• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The invention relates to a method for analyzing the operating state of a machine (M) comprising a learning step for filling a reference database (3) with one or more thresholds for one or more indicators. calculated from signals delivered by a sensor (2) associated with the machine, the learning step comprising the following operations implemented by a computer processing unit (10): - acquisition of signals characteristic of normal operation and abnormal operation of the machine; for each of the signals characteristic of normal operation, formation of at least one so-called deviation signal by implementing a mathematical operation whose attributes are the signal characteristic of normal operation and one of the characteristic signals normal or abnormal operation other than said signal characteristic of normal operation; for each of the deviation signals, calculating an indicator; determination of an indicator threshold representing a limit between normal operation and abnormal operation of the machine.
    公开号:FR3032273A1
    申请号:FR1550735
    申请日:2015-01-30
    公开日:2016-08-05
    发明作者:William Bense；Jean-Michel Boiteux；Audrey Dupont；Julien Christian Pascal Griffaton；Jerome Henri Noel Lacaille
    申请人:SNECMA SAS；
    IPC主号:
    专利说明:

    [0001] TECHNICAL FIELD The invention relates to the field of the monitoring of a machine, for example a motor such as a motor, such as a motor or a motor. aircraft engine. In particular, the invention relates to a method and a system for analyzing a machine, for example an acoustic or vibratory analysis, to detect, recognize or predict anomalies. STATE OF THE PRIOR ART A machine is a mechanical system subject to constraints that can cause wear of its components. It is therefore generally sought to monitor the state of a machine as effectively as possible in order to detect damages, to recognize these damages among a set of possible damages for the machine or to predict their appearance. If a machine, for example an aircraft engine, has the disadvantage of generating noise, it is however possible to imagine using this disadvantage to perform a diagnosis or prognosis of damage in a non-intrusive manner. Indeed, aircraft engines and rotating machines in general may have damage, some of which can be detected by the ear. US 2007/0255563 A1 discloses an aircraft turbojet engine monitoring system in which acoustic signals are recorded from the turbojet engine in operation by means of two microphones positioned under the turbojet engine nacelle. signals acquired from reference signals by means of a voice recognition algorithm. It is thus possible to identify signatures representative of the state of the turbojet engine among the acquired signals.
    [0002] Similarly, it has been envisaged to carry out such monitoring not using acoustic signals generated by the machine but vibratory signals traveling through the various components of the machine. In this regard, reference can be made to patent application WO 2011/054867 A1. Such monitoring systems exploit a database in which characteristics of the limit between signals representative of normal operation and signals representative of a signal are recorded. abnormal operation. More specifically, the database stores one or more thresholds for one or more indicators calculated from signals delivered by a sensor associated with the machine, for example by a microphone disposed within the machine. In particular, the database can be filled during a learning phase by statistically determining the thresholds of a certain number of indicators. The interrogation of the database from the indicators calculated from a signal delivered by the sensor is performed during a test phase to determine the normal or abnormal operating state of a machine tested. A difficulty in the learning phase lies in the fact that few data representative of normal operation but especially abnormal operation are available, which may be insufficient to perform statistics. This is the case, for example, when the learning is carried out by means of a test bench or a machine under development whose configuration may have to change. With a limited number of training data, in particular with a small number of abnormal data then available, the statistically estimated thresholds are not optimized, which limits the quality of the determination of the operating state in the test phase. However, it is desired to have a database of good quality to satisfy a good rate of good detection of anomalies and a reduced rate of false anomaly detection. DISCLOSURE OF THE INVENTION The invention aims to answer this problem and proposes for this purpose a method for analyzing the operating state of a machine, such as an aircraft engine, comprising a learning step for filling a reference database with one or more thresholds for one or more indicators calculated from signals delivered by one or more sensors associated with the machine, characterized in that the learning step comprises the following operations implemented implemented by a computer processing unit: - acquisition of signals characteristic of a normal operation of the machine and signals characteristic of abnormal operation of the machine; for each of the signals characteristic of normal operation, calculating at least one signal called deviation by implementing a mathematical operation whose attributes are the signal characteristic of normal operation and one of the characteristic signals normal or abnormal operation other than said signal characteristic of normal operation; for each of the deviation signals, calculating an indicator; determining, from the deviation signal indicators, a threshold of the indicator representing a limit between normal operation and abnormal operation of the machine; - recording the threshold of the indicator in the reference database. Some preferred but nonlimiting aspects of this method are as follows: it further comprises a step of testing the machine by means of a signal delivered by a sensor associated with the machine, the test step comprising the following operations: forming a test signal by carrying out said mathematical operation with attributes of the signal delivered by the sensor and a reference signal; o calculation of the test signal indicator; o comparison of the test signal indicator with the threshold of the indicator recorded in the reference database to determine the normal or abnormal operating state of the machine; the signals delivered by the sensor are transformed into frequency signals before calculating