• 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 present invention relates to a method for estimating the normal character or not of a value (yexec) measured by a sensor (20) of a physical parameter of an aircraft engine (1) for a value (xexec) d an operating parameter of said engine (1), characterized in that it comprises the implementation of steps of: (a) Calculation according to a regression model associated with said plurality of pairs (xi; yi) i∈ [1, n] of an estimated value (yexec) of the physical parameter; (b) Calculation of an associated residue (resexec); (c) calculating according to a variance model associated with said plurality of pairs (xi; yi) i∈ [1, n] of an estimated value (varexec) of a variance of the physical parameter; (d) Calculation of an anomaly score (Zscore) of the measured value (yexec) as a function of the residue (resexec), the estimated value (varexec) of the variance, and an average residue value (mean ) for said plurality of pairs (xi; yi) i∈ [1, n]; (e) Comparison of the anomaly score (Zscore) of the measured value (yexec) with a threshold (σ) in number of standard deviations; (f) if the confidence score (Zscore) is greater than said threshold (σ), signaling the measurement as abnormal on interface means (33).
    公开号:FR3019295A1
    申请号:FR1452650
    申请日:2014-03-27
    公开日:2015-10-02
    发明作者:Aurelie Gouby；Valerio Gerez
    申请人:SNECMA SAS；
    IPC主号:
    专利说明:

    [0001] TECHNICAL FIELD The present invention relates to the field of testing aeronautical equipment.
    [0002] More specifically, it relates to a method for estimating the normal character or not of a measured value of a physical parameter of a motor. STATE OF THE ART A test bench is a platform for measuring the performance of a test machine (typically an aircraft engine) during tests under controlled conditions in order to observe its behavior. Numerous data are acquired during such tests by sensors equipping the bench and / or the test machine, to be transferred to a dedicated database, also called test database and subsequently named database. trial. For simplicity we will talk about the sensors of the bench, including the sensors fitted to the engine as part of the tests.
    [0003] A test machine is usually either a prototype in the development phase that one wants to test (the acquired data are then used by consulting firms to improve the machines and to develop them), a finished product for which we want to check the specifications and the reliability (the acquired data are then used by the quality teams). Alternatively the test machine can be either a complete engine or a component of an engine for partial tests. However, due to defects of one or more sensors of a bench and / or the test machine, it frequently happens that acquired data have abnormal or aberrant values. The test data base is then "polluted" by these data obtained during a "failed" acquisition.
    [0004] This poses a problem for users of the database (in particular the design offices), who use all the data stored to perform performance comparisons. This data can also be used to read digital models from the engine to the bench. Moreover during the test itself it is important to quickly detect any defective sensor, as the decision to stop the test can then be made according to the gravity of the fault. Indeed, the tests are very expensive and it is important to optimize these and especially their rendering. French patent FR2965915 describes an example of real-time monitoring method of sensors of a test bench which makes it possible to signal a sensor failure, but it is incapable of detecting a drift of the quality of the measurements in itself. even. In addition, the known methods are inseparable from the test bench and do not allow to identify a posteriori abnormal measurements among the measurements stored in the test base. It would therefore be desirable to have a reliable, efficient, and reproducible way of controlling engine-related parameter measurements to easily identify an outlier in a set of measurements. PRESENTATION OF THE INVENTION According to a first aspect, the present invention proposes a method of estimating the normal character or otherwise of a value measured by a sensor of a physical parameter of an aircraft engine for a value of a parameter. of said engine, a plurality of pairs each defining a reference value of the physical parameter for a value of the operating parameter being stored in a database stored on data storage means, the method being characterized in that comprises implementing by step data processing means: (a) calculating according to a regression model associated with said plurality of pairs of an estimated value of the physical parameter for the value of the operating parameter ; (b) Calculation of an associated residue; (c) calculating according to a variance model associated with said plurality of pairs of an estimated value of a variance of the physical parameter for the value of the operating parameter; (d) calculating an anomaly score of the measured value as a function of the residue, the estimated value of the variance, and an average residue value for said plurality of pairs; (e) Comparison of the anomaly score of the measured value with a threshold in number of standard deviations; (f) if the confidence score is greater than said threshold, signaling the measurement as abnormal on interface means. The estimation of the variance as a function of the operating parameter makes it possible to dispense with a variation of the measurement uncertainty (dependent on the context of use). The constructed confidence interval (tolerance on the Z '',) around the regression model is thus more realistic than a constant variance and the detection of abnormal points all the more precise. The alarm model thus gains in efficiency.
