Method for monitoring rotating machine and associated apparatus and system

Abstract

An apparatus (4) for monitoring a rotating machine (2) comprising a plurality of rotating parts (2a, 2b, 2c) is proposed. The device comprises first determination means (5), selection means (8), second determination means (6), third determination means (7) and actuation means (10).

Description

Technical Field

[0001] The present invention relates to the monitoring of rotating machines.

[0002] More particularly, the present invention relates to methods and apparatus for monitoring rotating machines, and systems including such apparatus. Background Art

[0003] Typically, in order to detect faults in a rotating machine, for example to diagnose a bearing fault or an imbalance, a vibration analysis of the machine is performed.

[0004] Vibration analysis involves the comparison of lines in the estimated spectrum of the vibration signal with a typical large number of reference frequencies predicted by a machine vibration model (MVM) or an environmental model (EM).

[0005] Vibration signals are usually transmitted by sensors on rotating machines.

[0006] MVM comes from understanding the physical structure of the machine and its axis speed, and EM comes from understanding the environment surrounding the machine.

[0007] The physical structure of the rotating machine includes, for example, the characteristics of the shaft couplings, the type of bearings, the number of blades in the impeller, the number and type of gear teeth, and the number of gear teeth.

[0008] When the frequency of a line in the estimated spectrum coincides with the fault frequency predicted by the MVM, the corresponding component is considered to be faulty.

[0009] If the MVM is inaccurate, such as generating an incorrect reference frequency, the actual fault may be overlooked when the corresponding vibration frequency deviates from the frequency predicted by the MVM model, which may lead to machine failure.

[0010] Alternatively, mistaking benign vibration frequencies for fault frequencies can lead to unnecessary maintenance and production downtime.

[0011] One cause of inaccurate MVM is when a machine part is replaced with a different part without updating the specifications of the rotating machine, for example during maintenance.

[0012] Another reason may be that when commissioning a new vibration-based monitoring system on a rotating machine, the rotating machine specification data is often incomplete, with unknown gear tooth counts and / or bearing types.

[0013] Furthermore, building an MVM based on specifications or inspection of the rotating machine's physical structure is a manual and error-prone process.

[0014] Manual inspection of the spectrum may be performed to detect anomalies and errors in the MVM or EM.

[0015] This manual inspection is resource-intensive and time-consuming, and the results depend heavily on the operator’s awareness, knowledge of the machine, and attention to detail. Summary of the invention

[0016] Therefore, the present invention proposes to improve the reliability of vibration analysis of rotating machines by improving the robustness of MVM or EM by enhancing the detection of anomalies and errors in MVM or EM.

[0017] According to an aspect, a method for monitoring a rotating machine comprising a plurality of rotating components is proposed, wherein a first model comprises a first set of prediction spectra and a second model comprises a second set of prediction spectra.

[0018] The method includes:

[0019] - measuring the vibration of rotating machines to deliver the measured vibration signal,

[0020] - determine the frequency spectrum of the measured vibration signal, which includes multiple spectral lines,

[0021] - Selecting a spectral line from a plurality of spectral lines according to a predetermined selection rule, wherein the selected spectral line is a significant spectral line,

[0022] - determining significant spectral lines not included in the first set of predicted spectral lines and the second set of predicted spectral lines, the determined significant spectral lines are unidentified spectral lines,

[0023] - determine the waveform power of each unidentified spectral line,

[0024] - compare the determined power of each unidentified spectral line to a power threshold, and

[0025] - If the determined power of at least one unidentified spectral line is equal to or greater than the power threshold, an alarm is triggered.

[0026] The method allows automatic detection of modeling errors of the first and second models to avoid catastrophic failures or unnecessary maintenance and production stops due to misclassification of spectral lines.

[0027] Automatic detection runs during normal operation, which allows machine uptime to be maintained.

[0028] The first model includes a machine vibration model (MVM) and the second model includes an environmental model (EM) of the rotating machine.