the difference signals; the signals delivered by the sensor are sampled over a measurement period during which the engine speed of the machine is variable, the signals thus sampled are synchronized as a function of the variations of the engine speed over the measurement period, and the sampled signals; synchronized are transformed into frequency signals to obtain frequency lines ordered according to the rotation speed of the shaft; the one or more indicators of a signal comprise one or more indicators from a statistical moment of the signal and the energy of the signal; the calculation of an indicator of a signal is carried out by counting the number of points of the signal subtracted from a signal characteristic of normal operation present outside an envelope of said signal characteristic of normal operation; the calculation of an indicator of a signal is carried out by counting the number of peaks among n peaks of the signal characteristic of normal operation which coincide with a peak among p peaks of the signal subtracted from said signal characteristic of normal operation; the learning step comprises a secondary phase in the course of which a signal is formed characteristic of normal operation by combining several signals characteristic of normal operation; the sensor delivers an acoustic signal or a vibratory signal. The invention also relates to a computer program product comprising code instructions for executing the steps of the method, when said program is executed on a computer. And it extends to a system for analyzing the operating state of a machine, such as an aircraft engine, comprising a module for acquiring a signal delivered by a sensor associated with the machine and a reference database in which one or more thresholds are recorded for one or more indicators calculated from signals delivered by the sensor associated with the machine, characterized in that it further comprises: a signal calculation module; deviation configured to form, for each of the signals characteristic of normal operation, at least one so-called deviation signal by implementing a mathematical operation having as attributes the signal characteristic of normal operation and one signals characteristic of normal or abnormal operation other than said signal characteristic of normal operation; an indicator calculation module configured to calculate, for each of the deviation signals, an indicator; an indicator threshold determination module configured to determine, from the deviation signal indicators, a threshold of the indicator representative of a limit between normal operation and abnormal operation of the machine and to record the threshold of the indicator in the reference database.
    [0003] BRIEF DESCRIPTION OF THE DRAWINGS Other aspects, objects, advantages and characteristics of the invention will appear better on reading the following detailed description of preferred embodiments thereof, given by way of non-limiting example, and made in reference to the accompanying drawings in which: - Figure 1 illustrates an example of hardware means implemented in a possible embodiment of the system according to the invention; FIG. 2 represents the various operations implemented in the learning step of the method according to the invention; FIGS. 3a and 3b illustrate the data available for the calculation of the indicators according to a conventional implementation and according to an implementation in accordance with the invention; FIG. 4 illustrates the spectrum of a signal characteristic of a healthy operation of the machine and two signals resulting from the subtraction from said signal characteristic of sound operation respectively of a signal representative of normal operation and of a signal signal representative of abnormal operation; FIGS. 5a and 5b illustrate the calculation of an exemplary example that can be realized in the context of the invention.
    [0004] DETAILED DESCRIPTION OF PARTICULAR EMBODIMENTS FIG. 1 illustrates an example of material means implemented in various possible embodiments of the system and method for analyzing the operating state of a machine M tested according to the invention.
    [0005] The machine tested M may be an air or land vehicle engine, for example an aircraft engine as shown schematically in FIG. 1. The invention is however not limited to such an illustrative example, but extends from in a general way to the study of any mechanical system generating noise or vibrations. At least one sensor 2 is associated with the machine M. It is for example a microphone positioned within the machine M, for example inside the nacelle of an aircraft engine, or directly on the inner face of the nacelle is on the engine. It can also be a vibration sensor, such as an accelerometer or a strain gauge, preferably positioned on the engine. A plurality of sensors is preferably used, for example about ten sensors, which makes it possible to distribute them among the various components of the machine, for example between the fan casing, the main casing and the gas ejection cone. of an aircraft engine. The sensor or sensors 2 are ideally located near the components of the machine to be monitored. However, a sensor is not necessarily disposed on the monitored component, and instead one can use a room sensor that has the advantage of allowing the monitoring of several components. In particular, a sensor can be placed near one of the most critical components, which does not prevent monitoring of other components. The system comprises a computer processing unit 10 equipped with an acquisition module 1 of at least one signal delivered by at least one sensor 2 associated with the machine M. The processing unit 10 also comprises a reference database 3 in which are recorded one or more thresholds for one or more indicators calculated from signals delivered by at least one sensor 2 associated with the machine. In test phase, the passing or not of this or these thresholds by the indicator or indicators calculated from signals corresponding to a machine tested can determine whether the machine operates normally or abnormally. In certain embodiments of the invention, the detection of anomalies is accompanied by a classification of the detected anomaly (between, for example, a class of defective turbine engines and another class of defective compressor engines). The system and method according to the invention more precisely make it possible to detect anomalies from acoustic / vibratory symptoms while being optimized to operate with little learning data to constitute the reference database.
