Method of collecting data for detecting bearing damage
By collecting and processing the vibration signals of the bearings, identifying and storing the spectral line information after synchronous resampling, the problem of large amount of data in bearing damage detection in the prior art is solved, and the data volume is reduced and the diagnosis is simplified.
Patent Information
- Application Number
- CN202080043717.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2019-06-27
- Filing Date
- 2020-06-18
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2040-06-18
AI Technical Summary
In the prior art In the detection of bearing damage, airborne components need to perform a large amount of data preprocessing to reduce the amount of data transmitted, resulting in high development costs and limited fault identification capabilities. It is necessary to ensure consistency between airborne and ground components, and there is room for optimization. .
By collecting the vibration signals of the bearing, determining the spectrum diagram, detecting and retaining the peak after synchronous resampling, identifying the spectral lines, and storing their information items in the record to reduce the amount of data and distinguishing mechanical phenomena from noise, and optimizing data transmission.
It effectively reduces the amount of data transmitted, improves the collection of relevant information for bearing damage detection, reduces the computing and storage burden, and simplifies the diagnostic process.
Smart Images

Figure CN114096824B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of component monitoring and fault detection. The present invention more particularly relates to a method for collecting data for detecting bearing damage. The method is applicable to rotating machines, such as turbine engines used on aircraft. Background Art
[0002] The monitoring of aircraft turbine engine components usually requires a component on the turbine engine that collects the signals emitted by the parts to be monitored and extracts the relevant data from these signals, and a component on the ground that processes these data to achieve the diagnosis of the relevant components. In order to perform the diagnosis in real time, the data extracted by the onboard component is directly transmitted to the ground component. Therefore, reducing the amount of data transmitted has always been a concern.
[0003] As part of bearing damage detection, reducing the amount of data transmitted means that the onboard component must perform many data preprocessing operations. Currently, most of the diagnostics are performed on the onboard component. This has several disadvantages: the development cost of optimizing the onboard component is high, the limitations related to the onboard facts hinder the algorithmic capabilities of fault identification, the consistency between the onboard component and the ground component must always be ensured, etc.
[0004] The applicant's patent application FR2952177A1 describes a method for detecting bearing roller damage. Although this method is satisfactory, there is still room for improvement, which can at least partially overcome the above-mentioned disadvantages. Summary of the invention
[0005] To this end, the present invention relates to a method for collecting data for detecting bearing damage, comprising the following steps:
[0006] (a) Obtain vibration signals from the bearing over a period of time;
[0007] (b) determining a frequency spectrum of the vibration signal at different times during the time period;
[0008] (c) detecting the peak value on each spectrum graph;
[0009] (d) among the detected peaks, retaining the peaks that will appear at the same frequency and with a predetermined tolerance in the spectrogram corresponding to the consecutive moments and transformed by synchronous resampling at the bearing speed;
[0010] (e) identifying the spectral lines corresponding to the retained peaks;
[0011] (f) Storing items of information representative of said spectrum in a record which can be used by a unit for detecting bearing damage.
[0012] This method allows data to be collected, usually in flight, and prepared in the form of a record for transmission to a detection unit, usually arranged on the ground.
[0013] The bearing may be a turbine engine bearing, such as a roller, such as a roller rotatably supporting at least one rotating shaft. In general, the bearing may support any rotating component.
[0014] The vibration signal reflects the evolution over time of a quantity representative of the bearing vibrations (e.g. acceleration). It is therefore a time signal. The vibration signal may be obtained during the method or may have been previously recorded and recovered by the method during step (a).
[0015] The time period refers to the time period during which the bearing operates, and even the time period during which the bearing speed may vary.
[0016] The spectrogram represents the amplitude of a signal as a function of its frequency. It is the transformation of the vibration signal at a given moment. The spectrogram can be obtained by Fourier transformation of the vibration signal or other transformations known to those skilled in the art, such as by wavelet or calculation of the power spectral density (PSD).