    [0005] According to other advantageous and nonlimiting features: the method comprises a prior phase of processing said plurality of pairs of the database, comprising the implementation by data processing means of steps of: (a0) determining said regression model associated with said plurality of pairs by a regression modeling the value y of the physical parameter as a function of the value x of the operating parameter from the set {xi; yi}, elitiq, where xi; y, denotes the values of a couple stored in the database; (a1) for each of the pairs, calculating an estimated value of the physical parameter and an associated residual; (a2) calculating the average of said residues; (a3) computing on a sliding window of size w a set of residual variance values each associated with a value of the operating parameter of a pair; (a4) determining said variance model associated with said plurality of pairs by a regression modeling the var value of the residual variance as a function of the x value of the operating parameter from the set {xi; vars} where vars denotes a value i ° EL [tn-vvr of calculated residual variance and xj the value of the associated operating parameter. the prior phase comprises a step (a5) of determining, as a function of said determined variance model, of a zone of confidence around said determined regression model, and of display on the interface means of said zone of confidence; the confidence zone is defined by an upper envelope of formula f (x) + 6 x / g (x), and a lower envelope of formula f (x) ax / g (x), where f is the model of regression and g is the variance model; , - the anomaly score is obtained by the formula Z s'resexec-meanl or re = r V varexec resexec is the residual associated with the measured value of the physical parameter, 25 Var, the estimated value of the variance, and mean is the average residue value for said plurality of pairs; said physical parameter is chosen from a pressure, a temperature, a flow velocity of a fluid, and a noise level associated with the engine; said operating parameter is chosen from a speed and a fuel flow rate associated with the engine; the engine is arranged in a test bench comprising the sensor, step (e) comprising stopping the test bench if the measurement is signaled as abnormal; step (e) comprises adding the pair formed of the measured value of the physical parameter and the value of the operating parameter associated with said couples database if the measurement is not reported as abnormal. According to a second aspect, the invention relates to an equipment for estimating the normal character or otherwise of a value measured by a sensor of a physical parameter of an aircraft engine for a value of an operating parameter of said engine, comprising data processing means 15, data storage means storing in a database a plurality of pairs each defining a reference value of the physical parameter for a value of an operating parameter, and means for storing data; interface, the equipment being characterized in that the data processing means are configured to implement: a calculation module according to a regression model associated with said plurality of pairs of an estimated value of physical parameter for the value of the operating parameter; A module for calculating an associated residue; A calculation module according to a variance model associated with said plurality of pairs of an estimated value of a variance of the physical parameter for the value of the operating parameter; A module for calculating an anomaly score of the value measured as a function of the residue, the estimated value of the variance, and an average residue value for said plurality of pairs; A module for comparing the anomaly score of the measured value with a threshold in number of standard deviations; A module for transmitting an alarm signal on the interface means signaling the measurement as abnormal if the confidence score is greater than said threshold u.
    [0006] According to other advantageous and nonlimiting features: the data processing module is furthermore configured to implement: a module for determining said regression model associated with said plurality of pairs by a regression modeling the value y of the parameter; physical as a function of the x value of the operating parameter from the set {xi; yi}, elitiq, where xi; y, denotes the values of a couple stored in the database; A calculation module for each of the pairs of an estimated value of the associated physical parameter and a residual; A module for calculating the average of said residues; A calculation module on a sliding window of size w of a set of residual variance values each associated with a value of the operating parameter of a pair; A module for determining said variance model associated with said plurality of pairs by a regression modeling the var value of the residual variance as a function of the value x of the operating parameter from the set {xi; vara>} where var jE [[1, n-vvr denotes a calculated residual variance value and xj the value of the associated operating parameter.
    [0007] According to a third aspect, the invention relates to a system comprising: a test bench comprising a sensor and adapted to receive an aircraft engine; - Equipment according to the second aspect of the invention for estimating the normal character or not of a value measured by the sensor of a physical parameter of said aircraft engine for a value of an operating parameter of said engine. According to a fourth and a fifth aspect, the invention relates to a computer program product comprising code instructions for the execution of a method according to the first aspect of the invention for estimating the normal character or not of a a value measured by a sensor of a physical parameter of an aircraft engine for a value of an operating parameter of said engine; and computer-readable storage means on which a computer program product comprises code instructions for executing a method according to the first aspect of the invention for estimating the normal or non-normal character of a value measured by a sensor of a physical parameter of an aircraft engine for a value of an operating parameter of said engine.
    [0008] PRESENTATION OF THE FIGURES Other features and advantages of the present invention will appear on reading the description which follows of a preferred embodiment. This description will be given with reference to the accompanying drawings, in which: FIG. 1 represents an example of an environment in which the method according to the invention is implemented; FIGS. 2a-2b illustrate the steps of two phases of an example of the method according to the invention; FIGS. 3a-3f represent displays on a data interface obtained during the various steps of the method according to the invention.