    [0006] With reference to FIG. 2, the learning step comprises the following operations implemented by different modules 4, 5, 6 of the computer processing unit 10. The learning step comprises first of all an operation Acquiring "ACQ", by means of the acquisition module 1 previously presented, signals characteristic of normal operation of the machine and signals characteristic of abnormal operation of the machine. These signals are recorded in the reference database 3. The signals characteristic of normal operation of the machine are for example acquired during the first flights of the machine when the components thereof have the maximum of chances of to be healthy. Conversely, the signals characteristic of abnormal operation of the machine are acquired by means of a machine known, for example by expertise, that it has an anomaly. The learning step then comprises, after a possible operation of filtering and normalizing the acquired signals which will be presented subsequently, an operation "SOUS" of calculation, for each of the signals characteristic of normal operation, from minus a so-called deviation signal by carrying out a mathematical operation whose attributes are the signal characteristic of normal operation and one of the signals characteristic of normal or abnormal operation other than said characteristic signal of an operation. normal. This operation "SO" is implemented by a difference signal calculation module 4. The mathematical operation can be performed between the signal characteristic of normal operation and each of the characteristic signals of normal or abnormal operation other than said signal characteristic of normal operation. In an exemplary embodiment that will be retained in the remainder of the description, the mathematical operation is a subtraction. The "SOUS" operation then preferably consists, for each of the signals characteristic of normal operation, of forming deviation signals by subtraction from the signal characteristic of normal operation of each of the signals characteristic of normal or abnormal operation. other than said signal characteristic of normal operation. In one possible embodiment, the operation "SOUS" also comprises for each of the signals characteristic of an abnormal operation, the calculation of at least one signal called deviation by implementation of a mathematical operation having as attributes the characteristic signal of abnormal operation and one of the characteristic signals of abnormal operation other than said signal characteristic of abnormal operation. Following the operation "SO", the learning step comprises an "INDC-SOUS" operation, implemented by an indicator calculation module 5, consisting for each of the deviation signals, to calculate one or several indicators. This "INDC-SOUS" operation is followed by an operation for estimating the "CAL-TSH" damage thresholds, implemented by an indicator threshold determination module 6, consisting, for each of the signal indicators of difference, to determine a threshold of the indicator representative of a limit between normal operation and abnormal operation of the machine. The threshold makes it possible to discriminate the deviation signals formed from the signal characteristic of normal operation and from another signal characteristic of normal operation of the deviation signals formed from the signal characteristic of normal operation and a signal characteristic of abnormal operation. As examples, the "CAL-TSH" operation for estimating damage thresholds can be performed by calculating the spatial median between the first and the last indicator, or by excluding the outliers, or by resorting to to a carrier vector machine, or to a neural network, or to decision trees.
    [0007] The learning step then comprises an "ENG" operation, implemented by the indicator threshold determination module 6, consisting in recording the threshold or thresholds thus calculated in the reference database 3. FIGS. and 3b illustrate the advantage of the learning phase according to the invention in the context of an example where at the end of the acquisition operation "ACQ", five signals characteristic of normal operation Normal 1 to Normal 5 and two signals characteristic of an abnormal operation abnormal 1 and abnormal 2 are recorded in the database 3. As represented in FIG. 3a, a conventional algorithm assigns a score to each signal (via the calculation of an indicator, eg its variance) and seeks to differentiate the five healthy signal scores from the two anomalous signal scores to calculate a limit or threshold between the normal signal scores and the abnormal signal scores. In this way, during the test phase, to classify an unknown signal, the score of this unknown signal is calculated and assigned the normal / abnormal status according to its position with respect to the threshold determined during the step of learning. In the context of the invention, and as shown in FIG. 3b, pairs of signal signals are formed all identically: a first healthy signal, and a healthy or abnormal comparison signal. In an exemplary embodiment, the comparison signal is subtracted from the first sound signal to form a deviation signal. This time we assign a score to each pair of signals (deviation signal) and we try to differentiate the twenty pairs of healthy signal type subtracted from the first healthy signal of the ten pairs of abnormal signal type subtracted from the first sound signal. It should be noted that a signal is not compared to itself, hence the number of twenty pairs of sound signal signals subtracted from a first sound signal and the unused crossed bars of the table of Figure 3b.
