Device for monitoring the state of damage of a power transmission

Abstract

According to one aspect, the invention relates to a method for monitoring the state of health of a planetary gear train equipping a rotary machine and suitable for transmitting power over a line of shafts of the rotary machine, the method comprising the following steps: - acquiring, by means of a vibration sensor, a vibratory signal of the rotary machine, the vibratory signal including vibrations generated during the transmission of power by the planetary gear train; - constructing a measurement vector and a transition matrix of a phenomenological vibration model; - recursively estimating a possible defect vibration signature from the measurement vector, the transition matrix and the acquired vibration signal, the possible defect vibration signature taking into account a modulation overlay effect; and - determining a distance by comparing the possible defect vibration signature with a reference signature.

Description

[0001]DESCRIPTION TITLE: Device for monitoring the damage status of a power transmission TECHNICAL FIELD OF THE INVENTION The technical field of the invention is that of monitoring the health status of mechanical components used for power transmission. The present invention relates to a method and a device for monitoring the health status of an epicyclic gear train. TECHNOLOGICAL BACKGROUND OF THE INVENTION Shaft lines integrated into rotating machines, for example an aircraft engine, are conventionally equipped with various mechanical parts or components, such as bearings and gears. Among these equipments, epicyclic gear trains, also called planetary reducers, are mechanical parts comprising several concomitant gears. An example of an epicyclic gear train comprising 4 planetary gears 13 and a sun gear 14 is proposed in Figure 1. The operation of an epicyclic gear traingenerates complex vibration signals subject to modulation overlap. Monitoring the health of such a component, to detect excessive and premature degradation, is therefore not easy to implement. However, it is essential to ensure good mechanical strength and the service life of the shaft line equipped with it in order to avoid operating anomalies in the systems in which they are integrated. The modulation phenomenon is already known for parallel axis gears. This phenomenon is linked to the amplitude and / or phase modulation of the fault frequency by the meshing frequency. In the case of an epicyclic gear train, the modulation is much more complex due to the presence of several elements within the same train. For example, when the planet carrier is rotated, while keeping the crown stationary, there will be modulation of the rotation frequency of the sun gear and the rotation frequency of the planets by the frequencyrotation of the planet carrier. This modulation is all the more complex when the train includes several planets. In the case of an epicyclic train, we speak of multi-modulation. "Modulation overlap" is understood to mean the phenomenon by which a frequency of a defect on a gear appears, in the spectrum of the associated vibration signal, at an erroneous location because the modulation frequency of the gear is higher than the frequency of the defect. An analogy of this phenomenon can be made in optics: when an observer looks with the naked eye at a wheel of a vehicle whose rotation frequency is higher than the sampling frequency of the eye, the observer has the impression that the wheel is rotating in the opposite direction of its rotation. This phenomenon is specific to epicyclic trains for which there is always a modulation frequency greater than the frequency of the defect, compared to gears with parallel axes. In addition, the multiplicity ofmodulation sources for an epicyclic gear train generates a much more complex overlap than that which can occur with a parallel-axis gear. Approaches for monitoring gears are known from the state of the art based on the estimation of amplitude and / or phase modulations of the meshing, which are signatures revealing the health of the gears (PD McFadden, 'Detecting fatigue cracks in gears by amplitude and phase demodulation of the meshing vibration', 1986; US6526356B1; US6898975B2; EP2434266A2; US9797808B2; US8963733B2). These techniques focus on the estimation of modulations around the meshing or its harmonics and are relevant in the case of non-overlap of the modulations between them and for a shaft operating at a stationary operating regime. However, they are not suitable for the case of epicyclic trains whose vibration signals are subject to spectral folding, generating the overlap ofmodulations. In particular, these approaches do not take into account the interactions between the natural modulations of the train and the modulations linked to its damage, nor the masking of gear modulations by noise. Indeed, in the vibration signals of an epicyclic gear train there are not only specific frequencies linked to the damage of the gear but also frequencies modulating the meshing, for example: the rotation frequencies of the planet carrier, the frequencies of a defect, the interaction between the defect and the variation of position of the defect relative to the fixed sensor, etc. Furthermore, these approaches are not adapted to the case where the operating regime of one of the connected shafts of the epicyclic gear train is non-stationary, which is nevertheless the case in aeronautical applications. There is therefore