    [0009] DETAILED DESCRIPTION With reference to FIG. 1, the present method is a method of estimating the normal character or otherwise of a value v exec measured by a sensor 20 of a physical parameter of an aircraft engine 1 for a value xexec of an operating parameter of said engine 1. The engine 1 is typically all or part of a turbomachine, in particular double flow. The present method can be applied to any measurement monitoring (including during the life of the engine), but preferably, one will take the example of test measurements: the engine 1 is placed in a bench test 2 to which the sensor 20 is connected. The test bench 2 is designed to simulate a real-world operation of the engine 1. The objective of the present invention is to validate or not a measurement made during a test. Thus a normal character of a measure is directly related to its validity. A measure declared invalid will be considered abnormal. The operating parameter (x value) is an "explanatory" or "predictive" variable, as opposed to the measured physical parameter (y value) which is an "explain" or "predict" variable. In other words, the value x of the operating parameter is the cause, when the value y of the physical parameter is the consequence. More precisely, the operating parameter is a controlled value associated with the engine 1 on which either a user can act or the environment influences. In other words, it's a command. In the remainder of the present description, the example of the engine speed (that is to say the number of rotations made by a rotor of the engine 1 per unit time) will be taken, but it will be understood that many other parameters of operation, such as the fuel flow injected into the engine 1, a temperature of the fuel injected into the engine 1, an ambient pressure around the engine 1, and an ambient temperature around the engine 1, can be monitored. This parameter is an input parameter chosen for the motor.
    [0010] The physical parameter is representative of a physical quantity characteristic of an expected behavior of the engine 1 in response to the application of the operating parameter, the physical quantity at which the sensor 20 is adapted. It is understood that a plurality of sensors 20 adapted for different physical quantities can be provided. In particular, this physical parameter can be a pressure at a point of the engine 1, an internal temperature at a point of the engine 1, the speed of flow of a fluid at a point of the engine 1, a noise level generated by the engine. engine 1, a fuel density in the engine 1, etc. Those skilled in the art will choose the type of physical quantity to be measured according to the objectives of the test. In the remainder of the present description, the example will be taken of a pressure measured by a pressure sensor 20. A pair (x; y) designates a point acquisition: for a value x of the operating parameter, a value y of The physical parameter is measured by the sensor 20. A plurality of pairs (xi; yi), Elitnii each defining a reference value y, of the physical parameter for a value x, of the operating parameter are stored in a data base itself. The term "reference" values means that they are acquired during stabilized test phases, i.e. they can be considered as normal. Any abnormal values have already been removed from the database. The data storage means 32 (which is typically a hard disk) and data processing means 31 (for example a processor) are either those of a piece of equipment 3 connected to the sensor 20 (as shown in FIG. 1). The equipment 3 (or other equipment) is equipped with means of interfaces 33 (such as a keyboard and a screen) for interacting with the database, and in particular for the display of results (see below).
    [0011] The pair (x, x ', Yexec) denotes the "monitored" measure, that is to say the one for which one seeks to know if the measured Yexec value is normal or not. This monitored measurement can be both a measurement obtained in real time (in particular in an operation during the life of the engine) or a measurement in deferred time (couple (xexec; Yexec) already stored in the database and placed in waiting). Learning Phase The present method comprises two phases. The first is a learning phase and the second is an execution phase. Preferably, the learning phase is implemented beforehand so as to create the models that will be described later (and if necessary store them on the data storage means 32), and the execution phase is then implemented at each new acquisition of a measure. It is the execution phase which allows the estimation of the normal character or not of a measured value v exec, object of the invention. The learning phase can be restarted from time to time to update the 20 models. Alternatively, it is quite possible not to do prior learning and determine models in real time at each implementation of the execution phase. The two phases will be described in the following description. The learning phase can be viewed as a set of processing steps of only the data of the database (ie, independently of the torque (xexec; v 11 exec ,, - Referring to FIG. 2a (which uses the example of 30 speed / pressure parameters), the learning phase begins with a step (a0) of determining a regression model associated with said plurality of pairs (xi; yi) ie [[1, n1 by modeling regression y in function of x on the set of values of the pairs (xi, yi) i, litnp This regression model will be used in the execution phase Regression refers to a set of statistical methods well known to those skilled in the art to analyze the relation of a variable (here y) with respect to one or more others (here x) The step (a0) consists of other terms to determine a function f making it possible to better approximate the values yi as a function values xi, for a given type of link. nonne, polynomial, exponential, logarithmic, etc. The choice of the type of link used is advantageously made according to the shape of the curve and can be done automatically by optimization by maximizing a coefficient of determination, for example as described in the patent application FR2939928.
