Detection of electrical arcs in an electrical circuit of an aeronautical system by creating a pattern database

Abstract

The invention relates to a method for detecting electric arcs in an electrical circuit of an aeronautical system, comprising a first phase (P1) of discovering recurring patterns in the behavior of the electric current on a line of said electrical circuit, and storing said recurring patterns in a pattern database, and a second diagnostic phase (P2) comprising comparing the behavior of said electric current on said line with recurring patterns stored in said pattern database, and detecting an electric arc when said behavior does not correspond to a recurring pattern stored in said pattern database. Figure for the abstract: Fig. 1

Description

Title of the invention: Detection of electric arcs in an electrical circuit of an aeronautical system by creating a pattern base FIELD OF THE INVENTION

[0001] The present invention relates to the detection of faulty electrical arcs on an electrical line and applies in particular to the field of aeronautics and on-board systems.

[0002] An electric arc is a self-sustaining discharge with high current.

[0003] Two types of electrical arcs can be distinguished that may occur in an aeronautical environment: breaking arcs, present in contactors when a circuit is opened or closed, and fault arcs that can appear unexpectedly on all components of the electrical chain. The latter can cause severe damage to equipment, systems and even the structure of the aircraft. This is therefore an important point of vigilance.

[0004] Furthermore, fault arcs can be of two types depending on their position on the electrical circuit: parallel arcs and series arcs.

[0005] On systems embedded in current aircraft, these problems are partly resolved, or mitigated, by several factors: - Distributed voltage levels are generally a maximum of 230V to 400V, which reduces the impact of electric arcs; - The distribution of an alternating waveform (AC) also helps promote self-extinction of the arc at each half-period, when the voltage passes through 0. - In continuous mode, voltage levels are generally lower, around 28 V. - Passive protections can be used to limit the consequences of possible fault arcs: choice of materials resistant to these phenomena, distancing of the different elements, etc. - Finally, active protections can also be provided, such as arc detection mechanisms and line opening in the event of an arc being detected. The active protections currently used can detect most parallel arcs but not series arcs for which no active protection has been deployed to date.

[0006] However, recent work around aircraft electrification proposes a distribution of continuous waveforms (DC) with high voltage levels, which can reach one kilovolt. This paradigm shift obviously calls into question the passive protection strategy previously mentioned.

[0007] Indeed, climate change is a major concern for many legislative and regulatory bodies around the world. Various restrictions on carbon emissions have been, are being, or will be adopted by various States. In particular, an ambitious standard applies both to new types of aircraft and to those currently in circulation, requiring the implementation of technological solutions in order to make them compliant with current regulations. Civil aviation has been mobilizing for several years now to make a contribution to the fight against climate change.

[0008] Technological research efforts have already made it possible to significantly improve the environmental performance of aircraft. The Applicant takes into consideration the factors impacting all phases of design and development to obtain less energy-intensive, more environmentally friendly aeronautical components and products whose integration and use in civil aviation have moderate environmental impacts with the aim of improving the energy efficiency of aircraft. Consequently, the Applicant is constantly working to reduce its climate impact by using methods and operating virtuous development and manufacturing processes and minimizing greenhouse gas emissions to the minimum possible to reduce the environmental footprint of its activity.

[0009] This sustained research and development work thus covers new generations of aircraft engines, the lightening of aircraft, in particular through the materials used and lighter on-board equipment, aeronautical biofuels and the development of the use of electrical technologies to ensure propulsion.

[0010] Thus, in the context of the increase in voltage and electrical power levels of aircraft networks, the series electric arc must also be taken into account, in the same way as the parallel arc, because the damage associated with the generation of higher power faults subjected to a direct voltage risks damaging the aircraft and its on-board systems, and could even impact the safety of passengers and crew.

