Search

Complex systems are susceptible to many types of anomalies, faults, and abnormal behavior caused by a variety of off-nominal conditions that may ultimately result in major failures or catastrophic events. Early and accurate detection of these anomalies using system inputs and outputs collected from sensors and smart devices has become a challenging problem and an active area of research in many application domains. In this article, we present a new Bayesian hierarchical framework that is able to model the relationship between system inputs (sensor measurements) and outputs (response variables) without imposing strong distributional/parametric assumptions while using only a subset of training samples and sensor attributes. Then, an optimal cost-sensitive anomaly detection framework is proposed to determine whether a sample is an anomalous one taking into consideration the trade-off between misclassification errors and detection rates. The model can be used for both supervised and unsupervised settings depending on the availability of data regarding the behavior of the system under anomaly conditions. The unsupervised model is particularly useful when it is prohibitive to identify in advance the anomalies that a system may present and where no data are available regarding the behavior of the system under anomaly conditions. A Bayesian hierarchical setting is used to structure the proposed framework and help with accommodating uncertainty, imposing interpretability, and controlling the sparsity and complexity of the proposed anomaly detection framework. A Markov chain Monte Carlo algorithm is also developed for model training using past data. The numerical experiments conducted using a simulated data set and a wind turbine data set demonstrate the successful application of the proposed work for system response modeling and anomaly detection.

Wind energy is one of the largest sources of renewable energy in the world. To further reduce the operations and maintenance (O&M) costs of wind farms, it is essential to be able to accurately pinpoint the root causes of different failure modes of interest. An example of such a failure mode that is not yet fully understood is white etching cracks (WEC). This can cause the bearing lifetime to be reduced to 5–10% of its design value. Multiple hypotheses are available in literature concerning its cause. To be able to validate or disprove these hypotheses, it is essential to have historic high-frequency measurement data (e.g., load and vibration levels) available. In time, this will allow linking to the history of the turbine operating data with failure data. This paper discusses the dynamic loading on the turbine during certain events (e.g., emergency stops, run-ups, and during normal operating conditions). By combining the number of specific events that each turbine has seen with the severity of each event, it becomes possible to assess which turbines are most likely to show signs of damage.

Digitally enhanced services for wind power could reduce operations and maintenance costs as well as the levelized cost of energy. Therefore, there is a continuous need for advanced monitoring techniques, which can exploit the opportunities of internet of things and big data analytics, revolutionizing the future of the energy sector. The heart of wind turbines is a rather complex epicyclic gearbox. Among others, extremely critical gearbox components, which are often responsible for machinery stops, are the rolling element bearings. The vibration signatures of bearings are rather weak compared to other components, such as gears, and as a result, an extended number of signal processing techniques and tools have been proposed during the last decades, focusing toward accurate, early, and on time bearing fault detection with limited false alarms and missed detections. Envelope analysis is one of the most important methodologies, where an envelope of the vibration signal is estimated usually after filtering around a frequency band excited by impacts due to the bearing faults. Different tools, such as Kurtogram, have been proposed in order to accurately select the optimum filter parameters (center frequency and bandwidth). Cyclic spectral correlation (CSC) and cyclic spectral coherence (CSCoh), based on cyclostationary analysis, have been proved as very powerful tools for condition monitoring. The monitoring techniques seem to have reached a mature level in case a machinery operates under steady speed and load. On the other hand, in case the operating conditions change, it is still unclear whether the change of the monitoring indicators is due to the change of the health of the machinery or due to the change of the operating parameters. Recently, the authors have proposed a new tool called improved envelope spectrum via feature optimization-gram (IESFOgram), which is based on CSCoh and can automatically select the filtering band. Furthermore, the CSCoh is integrated along the selected frequency band leading to an improved envelope spectrum (IES). In this paper, the performance of the tool is evaluated and further extended on cases operating under different speeds and different loads. The effectiveness of the methodology is tested and validated on the National Renewable Energy Laboratory (NREL) wind turbine gearbox vibration condition monitoring benchmarking dataset, which includes various faults with different levels of diagnostic complexity as well as various speed and load operating conditions.