    [0012] Figure 3a illustrates the cloud of points formed by the n couples {xi; Yi} iE [[1, n1 and the model obtained by regression. In a second step (a1), the learning phase comprises the computation, for each of the pairs, of an estimated value 9i of the physical parameter and of an associated residue resi. The residue is the difference between an estimated value and a measured value. From the regression model, we obtain these values simply by the formulas 9i = f (xi), and resi = Yi - i, for E Q1, rd. Figure 3b illustrates the model residues obtained. In a step (a2), the average of said residues is calculated: mean = Once the regression model has been created, the data processing means 31 determine another model used in the execution phase: it is the model of variance. This model also associated with said plurality of pairs (xi; Yi) iEQ1, n1 is typically calculated in two consecutive steps. In a step (a3) is computed a set of values (vari) ie [[1, n-w + 1]] of residual variance, set of values which will allow the implementation of a regression in the step ( a4) to obtain the variance model. In a known way, the variance is calculated as the average of the residuals squared. Obtaining a plurality of variance values is done through a "sliding window" of size w (w is a predetermined parameter of the algorithm, it is preferably chosen to be of sufficient size to avoid a problem of squeezing the tube of confidence, see below, for example, 10% of the number of samples can be taken). More precisely, each value vars of variance is computed on a subset (indices jaj + w-1) of the set of pairs (xi; v biE [[1, n1] - In particular, vars = = 7-1 (resi) 2. Similar to what is done in step (a0), the variance model associated with said plurality of pairs (xi; yi) ielitnii is determined in step (a4) by regression modeling var in function In other words, in this step, the data processing means 31 determine the function g such that vari = g (xj), for j E 111, n-w + Figure 3c illustrates the scatterplot formed by the n - w + 1 couples {xi; vari} jEL [1, n-vv + 1]] and the model obtained by regression.
    [0013] In an optional step (a5), the learning phase may include defining a "confidence tube" (i.e., a confidence zone) around the regression model. This tube consists of an upper envelope and a lower envelope. Between these two envelopes, a point associated with a pair (x; y) is considered normal, and not outside, as we will see later. The confidence tube is defined by an amplitude difference la x .jgr I compared to the regression model, where a threshold has a standard deviation number (see below). As explained before, this tube defines a variance-dependent confidence interval, which increases the realism of the results: the number of false negatives and false positives is substantially reduced.
    [0014] The equation of the upper envelope is given by the formula f (x) + 6 x / g (x), and that of the lower envelope by the formula f (x) - 6 x I g (x). The confidence tube is preferably added to a representation of the cloud of points formed by the n pairs (xi; v iE [[1, n1 and of the regression model (of the type of FIG. 3a) so as to give the Figure 3d As will be seen later, this tube makes it possible to anticipate the result of the execution phase and to illustrate in a very visual way (and understandable even by the uninitiated) whether a measurement is normal or not. It should be noted that the distribution of the points (xi; v iE [[1, n1 is not equal according to the parameter x.In the example where the operating parameter is the regime, there is indeed a lot more measurement at high speed at low speed.This unequal distribution can cause a flare of the tube of confidence at its origin, as shown in the example of Figure 3e, if the parameter w is poorly chosen. w is chosen high enough for the function g to be increasing. steps (a3) then (a4) can be iteratively repeated with increasing values of w until this condition is verified. Execution phase As explained above, the learning phase represents a preparatory work to accelerate the execution phase (which corresponds to the core of the present method according to the invention). The learning phase can alternatively be performed "at the same time" as the execution phase. In the description of this part, reference will be made to all the associated formulas described above.
    [0015] This phase makes it possible to estimate the normal character or not of a measured value v exec of the physical parameter for an xexec value of the operating parameter. This phase is illustrated in Figure 2b. If the learning phase has been implemented previously, the models can be loaded from the data storage means 32, as illustrated in this figure. In a step (a), an estimated value V 'exec of the physical parameter is calculated for the value xexec of the operating parameter as a function of the regression model associated with said plurality of pairs (xi; Y i) iE [[1, n1 (model determined if necessary in the learning phase). This calculation is done as previously explained by the formula V 'exec = f (xexec) In the following step (b), the associated Resexec residue is computed: resexec - Yexec Yexec In step (c) (which can be done before one and / or the other of steps (a) and (b)), an estimated value varexec of the variance of the physical parameter is calculated for the value xexec of the operating parameter as a function of the variance model associated with said a plurality of pairs (xi; Y i) iE [[1, n1 (model determined, if necessary, in the learning phase). This calculation is done as previously explained by the formula varexec = 9 (xexec). The residual, the estimated value varexec of the variance, and the average residual value mean (calculated if necessary during the learning phase) for said plurality of pairs (xi; Yi) iEQtnii enable the data processing means of calculating in a step (d) a Zscore anomaly score of the measured value v exec- Preferably, the Iresexec-meanl confidence score is given by the formula Z score = r. The higher this score is y varexec, the more likely the measurement is to be abnormal.
    [0016] In a step (e), the Zscore anomaly score of the measured value Y exec is compared with the threshold a mentioned above (threshold expressed in number of standard deviations, for example three to six standard deviations). If the Zscore confidence score is greater than said threshold 6, the measurement is reported as abnormal on the interface 33 in a step (f). An alarm can be triggered if it is a real-time test on a test bench 2, and the latter is stopped (the current test is not valid and the anomaly must be analyzed before the test bench can be reused). The couple (x exec; Y exec) is not added to the database (or deleted if it was waiting for verification). Otherwise, the measurement is considered normal, and the pair (xexec; vexec) joins the reference values of the database. The learning phase can possibly be restarted to update the models.
