This presentation relates to the field of component monitoring and failure detection. This presentation relates more specifically to a data acquisition method for detecting damage to a bearing. Such a method finds its applications on rotating machines, for example turbomachines used on aircraft.

TECHNOLOGICAL BACKGROUND

Monitoring the components of an aircraft turbine engine generally requires a part on board the turbine engine, which acquires signals emanating from the components to be monitored and extracts relevant data from these signals, and a part on the ground, which processes these data to carry out a diagnosis of the components concerned. To perform real-time diagnostics, the data extracted by the on-board part is directly transmitted to the ground part. Reducing the volume of data transmitted is therefore a constant concern.

In the context of the detection of damage to a bearing, reducing the volume of data transmitted nevertheless implies that the on-board part must carry out numerous pre-processing of the data. Currently, a large part of the diagnosis is therefore carried out on the on-board part. This has several drawbacks: the development costs are high to optimize the on-board part, the constraints linked to being on-board restrict the algorithmic capacity for identifying faults, consistency must be ensured at all times between the on-board part and the ground part, etc.

The patent application FR 2 952 177 Al, of the Applicant, describes a method for detecting damage to a bearing bearing.

Although this method is satisfactory, there remains room for improvement in order to overcome, at least in part, the drawbacks mentioned above.

PRESENTATION OF THE INVENTION

To this end, this presentation relates to a data acquisition method for detecting damage to a bearing, comprising the following steps:

(a) obtaining a vibration signal from the bearing over a period of time;

(b) determining spectrograms of said vibrational signal at different instants of the time period;

(c) detecting the peaks on each spectrogram;

(d) among the peaks detected, retaining the peaks which, in spectrograms corresponding to successive instants and transformed by resampling synchronous with a rotational speed of the bearing, would be present at the same frequency, with a predetermined tolerance;

(e) identifying spectral lines corresponding to the selected peaks;

(f) storing information representative of said spectral lines in a ratio usable by a bearing damage detection unit.

This method makes it possible to acquire the data, typically in flight, and to prepare them in the form of a report to be transmitted to a detection unit, typically placed on the ground.

The bearing can be a turbomachine bearing, for example a bearing such as a bearing supporting in rotation at least one rotary shaft. The bearing can support, more generally, any rotating component.

The vibration signal translates the evolution of a quantity representative in particular of the vibrations of the bearing, for example an acceleration, as a function of time. It is therefore a time signal. The vibration signal can be acquired during the process or previously recorded and recovered by the process during step (a).

The time period is a period of time during which the bearing operates, or even during which the speed of rotation of the bearing can vary.

A spectrogram represents the amplitude of a signal as a function of frequency. It is a transformation of the vibratory signal at a given moment. A spectrogram can be obtained by Fourier transformation of the vibration signal or by other transformations known to those skilled in the art, for example by wavelets or calculation of power spectral density (PSD).

The peaks detected can be local extrema, in particular local maxima and/or local minima.

Synchronous resampling is a technique known to those skilled in the art, which aims to transform a datum (a signal) to eliminate therein the influence of the speed of rotation of the bearing. This technique is for example described in patent application FR 2 952 177 A1 mentioned above.

In step (d), among the previously detected peaks, those are retained which, in spectrograms corresponding to successive instants and transformed by resampling synchronous with a rotational speed of the bearing, are present at the same frequency if such is the case, knowing that it is possible not to find peaks answering this condition. This step can include the synchronous resampling of said spectrograms, then the detection of alignments of peaks at the same frequency. Alternatively, the alignment of the peaks at the same frequency, also called peak redundancy, can be detected by other methods, for example image processing methods, which do not require the synchronous resampling to be performed explicitly. In all cases,

The tolerance used to detect this presence can be a frequency and/or amplitude tolerance. Alternatively or in addition, the tolerance can consist of accepting the absence of a peak at one or more instants in a series of spectrograms in which the peak is also mostly present.

For step (d), for example spectrograms corresponding to at least two successive instants, preferably three, four, five or more, are considered.

