Abstract: To address the problem that uniform or random microphone layouts in compressive-sensing-based duct acoustic mode identification may lead to an ill-conditioned transfer matrix—thereby degrading inversion stability and identification accuracy—two microphone placement optimization methods for compressor casing-wall measurements are proposed: Effective Independence (EfI) and Singular Value Monitoring (SVM). The former iteratively removes redundant measurement points based on their contribution to the linear independence of the transfer matrix, while the latter mitigates matrix ill-conditioning by monitoring the singular value distribution and condition number. These optimized layouts are further integrated with Bayesian compressive sensing to identify dominant azimuthal acoustic modes in ducts. Simulation results show that both methods significantly improve the numerical properties of the transfer matrix and maintain high identification accuracy even when the number of microphones is reduced by approximately 50%. Experimental results from compressor tests further demonstrate that, using only 16 microphones, the optimized layouts reduce the dominant-mode identification error by more than 1.5 dB compared with random layouts; among them, the SVM-based method exhibits superior robustness under broadband excitation scenarios.