The localized failures of gears introduce cyclic-transient impulses in the measured gearbox vibration signals. These impulses are usually identified from the sidebands around gear-mesh harmonics through the spectral analysis of cyclo-stationary signals. However, in practice, several high-powered applications of gearboxes like wind turbines are intrinsically characterized by nonstationary processes that blur the measured vibration spectra of a gearbox and deteriorate the efficacy of spectral diagnostic methods. Although order-tracking techniques have been proposed to improve the performance of spectral diagnosis for nonstationary signals measured in such applications, the required hardware for the measurement of rotational speed of these machines is often unavailable in industrial settings. Moreover, existing tacho-less order-tracking approaches are usually limited by the high time-frequency resolution requirement, which is a prerequisite for the precise estimation of the instantaneous frequency. To address such issues, a novel fault-signature enhancement algorithm is proposed that can alleviate the spectral smearing without the need of rotational speed measurement. This proposed tacho-less diagnostic technique resamples the measured acceleration signal of the gearbox based on the optimal warping path evaluated from the fast dynamic time-warping algorithm, which aligns a filtered shaft rotational harmonic signal with respect to a reference signal assuming a constant shaft rotational speed estimated from the approximation of operational speed. The effectiveness of this method is validated using both simulated signals from a fixed-axis gear pair under nonstationary conditions and experimental measurements from a 750-kW planetary wind turbine gearbox on a dynamometer test rig. The results demonstrate that the proposed algorithm can identify fault information from typical gearbox vibration measurements carried out in a resource-constrained industrial environment.

Operation and maintenance costs are a major driver for levelized cost of energy of wind power plants and can be reduced through optimized operation and maintenance practices accomplishable by various prognostics and health management (PHM) technologies. In recent years, the wind industry has become more open to adopting PHM solutions, especially those focusing on diagnostics. However, prognostics activities are, in general, still at the research and development stage. On the other hand, the industry has a request to estimate a component’s remaining useful life (RUL) when it has faulted, and this is a key output of prognostics. Systematically presenting PHM technologies to the wind industry by highlighting the RUL prediction need potentially helps speed up its acceptance and provides more benefits from PHM to the industry. In this paper, we introduce a PHM for wind framework. It highlights specifics unique to wind turbines and features integration of data and physics domain information and models. The output of the framework focuses on RUL prediction. To demonstrate its application, a data domain method for wind turbine gearbox fault diagnostics is presented. It uses supervisory control and data acquisition system time series data, normalizes gearbox temperature measurements with reference to environmental temperature and turbine power, and leverages big data analytics and machine-learning techniques to make the model scalable and the diagnostics process automatic. Another physics-domain modeling method for RUL prediction of wind turbine gearbox high-speed-stage bearings failed by axial cracks is also discussed. Bearing axial cracking has been shown to be the prevalent wind turbine gearbox failure mode experienced in the field and is different from rolling contact fatigue, which is targeted during the bearing design stage. The method uses probability of failure as a component reliability assessment and RUL prediction metric, which can be expanded to other drivetrain components or failure modes. The presented PHM for wind framework is generic and applicable to both land-based and offshore wind turbines.

As global wind capacity continues to grow, the need for accurate operational analyses of a rapidly growing fleet of wind power plants has increased in proportion. The wind energy industry at present, however, is not ideally positioned to address this need. First, there is a lack of best practices and limited published standards for performing operational analyses. Second, operational data and methods are typically proprietary and not shared among the wind energy community. Consequently, there is considerable duplication of effort in developing methods as well as uncertainty in the calculated metrics. To address these problems, the National Renewable Energy Laboratory has publicly released OpenOA, an open-source code base for performing operational analyses on wind plant data. The intent of OpenOA is to provide a framework in which best practices can be developed, refined, and disseminated. Ultimately, such collaboration is expected to lead to a working example (i.e. reference implementation) of methods from which a published standard may develop. This article provides an overview of OpenOA, highlighting its release as a public repository, modular software architecture, current functionality, and planned functionality in subsequent releases. It is our goal for OpenOA to evolve into an indispensable tool for performing operational analyses that is used and supported by a large community of wind energy experts.

