Impact fault detection for marine current turbines blade via MEGK-means and PCA under variable marine conditions

Abstract

Abstract The advancement of the marine current turbine (MCT) technology has the potential to aid China in achieving its carbon peak and carbon neutrality goals. The performance of MCTs is influenced by fluctuating water velocities and erratic turbulence. The blades of MCTs are susceptible to cracking due to prolonged exposure to seawater. Unpredictable changes in marine currents contribute to unstable working environments. This study presents multiple envelope geometrical K-means to categorize stator current readings and develop fault detection models. It first constructs an envelope geometric feature matrix by using the extracted modulus signal. Then, the matrix is used to select the initial center for clustering, and a multi-scale principal component analysis is performed under each working condition to decrease data dimensionality. Finally, T 2 and squared prediction error (SPE) serve as the projection of the sample vector on the principal element space and the remaining subspace, respectively, and can be utilized to monitor fault cases. The experimental findings demonstrate that the proposed method has excellent recognition capabilities and detection accuracy for the impact faults of MCTs under variable working conditions.