    [0017] It should be noted that an abnormal measurement will be represented outside the confidence tube mentioned above, as for example shown in Figure 3f. In particular, if Zscore score> 6, then Ires exec (- Y exec 9exec) - meani> 6 x varexec (= the half-diameter of the pipe of confidence), which means that the associated point is out of the pipe of confidence. Indeed, the residues follow a Normal law of mean mean. The test (comparison with the Zscore) consists in seeing if it is likely that the current observation is coming from this same law. If we trace the distribution of a normal distribution, we can see that beyond six standard deviations, we are at the tail of the distribution. The probability of observing something normal beyond this standard deviation threshold is of the order of 10-9. Equipment and system The equipment 3 (shown in FIG. 1) for the implementation of the method just described (estimation of the normal character or otherwise of a vexec value measured by a sensor 20 of a physical parameter d an aircraft engine 1 for an xexec value of an operating parameter of said engine 1) comprises data processing means 31, data storage means 32, and interface means 33.
    [0018] The data storage means 32 store in a database the plurality of pairs (xi) each defining a reference value yi of the physical parameter for a value xi of an operating parameter The data processing means 31 are configured to implement: a computation module according to a regression model associated with said plurality of pairs (xi; v of an estimated value biEL [tni dexec of the physical parameter for the xexec value of the operating parameter; A module for calculating an associated resexec residue; a calculation module according to a variance model associated with said plurality of pairs (xi; vi) iE [[1, n1 of an estimated value varexec of a variance of the physical parameter for the xexec value of the operating parameter - a module for calculating a Zscore anomaly score of the measured value vexec according to the resexec residue, the estimated Varexec value of the variance, and a val average residual value for said plurality of pairs (xi; v biEL [tni; A module for comparing the Zscore anomaly score of the measured value v exec with a threshold 6 in number of standard deviations; - A module for transmitting an alarm signal on the interface means 33 signaling the measurement as abnormal if the Zscore confidence score is greater than said threshold u.
    [0019] If the equipment 3 also implements the learning phase, then the data processing module 31 is further configured to implement: a module for determining said regression model associated with said plurality of pairs (xi; v by a regression modeling r if iE [[1, n1 y as a function of x from the set {xi; y1E111,7,11 (this module is optionally configured to store the regression model on the means of data storage 32 for future use); - a calculation module for each of the pairs (xi; v iE [[1 of a, n1 estimated value 9i of the physical parameter and of a residue resi associated; - a calculation module of the average of said residual residues - A computation module on a sliding window of size w of a set of values (vari) jelitn_wil of residual variance - A module for determining said variance model associated with said plurality of pairs (xi v by a model regression r if iE [[1, n1 var according to x from the set {xi; . (If this is the case, the module is optionally configured to store the variance model on the data storage means 32 for future use). The equipment 3 is registered as explained in a preferred manner in a system further comprising a test bench 2 comprising the sensor 20 measuring the value v exec of the physical parameter and adapted to receive the engine 1. Computer program product According to fourth and fifth aspects, the invention relates to a computer program product comprising code instructions for the execution (on data processing means 31, in particular those of the equipment 3) of a method. according to the first aspect of the invention for estimating the normal character or not of a value vd exec measured by a sensor 20 of a physical parameter of an aircraft engine 1 for an xexec value of an operating parameter said engine 1, as well as storage means readable by a computer equipment (for example the data storage means 32 of this equipment 3) on which this computer program product is found.
    权利要求:
    Claims (14)
    [0001]
    REVENDICATIONS1. Method for estimating whether or not a lvexec 1 value is measured by a sensor (20) of a physical parameter of an aircraft engine (1) for a value (xexec) of an operating parameter of said motor (1), a plurality of pairs (xi; v iE [[1, n1 each defining a reference value (yi) of the physical parameter for a value (xi) of the operating parameter being stored in a database stored on data storage means (32), the method being characterized in that it comprises the implementation by data processing means (31) of steps of: (a) calculation according to a data model; regression associated with said plurality of pairs (xi; v iE [[1, n1 of an estimated value (9exec) of the physical parameter for the value (xexec) of the operating parameter; (b) Calculation of a residual (resexec) (c) Calculation according to a variance model associated with said plurality of pairs (xi; v iE [[1, n1 of a an estimated value (varexec) of a variance of the physical parameter for the value (xexec) of the operating parameter; (d) Calculation of an anomaly score (Zscore) of the measured value lv 1 in exec - function of the residue (resexec), the estimated value (varexec) of the variance, and a mean value (mean ) of residue for said plurality of pairs (xi; v iE [[1, n1; (e) Comparison of the abnormal score (Zscore) of the measured value (Yexec) with a threshold (a) in number of standard deviations (f) if the confidence score (Zscore) is greater than said threshold (a), signaling the measurement as abnormal on interface means (33).
