• 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:
    A method of monitoring a degradation of an onboard device of an aircraft, implemented by a computer, the degree of degradation of the onboard device being defined by an abnormality score formed by the count of occurrences of abnormalities detected by a device control system, the monitoring method comprising a step of comparing an abnormality score obtained for an observation sequence of length (n, t) given to a decision threshold (ks) and a step of issuing an alarm in case of reaching or exceeding the decision threshold (ks), the decision threshold (ks) being determined automatically for a given alarm probability Pa, corresponding to the probability that a alarm is emitted during the monitoring process while the onboard device is healthy, by means of the following steps: - a step of obtaining an abnormality score (r) on at least one corresponding reference sequence t flights of the aircraft without degradation and length (m, tc) equal to a plurality of lengths (n, t) of observation sequences; a step of adjusting a discrete probability law to find the abnormality score (r) obtained on the reference sequence; a step of calculating the decision threshold (ks) such that by applying the discrete probability law adjusted during the preceding step to an observation sequence having the given length (n), the probability that a score an anomaly greater than or equal to the decision threshold (ks) occurs either less than the given alarm probability Pa, or an elementary probability function of Pa in the case where a confirmation strategy is used.
    公开号:FR3022997A1
    申请号:FR1455921
    申请日:2014-06-25
    公开日:2016-01-01
    发明作者:Jean-Remi Andre Masse；Pierre Gerald Lalonde；Aurore Fabienne Paule Humeau；Armand Dariouche Alimardani；Guilhem Alcide Auguste Foiret
    申请人:SNECMA SAS；
    IPC主号:
    专利说明:

    [0001] GENERAL TECHNICAL FIELD AND PRIOR ART The present invention relates to the field of monitoring a degradation of an aircraft onboard device, in particular on a turbomachine. More particularly, the invention relates to devices such as, for example, a measurement chain, for which a health indicator is linked to the occurrence or not of events.
    [0002] To monitor an onboard device, it is known to form an indicator that is characteristic of a degradation of the onboard device. This indicator is known to those skilled in the art under the name of abnormality score. Typically, an abnormality score is formed from measurements of physical parameters of the onboard device such as, for example, a geometric position, a control current, an opening angle, a temperature, etc. The abnormality score is characteristic of the degree of degradation damage. Preferably, an abnormality score is formed for each flight of the aircraft. In order to determine whether the onboard device is indeed degraded, the monitoring method comprises a step of comparing an abnormality score obtained for a given flight of the aircraft with a decision threshold and a step of issuing an alarm. if the decision threshold is exceeded. Thus, by following the evolution of the abnormality score, it is detected whether the degree of degradation increases and we can anticipate the risk of failure of the onboard device and improve the management of maintenance operations.
    [0003] The setting of the decision threshold for a given degradation is crucial, given, on the one hand, that a decision threshold that is too low induces the issuance of frequent alarms while the degradation is not proven (false alarm) and on the other hand, a decision threshold that is too high inhibits the transmission of alarms while the degradation is proven (non-detection).
    [0004] In a conventional manner, for each degradation that it is desired to detect, the value of the decision threshold is defined empirically. In order to ensure maximum security, the value of the decision thresholds is generally under evaluated to minimize the risk of non-detection. As a result, the number of false alarms remains high, which presents a drawback for the airlines that are forced to frequently implement a maintenance operation while the aircraft aeronautical device is not degraded. To eliminate this inconvenience, airlines set specifications for manufacturers of on-board devices to limit the risk of error. Given the precision required, any empirical definition of a decision threshold for a fault monitoring method is then prohibited.
    [0005] The patent application FR 1254506 provides a solution to this need, and describes a method of monitoring a degradation of an onboard device of an aircraft, implemented by a computer, the degree of degradation of the onboard device being defined by an abnormality score formed from physical parameter measurements of the on-board device, the monitoring method comprising a step of comparing an abnormality score obtained for a given flight of the aircraft with a decision threshold and a step of transmitting an alarm if the decision threshold is exceeded, the decision threshold being determined automatically for a given alarm probability Pa, corresponding to the probability that an alarm will be emitted during the monitoring process then that the onboard device is healthy, by means of the following steps: a step of calculating a plurality of abnormality scores for a plurality of flights of the aircraft without eroding to obtain a distribution of the probability density of the abnormality score, the distribution being specific to the physical nature of the onboard device; a step of adjusting the distribution by a non-parametric estimator of the probability density so as to obtain a continuous adjusted distribution function; a step of calculating a continuous adjusted distribution function from the continuous adjusted distribution function; and a step of reading the antecedent of the continuous adjusted distribution function for a given value (1-Pa in the previous application), the antecedent corresponding to the decision threshold.
    [0006] The patent application FR 1358593 provides an improvement to the previous invention by introducing a confirmation strategy "k among n", for which an alarm is issued only if there are at least k threshold overruns in n successive flights .
    [0007] This method makes it possible to determine the decision threshold automatically with a high accuracy according to the alarm probability Pa corresponding to the requirements of the airlines. Such a method is reliable compared to the methods of the prior art. This makes it possible to accurately detect any degradation of an onboard device and to anticipate any failure of the latter during the monitoring.
    [0008] However, such a method is generally not applicable when the distribution of the anomaly score is not continuous but discrete. By way of example, the patent application filed in France on September 15, 2001 under the number 2980266, of the company SNECMA, discloses a system for monitoring a measurement chain of a turbojet engine for which the d Abnormality is defined by counting in successive time increments the number of "unwanted" transitions of the health word "OK" to another health word indicating a malfunction. Sometimes this count is almost always zero. In the example presented in Figure 1, out of a total of 750 undegraded flights, one flight had a degradation and another eighteen, all other flights had no adverse transition. It is then no longer appropriate to adjust a continuous distribution to the histogram obtained in FIG. 1 and to deduce a threshold for the count, by adapting the preceding methods.