    [0008] The invention is thus based on the idea of multiplying the number of signals from which the indicators are calculated and the thresholds are determined, using more deviation signals than the acquired signals. For example, we come to compare the two-by-two acquired signals and classify this difference rather than classify each of the acquired signals taken separately.
    [0009] By using deviation signals, obtained for example by comparing the two signals to two, we multiply the number of signals that we will seek to differentiate to determine the threshold. Using the example above, instead of determining a threshold by seeking to differentiate five healthy signals from two abnormal signals, a threshold is determined by seeking to differentiate twenty difference signals corresponding to the twenty healthy signal-type couples subtracted from a first healthy signal of ten deviation signals corresponding to the ten pairs of abnormal signal type subtracts a first healthy signal. It follows from this increase in the number of available signals a more precise determination of the threshold.
    [0010] Of course, once the learning step is over, the test of a machine can be performed. This test is performed by using indicators defined in the same manner as the indicators calculated from the deviation signals. Thus, a step of testing a machine M by means of a signal delivered by the sensor associated with the machine under test comprises the following operations: - formation, by the difference signal calculation module 4, of a signal of test developed by implementing said mathematical operation, for example a subtraction, with as attributes the signal delivered by the sensor and a reference signal previously recorded in the reference database during the learning step; calculating, by the indicator calculation module 5, one or more indicators of the test signal; - comparing, by an abnormality detection module 7, the indicator or indicators of the test signal at the corresponding threshold recorded in the reference database to determine the normal or abnormal operating state of the machine tested M.
    [0011] The reference signal is typically a signal characteristic of normal operation of the machine. It may also be a signal characteristic of an abnormal operation of the machine, for example when trying to identify a failure signature. The reference signal characteristic of normal (respectively abnormal) operation may be one of the signals characteristic of normal (respectively abnormal) operation used during the learning phase, or may be a signal developed from the one and / or the other characteristic signals of normal operation (respectively abnormal) used during the learning phase, such as for example the average of the signals characteristic of normal operation (respectively abnormal) used during of the learning phase. In the context of the invention, the signals delivered by the sensor are preferably converted into frequency signals prior to the formation of deviation signals and test signals. The comparison of the signals between them and the calculation of the indicator (s) is thus carried out in the spectral domain. FIG. 4 thus shows the spectrum S of a reference signal characteristic of a normal operation of the machine, the spectrum AS1 of the signal corresponding to the difference between the reference signal and another signal representative of an operation. normal of the machine, and the AS2 spectrum of the signal corresponding to the difference between the reference signal and another signal representative of abnormal operation of the machine. In one embodiment, the learning step and the testing step comprise a filtering operation (denoising) and normalization of the acquired signals implemented respectively before the formation of the deviation signals and before the formation of the signals. test.
    [0012] This filtering can be achieved by means of a Kalman filter or a particulate filter. Alternatively, it may take the form of a Fourier transform, a principal component analysis, a wavelet transform, or any method of normalizing vibration signals. In another embodiment, this filtering takes the form of a projection in the field of orders, which makes it easier to read the spectrograms and in particular the important energy lines. According to this embodiment, the signals delivered by the sensor are sampled over a measurement period during which the engine speed of the machine is variable, the signals thus sampled are synchronized as a function of the variations of the engine speed over the measurement period. , and the synchronized sampled signals are transformed into frequency signals to obtain frequency lines ordered according to the engine speed. Further details on this projection in the field of orders can be found in patent application WO 2011/054867 A1 cited above. It will be noted here that it consists in performing a variable-regime measurement, in sampling the signal measured at a constant frequency and then in resampling it at a frequency proportional to the speed. This requires determining an angular travel curve, for example by integrating the curve representing the evolution of the engine speed as a function of time, then projecting the signal delivered by the sensor on the angular travel curve by performing a new sampling which consists in taking regularly spaced points on the angular course curve. The description below is concerned with the processing of deviation signals and test signals for calculating one or more indicators of these signals. In general, the one or more indicators must make it possible to reduce the very dense information contained in the signals, while keeping enough characteristic information to discriminate the healthy signals from the abnormal signals. Examples of indicators of deviation signals and test signals include a statistical moment of these signals, such as variance, dissymmetry (skewness), or kurtosis, or the energy of these signals. In all cases, it is preferable that the performance of the sensors (dynamic, bandwidth, etc.) is in line with the type of indicator implemented. Another example of indicator calculation consists of counting the number of points of the signal compared to a signal characteristic of normal operation present outside an envelope of said signal characteristic of normal operation. To do this, we draw an envelope around the sound signal (for example by shifting it by a fixed distance d upwards and downwards, or by calculating its spectral envelope) and we count the number of points of the signal compared coming out of this envelope. Another example of indicator computation is to count the number of peaks among n peaks of the signal characteristic of normal operation that coincide with a peak among p peaks of the signal compared to said signal characteristic of normal operation. The highest peaks of the healthy signal and the highest peaks of the compared signal are preferably retained. Alternatively, the local minima can be selected for peaks, which makes it possible to detect not the energy maxima but the sudden drops. In practice, the peaks may be slightly shifted from one signal to the other, and the algorithm is constructed to tolerate then a small offset of the comb, that is to say, of all the peaks. Advantageously, n and p are chosen to be different so as not to construct a symmetrical indicator array and thus not to reduce the number of data available during the learning phase to determine the threshold on the indicator representative of a limit. between normal operation and abnormal operation of the machine. FIG. 5a shows the case where a signal Si is taken as the reference signal to which a signal S2 is compared, and in FIG. 5b the case where the signal S2 is taken as the reference signal to which the signal S2 is compared. . It can be seen from FIG. 5a that if the highest peaks of each of the signals S1 and S2 are compared, the number of common peaks is three. On the other hand, if one compares the highest peaks of the reference signal Si with the 6 highest peaks of the compared signal S2, then the number of peaks in common is four.
