Method, system and computer program for the acoustic analysis of a machine

Patent metadata

Field	Value
Publication number	US-10495546-B2
Application number	US-201415028284-A
Country	US
Kind code	B2
Filing date	Oct 9, 2014
Priority date	Oct 11, 2013
Publication date	Dec 3, 2019
Grant date	Dec 3, 2019

Abstract

A method for the acoustic analysis of a machine (M) including the acquisition of at least one acoustic signal supplied by at least one microphone ( 7 ) positioned in the machine, characterized in that it comprises the following steps: separation of at least one acoustic signal into a plurality of sound sources, the signal being modelled as a mixture of components, each one corresponding to a sound source; for at least one separated sound source, determination of a characteristic acoustic signature; comparison of at least one characteristic acoustic signature with at least one reference acoustic signature recorded in a reference database ( 5 ).

First claim

What is claimed is:

1. A method of monitoring an aircraft engine comprising:
acquiring at least one acoustic signal supplied by at least one microphone positioned in the aircraft engine;
performing non-blind source separation on the at least one acquired acoustic signal so as to separate the at least one acquired acoustic signal into a plurality of sound sources, said at least one acquired acoustic signal being modelled as a mixture of components, each one corresponding to a sound source;
for at least one of the separated sound sources, determining a characteristic acoustic signature;
comparing the at least one characteristic acoustic signature with at least one reference acoustic signature recorded in a reference database; and
detecting a defect of the aircraft engine based on a result of the comparison,

wherein performing the non-blind source separation on the at least one acquired acoustic signal comprises:
determining at least one residual signal by filtering the at least one acquired acoustic signal to remove frequencies corresponding to known resonances of components of the engine, and
performing source separation on the at least one residual signal based on a modelling of the at least one residual signal as a mixture of independent sound sources, thereby estimating each one of the independent sound sources from the at least one residual signal.

2. The method according to claim 1 , wherein each reference acoustic signature corresponds to an acoustic signature characteristic of a defect of the aircraft engine, and the defect of the aircraft engine is detected when a difference between a characteristic acoustic signature of a separated sound source and the reference acoustic signature characteristic of said defect is below a threshold.

3. The method according to claim 1 , further comprises determining at least one reference acoustic signature which comprises:
an acquisition of at least one reference acoustic signal supplied by at least one microphone positioned in at least one reference aircraft engine;
a separation of at least one reference acoustic signal into a plurality of reference sound sources;
for at least one of the separated reference sound sources, a determination of a reference acoustic signature;
a recording in the reference database of the reference acoustic signature of at least one separated reference sound source.

4. The method according to claim 3 , wherein at least one reference aircraft engine is a defect-free aircraft engine, and the defect of the aircraft engine is detected when a difference between a characteristic acoustic signature of a separated sound source and an acoustic signature recorded in the reference database is above a threshold.

5. The method according to claim 4 , wherein at least one reference aircraft engine includes at least two aircraft engines, said method including a calculation of differences between a fingerprint vector of the aircraft engine under acoustic analysis composed of at least one characteristic acoustic signature of at least one separated sound source and fingerprint vectors each constituted of at least one reference acoustic signature recorded in the reference database corresponding to one of the reference aircraft engines.

6. The method according to claim 3 , wherein at least one reference aircraft engine is the same aircraft engine as the aircraft engine under acoustic analysis considered earlier or an aircraft engine of the same type having the same operating history, and the defect of the aircraft engine is detected when a difference between a fingerprint vector of the aircraft engine composed of at least one characteristic acoustic signature of at least one separated sound source and a fingerprint vector composed of at least one reference acoustic signature recorded in the reference database is above a threshold.

7. The method according to claim 1 , wherein the determination of a characteristic acoustic signature of a separated sound source includes a plotting of a spectrogram of said separated sound source, an identification of intensity peaks by thresholding of the spectrogram, a calculation of distances between said intensity peaks.

8. The method according to claim 1 , wherein the non-blind source separation of the at least one acquired acoustic signal into a plurality of sound sources is carried out by a computer processor configured to implement an independent component analysis.

9. A non-transitory computer-readable medium having stored thereon a program product including code instructions for the execution of the steps of the method according to claim 1 , when said program is run on a computer.

10. The method according to claim 1 , wherein the non-blind source separation comprises filtering, from the at least one acoustic signal, previously known resonances of components of the aircraft engine.

11. The method according to claim 1 , further comprising, prior to performing non-blind source separation on the at least one acoustic signal, a step of filtering the at least one acquired acoustic signal with a bank of Gabor filters.

12. A system for the acoustic analysis of an aircraft engine, including means for acquiring at least one acoustic signal supplied by at least one microphone positioned in the aircraft engine, and a reference database in which is recorded at least one reference acoustic signature, the system further comprising:
a module for performing non-blind source separation on the at least one acoustic signal so as to separate the at least one acoustic signal into a plurality of sound sources, said at least one acoustic signal being modelled as a mixture of components each one corresponding to a sound source;
a module for determining an acoustic signature configured to determine at least one characteristic acoustic signature of at least one of the separated sound source;
a module for comparing acoustic signatures configured to compare at least one characteristic acoustic signature with at least one reference acoustic signature recorded in the reference database; and
a module for detecting a defect of the aircraft engine based on a comparison result provided by the module for comparing,

wherein performing the non-blind source separation on the at least one acoustic signal comprises:
determining at least one residual signal by filtering the at least one acoustic signal to remove frequencies corresponding to known resonances of components of the engine, and
performing source separation on the at least one residual signal based on a modelling of the at least one residual signal as a mixture of independent sound sources, thereby estimating each one of the independent sound sources from the at least one residual signal.

Assignees

• Snecma
• Safran Aircraft Engines

Inventors

• Bense William
• Lacaille Jérôme
• Gerez Valerio

Classifications

• G01N29/14

  using acoustic emission techniques {(echo of particles G01N29/046; measuring mechanical vibrations or acoustic waves in solids in general G01H1/00)} · CPC title

• G01M15/14 Primary

  Testing gas-turbine engines or jet-propulsion engines · CPC title

• G01N29/449

  Statistical methods not provided for in G01N29/4409, e.g. averaging, smoothing and interpolation · CPC title

• G01N29/46

  by spectral analysis, e.g. Fourier analysis {or wavelet analysis (spectral signal processing per se G06F17/14)} · CPC title

• G01N29/4436

  with a reference signal (amplitude comparison G01N29/48) · CPC title