    [0009] There is thus a need to determine a strategy for monitoring a degradation of an aircraft onboard device from a count of events with a low occurrence of occurrence to automatically trigger an alarm by respecting reliable and accurate a given alarm probability Pa.
    [0010] The invention proposes a simple and effective solution to this problem. GENERAL PRESENTATION OF THE INVENTION To this end, the invention proposes a method of monitoring a degradation of an onboard device of an aircraft, implemented by a computer, the degree of degradation of the onboard device being defined by a an abnormality score formed by counting occurrences of anomalies observed by a control system of the device, the monitoring method comprising a step of comparing an abnormality score obtained for an observation sequence of given length with a decision threshold ks and a step of issuing an alarm in the event of reaching or exceeding the decision threshold ks, the decision threshold ks being determined automatically for a given alarm probability Pa, corresponding to the probability that an alarm is emitted during the monitoring process while the onboard device is healthy, by means of the following steps: - a step of obtaining an abnormality score s ur at least one reference sequence corresponding to flights of the aircraft without degradation and of length equal to a plurality of lengths of observation sequences; a step of adjusting a discrete probability law making it possible to retrieve the abnormality score obtained on the reference sequence; a step of calculating the decision threshold ks such that, by applying the discrete probability law adjusted during the preceding step to an observation sequence having the given length, the probability that a higher anomaly score or equal to the decision threshold (ks) occurs either less than an elementary probability Peac of threshold exceeding obtained from the given alarm probability Pa by the formula: Peac = Bs, k-s + 1 (Pa) in which q is a number of observation sequences, p a number of times where the threshold ks has been exceeded for q successive sequences, and Ba-s.4.1 the inverse beta Eulerian distribution function of parameters S and N-S + 1, N and S may be equal to one.
    [0011] In other words, the value of the threshold ks is the number of occurrences of an anomaly during an observation sequence from which the hypothesis that a healthy device with a probability of anomaly occurrence represented by the discrete probability law that has been adjusted on the reference sequence is not tenable with a probability greater than Pa, or Peac. So, if the number of occurrences of anomalies on an observation sequence reaches the decision threshold ks, either once or p times in a strategy with confirmation, we can conclude that the device no longer follows a corresponding behavior. to a device without degradation with a risk of being wrong lower than Pa. With laws of known probabilities, it is possible to calculate this decision threshold ks by using the expression of their inverse distribution law. This method is particularly advantageous when the reliability of the device results in the fact that the detection of anomalies corresponds to the counting of rare events, when the identification of a continuous law is impossible on the histogram of the counts.
    [0012] According to a preferred aspect of the invention, the discrete probability law used represents a phenomenon whose occurrences have a given probability per time unit, the second step corresponding to an estimation of the probability per unit of time in the discrete probability law to find the abnormality score obtained on the reference sequence. Said probability law can be a binomial law or a Poisson law, parameterized by the probability per unit of time and a sequence length. In the previous case, the estimation of the probability per unit of time of occurrence of anomalies for the system without degradation in the discrete probability law can be equal to an increase of this probability per unit of time with a given confidence for the abnormality score in the reference sequence. Preferably, the estimate of the probability per unit time is equal to the maximum likelihood corresponding to the abnormality score obtained on the reference sequence. Advantageously, an observation sequence corresponds to several successive flights of the aircraft. The method according to the invention may comprise a step of predetermining the given length for the observation sequences. The greater the length of the observation sequence, the greater the detection threshold ks will be and the better will be respected the alarm probability Pa or the elementary probability Peac. On the other hand, the degradation phenomenon before failure will have to be slower to be detected in time. The choice of the length of an observation sequence can therefore be based on operational considerations of the time of deposition and the dynamic of physical degradation until failure.
    [0013] This length of the observation sequence can be advantageously chosen so that the decision threshold (ks) is between three and ten. In the example presented, the length of the observation sequence is 60000 time increments and the value of the decision threshold calculated by the method takes the value 6.
    [0014] According to a preferred aspect of the invention, the alarm probability Pa is obtained by the following formula: Pe Pd Pa - - (1p) apn 0 - Pe) (1- Pd) in which: Pe corresponds to the probability that the onboard device is healthy while an alarm is issued; Pd is the probability of degradation for an observation sequence; and (1 -, 6) a corresponds to the probability of detection of a priori degradation. Advantageously, the alarm probability Pa is determined according to a probability of error Pe which corresponds to the practical requirements of the airlines who wish to reduce the number of unnecessary maintenance steps of the on-board devices. More preferably, the method comprises: a step of implementing the method for monitoring a degradation of the on-board device, with the decision threshold (ks) previously determined, for a plurality of abnormality scores formed for a plurality of observation sequences with degradation to deduce a probability of detecting a posterior decay (1-, 6) poenon, a step of determining a new alarm probability Pa as a function of the detection probability a posteriori (1, 6) poenon degradation, a step of determining a new decision threshold (ks) refined from the new alarm probability. These steps advantageously make it possible to refine the value of the decision threshold quickly and reliably. PRESENTATION OF THE FIGURES The invention will be better understood on reading the description which will follow, given solely by way of example, and referring to the appended drawings in which: FIG. 1 represents the distribution of abnormality scores obtained for a given embedded measurement device of an aircraft for a plurality of flights without degradation, FIG. 2 is a schematic diagram of a monitoring method comprising steps of automatic determination of the decision threshold according to the invention; FIG. 3 is a schematic diagram of a variant of the monitoring method comprising steps for automatic determination of the decision threshold according to the invention.