A spectral line, or more simply line, is a spectrogram line. The spectral lines identified in step (e) can be the lines of each of the spectrograms (resampled or not) determined in step (b), but also, alternatively or in addition, lines of an average spectrogram cumulating the information from said spectrograms. For example, if the spectrograms of step (b) are obtained by Fourier transform, the lines of step (e) can be lines of the PSD calculated from the average of the Fourier transforms. Thus, it is possible to identify a single line for each frequency concerned, independently of the instants of the time period for which the Fourier transforms have been determined. Step (e) mentions the term "rays" in the plural,

The representative information stored in the report can be a description of the spectral lines themselves, for example their frequency and/or their amplitude, or any information making it possible to describe the identified lines, within a certain precision. The precision is a function of the rate of loss of information accepted for the acquisition process.

By virtue of the proposed acquisition method, the report contains relatively little data, and is therefore easy to send to a bearing damage detection unit which would be on the ground. Moreover, the analysis of the redundancy of the peaks in the successive spectrograms makes it possible to distinguish from the noise the lines due to mechanical phenomena. Thus, the acquisition method performs an effective selection of the relevant information, without presupposing the location of the lines of interest for diagnosing damage to the bearing.

In certain embodiments, the acquisition method comprises, before storing the information representative of the spectral lines in step (f), the fact of coding a characteristic of the spectral lines with a different precision depending on whether the value of this characteristic is low or high, and including this encoded characteristic in said representative information. This difference in precision is sometimes called “multi-resolution”. The characteristic concerned may be the amplitude, the frequency or any other quantity representative of these. Thus, for example, the amplitude of the spectral lines of relatively low amplitude is encoded with a different precision from the amplitude of the spectral lines of relatively high amplitude. Typically, better accuracy can be expected for the low amplitude lines than for the high amplitude lines. A multiresolution coding can be obtained by applying to the characteristic of each line a function which expands the zones where better precision is desired and contracts the zones where one can be satisfied with a less good precision. The function can be a nonlinear transformation, for example a logarithm.

In some embodiments, the spectral lines corresponding to nominal operation of the bearing are ignored. Indeed, since the ratio aims to detect damage, it is not useful to include therein lines corresponding to nominal operation of the bearing. These lines can be ignored in several ways, for example by subtracting them from the spectrograms, by not retaining the corresponding peaks, or by not including them in the report when the lines are identified. This leads to a gain in calculation time, a gain in space in the report and possibly a simplification of the identification of the relevant lines. The nominal lines can be known beforehand, by means of real tests, calculations or simulations, and prerecorded.

In the technical field considered, reference is sometimes made to cleaned data, as opposed to complete data, to designate the data from which the spectral lines which are still present during nominal operation of the bearing have been removed. These lines could indeed mask other potential lines to be extracted.

In some embodiments, the spectral lines identified in step (e) are subtracted from the spectrograms, and steps (d) and (e) are repeated. In other words, steps (d) and (e) can be carried out iteratively by subtracting each time the identified spectral lines, to facilitate the detection of possibly less visible peaks. This repetition can be carried out until no more peaks are retained which satisfy the search conditions, that is to say the aforementioned tolerance. At least the last occurrence of steps (d) and (e), or even each occurrence, is directly or indirectly followed by step (f) so as to store information representative of the lines identified in the report.

In some embodiments, after repeating steps (d) and

(e), the remaining peaks whose amplitude is greater than a predetermined threshold are retained, and the corresponding spectral lines are identified. Indeed, after steps (d) and (e), the information remaining in the spectrograms is either noise or complex relevant information. In order to limit the loss of useful information, the remaining peaks whose amplitude is greater than a predetermined threshold are retained. At the stage

(f), information representative of the corresponding lines can be stored in the report. Thus, despite the possible complexity of the

phénomène mécanique ayant donné naissance à ces pics, ils pourront faire l'objet d'une analyse complète au sol. Par suite, dans ces modes de réalisation, le procédé prend aussi en considération les facteurs complexes d'endommagement du palier.