    [0013] In FIG. 5b, whether the same number of peaks as for the reference signal is taken or not for the signal compared, in both cases, there are three peaks in common on the five highest peaks. of the reference signal. By taking n and p different, different scores are obtained depending on whether S2 is compared to 51 or 51 to S2. This makes it possible to diversify the distribution of the scores and thus to improve the description of the space of the abnormal data. Another example of an indicator that can be calculated directly from acquired signals is based on the theory of extreme values. This is done by dividing the spectrum of the signals into numerous small frequency intervals (eg 400 intervals). On each interval, the extrema distribution of the amplitudes (maximum and minimum) of all the sound reference spectra is modeled using the extreme value theory. A tolerance threshold is set at the tail of each of the two distributions beyond, respectively below, from which the points of the spectrum tested will be considered abnormal. The number of abnormal points is taken as an indicator. In one embodiment of the invention, the detection of anomalies performed during a test step of a machine is accompanied by a classification of the anomaly detected (for example between a class of turbine engines defective and another class of defective compressor engines). Several mathematical methods can be used to automate this classification, including discriminant analysis (where each class of anomaly is modeled by a Gaussian law), data partitioning methods (for example, k-means) supervised learning using support vector machines (SVMs), etc. In one embodiment of the invention, there are several intelligence phases of the reference database.
    [0014] In a primary phase, such as the use of a test bench or the development of the machine, few healthy and abnormal learning data are available. The database is then filled by implementing the learning step as described above. In a secondary phase, corresponding for example to the beginning of operation of the machine, a lot of sound learning data is available, but little abnormal data. The learning step can then be modified to form a signal characteristic of normal operation by combining several distinct signals characteristic of normal operation (average or median of the sound reference signals of the learning base for example) . Such a combination is in fact more representative of the normal operation of the machine than an isolated sound signal. The test step is then similarly modified to calculate an indicator of the signal by subtracting from said combination the acquired signal of the tested machine. In a later phase, corresponding for example to a mature operation of the machine, many learning data, both healthy and abnormal are available, moreover for each type of possible anomaly. In this phase, it is possible to conventionally assign an indicator score to each signal (and no longer necessarily to signal pairs) since the data are then sufficiently numerous. The normal or abnormal data (and the type of anomaly) are classified according to the score of the indicators. The invention is of course not limited to the method as described above, and extends as understood also to a system configured to implement this method, and in particular to an analysis system of the invention. the operating state of a machine, such as an aircraft engine, comprising a signal acquisition module delivered by a sensor associated with the machine and a reference database in which one or more thresholds for one or more indicators calculated from signals delivered by the sensor associated with the machine, characterized in that it further comprises: a difference signal calculation module 4 configured to form, for each of the characteristic signals of normal operation, at least one so-called difference signal by implementing a mathematical operation whose attributes are the signal characteristic of normal operation and one of the characteristic signals of a normal or abnormal operation other than said signal characteristic of normal operation; an indicator calculation module configured to calculate, for each of the deviation signals, an indicator; an indicator threshold determination module configured to determine, from the deviation signal indicators, a threshold of the indicator representative of a limit between normal operation and abnormal operation of the machine and to record the threshold of the indicator in the reference database. And the invention also extends to a software implementation of the method, and thus to a computer program product comprising code instructions for executing the steps of the method according to the invention when said program is executed on a ordinateur.30
    权利要求:
    Claims (13)
    [0001]
    REVENDICATIONS1. A method of analyzing the operating state of a machine (M), such as an aircraft engine, comprising a learning step for filling a reference database (3) with one or more thresholds for one or more indicators calculated from signals delivered by one or more sensors (2) associated with the machine, characterized in that the learning step comprises the following operations implemented by a computer processing unit (10): - acquisition (ACQ) of signals characteristic of normal operation of the machine and signals characteristic of abnormal operation of the machine; for each of the signals characteristic of normal operation, calculating (SOUS) at least one signal called deviation by implementing a mathematical operation whose attributes are the signal characteristic of normal operation and one of signals characteristic of normal or abnormal operation other than said signal characteristic of normal operation; for each of the deviation signals, calculation of an indicator (INDC-SOUS); determination (CAL-TSH), from the deviation signal indicators, of a threshold of the indicator representative of a limit between normal operation and abnormal operation of the machine; - recording (ENG) of the threshold of the indicator in the reference database (3).
