Performance analysis for mode decomposition methods in an axial fan

Research output: Contribution to journalArticlepeer-review

Abstract

To analyze the performance of mode decomposition methods in rotating machinery such as turbofan engines, this paper begins with a series of tests using the experimental database of the TA36 single-stage compressor on the classic circumferential mode decomposition method (CM) and the compressive sensing method (CSM). While CM is widely used in industry, its uncertainties under limited microphone conditions have not been systematically evaluated, and CSM has rarely been validated in realistic fan experiments. This study aims to address these gaps. In addition, we focus on a realistic compressor with non-uniform inlet conditions, where the resulting circumferential non-uniformity of the acoustic field can significantly affect mode decomposition results. It is found that the uncertainties in CM results are relatively large, even when the system of equations is overdetermined. The solutions exhibit good convergence as the number of microphones increases. The error bands of the TS mode sound pressure level (SPL) are reduced to ± 0.5 dB for CM 20 , which is then regarded as a reference. The analyses show the dependence of the decomposition results on the mode dominance ratio (MDR), which can be preliminarily estimated from the circumferential distribution of microphone SPL in the tests. Based on this, we investigated a series of factors that may influence the performance of CSM, including the number of microphones, signal-to-noise ratio (SNR), MDR, the absolute angle of array arrangement, and the masking of low-reliability signals.

Original languageEnglish
Article number111147
JournalApplied Acoustics
Volume243
DOIs
StatePublished - 5 Feb 2026

Keywords

  • Circumferential mode decomposition
  • Compressive sensing method
  • Compressor noise
  • Mode dominance ratio (MDR)
  • Signal-to-noise ratio (SNR)
  • Turbofan engines

Fingerprint

Dive into the research topics of 'Performance analysis for mode decomposition methods in an axial fan'. Together they form a unique fingerprint.

Cite this