Dans certains modes de réalisation, l'information représentative comprend l'évolution de l'amplitude de la raie spectrale en fonction de la vitesse de rotation du palier. Cette évolution peut être comprise pour au moins une raie, pour au moins une majorité de raies ou pour chaque raie. L'évolution de l'amplitude d'une raie peut être déterminée simplement à partir des spectrogrammes correspondant aux instants successifs, et inclure cette information dans le rapport est utile pour le diagnostic d'endommagement. On comprend que, dans ces modes de réalisation, la vitesse du palier varie au cours de la période de temps considérée.

Dans certains modes de réalisation, l'information représentative comprend des groupes comportant une première information correspondant à une raie spectrale de référence et une ou plusieurs deuxièmes informations correspondant chacune à une raie spectrale secondaire par rapport à la raie spectrale de référence. Afin de diminuer encore la taille du rapport et d'optimiser ainsi son envoi, il est possible de regrouper les raies, par exemple par harmoniques, afin d'éviter d'inclure l'information de position de chaque raie individuellement. Une première raie est identifiée comme raie spectrale de référence et l'information y relative est une information d'un premier type (première information), par exemple suffisant à elle-seule à identifier la raie de référence. Les raies suivantes sont prises en compte dans le rapport grâce à une information d'un deuxième type (deuxième information), chaque deuxième information étant idéalement moins volumineuse que la première information, et définissant chacune de ces raies par rapport à la raie de référence. On dit parfois que les raies spectrales sont regroupées en peignes, en raison de la ressemblance visuelle entre un peigne et une multitude de raies.

In certain embodiments, the representative information comprises, for each spectral line already known during a previous operation of the step, only an amplitude indicator, and for each spectral line not known during a previous operation of the step, at least the frequency of the spectral line. In this way, it is possible to further reduce the size of the report, the complete characterization of the lines being able to be reconstructed on the basis of a known previous operation of the bearing and of the present report, in which each representative piece of information includes, when it is possible, only the difference in amplitude of a line between the previous operation and the present operation. L'

In certain embodiments, before being stored, the representative information is encoded by means of a pre-established dictionary before the implementation of said acquisition method. By pre-established, it is understood that the dictionary is established on the basis of prior tests or simulations. As this dictionary is known and does not depend on the execution of the method, it is not necessary to calculate it each time nor to transmit it, which further reduces the volume of data exchanged between the on-board part and the part at the floor.

This presentation also relates to a data acquisition unit for detecting damage to a bearing, in particular for a turbomachine, the acquisition unit comprising

(a) a module for obtaining a vibration signal from the bearing over a period of time;

(b) a module for determining the spectrograms of said vibration signal at different instants of the time period;

(c) a peak detection module on each spectrogram;

(d) a selection module configured to, among the peaks detected by the detection module, retain the peaks which, in spectrograms corresponding to successive instants and transformed by resampling synchronous with a rotation speed of the bearing, would be present at a same frequency, with a predetermined tolerance;

(e) a spectral line identification module corresponding to the selected peaks;

(f) a storage module, configured to store information representative of said spectral lines in a report usable by a bearing damage detection unit.

The data acquisition unit can be configured to implement all or part of the characteristics of the data acquisition method previously described.

This presentation also relates to a turbine engine, in particular an aircraft turbine engine, comprising a bearing, an acquisition unit as previously described and a communication unit configured to send the report to a remote bearing damage detection unit .

In a particular embodiment, the different steps of the data acquisition method are determined by computer program instructions.

Consequently, this presentation is also aimed at a program on an information medium, this program being capable of being implemented in an acquisition device or more generally in a computer, this program comprising instructions adapted to the implementation implementation of the steps of a data acquisition method as described above.

This program may use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code, such as in partially compiled form, or in any other desirable form.

This presentation also aims at an information medium readable by a computer or by a microprocessor, and comprising instructions of a program as mentioned above.

The information carrier can be any entity or device capable of storing the program. For example, the medium may comprise a storage medium, such as a ROM, for example a CD ROM or a microelectronic circuit ROM, or else a magnetic recording medium, for example a diskette (floppy disk) or a disk hard. On the other hand, the information medium can be a transmissible medium such as an electrical or optical signal, which can be conveyed via an electrical or optical cable, by radio or by other means. The program according to the present description can in particular be downloaded from a network of the Internet type.