a need for a means of monitoring an epicyclic gear train that is robust to the recovery of modulations in operating conditionsstationary and non-stationary operation of the rotating machine. SUMMARY OF THE INVENTION The invention provides a solution to the problems mentioned above, by making it possible to monitor the health status of epicyclic gear trains equipped on a high-power transmission system, for example on a rotating machine. A first aspect of the invention relates to a method for monitoring the health status of an epicyclic gear train equipped on a rotating machine and adapted to carry out a power transmission on a shaft line of said rotating machine, the method comprising the following steps: - Acquisition by a vibration sensor of a vibration signal from the rotating machine, the vibration signal comprising vibrations generated during the transmission of power by the epicyclic gear train; - Construction of a measurement vector and a transition matrix of a phenomenological vibration model, this model being based on a series decompositionof the vibration signal taking into account interactions of different vibration sources of the epicyclic gear train; - Estimation of a vibration signature of a possible defect from the measurement vector, the transition matrix and the acquired vibration signal, the vibration signature of a possible defect taking into account a modulation overlap effect; - Determination of a distance by comparing the vibration signature of a possible defect with a reference signature. Thanks to the invention, it is possible to reliably and robustly determine the presence of a defect on one or more elements of the epicyclic gear train. Indeed, thanks to the phenomenological modeling of the vibration signal of the gear and the recursive estimation of the parameters of the constructed model, it is possible to estimate the different modulation components carrying information on the health status of a gear as well as their mutual interaction. The modulations are thus estimatedby a deterministic approach thanks to the phenomenological vibration model and a priori knowledge of the kinematics of the power transmission, contained in the measurement vector and the transition matrix. In addition, the modeling takes into account the interactions between the modulations generated by the power transmission within the epicyclic gear train, coming from the various vibration sources that are the different elements of the epicyclic gear train (planets, planet carrier, solar and crown), making the approach robust to modulation overlap. Advantageously, the proposed solution is valid for both a stationary and non-stationary regime of the rotating machine. In addition, taking into account the different modulation sources makes it possible to be robust to noise and peaks not related to the power transmission. Furthermore, the method can be used in real time, it can be used to monitor the progression of damage, for example thepropagation of a crack from a tooth through the gear. In addition to the characteristics which have just been mentioned in the preceding paragraph, the method according to the first aspect of the invention may have one or more complementary characteristics among the following, considered individually or according to all technically possible combinations. In one embodiment, the vibration signal is acquired during an acquisition duration, the acquisition duration being at least as long as a duration corresponding to a predetermined number of rotation cycles of a shaft of the rotating machine connected to the epicyclic gear train. Thanks to this embodiment it is possible to carry out monitoring of the epicyclic gear train in real time, by successive repeated implementations of the method according to the first aspect of the invention. In one embodiment, in the acquisition step a rotation speed of the shaft of the machine is also measuredrotating connected to the epicyclic gear train. Thanks to this embodiment it is possible to have a reference speed to construct the phenomenological vibration model. In one embodiment, the measurement vector is constructed from kinematic data of the epicyclic gear train and the shaft to which the epicyclic gear train is connected and from parameters of the phenomenological vibration model. In one embodiment, the transition matrix is ​​an identity matrix whose size depends on the parameters of the phenomenological vibration model. In one embodiment, the step of estimating the vibration signature of a possible defect comprises the following two sub-steps: - Recursive estimation of an estimated vector of the parameters of the model of the acquired signal, the estimation being a recursive estimation carried out by means of a Kalman filter, the Kalman filter taking as input the acquired vibration signal, the transition matrix and the measurement vector; -Reconstruction of the vibration signature of a possible fault from the estimated vector of the parameters of the acquired signal model. In one embodiment, the distance is a difference between a standard deviation of a possible fault indicator calculated for the vibration signature of a possible fault and a standard deviation of the possible fault indicator calculated for the reference signature. A second aspect of the invention relates to a device for monitoring the health status of an epicyclic gear train, the device comprising: - An acquisition module comprising at least the vibration sensor and configured to implement