An ESPRIT-Based Dual-Sensor Method for Monitoring Blade Vibration Frequencies

Blade tip timing (BTT) technology allows for prolonged and comprehensive monitoring of all blades. However, BTT signal is undersampled signal, Fourier transform is not enough for the extraction of blades vibration frequency. In engineering applications, the space constraints of aircraft engines make it difficult to install multiple sensors. Therefore, achieving blade frequency monitoring with a reduced number of sensors is critically important. This article introduces an estimation of signal parameters via rotational invariance technique (ESPRIT)-based dual-sensor method for monitoring blade vibration frequencies. The method utilizes the rotational invariance property of the subspace formed by two-channel data to determine the signal vibration frequency and it overcomes the challenge of frequency identification with limited sensors and obtains the time-frequency result of blade vibration using only two sensors. Besides, it solves the problem of the traditional ESPRIT algorithm being usable only at constant speed. A dynamic model of rotating blades is established to calculate the synchronous responses of the blades. The BTT numerical simulator samples the undersampled signals of the blade under different excitation forces. The analysis of the undersampled signals proves the effectiveness of the method. Additionally, experiments involving a five-blade disk conducted to validate the method, and the time-frequency results of the analysis results are compared with those of the strain gauge, and the frequency error is less than 5%. The proposed method is compared with multiple signal classification (MUSIC) and OMP methods by processing the same 400-s data segment. The OMP takes 264 s, MUSIC takes 168 s, while the improved ESPRIT algorithm takes only 56 s. This method effectively addresses the challenge of blade frequency identification with limited sensors due to space constraints and enables real-time online monitoring of blade vibrations.