BRIEF DESCRIPTION OF DRAWINGS

The invention and its advantages will be better understood on reading the following detailed description of embodiments given by way of non-limiting examples. This description refers to the accompanying drawings, in which:

[Fig. 1] Figure 1 is a diagram illustrating a data acquisition method according to one embodiment.

[Fig. 2] FIG. 2 illustrates the separation of the lines of the background spectrum according to one embodiment.

[Fig. 3] FIG. 3 illustrates the detection of peaks according to one embodiment.

[Fig. 4] Figure 4 illustrates retaining some of the detected peaks according to one embodiment.

[Fig. 5] Figure 5 is a diagram schematically representing a turbine engine comprising a data acquisition unit.

DESCRIPTION OF EMBODIMENTS

La figure 1 illustre un procédé d'acquisition de données selon un mode de réalisation. Ce procédé peut être utilisé dans le cadre de la détection de l'endommagement d'un palier. Globalement, le principe de la détection est de mettre en lumière l'endommagement d'un palier en mesurant l'influence que cet endommagement a sur les vibrations d'un élément tournant de la turbomachine, certaines fréquences pouvant être reliées de manière plus ou moins complexe à certains types d'endommagements. Le procédé de la figure 1 permet plus spécifiquement d'acquérir les données vibratoires pertinentes qui pourront, ensuite, servir à l'appréciation de l'endommagement du palier.

A cette fin, le procédé comprend une étape S10 au cours de laquelle un signal vibratoire est acquis sur une période de temps. Le signal vibratoire peut traduire les vibrations du palier ou d'un élément tournant solidaire en rotation d'une partie du palier. Le signal vibratoire peut être acquis par exemple au moyen d'un accéléromètre, d'une jauge de contrainte ou de tout capteur adapté.

Le palier peut également être équipé d'un ou plusieurs tachymètres pour mesurer la vitesse de rotation, aussi appelée régime, des éléments tournants, notamment dudit palier. Ces capteurs de vitesse et de vibration peuvent être reliés à un calculateur qui effectue l'acquisition des signaux et les transmet ensuite, de préférence numériquement, à une unité de calcul.

La suite du procédé peut être appliquée pour chaque signal vibratoire, typiquement issu de chaque accéléromètre. Sans perte de généralité, on ne détaillera par la suite que le traitement d'un signal vibratoire.

A l'étape S12, optionnelle, on effectue un rééchantillonnage synchrone du signal vibratoire. Comme les fréquences à surveiller pour détecter un endommagement du palier dépendent des différentes vitesses de rotation, il est utile, afin de simplifier son traitement, de rééchantillonner le signal vibratoire de façon synchrone aux vitesses de rotation en jeu, ce qui peut inclure les vitesses mesurées directement par les tachymètres mais aussi toute combinaison linéaire de ces vitesses : par exemple, pour un palier interarbre, la vitesse de rotation du palier sera la somme ou la différence de la vitesse des deux arbres par rapport à une partie considérée comme fixe de la turbomachine.

Le rééchantillonnage synchrone, connu en soi et décrit par exemple dans la publication FR 2 952 177 Al précitée, permet d'éliminer l'influence de la vitesse de rotation. Il peut donc être fait un rééchantillonnage synchrone du même signal vibratoire pour chaque régime de rotation, c'est-à-dire par rapport à chaque composant tournant cinématiquement indépendant, afin d'éliminer sélectivement l'influence de telle ou telle rotation.

Comme indiqué précédemment, on détermine ensuite des spectrogrammes dudit signal vibratoire à différents instants de la période de temps (étape S14). En l'occurrence, on détermine des spectrogrammes pour chaque signal vibratoire rééchantillonné. Par exemple, on peut calculer des transformées de Fourier du signal vibratoire, éventuellement au moyen de techniques connues en elles-mêmes.