    [0002]
    2. Method according to claim 1, further comprising a step of testing the machine by means of a signal delivered by a sensor associated with the machine, the testing step comprising the following operations: - forming a signal of testing by carrying out said mathematical operation with as attributes the signal delivered by the sensor and a reference signal; - calculation of the test signal indicator - comparison of the test signal indicator with the threshold of the indicator stored in the reference database to determine the normal or abnormal operating state of the machine.
    [0003]
    3. Method according to one of claims 1 and 2, wherein the signals delivered by the sensor are converted into frequency signals prior to calculating the deviation signals.
    [0004]
    4. The method according to claim 3, wherein the signals delivered by the sensor are sampled over a measurement period during which the engine speed of the machine is variable, in which the sampled signals are synchronized as a function of the variations of the engine speed. over the measurement period, and in which the synchronized sampled signals are converted into frequency signals to obtain frequency lines ordered according to the rotational speed of the shaft.
    [0005]
    5. Method according to one of claims 1 to 4, wherein the one or more indicators of a signal comprise one or more indicators from a statistical moment of the signal and the signal energy.
    [0006]
    6. Method according to one of claims 1 to 4, wherein the calculation of an indicator of a signal is performed by counting the number of points of the signal subtracted from a signal characteristic of normal operation present outside. an envelope of said signal characteristic of normal operation.
    [0007]
    7. Method according to one of claims 1 to 4, wherein the calculation of an indicator of a signal is carried out by counting the number of peaks among n peaks of the signal characteristic of normal operation which coincide with a peak among p peaks of the signal subtracted from said characteristic signal of normal operation.
    [0008]
    8. Method according to one of claims 1 to 7, wherein the learning step comprises a secondary phase during which a characteristic signal is formed of normal operation by combining several signals characteristic of normal operation.
    [0009]
    9. Method according to one of claims 1 to 8, wherein the sensor (2) delivers an acoustic signal or a vibratory signal.
    [0010]
    10. Method according to one of claims 1 to 9, wherein the calculation (SOUS) of at least one deviation signal is performed by carrying out said mathematical operation between the signal characteristic of normal operation and each signals characteristic of normal or abnormal operation other than said signal characteristic of normal operation.
    [0011]
    11. Method according to one of claims 1 to 10, wherein said mathematical operation is a subtraction.
    [0012]
    A computer program product comprising code instructions for performing the steps of the method according to one of claims 1 to 11, when said program is executed on a computer.
    [0013]
    13. System for analyzing the operating state of a machine (M), such as an aircraft engine, comprising a module (1) for acquiring a signal delivered by a sensor associated with the machine and a reference database (5) in which are stored one or more thresholds for one or more indicators calculated from signals delivered by the sensor associated with the machine, characterized in that it further comprises: - a module for calculating deviation signals (4) configured to form, for each of the signals characteristic of normal operation, at least one so-called deviation signal by implementing a mathematical operation whose attributes are the characteristic signal of normal operation and one of the characteristic signals of normal or abnormal operation other than said signal characteristic of normal operation - an indicator calculation module (5) configured to calculate, for each of the deviation signals , an indicator ; an indicator threshold determination module (6) configured to determine, from the deviation signal indicators, a threshold of the indicator representative of a boundary between normal operation and abnormal operation of the machine and to record the threshold of the indicator in the reference database (3).
    类似技术:
    公开号 | 公开日 | 专利标题
    EP3250974B1|2021-09-08|Method, system and computer program for learning phase of an acoustic or vibratory analysis of a machine
    CA2813556C|2019-01-15|System for monitoring an engine test bed
    EP2376988B1|2013-07-10|Standardisation of data used for monitoring an aircraft engine
    EP2693176B1|2019-11-06|Method for detecting defects of a bearing by vibrational analysis
    FR2939170A1|2010-06-04|DETECTION OF ANOMALY IN AN AIRCRAFT ENGINE.