[0017] The detected peaks may be local extrema, in particular local maxima and / or local minima.
[0018] Synchronous resampling is a technique known to those skilled in the art, the purpose of which is to transform data (signals) so as to eliminate the influence of the bearing rotation speed. This technique is described, for example, in the aforementioned patent application FR2952177A1.
[0019] In step (d), from the previously detected peaks, those peaks are retained that appear at the same frequency in the spectrogram corresponding to consecutive moments and transformed by synchronous resampling at the bearing speed, knowing that it is possible that no peak meeting this condition will be found. This step may comprise synchronous resampling of the spectrogram and then detecting peak alignment at the same frequency. Alternatively, peak alignment at the same frequency, also known as peak redundancy, may be detected by other methods that do not require explicit synchronous resampling, such as image processing methods. In all cases, if synchronous resampling is performed, the retained peaks are characterized in that they will be aligned, i.e. appear at the same frequency.
[0020] The tolerance used to detect such an occurrence may be a tolerance of frequency and / or amplitude.Alternatively or additionally, the tolerance may include accepting the absence of a peak at one or more times in a series of spectrograms in which the peak is mostly present.
[0021] For example, for step (d), spectrograms corresponding to at least two, preferably three, four, five or more consecutive moments are considered.
[0022] Spectral lines, or more simply lines, are spectral lines. The spectral lines identified in step (e) may be spectral lines of each spectrogram (resampled or not) determined in step (b), but may also be spectral lines of an averaged spectrogram accumulating information items from said spectrograms. For example, if the spectrogram of step (b) was obtained by Fourier transformation, the spectral lines of step (e) may be spectral lines of a PSD calculated from the average of the Fourier transformations. Thus, a unique spectral line may be identified for each relevant frequency, independently of the instant of the time period over which the Fourier transformation was determined.
[0023] Step (e) mentions the term "spectral lines" in the plural form, but it goes without saying that the expression also includes the singular.
[0024] The representative information items stored in the record may be descriptions of the spectral lines themselves, such as their frequencies and / or amplitudes, or any information item that allows the identified spectral lines to be described within a certain accuracy. The accuracy depends on the acceptable information loss rate of the acquisition method.
[0025] Thanks to the proposed acquisition method, the records contain relatively little data and can therefore be easily sent to the unit on the ground for the detection of bearing damage. Moreover, the redundant analysis of the peaks in the continuous spectrograms allows to distinguish the spectral lines due to mechanical phenomena from the noise. Thus, the acquisition method allows to effectively select the relevant information without presupposing the location of the spectral lines of interest for the diagnosis of bearing damage.
[0026] In some embodiments, the acquisition method includes, before storing the information item representing the spectral line in step (f), encoding the feature of the spectral line with different precision depending on whether the value of the feature is low or high, and including the encoded feature in the representative information item. This difference in precision is sometimes called "multi-resolution". The relevant characteristics may be amplitude, frequency or any other quantity representing the latter. Thus, for example, the amplitude of a relatively low-amplitude spectral line is encoded with a different precision than the amplitude of a relatively high-amplitude spectral line. In general, better precision can be provided for low-amplitude spectral lines than for high-amplitude spectral lines. Multi-resolution encoding can be obtained by applying a function to the characteristics of each spectral line that expands areas where higher precision is required and shrinks areas where poorer precision may be required. The function may be a nonlinear transformation, such as a logarithm.
[0027] In some embodiments, spectral lines corresponding to normal operation of the bearing are ignored. In fact, for a record of detected damage, it is useless to include spectral lines corresponding to normal operation of the bearing. These spectral lines can be ignored in a variety of ways, such as by subtracting them from the spectrogram, not retaining the corresponding peaks, or not entering them into the record when identifying the spectral lines. This results in a saving of computing time, a saving of space in the record, and a possible simplification of the identification of relevant spectral lines. The nominal spectral lines can be known in advance by actual testing, calculation or simulation, and recorded in advance.
