Ponente
Descripción
This work presents the design, construction, and experimental characterization of a scale offshore wind turbine mounted on a jacket-type support structure, aimed at the detection of structural faults using machine learning techniques. The structure was fabricated using A36 structural steel and features a modular, disassemblable design that allows for the controlled introduction of structural faults, facilitating experimental validation and fault simulation under realistic conditions. The assembly includes key components of a typical wind turbine system: the foundation, tower, nacelle, and rotor.
To enable continuous structural health monitoring, triaxial inertial sensors (MPU6050) were integrated with ESP32 microcontrollers, forming a wireless data acquisition system using the ESP-NOW communication protocol. A comprehensive dataset was collected under four defined structural conditions: healthy state, loose bolts, corrosion, and crossbeam damage. These conditions were evaluated across four amplitude levels to simulate varying wind intensities, resulting in a total of 8,850 data samples recorded at a sampling rate of 100 Hz.
The acquired data was preprocessed through normalization and analyzed using dimensionality reduction techniques, specifically Principal Component Analysis (PCA). The PCA results demonstrated clear patterns that differentiate between structural states, underscoring its effectiveness for future classification tasks. This study lays the groundwork for the development of intelligent structural monitoring systems, with potential applications in predictive maintenance strategies for offshore wind energy infrastructure.
Keywords: Offshore wind turbine, jacket-type structure, structural health monitoring, machine learning, PCA, ESP32, wireless sensors, predictive maintenance.