    [0015] DESCRIPTION OF ONE OR MORE EMBODIMENTS AND IMPLEMENTATION The automatic determination of a decision threshold will be presented for a method of monitoring a degradation of an on-board device of an aircraft as known from FIG. prior art, for example, by the patent application under the number 1358593 of the company SNECMA. In a known manner, the degree of degradation of the onboard device is defined by an abnormality score obtained by counting during "clock ticks", determining regular unit increments of time, of a number of unwanted transitions of an indicator. state of the device established by the monitoring system. In the case described, the state of the device can take two values and can therefore be coded in a binary manner signaling the possible occurrence of an anomaly. Preferably, the abnormality score is formed by an onboard computer of the aircraft and connected to the onboard device by a communication link. To take into account the fact that the operation of the device may present anomalies even when it is not degraded, the method corresponds, as in the state of the art, to a strategy for emitting an alarm only with an alarm probability Pa corresponding to a false alarm proportion Pfa imposed. When transmitting an alarm, a maintenance step of the onboard device must be implemented by the airline to avoid a failure of said onboard device. Probability of alarm 30 Probability of alarm means the probability that an alarm will be emitted during the monitoring process while the onboard device is healthy, that is to say, without proven degradation. In other words, an alarm probability Pa of 1% corresponds to the emission of an alarm for 100 flights without degradation. In practice, the alarm probability Pa is not directly known. Indeed, the specifications of the airlines impose as a criterion a probability of error Pe which corresponds to the probability that the onboard device is healthy while an alarm is emitted. In other words, an error probability Pe of 1% corresponds to a healthy onboard device for 100 alarms issued. In practice, when an alarm is emitted by the monitoring method, a maintenance operation is implemented by the airline. A limitation of the probability of error Pe thus allows airlines to limit the number of maintenance operations of an on-board device that is healthy, such operations generating unnecessary operating costs. Preferably, the alarm probability Pa is determined as a function of the probability of error Pe by means of the conditional probability formula (F1) below. (F1) Pa = Pe Pd (1 P) has orlon 0 Pe) '(1-Pd) The formula (F1) has the following parameters: p corresponds to the probability of non-detection of a degradation for a given flight, 1-p then corresponding to the probability of detecting a degradation; and Pd is the probability of degradation for a given flight. The formula (F1) is deduced from the conditional probability equations in which the probability Pa corresponds to an alarm knowing the sound device (Pa = P (1Sain alarm)) and the probability Pe corresponds to a sound device knowing that an alarm is emitted (Pe = P (Healthy Alarm)).
    [0016] P (Sound Alarm) P (Sound Alarm) P (Sound) P (Sound Alarm) - P (Alarm) P (Sound)) P (Alarm) P (Sound) P (Sound) = P (Sound Alarm) - - Pa P (Alarm) P (Alarm) In other words 25P (Alarm) = Pa P (Healthy) (1 - Pd) = Pa P (Healthy Alarm) Pe In addition, as P (Alarm) = P (Alarm n Healthy) + P (Alarm n Degraded) P (Alarm) = P (Sound Alarm). P (Healthy) + P (Gradient Alarm) .P (Degraded) With P (Degraded) = Pd P (Healthy) = 1 - Pd P (Degraded Alarm) = 1- / 3 P (Healthy Alarm) = Pa P ( HealthyAlarm) = Pe P (Alarm) = P (Healthy Alarm) .P (Healthy) + P (Gradient Alarm) .P (Degraded) P (Alarm) = Pa - (1 - Pd) + (1-, 6) - Pd P (Alarm) = Pa - (1 - Pd) + (1-, 6) - Pd Thanks to the two equalities relating to P (Alarm), we then deduce (F1).
    [0017] In the formula (F1), the probability of a degradation for a given flight Pd is known by experimentation or estimation and may, for example, be of the order of 10-7. The probability of detecting a degradation (1-, q) a mon is set "a priori" to 1 and refined by iteration as will be detailed later to improve the precision of the decision threshold.
    [0018] By way of example, an alarm probability Pa of the order of 5.10E-8 is conventionally obtained for a probability of error Pe required of the order of 5%, a probability of detection of a degradation (1 - j6) to my of the order of 1 and a probability of a degradation for a flight Pd of the order of 10-6.
    [0019] Discrete Abnormality Score and Discrete Probability Law In the present method, an observation sequence length, constituted on the example of a number n of counting increments and a threshold ks on the number k of appearances, is determined. anomalies, corresponding in the example to unwanted transitions, during a sequence, to trigger the alarm. The length of an observation sequence may correspond to one flight or to several consecutive flights. The abnormality score on a sequence is defined as the number k of occurrences of anomalies. However, it is no longer determined, as in the examples cited, a threshold on the value of the abnormality score assumed continuous. The method refers to a discrete probability law for representing the occurrences of anomalies and evaluating the threshold ks from which an alarm can be triggered by respecting an alarm probability Pa. In a first exemplary embodiment, it is used a law of probability represented by a binomial distribution. The binomial distribution of parameters n and p corresponds to the fact of renewing on a sequence n times independently the pull of an event that has a probability p to occur and a probability (1-p) not to occur. We then count the number of occurrence of the event, here an anomaly, and we call X the random variable indicating this number of appearances on the sequence of n draws.