    WO2015052438A1|2015-04-16|Method, system and computer program for the acoustic analysis of a machine
    EP3732457B1|2021-09-29|Method and device for monitoring a bearing equipping a rotary device
    FR3052273A1|2017-12-08|PREDICTION OF TROUBLES IN AN AIRCRAFT
    FR3005732A1|2014-11-21|METHOD AND SYSTEM FOR VIBRATION ANALYSIS OF AN ENGINE
    FR2953289A1|2011-06-03|METHOD AND DEVICE FOR MONITORING TORSION VIBRATIONS OF A ROTARY SHAFT OF A TURBOMACHINE.
    EP3658881B1|2021-07-07|Method and device for searching for a defect capable of affecting a rotating mechanical power transmission device
    EP3215903A1|2017-09-13|Tool for validating a system for monitoring an aircraft engine
    WO2004038346A1|2004-05-06|Method and test device comprising a vibratory signal sensor
    KR20200075148A|2020-06-26|AI system and pre-conditioning method in use with noise data for detecting noise source
    FR3015670A1|2015-06-26|DEVICE FOR DETECTING FAILURE PREMIUMS OF A MECHANICAL SYSTEM
    FR3035215A1|2016-10-21|METHOD FOR DETECTION OF TURBOMACHINE ANOMALY BY AUTOMATED VIBRATION ANALYSIS
    FR3036186A1|2016-11-18|METHOD, SYSTEM AND COMPUTER PROGRAM FOR MONITORING THE WEAR OF A WORKPIECE USING A CLASSIFICATION SHAFT
    FR3084046A1|2020-01-24|METHOD FOR MONITORING THE CONDITION OF AN OVERHEAD CONTACT LINE AND ASSOCIATED PROGRAM DEVICE
    FR3096133A1|2020-11-20|Extraction of a vibro-acoustic component generated by a mechanical source in variable regime
    FR3003986A1|2014-10-03|METHOD FOR DETECTING AND TAKING INTO ACCOUNT A CHANGE ABRUPTED FROM AT LEAST ONE DEVICE MONITORING INDICATOR.
    CA3140324A1|2020-12-30|Method for acquiring data for detecting damage to a bearing
    FR3080450A1|2019-10-25|METHOD AND DEVICE FOR MONITORING A GEAR SYSTEM
    WO2019197771A1|2019-10-17|Method and device for monitoring a rotary machine
    同族专利:
    公开号 | 公开日
    RU2017130315A|2019-03-01|
    CN107209512A|2017-09-26|
    US10551830B2|2020-02-04|
    EP3250974B1|2021-09-08|
    CA2975208A1|2016-08-04|
    WO2016120566A1|2016-08-04|
    EP3250974A1|2017-12-06|
    FR3032273B1|2019-06-21|
    CN107209512B|2019-11-22|
    RU2704073C2|2019-10-23|
    RU2017130315A3|2019-07-17|
    US20180017961A1|2018-01-18|
    BR112017015786A2|2018-03-27|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    FR2947335A1|2009-06-25|2010-12-31|Mathieu Plachot|Systematic spectral monitoring method for e.g. jet engine of Boeing 737, has performing systematic statistical search of correlation between functional abnormality and frequential spectrum abnormality of any periodic phenomenon|
    JP2014059910A|2009-08-28|2014-04-03|Hitachi Ltd|Method for monitoring state of facility|FR3086059A1|2018-09-18|2020-03-20|Cartesiam|METHOD FOR MONITORING THE OPERATION OF A VIBRATION GENERATING MACHINE AND DEVICE FOR CARRYING OUT SUCH A METHOD|TW490357B|1999-05-27|2002-06-11|Sanyo Electric Co|Method and device for detecting abnormities of the cutting tool of a cutting machine|
    US20070255563A1|2006-04-28|2007-11-01|Pratt & Whitney Canada Corp.|Machine prognostics and health monitoring using speech recognition techniques|
    DE102006054603A1|2006-11-20|2008-05-21|Robert Bosch Gmbh|Diagnosing e.g. components damages detecting, method for e.g. diesel engine of motor vehicle, involves recording sound of engine using microphone, and converting sound into electrical signal for diagnosing operating condition of engine|
    US7908072B2|2007-06-26|2011-03-15|General Electric Company|Systems and methods for using a combustion dynamics tuning algorithm with a multi-can combustor|
    JP5048625B2|2008-10-09|2012-10-17|株式会社日立製作所|Anomaly detection method and system|
    FR2939170B1|2008-11-28|2010-12-31|Snecma|DETECTION OF ANOMALY IN AN AIRCRAFT ENGINE.|
