Method and apparatus for diagnosing a mechanism

• US 20070260656A1
• Filed: 09/27/2006
• Published: 11/08/2007
• Est. Priority Date: 05/05/2006
• Status: Active Grant

• Current Assignee: Airbus Helicopters
• Original Assignee: Eurocopter
• Inventors: Wiig, Johan
• Granted Patent: US 7,430,483 B2
• Time in Patent Office: Days
• US Class Current: 708/200
• CPC Class Codes: G01H 1/003 of rotating machines G01H1/10 takes precedence
• CPC Class Codes: G01M 13/028 Acoustic or vibration analysis
• CPC Class Codes: G05B 23/0221 Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods

Abstract

The present invention relates to a method of analyzing variations in an indicator (i) of the behavior of a mechanism, the method comprising the following operations:

• • (OP1, OP2) extracting noise (w) and any discontinuities (b, s) from the indicator (i) in order to obtain a trend (c);
  • (OP8) comparing variations in the trend (c) with one or more reference values; and
  • (OP10) indicating a fault of the mechanism as a function of the result of the comparison(s) (OP8).

Claims

1. A method of analyzing variations in an indicator (i) of the behavior of a mechanism, the method being characterized in that it comprises the following operations:

(OP1, OP2) extracting noise (w) and any discontinuities (b, s) from the indicator (i) in order to obtain a trend (c);

(OP8) comparing variations in the trend (c) with one or more reference values; and

(OP10) indicating a fault of the mechanism as a function of the result of the comparison(s) (OP8).

2. A method according to claim 1, in which, in order to extract any discontinuities, the following operations are performed:

(OP1) searching for and subtracting out any outlier data (s) in the indicator data (i); and

(OP2) extracting any edge data (b) from the data (i-s) obtained from operation (OP1).

3. A method according to claim 2, in which, in order to find the times of the outlier data (s):

a moving median value (mm(i)) of the indicator data is subtracted from the indicator data (i) in order to obtain noisy data (i′

); and

the time(s) is/are identified for which the ratio of the amplitude of the noisy data (i′

) to the indicator data level (wrms(i)) is greater than a threshold (ts).

4. A method according to claim 2, in which, in order to identify the presence of an edge (b) in the data (i-s), variations in the scale coefficients (d′

) corresponding to a wavelet transform of variations in the data (i-s) are analyzed.

5. A method according to claim 4, in which a potential edge (b) is identified in the data (i-s) when the ratio b_p of a mean (abs(mean(d′

))) of the sale coefficients to the standard deviation (std(d′

)) of the variations of these coefficients (d′

), exceeds a first determined value t_p, and when the ratio b_m of said mean (abs(mean(d′

))) of the scale coefficients to the level (wrms(i-s)) of the data (i₁=i−

s) exceeds a second determined value t_m.

6. A method according to claim 1, in which, the following operations are also performed:

(OP3) calculating the level (wrms(w)) of the noi.se (w);

(OP7) comparing variations in the level (wrms(w)) of the noise with one or more reference values; and

(OP9) indicating a fault of the mechanism as a function of the results of the comparison(s) performed in operation (OP7).

7. A method according to claim 6, in which, in order to compare variations in the noise level (wrms) with reference values, the following operations are performed:

(OP4) calculating scale variation data of a wavelet transform of the noise level variations (wrms(w)); and

(OP7) comparing the scale variation data of the wavelet transform of the noise level (wrms(w)) with the reference values.

8. A method according to claim 4, in which use is made of the Haar wavelet and of the scale function associated with the Haar wavelet.

9. A method according to claim 2, in which, when an edge is detected, the time of the edge is compared with historical data concerning maintenance of the mechanism.

10. A method according to claim 1, in which the amplitude of the data (i-s) is compared (OP9) with at least one third reference value.

11. A method according to claim 5, in which a fault of the mechanism is diagnosed as a function of the results of the comparisons.

12. A method according to claim 1, further including the following successive prior operations:

normalizing variations of the indicator with reference to simultaneous variations of an operating parameter (PA) of the mechanism; and

then, in a training stage, determining reference values or thresholds for the rate of variation of at least one component (w, b, s, c) of the indicator (i) from the variations of the indicator during a period of operation without any fault of the mechanism.

13. A method according to claim 12, in which the normalization operation includes parametric modeling of the indicator or of measurements from the indicator is extracted.

14. A method according to claim 12, in which the normalization operation comprises filtering indication variation data.

15. A method according to claim 1, in which the mechanism is a rotorcraft transmission mechanism, the indicator is derived from vibration measurements performed on the rotorcraft in flight, and the operating parameter (ias) is the air speed of the rotorcraft.

16. Apparatus for diagnosing a mechanism as a function of data (i) generated from vibration measurements performed on the mechanism, the apparatus being characterized in that it is programmed to perform the operations of a method according to claim 1.

17. A program for diagnosing a mechanism as a function of data (i) derived from vibration measurements performed on the mechanism, the program being characterized in that it is arranged (structured and organized) to perform the operations of a method according to claim 1.

18. A program according to claim 17, arranged to perform the following operations:

(OP1) searching for and separating out any outlier data (s) in the data (i);

(OP2) extracting noise data (w) and any edge data (b) from the data (i-s) obtained from operation (OP1), in order to obtain trend data (c);

(OP3) calculating noise level data (wrms(w)) from the noise data;

(OP4) calculating data concerning the rate of variation of the noise level data over time;

(OP5) calculating data (d3 to d8) concerning the rate of variation of the trend data over time;

(OP6) generating edge detection data, and where applicable calculating an edge time from the edge data, and comparing the edge time with historical data concerning mechanism maintenance;

(OP7) comparing the rate of variation data for the noise level data over time with first reference values;

(OP8) comparing the rate of variation data for the trend data over time with second reference values;

(OP9) comparing the amplitude of the data (i-s) obtained at the end of operation (OP1) with at least one third reference value; and

(OP10) generating data indicative of a fault being detected in the mechanism as a function of the results of the comparisons (OP6, OP7, OP8, OP9).

19. A program according to claim 17, in which the data (i) generated from the vibration measurements taken on the mechanism during a period of use of the mechanism corresponds to variations:

in the acceleration (energy) level measured on the mechanism at a synchronization frequency (i.e. a frequency of rotation of an element of the mechanism), or a multiple of said frequency;

or

in the acceleration level measured on the mechanism in a frequency band;

or

in an indicator of the degree of flattening of a spectrum of the vibratory measurements, such as a kurtosis.

20. A method according to claim 1, in which, in order to compare the trend variations with reference values, the following operations are performed:

(OP5) calculating scale variation data of a wavelet transform of the trend data (c); and

(OP8) comparing the scale variation data of the wavelet transform of the trend data with reference values.