[0011] There are some proposals aimed at detecting an electric arc on a power line in order to isolate it (by cutting the circuit). However, these proposals must respect constraints of: - Reliability: dangerous electrical phenomena must be systematically detected under all circumstances. In other words, we are looking for a high true positive rate, close to 100%. - Robustness: the active protection mechanism must be immune to any event other than the one for which it was designed. In other words, the false positive rate must be as low as possible, i.e. close to 0%.

[0012] So far, these proposals show shortcomings either in terms of reliability or in terms of robustness.

[0013] For example, we can cite patent application EP3959525 which proposes arc detection by the combination of two typical behaviors: high scanning speed at the time of the start of the arc and high frequency behavior. This proposal can therefore allow the detection of electric arcs but it can also detect as electric arcs normal behaviors linked to the load connected to the power line. The rate of false positives is therefore high.

[0014] The same applies to patent application WO202143027 which describes a method based on machine learning. Summary of the invention

[0015] There is therefore a need to improve the current state-of-the-art proposals.

[0016] The invention thus aims in particular to improve the performance of the detection of fault electric arcs on a power line. In particular, it makes it possible to discriminate these electric arcs from transient phenomena linked to the load, and therefore to very significantly improve the rate of false alerts. In other words, the invention makes it possible to significantly increase the rate of true positives and also significantly reduce the rate of false positives.

[0017] According to a first aspect, the present invention can be implemented by a method for detecting electric arcs in an electrical circuit of an aeronautical system, comprising: - a first phase of discovering recurring patterns in the behavior of the electric current on an electric line of said electric circuit, and storing said recurring patterns in a pattern database, and, - a second diagnostic phase comprising a comparison of the behavior of said electric current on said electric line with recurring patterns stored in said pattern base, and a detection of an electric arc when said behavior does not correspond to a recurring pattern stored in said pattern base.

[0018] According to preferred embodiments, the invention comprises one or more of the following features which can be used separately or in partial combination with each other or in total combination with each other: - a time threshold triggers a switch from said first phase to said second phase. - said time threshold corresponds to a duration since the first power-up of said aeronautical system greater than or equal to 15 hours and less than or equal to 35 hours, preferably equal to 25 hours. - said first phase comprises a step of acquiring measurement data relating to said electric current, and of storing said measurement data to form a time series, as well as a step of recognizing recurring patterns within said time series of measurements. - said step of recognizing recurring patterns implements a k-pattern method. - said second phase comprises a step of searching for a transient phenomenon in said electric current, and a step of comparing said transient phenomenon with said stored recurring patterns. - a time window corresponding to said transient phenomenon is extracted, a distance is calculated between said time window and each of said stored recurring patterns, then the presence of an electric arc is detected as a function of all the distances, the calculated distance preferably being a z-normalized Euclidean distance.

[0019] Another aspect of the invention relates to a computer program comprising instructions for implementing a method as previously described when executed on an information processing platform.

[0020] Another aspect of the invention relates to a detection device for detecting electric arcs in an electrical circuit of an aeronautical system, comprising - a functional discovery module adapted for discovering recurring patterns in the behavior of the electric current on an electric line, and storing said recurring patterns in a pattern base, and, - a functional diagnostic module adapted for comparing the behavior of said electric current on said electric line with recurring patterns stored in said pattern base and for detecting an electric arc when said behavior does not correspond to a recurring pattern stored in said pattern base.

[0021] Another aspect of the invention relates to a monitoring device, in particular of the semiconductor power controller type, comprising a detection device as previously defined.

[0022] Another aspect of the invention relates to an aircraft comprising at least one detection device as previously defined.

[0023] According to preferred embodiments, this device may comprise one or more of the features previously mentioned in relation to the method, which can be used separately or in partial combination with each other or in total combination with each other.

[0024] Other characteristics and advantages of the invention will appear on reading the following description of a preferred embodiment of the invention, given by way of example and with reference to the appended drawings. BRIEF DESCRIPTION OF THE FIGURES

[0025] The accompanying drawings illustrate examples of the invention: [Fig.l] schematically illustrates an example of a chronology of implementation of the detection method situating a first and a second phase, according to an embodiment of the invention,

[0026] [Fig.2a] and [Fig.2b] schematically illustrate a functional architecture making it possible to implement a detection device according to two embodiments of the invention.

