A deep boosted transfer learning method for wind turbine gearbox fault detection

Jamil, Faras; Verstraeten, Timothy; Nowé, Ann; Peeters, Cédric; Helsen, Jan

doi:10.1016/j.renene.2022.07.117

Cited by 42 publications

(17 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Both types can be found in the state-of-the-art literature. Deep neural networks are used in (Black et al, 2022), (Jamil et al, 2022) (transfer learning), (Mazidi et al, 2017), (Verma et al, 2022), (Meyer, 2021) (multi-target neural network), (Turnbull et al, 2021), (Cui et al, 2018), (Sun et al, 2016), (Bangalore and Tjernberg, 2015) (NARX), (Bangalore and Tjernberg, 2014), (Li et al, 2014), (Kusiak and Verma, 2012), (Kusiak and Li, 2011), (Zaher et al, 2009). Another popular type of model is the Autoencoder (AE).…”

Section: Normal Behavior Modeling Algorithmsmentioning

confidence: 99%

Overview of normal behavior modeling approaches for SCADA-based wind turbine condition monitoring demonstrated ondata from operational wind farms

Chesterman

Verstraeten

Daems

et al. 2023

Preprint

View full text Add to dashboard Cite

Abstract. Condition monitoring and failure prediction for wind turbines is currently a hot research topic. This follows from the fact that investments in the wind energy sector have increased dramatically due to the transition to renewable energy production. This paper reviews and implements several techniques from state-of-the-art research on condition monitoring for wind turbines using SCADA data and the Normal Behavior Modelling framework. The first part of the paper consists of an in-depth overview of 5 the current state-of-the-art. In the second part, several techniques from the overview are implemented and compared using data (SCADA and failure data) from five operational wind farms. To this end, 6 demonstration experiments are designed. The first 5 experiments test different techniques for the modeling of the normal behavior. The sixth experiment compares several techniques that can be used for identifying anomalous patterns in the prediction error. The paper concludes with several directions for future work.

show abstract

Section: Normal Behavior Modeling Algorithmsmentioning

confidence: 99%

Overview of normal behavior modeling approaches for SCADA-based wind turbine condition monitoring demonstrated ondata from operational wind farms

Chesterman

Verstraeten

Daems

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…DL methods have become popular among researchers in the field of fault detection. However, their performance depends on the availability of big data sets [10]. To overcome this problem researchers started applying TL to achieve good performance from small available datasets, by leveraging multiple prediction models over similar machines and working conditions.…”

Section: Introductionmentioning

confidence: 99%

“…As shown in the study, TL enables machine learning models to transfer learned knowledge from source domains to a target domain to improve the performance of the target learning function. In contrast, the source and target domains have different data distributions [10]. Moreover, the source domain data samples can be transferred to improve the learning of the target model [19].…”

Section: Introductionmentioning

confidence: 99%

“…In contrast with the negative transfer, the valuable information from the source is positive, as addressed e.g. in [10]. Combining DL with TL to train nonlinear high-dimensional DTL models with a small data-set is also possible.…”

Section: Introductionmentioning

confidence: 99%

“…TL is also classified into four categories: instance-based DTL, network-based DTL, mapping-based DTL, and adversarial-based DTL. In order to address the negative transfer problem, the work [10] proposed, e.g., an instance-based DTL method.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Transfer Learning for Fault Detection with Application to Wind Turbine SCADA Data

Simani¹,

Farsoni²,

Castaldi³

2023

J Energ Power Technol

View full text Add to dashboard Cite

The installed wind power capacity is growing worldwide. Remote condition monitoring of wind turbines is employed to achieve higher uptimes and lower maintenance costs. Machine learning models can detect developing damages in wind turbines. Therefore, this paper demonstrates that cross–turbine transfer learning can drastically improve the accuracy of fault detection models in turbines with scarce SCADA data. In particular, it shows that combining the knowledge from turbines with scarce and turbines with plentiful data enables earlier detection of faults than prior art methods. Training fault detection models require large amounts of past and present SCADA data but these data are often unavailable or not representative of the current operation behavior. Newly commissioned wind farms lack SCADA data from the previous operation. Due to control software updates or hardware replacements, older turbines may also lack representative SCADA data. After such events, a turbine’s operation behavior can change significantly so its SCADA data no longer represent its current behavior. Therefore, the work highlights how to reuse and transfer knowledge across wind turbines to overcome this lack of data and enable the earlier detection of faults in wind turbines.

