Detection of Jacket Offshore Wind Turbine Structural Damage using an 1D-Convolutional Neural Network with a Support Vector Machine Layer. Issue 3 (1st May 2022)

• Record Type: Journal Article
• Authors: Tutivén, Christian; Moreno, Sueanny; Vidal, Yolanda; Benalcázar, Carlos
• Is Part Of: Journal of physics. Volume 2265:Issue 3(2022)
• Publication Date: 2022-05-01
• Subjects: Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/2265/3/032088
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: Because offshore wind turbines, particularly their foundations, operate in hostile environments, implementing a structural health monitoring system is one of the best ways to monitor their condition, schedule maintenance, and predict possible fatal failures at lower costs. A novel strategy for detecting damage in offshore wind turbine jacket foundations is developed in this work, based on a vibration monitoring methodology that reshapes the data into a multichannel array, with as many channels as correlated sensors with the predicted variable, a 1-D deep convolutional neural network to extract temporal features from the monitored data, and a support vector machine as a final classification layer. The obtained model allows the detection of three types of bar states: healthy bar, cracked bar, and bar with an unlocked bolt.