A l'étape S16, optionnelle, on supprime des spectrogrammes les raies correspondant à un fonctionnement nominal du palier. En effet, ces raies pourraient complexifier la détection de certaines raies vibratoires pertinentes. En outre, il n'est pas pertinent de les transmettre pour diagnostiquer l'endommagement, puisque ces raies sont associées à un fonctionnement normal du palier non endommagé. Supprimer ces raies directement dans les spectrogrammes, avant traitement, permet d'alléger les traitements qui suivent. Toutefois, comme indiqué précédemment, ceci n'est qu'un exemple du cas plus général consistant à ignorer les raies spectrales correspondant à un fonctionnement nominal du palier.

A l'étape S18, optionnelle, une PSD du signal vibratoire est calculée. Le calcul de la PSD peut utiliser une méthode connue en soi, par exemple la méthode de Welsch qui utilise les transformées de Fourier calculées à différents instants, qui en l'occurrence ont déjà été calculées à l'étape S14 : ce sont les spectrogrammes. La PSD fournit un spectre avec un bruit réduit, ce spectre étant un spectre moyen sur la période de temps considérée.

A l'étape S20, optionnelle, pour chaque spectrogramme, on sépare les raies du fond de spectre. Les raies sont les évolutions accidentées du spectrogramme (fortes variations), tandis que le fond de spectre représente l'évolution de fond du signal (faibles variations). La détermination d'une délimitation entre les raies et le fond de spectre est à la portée de l'homme du métier. Le fait de séparer les raies du fond de spectre permet de compresser plus efficacement les données issues de chacun de ces éléments.

La figure 2 illustre cette étape au moyen de différents graphes montrant l'amplitude des vibrations en fonction de la fréquence : le graphe (a) représente le logarithme en base 10 du spectrogramme, le graphe (b) représente les raies extraites du graphe (a) et le graphe (c) représente le logarithme en base 10 du fond de spectre. Chaque graphe illustre l'amplitude du signal en fonction de la fréquence. Comme on le voit sur le graphe (c), le fond de spectre est une information à faible variabilité qui peut être approximée efficacement par quelques points le long de la courbe. Ces points peuvent être traités et transmis séparément. La suite du procédé de la figure 1 s'applique aux raies extraites illustrées sur le graphe (b).

Alternativement, il serait possible de ne pas différencier les raies du fond de spectre à ce stade, bien que cela complique la détection des pics qui a lieu à l'étape suivante.

A l'étape S22 en effet, on détecte les pics sur chaque spectrogramme. Comme indiqué précédemment, le terme « pics » est générique et désigne aussi bien les pics hauts (sommets) que les pics bas (vallées), c'est-à-dire les maxima locaux et les minima locaux. Il est bien entendu possible de détecter seulement les maxima locaux, seulement les minima locaux, ou l'un et/ou l'autre.

La figure 3 illustre cette étape. La courbe représente l'amplitude du spectrogramme en fonction de la fréquence. Les pics ou extrema locaux sont repérés, sur cette courbe, par de petits cercles, respectivement des cercles pleins pour les pics et des cercles vides pour les vallées. La détection de pics d'une courbe est, en soi, connue de l'homme du métier. Cette détection de pics permet d'extraire du spectrogramme les informations pertinentes, les pics dans un sens ou dans l'autre permettant de distinguer les données dues à l'endommagement du palier d'une part et le bruit d'autre part.

Le principe de l'étape suivante S24 est de retenir, parmi les pics détectés précédemment, ceux qui ont une régularité ou une redondance dans le temps. Pour ce faire, dans le présent mode de réalisation, on superpose les spectrogrammes correspondant à des instants successifs et transformés par rééchantillonnage synchrone comme exposé précédemment. Cette superposition est illustrée sur la figure 4, qui présente en l'occurrence onze spectrogrammes pris à onze instants successifs de haut en bas. Ces spectrogrammes sont disposés de sorte que leurs axes des abscisses, indiquant la fréquence, soit communs.

Les spectrogrammes en question peuvent être, par exemple, les transformées de Fourier calculées précédemment, ou encore des PSD intermédiaires, c'est-à-dire des PSD calculées à partir desdites transformées de Fourier sur une sous-période de la période de temps totale. Un tel regroupement permet de diminuer le nombre de spectrogrammes et donc d'alléger la charge de calcul.