    [0020] The probability that the random variable X takes a value k between 0 and n on the sequence is then given by the binomial distribution: P (X = k) = [k. (1-p) "The distribution function of the law of the binomial Fn, p provides the probability that the random variable X takes a value less than or equal to k for a sequence of n draws .The value Fn, p (k) the distribution function of the binomial law in X = k can also be expressed using a Beta Eulerian distribution function of the probability of occurrence p by pull and of parameters k + 1 and nk. is well suited to the case described for which the counting of possible anomalies is made at regular intervals of unit duration, provided by the "top clock" of the surveillance system during a flight.
    [0021] In a second exemplary embodiment, a Poisson probability law is used. The parameter Poisson's law (λt) corresponds to the behavior of the number of anomalies occurring in an observation sequence whose length is the time t, if these anomalies occur with a mean frequency At known and independently of the time passed between two occurrences of anomalies. This frequency corresponds to a probability of occurrence per unit of time. The probability that there are k occurrences of the anomaly on a sequence of observation of duration is then given, in the case where t is a multiple of the increment of time for which the frequency A is expressed, by: (At .t) k P (X = k) = exp (-At). k! The distribution function of the Poisson law FA! provides the probability that the random variable X takes a value less than or equal to k for a sequence of length t. The value FA! (k) of distribution function of the Poisson distribution in X = k can also be expressed by a Gamma Eulerian distribution function of the frequency λ of occurrence over the duration t, having for parameters k + 1 and t .
    [0022] The Poisson's law is a passage to the limit of the binomial law when the number of prints is very large. It is therefore also suitable for the case described when the number "tops clock" is very large. It is also appropriate in the case of a device whose monitoring system would report anomalies only when they occur, that is to say at irregular intervals. In this case, the Poisson law makes it possible to perform the calculations over a time interval that is not a multiple of increments between successive counts. Example of a first embodiment of the method The method will now be described using the binomial law, the monitoring system indicating a binary anomaly occurrence result or not for regular time increments.
    [0023] With reference to FIG. 2, the method begins with a first step El of extracting a plurality of anomaly scores for a plurality of flights without degradation with the monitored device. In particular, this step El makes it possible to create at least one reference sequence consisting of a large number of flights without degradation, which corresponds to a very large number m of detection time increments of the anomaly indicator for a number of flights. device without degradation, and to constitute an abnormality score r on this reference sequence, equal to the sum of occurrences of anomalies.
    [0024] In the example corresponding to the monitoring of a device described by the patent application under number 1358593, illustrated in FIG. 1, the results were observed for 750 flights without degradation, each flight comprising 1200 transient phase increments. , corresponding to the duration during which the device is requested. This makes it possible to construct a reference sequence of length m, m being equal to 750 × 1200 = 90 000 time increments, for which an abnormality score r has been found, r being equal to 1 + 18 = 19. A second step E2 of the method, before starting the monitoring, is to seek an estimate of the probability p of occurrence of an anomaly at each time increment, using the reference sequence.
    [0025] The estimate 15 is obtained by taking an upper bound / 5 'of p with a degree of confidence a. Its value is obtained by the formulas: / 5 '= 1 - Fm, p (r) = a} = (a) Fm, p is the binomial distribution function of parameters m and p, where m is the number of increments time of the reference sequence and p a probability of occurrence, applied to the abnormality score r found for the reference sequence. which is the inverse Eulerian distribution function Beta inverse of parameters r + 1 and m-r makes it possible to directly calculate the estimate / 5 'with a degree of confidence a. For the process it is thus possible to use a = 90% or a = 50%. It is also possible to use a value a = 44% which corresponds to the maximum of likelihood. In this case, the estimate / 5 'is directly provided by the ratio r / m. With the values of the monitoring sequence given as an example in FIG. 1, of length m = 90 000 increments with an abnormality score r = 19, we obtain: / 590% = 2.88 10-5, / 550% = 2,19 10-5 and the maximum likelihood estimate, L .--) 111, = 77i P44% = 2,11,105. The method then comprises a third step E3 in which the threshold ks is determined on the number k of occurrences of anomalies in an observation sequence of length n time increments, for which it can be considered that the probability of occurrence of anomalies is greater than the estimate p of the probability of occurrence of d anomalies for a device without degradation, made in the previous step E2 from the reference sequence, with a probability of being erroneous equal to the value Pa defined by the formula F1. For this purpose, the ks threshold for detecting degradation on the number k of occurrences of anomalies during an observation sequence of n increments is determined such that ks is the smallest integer for which the distribution function of the probability law on an observation sequence, that is to say, the probability that there are k anomalies for any k less than ks on this sequence, approaches the unit unless the probability alarm given Pa.