[0028] In the technical field considered, sometimes cleaned data are considered, rather than complete data, to specify data from which the lines that are always present during normal operation of the bearing are removed. These lines can indeed mask other potential lines to be extracted.
[0029] In some embodiments, the spectral lines identified in step (e) are subtracted from the spectrogram, and steps (d) and (e) are repeated. In other words, steps (d) and (e) may be performed iteratively, each time subtracting the identified spectral lines to facilitate the detection of peaks that may be less obvious. This repetition may be performed until there are no more peaks that meet the search criteria, i.e., the above tolerance is retained. At least the last occurrence of steps (d) and (e), or even each occurrence, is followed directly or indirectly by step (f) to store an information item representing the spectral lines identified in the recording.
[0030] In some embodiments, after repeating steps (d) and (e), the remaining peaks with amplitudes greater than a predetermined threshold are retained and the corresponding spectral lines are identified. In fact, after steps (d) and (e), the remaining information items in the spectrum diagram are either noise or complex related information. In order to limit the loss of useful information, the remaining peaks with amplitudes greater than a predetermined threshold are retained. In step (f), the information items representing the corresponding spectral lines can be stored in the record. Therefore, although the mechanical phenomena that produce these peaks may be complex, they can be the subject of a complete analysis on the ground. Therefore, in these embodiments, the method also takes into account the complex factors of bearing damage.
[0031] In some embodiments, the representative information item comprises the evolution of the amplitude of the spectral line with the rotational speed of the bearing. This evolution can be understood for at least one spectral line, at least a majority of the spectral lines or each spectral line. The evolution of the amplitude of the spectral line can be determined simply from the spectrogram corresponding to successive moments, and the inclusion of this information item in the record is useful for the diagnosis of damage. It should be understood that in these embodiments, the speed of the bearing varies within the period of time considered.
[0032] In some embodiments, the representative information items include a group, the group including a first information item corresponding to a reference spectral line, and one or more second information items, each second information item corresponding to a secondary spectral line relative to the reference spectral line. In order to further reduce the size of the record and thus optimize its transmission, the spectral lines can be grouped together, for example by harmonics, to avoid including a position information item for each spectral line separately. The first spectral line is identified as a reference spectral line and the information items associated with it are information items of the first type (first information items), for example, which are sufficient in themselves to identify the reference spectral line. Due to the second type of information items (second information items), the following spectral lines are considered in the record, each second information item is ideally less than the first information item, and each of these spectral lines is defined relative to the reference spectral line. Sometimes people say that the spectral lines are grouped together in a frequency comb due to the visual similarity between the frequency comb and the plurality of spectral lines.
[0033] In some embodiments, for each spectral line known during previous operation of the bearing, the representative information item comprises only an amplitude indicator and, for each spectral line not known during previous operation of the bearing, at least the frequency of the spectral line. In this way, the size of the record can be further reduced and the complete characteristics of the spectral lines can be reconstructed based on the known previous operation of the bearing and given the record, wherein, if possible, each representative information item comprises only the amplitude deviation of the spectral line between the previous operation and the current operation. The representative information item can be a complete information item for a previously unknown spectral line.
[0034] In some embodiments, before storage, the representative information items are encoded by a codebook pre-established before the acquisition method is implemented. Pre-established means that the codebook is established on the basis of previous tests or simulations. Since the codebook is known and does not depend on the execution of the method, it is not necessary to calculate or transmit the codebook each time, which further reduces the amount of data exchanged between the onboard component and the ground component.
[0035] The present disclosure also relates to a data acquisition unit for detecting bearing damage, in particular a data acquisition unit for a turbine engine, the acquisition unit comprising:
[0036] (a) a module for obtaining a vibration signal from a bearing over a period of time;
[0037] (b) a module for determining a frequency spectrum of the vibration signal at different times during the time period;
[0038] (c) a module for detecting peaks on each spectrum graph;
[0039] (d) a selection module configured to retain, among the peaks detected by the detection module, peaks that will appear at the same frequency and have a predetermined tolerance in a spectrum diagram corresponding to consecutive moments and transformed by synchronous resampling at the bearing speed;
[0040] (e) a module for identifying spectral lines corresponding to retained peaks;
[0041] (f) a storage module configured to store items of information representative of said spectral lines in a record usable by a unit for detecting bearing damage.