the acquisition step of the method; - A processing module configured to implement the steps of constructing a measurement vector and a transition matrix, estimating a vibration signature of a possible fault, determining a distance and issuing an alert. This second aspect according to the invention makes it possible to easilyimplement the method according to the first aspect by means of a simple device. A third aspect of the invention relates to a computer program product comprising instructions which, when the program is executed on a computer, cause the latter to implement the steps of the method according to the first aspect. A fourth aspect of the invention relates to a computer-readable medium comprising instructions which, when executed by a computer, cause the latter to implement the steps of the method according to the first aspect. The invention and its various applications will be better understood upon reading the following description and examining the accompanying figures. BRIEF DESCRIPTION OF THE FIGURES The figures are presented for information purposes only and in no way limit the invention. - Figure 1 is an illustration of an epicyclic gear train. - Figure 2 is a block diagram illustrating the sequence of steps of a methodaccording to the invention. - Figure 3 is a vibration signal of the epicyclic gear train acquired during the execution of the method. - Figure 4 is a spectrum of the acquired vibration signal. - Figure 5 is a spectrum of an estimation around the fundamental meshing. - Figure 6 is a spectrum of the signature of a solar without taking into account the modulation effect by a planet carrier. - Figure 7 is the spectrum of the estimation of the signature of the solar with a meshing effect and a modulation effect of the 1st harmonic of the planet carrier. - Figure 8 is the spectrum of the signature with the taking into account the effect of the modulation of the 4th harmonic of the planet carrier. - Figure 9 is a graph representing an evolution of a possible fault indicator. DETAILED DESCRIPTION Unless otherwise specified, the same element appearing in different figures has a single reference. A first aspect of the invention relates to a method for monitoring the state ofhealth of an epicyclic gear train equipped on a rotating machine. The epicyclic gear train is adapted to carry out a power transmission on a shaft line of said rotating machine. The term "rotating machine" means a motor which transforms the energy supplied to it into a rotary movement, for example through a shaft line. In the context of the invention, this particularly concerns aircraft such as airplanes or helicopters, but it can also concern wind turbine engines, motors of rolling vehicles, etc. Figure 1 is a schematic representation of the epicyclic gear train 10. This comprises several elements: a ring gear 11, a planet carrier 12, four planets 13 and a sun gear 14. Each element of the epicyclic gear train 10 is connected to one of the shafts of the rotating machine. The stress on the epicyclic gear train 10 by the rotation of one of its elements generates vibrations coming from each of the elements and possible defects. Thesevibrations can be captured by a vibration sensor which will produce a vibration signal comprising the vibrations produced by each of the aforementioned sources as well as noise. This may be noise coming from other parts of the rotating machine or noise linked to the environment of said machine. By "defect" is meant a discontinuity in the properties of the material making up a part or an inspected object, in this case the epicyclic gear train 10. This discontinuity results from an anomaly present in the material. This anomaly can have various origins and be of varied nature. These anomalies are mainly the consequence of hazards which occur during the manufacture of the part. These anomalies also occur quite frequently during the use of the part or its handling: the material may, for example, have been weakened during the manufacturing process and its use, generating high local stresses at thelevel of the weakened zone, or following an impact, causes a defect. The term "defect" therefore covers all forms of anomalies that the material may undergo: material defect, inclusion, crack, porosity, corrosion, alteration of the material properties, etc. In particular, the case of a tooth defect is considered here. In general terms, consider that the epicyclic gear train 10 comprises ^ ^ planets 13 with ^ ^ teeth, the planet holder 12, the solar 14 with ^ ^ teeth and crown 11 with ^ ^ teeth. The angular velocity of the planet carrier 12, the angular velocity of the solar 14, the angular velocity of the corona 11 and the angular velocity of one of the planets 13 are respectively noted ^ ^^ , ^ ^ , ^ ^ and ^ ^ . Similarly, the angular velocities of a defect on the planet carrier 12, on the solar 14, on the corona 11 and on one of the planets 13 are respectively noted ^^ ^^ ^ ^ , ^^ ^ ^ ^ and ^ ^ ^^^ . By generalization, ^ ^^^^^is the angular velocity of the defect regardless of the element on which it is located. It is considered that the vibration signal of the epicyclic gear train 10 comprises two components: - A natural component linked to the engagement of the teeth of the planets with those of the crown or between those of the sun; - An abnormal component linked to the presence of the defect on one of the elements. The amplitude associated with each of the components is modulated due to the mobility of the