Acoustic inspection system with unmanned aerial vehicles for wind turbines structure health monitoring

Wind energy is considered as one of the most important renewable energies in the world, employing larger and more complex wind turbines. They need novel condition monitoring systems to ensure the reliability, availability, safety and maintainability of the main components of the wind turbines. It le...

Descripción completa

Detalles Bibliográficos
Autores: García Márquez, Fausto Pedro, Bernalte Sánchez, Pedro José, Segovia Ramírez, Isaac
Tipo de recurso: artículo
Fecha de publicación:2021
País:España
Institución:Universidad de Castilla-La Mancha
Repositorio:RUIdeRA. Repositorio Institucional de la UCLM
OAI Identifier:oai:ruidera.uclm.es:10578/28538
Acceso en línea:https://doi.org/10.1177/14759217211004822
http://hdl.handle.net/10578/28538
Access Level:acceso abierto
Palabra clave:Wind turbine
maintenance management
acoustic analysis
unmanned aerial vehicle
structural health monitoring
Descripción
Sumario:Wind energy is considered as one of the most important renewable energies in the world, employing larger and more complex wind turbines. They need novel condition monitoring systems to ensure the reliability, availability, safety and maintainability of the main components of the wind turbines. It leads to early fault detection, increasing the productivity and minimizing the maintenance costs and downtimes. This article proposes a novel non-destructive testing system to analyse acoustically rotatory devices of wind turbines. It captures the noise emitted by the devices using an acoustic condition monitoring system embedded in an unmanned aerial vehicle. The signal acquired is sent to ground computer station for recording and analysing the data. It uses a test rig, previously validated, to carry out a set of experiments to simulate the main faults. A signal processing method is done by wavelet transforms that filters and analyses the energy patterns of the signals. The results are analysed qualitatively and quantitatively considering different scenarios. A statistical analysis is developed to compare the numerical results provided by different wavelet transform families and convolutional neural network. It is concluded that Symlets and Daubechies families report equivalent results for this case study. The accuracies of the results are more than 75%, reaching up to 100%. The approach is validated employing Friedman test.