Wind farms are an indispensable driver toward renewable and nonpolluting energy resources. However, as ideal sites are limited, placement in remote and challenging locations results in higher logistics costs and lower average wind speeds. Therefore, it is critical to increase the reliability of the turbines to reduce maintenance costs. Robust implementation requires a thorough understanding of the loads subject to the turbine's control. Yet, such dynamically changing multidimensional loads are uncommon with other machinery, and generally underresearched. Therefore, a multitiered approach is proposed to investigate the load spectrum occurring in wind farms. Our approach relies on both fundamental research using controllable test rigs, as well as analyses of real-world loading conditions in high-frequency supervisory control and data acquisition data. A method is introduced to detect operational zones in wind farm data and link them with load distributions. Additionally, while focused research further investigates the load spectrum, a method is proposed that continuously optimizes the farm's control protocols without the need to fully understand the loads that occur. A case of gearbox failure is investigated based on a vast body of past experiments and suspect loads are identified. Starting from this evidence on the cause and effects of dynamic loads, the potential of our methods is shown by analyzing real-world farm loading conditions on a steady-state case of wake and developing a preventive row-based control protocol for a case of cascading emergency brakes induced by a storm., 24 pages, 8 figures

Despite the wind industry's dramatic development during the past decade, it is still challenged by premature turbine subsystem/component failures, especially for turbines rated above 1 MW. Because a crane is needed for each replacement, gearboxes have been a focal point for improvement in reliability and availability. Condition monitoring (CM) is a technique that can help improve these factors, leading to reduced turbine operation and maintenance costs and, subsequently, lower cost of energy for wind power. Although technical benefits of CM for the wind industry are normally recognized, there is a lack of published information on the advantages and limitations of each CM technique confirmed by objective data from full-scale tests. This article presents first-hand oil and wear debris analysis results obtained through tests that were based on full-scale wind turbine gearboxes rated at 750 kW. The tests were conducted at the 2.5-MW dynamometer test facility at the National Wind Technology Center at t...

This article describes an interdisciplinary methodology to calculate the probability of failure for bearing axial cracking, the dominant failure mode in the intermediate and high-speed stages of many wind turbine gearboxes. This approach is mainly a physics-domain method with needed inputs from the data domain. The gearbox and bearing design along with operations data and component failure records from a wind power plant provide the input to physics-based models and define axial cracking damage metrics. The physics-domain models predict the bearing loads and sliding velocities, which are the essential elements for quantifying the accumulated frictional energy. Both accumulated frictional energy and electrical energy generation are proposed as damage metrics for bearing axial cracking. A first-order reliability method is then used to compare the proposed damage metrics to failure threshold functions and calculate the probability of failure of each individual bearing. Although the probability of failure for the failed turbines is not separated from the population, a feature engineering analysis shows the potential of frictional energy as a damage metric when combined with roller loads, bearing sliding speed, lubricant type, and terrain features. Through statistical analysis of historical data, the proposed methodology enables reliability assessment of axial cracking in individual wind turbine bearings and connects the reliability forecast with turbine design and operations.

This paper presents a scalable, autonomous framework for diagnostics of wind turbine gearbox failures using a multi-feature supervisory control and data acquisition data set spanning multiple years. Because of the size of the data set studied, all algorithms were constructed to be scalable using a Spark-Hadoop data framework. An unsupervised approach was used to detect significant failures based on predetermined criteria. The initial criteria selected were an abnormal spike in turbine component temperature followed by a turbine power off, which helps reduce the number of potential false alarms. To detect abnormal spikes in component temperature, a model was introduced to adjust the temperature data for effects caused by ambient temperature and normal temperature increases when loaded. This study evaluates methods for normalizing temperature sensor data to identify anomalies. The models that performed the best were a linear regression and a multivariate polynomial regression. The proposed process for finding failures has tunable parameters that can be adjusted to be more or less sensitive. The combination of sensor data normalization and application of these criteria is successful in finding turbine failures resulting in downtime. The proposed methods can operate without failure or maintenance logs and can be utilized for offline analysis of large high-resolution data sets.

Wind turbine is a complex system composed of a variety of subsystems such as blade, gear box, drive train, generator, power converters, control unit etc., which brings challenge to do real-time monitoring, diagnosis and resilient control from a system level. Furthermore, wind farms are generally composed of tens and hundreds of wind turbines, which are even more challenging to do monitoring, prognosis and resilient control due to the nature of multi-agents and distribution. This special issue aims to provide a platform for academic and industrial communities to report recent results and emerging research direction in real-time monitoring, fault diagnosis, prognosis and resilient control and design for wind turbine systems.

The global wind industry has seen tremendous growth during the past two decades. However, the industry is challenged by premature component failures, which lead to increased turbine downtime and subsequently, cost of energy for wind power. To mitigate the impacts from these failures, the wind industry has been exploring various areas for improvements ranging from product design, new materials or lubricants, to operation and maintenance (O&M) practices. Condition-based maintenance or prognostics and health management (PHM) has been explored as one enabling technology for improving O&M practices. This chapter provides a brief overview of wind turbine PHM with a focus on operational data mining and condition monitoring of drivetrains. Some future research and development opportunities in wind turbine PHM are also briefly discussed.