Comme il ressort de la figure 4, certains pics sont présents à une même fréquence, avec une tolérance prédéterminée, sur plusieurs spectrogrammes successifs. Sur la figure 4, ces pics ont été identifiés grâce à des traits verticaux 50, 52, qui matérialisent le fait que la fréquence de ces pics, après rééchantillonnage synchrone, est sensiblement la même. Ainsi, on parle également d'alignement des pics. La tolérance prédéterminée mentionnée précédemment peut porter sur la position du pic mais également sur sa présence, des occurrences manquantes de pics pouvant être tolérées (voir par exemple le trait 50, deuxième spectrogramme en partant du bas). Les critères de tolérance peuvent être ajustés par l'homme du métier selon la précision et le taux de compression recherchés pour le procédé.

In FIG. 4, other curves 54, 56 illustrate the persistence of other peaks over time; these peaks would be present at the same frequency (“aligned”) if the spectrograms were resampled synchronously with another rotation regime. Spotting these peaks can be done explicitly by overlaying the resampled spectrograms synchronously with another regime, as detailed previously about an example rotation regime, but can also be done without explicitly performing this detection of alignment and ad hoc resampling, for example by means of image processing algorithms.

Thus, whether the synchronous resamplings are made explicitly or not, we retain those of the detected peaks which, in spectrograms corresponding to successive instants and transformed by synchronous resampling with a rotation speed of the plateau, would be present at the same frequency, with a predetermined tolerance.

Next, in step S26, for each series of peaks retained, the spectral line corresponding to these peaks is identified. The spectral line can be the line of the PSD which has the same frequency as the peaks retained. This line is characterized by its frequency, its amplitude, but also the accelerometer from which it comes and the rotation regime for which the peaks are aligned.

Optionally, in step S28, the evolution of the amplitude of the spectral line is determined as a function of time, therefore of the speed of rotation of the bearing which, in this embodiment, varies in the period of time considered. Indeed, the spectral lines are present in certain regime ranges. Consequently, depending on the evolution of the rotational speed of the bearing, the amplitude of certain lines increases (for example the lines corresponding to curve 54 in FIG. 4) while the amplitude of others decreases (for example example the lines corresponding to the curves 50, 52 in FIG. 4). Generally, the lines which evolve differently are not linked to the same phenomenon. Thus, the fact that the information representative of the lines (see below) includes the evolution of the The amplitude of the line as a function of the bearing rotation speed is useful for diagnosing bearing damage. It is noted that this evolution can be stored in a more or less precise and detailed manner, ranging for example from the simple direction of variation to quantified information. According to one example, the evolution can be obtained by regression from the spectrograms.

At step S30, the data acquired so far, namely the spectrograms, the lines, the corresponding regimes, the amplitudes, the evolutions, etc., are kept. for their further processing. The detection of other peaks and therefore potentially of other lines can be carried out by repeating steps S24 and S26, possibly accompanied by step S28, as shown schematically by the loop of FIG. 1.

During this loop, in order to facilitate the detection of other peaks, the spectral lines identified at this stage can, optionally, be subtracted from the spectrograms (step S32). For example, if one wishes to subtract a line corresponding to a vertex (high peak), it is possible to replace the vertex by an interpolation between the valleys (low peaks) directly surrounding this vertex. According to another example, each amplitude between the two valleys framing the apex of a spectrogram is replaced by interpolation with the corresponding frequencies of the spectrogram corresponding to the previous instant and of the spectrogram corresponding to the following instant.

After these steps, the information not extracted from the spectrograms is either noise or information that has not been detected due to its complex structure. Optionally, in order to limit the risk of loss of useful information, in step S34, the remaining peaks whose amplitude is greater than a predetermined threshold are retained, and the corresponding spectral lines are identified. This threshold can be either fixed, for example determined according to the desired compression rate, or dynamic, for example determined according to the quantity of data already identified as relevant (the spectral lines retained) and the maximum size desired for the report final.

It is considered, at this stage, that the relevant information has been extracted from the vibration signal. In order to reduce the size of the information to be stored in the report, one or more processes, optional and independent of each other, can be performed.