    [0002]
    2. Method according to claim 1, comprising a prior phase of processing said plurality of pairs (xi; v iE [[1, n1 of the database, including the implementation by data processing means (31) of steps of: (a0) determining said regression model associated with said plurality of pairs (xi; Y) iE [[1, n1 by a regression modeling the value y of the physical parameter as a function of the value x of the operating parameter from the set {xi; yi} ielitnp where xi; yi denotes the values of a couple stored in the database; (al) for each of the pairs (xi; yi) i, litnp calculation of a value estimated (9i) of the physical parameter and of an associated residual (resi); (a2) calculation of the mean (mean) of said residuals (resi); (a3) computation on a sliding window of size w of a set of values (vari) JEL [tn-w + 11 residual variance each associated with a value (xj) of the operating parameter of a couple (xj; j) jE [[1, n-w + 11 (a4) determining said variance model associated with said plurality of pairs (xi; yi) ie [[1, n1 by a regression modeling the var value of the residual variance as a function of the x value of the operating parameter from the set {xi; var}, where var-, jE [[1, n-vv + 1]] designates a calculated residual variance value and xj the value of the associated operating parameter.
    [0003]
    3. The method according to claim 2, wherein the prior phase comprises a step (a5) of determining according to said determined variance model of a confidence zone around said determined regression model, and displaying on the means of interface (33) of said zone of confidence.
    [0004]
    The method according to claim 3, wherein the confidence zone is defined by an upper envelope of formula f (x) + x / g (x), and a lower envelope of formula f (x) - 6 x / g ( x), where f is the regression model and g is the variance model. Using the physical parameter, varex 'the estimated value of the variance, and me an is the average residue value for said plurality of pairs (xi; v iE [[1, n1 -
    [0005]
    5. Method according to one of claims 1 to 4, wherein the anomaly score (Zscore) is obtained by the formula Zscore = Iresexec-mean where resexec is the residue associated with the measured value of
    [0006]
    The method of one of claims 1 to 5, wherein said physical parameter is selected from associated pressure, internal temperature, fluid flow velocity, noise level, and fuel density. to the engine (1).
    [0007]
    The method according to one of claims 1 to 6, wherein said operating parameter is selected from a speed, fuel flow, fuel temperature, ambient pressure, and ambient temperature associated with the engine (1). .
    [0008]
    8. Method according to one of claims 1 to 7, wherein the motor (1) is disposed in a test bench (2) comprising the sensor (20), step (e) comprising stopping the bench test (2) if measurement 25 is reported as abnormal.
    [0009]
    9. Method according to one of claims 1 to 8, wherein step (e) comprises adding the pair (x, x 'Yexec) formed of the measured value of the physical parameter and the value of the parameter 30. operation associated with said couples database (xi; v iE [[1, n1 if the measurement is not reported as abnormal.
    [0010]
    10. Equipment (3) for estimating whether or not a value is normal (vexec, 1 measured by a sensor (20) of a parameter, physics of an aircraft engine (1) for a value (xexec ) an operating parameter of said engine (1), comprising data processing means (31), data storage means (32) storing in a database a plurality of pairs (xi; yi) ielitnii defining each a reference value yi of the physical parameter for a value xi of an operating parameter, and interface means (33), the equipment (3) being characterized in that the data processing means (31) are configured to implement: - a computation module according to a regression model associated with said plurality of pairs (xi; v of an estimated value biEL [tni dexec of the physical parameter for the xexec value of the operating parameter - A module for calculating a residue (resexec) associated - A module of calculating according to a variance model associated with said plurality of pairs (xi; vi) iE [[1, n1 of an estimated value (varexec) of a variance of the physical parameter for the value (xexec) of the operating parameter; - A module for calculating an anomaly score (Zscore) of the measured value (vexec) according to the residue (resexec), the estimated value (varexec) of the variance, and a mean value (mean) residue for said plurality of pairs (xi; vi) iE [[1, n1]; A module for comparing the anomalous score (Zscore) of the measured value (v) with a threshold (a) in number of standard deviations; A module for transmitting an alarm signal on the interface means (33) signaling the measurement as abnormal if the confidence score (Zscore) is greater than said threshold (a).
    [0011]
    11. Equipment according to claim 10, in which the data processing module (31) is further configured to implement: a module for determining said regression model associated with said plurality of pairs (xi; v by a regression); modeling r if iE [[1, n1 the value y of the physical parameter as a function of the value x of the operating parameter from the set {xi; Yi} iE111,1 / 11 where Xi; yi denotes the values of a pair stored in the database; a calculation module for each of the pairs (xi; v iE [[1 of a, n1 estimated value (9i) of the physical parameter and of a residue (resi) associated therewith; A module for calculating the mean (mean) of said residues (resi); - a computation module on a sliding window of size w of a set of values (Vari) JEL [1, n-w + 11 of residual variance, each associated with a value (x1) of the operating parameter of a pair (x1; vj) jElltn-w + 11; - a module for determining said variance model associated with said plurality of pairs (xi; v by a regression modeling r if iE [[1, n1 var as a function of x from the set {xi; vara>} jEL [1, n-vv + 11, where vars denotes a computed residual variance value and xj the value of the associated operating parameter.
