Bearings faults and limits in wind turbine generators
The detection of sudden faults in wind turbine generator (WTG) is a complex task, especially in bearings. Usually, the evaluation of methodologies such as vibration, ultrasound, and bearing temperatures are widely used in predictive maintenance, an important aspect for the traditional approach, in w...
Main Author: | |
---|---|
Format: | Article |
Language: | English |
Published: |
Elsevier
2024-03-01
|
Series: | Results in Engineering |
Subjects: | |
Online Access: | http://www.sciencedirect.com/science/article/pii/S2590123024001440 |
_version_ | 1797247684355031040 |
---|---|
author | Ricardo Manuel Arias Velásquez |
author_facet | Ricardo Manuel Arias Velásquez |
author_sort | Ricardo Manuel Arias Velásquez |
collection | DOAJ |
description | The detection of sudden faults in wind turbine generator (WTG) is a complex task, especially in bearings. Usually, the evaluation of methodologies such as vibration, ultrasound, and bearing temperatures are widely used in predictive maintenance, an important aspect for the traditional approach, in wind turbine fault detection, is the limited analysis with a single variable as vibration, or temperature. For instance, these sensors detect 5–20% of torsional vibration in the drivetrain and 55% has a failure due to lubricant problem, 20% for solid contamination or 9% for the incorrect application of the bearing. Consequently, to solve this limitation and failures modes, this research evaluated the limits and focused on the early detection of bearing faults in wind generators; it utilized a multi-stage approach involving Random Forest, XGBoost, Light XGB, and Logistic Regression, followed by probability scores and optimal features with a search grid validation, in addition, validated results through finite element modeling, Boroscopy, and vibration analysis. Hence, the database considers faults for bearings, gearboxes, and for normal operation; with the vibration analysis regarding 8,711,808 samples used for validating process. The result of the study is the detection of sudden failures five days before vibration analysis with high classification accuracy of 99.994%, recall of 99.982%, F1 score of 98.124%, kappa 99.330%, and a test set time of 22.82 s. This new approach provides an early detection of sudden failures compared to traditional vibration analysis in bearings and gearboxes. |
first_indexed | 2024-03-08T02:00:54Z |
format | Article |
id | doaj.art-780eedf2caf141b3a0a44954adef5bbf |
institution | Directory Open Access Journal |
issn | 2590-1230 |
language | English |
last_indexed | 2024-04-24T20:02:36Z |
publishDate | 2024-03-01 |
publisher | Elsevier |
record_format | Article |
series | Results in Engineering |
spelling | doaj.art-780eedf2caf141b3a0a44954adef5bbf2024-03-24T07:01:04ZengElsevierResults in Engineering2590-12302024-03-0121101891Bearings faults and limits in wind turbine generatorsRicardo Manuel Arias Velásquez0Corresponding author. Universidad Tecnológica del Perú, Lima, Perú.; Universidad Tecnológica Del Perú, PeruThe detection of sudden faults in wind turbine generator (WTG) is a complex task, especially in bearings. Usually, the evaluation of methodologies such as vibration, ultrasound, and bearing temperatures are widely used in predictive maintenance, an important aspect for the traditional approach, in wind turbine fault detection, is the limited analysis with a single variable as vibration, or temperature. For instance, these sensors detect 5–20% of torsional vibration in the drivetrain and 55% has a failure due to lubricant problem, 20% for solid contamination or 9% for the incorrect application of the bearing. Consequently, to solve this limitation and failures modes, this research evaluated the limits and focused on the early detection of bearing faults in wind generators; it utilized a multi-stage approach involving Random Forest, XGBoost, Light XGB, and Logistic Regression, followed by probability scores and optimal features with a search grid validation, in addition, validated results through finite element modeling, Boroscopy, and vibration analysis. Hence, the database considers faults for bearings, gearboxes, and for normal operation; with the vibration analysis regarding 8,711,808 samples used for validating process. The result of the study is the detection of sudden failures five days before vibration analysis with high classification accuracy of 99.994%, recall of 99.982%, F1 score of 98.124%, kappa 99.330%, and a test set time of 22.82 s. This new approach provides an early detection of sudden failures compared to traditional vibration analysis in bearings and gearboxes.http://www.sciencedirect.com/science/article/pii/S2590123024001440BearingTemperatureVibrationWind turbine |
spellingShingle | Ricardo Manuel Arias Velásquez Bearings faults and limits in wind turbine generators Results in Engineering Bearing Temperature Vibration Wind turbine |
title | Bearings faults and limits in wind turbine generators |
title_full | Bearings faults and limits in wind turbine generators |
title_fullStr | Bearings faults and limits in wind turbine generators |
title_full_unstemmed | Bearings faults and limits in wind turbine generators |
title_short | Bearings faults and limits in wind turbine generators |
title_sort | bearings faults and limits in wind turbine generators |
topic | Bearing Temperature Vibration Wind turbine |
url | http://www.sciencedirect.com/science/article/pii/S2590123024001440 |
work_keys_str_mv | AT ricardomanuelariasvelasquez bearingsfaultsandlimitsinwindturbinegenerators |