gear members. The amplitude of each component in the vibration signal is therefore more or less strong depending on the relative position of the point of contact between the teeth, which are mobile, with the sensor, which is fixed. Explicitly, the meshing signal ^ in the presence of the defect can be modeled using a phenomenological model, in discrete time, in the form ^ ^ [Math.1] ^ ^^^^^,^!" ^^^where: - ) is a number of harmonics of the meshing, defined by an operator according to the desired precision of the model, - ^ ^ is a weighting function resulting from the position of the contact point of the *-th planet relative to the position of the fixed sensor, - ^ ^,^^^^^ is the signal generated by the defect in contact with the *-th planet, - ^ ^,^,%&^' is the +-th meshing harmonic generated by the *-th planet, - ( is a measurement noise, characterized by a variance , and containing the sensor noise, the structure noise and the unmodeled vibration components, - And ^ is the index of the discrete time or that of the signal sample. Four stress configurations of the epicyclic gear train 10 are to be considered: - Configuration 1: the crown is fixed and the other gear members are mobile; in this case, the weighting function ^ ^is non-zero. - Configuration 2: the planet carrier is fixed and the other gear components are mobile with the planets rotating around their axis of rotation; in this case, the weighting function ^ ^ is zero. - Configuration 3: the solar is fixed and the other gear components are mobile; in this case, the weighting function ^ ^ is non-zero. - Configuration 4: all the gear components are mobile; in this case, the weighting function ^ ^ is non-zero. In the following, only configurations 1, 3 and 4, where the weighting function is non-zero, are considered. In the second configuration, the weighting function ^ ^ is zero and the epicyclic gear train 10 is treated as a parallel-axis gear. In configurations 1, 3 and 4, the weighting function is considered to be periodic at the rotation period of the planet carrier - ^^ Therefore, the function ^ ^ can be approximated by a trigonometric series with coefficients [Math. 2^^8,2!" , : - 9 the number of harmonics contained in the weighting function ^ ^ , defined by the operator according to the precision of the desired model, - 1 2,^ an amplitude of the weighting function for the :-th harmonic of the *-th planet, - ^ ^ the inverse sampling period of the sampling frequency; ^ , - ^ ^ a time shift between the vibration of the *-th planet and that of the (* < 1)-th planet and is equal to ^ ^ ^ = ^ / ^ ^^ with = ^ the angular phase shift between two planets In vector form, the weighting function is written as [Math.3] ^ ^ ^^^ ^ ? A @7 ^^^B ^ ^^^, with: - [Math. ^ - Et B^ ^ ^ ^ ^ ^1^8,^ ^^^ ⋯ ∈ The meshing between the teeth being periodic at the meshing period, the +-th meshing harmonic of the *-th planet is written [ Math. : - H ^,^ the complex envelope of the meshing considered; this envelope contains the amplitude and the phase of the meshing. For a fault on the epicyclic gear train 10, the characteristic signal of the fault is periodic at the period of the fault - ^^^^^ ^ . / 0 LMNOP , where ^ ^^^^^ is the angular velocity of the defect. As before, this signal can be expressed in the form of a trigonometric series such as ∈ , -[Math. 5b] S^ ^ ^ ^ ^ ^Q^R,^ ⋯ QR,^^A ∈ ℂ .RG^ , - And T is the harmonic number of the characteristic signal of the fault, defined by the operator according to the desired precision of the model. ^ In vector form, this signal is written [Math.7] ^ ^ ^ ^ ^ ^ ^ ^ ^^^ ? A ^ where: and In a compact writing, the vibration signal ^ can thus be written [Math.9] ^^^^ ^ ? A ^^^W^^^ ^ (^^^, with: - [Math. 9a] ? ^ ^ ^ ^ ^ ? A ^ ^ ^^^ … of measurement, - [Math. of the model parameters. The vector of the model parameters W implicitly contains the information related to the health status of the gear components. Given the slow variation of the parameters of the vector W, it is possible to apply a smoothing constraint to the vector of the model parameters. The smoothing constraint is of order greater than or equal to 1. Preferably, the smoothing constraint is of order 1, and its explicit writing is [Math.10] W^^ ^ 1^ ^ W^^^ ^ [^^^, with - [ a Gaussian white noise vector with covariance matrix \ ; the size of the white noise vector is the same as that of the vector of the model parameters. In this case, the transition matrix ] is the identity matrix of size 2)^ ^ ^ 1 ^ 2 ^ T ^ 9 ^ ^ 4T9 ^ G 2)^ ^ ^ 1 ^ 2 ^ T ^ 9 ^ ^ 4T9 ^. Alternatively, the size of the transition matrix ] can be different if the order of the smoothing constraint is greater than 1. For example, for a smoothing constraint of order 2, the equation [Math. 10] becomes W^^ ^ 1^ ^ 2W^^^ < W^^ < 1^ ^ [^^^ and the transition matrix ^ ` 0 1 The equations [Math. respectively the measurement equation and the equation of state for the vibration signal ^. Advantageously, the above phenomenological model can be applied to epicyclic gear trains as well as to parallel-axis gears, which are explicitly integrated into this vibration model. According to the above, the challenge of the present invention is therefore to produce an estimate of the vector of the parameters of the model of the acquired signal, denoted Wc, from the construction of the transition matrix ] of said signal and a determination of the measurement vector ?. The interest is then to extract one or more vibration signatures of possible defect