[0029] Preferably, determining the frequency spectrum of the measured vibration signal comprises performing a non-parametric frequency spectrum analysis method or a parametric frequency spectrum analysis method.

[0030] Advantageously, the predetermined selection rule comprises a model order selection rule.

[0031] Preferably, determining the significant spectral lines not included in the first set of predicted spectral lines and the second set of predicted spectral lines comprises:

[0032] - for each significant spectral line and each rotating component and each periodic external disturbance, determining a likelihood function from a first model of the frequencies generated by said rotating component in rotation and from a second model of the frequencies of a second set of predicted spectral lines,

[0033] - for each likelihood function, evaluating said likelihood function at the frequency of each significant spectral line to obtain the likelihood probability at the frequency generated by said rotating component in rotation and at the frequency of the periodic external disturbance,

[0034] - for each likelihood probability, compare said likelihood probability to the classification threshold,

[0035] - When the likelihood probability is less than a classification threshold, the significant spectral line is determined to be not included in the first group of predicted spectral lines and the second group of predicted spectral lines.

[0036] Advantageously, when the likelihood probability is equal to or greater than a classification threshold, the significant spectral line is considered to be included in the first set of predicted spectral lines or the second set of predicted spectral lines.

[0037] If it is determined that one or more of the significant spectral lines are included neither in the first set of predicted spectral lines nor in the second set of predicted spectral lines, the first and second models are considered unreliable.

[0038] In contrast, if all of the significant spectral lines are determined to be included in the first set of predicted spectral lines or the second set of predicted spectral lines, the first and second models are considered to be reliable.

[0039] Preferably, when for a rotating component, the a priori probability that the frequency generated by the rotating component during rotation is included in the measured vibration signal is known, and the a priori probability that the frequency of the second group of predicted spectral lines is included in the measured vibration signal is known, determining the significant spectral lines not included in the first group of predicted spectral lines and the second group of predicted spectral lines comprises:

[0040] - for each significant spectral line and each rotating component and each periodic external disturbance, determining a likelihood function from a first model of the frequencies generated by said rotating component in rotation and from a second model of the frequencies of a second set of predicted spectral lines,

[0041] - for each likelihood function, evaluating said likelihood function at the frequency of each significant spectral line to obtain the likelihood probability at the frequency generated by said rotating component in rotation and at the frequency of the periodic external disturbance,

[0042] - determining a first probability that the frequency of the significant spectral line is included in the measured vibration signal,

[0043] - for each likelihood probability, determining the posterior probability produced by the rotation means in the rotation at the frequency of the significant line or at the frequency of the second set of predicted lines, the posterior probability being equal to the likelihood probability multiplied by the a priori probability and the product divided by the first probability,

[0044] - When the posterior probability is less than the classification threshold, determining that the significant spectral line is not included in the first group of predicted spectral lines and the second group of predicted spectral lines.

[0045] Advantageously, when the posterior probability is equal to or greater than the classification threshold, the significant spectrum line is considered to be included in the first set of predicted spectrum lines or the second set of predicted spectrum lines, and the first and second models are considered to be reliable.

[0046] Preferably, the likelihood function is a Gaussian mixture model, the mode of which is centered on the frequency generated by the rotating component in rotation or the frequency generated by the periodic external disturbance.

[0047] According to another aspect, a device for monitoring a first model of a rotating machine comprising a plurality of rotating components and a second model of a periodic external disturbance is proposed, the first model comprising a first set of predicted spectra and the second model comprising a second set of predicted spectra.