    [0026] This results in the formula: (F2) ks = Inf {ic I1 - Fne (k - 1) <Pa} = inf {k I k, n-k + 1 (P) Pa} where Fne is the binomial distribution function of parameters, the number of increments n chosen for the observation sequence and the estimate ij of the probability of occurrence of the anomaly at each increment for a device without degradation, obtained during the preceding step E2. k, n-k + 1 is the Beta Eulerian distribution function of parameters k and n-k + 1, applied to the p estimate of the probability of occurrence. In other words, ks is the number of occurrences of an anomaly among the n time increments of the observation sequence from which the assumption that the true probability p in time increment of occurrence of anomalies is equal to l estimate p is not tenable with a probability greater than Pa. Therefore, if k is greater than or equal to ks, we can conclude that the probability p per time increment of occurrence of anomalies exceeded the estimate p for a device without degradation with a probability of being wrong less than Pa.40 The choice of the number n of increments for an observation sequence is the result of a compromise. This number must not be too small, which then leads to a value of ks that is too low and too much inaccuracy on the probability of alarm being respected. Conversely, this number should not be too large because there is a risk of allowing degradation of the monitored device to develop. The choice of the length n of the observation sequence can be made from the knowledge of the device, by simulations before implementing the method in the onboard computer. In a variant of the method, an additional step can be implemented to refine the length of the observation sequence as a function of the values of the threshold ks calculated in step E3, so, for example, that this threshold takes a value between a few units and ten. In practice, for the surveillance of the devices targeted by the aeronautical applications, the length n of increments of the observation sequence corresponds to a few successive flights, the number nv of flight being preferably between three and ten. For the device already cited for illustration, a sequence of observation of five successive flights corresponds to n = 5 x 1200 = 6000 observation increments. By retaining the objective value of 5.10-e for the alarm probability Pa, the application of the formula F2 results in a value of ks equal to six. (ks = 6) The method then comprises a step E4 of monitoring the device in operation which follows sequences of successive observations of n increments, for example during a number nv successive flights. In this step E4, an alarm E5 is triggered if the number of occurrences of anomalies reaches the value ks during an observation sequence. Example of a second embodiment of the method In this second embodiment, a Poisson distribution is used, the monitoring system indicating a binary result of anomaly or not for regular time increments, as in the first mode. The first step El of extracting a plurality of anomaly scores for a plurality of flights without degradation with the monitored device is here identical to that implemented in the first mode. Here, the reference sequence of m time increments has a duration equal to tc. The increments being equal unit duration on the example, the duration tc of the sequence can be counted in number of increments.
    [0027] The second step E2 of the method, before starting the monitoring, consists in seeking an estimate of an average frequency of occurrence of an anomaly at each increment of time, by using the reference sequence. The estimate is obtained by taking an upper bound of A with a degree of confidence a. For this implementation of the method, it is thus possible to use, as in the first embodiment, a = 90%, a = 50% or a = 44% which corresponds to the maximum likelihood. With the values of the monitoring sequence given as an example in FIG. 1, of length tc = 90,000, counted in time increments, with an abnormality score r = 19, we obtain 2.88 1e, 2.19. -5 and the maximum likelihood estimate: 2.11 10-5. In the third step E3 of this embodiment of the method the threshold ks is determined on the number of occurrences k of anomalies in an observation sequence of duration t, for which it can be considered that the probability of occurrence Anomalies are greater than the estimate of the probability of occurrence of anomalies for a device without degradation from the reference sequence, with a probability of being erroneous equal to the value Pa defined by the formula F1. that, the detection threshold ks on the number k of occurrences of anomalies during the duration t of an observation sequence is determined, in a manner similar to that of the first embodiment, as being the smallest integer such that the distribution function of the probability law 25 on an observation sequence approaches the unit unless the given alarm probability Pa. The same arguments as in the first embodiment perm conclude that if k is greater than or equal to ks, the average occurrence frequency of an anomaly at each increment has exceeded the estimate for a non-degradation device with a probability of being less than Pa. The choice of n number of increments for an observation sequence is done in the same way as in the first embodiment described. In practice, for the surveillance of the devices targeted by the aeronautical applications, the length of the observation sequence corresponds to a few successive flights, their number n being preferably between three and ten. For the example already cited in illustration, an observation sequence of five successive flights corresponds to a duration t equal to 6000 observation increments. By retaining the objective value of 5.108 for the alarm probability Pa, the calculation of ks results in a value equal to six. (ks = 6) The similarity of the results obtained in the two embodiments is consistent with the fact that the two probability laws used are equivalent over a very large number of increments. Variant Using a Confirmation Strategy of p Threshold Exceedances for q Sequences The two previous embodiments have been presented for the transmission of an alarm as soon as a threshold has been exceeded on a sequence. Alternatively, an "S among N" confirmation strategy can be applied to these embodiments, regardless of the discrete probability law used to retrieve the abnormality score r on a sequence. In this strategy, an alarm is emitted if the number of anomaly occurrences exceeds S times the threshold ks for N consecutive sequences. On the other hand, the value of the threshold ks corresponds here to an estimate Peac of the probability of elementary alarm on a sequence repeated N times, to observe S times the exceeding of threshold with a non-degraded device having a probability of alarm Pa by sequence . To do this, one can consider the number k of occurrences of an anomaly per sequence as a random variable from a Bernoulli distribution, taking the sequences as increments. In this case, Peac can be estimated by the following formula: (F4) Peac = B, k-, s + 1 (Pa) in which N is the number of observation sequences, S the minimum number of times the threshold ks has been exceeded repeatedly to confirm the persistence of a detection signal, and Bs, k _5 + 1 the Eulerian distribution function Beta inverse of parameters S and N-S + 1.
    [0028] If N and S are both one, we find Peac = Pa. In this variant, with reference to FIG. 3, the method comprises a step E7, corresponding to the choice of N and S for calculating the value Peac that will be used at the time. place of Pa in step E3 to calculate the threshold ks in order to respect the alarm probability Pa with the confirmation strategy, then in step E4, in which an alarm is emitted after having S observed an overrun of threshold in N successive sequences. The higher the values of N and S chosen in step E7, the greater the detection threshold for confirmation by p overruns among q will be reliable. On the other hand, the degradation phenomenon before failure will have to be slower to be detected in time. The choice of N is therefore based on operational considerations of time to deposition and dynamics of physical degradation until failure. Once chosen N, S can be in turn to maximize the probability of detection. In addition, generally, a sequence will correspond to a flight.