and then to compare and analyze these vibration signatures with one or more reference signatures d e&^in order to detect the presence of the possible defect(s) on one or more elements of the epicyclic gear train 10. The method 100 for monitoring the health status of an epicyclic gear train is shown diagrammatically in FIG. 2. The method 100 comprises five steps numbered from 101 to 105. The first step 101 is a step of acquiring the vibration signal ^ by means of a vibration sensor. The vibration sensor is, for example, a displacement sensor, a speed sensor or an accelerometer. Preferably, the vibration sensor is an accelerometer based on piezoelectric technology. The vibration signal is acquired at the sampling frequency; ^. Preferably, the sampling frequency is at least twice as high as the maximum number of meshing harmonics considered for the epicyclic gear train. The acquisition time of the vibration signal is at least as long as a time corresponding to a predetermined number of rotation cycles of a shaft of the rotating machine connected to the epicyclic gear train 10. The predetermined number of rotation cycles of the shaft is greater than 1 and may be an integer or real. Preferably, the predetermined number of rotation cycles of the shaft is chosen so as to cover enough rotation cycles to guarantee a robust and reliable analysis of the signal and to limit the size of the signal, thus allowing a rapid analysis of said signal and real-time monitoring of the epicyclic gear train 10. The acquisition time is therefore advantageously short to allow the repetition of the implementation of the method 100 at a real-time rate.The acquisition duration is at least as long as the duration corresponding to the predetermined number of rotation cycles of the shaft connected to the epicyclic gear train 10 which has the slowest rotation speed. The duration of the signal can, moreover, be fixed by a maximum number of samples - to be acquired, at the sampling frequency;. ^ The vibration sensor is placed on or near the rotating machine. Preferably, the vibration sensor is placed near the shaft connected to the epicyclic gear train 10, for example on a frame of said shaft. The acquisition step 101 may also concern the measurement of a rotation speed of the shaft connected to the epicyclic gear train 10. The rotation speed is noted ^ ". The rotational speed of the shaft can be obtained directly by means of a speed sensor, for example a tachometer. It is also possible to use another type of speed sensor, providing a square, sinusoidal or series of pulses speed signal. This speed signal is then processed to estimate the rotational speed of the shaft. This processing can be carried out by: - ​​Detection of rising edge instants of the square speed signal; - Estimation of the instantaneous frequency of the sinusoidal speed signal; - Time localization of the pulses. Preferably, the rotational speed of the shaft is measured simultaneously with the vibration signal, throughout the acquisition period. Alternatively, the rotational speed can be determined from the operating speed of the rotating machine, for which the rotational speeds of the different shafts as a function of its operating speed are known.The second step 102 is a step of constructing the measurement vector ? and the transition matrix ]. The interest of constructing the measurement vector ? is to model the frequency localization of the frequencies of interest, in particular the meshing frequency, the gear frequencies and those of a possible defect. The interest of constructing the transition matrix ] is to describe the type of variation of the amplitude of the damage signatures or the amplitudes of the modulations, in particular to specify whether this variation is fast or slow over time. The transition matrix ] is constructed according to the smoothing constraint chosen for the model. In this case, for a smoothing constraint of order 1, the transition matrix ] is the identity matrix of size 2)^. ^ ^1 ^ 2^T ^ 9^ ^ 4T9 ^ G 2)^ ^ ^ 1 ^ 2 ^ T ^ 9 ^ ^ 4T9 ^, as described above. The measurement vector ? is constructed from known data of the kinematics of the shaft and the epicyclic gear train 10, according to equations Math.1 to 9. In this case, the measurement vector ? is constructed from: - the number 9 of harmonics contained in the weighting function ^ ^ , - of the number T of harmonics of the characteristic signal of the fault, - of the number ) of harmonics of the meshing, - of the angular speed of the fault - of the sampling period ^ ^ - time lag ^ ^ between the vibration of the i-th planet and that of the (i-1)-th planet, - of the weighting function ^ ^ , - rotation frequencies of the crown, the solar, the planets and the planet carrier, ; ^ , ; ^ , ; ^ , And ; ^^respectively. The rotation frequencies of the crown, the solar, the planets and the planet carrier, coming from the kinematics of the epicyclic train, are known, for example by reading Table 1 below. [Table 1] C ^ ^ ^ ;^ ^ ^ ^^ ^ ^ ^^ ^ ^ ^ ^ ^ ^;^ ^ ; ^^ ^ 0 ; ^ ^ ^ ^ < ; ^ < ^ ; ^ ^ ^ ; ^ It should be noted that the angular velocity of the defect ^ ^^^^^ depends on the element on which it is located, such that ^ ^^^^^ ^ 2j; k ^^^ , with ; k ^^^ the frequency of the defect on element l ∈ m ^, n, o p . The input frequency; &f^eé& of the epicyclic gear train 10 is determined from the rotational speed of the shaft ^ "or is estimated from the vibration signal. Step 103 is then a step of estimating the vibration signature(s) of