show abstract

Fleet-wide analytics on field data targeting condition and lifetime aspects of wind turbine drivetrains

Daems

Peeters

Matthys

et al. 2023

Forsch Ingenieurwes

Self Cite

View full text Add to dashboard Cite

Drivetrain failures result in the largest downtime per failure among the different turbine components. To minimize O&M costs, it is therefore essential to be able to anticipate failure events sufficiently in advance such that scheduled maintenance can take place. Moreover, a root cause for the failure should be identified, allowing to incorporate this knowledge in future design iterations, thereby increasing the reliability of the machine. In offshore wind energy, high-frequency SCADA (1 Hz) and vibration data (>20 kHz) are becoming increasingly available to monitor machine performance and health. This paper presents a twofold approach for monitoring drivetrain health and load history based on these two data sources. First, SCADA data is used to extract different design load cases (DLCs), described in IEC 61400-3. Second, vibration data is used for advanced signal analysis to detect potential incipient bearing or gear defects in the drivetrain. It is shown that the efficacy of this vibration analysis is further enhanced by combining it with operating condition information from the SCADA data. Flottenweite Analyse von Felddaten zur Untersuchung von Zustands-und Lebensdaueraspekten von Antriebssträngen in Windenergieanlagen ZusammenfassungAusfälle des Antriebsstrangs verursachen von allen Turbinenkomponenten die größte Ausfallzeit pro Ausfall. Um die Betriebs-und Wartungskosten zu minimieren, ist es daher von entscheidender Bedeutung, dass Ausfallereignisse so frühzeitig erkannt werden, dass eine planmäßige Wartung durchgeführt werden kann. Darüber hinaus sollte die Fehlerursache identifiziert werden, um dieses Wissen in zukünftige Konstruktionsiterationen einfließen zu lassen und so die Zuverlässigkeit der Anlage zu erhöhen. In der Offshore-Windenergie werden zunehmend hochfrequente SCADA-(1 Hz) und Schwingungsdaten (>20 kHz) zur Überwachung der Maschinenleistung und des Maschinenzustands verfügbar. In diesem Beitrag wird ein zweifacher Ansatz zur Überwachung des Zustands des Antriebsstrangs und des Lastverlaufs auf der Grundlage dieser beiden Datenquellen vorgestellt. Erstens werden SCADA-Daten verwendet, um verschiedene Auslegungslastfälle (DLCs) zu extrahieren, die in IEC 61400-3 beschrieben sind. Zweitens werden die Schwingungsdaten für eine erweiterte Signalanalyse verwendet, um mögliche beginnende Lager-oder Getriebeschäden im Antriebsstrang zu erkennen. Es wird gezeigt, dass die Wirksamkeit dieser Schwingungsanalyse durch die Kombination mit Betriebszustandsinformationen aus den SCADA-Daten weiter verbessert wird.

show abstract

A deep boosted transfer learning method for wind turbine gearbox fault detection

Cited by 42 publications

References 41 publications

Overview of normal behavior modeling approaches for SCADA-based wind turbine condition monitoring demonstrated ondata from operational wind farms

Overview of normal behavior modeling approaches for SCADA-based wind turbine condition monitoring demonstrated ondata from operational wind farms

Transfer Learning for Fault Detection with Application to Wind Turbine SCADA Data

Fleet-wide analytics on field data targeting condition and lifetime aspects of wind turbine drivetrains

Contact Info

Product

Resources

About