According to a first processing, the identified spectral lines can be grouped together (step S36). In particular, each group of lines can comprise a first item of information corresponding to a reference spectral line and one or more second items of information each corresponding to a secondary spectral line with respect to the reference spectral line. The second information is less voluminous than the first information but is sufficient to completely identify each secondary spectral line from the reference spectral line. For example, we start from an identified line of highest frequency, and we divide its frequency by whole numbers. If the frequencies obtained correspond to one or more other lines selected, possibly

with a certain tolerance, then all these lines can be grouped together: “combs” are thus reconstructed comprising a fundamental line, or more generally reference line, and one or more harmonics, or more generally secondary lines. The groups or combs can be ordered according to their size or according to the divisors used, in order to avoid coding, for each comb, the type of comb.

According to a second processing, at step S38, to further reduce the size of the data to be stored in the report, in this case to avoid storing each time the position (that is to say the frequency) of each spectral line, it is possible to compare the identified lines with the spectral data of a previous operation taken under the same operating conditions. Thus, the representative information comprises, for each spectral line already known during a previous operation of the step, only an amplitude indicator, for example the amplitude or the deviation in amplitude with respect to the previous operation. For each spectral line not known during a previous operation of the stage, the

Il est possible de combiner les différentes approches proposées ici et, par exemple, d'envoyer la fréquence et la position de toutes les raies (c'est-à-dire sans l'étape S38) pour un certain régime et d'envoyer uniquement un indicateur d'amplitude (c'est-à-dire avec l'étape S38) pour les autres régimes. D'un fonctionnement à l'autre, on alterne de sorte qu'un spectre de référence (sans l'étape S38) est régulièrement envoyé pour chaque régime, tout en conservant une taille acceptable sur l'ensemble des données.

According to a third processing, in step S40, a characteristic of the spectral lines is coded with a different precision depending on whether the value of this characteristic is low or high. In this case, this step is carried out by applying a logarithm, for example in logarithm to base 10. Any other logarithm would be suitable. Those skilled in the art can also choose another non-linear function which crushes the ranges requiring little precision and expands the ranges requiring better precision. Thus, the characteristic, for example here the amplitude or more generally the amplitude indicator, takes up less storage space. In addition, the drop in resolution in certain ranges favors a greater repetition of values, therefore increasing the

According to a fourth processing, at step S42, the data can be compressed using compression algorithms known per se, for example a lossless compression algorithm of the LZW type. The coding dictionary used in such algorithms can be pre-established, so that it is not useful to transmit it. Furthermore, this dictionary can be specific to the lines and different from the dictionary used to compress the information relating to the background of the spectrum.

These processing operations, whether or not they are applied, result in information which is representative of the spectral lines identified at step S26, or even at step S34, in the sense that it is possible to reconstruct said lines, at least with a certain tolerance, based on representative information.

This representative information is stored, in step S44, in a report usable by a bearing damage detection unit.

Optionally, this report can be transmitted, in step S46, to a detection unit, typically a unit on the ground which will carry out the diagnosis of damage on the basis of the report, whereas the whole of the processing described hitherto had was performed by an on-board device. Alternatively, this report could be stored awaiting a return of the aircraft to the ground, and the stored report(s) would be downloaded once the aircraft is on the ground.

FIG. 5 schematically shows a turbomachine 100 comprising a data acquisition unit 10 for detecting damage to a bearing. The data acquisition unit 10 here has the hardware architecture of a computer. It comprises in particular a processor 12, a read only memory 13, a random access memory 14, a non-volatile memory 15 and means of communication 16 with the accelerometer 11 enabling the data acquisition unit 10 to obtain the measurements carried out by the accelerometer 11. The estimation device 10 and the accelerometer 11 are for example connected by a digital data bus or a serial interface (eg USB interface (Universal Serial Bus)) or wireless known in self.

The read only memory 13 of the estimation device 10 constitutes a recording medium in accordance with the present description, readable by the processor 12 and on which is recorded a computer program in accordance with the present description, comprising instructions for the execution of the steps of a data acquisition method described above with reference to Figures 1 to 4.