a possible defect. The defect may correspond to a tooth defect of one of the elements of the epicyclic gear train 10. Step 103 of estimating vibration signatures of a possible defect comprises two sub-steps 103a and 103b. Sub-step 103a is a step of estimating the estimated vector c W of the parameters of the model of the acquired signal. The estimation is preferably carried out recursively by means of a Kalman filter, for example according to the Rauch–Tung–Striebel variant. The advantage of using such a Kalman filter is to benefit from its recursiveness and its applicability in real time. It is alternatively possible to use other robust estimators such as an LMS filter (in English, “Least Mean-Squares”) or a synthesis q r. As input, the Kalman filter uses the transition matrix ], the measurement vector ? and the acquired signal ^. As output, the Kalman filter provides the estimatec W of the vector of the parameters of the acquired signal model. The Kalman filter uses as parameter the initialization of the estimated vector Wc^1^, a covariance matrix s^1^ of an initialization error, the covariance \ of the state noise and the variance , of the measurement noise. The advantage of using such an estimator is that it allows the noise to be filtered and the estimate to be smoothed in the event of an error during the filtering phase. Then, sub-step 103b is a step of reconstructing the vibration signature(s) of a possible defect t d. The reconstruction is carried out using the previous equations [Math. 1 to 9] and from the estimate Wc of the vector of the parameters of the acquired signal model. Each vibration signature of possible defect t d is a constructed matrix such that d t^ - - -wx^,^^w? ^ ^∑ ^ ^̂ ^,^, %&^' ^^ ^ 1^ ⋯ ∑ ^ ^̂ ^,^, %&^' ^^ ^ -^ ^ y , - The vectors uc , wx ^^^^^ , wx ^,%&^' wx^f^&e^^^^hf are of size - or have a number of samples equal to -. The vector wx z{|}~ expresses the estimation of the vibration signature of the possible defect. The vector wx^f^&e^^^^hfexpresses the interaction between the defect and the weighting function due to the fixed position of the sensor relative to the variable position of each planet. The reconstruction of the vector uc can be carried out in the following way: for each sample ^, set to zero the elements of the measurement vector ? except those corresponding to the vector ? u^ . The vectors w x ^^^^^ , w x ^,%&^' and w x^f^&e^^^^hf can be reconstructed in the same way as for the vector uc . Step 104 is then a step of comparing the vibration signature(s) of possible defect t d with the reference signature d ^^z . If no reference signature d ^^z is not available, the or at least one of the vibration signatures of possible defect t d then becomes the reference signature(s) d ^^z . Preferably, the vibration signature of possible defect t d is that of a healthy epicyclic gear train, i.e. without defects. The comparison provides a distance between each of the vibration signatures of possible defects t d and the reference signature d ^^z. The distance is determined as a difference between a standard deviation of a possible default indicator and a standard deviation of a reference indicator. The possible default indicator and the reference indicator are of the same nature, i.e. they are obtained from the same mathematical formula. In particular, the indicators may be energy indicators, for example an effective value of the signature, and / or a statistical indicator, for example a kurtosis. For example, the effective value of the signature w x z{|}~ is ^ &^^ ^ ^^ ^ that wx z{|}~ ^ ^ ^ ^ ∑ ^ ^̂ ^, ^^^^^ ^ ^ ^ . In this case, the distance is the between the standard deviation of the effective value of the vibration signature of possible defect t d and the standard deviation of the effective value of the reference signature d ^^z. Step 105 is, finally, a step of issuing an alert as a function of the distance calculated in the previous step. The alert is triggered when the absolute value of the distance is greater than or equal to an alert threshold. This alert threshold may be an integer or real multiple of the standard deviation of the reference indicator. For example, this threshold is equal to three times the absolute value of the standard deviation of the reference indicator. Furthermore, an alarm may be triggered when the distance exceeds an alarm threshold. This alarm threshold may be an integer or real multiple of the standard deviation of the reference indicator. For example, this threshold is equal to six times the absolute value of the standard deviation of the reference indicator. The alert and / or the alarm make it possible to inform the operator about the state of the rotating machine, in particular that the fault is detected on one of the epicyclic gear train elements 10.From this alert and / or the alarm, the operator can decide to trigger a maintenance operation in order to correct the detected fault. The alert informs of a minor damage which does not require stopping the machine while the alarm informs of a severe damage requiring stopping the machine. Once the alert and / or the alarm has been issued, it is possible to repeat the execution of the method 100 to acquire a new vibration signal, according to step 101, and to carry out the analysis of said new signal according to steps 102 to 105. In the case where no alert or alarm is issued, it is also possible to repeat the execution of the method 100 to acquire the new vibration signal, according to step 101, and