[0048] The equipment includes:

[0049] - a first determining device configured to determine a frequency spectrum of a measured vibration signal, the frequency spectrum comprising a plurality of spectral lines, the measured vibration signal comprising a measured vibration of the rotating machine,

[0050] - a selection device configured to select a spectral line from a plurality of spectral lines according to a predetermined selection rule, the selected spectral line being a significant spectral line,

[0051] - a second determining means, configured to determine significant spectral lines not included in the first set of predicted spectral lines and the second set of predicted spectral lines, the determined significant spectral lines being unidentified spectral lines,

[0052] - a third determining means configured to determine the waveform power of each unidentified spectral line,

[0053] - a comparison device configured to compare the determined power of each unidentified spectral line with a power threshold, and

[0054] - an actuation device configured to trigger an alarm if the determined power of at least one unidentified spectral line is equal to or greater than a power threshold.

[0055] According to another aspect, a system is proposed, which includes a rotating machine comprising multiple rotating parts, a measuring device configured to measure vibrations of the rotating machine and transmit measured vibration signals, a first model of the rotating machine comprising multiple rotating parts, a second model of periodic external disturbances, and a device as defined above connected to the measuring device, the first model including a first set of predicted spectral lines and the second model including a second set of predicted spectral lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Other advantages and features of the present invention will appear, but are in no way limiting, from an examination of the detailed description of embodiments and the accompanying drawings, in which:

[0057] Figure 1 An example of a system according to the invention is schematically shown;

[0058] Figure 2 An example of a method for monitoring a rotating machine according to the invention is schematically shown;

[0059] Figure 3 An example of a frequency spectrum of a vibration signal measured according to the present invention is schematically shown,

[0060] Figure 4 Schematically showing an example of a significant spectral line according to the present invention,

[0061] Figure 5 schematically shows an example of a likelihood function according to the present invention, and

[0062] Figure 6 A graphical example of comparing an unidentified spectral line with a power threshold according to the invention is schematically shown. DETAILED DESCRIPTION

[0063] refer to Figure 1 , which schematically represents a system 1 comprising a rotating machine 2, a measuring device 3, a first model M1 of the rotating machine 2, a second model M2 of a periodic external disturbance and a device 4 for monitoring the first model M1 and the second model M2.

[0064] The first model M1 comprises a machine vibration model and the second model M2 comprises an environmental model of the rotating machine.

[0065] The machine 2 comprises a plurality of rotating parts 2a, 2b, 2c driven, for example, by a shaft 2d of the machine 2.

[0066] The rotating parts 2a, 2b, 2c may be shaft couplings, bearings, or impellers.

[0067] The characteristics of each rotating part 2a, 2b, 2c are known, so the frequency f generated by the rotation of the rotating part 2a, 2b, 2c is2a , f 2b , f 2c is known or can be derived from its physical structure and the speed of shaft 2d.

[0068] The first model M1 includes a first set of predicted spectral lines, and the second model M2 includes a second set of predicted spectral lines.

[0069] The second model M2 simulates the periodic external disturbances that may occur in the measurement, such as the frequency f d Equivalent to electromagnetic interference of, for example, 50 Hz or 100 Hz.

[0070] Periodic external disturbances are generated by devices in the environment of the machine 2. The speed of the shaft 2d may be measured by a speed sensor and may have some variation.

[0071] The measuring device 3 is used to measure the vibration of the machine 2 and convert the measured vibration signal S M Passed to device 4.

[0072] The measuring device 3 may comprise a vibration sensor.

[0073] A set of frequencies SET includes the frequencies f generated by the rotation of the rotating parts 2a, 2b, 2c 2a , f 2b , f 2c , the frequency of periodic external disturbance f d and its harmonics constitute the expected set of frequencies of the machine 1 found in vibration measurements, assuming that the physical structure of the rotating machine is precisely known.

[0074] The device 4 comprises first determining means 5 , second determining means 6 , third determining means 7 , selecting means 8 , comparing means 9 , actuating means 10 , a memory 11 storing the set SET and processing means 12 .

[0075] The processing means 12 is intended to realize the first determining means 5 , the second determining means 6 , the third determining means 7 , the selecting means 8 , the comparing means 9 , the actuating means 10 and the memory 11 , and may include a processing unit.

