Powered by OpenAIRE graph
Found an issue? Give us feedback
addClaim

Wind turbine blade inspection

Authors: Conceição, Pedro Santos da;

Wind turbine blade inspection

Abstract

Devido a preocupações climatéricas, energia renovável é uma importante fonte energética. Energia eólica desempenha um importante papel na produção de energias renováveis e é produzida por geradores eólicos. Estes estão sujeitos a danos e requerem inspeção regular. Esta tarefa é realizada manualmente e é um procedimento demorado, perigoso e dispendioso. Por estas razões, uma abordagem diferente e automática é necessária. Nos recentes anos, a área da Inteligência Artificial tem crescido muito e já desempenha um papel importante no nosso dia-a-dia. Deep Learning, um dos ramos da Inteligência Artificial, usa dados de forma a aprender padrões nos quais se baseia para tomar decisões. É já usada em muitas aplicações provando ser uma alternativa a ter em conta para desempenhar diferentes tarefas. Neste trabalho, propomos um modelo capaz de detectar áreas com dano nas pás de geradores eólicos baseado em Deep Learning. O dataset disponível é composto por oitenta imagens de p as de geradores eólicos tiradas por especialistas. As áreas com dano nestas imagens foram anotadas por um especialista. Testamos duas arquitecturas diferentes para segmentação de objectos, para segmentar as p as nas imagens, e duas arquitecturas diferentes para deteção de objectos, para colocar uma bounding box na área com dano. Quanto à segmentação, U-Net e DeepLabv3 produziram resultados promissores para serem usados em aplicações no mundo real. No entanto, a segmentação mostrou não ser totalmente correcta com algumas áreas das pás a não serem correctamente segmentadas sendo necessária mais pesquisa futura. Quanto á deteção, RetinaNet teve um desempenho superior que a Faster R-CNN. No entanto, é recomendável continuar o desenvolvimento do modelo de modo a aumentar a eficácia do mesmo.

Due to climate concerns, renewable energy is an important energy resource. Wind energy plays a big role in renewable energy production and is produced by wind turbines. These are prone to damage and require inspection at all time. This task is done manually and it is a time-consuming, dangerous and expensive procedure. Because of this reasons, a di erent and automatic approach is needed. In recent years, Arti cial Intelligence has improved a lot and already takes an important role in our daily lives. Deep Learning, one of Arti cial Intelligence's elds, uses data to learn patterns for use in decision making. It is already used in many applications and has proven to be reliable in di erent tasks. In this work we propose a model capable of detecting damaged areas in wind turbine blades based in Deep Learning. The available dataset was made by eighty wind turbine blade images taken by experts. The damaged areas in these images were annotated by an expert. We try two di erent architectures for object segmentation, to segment the blades in the images, and two di erent architectures for object detection, to place a bounding box around the damaged area. Regarding object segmentation, U-Net or DeepLabv3 produced promising results to be used in a real-world application. Although, the segmentation revealed to be faulty with several areas of the blades not being correctly segmented and further research is recommended. Regarding object detection, RetinaNet performed better than Faster R-CNN. However, it is recommended to continue the research in order to increase the accuracy of the model.

Mestrado em Engenharia de Automação Industrial

Country
Portugal
Related Organizations
  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Upload OA version
Are you the author of this publication? Upload your Open Access version to Zenodo!
It’s fast and easy, just two clicks!