    FR2941049B1|2009-01-13|2011-02-11|Snecma|METHOD AND SYSTEM FOR MONITORING VIBRATION PHENOMENA IN AN AIRCRAFT GAS TURBINE ENGINE IN OPERATION|
    JP5431235B2|2009-08-28|2014-03-05|株式会社日立製作所|Equipment condition monitoring method and apparatus|
    JP5364530B2|2009-10-09|2013-12-11|株式会社日立製作所|Equipment state monitoring method, monitoring system, and monitoring program|
    FR2952177B1|2009-11-04|2012-06-01|Snecma|METHOD FOR DETECTING DAMAGE TO AT LEAST ONE BEARING BEARING OF AN ENGINE|
    FR2971595B1|2011-02-15|2013-03-22|Snecma|MONITORING AN AIRCRAFT ENGINE TO ANTICIPATE MAINTENANCE OPERATIONS|
    US20120330577A1|2011-06-22|2012-12-27|Honeywell International Inc.|Vibration severity analysis apparatus and method for rotating machinery|
    EP2752722B1|2011-08-31|2019-11-06|Hitachi Power Solutions Co., Ltd.|Facility state monitoring method and device for same|
    JP6076751B2|2013-01-22|2017-02-08|株式会社日立製作所|Abnormality diagnosis method and apparatus|
    FR3005732B1|2013-05-17|2016-10-07|Snecma|METHOD AND SYSTEM FOR VIBRATION ANALYSIS OF AN ENGINE|
    RU2545250C2|2013-06-10|2015-03-27|Владимир Николаевич Иванов|Device for diagnostics and control of technical state of mechanisms of transport and stationary systems|
    JP6216242B2|2013-12-13|2017-10-18|株式会社日立ハイテクノロジーズ|Anomaly detection method and apparatus|EP3188038B1|2015-12-31|2020-11-04|Dassault Systèmes|Evaluation of a training set|
    US10847173B2|2018-02-13|2020-11-24|Intel Corporation|Selection between signal sources based upon calculated signal to noise ratio|
    CN108827632A|2018-03-30|2018-11-16|刘文飞|A kind of determination method of communications equipment room equipment health status|
    WO2020109071A1|2018-11-27|2020-06-04|Tetra Laval Holdings & Finance S.A.|A method for condition monitoring of a cyclically moving machine component|
    法律状态:
    2016-01-12| PLFP| Fee payment|Year of fee payment: 2 |
    2016-08-05| PLSC| Publication of the preliminary search report|Effective date: 20160805 |
    2017-01-13| PLFP| Fee payment|Year of fee payment: 3 |
    2017-11-10| CD| Change of name or company name|Owner name: SNECMA, FR Effective date: 20170713 |
    2017-12-21| PLFP| Fee payment|Year of fee payment: 4 |
    2019-12-19| PLFP| Fee payment|Year of fee payment: 6 |
    2020-12-17| PLFP| Fee payment|Year of fee payment: 7 |
    2021-12-15| PLFP| Fee payment|Year of fee payment: 8 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1550735|2015-01-30|
    FR1550735A|FR3032273B1|2015-01-30|2015-01-30|METHOD, SYSTEM AND COMPUTER PROGRAM FOR LEARNING PHASE OF ACOUSTIC OR VIBRATORY ANALYSIS OF A MACHINE|FR1550735A| FR3032273B1|2015-01-30|2015-01-30|METHOD, SYSTEM AND COMPUTER PROGRAM FOR LEARNING PHASE OF ACOUSTIC OR VIBRATORY ANALYSIS OF A MACHINE|
    EP16707846.8A| EP3250974B1|2015-01-30|2016-01-28|Method, system and computer program for learning phase of an acoustic or vibratory analysis of a machine|
    PCT/FR2016/050176| WO2016120566A1|2015-01-30|2016-01-28|Method, system and computer program for learning phase of an acoustic or vibratory analysis of a machine|
    US15/545,837| US10551830B2|2015-01-30|2016-01-28|Method, system and computer program for learning phase of an acoustic or vibratory analysis of a machine|
    CA2975208A| CA2975208A1|2015-01-30|2016-01-28|Method, system and computer program for learning phase of an acoustic or vibratory analysis of a machine|
    CN201680007820.0A| CN107209512B|2015-01-30|2016-01-28|Method, system and the computer program in the study stage of the acoustics or vibration analysis for machine|
    RU2017130315A| RU2704073C2|2015-01-30|2016-01-28|Method and system for training acoustic or vibration analysis of machine|
    BR112017015786-1A| BR112017015786A2|2015-01-30|2016-01-28|method of analyzing a machine's operating state, computer program product and system for analyzing a machine's operating state|
    [返回顶部]