[0076] Figure 2 An example of a method for monitoring a first model M1 and a second model M2 implementing a device 4 is schematically shown.

[0077] During a step 20, the measuring device 3 measures the vibrations of the machine 2 and transmits a measured vibration signal S comprising the measured vibrations. M .

[0078] During step 21, the first determination means 5 determine the measured vibration signal S M spectrum.

[0079] The first determination device 5 can implement a non-parametric spectrum analysis method, such as Fast Fourier Transform FFT, or implement a parametric spectrum analysis method, such as estimating signal parameters through the rotation invariant technique ESPRIT or the autoregressive model AR.

[0080] It is assumed below that the first determining means 5 implements a Fast Fourier Transform FFT algorithm.

[0081] Figure 3 The measured vibration signal S determined by the first determination device 5 is schematically shown. M Example of a spectrum.

[0082] The spectrum includes multiple spectral lines from C1 to C16.

[0083] During a step 21, the selection means 8 select a spectral line from a plurality of spectral lines according to a predetermined selection rule.

[0084] The selected spectral lines are significant spectral lines.

[0085] The predetermined selection rule may include a model order selection rule, such as the Akaike information criterion disclosed in the document "A new look at the statistical model identification", IEEE Transactions on Automatic Control, 19(6): 716–723, Akaike H. (1974).

[0086] Figure 4 Examples of significant spectral lines C6, C10, C12 determined by the selection device 8 according to frequency are schematically shown.

[0087] During a step 22 , the second determination means 6 determine significant spectral lines which are not included in the first set of predicted spectral lines of the first model M1 and in the second set of predicted spectral lines of the model M2 .

[0088] The second determination means 6 implements a model-based hypothesis test on the frequencies of the significant spectral lines.

[0089] The significant lines identified are the unidentified lines.

[0090] A first method of determining significant spectral lines that are not included in a first set of predicted spectral lines of a first model M1 and a second set of predicted spectral lines of a model M2 is described.

[0091] The first method is implemented by the second determination device 6 .

[0092] For each significant spectral line C6, C10, C12 and each rotating component 2a, 2b, 2c and each periodic external disturbance, the second determination means 6 predict the frequency (f) of the spectral line from the second set. d ) and a first model M1 and a second model M2 of frequencies (f2a, f2b, f2c) generated by the rotation of the rotating components (2a, 2b, 2c) to determine the likelihood function p.

[0093] For each likelihood function p, the second determination means 6 evaluates the frequency f of the significant lines C6, C10, C12 C6 ,f C10 , f C12 The likelihood function p at the position to obtain the frequency f generated by the rotating component in the rotation 2a , f 2b , f 2c Or the frequency of periodic external disturbance f d The likelihood probability.

[0094] The likelihood probability is recorded as (rotating part), where i is equal to 6, 10, 12, and j is equal to a, b, c in this example, or (Periodic external disturbance).

[0095] For each likelihood probability, the second determining means 6 converts the likelihood probability or Compare with the classification threshold τ.

[0096] The classification threshold τ is chosen in the range (0, 1), which balances the risk of false alarms (small τ) and the risk of missed detections (large τ).

[0097] When the likelihood probability or When it is smaller than the classification threshold, the significant spectrum line is determined to be not included in the first group of predicted spectrum lines or the second group of predicted spectrum lines, and the first and second models are considered to be reliable.

[0098] When the likelihood probability or When it is equal to or greater than the classification threshold τ, the significant spectrum line Ci is considered to be included in the first group of predicted spectrum lines or the second group of predicted spectrum lines, and the first and second models are considered to be reliable.

[0099] The likelihood function may be a Gaussian mixture model, the modes of which are centered on the frequencies generated by the rotating components 2a, 2b, 2c in rotation or by periodic external disturbances.

