Building upon this framework, we frame the problem of optimal sensor placement in terms of maximizing a GNNs ability to reconstruct the surrounding flow field given variable partial airfoil surface pressure. This is a challenging optimization problem, which depends not only on the shape of the airfoil but also on the external operational and environmental conditions. We show how the expressivity and extrapolation capabilities of GNNs can be leveraged such that given an arbitrary airfoil shape, a set of input conditions and some physical sensor constraints, an optimal sensing patch location is selected such that a good estimation of the surrounding flow is obtained. Finally, by incorporating elements from Graph Sampling Theory, we highlight how to properly condition the learning objective for the reconstruction task., {"references":["Barber, S., Deparday, J., Marykovskiy, Y., Chatzi, E., Abdallah, I., Duthé, G., ... & Müller, H. (2022). Development of a wireless, non-intrusive, MEMS-based pressure and acoustic measurement system for large-scale operating wind turbine blades. Wind Energy Science, 7(4), 1383-1398.","Duthé, G., Abdallah, I., Barber, S., & Chatzi, E. (2023). Graph Neural Networks for Aerodynamic Flow Reconstruction from Sparse Sensing. arXiv preprint arXiv:2301.03228."]}

The Digital Twin (DT) conceptual model was initially introduced in the context of product life-cycle management by Grieves [2], and later adopted for Structural Health Monitoring (SHM) applications in various domains, including wind energy. The widespread adoption of DT-based systems can be attributed to a promise of increased accuracy of predictive capabilities and more rapid and informed decision-making [1]. In this work, we propose a classification for different types of DTs and demonstrate its value as a part of application ontology in the wind energy context. The classification is formalized and published as a taxonomy using a Simple Knowledge Organization System (SKOS) [8] data model. Since the introduction of the DT model, a broad spectrum of DT manifestations has emerged. The term itself takes on different interpretations, depending on the context and specific use case. For example, the Aerospace Industries Association, in their position paper, proposes a general definition of a DT as a "virtual representation of a connected physical asset" along with examples and added value of 17 different DT types [1]. This is, partially, "by design," as Grieves himself indicates only two essential attributes of a DT concept: duality and strong similarity [2]. At the same time, while developing a practical DT implementation, it becomes increasingly difficult to navigate the space of DTs and select appropriate underlying algorithms, software packages, and technology stacks. The proposed classification system is based on the differences present in the three components of a DT: the physical, virtual, and connection system. The taxonomy design follows principles such as clarity, coherence, extendibility, minimal encoding bias, and minimal ontological commitment, proposed by Gruber [3]. It builds upon existing and established taxonomies for DT components as well as established terminology for the DT types in the published literature. While prior work in this area has been conducted by [6, 7], [4], and [5], this research formalizes and evaluates the results from an ontological point of view. Importantly, this work does not aim to deliver a unified definition or claims to be a uniquely correct and static classification, but rather offers a pragmatic approach to the adoption of the DT conceptual model. The choice of the SKOS implementation allows for publishing, sharing, and linking of the proposed taxonomy, and invites further community discussion on the topic.

Data from field tests of Aerosesne measurement system installed on Avneta AV-7 wind turbine, located in Taggenberg (ZH), Switzerland. Dataset contains intermittent 10 minute long measurement sessions from 40 barometers, and IMU system installed on the wind turbine blade at r/R = 0.740625. The measurement campaign was performed between 01.DEC.2022 and 01.APR.2023. All the samples are timestamped with UTC time. Wind turbine, sensor location and sensor properties metadata is provided as JSON files.

Research question: Can sectional aerodynamic pressure over an airfoil be used to detect structural damage (cracks) on an aeroelastic structure (wind turbine blade)?

Data from wind tunnel tests of Aerosesne measurement system installed on NACA 0012 at ETH Zurich wind tunnel.

Data from wind tunnel tests of Aerosesne measurement system installed on NACA63418 at EPFL wind tunnel.