    [0029] Refinement of the decision threshold value A determination of the decision threshold in which the alarm probability Pa is known or estimated from the probability of error Pe has been previously presented. When the alarm probability Pa is estimated, it is possible to carry out, optionally, a step of refining the decision threshold S by refining the probability of detecting a degradation (1- p) as illustrated in the schematic diagrams. FIGS. 2 and 3. As illustrated in FIG. 2, in an evaluation step E6, the monitoring method is implemented with the decision threshold ks as previously determined for a plurality of observation sequences obtained during flights with degradation, preferably, obtained by simulation. During the monitoring process, the abnormality scores rd obtained on the observation sequences are compared with the decision threshold ks, which makes it possible to deduce "a posteriori" the probability of detection of a degradation (1- p). Indeed, it suffices to observe the number of alarms Na emitted with respect to the number of observation sequences corresponding to the flights with degradation monitored.
    [0030] As illustrated in FIG. 2, by iteration, in the determination steps of the decision threshold, the value (1-, q) a is replaced by the value (1-, 6) obtained in the course of the refining to obtain a new value of the decision threshold S more precise. The refining step can be iterated in order to converge towards the value of the decision threshold S the most accurate. In one implementation of the invention, the method of monitoring a degradation of an aircraft on-board device, the method for automatically determining a decision threshold and the method for generating damages are implemented. by a computer, preferably by a plurality of processors of the computer. By way of example, the monitoring method is used to monitor a measurement chain on the turbojet engine.
    [0031] The decision threshold determining processor receives a plurality of abnormality scores without degradation of the monitoring processor and determines, for a given error probability Pe and a probability of detection "a priori" of a degradation by the method of surveillance, the decision threshold ks. Once the decision threshold ks is determined, it is transmitted to the monitoring processor, which can then compare the abnormality scores calculated on the observation sequences of the same duration (n, t) as that used in the monitoring method. said decision threshold ks for monitoring the evolution of the degradation of the measuring chain on the turbojet engine. To refine the value of the decision threshold ks, the degrade generation processor simulates degraded flight data that is submitted to the monitoring processor, which transmits a certain number of alarms according to the data received, which makes it possible to deduce therefrom a posteriori "the probability of detection of degradation by the monitoring method. This value is then communicated to the decision threshold determination processor ksi which provides a new decision threshold value ks for the new detection probability obtained.
    [0032] The process is iterated until a convergence of the value of the decision threshold ks is achieved. In practice, a satisfactory convergence is obtained from two iterations. The invention has been presented for a measurement chain on the turbojet engine, but it applies to any on-board device of an aircraft.
    权利要求:
    Claims (10)
    [0001]
    REVENDICATIONS1. A method of monitoring a degradation of an on-board device of an aircraft, implemented by a computer, the degree of degradation of the on-board device being defined by an abnormality score formed by the count of occurrences of anomalies detected by a control system of the device, the monitoring method comprising a step of comparing an abnormality score obtained for an observation sequence of length (n, t) given to a decision threshold (ks) and a step of issuing an alarm in the event of reaching or exceeding the decision threshold (ks), the decision threshold (ks) being determined automatically for a given alarm probability Pa, corresponding to the probability that an alarm transmitted during the monitoring process while the onboard device is healthy, by means of the following steps: a step of obtaining an abnormality score (r) on at least one reference sequence corresponding to flights of the aircraft without degradation and of length (m, tc) equal to a plurality of lengths (n, t) of observation sequences; a step of adjusting a discrete probability law to find the abnormality score (r) obtained on the reference sequence; a step of calculating the decision threshold (ks) such that by applying the discrete probability law adjusted during the preceding step to an observation sequence having the given length (n), the probability that a score of anomaly greater than or equal to the decision threshold (ks) occurs either less than an elementary probability Peac of threshold exceeding obtained from the given alarm probability Pa by the formula: Peac = (Pa) in which N is a number of observation sequences, S a number of times where the threshold ks will have been exceeded for N successive sequences, and the inverse beta Eulerian distribution function of parameters S and N-S + 1, N and S which may be equal has a.
    [0002]
    2. Method according to claim 1, characterized in that the discrete probability law represents a phenomenon whose occurrences have a probability (p, λ) per time unit, the second step corresponding to an estimate of the probability (p, A) per unit time in the discrete probability law to find the abnormality score (r) obtained on the reference sequence.
    [0003]
    3. Method according to claim 2, wherein the discrete probability law is a binomial law or Poisson's law, parameterized by a probability per unit of time and a sequence length.
    [0004]
    4. Method according to one of claims 2 to 3, for which the estimate of the probability (p, λ) per unit of time of occurrence of anomalies for the system without degradation in the discrete probability law is equal to one increasing this probability per unit of time with a given confidence (a) for the abnormality score (r) in the reference sequence.
    [0005]
    5. Method according to one of claims 2 to 3, for which the estimate of the probability (p, λ) per unit of time of occurrence of anomalies for the system without degradation is equal to the maximum likelihood of the probability ( p, A) per unit corresponding to the abnormality score obtained on the reference sequence.
    [0006]
    6. Method according to one of claims 1 to 5, wherein the length (n, t) of an observation sequence is defined in number (n) of observation increments of unit duration.
    [0007]
    7. Method according to one of claims 1 to 6, wherein an observation sequence corresponds to several successive flights of the aircraft.
    [0008]
    8. Method according to one of claims 1 to 7, comprising a step of presetting the length (n, t) of the observation sequences. 20
    [0009]
    9. Method according to one of claims 1 to 8, wherein the alarm probability Pa is obtained by the following formula: ed Pa = P (1/6) a priori - Pe (1 P-Pd) in which: Pe is the probability that the on-board device is healthy while an alarm is emitted; Pd is the probability of degradation for a given observation sequence; and (1- /)., ',' 'corresponds to the probability of detection of a priori degradation. 30
    [0010]
    The method according to claim 9, comprising: a step of implementing the method for monitoring a degradation of the on-board device, with the decision threshold (ks) previously determined, for a plurality of abnormality scores formed for a plurality of observation sequences with degradation to deduce a probability of detecting a posteriori (1) degradation apoeeriori a step of determining a new probability of alarm -Pa as a function of the probability of detection of a posteriorly (1-6) a posteriori a step of determining a new decision threshold (ks) refined from the new alarm probability.