[0100] The frequencies generated by each rotating part 2a, 2b, 2c include the fundamental frequency f 2a , f2b , f 2c and its harmonics n. f 2a ,nf 2b , nf 2c , where n is an integer greater than or equal to 2.

[0101] Figure 5 Schematically illustrates the likelihood function .

[0102] Likelihood function The value is greater than the frequency f C10 The classification threshold τ of the significant spectrum line C10.

[0103] The significant spectral line C10 is identified as the rotation frequency f of the rotating component 2a. 2a harmonics.

[0104] It is considered that the significant spectrum line C10 is included in the first group of predicted spectrum lines or the second group of predicted spectrum lines, and the first and second models are considered to be reliable.

[0105] Likelihood function The value is less than the frequency f C6 , f C12 The classification threshold τ of the significant spectral lines C6 and C12.

[0106] The significant spectral lines C6, C12 are determined not to be included in the first set of predicted spectral lines and the second set of predicted spectral lines.

[0107] The second method performed by the second determination means 6 is used to determine significant spectral lines which are not included in the first set of predicted spectral lines of the first model M1 and the second set of predicted spectral lines of the second model M2.

[0108] The second method is implemented by the second determination device 6 .

[0109] When for the rotating parts 2a, 2b, 2c, the prior probability of the frequency generated by the rotating parts in rotation is and the prior probability that the frequencies of the second set of predicted spectral lines are included in the measured vibration signal When is known, the second method is implemented.

[0110] Prior probability The selection can be based on experience or specific machine knowledge.

[0111] The second method uses a Bayesian approach to improve the average performance of vibration source classification.

[0112] The group of significant lines is considered to include the lines C6, C10, C12 determined as defined above.

[0113] For each significant spectral line C6, C10, C12 and each rotating component 2a, 2b, 2c or periodic external disturbance, the second determination means 6 determines the likelihood probability defined above or .

[0114] The second determining means 6 determines the frequencies f of the significant spectral lines C6, C10, C12 ci Included in the measured vibration signal S M The first probability .

[0115] The second determining means 6 also determines the frequencies f of the significant spectral lines C6, C10, C12 c6 , f c10 , f c12 The frequency f generated by the rotating parts 2a, 2b, 2c during rotation 2j The posterior probability Or the frequencies f of the second set of predicted spectral lines d of .

[0116] Posterior probability Equal to the likelihood probability Multiply by the prior probability , the product divided by the first probability .

[0117] (1)

[0118] Or the posterior probability Equal to the likelihood probability Multiply by the prior probability , the product divided by the first probability .

[0119] (2)

[0120] The second determining means 6 converts the posterior probability or Compare with the classification threshold τ.

[0121] When the posterior probability or When it is less than the classification threshold τ, it is determined that the significant spectrum line Ci is not included in the first group of predicted spectrum lines and the second group of predicted spectrum lines.

[0122] When the posterior probability or When it is equal to or greater than the classification threshold τ, the significant spectrum line Ci is considered to be included in the first group of predicted spectrum lines or the second group of predicted spectrum lines, and the first and second models are considered to be reliable.

[0123] The likelihood function may be a Gaussian mixture model whose mode is centered on the frequency generated by the rotating components 2a, 2b, 2c in their rotation, or on the frequency f of a periodic external disturbance d As the center.

[0124] When it is determined that the unidentified spectral lines C6, C12 are included in the first group of predicted spectral lines and the second group of predicted spectral lines, in step 23, the third determination device 7 determines the waveform power of each unidentified spectral line C6, C12.

[0125] When the measured vibration signal S is determined by FFT or ESPRIT in step 21 M When the spectrum of the measured vibration signal S is determined by the third determining means 7 from the unidentified spectral lines C6, C12, for example by sinusoidal amplitude estimation, or when the measured vibration signal S is determined by AR in step 21. M When the spectrum is plotted, it is determined by the pole amplitudes.