The global wind industry has witnessed exciting developments in recent years. The future will be even brighter with further reductions in capital and operation and maintenance costs, which can be accomplished with improved turbine reliability, especially when turbines are installed offshore. One opportunity for the industry to improve wind turbine reliability is through the exploration of reliability engineering life data analysis based on readily available data or maintenance records collected at typical wind plants. If adopted and conducted appropriately, these analyses can quickly save operation and maintenance costs in a potentially impactful manner. This chapter discusses wind turbine reliability by highlighting the methodology of reliability engineering life data analysis. It first briefly discusses fundamentals for wind turbine reliability and the current industry status. Then, the reliability engineering method for life analysis, including data collection, model development, and forecasting, is presented in detail and illustrated through two case studies. The chapter concludes with some remarks on potential opportunities to improve wind turbine reliability. An owner and operator’s perspective is taken and mechanical components are used to exemplify the potential benefits of reliability engineering analysis to improve wind turbine reliability and availability.

Condition monitoring helps reduce the operations and maintenance costs by providing information about the physical condition of wind power systems. This study proposes to use a statistical method for effective condition monitoring. The turbine operation is significantly affected by external weather conditions. We model the wind turbine response as a function of weather variables, using a nonparametric regression method named least squares support vector regression. In practice, online condition monitoring of wind power systems often relies on datasets contaminated with outliers. This study proposes to use a weighted version of least squares support vector regression that provides a formal procedure for removing the outlier effects. We determine the decision boundaries to distinguish faulty conditions from normal conditions by examining the variations in the operational responses that are significantly affected by external weather. The results show that the proposed method effectively detects anomalies.

Improving the availability of wind turbines is critical for minimising the cost of wind energy, especially offshore. The development of reliable and cost-effective gearbox condition monitoring systems (CMSs) is of concern to the wind industry, because the gearbox downtime has a significant effect on the wind turbine availabilities. Timely detection and diagnosis of developing gear defects is essential for minimising an unplanned downtime. One of the main limitations of most current CMSs is the time consuming and costly manual handling of large amounts of monitoring data, therefore automated algorithms would be welcome. This study presents a fault detection algorithm for incorporation into a commercial CMS for automatic gear fault detection and diagnosis. Based on the experimental evidence from the Durham Condition Monitoring Test Rig, a gear condition indicator was proposed to evaluate the gear damage during non-stationary load and speed operating conditions. The performance of the proposed technique was then successfully tested on signals from a full-size wind turbine gearbox that had sustained gear damage, and had been studied in a National Renewable Energy Laboratory's (NREL) programme. The results show that the proposed technique proves efficient and reliable for detecting gear damage. Once implemented into the wind turbine CMSs, this algorithm can automate the data interpretation, thus reducing the quantity of the information that the wind turbine operators must handle.

Equally spaced planetary gearboxes are important power-train components for varied engineering systems. Their failures can result in significant capital losses and pose safety concerns. The vibration measurements perceived by a sensor mounted on the gearbox housing can provide valuable diagnostic information without normal gearbox operation interference. However, such vibration based monitoring techniques are difficult to implement in planetary gearboxes because of the complex nature of measured vibration spectra that is a result of planets revolving with respect to the stationary sensors mounted on the gearbox housing. Previous research with simulations and experiments using such measurements has reported distinct sideband patterns in the resulting vibration spectra, which differ significantly from the spectra of a normal fixed-axis/parallel gear pair system. In this paper, Fourier series analysis is used to explain these distinct sideband patterns that contain rich diagnostic information. The results obtained are useful to understand the cause of the observed vibration behavior in both healthy and faulty planetary gearboxes and identify the locations of additional frequency components introduced by the damaged gear in a complex measured vibration spectrum. Thus, the formulation presented in this paper can assist in developing robust feature extraction algorithms for early detection of planetary gearbox failures. The theoretical derivations presented in this paper are validated by both dynamic simulations and experiments on a dynamometer test bed using a 750 kW gearbox damaged during its operation while installed in a wind turbine. The predicted frequencies for observed faults in the annulus and sun gears of the gearbox are vividly presented in the experimentally measured frequency spectrum.