    类似技术:
    公开号 | 公开日 | 专利标题
    EP3161568B1|2019-01-16|Method of monitoring a degradation of a device on board an aircraft including the determination of a counting threshold
    EP2756280B1|2018-07-11|System for monitoring a measurement chain of a turbojet engine
    EP2850499B1|2016-11-23|Method for monitoring a degradation of an on-board device of an aircraft with automatic determination of a decision threshold
    CA2828704C|2019-03-26|Method and device for monitoring a servo-valve actuation system
    FR2949220A1|2011-02-25|METHOD AND SYSTEM FOR DETECTING THE INGESTION OF AN OBJECT BY AN AIRCRAFT TURBOJUSTER DURING A MISSION
    CA2751114C|2018-05-15|Redundant signal processing system, associated method, and aircraft equipped with such a system
    EP2932226B1|2019-03-13|Method and device for acoustically detecting a malfunction of a motor having an active noise control
    EP1820170B1|2008-04-16|Suppression of false alarms among alarms produced in a monitored information system
    WO2016071605A1|2016-05-12|Tool for validating a system for monitoring an aircraft engine
    EP3631517B1|2021-03-31|Method and device for processing neutron flux signals, associated computer program product and control system
    WO2019129568A1|2019-07-04|Processing method for multi-target detection, characterisation and tracking, and associated device
    FR3010448A1|2015-03-13|METHOD FOR MONITORING A DEGRADATION OF AN AIRCRAFT DEVICE OF AN AIRCRAFT WITH AUTOMATIC DETERMINATION OF A DECISION THRESHOLD
    FR2932583A1|2009-12-18|Failure self-detection method for embarked avionic system of aircraft, involves detecting failure of system by implementing boundary defining sub-space where representation of data is associated to detection of failure, of space
    EP3236280B1|2020-01-01|Automatic calculation of a dimension of a mobile platform
    FR3073348B1|2019-10-04|SYSTEM AND METHOD FOR PROCESSING AND SECURING A SIGNAL TRANSMITTED BY A SENSOR
    EP3306350B1|2019-05-15|Detection module of a disturbance signal while initial receiving by a receiver of navigation information, receiver comprising such module, associated method and computer program product
    EP2939179A1|2015-11-04|Method for segmenting an image of an eye and associated device
    EP2082940A1|2009-07-29|Device and method for detecting a particular status of a vehicle
    FR3046265A1|2017-06-30|SYSTEM FOR MONITORING AN INDUSTRIAL INSTALLATION; ASSOCIATED CONFIGURATION AND MONITORING METHODS
    FR3003986A1|2014-10-03|METHOD FOR DETECTING AND TAKING INTO ACCOUNT A CHANGE ABRUPTED FROM AT LEAST ONE DEVICE MONITORING INDICATOR.
    FR3102270A1|2021-04-23|Method and device for detecting an intrusion on a vehicle data bus
    EP3729003A1|2020-10-28|Detection and accommodation of intermittent open circuits on an engine sensor of an aircraft turbine engine
    FR2987965A1|2013-09-13|Method for analyzing messages transmitted in interconnected network, involves developing alarm signal if level of occurrence in reference table corresponding to scheduling of fresh sequence of messages is lower than given threshold
    同族专利:
    公开号 | 公开日
    CA2953251C|2017-12-05|
    CN106662868A|2017-05-10|
    RU2017100252A|2018-07-26|
    EP3161568A1|2017-05-03|
    RU2677757C2|2019-01-21|
    BR112016030086A2|2017-08-22|
    WO2015197944A1|2015-12-30|
    FR3022997B1|2016-06-10|
    US20170160734A1|2017-06-08|
    EP3161568B1|2019-01-16|
    US9983577B2|2018-05-29|
    CA2953251A1|2015-12-30|
    RU2017100252A3|2018-12-05|
    CN106662868B|2019-04-26|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US20050143873A1|2003-05-06|2005-06-30|Edward Wilson|Model-based fault detection and isolation for intermittently active faults with application to motion-based thruster fault detection and isolation for spacecraft|
    US20070028219A1|2004-10-15|2007-02-01|Miller William L|Method and system for anomaly detection|
    US8700550B1|2007-11-30|2014-04-15|Intellectual Assets Llc|Adaptive model training system and method|
    EP2075474A2|2007-12-28|2009-07-01|United Technologies Corporation|Degraded actuator detection|
    US20110307743A1|2010-06-11|2011-12-15|Asif Khalak|False alarm mitigation|
    US20130024179A1|2011-07-22|2013-01-24|General Electric Company|Model-based approach for personalized equipment degradation forecasting|
    WO2013171427A1|2012-05-16|2013-11-21|Snecma|Method for monitoring a degradation of an on-board device of an aircraft with automatic determination of a decision threshold|
    WO2014064396A2|2012-10-26|2014-05-01|Snecma|System for monitoring a set of components of a device|
    FR1254506A|1960-01-13|1961-02-24|Creusot Forges Ateliers|Improvements to lifting equipment such as overhead cranes or gantries|
    FR1358593A|1963-03-07|1964-04-17|Tecalemit|Advanced injector for supplying internal combustion engines|