to carry out the analysis of said new signal according to steps 102 to 105. In an alternative, the monitoring is translated into graphical form.In this case, the evolution of the different indicators is displayed, as well as the alert and / or alarm thresholds, throughout the monitoring, by the repeated execution of the method 100. A second aspect according to the invention relates to a device for monitoring the health status of the epicyclic gear train 10. The device comprises software and hardware means for implementing the method 100. In particular, the monitoring device comprises an acquisition module comprising the vibration sensor, a signal conditioner, an analog-digital converter, a volatile and / or non-volatile memory and a processor. Instructions are included in the memory of the acquisition module which, when executed by the processor, allow the implementation of the acquisition step 101 of the method 100, for the acquisition of the vibration signal and, if necessary, the rotation speed of the shaft connected to the epicyclic gear train 10.The monitoring device also comprises a processing module, comprising a processor and a volatile or non-volatile memory. The memory of the processing device comprises instructions which, when executed by the processor, enable the implementation of steps 102 to 105 of the method 100. The processing module may also comprise display means, such as a screen and a graphical interface, for translating the monitoring into graphical form. The acquisition module and the processing module may be implemented in two different devices. Two examples are proposed below to demonstrate the performance and usefulness of the method 100. The first example concerns the monitoring of an epicyclic train on a measuring bench. The second example concerns the monitoring of a progression of damage. In the first example, the crown 11 has ^. ^ ^ 96 teeth and is fixed. The gear input is the sun 14 with ^^ ^ 34 teeth and the output is the planet carrier 12 with ^ ^ ^ 5 planets 13 of ^ ^ ^ 31 teeth. The vibration signal is acquired at a sampling frequency; ^ ^ 51.2 kHz. On this bench, a seizure on the solar 14 was observed. Method 100 is therefore applied to extract the signature of the defect of the solar 14, the period of which is equal to the rotation period of said solar 14. Figure 3 is an extract of the vibration signal ^ acquired as well as the rotation frequency; ^^of the planet carrier 12, expressed as a function of time t. Figure 4 shows the spectrum ^ of the vibration signal ^ as a function of the machine orders ^T, the reference of which here is the planet carrier 12. Method 100 makes it possible to extract the signature of the damage to the solar 14 from the vibration signal without taking into account the modulation induced by the planet carrier 12. The harmonic number of the signature of the solar 14 is set, for illustration, to M=3. Figure 5 shows the spectrum of the raw signal as well as the signature of the solar determined from the estimation of the vector of the parameters of the associated model. Figure 6 shows the spectrum of the signature of the solar 14 without taking into account the effect of the modulation by the planet carrier 12. The prominence of the peaks linked to the order of the solar defect, which is at 2.8235, and its harmonics, are clearly distinguishable.However, this estimation does not take into account the effect of the modulation generated by the planet carrier 12. In figure 7 is represented the spectrum of the estimation of the signature of the solar 14 with the effect of the meshing and the modulation effect of the harmonic 1 of the planet carrier 12. In figure 8 is represented the spectrum of the signature with the taking into account of the effect of the modulation of the planet carrier 12 when the first harmonic of the frequency of the solar 14 is considered, and that the latter is modulated by the harmonic 4 of the planet carrier 12. It can therefore be observed that taking into account the interaction between the frequencies of the signature and those of the planet carrier 12 makes it possible to better explain the peaks in the spectrum and thus to better reflect the real state of the teeth of a power transmission by epicyclic train 10.In the second example, method 100 is applied to vibration data of damage to the epicyclic gear train 10 with damage propagation. During the test, a crack was detected in the tooth root of a planet 13, which then propagated across the entire width of the gear body. To apply method 100, the following parameters are chosen: + ^ 1, T ^ 4, 9 ^ 0, , ^ 10, ^ ^ 10. ^^G ^, ^ being the identity matrix of appropriate size. The initializations for the Kalman filter are made by random selection following a Gaussian distribution. The indicator of possible defect is here the RMS (root-mean-square) value, applied to the signature of the defect whose fundamental frequency is that of the defect of planet 13. The evolution of the indicator of possible defect is displayed in Figure 9. The threshold is set here, illustratively, at 2 times the standard deviation (indicated by the notation ^ ^ 2^ where ^ is the mean value of the indicator and ^ is the standard deviation) of the same indicator in the absence of damage to the gear, which is the case before the 350 e measurement. It is possible to observe a clear increase in the effective value, reflecting a propagation of the damage over the entire body of the gear, as was observed during the test.

