Method and apparatus for monitoring and analyzing vibrations in rotary machines

• US 9,874,472 B2
• Filed: 02/17/2010
• Issued: 01/23/2018
• Est. Priority Date: 02/18/2009
• Status: Active Grant

• Current Assignee: Rolls-Royce PLC (Rolls-Royce Holdings Plc)
• Original Assignee: Rolls-Royce PLC (Rolls-Royce Holdings Plc)
• Inventors: Tarassenko, Lionel, Clifton, David A, King, Dennis, King, Steven P, Ault, David J
• Primary Examiner(s): Toatley, Gregory J
• Assistant Examiner(s): STIFTER JR, TERENCE E
• Application Number:
  US13/201,082
• Publication Number:
  US 20120035885A1
• Time in Patent Office: 2,897 Days
• Field of Search: 702183
• CPC Class Codes: G01H 1/006 of the rotor of turbo machines
• CPC Class Codes: G01H 3/08 Analysing frequencies present in complex vibrations, e.g. comparing harmonics present acoustic presence detection G01V1/001
• CPC Class Codes: G01M 13/028 Acoustic or vibration analysis
• CPC Class Codes: G01M 13/045 Acoustic or vibration analysis
• CPC Class Codes: G01M 15/12 by monitoring vibrations
• CPC Class Codes: G01N 2291/2693 Rotor or turbine parts
• CPC Class Codes: G01N 29/46 by spectral analysis, e.g. Fourier analysis or wavelet analysis spectral signal processing per se G06F17/14

Abstract

Vibration amplitudes are recorded as a function of rotation speed and of frequency and the data is analyzed to estimate a noise floor amplitude threshold for each of a plurality of different speed and frequency sub-ranges. On the basis of training data known to be normal speed-frequency areas which contain significant spectral content in normal operation are deemed “known significant spectral content”, so that during monitoring of new data points which correspond to significant vibration energy at speeds and frequencies different from the known significant spectral content can be deemed “novel significant spectral content” and form the basis for an alert. The estimation of the noise floor is based on a probabilistic analysis of the data in each speed-frequency area and from this analysis an extreme value distribution expressing the probability that any given sample is noise is obtained.

Claims

1. A method of monitoring vibration amplitude and frequency in a rotary machine, comprising:

receiving measurements of vibration amplitude and rotation speed of the machine from at least one sensor;

recording as a data point each amplitude measurement as a function of vibration frequency and rotational speed;

storing a plurality of different amplitude thresholds each representing a noise floor for a respective one of a first set of sub-ranges of vibration frequency and rotational speed,

wherein each sub-range includes a sub-range of speeds and a sub-range of frequencies,

wherein the different amplitude thresholds are set in a training process using training data points comprising vibration data from a run of the rotary machine deemed to be normal, and

wherein the different amplitude thresholds are set by fitting an amplitude distribution to the training data points in each sub-range of vibration frequency and rotational speed;

comparing each amplitude measurement to the amplitude threshold for a corresponding sub-range of vibration frequency and rotational speed for the data points within which the amplitude measurement falls;

determining the amplitude measurement to represent noise if it falls below the amplitude threshold and as significant spectral content if it is above the amplitude threshold;

using the amplitude measurement determined to be significant spectral content, assessing abnormalities of at least a component of the rotary machine; and

when an abnormality is assessed, generating a signal configured to indicate the abnormality of at least a component of the rotary machine.

2. A method according to claim 1 wherein the different amplitude thresholds each representing a noise floor for a respective one of the first set of sub-ranges of vibration frequency and rotational speed are further set by

calculating an extreme value distribution for said amplitude distribution and

setting as said amplitude threshold an amplitude value representing a preset probability of being a maximum amplitude value that would be obtained for noise.

3. A method according to claim 2 wherein the preset probability is a probability above 0.99.

4. A method according to claim 2 wherein the amplitude distribution is a Gamma distribution.

5. A method according to claim 2 wherein the amplitude distribution is fitted after excluding training data points not representing background noise from the data in each sub-range of vibration frequency and rotational speed.

6. A method according to claim 2 further comprising:

comparing each amplitude measurement to the extreme value distribution obtained from the training data points to read-off a probability value for that amplitude measurement;

calculating from the probability value a novelty value as;

novelty value=−

log₁₀{1−

P_e(x)}, where P_e(x) is the extreme value distribution as a function of amplitude x; and

outputting the novelty value.

7. A method according to claim 1 further comprising:

among a second set of sub-ranges of vibration frequency and rotational speed, determining which ones of the sub-ranges of the second set contained, during a training process using vibration data from a run of the rotary machine deemed to be normal, more than a preset number of training data points whose amplitude measurement was above the noise floor, and defining these ones of the sub-ranges of the second set of sub-ranges of vibration frequency and rotational speed as containing known significant spectral content; and

for each amplitude measurement determined to represent significant spectral content, determining the amplitude measurement to represent known significant spectral content when the amplitude measurement is within a sub-range of the second set of sub-ranges of vibration frequency and rotational speed that contains known significant spectral content and to represent novel significant spectral content when the amplitude measurement is within a sub-range of the second set of sub-ranges that does not contain known significant spectral content.

8. A method according to claim 7 wherein the first set and the second set of sub-ranges of vibration frequency and rotational speed are the same.

9. A method according to claim 7 wherein an alarm condition is triggered if a threshold number of amplitude measurements representing novel significant spectral content are found or if amplitude measurements representing novel significant spectral content persist for a predetermined time.

10. A method according to claim 1 wherein the frequency range is divided into 30 to 120 sub-ranges and the rotational speed range is divided into 5 to 40 sub-ranges.

11. A method according to claim 1 wherein the frequency range is divided into 50 to 100 sub-ranges and the rotational speed range is divided into 10 to 30 sub-ranges.

12. Apparatus for monitoring vibration amplitude and frequency in a rotary machine, comprising inputs for receiving vibration measurements from a vibration sensor and measurements of the rotation speed of the machine from a tachometer, and a data processing system adapted to execute the method of claim 1.

13. A control system for a rotary machine comprising apparatus according to claim 12.

14. A method of monitoring vibration amplitude and frequency in a rotary machine, comprising the steps of:

receiving measurements of the vibration amplitude and the rotation speed of the machine from a run of the rotary machine from at least one sensor;

recording as a data point each amplitude measurement as a function of vibration frequency and rotational speed;

comparing each amplitude measurement to an amplitude threshold for a sub-range of vibration frequency and rotational speed for the data points within which the amplitude measurement falls,

wherein the amplitude threshold represents a noise floor for the sub-range of vibration frequency and rotational speed,

wherein the sub-range includes a sub-range of speeds and a sub-range of frequencies,

wherein the amplitude threshold is set in a training process using training data points comprising vibration data from a run of the rotary machine deemed to be normal, and

wherein the amplitude threshold is set by fitting an amplitude distribution to the training data points in the sub-range of vibration frequency and rotational speed;

determining a set of sub-ranges in which the amplitude measurement is above the amplitude threshold;

defining sub-ranges from the set of sub-ranges as known significant spectral content if more than a preset number of data points have an amplitude measurement above the amplitude threshold;

defining sub-ranges from the set of sub-ranges as novel significant spectral content if there is less than a preset number of data points having an amplitude measurement above the amplitude threshold;

using the sub-ranges defined to be novel significant spectral content, assessing abnormalities of at least a component of the rotary machine; and

when an abnormality is assessed, generating a signal configured to indicate the abnormality of at least a component of the rotary machine.