    RU2052847C1|1992-05-12|1996-01-20|Ловчиков Сергей Ликарионович|Device for processing flight information|
    RU2205441C1|2001-12-13|2003-05-27|Федеральное государственное унитарное предприятие "Центральный научно-исследовательский институт "Гранит"|Test complex for check of on-board systems of unmanned flying vehile|
    US6892163B1|2002-03-08|2005-05-10|Intellectual Assets Llc|Surveillance system and method having an adaptive sequential probability fault detection test|
    US7933754B2|2006-12-07|2011-04-26|General Electric Company|System and method for damage propagation estimation|
    US7809513B2|2007-04-16|2010-10-05|Acellent Technologies, Inc.|Environmental change compensation in a structural health monitoring system|
    US8437904B2|2007-06-12|2013-05-07|The Boeing Company|Systems and methods for health monitoring of complex systems|
    RU2386569C2|2008-07-02|2010-04-20|Общество с ограниченной ответственностью "Авиакомпания Волга-Днепр"|System for support of decision making by aircraft crew aimed at prevention of abnormal situations|
    US20100030519A1|2008-07-31|2010-02-04|Collier Jarrell D|System for Real-Time Object Damage Detection and Evaluation|
    GB0818544D0|2008-10-09|2008-11-19|Oxford Biosignals Ltd|Improvements in or relating to multi-parameter monitoring|
    FR2980266B1|2011-09-15|2014-06-13|Snecma|MONITORING SYSTEM OF A MEASUREMENT CHAIN OF A TURBOJETACTOR|
    FR2983528B1|2011-12-05|2014-01-17|Snecma|METHOD FOR MONITORING A MEASUREMENT CHAIN OF A TURBOJET ENGINE|
    FR2989500B1|2012-04-12|2014-05-23|Airbus Operations Sas|METHOD, DEVICES AND COMPUTER PROGRAM FOR AIDING THE TROUBLE TOLERANCE ANALYSIS OF AN AIRCRAFT SYSTEM USING REDUCED EVENT GRAPHICS|
    FR3010448B1|2013-09-06|2015-08-21|Snecma|METHOD FOR MONITORING A DEGRADATION OF AN AIRCRAFT DEVICE OF AN AIRCRAFT WITH AUTOMATIC DETERMINATION OF A DECISION THRESHOLD|
    FR3022997B1|2014-06-25|2016-06-10|Snecma|METHOD OF MONITORING A DEGRADATION OF AN AIRCRAFT DEVICE OF AN AIRCRAFT INCLUDING THE DETERMINATION OF A COUNTING THRESHOLD|FR3022997B1|2014-06-25|2016-06-10|Snecma|METHOD OF MONITORING A DEGRADATION OF AN AIRCRAFT DEVICE OF AN AIRCRAFT INCLUDING THE DETERMINATION OF A COUNTING THRESHOLD|
    US10403056B2|2014-12-08|2019-09-03|Nec Corporation|Aging profiling engine for physical systems|
    US10616040B2|2015-11-10|2020-04-07|Telefonaktiebolaget Lm Ericsson |Managing network alarms|
    FR3065995B1|2017-05-05|2019-07-05|Safran Aircraft Engines|SYSTEM AND METHOD FOR MONITORING A TURBOMACHINE WITH ANOMALY DETECTION CORRECTED BY WEAR FACTOR|
    EP3945384A1|2020-07-30|2022-02-02|Siemens Energy Global GmbH & Co. KG|Continuous flow engine self optimizing control method and system|
    法律状态:
    2015-06-12| PLFP| Fee payment|Year of fee payment: 2 |
    2016-01-01| PLSC| Search report ready|Effective date: 20160101 |
    2016-05-06| RM| Correction of a material error|Effective date: 20160404 |
    2016-06-08| PLFP| Fee payment|Year of fee payment: 3 |
    2017-04-26| 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 |
    2018-06-05| PLFP| Fee payment|Year of fee payment: 5 |
    2019-05-22| PLFP| Fee payment|Year of fee payment: 6 |
    2020-05-20| PLFP| Fee payment|Year of fee payment: 7 |
    2021-05-19| PLFP| Fee payment|Year of fee payment: 8 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1455921A|FR3022997B1|2014-06-25|2014-06-25|METHOD OF MONITORING A DEGRADATION OF AN AIRCRAFT DEVICE OF AN AIRCRAFT INCLUDING THE DETERMINATION OF A COUNTING THRESHOLD|FR1455921A| FR3022997B1|2014-06-25|2014-06-25|METHOD OF MONITORING A DEGRADATION OF AN AIRCRAFT DEVICE OF AN AIRCRAFT INCLUDING THE DETERMINATION OF A COUNTING THRESHOLD|
    US15/320,713| US9983577B2|2014-06-25|2015-06-16|Method of monitoring a degradation of a device on board an aircraft including the determination of a counting threshold|
    CA2953251A| CA2953251C|2014-06-25|2015-06-16|Method of monitoring a degradation of a device on board an aircraft including the determination of a counting threshold|
    RU2017100252A| RU2677757C2|2014-06-25|2015-06-16|Method of monitoring degradation of device on board aircraft including determination of counting threshold|
    BR112016030086A| BR112016030086A2|2014-06-25|2015-06-16|process of monitoring a degradation of an aircraft's embedded device that includes determining a count limit|
    PCT/FR2015/051586| WO2015197944A1|2014-06-25|2015-06-16|Method of monitoring a degradation of a device on board an aircraft including the determination of a counting threshold|
    CN201580033762.4A| CN106662868B|2014-06-25|2015-06-16|What the degeneration for the airborne equipment to aircraft was monitored includes the method for determining count threshold|
    EP15733842.7A| EP3161568B1|2014-06-25|2015-06-16|Method of monitoring a degradation of a device on board an aircraft including the determination of a counting threshold|
    [返回顶部]