L_1/2-norm Regularization for Detecting Aero-engine Fan Acoustic Mode

Compressive sensing provides an effective approach to detect the azimuthal acoustic modes of an aero-engine fan, with fewer microphones required than the conventional method. The paper proposes a L1/2-norm regularization based compressive sensing method to recognize the tonal acoustic modes, with an improvement of detection accuracy and significant robustness to the background noise interference on different conditions. Specifically, the iterative half thresholding algorithm with a K-sparsity strategy is introduced to solve the non-convex L1/2-norm regularized problem conveniently and efficiently. Meanwhile, the regularization parameter updates adaptively during the calculation to avoid the tuning work. A further acoustic test is conducted on a 3.5-stage aero-engine fan, where the effectiveness of the proposed method is validated by two cases where the blade-tip speed is subsonic and supersonic, respectively. Experimental results demonstrate that the proposed approach outperforms the classical L1-norm regularization under both operating conditions, enhancing accuracy and reducing microphone number.