Wind turbine fatigue estimation is based on time-consuming Monte Carlo simulations for various wind conditions, followed by cycle-counting procedures and the application of engineering damage models. The outputs of the fatigue simulations are large in volume and of high dimensionality, as they typically consist of estimates on finite-element computational meshes. The strain and stress tensor time series, which are the primary quantities of interest when considering the problem of fatigue estimation, are dictated by complex vibration characteristics due to the coupled effect of aerodynamics, structural dynamics, geometrically non-linear mechanics, and control. A Variational Auto-Encoder (VAE) is trained in order to model the probability distribution of the accumulated fatigue on the root cross-section of a simulated wind turbine blade. The VAE is conditioned on historical data that correspond to coarse wind-field measurement statistics, such as mean hub-height wind speed, standard deviation of hub-height wind speed and shear exponent. In the absence of direct measurements of structural loads, the proposed technique finds applications in making long-term probabilistic deterioration predictions from historical Supervisory, Control, and Data Acquisition (SCADA) data, while capturing the inherent aleatoric uncertainty due to the incomplete information on strain time series of the wind turbine structure, when only SCADA data statistics are available.

Leading edge surface erosion is an emerging issue in wind turbine blade reliability, causing a reduction in power performance, aerodynamic loads imbalance, increased noise emission, and, ultimately, additional maintenance costs, and, if left untreated, it leads to the compromise of the functionality of the blade. In this work, we first propose an empirical spatio-temporal stochastic model for simulating leading edge erosion, to be used in conjunction with aeroelastic simulations, and subsequently present a deep learning model to be trained on simulated data, which aims to monitor leading edge erosion by detecting and classifying the degradation severity. This could help wind farm operators to reduce maintenance costs by planning cleaning and repair activities more efficiently. The main ingredients of the model include a damage process that progresses at random times, across multiple discrete states characterized by a non-homogeneous compound Poisson process, which is used to describe the random and time-dependent degradation of the blade surface, thus implicitly affecting its aerodynamic properties. The model allows for one, or more, zones along the span of the blades to be independently affected by erosion. The proposed model accounts for uncertainties in the local airfoil aerodynamics via parameterization of the lift and drag coefficients’ curves. The proposed model was used to generate a stochastic ensemble of degrading airfoil aerodynamic polars, for use in forward aero-servo-elastic simulations, where we computed the effect of leading edge erosion degradation on the dynamic response of a wind turbine under varying turbulent input inflow conditions. The dynamic response was chosen as a defining output as this relates to the output variable that is most commonly monitored under a structural health monitoring (SHM) regime. In this context, we further proposed an approach for spatio-temporal dependent diagnostics of leading erosion, namely, a deep learning attention-based Transformer, which we modified for classification tasks on slow degradation processes with long sequence multivariate time-series as inputs. We performed multiple sets of numerical experiments, aiming to evaluate the Transformer for diagnostics and assess its limitations. The results revealed Transformers as a potent method for diagnosis of such degradation processes. The attention-based mechanism allows the network to focus on different features at different time intervals for better prediction accuracy, especially for long time-series sequences representing a slow degradation process., Energies, 14 (21), ISSN:1996-1073