Claims

CLAIMS

1. Method (100) for monitoring the health of an epicyclic gear train (10) equipped on a rotating machine and adapted to carry out a power transmission on a shaft line of said rotating machine, the method (100) comprising the following steps: - Acquisition (101) by a vibration sensor of a vibration signal from the rotating machine, the vibration signal comprising vibrations generated during the transmission of power by the epicyclic gear train (10); - Construction (102) of a measurement vector and a transition matrix from a phenomenological vibration model, this model being based on a Fourier series decomposition of the vibration signal taking into account interactions of different vibration sources of the epicyclic gear train;- Estimation (103) of a vibration signature of a possible defect from the measurement vector, the transition matrix and the acquired vibration signal, the vibration signature of a possible defect taking into account a modulation overlap effect;- Determination (104) of a distance by comparison of the vibration signature of a possible fault with a reference signature.

2. Method (100) according to the preceding claim, in which the vibration signal is acquired during an acquisition duration, the acquisition duration being at least as long as a duration corresponding to a predetermined number of rotation cycles of a shaft of the rotating machine connected to the epicyclic gear train (10).

3. Method (100) according to the preceding claim, in which, in the acquisition step, a rotation speed of the shaft of the rotating machine connected to the epicyclic gear train (10) is also measured.

4. Method (100) according to one of claims 2 and 3, in which the measurement vector is constructed from kinematic data of the train; epicyclic gear (10) and the shaft to which the epicyclic gear train (10) is connected and from parameters of the phenomenological vibration model.

5. Method (100) according to one of the preceding claims, in which the transition matrix is ​​an identity matrix whose size depends on the parameters of the phenomenological vibration model.

6. Method (100) according to one of the preceding claims, in which the step of estimating (103) the vibration signature of a possible fault comprises the following two sub-steps: - Recursive estimation (103a) of an estimated vector of the parameters of the model of the acquired signal, the estimation being a recursive estimation carried out by means of a Kalman filter, the Kalman filter taking as input the acquired vibration signal, the transition matrix and the measurement vector; - Reconstruction (103b) of the vibration signature of a possible fault from the estimated vector of the parameters of the model of the acquired signal.

7. Method (100) according to one of the preceding claims, wherein the distance is a difference between a standard deviation of a possible fault indicator calculated for the possible fault vibration signature and a standard deviation of the possible fault indicator calculated for the reference signature.

8. Device for monitoring the health status of an epicyclic gear train (10) for implementing the method according to any one of the preceding claims, the device comprising: - An acquisition module comprising at least one vibration sensor and configured to implement the acquisition step (101) of the method (100); - A processing module configured to implement the steps of constructing (102) a measurement vector and a transition matrix, estimating (103) a possible fault vibration signature, determining (104) a distance and issuing (105) an alert.

9. A computer program product comprising instructions which, when executed on a computer, cause the computer to carry out the steps of the method (100) according to any one of claims 1 to 7.

10. A computer-readable recording medium comprising instructions which, when executed by a computer, cause the computer to carry out the steps of the method (100) according to any one of claims 1 to 7.

Images