[0042] The data acquisition unit may be configured to implement all or part of the features of the above data acquisition method.
[0043] The present disclosure also relates to a turbine engine, in particular an aircraft turbine engine, comprising a bearing, an acquisition unit as described above, and a communication unit configured to send the record to a remote unit for detecting bearing damage.
[0044] In certain embodiments, the various steps of the data collection method are determined by computer program instructions.
[0045] Therefore, the present disclosure also relates to a program on an information medium, capable of being implemented in a collection device or more generally in a computer, the program comprising instructions suitable for implementing the steps of the data collection method described above.
[0046] The program may use any programming language, and may be in the form of source code, object code, or an intermediate code between source code and object code, such as a partially compiled form, or in any other desired form.
[0047] The present disclosure also relates to an information medium readable by a computer or a microprocessor and comprising the above-mentioned program instructions.
[0048] The information medium may be any entity or device capable of storing a program. For example, the medium may include a storage device, such as a ROM, such as a CD ROM or a microelectronic circuit ROM, or a magnetic recording device, such as a floppy disk or a hard disk.
[0049] On the other hand, the information medium may be a transmissible medium such as an electrical or optical signal, which may be transmitted by electrical or optical cable, radio or other means.The program according to the present disclosure may in particular be downloaded from a network of the Internet type. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] The invention and its advantages will be better understood by reading the following detailed description of embodiments given by way of non-limiting example. This description refers to the accompanying drawings, in which:
[0051] Figure 1is a diagram illustrating a method for collecting data according to one embodiment;
[0052] Figure 2 shows the separation of spectral lines from spectral background according to one embodiment;
[0053] Figure 3 Detection of a peak value according to one embodiment is shown;
[0054] Figure 4 shows that some detected peaks are retained according to one embodiment;
[0055] Figure 5 is a diagram schematically showing a turbine engine including a data acquisition unit. DETAILED DESCRIPTION
[0056] Figure 1 A method for collecting data according to one embodiment is shown. The method can be used to detect bearing damage. In general, the detection principle is to highlight bearing damage by measuring its effect on the vibration of the rotating elements of the turbine engine, and certain frequencies may be more or less complexly related to certain types of damage. Figure 1 The method allows more specific acquisition of relevant vibration data, which can then be used to assess bearing damage.
[0057] To this end, the method comprises a step S10 of collecting a vibration signal over a period of time. The vibration signal can be converted into vibration of the bearing or a rotating element rotatably coupled to a part of the bearing. The vibration signal can be obtained, for example, by an accelerometer, a strain gauge or any suitable sensor.
[0058] The bearing may also be equipped with one or more tachometers for measuring the rotational speed, also called velocity, of the rotating element, in particular the bearing. These speed and vibration sensors may be connected to a computer which acquires the signals and then transmits them preferably digitally to a computing unit.
[0059] The rest of the method can be applied to each vibration signal, typically originating from each accelerometer. Without loss of generality, only the processing of the vibration signals is detailed below.
[0060] In optional step S12, the vibration signal is synchronously resampled. Since the monitoring frequency for detecting bearing damage depends on different rotational speeds, in order to simplify its processing, it is useful to synchronously resample the vibration signal at the relevant rotational speeds, which may include the speed directly measured by the tachometer, but also any linear combination of these speeds: for example, for an inter-shaft bearing, the rotational speed of the bearing will be the sum or difference of the speeds of the two shafts relative to the part of the turbine engine considered to be stationary.