Identifying the extent of erosion on the leading edge of wind turbine blades is becoming increasingly important in the wind energy industry as blades become larger and more flexible. Leading edge erosion (LEE) can result from abrasive airborne particles or weather conditions, and can impact the Annual Energy Production of a MW-scale wind turbine on the order of 5% (Langel et al., 2015). Current methods for identifying LEE involve manual (Nielsen et al., 2020) or drone-based visual inspection (Shihavuddin et al., 2019), electrical signal analysis (He et al., 2020) or vibration monitoring (Skrimpas et al., 2016) , methods which either require the turbine to be shut down or are limited for continuous monitoring (Du et al., 2020). In this work, we propose a data-driven model to predict the state of degradation of the leading edge of a 2D airfoil via aerodynamic pressure coefficient learning, under the influence of various uncertain inputs and parameters. The learning-based algorithm is trained on a novel dataset, comprised of pressure coefficient distributions that are generated by 2D steady-state Computational Fluid Dynamics (CFD) simulations. Each individual simulation is executed with a different set of input parameters, constructed via probabilistic sampling, with distributions designed to mimic operational conditions of a wind turbine. The airfoil LEE status can be broadly sorted into four categories corresponding to the severity of the degradation (Sareen et al., 2014): (1) undamaged, (2) presence of pits, (3) presence of pits and gouges and (4) presence of pits, gouges, and delamination. This categorization forms the prediction goal of the learning algorithm. In our CFD setup, which is based on the k-Omega SST Reynolds-averaged Navier–Stokes turbulence model using OpenFOAM, the surface conditions of the leading edge are emulated via the use of rough wall-functions, based on previous work by (Knopp et al., 2009). By adjusting the sand-grain roughness and the spatial extent of the rough patch, we can simulate all degradation categories. We validate this approach by comparing our simulation results in terms of lift, drag and pressure coefficients to experimental data and CFD simulation data from the literature (Maniaci et al., 2016). We aim to train a model which is robust to uncertain conditions and which can generalize to any type of airfoil. This requires a dataset established across a wide range of airfoil shapes and flow conditions. To increase robustness to shape warping induced by blade deflections, we propose a novel generative learning method to produce unique, random yet coherent, airfoil geometries, which are subsequently used as the basis for the CFD simulation meshes. As a result, during training, the learning algorithm is never exposed to multiple samples with the exact same airfoil geometry which allows for better generalization. By using random probabilistic sampling to draw the airfoil geometry and the other CFD flow condition inputs, we create an ensemble of unique and diverse pressure coefficient curves The deep learning algorithm we propose admits as inputs the pressure coefficient distribution and geometry of an airfoil, and outputs one of four degradation labels. A neural network is chosen as the data-driven model, although we test and compare multiple different network types and architectures. We show classification scores for all models and discuss their suitability for the task at hand, as well as potential improvements., {"references":["Du, Y., Zhou, S., Jing, X., Peng, Y., Wu, H., Kwok, N., 2020. Damage detection techniques for wind turbine blades: A review. Mech. Syst. Signal Process. 141, 106445. https://doi.org/10.1016/j.ymssp.2019.106445","He, L., Attia, M., Hao, L., Fang, B., Younsi, K., Wang, H., 2020. Remote Monitoring and Diagnostics of Blade Health in Commercial MW-Scale Wind Turbines Using Electrical Signature Analysis (ESA), in: 2020 IEEE Energy Conversion Congress and Exposition (ECCE). Presented at the 2020 IEEE Energy Conversion Congress and Exposition (ECCE), pp. 808–813. https://doi.org/10.1109/ECCE44975.2020.9235984","Knopp, T., Eisfeld, B., Calvo, J.B., 2009. A new extension for k–ω turbulence models to account for wall roughness. Int. J. Heat Fluid Flow 30, 54–65. https://doi.org/10.1016/j.ijheatfluidflow.2008.09.009","Langel, C.M., Chow, R., Hurley, O.F., Van Dam, C. (CP) P., Maniaci, D.C., Ehrmann, R.S., White, E.B., 2015. Analysis of the Impact of Leading Edge Surface Degradation on Wind Turbine Performance, in: 33rd Wind Energy Symposium. Presented at the 33rd Wind Energy Symposium, American Institute of Aeronautics and Astronautics, Kissimmee, Florida. https://doi.org/10.2514/6.2015-0489","Maniaci, D.C., White, E.B., Wilcox, B., Langel, C.M., van Dam, C.P., Paquette, J.A., 2016. Experimental Measurement and CFD Model Development of Thick Wind Turbine Airfoils with Leading Edge Erosion. J. Phys. Conf. Ser. 753, 022013. https://doi.org/10.1088/1742-6596/753/2/022013","Nielsen, M.S., Nikolov, I., Kruse, E.K., Garnæs, J., Madsen, C.B., 2020. High-Resolution Structure-from-Motion for Quantitative Measurement of Leading-Edge Roughness. Energies 13, 3916. https://doi.org/10.3390/en13153916","Sareen, A., Sapre, C.A., Selig, M.S., 2014. Effects of leading edge erosion on wind turbine blade performance: Effects of leading edge erosion. Wind Energy 17, 1531–1542. https://doi.org/10.1002/we.1649","Shihavuddin, A.S.M., Chen, X., Fedorov, V., Nymark Christensen, A., Andre Brogaard Riis, N., Branner, K., Bjorholm Dahl, A., Reinhold Paulsen, R., 2019. Wind Turbine Surface Damage Detection by Deep Learning Aided Drone Inspection Analysis. Energies 12, 676. https://doi.org/10.3390/en12040676","Skrimpas, G.A., Kleani, K., Mijatovic, N., Sweeney, C.W., Jensen, B.B., Holboell, J., 2016. Detection of icing on wind turbine blades by means of vibration and power curve analysis. Wind Energy 19, 1819–1832. https://doi.org/10.1002/we.1952"]}