    [0012]
    12. System comprising: - a test bench (2) comprising a sensor (20) and adapted to receive an engine (1) aircraft; - Equipment (3) according to one of claims 10 and 11 for estimating the normal character or not of a value lvexec 1 measured by the sensor (20) of a physical parameter of said engine (1) aircraft for a value (xexec) of an operating parameter of said motor (1).
    [0013]
    13. Computer program product comprising code instructions for the execution of a method according to one of claims 1 to 9 for estimating the normal character or not of a Yexec value measured by a sensor (20) d a physical parameter of an aircraft engine (1) for an xexec value of an operating parameter of said engine (1).
    [0014]
    14. A storage medium readable by a computer equipment on which a computer program product comprises code instructions for the execution of a method according to one of claims 1 to 9 for estimating the normal character or not of a Yexec value measured by a sensor (20) of a physical parameter of an aircraft engine (1) for an xexec value of an operating parameter of said engine (1).
    类似技术:
    公开号 | 公开日 | 专利标题
    EP3123139B1|2018-05-23|Method for estimating the normal or abnormal character of a measured value of a physical parameter of an aircraft motor
    EP2376989B1|2019-10-16|Identification of defects in an aircraft engine
    EP2368161B1|2016-11-09|Detection of anomalies in an aircraft engine
    EP3126659B1|2019-12-11|Method and device for monitoring a parameter of a rocket engine
    EP2376988B1|2013-07-10|Standardisation of data used for monitoring an aircraft engine
    FR2965915A1|2012-04-13|SYSTEM FOR MONITORING AN AIRCRAFT ENGINE TEST BENCH
    EP3215903B1|2019-12-18|Tool for validating a system for monitoring an aircraft engine
    FR2957170A1|2011-09-09|Equipment monitoring system designing tool for engine of aircraft, involves processing unit evaluating quantification of monitoring system based on quality value of result corresponding to output quality value associated with output module
    FR3012636A1|2015-05-01|METHOD FOR NON-REGRESSION OF A DESIGN TOOL OF AN AIRCRAFT ENGINE MONITORING SYSTEM
    FR3025337A1|2016-03-04|SYSTEM FOR CREATING AND DEPLOYING A MODEL OF INFERENCE
    FR3035232A1|2016-10-21|SYSTEM FOR MONITORING THE HEALTH CONDITION OF AN ENGINE AND ASSOCIATED CONFIGURATION METHOD
    EP3066534B1|2021-02-17|Method and device for characterising a signal
    FR2979013A1|2013-02-15|Method for monitoring test bench of component of aircraft engine, involves calculating set of parameters of engine component from valid data, and constructing curves representing real-time monitoring of parameters
    EP3454034B1|2022-03-16|Method and device for measuring a thrust margin of a turbomachine
    FR3075359B1|2019-11-22|DETECTION AND ACCOMODATION OF INTERMITTENT OPEN CIRCUITS ON A MOTOR SENSOR OF AN AIRCRAFT TURBOMACHINE
    WO2021180726A1|2021-09-16|Method and system for the non-destructive inspection of an aeronautical component
    FR3107979A1|2021-09-10|Method for training an emission prediction model of an internal combustion engine
    FR3012882A1|2015-05-08|TECHNICAL TEST METHOD
    WO2022043629A1|2022-03-03|Method for determining a fault for a set of sensors
    FR3018546A1|2015-09-18|METHOD FOR MONITORING THE CONDITION OF AN ENGINE BY MONITORING EXHAUST GASES
    同族专利:
    公开号 | 公开日
    RU2686252C2|2019-04-24|
    WO2015145085A1|2015-10-01|
    CN106233115B|2018-02-23|
    CN106233115A|2016-12-14|
    FR3019295B1|2016-03-18|
    CA2943397A1|2015-10-01|
    US10060831B2|2018-08-28|
    EP3123139A1|2017-02-01|
    RU2016142123A3|2018-09-24|
    US20170176292A1|2017-06-22|
    EP3123139B1|2018-05-23|
    CA2943397C|2019-10-15|
    RU2016142123A|2018-04-28|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US20050049832A1|2003-08-29|2005-03-03|Dimitry Gorinevsky|Trending system and method using monotonic regression|
    FR2983528A1|2011-12-05|2013-06-07|Snecma|Method for monitoring measuring equipment of turbojet of aircraft, involves measuring position of fuel metering device of turbojet, and standardizing degradation indicator depending on position of fuel metering device|WO2017140286A1|2016-02-17|2017-08-24|MTU Aero Engines AG|Method for determining the influence of an inner test stand on a gas turbine operated in the inner test stand|JP3593886B2|1998-06-17|2004-11-24|日産自動車株式会社|Active vibration control device for vehicles|