[0061] Synchronous resampling, known per se and described for example in the aforementioned publication FR 2 952 177 A1, allows the effects of the rotational speed to be eliminated. Synchronous resampling of the same vibration signal can thus be performed for each rotational speed, that is to say for each kinematically independent rotating part, in order to selectively eliminate the effects of a specific rotation.
[0062] As previously described, the spectrogram of the vibration signal is then determined at each moment in the time period (step S14). In this case, the spectrogram of each resampled vibration signal is determined. For example, the Fourier transform of the vibration signal can optionally be calculated by techniques known per se.
[0063] In optional step S16, the spectral lines corresponding to normal operation of the bearing are removed from the spectrogram. In fact, these spectral lines may complicate the detection of some relevant vibration spectral lines. Moreover, it is irrelevant to transmit these spectral lines to diagnose damage, because these spectral lines are related to the normal operation of the undamaged bearing. Removing these spectral lines directly in the spectrogram before processing can alleviate the following processing operations. However, as mentioned above, this is only an example of the more general case of ignoring the spectral lines corresponding to normal operation of the bearing.
[0064] In an optional step S18, the PSD of the vibration signal is calculated. The calculation of the PSD can use methods known per se, such as the Welsch method, which uses Fourier transforms calculated at different moments, in this case already calculated in step S14: these are spectrograms. The PSD provides a noise-reduced spectrum, which is the average spectrum over the period of time considered.
[0065] In optional step S20, for each spectrogram, the spectral lines of the spectral background are separated. The spectral lines are the jagged evolution of the spectrogram (large changes), while the spectral background represents the background evolution of the signal (small changes). The determination of the boundary between the spectral lines and the spectral background is within the capabilities of those skilled in the art. Separating the spectral lines from the spectral background can more effectively compress the data generated by each of these elements.
[0066] Figure 2 This step is illustrated by different graphs showing the amplitude of the vibration as a function of frequency: graph (a) shows the logarithm (base 10) of the spectrum graph, graph (b) shows the spectrum line extracted from graph (a), and graph (c) shows the logarithm (base 10) of the spectrum background. Each graph illustrates the relationship between the amplitude of the signal and the frequency. As can be seen in graph (c), the background of the spectrum is information with low variability and can be effectively approximated by a few points on the line. These points can be processed and transmitted individually. Figure 1 The rest of the method is applicable to the extracted spectrum shown in Figure (b).
[0067] Alternatively, the spectral lines may not be distinguished from the spectral background at this stage, although this would complicate the detection of the peaks that occur in the next step.
[0068] In fact, in step S22, peak values are detected on each spectrum graph. As mentioned above, the term "peak value" refers to the peak (top) and the valley (bottom), that is, the local maximum and the local minimum. Of course, only the local maximum, only the local minimum, or both can be detected.
[0069] Figure 3 This step is shown. The line represents the amplitude of the spectrogram as a function of the frequency. On this line, local peaks or extreme values are marked by small circles, with solid circles representing peaks and hollow circles representing valleys, respectively. The detection of peaks in a line is known per se to a person skilled in the art. Such peak detection allows the extraction of relevant information items from the spectrogram, peaks in one direction or the other allowing the distinction between data due to bearing damage on the one hand and data due to noise on the other.
[0070] The principle of the next step S24 is to retain those peaks of the previously detected peaks that have regularity or redundancy over time. To this end, in this embodiment, the spectrograms corresponding to consecutive moments and transformed by synchronous resampling as described above are superimposed. This superposition Figure 4 This is illustrated in , which shows 11 spectrograms taken at 11 consecutive moments from top to bottom. The spectrograms are placed together so that their axes (representing frequency) are common.
[0071] The spectrograms in question may be, for example, previously calculated Fourier transforms, or intermediate PSDs, ie PSDs calculated from said Fourier transforms over sub-periods of the total time period. Such grouping allows reducing the number of spectrograms and thus easing the computational burden.