[0126] The comparison means 9 compare the determined power of each unidentified spectral line C6, C12 with a power threshold value γ.

[0127] The choice of the power threshold γ is a trade-off between the risk of false alarms (small γ) and the risk of missed detections (large γ).

[0128] Figure 6 A graphical example of a comparison of unidentified spectral lines with a power threshold γ is schematically illustrated.

[0129] The waveform power of the unidentified spectrum line C6 is less than the power threshold γ, and the waveform power of the unidentified spectrum line C12 is greater than the power threshold γ.

[0130] When at least the waveform power of at least one unidentified spectrum line C12 is greater than the power threshold γ (step 24 ), in step 25 , the actuator 10 triggers an alarm.

[0131] If the waveform power of the unidentified spectral line is less than the power threshold γ (step 24), the method continues at step 20. The device 4 allows detecting modeling errors of the first model M1 and the second model M2 to avoid catastrophic failures or unnecessary maintenance and production stops due to misclassification of spectral lines.

[0132] Automatic detection runs during normal operation and allows machine uptime to be maintained.

[0133] Statistical and model-based hypothesis testing of significant spectral line frequencies allows a systematic and comprehensive search using all available machine and environmental knowledge.

[0134] Similar corresponding manual analysis can be cumbersome, especially on complex machines.

Claims

1. A method for monitoring a first model of a rotating machine (2) comprising a plurality of rotating parts (2a, 2b, 2c) and a second model of a periodic external disturbance, the first model comprising a first set of prediction spectra and the second model comprising a second set of prediction spectra, the method comprising: - Measuring the vibration of a rotating machine (2) to transmit a measured vibration signal (S M ), - Determine the measured vibration signal (S M ), which includes multiple spectral lines (C1, C2, C3, C4, C5, C6,C7, C8, C9, C10, C11, C12, C13, C14, C15, C16), - Select a spectrum line from multiple spectrum lines according to a predetermined selection rule, and the selected spectrum line is a significant spectrum line (C6, C10, C12), - Determine the significant lines (C6, C12) that are not included in the first set of predicted lines and the second set of predicted lines. The determined significant lines are the unidentified lines (C6, C12). - determine the waveform power of each unidentified spectral line (C6, C12), - compare the determined power of each unidentified spectral line (C6, C12) with a power threshold (γ), and - If the determined power of at least one unidentified spectral line (C12) is equal to or greater than the power threshold (γ), an alarm is triggered.

2. The method according to claim 1, wherein: Determine the measured vibration signal (S M )'s spectrum includes performing a non-parametric spectrum analysis method or a parametric spectrum analysis method.

3. The method according to claim 1 or 2, wherein: The predetermined selection rule comprises a model order selection rule.

4. The method according to any one of claims 1 to 3, wherein: Determining the significant spectral lines (C6, C12) not included in the first set of predicted spectral lines and the second set of predicted spectral lines comprises: - for each significant spectral line (C6, C10, C12) and each rotating component (2a, 2b, 2c) and each periodic external disturbance, from the frequencies (f2a, f2b, f2c) generated by said rotating component (2a, 2b, 2c) in rotation and the frequencies (f d )'s first model and the second model determine the likelihood function, - for each likelihood function, evaluate the likelihood function at the frequency of each significant spectral line to obtain the frequency (f 2a , f 2b , f 2c ) and the frequency of periodic external disturbance (f d ), - for each likelihood probability, compare the likelihood probability to the classification threshold (τ), - When the likelihood probability is less than a classification threshold (τ), the significant spectral lines (C6, C12) are determined to be not included in the first set of predicted spectral lines and the second set of predicted spectral lines.

5. The method according to claim 4, wherein: When the likelihood probability is equal to or greater than the classification threshold (τ), the significant spectrum line is considered to be included in the first group of predicted spectrum lines or the second group of predicted spectrum lines, and the first and second models are considered to be reliable.