Operational managers of wind turbines usually monitor a big fleet of turbines and thus need highly condensed information to identify underperforming turbines and to prioritize their work. Key performance indicators (KPIs) are a solid and frequently used tool for this purpose. However, the KPIs used in the wind industry are not unified to date, which makes comparison in the industry difficult. Further, comprehensive standards on a set of KPIs for the wind industry are missing. This article identifies and recommends KPIs and provides detailed definitions to make KPIs comparable and to enable benchmarking. The starting point of this work is an industry survey with 28 participants intended to identify commonly used KPIs, collect various possible definitions, and prioritize them. Out of a total of 50 KPIs, we discuss in a next step 33 selected KPIs on performance, maintenance, and reliability in detail and recommend definitions, most of which are based on international standards. As a result, operators can easily use these recommendations to base their system of KPIs. By using this unified set of KPIs, operators can be well-prepared to conduct industrywide comparisons and benchmarks. The survey and this article will also serve as a basis for committee work of the FGW e.V. to develop a corresponding technical guideline.

Renewable electricity generation from wind turbines presents unique challenges in reliability of their mechanical systems compared to utility-scale fossil fuelled generation. The number of turbines and their distribution make maintenance a difficult and infrequent task, and the peak loading events (caused by unsteady winds and grid faults) and harsh environmental conditions (temperature fluctuation and moisture) result in extreme operation. Consequently, wind turbine systems present one of the more challenging current practical tribological problems. Contact failures in gear and bearing components have been the source of costly repairs and downtime of the turbine's drivetrain and actuators. Gearboxes are commonly faced with abrasive wear, micropitting, scuffing, and macropitting issues, while the nacelle yaw and blade pitch mechanisms commonly suffer from false brinelling and fretting damage. This paper examines contact failures of bearings from wind turbine gearboxes. Scanning electron microscopy is used to observe the subsurface material microstructure. Areas of localized microstructural change, known as white etching area (WEA), are observed. Several theories as to the cause of the WEA are examined. Results of this analysis will help guide the development of lasting solutions to a pressing problem in wind turbines.

This paper presents effective and accurate approaches in vibration-based wind turbine drivetrain component condition monitoring. Detailed spectral analysis and acceleration enveloping techniques were used to effectively extract the gear and bearing damage features. Synchronous analysis was used to accurately detect specific damage features during constantly varying operational conditions. A typical wind turbine gearbox amplifies shaft speed two orders of magnitude from the rotor to the generator. To account for all necessary vibration signatures, synchronous sampling must be carried out for multiple revolutions and at a relatively high rate. The synchronous sampling used in this paper was carried out in the digital domain after both the keyphasor and the vibration signals were digitized at high sampling rate and high sampling resolution analog-to-digital conversion. Sometimes, the shaft speed is provided in a speed time history format, as in the case of the National Renewable Energy Laboratory (NREL) Round Robin project. To carry out synchronous sampling using the speed time history, a unique synthesized synchronous sampling technique was adopted. The approach presented in this paper was realized using MATLAB (MathWorks, Natick, MA) codes and then validated with a wind turbine field case. It was also applied to the NREL wind turbine drivetrain condition monitoring Round Robin project. The identified damage results using the techniques discussed were compared with post-test inspection results with good correlation. Copyright © 2013 John Wiley & Sons, Ltd.

The research described was performed with diagnostic tools used to detect damage to dynamic mechanical components in a wind turbine gearbox. Different monitoring technologies were evaluated by collecting vibration and oil-debris data from tests performed on both a ‘healthy’ gearbox and a damaged gearbox that were mounted on a dynamometer test stand at the National Renewable Energy Laboratory (NREL). The damaged gearbox tested had been removed from the field after it experienced component damage because of two events that resulted in the loss of oil. The gearbox was re-tested under controlled conditions by using the NREL dynamometer test stand. Preliminary results indicate that oil-debris and vibration data can be integrated to improve the assessment of the health of the wind turbine gearbox. Copyright © 2012 John Wiley & Sons, Ltd.

The pre-construction energy generation of a wind farm (P50) is difficult to estimate and evaluate. This paper presents a methodology to measure the accuracy of the p50 prediction, which we call the Historical Validation Survey (HVS), for several wind farms in the continental United States. Our results indicate that there is a bias between predicted and measured energy, even when controlling for factors like grid curtailment and resource variability. We also find that our results depend on the assumptions we make during analysis, which we quantify with a sensitivity analysis. This method allows the estimation of uncertainty we have in our findings. When we account for reasonable ranges of model assumptions, we find that, in the most optimistic case, there is still a bulk −5.5% bias when estimating pre-construction energy generation. When controlling for grid curtailment this number reduces to a range of −3.5 to −4.5%.