[0072] from Figure 4 It can be seen in Figure 1 that on several consecutive spectrograms, some peaks appear at the same frequency with a predetermined tolerance. Figure 4 In the spectrum, these peaks are identified by vertical dashed lines 50, 52, which reflects the fact that the frequencies of these peaks are substantially the same after synchronous resampling. Therefore, it is also called peak alignment. The predetermined tolerance mentioned above can be related to the position of the peak, but also to its presence, and the absence of the peak can be tolerated (see for example dashed line 50, the second spectrum from the bottom). A person skilled in the art can adjust the tolerance criteria according to the required accuracy and compression rate of the method.
[0073] exist Figure 4, other lines 54, 56 illustrate the persistence of other peaks over time; if the spectrogram is resampled synchronously with another rotational speed, these peaks will appear at the same frequency ("aligned"). The identification of these peaks can be done explicitly by superimposing the resampled spectrogram synchronously with another speed, as detailed above in the example with rotational speed, but can also be done without explicitly performing such alignment detection and temporal resampling, for example using image processing algorithms.
[0074] Therefore, whether or not synchronous resampling is explicitly performed, those detected peaks are retained that will appear at the same frequency and with a predetermined tolerance in the spectrogram corresponding to consecutive moments and transformed by synchronous resampling at the bearing speed.
[0075] Then, in step S26, for each series of retained peaks, the spectral line corresponding to these peaks is identified. The spectral line may be a spectral line of the PSD having the same frequency as the retained peaks. This spectral line is characterized by its frequency, amplitude, but also by the accelerometer from which it comes and the rotational velocity with which the peaks are aligned.
[0076] Optionally, in step S28, the evolution of the amplitude of the spectral line over time is determined, thereby determining the evolution of the bearing speed, which in this embodiment varies within the time period considered. In practice, the spectral lines exist in certain speed ranges. Therefore, depending on the evolution of the bearing speed, the amplitude of certain spectral lines increases (e.g. Figure 4 The spectral line corresponding to the middle line 54), while the amplitudes of other spectral lines decrease (e.g. Figure 4 Spectra corresponding to the dashed lines 50, 52 in the middle). In general, spectral lines evolving in different ways are not related to the same phenomenon. Therefore, the information item representing the line (see below) includes the evolution of the amplitude of the spectral line with the bearing speed, a fact that is of great interest for diagnosing bearing damage. It should be noted that this evolution can be stored more or less accurately and in detail, for example from a simple direction of change to quantitative information. According to an example, the evolution can be obtained by regression from the spectrum.
[0077] In step S30, the data acquired so far, i.e. the spectrogram, spectral lines, corresponding speeds, amplitudes, evolutions, etc., are retained for further processing. Further peaks and thus possibly further spectral lines may be detected by repeating steps S24 and S26 (optionally in combination with step S28), e.g. Figure 1 The cycle is shown.
[0078] During this cycle, in order to facilitate the detection of other peaks, the spectral lines identified at this stage can be optionally subtracted from the spectrogram (step S32). For example, if it is necessary to subtract the line corresponding to the top (high peak), the top can be replaced by interpolation between the valleys (low peaks) that directly constitute the top. According to another example, each amplitude between the two valleys that constitute the top of the spectrogram is replaced by the corresponding frequency interpolation of the spectrogram corresponding to the previous moment and the spectrogram corresponding to the next moment.
[0079] After these steps, the information items not extracted from the spectrogram are either noise or information items that have not been detected due to their complex structure. Optionally, in order to limit the risk of loss of useful information items, 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 fixed, for example determined according to the required compression rate, or dynamic, for example determined according to the amount of data identified as relevant (spectral lines retained) and the maximum size required for the final record.
[0080] At this point, the relevant information items are considered to have been extracted from the vibration signal.In order to reduce the size of the information items to be stored in the record, one or more optional and mutually independent processing operations may be performed.
[0081] According to the first processing, the identified spectral lines can be grouped together (step S36). In particular, each group of spectral lines can include a first information item corresponding to a reference spectral line, and one or more second information items, each second information item corresponding to a secondary spectral line relative to the reference spectral line. The second information items are fewer than the first information items, but sufficient to fully identify each secondary spectral line from the reference spectral line.