[0027] [Fig. 3] schematically illustrates a functional architecture of a detection device according to one embodiment of the invention.

[0028] [Fig.4] schematically illustrates a flowchart of a detection method according to one embodiment of the invention.

[0029] [Fig.5a] and [Fig.5b] represent two examples of behavior of the electric current in a supervised electric line, according to implementations of the invention.

[0030] [Fig.6a] and [Fig.6b] illustrate two examples of transient phenomena introduced by electric arcs, according to implementations of the invention.

[0031] DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

[0032] The proposed method is based on two distinct phases, linked together in time, as illustrated in [Fig.l].

[0033] This principle is itself based on the idea that the aeronautical systems of aircraft function a priori correctly during the first hours of flight and that over time, malfunctions are statistically more likely to occur.

[0034] Also, it can be assumed that during a time slot [t0, tj corresponding to the first hours of flight, the behavior of the electric current of a monitored electric line is normal, that is to say free of fault electric arcs. These are only likely to occur in a second time slot, starting at time tb forming a time threshold.

[0035] The instant to corresponds to the first power-up of the aeronautical system (which may be all or part of the aircraft). We can assume t0=0.

[0036] According to this embodiment, this time threshold tl therefore corresponds to a time elapsed since the instant t0 of the putting into service of the aircraft.

[0037] The value of the instant ti may depend on the type of aircraft and may be configurable. A possible value may be greater than or equal to 15 hours and less than or equal to 35 hours, and may be, for example, 25 hours, which corresponds to the time after which the aircraft must undergo preventive maintenance according to certain regulations.

[0038] The method comprises a first phase Pb which may correspond to this first time slot, and during which recurring patterns are searched for and stored in a pattern database. The behavior of the aircraft being assumed to be free of malfunction, these recurring patterns correspond to the nominal operation of the monitored power line, i.e. to the electrical behavior of the load (starting of an engine, for example...).

[0039] This period [t0; tj] must be long enough to allow a wide variety of recurring patterns to be discovered: some may in fact only appear infrequently and it is therefore necessary to study the electrical behavior for a long enough time to identify them.

[0040] It should also be noted that, according to embodiments, the identification of a pattern is not carried out on a single occurrence but on a plurality of occurrences, representative of a non-accidental electrical phenomenon and likely to occur again in the future. For this reason also, the period [t0; tj must be sufficiently extended to allow the identification of several occurrences of the same pattern.

[0041] Thus, the method learns the behavior of the power line without making any assumptions. It consists in fact of determining all the recurring patterns occurring on the power line over time, so that the specificities of the line (in particular its electrical load) can be taken into account. It is therefore not necessary to provide specific or parameterized identification mechanisms depending on the type of line, or a type of load, etc. On the contrary, the method is based on a self-discovery mechanism.

[0042] A second phase, P2, can correspond to the time range starting at time ti. It consists of detecting abnormal electrical behaviors, in particular fault electric arcs. This detection is valid for both series arcs (current drops) and parallel arcs (current surge).

[0043] This diagnostic phase P2 includes a comparison of the behavior of the electric current on the power line with recurring patterns stored in the pattern database. If this behavior does not correspond to a pattern stored in the pattern database, then the presence of a fault arc is detected.

[0044] Thus, the method makes it possible to learn a nominal behavior (in phase Pi) in order to then be able (in phase P2) to discriminate normal electrical artifacts. and not characteristic of the occurrence of an electric arc, compared to artifacts corresponding to electric arcs.

[0045] It is thus possible to improve the rate of detection of fault arcs (rate of true positives) but also considerably reduce the rate of false positives.

[0046] Phases Pi and P2 of the method can be implemented using a device for obtaining measurement flows of the electric current transmitted on the line to be monitored. These measurements can be measurements of the intensity of the electric current.

[0047] Figures 2a and 2b illustrate two embodiments among several possible ones.