This computer program defines, in an equivalent manner, functional modules of the data acquisition unit 10 able to implement the steps of the data acquisition method. Thus, in particular, this computer program defines a module 10A for obtaining a vibration signal from the bearing over a period of time; a module 10B for determining the spectrograms of said vibration signal at different instants of the time period; a module 10C for detecting the peaks on each spectrogram; a selection module 10D configured to, among the peaks detected by the detection module, retain the peaks which, in spectrograms corresponding to successive instants and transformed by resampling synchronous with a rotation speed of the bearing, would be present at the same frequency , with a predetermined tolerance; a module 10E for identifying spectral lines corresponding to the selected peaks; a storage module 10F, configured to store information representative of said spectral lines in a report usable by a bearing damage detection unit. The functions of these modules have been described in more detail with reference to the steps of the data acquisition method.

Although the present description refers to specific embodiments, modifications can be made to these examples without departing from the general scope of the invention as defined by the claims. In particular, individual features of the different illustrated/mentioned embodiments can be combined in additional embodiments. Accordingly, the description and the drawings should be considered in an illustrative rather than restrictive sense.

CLAIMS

A method of acquiring data for detecting bearing damage, comprising the steps of: (a) obtaining (S10) a vibration signal from the bearing over a period of time;

(b) determining (S14) spectrograms of said vibrational signal at different instants of the time period;

(c) detecting (S22) the peaks on each spectrogram;

(d) among the peaks detected, retaining (S24) the peaks which, in spectrograms corresponding to successive instants and transformed by resampling synchronous with a rotational speed of the bearing, would be present at the same frequency, with a predetermined tolerance;

(e) identifying (S26) spectral lines corresponding to the retained peaks;

(f) storing (S44) information representative of said spectral lines in a ratio usable by a bearing damage detection unit.

2. Acquisition method according to claim 1, comprising, before storing the information representative of the spectral lines in step (f), the fact of coding (S40) a characteristic of the spectral lines with a different precision depending on whether the value of this characteristic is low or high, and including this encoded characteristic in said representative information.

3. Acquisition method according to claim 1 or 2, in which the spectral lines corresponding to nominal operation of the bearing are ignored (S16).

4. Acquisition method according to any one of claims 1 to 3, in which the spectral lines identified in step (e) are subtracted from the spectrograms (S32), and steps (d) and (e) are repeated .

5. Acquisition method according to claim 4, in which, after the repetition of steps (d) and (e), the remaining peaks whose amplitude is greater than a predetermined threshold are retained, and the corresponding spectral lines are identified (S34).

6. Acquisition method according to any one of claims 1 to

5, in which the representative information comprises the evolution of the amplitude of the spectral line as a function of the rotational speed of the bearing (S28).

7. Acquisition method according to any one of claims 1 to

6, in which the representative information comprises groups (S36) comprising a first information corresponding to a reference spectral line and one or more second information each corresponding to a secondary spectral line with respect to the reference spectral line.

8. Acquisition method according to any one of claims 1 to

7, in which the representative information comprises, for each spectral line already known during a previous operation of the stage, only an amplitude indicator (S38), and for each spectral line not known during a previous operation of the stage , at least the frequency of said spectral line.

9. Acquisition method according to any one of claims 1 to 8, in which, before being stored, the representative information is encoded by means of a pre-established dictionary before the implementation of said acquisition method ( S42).

10. Data acquisition unit (10) for detecting damage to a bearing, in particular for a turbomachine, the acquisition unit comprising

(a) a module for obtaining (10A) a vibration signal from the bearing over a period of time;

(b) a module (10B) for determining the spectrograms of said vibration signal at different instants of the time period;

(c) a detection module (10C) of the peaks on each spectrogram; (d) a selection module (10D) configured to, among the peaks detected by the detection module, retain the peaks which, in spectrograms corresponding to successive instants and transformed by resampling synchronous with a rotational speed of the bearing, would be present at the same frequency, with a predetermined tolerance;

(e) an identification module (10E) of spectral lines corresponding to the selected peaks;

(f) a storage module (10F), configured to store information representative of said spectral lines in a report usable by a bearing damage detection unit.