[0082] The starting point is, for example, an identified higher frequency line, whose frequency is divided by an integer. If the obtained frequency corresponds to one or more other retained lines, possibly with a certain tolerance, all these lines can be recombined: thus reconstructing a "frequency comb" comprising a fundamental line or more generally a reference line and one or more harmonics or more generally secondary lines. The groups or frequency combs can be ordered according to their size or according to the separator used, in order to avoid encoding the type of each frequency comb.
[0083] According to the second processing, in step S38, in order to further reduce the size of the data to be stored in the record, in this case, in order to avoid storing the position (i.e. frequency) of each spectral line each time, the identified spectral line can be compared with the spectral data of a previous operation performed under the same speed conditions. Thus, for each spectral line known during the previous operation of the bearing, the representative information item only comprises an amplitude indicator, such as an amplitude or an amplitude deviation relative to the previous operation. For each spectral line not known during the previous operation of the bearing, the representative information item may comprise at least the frequency of said spectral line and preferably also its amplitude.
[0084] It is possible to combine the different methods proposed here and, for example, send the frequencies and positions of all spectral lines for a certain speed (i.e. without step S38) and send only the amplitude indicators for other speeds (i.e. with step S38). Alternate from one operation to the next so that a reference spectrum is sent regularly for each speed (without step S38) while maintaining an acceptable size on all data.
[0085] According to the third processing, in step S40, the features of the spectral line are encoded with different precisions depending on whether the characteristic value of the spectral line is low or high. In this case, this step is performed by applying a logarithm, for example as a logarithm with a base of 10. Any other logarithm is suitable. A person skilled in the art can also choose other non-linear functions that reduce the range where low precision is required and expand the range where better precision is required. As a result, the features, such as the amplitude here or the more general amplitude indicator, take up less storage space. In addition, the reduction in resolution in certain ranges favors a greater repetition of values, thereby improving the compression efficiency, as will be described below.
[0086] According to the fourth process, in step S42, the data may be compressed using a compression algorithm known per se, such as an LZW type lossless compression algorithm. The codebook used in such an algorithm may be pre-established and therefore need not be transmitted. Furthermore, the codebook may be spectral line specific and different from the codebook used to compress information items associated with the spectral background.
[0087] These processing operations, whether they are applied or not, result in information items representative of the spectral lines identified in step S26 or even in step S34, in the sense that said spectral lines can be reconstructed at least with a certain tolerance based on the representative information items.
[0088] In step S44, the representative information item is stored in a record which can be used by a unit for detecting bearing damage.
[0089] Optionally, the record can be transmitted in step S46 to a detection unit, typically a ground unit, which will perform damage diagnosis based on the record, while all the processing described so far has been performed by the onboard equipment. Alternatively, the record can be stored before the aircraft returns to the ground, and once the aircraft is on the ground, the stored record will be unloaded.
[0090] Figure 5 A turbine engine 100 is schematically shown, comprising a data acquisition unit 10 for detecting bearing damage. 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 16 for communicating with an accelerometer 11, which allow the data acquisition unit 10 to obtain the measurements taken by the accelerometer 11. The data acquisition unit 10 and the accelerometer 11 are connected, for example, via a digital data bus known per se or a serial (for example a USB (Universal Serial Bus) interface) or wireless interface.
[0091] The read-only memory 13 of the data acquisition unit 10 constitutes a recording medium according to the present invention, which can be read by the processor 12 and on which the computer program according to the present invention is recorded, including a computer program for executing the previously referenced Figures 1 to 4 Instructions describing the steps of a method for collecting data.
