基于稀疏贝叶斯的航空发动机风扇声模态重构

To address the large number of sensors required by uniform circular arrays in duct acoustic mode reconstruction for aero-engines，and the amplitude underestimation problem of traditional L1-norm-based compressed sensing methods， a sparse Bayesian approach for fan noise modal reconstruction was proposed. A hierarchical sparse Bayesian prior model was established and solved using a block coordinate descent algorithm， effectively characterizing and quantifying uncertainties in the measurement process. Furthermore，a non-dominated sorting genetic algorithm was employed to optimize array configuration and enhance reconstruction accuracy. Fan noise modal tests were conducted on a 3.5-stage aero-engine. Results showed that， under the same number of microphones， the sparse Bayesian method achieved lower reconstruction error than the L1-norm regularization method. Under low-speed condition，with an optimized layout of 6 sensors，the reconstruction error for circumferential mode order 5 was 0.01 dB. Under high-speed condition，with 8 optimally placed sensors，the reconstruction errors for mode orders 5 and −12 were 0.50 dB and 0.46 dB，respectively. The study demonstrated that the sparse Bayesian method significantly improved the accuracy of duct acoustic mode reconstruction with fewer sensors.