Vibration-based condition monitoring (CM) of geared utility-scale turbine drivetrains has been used by the wind industry to help improve operation and maintenance (O&M) practices, increase turbine availability, and reduce O&M cost. This study is a new endeavor that integrates the vibration-based CM technique with wind turbine gearbox modeling to investigate various gearbox design options. A team of researchers performed vibration-based CM measurements on a damaged wind turbine gearbox with a classic configuration, (i.e., one planetary stage and two parallel stages). We observed that the acceleration amplitudes around the first-order sidebands of the intermediate stage gear set meshing frequency were much lower than that measured at the high-speed gear set, and similar difference was also observed in a healthy gearbox. One factor for a reduction at the intermediate stage gear set is hypothesized to be the soft sun-spline configuration in the test gearbox. To evaluate this hypothesis, a multibody dynamic model of the healthy test gearbox was first developed and validated. Relative percent difference of the first-order sidebands — of the high-speed and intermediate stage gear-meshing frequencies — in the soft and the rigid sun spline configurations were compared. The results verified that the soft sun-spline configuration can reduce the sidebands of the intermediate stage gear set and also the locating bearing loads. The study demonstrates that combining vibration-based CM with appropriate modeling can provide insights for evaluating different wind turbine gearbox design options.

This paper presents a systematic approach to determining optimal sensor locations with the goal to improve data-acquisition quality and the effectiveness in machine condition monitoring. The presented method ranks an initial set of candidate sensor locations based on their respective effective-independence (EfI) value, which is a measure for the contribution of each location to the measurement data matrix. Through an iterative procedure, locations having relatively low EfI values are progressively eliminated from the candidate set. The remaining locations with high EfI values have shown to be more effective in providing sensing coverage of the machine component being monitored. The method was applied to selecting locations for four accelerometers to monitor a rolling element bearing with a localized defect. Experimental tests have confirmed the effectiveness of the method in improving sensing quality

This paper presents a systematic investigation of the effect of sensor location on the data quality and subsequently, on the effectiveness of machine health monitoring. Based on an analysis of the signal propagation process from the defect location to the sensor, numerical simulations using finite element modeling were conducted on a bearing test bed to determine the signal strength at several representative sensor locations. The results showed that placing sensors closely to the machine component being monitored is critical to achieving high signal-to-noise ratio, thus improving the data quality. Using millimeter-sized piezoceramic plates, the obtained results were evaluated experimentally. A comparison with a set of commercial vibration sensors verified the developed structural dynamics-based sensor placement strategy. It further demonstrated that the proposed shock wave-based sensing technique provided an effective alternative to vibration measurement, while requiring less space for sensor installation.

Modern mathematics has commonly been utilized as an effective tool to model mechanical equipment so that their dynamic characteristics can be studied analytically. This will help identify potential failures of mechanical equipment by observing change in the equipment’s dynamic parameters. On the other hand, dynamic signals are also important and provide reliable information about the equipment’s working status. Modern mathematics has also provided us with a systematic way to design and implement various signal processing methods, which are used to analyze these dynamic signals, and to enhance intrinsic signal components that are directly related to machine failures. This special issue is aimed at stimulating not only new insights on mathematical methods for modeling but also recently developed signal processing methods, such as sparse decomposition with potential applications in machine fault diagnosis. Finally, the papers included in this special issue provide a glimpse into some of the research and applications in the field of machine fault diagnosis through applications of the modern mathematical methods.

The Gearbox Reliability Collaborative (GRC) at the National Wind Technology Center (NWTC) tested two identical gearboxes. One was tested on the NWTCs 2.5 MW dynamometer and the other was field tested in a turbine in a nearby wind plant. In the field, the test gearbox experienced two oil loss events that resulted in damage to its internal bearings and gears. Since the damage was not severe, the test gearbox was removed from the field and retested in the NWTCs dynamometer before it was disassembled. During the dynamometer retest, some vibration data along with testing condition information were collected. These data enabled NREL to launch a Wind Turbine Gearbox Condition Monitoring Round Robin project, as described in this report. The main objective of this project was to evaluate different vibration analysis algorithms used in wind turbine condition monitoring (CM) and find out whether the typical practices are effective. With involvement of both academic researchers and industrial partners, the project sets an example on providing cutting edge research results back to industry.