[0092] The computer program defines in an equivalent manner the functional modules of a data acquisition unit 10 capable of implementing the steps of the data acquisition method. Thus, in particular, the computer program defines a module 10A for obtaining a vibration signal from a bearing over a period of time; a module 10B for determining a spectrum of the vibration signal at different moments of the period of time; a module 10C for detecting peaks on each spectrum; a selection module 10D configured to retain, among the peaks detected by the detection module, peaks that will appear at the same frequency and have a predetermined tolerance in the spectrum corresponding to consecutive moments and transformed by synchronous resampling at the bearing speed; a module 10E for identifying the spectral line corresponding to the retained peak; a storage module 10F configured to store information items representing the spectral line in a record that can be used by a unit for detecting bearing damage. The functions of these modules are described in more detail with reference to the steps of the data acquisition method.
[0093] Although this description refers to specific exemplary embodiments, these examples may be modified without departing from the general scope of the invention as defined by the claims. In particular, the various features of the different illustrated / mentioned embodiments may be combined in additional embodiments. Therefore, the description and drawings should be understood as illustrative rather than restrictive.
Claims
1. A method for collecting data for detecting bearing damage, comprising the following steps: (a) obtaining a vibration signal from a bearing over a period of time (S10); (b) determining the frequency spectra of the vibration signal at different times within the period of time (S14); (c) detecting a peak value on each spectrum graph (S22); (d) retaining, among the detected peaks, peaks that will appear at the same frequency and have a predetermined tolerance in the frequency spectrum corresponding to the consecutive moments and transformed by synchronous resampling at the bearing speed (S24); (e) identifying a spectral line corresponding to the retained peak (S26); wherein the identified spectral line is subtracted (S32) from the spectrogram; and repeating steps (d) and (e); (f) Storing items of information representative of said spectrum in a record, said record being usable by a unit for detecting bearing damage (S44).
2. The acquisition method according to claim 1 includes encoding the characteristics of the spectral line with different precisions according to whether the value of the characteristic is low or high before storing the information item representing the spectral line in step (f), and including the encoded characteristic in the information item representing the spectral line.
3. The collection method according to claim 1, wherein: The spectrum corresponding to the normal operation of the bearing is ignored (S16).
4. According to any one of claims 1 to 3, after repeating steps (d) and (e), the remaining peaks with amplitudes greater than a predetermined threshold are retained and the corresponding spectral lines are identified (S34).
5. The collection method according to any one of claims 1 to 3, wherein: The information item representing the spectrum line includes the evolution of the spectrum line amplitude with the bearing rotation speed (S28).
6. The collection method according to any one of claims 1 to 3, wherein: The information items representing the spectral lines comprise a group (S36) comprising a first information item corresponding to a reference spectral line, and one or more second information items, each second information item corresponding to a secondary spectral line relative to the reference spectral line.
7. The collection method according to any one of claims 1 to 3, wherein: For each spectral line known during previous operation of the bearing, the item of information representative of said spectral line comprises only an amplitude indicator (S38) and, for each spectral line not known during previous operation of the bearing, at least the frequency of said spectral line.
8. The collection method according to any one of claims 1 to 3, wherein: Prior to storage, the information items representative of the spectral lines are encoded by means of a codebook previously established before implementing the acquisition method (S42).
9. A data acquisition unit (10) for detecting bearing damage, the data acquisition unit comprising: (a) a module (10A) for obtaining a vibration signal from a bearing over a period of time; (b) a module (10B) for determining a frequency spectrum of the vibration signal at different moments in the period of time; (c) a module (10C) for detecting peaks on each spectrum graph; (d) a selection module (10D) configured to retain, among the peaks detected by the module for detecting peaks on each spectrum graph, peaks that will appear at the same frequency and with a predetermined tolerance in spectrum graphs corresponding to consecutive moments and transformed by synchronous resampling at the bearing speed; (e) a module (10E) for identifying the spectral line corresponding to the retained peak and subtracting the identified spectral line from the spectrogram; (f) a storage module (10F) configured to store items of information representative of said spectral lines in a record capable of being used by a unit for detecting bearing damage, after repeating the retention performed by the selection module (10D) and the identification and subtraction performed by the module (10E).