[0048] In the embodiment of [Fig.2a], the measuring device 4 is positioned in upstream of the power line 3, close to the current source 1. In the embodiment of [Fig.2b], the measuring device 4 is positioned downstream of the power line 3, close to the load 2.

[0049] Generally speaking, the method presented can be integrated into monitoring devices generally present in the electrical cores of aircraft, such as for example semiconductor power controllers, of the SSPC type (for "Solid State Power Controller" in English). An SSPC consists of one or more semiconductor switching devices and associated semiconductor circuits for protection, action of control signals and provision of status information.

[0050] [Fig.3] illustrates a high-level functional architecture of a device for detection. As these are functional elements, the various functional modules represented can be arbitrarily combined or subdivided during a technical implementation.

[0051] The functions of this architecture will be explained in connection with [Fig.4] which represents an example of a flowchart of the process. Here again, the correspondence between these steps and the functional modules may vary depending on the concrete implementations of the principles of the process and the device proposed.

[0052] According to one embodiment of the detection device 4, a functional measurement module 41 is provided for carrying out measurements on the electrical line 3, for example measurements of the intensity of the electrical current.

[0053] A functional switching module 42 may be provided to switch these measurements to a functional discovery module 43 or a functional search module 45, depending on the phase of the method. This determination may be based on a current time t which may be compared to a time threshold ti between the discovery phase Pi and the diagnosis phase P2.

[0054] Thus, - if t < ti then the discovery functional module 43 is active and the search functional module 45 is inactive, and - if t > tb then the search functional module 45 is active, and the discovery functional module 43 is inactive.

[0055] In other words, the time threshold tl triggers the switch from the first phase to the second phase.

[0056] The discovery functional module 43 can thus be provided to discover recurring patterns in the behavior of the electric current, as measured, on the power line. The discovered recurring patterns can then be stored in a pattern database 44.

[0057] More precisely, in the example of [Fig.4], the proposed method may comprise a step SI of acquiring measurement data (coming from the measurement functional module 41). This measurement data is stored in a memory, not shown.

[0058] These measurement data thus constitute a time series.

[0059] A time series, or chronological series, is a sequence of numerical values ​​representing the evolution of a specific quantity over time

[0060] In step S2, the condition for switching to phase P2 is tested.

[0061] As long as the switching condition is not validated (i.e. it has not yet reached the time threshold ti), new measurements are acquired, and the method therefore loops back to step SL

[0062] When the switching condition is reached (t>ti), a step S3 of processing the stored measurement flow can be implemented before switching to diagnostic phase P2.

[0063] This step S3 aims to recognize recurring patterns within a time series of measurements.

[0064] A pattern can be defined as a time series that repeats itself (with a possible error rate) within a larger time series.

[0065] The concept of motif discovery in a time series was proposed in the article by P. Patel, E. Keogh, J. Lin and S. Lonardi, "Mining motifs in massive time series database" in 2002 IEEE International Conference on Data Mining, 2002 Proceedings, pages 370-377.

[0066] A time series T=(tl, t2, .. ..tn) of length n can be defined as an ordered sequence of n real values.

[0067] We can also define a subsequence Sy of a time series T, with 1 ❖ i ❖ n and 1 ❖ i+1 ❖ n, as a time series of length 1 consisting of the 1 successive real values ​​belonging to T and starting at index i: Sy=(ti, ti+i,..., ti+u).

[0068] Pattern discovery then amounts to searching for subsequences of greater length verifying a minimum distance between them. The distance can be a Euclidean distance or, more particularly, a z-normalized Euclidean distance.

[0069] We can define a z-normalized Euclidean distance between two subsequences such as

[0070] [Math.l]

[0071] We also define, for each time series, the mean q(1) q(2) respectively and the standard deviation o'9 o(2) respectively.

[0072] The z-normalized Euclidean distance ZED can then be written as

[0073] [Math.2]

[0074] Other distances can however also be used.