6. The method according to any one of claims 1 to 3, wherein for a rotating component (2a, 2b, 2c), the frequency generated by the rotating component (2a, 2b, 2c) during rotation is included in the measured vibration signal (S M ) is known, and the prior probability that the frequencies of the second group of predicted spectral lines are included in the measured vibration signal is known, determining the significant spectral lines (C6, C10, C12) not included in the first group of predicted spectral lines and the second group of predicted spectral lines comprises: - for each significant spectral line (C6, C10, C12) and each rotating component (2a, 2b, 2c) and each periodic external disturbance, from the frequency (f 2a , f 2b , f 2c ) and the frequency of the second set of predicted spectral lines (f d )'s first model and the second model to determine the likelihood function, - for each likelihood function, evaluate the likelihood function at the frequency of each significant spectral line to obtain the frequency (f 2a , f 2b , f 2c ) or the frequency of periodic external disturbance (f d ), - Determine the frequencies of the significant spectral lines (C6, C10, C12) included in the measured vibration signal (S M ), - For each likelihood probability, determine the frequency of the significant line (C6, C10, C12) or the frequency of the second set of predicted lines (f d ) is generated by the rotating parts (2a, 2b, 2c) during rotation (f 2a , f 2b , f 2c ), the posterior probability being equal to the likelihood probability multiplied by the prior probability and the product divided by the first probability, - When the posterior probability is less than the classification threshold (τ), the significant spectrum line (C6, C12) is determined to be not included in the first set of predicted spectrum lines and the second set of predicted spectrum lines.

7. The method according to claim 6, wherein: When the posterior probability is equal to or greater than the classification threshold (τ), the significant spectrum line is considered to be included in the first group of predicted spectrum lines or the second group of predicted spectrum lines, and the first and second models are considered to be reliable.

8. The method according to any one of claims 4 to 7, wherein: The likelihood function is a Gaussian mixture model whose mode is centered on the frequency generated by the rotating component in rotation or the frequency of periodic external disturbances.

9. A device (4) for monitoring a first model of a rotating machine (2) comprising a plurality of rotating parts (2a, 2b, 2c) and a second model of a periodic external disturbance, the first model comprising a first set of predicted spectra and the second model comprising a second set of predicted spectra, the device comprising: - a first determining device (5), which is configured to determine the measured vibration signal (S M ), the spectrum including a plurality of spectral lines (C1, C2, C3, C4, C5, C6, C7, C8, C9, C10, C11, C12, C13, C14, C15, C16), the measured vibration signal including the measured vibration of the rotating machine, - a selection device (8), which is configured to select a spectral line from a plurality of spectral lines (C1, C2, C3, C4, C5, C6, C7, C8, C9, C10, C11, C12, C13, C14, C15, C16) according to a predetermined selection rule, the selected spectral line being a significant spectral line (C6, C10, C12), - a second determining means (6), configured to determine significant spectral lines (C6, C10, C12) not included in the first set of predicted spectral lines and the second set of predicted spectral lines, the determined significant spectral lines being unidentified spectral lines (C6, C12), - a third determining means (7) configured to determine the waveform power of each unidentified spectral line (C6, C12), - comparison means (9) configured to compare the determined power of each unidentified spectral line (C6, C12) with a power threshold (γ), and - an actuation device (10) configured to trigger an alarm if the determined power of at least one unidentified spectral line (C12) is equal to or greater than a power threshold (γ).

10. A system (1), comprising: A rotating machine (2) comprising a plurality of rotating parts (2a, 2b, 2c) configured to measure vibration of the rotating machine (2) and transmit a measured vibration signal (S M ) of a measuring device (3), a first model of a rotating machine (2) comprising a plurality of rotating parts (2a, 2b, 2c), a second model of a periodic external disturbance, and an apparatus (4) according to claim 9 connected to the measuring device, the first model comprising a first set of predicted spectral lines and the second model comprising a second set of predicted spectral lines.

Images