    RU2282940C2|2001-12-03|2006-08-27|Юнайтид Текнолоджиз Копэрейшн|Complex system for controlling status of product from remote station|
    US7464533B2|2003-01-28|2008-12-16|General Electric Company|Apparatus for operating gas turbine engines|
    US7062370B2|2004-03-30|2006-06-13|Honeywell International Inc.|Model-based detection, diagnosis of turbine engine faults|
    DE102005054735B4|2005-11-17|2019-07-04|Robert Bosch Gmbh|Method and device for operating an internal combustion engine|
    US20090030752A1|2007-07-27|2009-01-29|General Electric Company|Fleet anomaly detection method|
    US20090043447A1|2007-08-07|2009-02-12|General Electric Company|Systems and Methods for Model-Based Sensor Fault Detection and Isolation|
    US20090112519A1|2007-10-31|2009-04-30|United Technologies Corporation|Foreign object/domestic object damage assessment|
    JP2010025024A|2008-07-22|2010-02-04|Toyota Motor Corp|Abnormality detector of temperature sensor|
    FR2939928B1|2008-12-15|2012-08-03|Snecma|STANDARDIZATION OF DATA USED FOR MONITORING AN AIRCRAFT ENGINE|
    DE102009002682B4|2009-04-28|2012-09-20|Airbus Operations Gmbh|Device and method for residual evaluation of a residual for detecting system errors in the system behavior of a system of an aircraft|
    FR2965915B1|2010-10-11|2013-08-02|Snecma|SYSTEM FOR MONITORING AN AIRCRAFT ENGINE TEST BENCH|
    US8219356B2|2010-12-23|2012-07-10|General Electric Company|System and method for detecting anomalies in wind turbines|CN106053087A|2016-05-12|2016-10-26|大工（青岛）新能源材料技术研究院有限公司|Method for assessing gas engine|
    CN106200621A|2016-08-17|2016-12-07|中国航空工业集团公司西安飞行自动控制研究所|A kind of flight control system normally works the acquisition methods of criterion|
    US10642852B2|2016-09-26|2020-05-05|Splunk Inc.|Storing and querying metrics data|
    US10587635B2|2017-03-31|2020-03-10|The Boeing Company|On-board networked anomaly detectionmodules|
    FR3075359B1|2017-12-18|2019-11-22|Safran Aircraft Engines|DETECTION AND ACCOMODATION OF INTERMITTENT OPEN CIRCUITS ON A MOTOR SENSOR OF AN AIRCRAFT TURBOMACHINE|
    CN108734371A|2018-02-12|2018-11-02|阿里巴巴集团控股有限公司|A kind of processing method, device and equipment for air control instruction|
    CN108563739B|2018-04-11|2021-05-04|平安科技（深圳）有限公司|Weather data acquisition method and device, computer device and readable storage medium|
    CN112378670B|2020-11-10|2021-10-15|北京航空航天大学|Rocket engine fault detection method based on improved particle filtering|
    法律状态:
    2015-03-13| PLFP| Fee payment|Year of fee payment: 2 |
    2016-02-24| PLFP| Fee payment|Year of fee payment: 3 |
    2017-03-08| PLFP| Fee payment|Year of fee payment: 4 |
    2017-11-10| CD| Change of name or company name|Owner name: SNECMA, FR Effective date: 20170713 |
    2018-02-20| PLFP| Fee payment|Year of fee payment: 5 |
    2020-02-20| PLFP| Fee payment|Year of fee payment: 7 |
    2021-02-19| PLFP| Fee payment|Year of fee payment: 8 |
    2022-02-18| PLFP| Fee payment|Year of fee payment: 9 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1452650A|FR3019295B1|2014-03-27|2014-03-27|METHOD FOR ESTIMATING THE NORMAL OR NON-MEASURED VALUE OF A PHYSICAL PARAMETER OF AN AIRCRAFT ENGINE|FR1452650A| FR3019295B1|2014-03-27|2014-03-27|METHOD FOR ESTIMATING THE NORMAL OR NON-MEASURED VALUE OF A PHYSICAL PARAMETER OF AN AIRCRAFT ENGINE|
    RU2016142123A| RU2686252C2|2014-03-27|2015-03-26|Method of estimating normal or abnormal value of measured value of physical parameter of aircraft engine|
    US15/129,279| US10060831B2|2014-03-27|2015-03-26|Method for assessing whether or not a measured value of a physical parameter of an aircraft engine is normal|
    CA2943397A| CA2943397C|2014-03-27|2015-03-26|Method for assessing whether or not a measured value of a physical parameter of an aircraft engine is normal|
    EP15717037.4A| EP3123139B1|2014-03-27|2015-03-26|Method for estimating the normal or abnormal character of a measured value of a physical parameter of an aircraft motor|
    CN201580016936.6A| CN106233115B|2014-03-27|2015-03-26|Estimate the whether normal method of measured value of the physical parameter of aircraft engine|
    PCT/FR2015/050785| WO2015145085A1|2014-03-27|2015-03-26|Method for assessing whether or not a measured value of a physical parameter of an aircraft engine is normal|
    [返回顶部]