[0075] We can then define the notions of correspondence and trivial correspondences between two subsequences, on the basis of a distance measure.

[0076] Two subsequences SySjj form a “correspondence” (or “matching” in English) if and only if [Math.3] ZED^S^, Sjj) < r€R , and - Sy Sjj do not form a trivial correspondence.

[0077] The radius r is a parameter that defines the proximity that two subsequences must have to be considered as "corresponding". For this reason, the term "R-matching" is used in English to define this correspondence.

[0078] Trivial matches are those that any subsequence forms with a sequence slightly shifted in time. These matches are excluded in pattern searching.

[0079] A possible definition of trivial correspondence can be: two subsequences SySjj of the same length 1 and of the same time series form a trivial correspondence ("trivial match" in English) if and only if they share at least 1 / 2 common indices of the time series T: (il / 2)^j^(i+E2).

[0080] The aforementioned article proposes a first method called “K-motifs”.

[0081] Given a time series T, a subsequence length n and a radius R, the most significant pattern in T is the subsequence Cl that has the largest number of non-trivial matches. The K-pattern is the largest set of subsequences of length 1 in which each subsequence forms an (“R-matching”) with all other subsequences in the set

[0082] Many other methods have been proposed to enable the discovery of patterns in time series depending on the nature or type of patterns to be discover.

[0083] For example, a method described in the article by M. Linardi, Y. Zhu, T. Papanas and E. Keogh, “Valmod: A suite for easy and exact detection of cariable length motifs in serial data”, in Proceedings of the 2018 International Conference on Management of Data, pages 1757-1760, 2018, consists of first determining a pair of subsequences forming a correspondence, then iteratively determining new subsequences forming correspondence with members of a set that is thus progressively increased.

[0084] According to one embodiment, a k-motiflet method is used. This method was described in the article by Patrick Schâfer and Ulf Leser, “Motiflets - Fast and Accurate Detection of Motifs in Time Series” in Woodstock '18: ACM Symposium on Neural Gaze Detection, June 03-05, 2018, Woodstock, New York, doi.org / 10.1145 / 1122445.1122456

[0085] The notion of extension (or "extent" in English) of a set S of patterns is defined as being the maximum value of the set of Euclidean distances between each pattern in the set, taken two by two. In other words:

[0086] [Math.4] mendsU max \ / / (Y; S®) esxs' v 7 7

[0087] We can then define the best k-motiflet as being the set S, of cardinality k, of subsequences of length 1 for which - All subsequences of S form two-by-two correspondences, - There is no set S' with extent(S') <extent(S) et qui remplit also these constraints.

[0088] The aforementioned article also describes algorithms for determining k-motiflets within a time series.

[0089] One of the advantages of an implementation based on a k-motiflet method is that it is independent of the radius r required for the Valmod and k-motif methods. Current analysis can thus be independent of the electrical monitoring system.

[0090] The article describes a possible method for automatically finding the values ​​of k and 1. This method involves looking at the value of each extent(Sk) for the entire acquisition. By doing this, the elbow points of the extent(Sk) indicate the changes in patterns throughout an acquisition. The number of different inclinations of the extent(Sk) therefore form an estimate of the parameter k. To determine the optimal value of parameter 1 (pattern length), one can look at what value of 1 the area above below the extension extent(Sk) is minimal.

[0091] Figures 5a and 5b represent two examples of behavior of the electric current in the supervised power line.

[0092] This behavior is captured by measurement data, for example the current intensity expressed in Amperes (A in the figures). All of this measurement data thus represents a time series where for each instant t corresponds a value of the current A.

[0093] In this example, the discovery phase reveals a pattern that appears at locations T1, T2, T3. This recurring pattern corresponds to a transient phenomenon of the current caused by the loads of the electrical network. It may, for example, be a regular or irregular ignition of an engine or other device of the aircraft, for example. In any case, it corresponds to the normal operation of the aircraft.

[0094] [Fig.5b] shows another example, in which a T4 motif is discovered.

