Publicação
Automatic defect detection in wind turbine blades: A Deep Learning Model Pipeline for Detection and Classification of Defects in Drone Images
| Resumo: | The decarbonisation of the energy industry is key in the global approach to the climate emergency. Wind energy in particular, plays an important role in transitioning the global energy system to more sustainable sources. To do so, it must optimise O&M costs through a data-driven, predictive maintenance approach. When it comes to the maintenance of a wind turbine, the manual inspection of blade photographs - often taken by drones - is a time-consuming and labour intensive process, yet it is fundamental to ensure the turbine’s continued energy production during its lifecycle. This thesis intends to purpose, test, and implement a defect detection pipeline based on the Faster-RCNN deep learning architecture, capable of detecting 6 distinct classes of blade defects to a level of accuracy suitable enough to deploy in A.I.-assisted inspections. |
|---|---|
| Autores principais: | Neves, Lara Souto das |
| Assunto: | Machine Learning Wind Turbines Turbine Blades Deep Learning Faster-RCNN TensorFlow Object Detection API Automatic Defect Detection Damage Detection Artificial Intelligence Optimising Wind O&M SDG 7 - Affordable and clean energy |
| Ano: | 2022 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | The decarbonisation of the energy industry is key in the global approach to the climate emergency. Wind energy in particular, plays an important role in transitioning the global energy system to more sustainable sources. To do so, it must optimise O&M costs through a data-driven, predictive maintenance approach. When it comes to the maintenance of a wind turbine, the manual inspection of blade photographs - often taken by drones - is a time-consuming and labour intensive process, yet it is fundamental to ensure the turbine’s continued energy production during its lifecycle. This thesis intends to purpose, test, and implement a defect detection pipeline based on the Faster-RCNN deep learning architecture, capable of detecting 6 distinct classes of blade defects to a level of accuracy suitable enough to deploy in A.I.-assisted inspections. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.