On the subspace selection and frequency identification of rotor blade vibration signal measured by blade tip timing

Anti-aliasing spectrum analysis is essential for rotor blade condition monitoring based on Blade Tip Timing (BTT). The Multiple Signal Classification (MUSIC) algorithm, which exploits the orthogonality between signal and noise subspaces, has been successfully applied for this purpose. However, conventional subspace selection methods relying on fixed thresholds are sensitive to variations in large eigenvalues. Furthermore, the complex disturbances during rotor operation and measurement complicate the identification of blade vibration characteristics. To overcome these challenges, this paper proposes Adaptive Subspace Separation (ASS) and Local Spectral Centroid (LSC) methods to improve the adaptability of subspace selection and the stability of frequency identification, respectively. The impacts of overestimating and underestimating the subspace dimensions on MUSIC's performance are derived mathematically. Simulation and experiments demonstrate the effectiveness of proposed approaches: ASS offers more accurate and stable subspace dimension selection and tracking, while LSC reduces the standard deviation of estimated frequencies by 30 percent.