[0095] The patterns discovered by the analysis of the stored time series are then stored in the pattern database 44.

[0096] Thus, at the end of the discovery phase, this pattern base 44 contains a dictionary of patterns corresponding to the transient phases of the electric current corresponding to normal operation of the electric circuit. As seen previously, the time threshold ti is chosen so that this dictionary contains all possible normal transients of the current during operation of the aircraft.

[0097] When the diagnostic phase is triggered, the discovery phase may be interrupted.

[0098] It is assumed that at this point there is no longer any guarantee that the electrical system will not present an arc fault, and the aim is then to detect them with the best possible precision, using the previously created pattern dictionary.

[0099] When the switching condition is met, that is to say for example when the time threshold ti is crossed, the switching functional module 42 switches the measurement data carried out by the measurement functional module 41 to the search functional module 45.

[0100] This functional search module 45 can implement a step S4 of searching for a transient phenomenon in the electric current.

[0101] It should be noted that when a fault arc occurs, it also generates transient current phenomena. This is valid for both series arcs (current drop) and parallel arcs (current surge).

[0102] Figures 6a and 6b illustrate two examples of such transient phenomena. [Fig.6a] illustrates two transients T5, T6 linked to series arcs (drop in intensity A). [Fig.6b] illustrates 3 transients T7, T8, T9 linked to parallel arcs (calls in intensity A).

[0103] Also, continuously (as represented by the loop in step S4 in [Fig.4]), we search for transient phenomena within the time series constituted by the acquired measurement data.

[0104] To do this, we can define a sliding window to analyze the time series, and a thresholding of the current.

[0105] For example, if the growth or decrease of the current exceeds a given threshold (for example 5%), then it can be considered that a transient phenomenon is detected. The sliding window located around this crossing time point can be extracted for the subsequent step S5 of comparing this detected phenomenon with the patterns stored in the pattern base 44. This step S5 can be implemented by a diagnostic functional module 46.

[0106] This step S5 aims to determine whether the detected transient phenomenon corresponds to normal behavior, linked to a load connected to the power line, or to an arc fault.

[0107] To do this, we can measure a distance between the data contained in the sliding window and the data contained in each pattern in the pattern base.

[0108] The distance may be a Euclidean distance, a normalized Euclidean distance, a DTW distance, etc.

[0109] Dynamic Time Warping (DTW) is an algorithm for measuring the similarity between two sequences that can vary over time.

[0110] The normalized Euclidean distance between the time window y and a pattern x; contained in the pattern base can be described by [YES] [Math.5] dnEuc{^ y) =

[0112] In this expression, the "tildes" (~) indicate the normalization of the quantities. For example:

[0113] [Math.6] ~ laughed X i LJ cr x ~ r iy[n] =

[0114] In these equations: - N represents the total amount of points in the pattern (i.e. the length of the subsequence that was determined as the pattern in the PI discovery phase); - Xi represents the pattern i within the pattern base 44. This same distance must therefore be estimated for each pattern x; of the pattern base.

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124]

[0125]

[0126]

[0127]

[0128] - y represents the analysis window extracted in step S4. It is also a subsequence of a time series. - qx qy represent the means of, respectively, the pattern x; and the analysis window y. - ox>oy represent the standard deviations of, respectively, the pattern x; and the analysis window y. Euclidean distance is sensitive to time shifts, expansion / contraction (of the pattern relative to the analysis window), outliers, and time delays. However, normalization increases sensitivity to amplitude changes. The analysis window can be planned to be of the same length as the recurring patterns stored in the pattern database. In the case of an implementation based on a k-motiflet method, this length 1 is determined when identifying recurring patterns. In the case of a transient phase induced by a load longer than the time window, detection and identification will still work on the basis of a sub-part of this transient. The dynamic time warping (DTW) algorithm is more robust against expansion / contraction. Different distances and different mechanisms and algorithms can implement this S5 step. For each time window extracted in step S4, a distance is therefore calculated with all the patterns stored in the pattern database 44. We can then determine a decision function based on this set of distances. In particular, we can write a decision function f such that: [Math.7] With : [Math. 8] 0, otherwise d(.Y(, y) is a distance measure between the pattern x; and the analysis window y. This distance can for example be the normalized Euclidean distance dnEUC(x / , y) described previously. M is the number of recurring patterns stored in the pattern base 44. S is a detection threshold. Its value can be predefined.

