Frequency Focus Sparse Bayesian Learning for Nonuniform Undersampled Blade Vibration Signal Parameter Identification

Rotating blades play a critical role as essential components of aeroengines, and monitoring their operational status is of utmost importance. The blade tip timing (BTT) technique has garnered significant attention due to its noninvasive nature, high reliability, all-blade measurement, and precision. However, the inherent characteristics and working properties of aeroengine blades result in highly undersampled data from BTT probes. Consequently, blade vibration parameter identification has become a significant challenge for the BTT technique. In this article, the frequency focus sparse Bayesian learning with fast marginal likelihood maximization approach is proposed to address this issue. First and foremost, the measured signals are expanded in the frequency domain using a redundant dictionary, and the frequency focus technique is employed to enhance the accuracy of frequency identification while reducing sparse parameters and enhancing sparsity. Second, fast marginal likelihood maximization is harnessed to improve convergence speed while preserving identification accuracy. Finally, the comprehensive simulation and experimental results demonstrate the accuracy and robustness of the proposed method in identifying frequency, magnitude, and phase precisely.