Tribology is the science and engineering of interacting surfaces in relative motion. It includes the study and application of the principles of friction, lubrication, and wear. It is an important phenomenon that not only impacts the design and operation of wind turbine gearboxes, but also their subsequent maintenance requirements and overall reliability. With the major growth and increasing dependency on renewable energy, mechanical reliability is an extremely important issue. The Wind Turbine Tribology Seminar was convened to explore the state-of-the-art in wind turbine tribology and lubricant technologies, raise industry awareness of a very complex topic, present the science behind each technology, and identify possible R&D areas. To understand the background of work that had already been accomplished, and to consolidate some level of collective understanding of tribology by acknowledged experts, the National Renewable Energy Laboratory (NREL), Argonne National Laboratory (ANL), and the U.S. Department of Energy (DOE) hosted a wind turbine tribology seminar. It was held at the Renaissance Boulder Flatiron Hotel in Broomfield, Colorado on November 15-17, 2011. This report is a summary of the content and conclusions. The presentations given at the meeting can be downloaded. Interested readers who were not at the meeting may wish to consultmore » the detailed publications listed in the bibliography section, obtain the cited articles in the public domain, or contact the authors directly.« less

This report will present the wind turbine drivetrain condition monitoring (CM) research conducted under the phase 1 and phase 2 Gearbox Reliability Collaborative (GRC) tests. The rationale and approach for this drivetrain CM research, investigated CM systems, test configuration and results, and a discussion on challenges in wind turbine drivetrain CM and future research and development areas, will be presented.

This report describes the Gearbox Reliability Collaborative's work at NREL to increase the reliability of wind turbine gearboxes.

High operations and maintenance costs for wind turbines reduce their overall cost effectiveness. One of the biggest drivers of maintenance cost is unscheduled maintenance due to unexpected failures. Continuous monitoring of wind turbine health using automated failure detection algorithms can improve turbine reliability and reduce maintenance costs by detecting failures before they reach a catastrophic stage and by eliminating unnecessary scheduled maintenance. A SCADA (Supervisory Control and Data Acquisition System) -data based condition monitoring system uses data already collected at the wind turbine controller. It is a cost-effective way to monitor wind turbines for early warning of failures and performance issues. In this paper, we describe our exploration of existing wind turbine SCADA data for development of fault detection and diagnostic techniques for wind turbines. We used a number of measurements to develop anomaly detection algorithms and investigated classification techniques using clustering algorithms and principal components analysis for capturing fault signatures. Anomalous signatures due to a reported gearbox failure are identified from a set of original measurements including rotor speeds and produced power.

Prediction of a bearing service life is traditionally achieved by empirical or physical models, which have their own strengths and limitations. In an effort to combine the strengths of these modeling approaches, this research investigates the concept of Multi-Time Scale Modeling (MTSM). Specifically, a MTSM strategy for bearing life prognosis is developed by correlating experimentally acquired bearing vibration data with physics based model of microscopic growth of crack size. The strategy is composed of a fast scale empirical model (e.g., root mean square value of vibration), a slow scale physical model (e.g., change of crack length over one loading cycle), and a model coupling mechanism (e.g., bidirectional mapping functions). The fast and slow scale models are obtained by polynomial regression analysis and using the concept of the Paris Law, respectively. The coupling mechanism is established through introduction of dynamic mass into the model. The improvement in bearing service life prediction, obtained by the presented MTSM strategy is experimentally validated.

This paper presents a structural dynamics-based sensor placement strategy for improving the quality of data acquisition and consequently, the effectiveness of machine health monitoring. Based on vibration measurements on a realistic bearing test bed, signal quality at several representative sensor locations was compared using the structural nodal signal-to-noise ratio as a criterion. It was found that placing sensors closely to the machine component to be monitored has significant effect on improving sensing quality. Using one custom-designed and four commercial accelerometers, results from the numerical analysis were experimentally verified.

The effectiveness and reliability of measurement techniques for bearing condition monitoring are affected by both the locations of the sensors and the signal processing algorithms selected for defect feature extraction. This chapter describes a structural dynamics-based sensor placement strategy by investigating the mechanisms of signal propagation from the source of its generation to the sensor location. Numerical simulation of a group of sensors for measuring vibration measurement on two custom-designed bearing test beds is presented, and an approach to optimizing the sensor placement based on the Effective Independence (EfI) method is introduced. The chapter then comparatively investigates several commonly employed signal processing techniques for feature extraction, such as wavelet transform-based signal enveloping, the Wigner-Ville distribution, and the wavelet packet transform, and evaluates performance using vibration signals measured from the bearing test beds.