[0129] The value taken by the decision function f directly gives the desired result: - If f=0, then the analysis window corresponds to a pattern x; among those stored in the pattern database. This is therefore normal electrical behavior. In the example of [Fig.4], the method then returns to step S4 to search for a new transient phenomenon. - If f=l, then the analysis window does not correspond to any pattern among those stored in the pattern database. This is therefore abnormal electrical behavior, i.e. an electric arc.

[0130] In the case where an electric arc is detected, a step S6 of processing the detection can be triggered. This step S6 can be implemented by a functional processing module 47.

[0131] This step may include emergency reactions such as cutting and isolating the supervised power line: an electric arc denotes abnormal behavior that may cause other electric arcs in the future, each of which may damage the system locally, including the aircraft, as seen previously. It is therefore advisable to protect against any potential damage by cutting the power line as soon as a first electric arc is detected.

[0132] Also, an alert can be triggered in order to warn a human operator who can remedy this fault (activation of a possible redundant system following the shutdown of the faulty one, interventions to repair the fault, etc.)

[0133] Of course, the present invention is not limited to the examples and embodiments described and shown, but is defined by the claims. It is in particular susceptible of numerous variants accessible to those skilled in the art.

Claims

Claims

1. Method for detecting electric arcs in an electrical circuit of an aeronautical system, comprising - a first phase (PI) of discovering recurring patterns in the behavior of the electric current on an electric line (3) of said electric circuit, and storing said recurring patterns in a pattern base (44), and, - a second phase (P2) of diagnosis comprising a comparison of the behavior of said electric current on said electric line (3) with recurring patterns stored in said pattern base (44), and a detection of an electric arc when said behavior does not correspond to a recurring pattern stored in said pattern base (44).

2. Method according to the preceding claim, in which a time threshold (ti) triggers a switch from said first phase (PI) to said second phase (P2).

3. Method according to the preceding claim, in which said time threshold (b) corresponds to a duration since the first powering up of said aeronautical system greater than or equal to 15 hours and less than or equal to 35 hours, preferably equal to 25 hours.

4. Method according to one of the preceding claims, in which said first phase (PI) comprises a step (SI) of acquiring measurement data relating to said electric current, and of storing said measurement data to form a time series, as well as a step (S3) of recognizing recurring patterns within said time series of measurements.

5. Method according to the preceding claim, wherein said step (S3) of recognizing recurring patterns implements a k-pattern method.

6. Method according to one of the preceding claims, in which said second phase (P2) comprises a step of searching (S4) for a transient phenomenon in said electric current, and a step of comparing (S5) said transient phenomenon with said stored recurring patterns.

7. Method according to the preceding claim, in which a time window corresponding to said transient phenomenon, a distance is calculated between said time window and each of said stored recurring patterns, then the presence of an electric arc is detected as a function of all the distances, the calculated distance preferably being a z-normalized Euclidean distance.

8. Computer program comprising instructions for implementing a method according to one of the preceding claims when executed on an information processing platform.

9. Detection device (4) for detecting electric arcs in an electrical circuit of an aeronautical system, comprising - a functional discovery module (43) adapted for discovering recurring patterns in the behavior of the electric current on an electric line (3), and storing said recurring patterns in a pattern base (44), and - a functional diagnostic module (46) adapted for comparing the behavior of said electric current on said electric line with recurring patterns stored in said pattern base (44) and for detecting an electric arc when said behavior does not correspond to a recurring pattern stored in said pattern base (44).

10. Aircraft comprising at least one detection device (4) according to claim 9.

Images