This paper presents a systematic investigation of the effect of sensor placement on the measurement data quality and subsequently, on the effectiveness of machine system health monitoring. First, signal propagation process from the defect location to the sensor was analyzed. Numerical simulations using finite element modeling was then conducted to determine the signal strength at several representative sensor locations. The analytical and numerical results showed that placing sensors closer to the component being monitored resulted in higher signal-to-noise ratio, thus improving the data quality. Using meso-sized piezoceramic chips, the obtained results were then experimentally evaluated. Comparisons with a set of commercial vibration sensors verified the developed sensor placement strategy. The presented study quantitatively justified why placing sensors closer to the machine being monitored is of critical importance to ensuring high quality data, and confirmed that the customized shock wave sensing approach can achieve comparable result for vibration measurement, but with a much less space requirement.

This paper investigates the architectural effect on adaptive neuro-fuzzy inference system (ANFIS) for machine condition assessment. The study was motivated by ANFIS’s limitation in adapting its architecture to map the modeled input output relationship. Based on the grid input space partition method, two elements in defining an ANFIS architecture were studied: the type of the membership function (MF) and the MF number assigned to ANFIS inputs. A new modeling accuracy index was introduced to address the limitation of the traditional root mean square error (RMSE) in describing the effect of the MF type. The analysis showed that wide core membership functions enabled a smaller RMSE than narrow core membership functions for machine defect severity classification. It is further shown that selecting appropriate MF number is critical to ensuring accuracy of ANFIS, considering the overfitting problem. These results were experimentally investigated on a bearing test bed, where defect severity classification and dynamic load estimation were evaluated. The experiments agreed well with the theoretical analysis.Copyright © 2004 by ASME

This paper presents a fuzzy data fusion scheme for bearing defect severity classification. Instead of using time-domain features (e.g. peak value, root mean square, kurtosis) of the original signal as the inputs, wavelet transform was applied to the original signals as a data pre-processor. Defect-related features were filtered out from the corresponding wavelet coefficients. Next, fuzzifications on individual input features were performed to obtain individual decisions. Finally, data fusion was peformed to achieve an overall bearing defect severity classification. Three data fusion methods: voting, weighted decision, and fuzzy inference, were investigated. Experimental results have shown that the presented fuzzy data fusion scheme is more effective than traditional approaches using raw signals as the input features. In particular, the presented scheme has resolved conflicting decisions arising from single feature-based classification, thus improving the reliability of defect severity assessment.Copyright © 2003 by ASME

To date the existing wind turbine condition monitoring technologies and commercially available systems have not been fully accepted for improving wind turbine availability and reducing their operation and maintenance costs. One of the main reasons is that wind turbines are subject to constantly varying loads and operate at variable rotational speeds. As a consequence, the influences of turbine faults and the effects of varying load and speed are coupled together in wind turbine condition monitoring signals. So, there is an urgent need to either introduce some operational condition de-coupling procedures into the current wind turbine condition monitoring techniques or develop a new operational condition independent wind turbine condition monitoring technique to maintain high turbine availability and achieve the expected economic benefits from wind. The purpose of this paper is to develop such a technique. In the paper, three operational condition independent criteria are developed dedicated for monitoring the operation and health condition of wind turbine generators. All proposed criteria have been tested through both simulated and practical experiments. The experiments have shown that these criteria provide a solution for detecting both mechanical and electrical faults occurring in wind turbine generators.

Advances in high-performance sensing technologies enable the development of wind turbine condition monitoring systems to diagnose and predict the system-wide effects of failure events. This paper presents a vibration-based two-stage fault detection framework for failure diagnosis of rotating components in wind turbines. The proposed framework integrates an analytical defect detection method with a graphical verification method to ensure diagnosis efficiency and accuracy. The efficacy of the proposed methodology is demonstrated with a case study using the gearbox condition monitoring Round Robin study dataset provided by the National Renewable Energy Laboratory (NREL). The developed methodology successfully detected five faults out of a total of seven with accurate severity levels and without producing any false alarm in the blind analysis. The case study results indicate that the developed fault detection framework is effective for analyzing gear and bearing faults in wind turbine drivetrain systems based on system vibration characteristics.