Detecting arcing in circuits of an aviation system by forming a pattern database

Abstract

The invention relates to a method for detecting an electric arc in an electrical circuit of an aerial system, the method comprising a first phase (P1) which identifies a recurring pattern in the behavior of the current on an electrical line of the electrical circuit and stores the recurring pattern in a pattern database, and a second diagnostic phase comprising comparing the behavior of the current on the electrical line with the recurring patterns stored in the pattern database and detecting an electric arc when said behavior does not correspond to a recurring pattern stored in the pattern database.

Description

Technical Field

[0001] This invention relates to the detection of arcing faults in electrical wires, and is particularly applicable to the field of aviation and airborne systems. Background Technology

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

[0003] Two types of electric arcs can be distinguished in an aviation environment: disconnecting arcs present in contactors when opening or closing circuits, and arcing faults that can occur unexpectedly across all components of the electrical chain. These arcing faults can cause severe damage to equipment, systems, and even aircraft structures. Therefore, this is an area requiring high vigilance.

[0004] Furthermore, depending on the location of the arc fault in the circuit, arc faults can be of two types: parallel arcs and series arcs.

[0005] In current airborne systems of aircraft, these problems are partially solved or mitigated through the following factors: - Distributed voltage levels are typically between 230 V and 400 V, which reduces the impact of electric arcs; - When the voltage passes through 0, the distribution of the alternating current (AC) waveform promotes the self-extinguishing of the arc at each half-cycle.

[0006] - In DC mode, the voltage level is typically lower, in the 28 V range.

[0007] Passive protection can be used to limit the effects of any arc fault: selecting materials that resist these phenomena, increasing the distance between different components, etc.

[0008] Finally, active protection can also be provided, such as arc detection and line-opening mechanisms in the event of arc detection. Currently used active protection enables the detection of most parallel arcs, but not series arcs; for series arcs, active protection has not yet been deployed. However, recent work on aircraft electrification has proposed direct waveform (DC) distributions with high voltage levels potentially reaching 1 kilovolt. This paradigm shift clearly poses a challenge to the passive protection strategies mentioned earlier.

[0009] Climate change is a major concern for many legislative and regulatory bodies worldwide. States have implemented, are implementing, and will implement various restrictions on carbon emissions. In particular, a challenging standard applies to both new and currently operating aircraft, meaning that technological solutions must be found to comply with existing regulations. For several years, the civil aviation industry has been taking action to contribute to addressing climate change.

[0010] Efforts in technological research have significantly improved the environmental performance of the aircraft. The applicant has considered factors influencing all stages of design and development to obtain more environmentally friendly, less energy-intensive aerospace components and products, whose integration and use in civil aviation have a moderate environmental impact, and aims to improve the energy efficiency of the aircraft. Therefore, the applicant has consistently strived to use environmentally friendly development and manufacturing processes and methods to reduce greenhouse gas emissions to the lowest possible level and minimize the environmental impact of its operations, thereby reducing its climate impact.

[0011] Therefore, this ongoing research and development effort simultaneously focuses on next-generation aircraft engines, particularly reducing aircraft weight through the materials used and lighter airborne equipment, aircraft biofuels, and the development and use of electrical technologies to ensure propulsion.

[0012] Therefore, when increasing the voltage level and electrical power in the aircraft circuitry, series arcs must also be considered in the same way as parallel arcs, because the hazards associated with the generation of higher power faults subjected to DC voltages risk damaging the aircraft and its onboard systems, and may even affect the safety of passengers and crew.

[0013] Some existing solutions aim to detect electric arcs on wires and isolate them (by shutting off the circuit). However, these solutions must be aware of the following limitations: - Reliability: It must be able to systematically detect hazardous electrical phenomena under all circumstances. In other words, a near 100% true positive rate is required.

[0014] - Robustness: Active protection mechanisms must not react to any events other than those they are designed to respond to. In other words, the false positive rate must be as low as possible, i.e., close to 0%.

[0015] So far, these proposals have shown shortcomings in terms of reliability or robustness.

[0016] For example, patent application EP3959525 proposes arc detection through a combination of two typical behaviors: high conversion rate at the onset of an arc and high-frequency behavior. Therefore, this scheme can detect arcs, but it also detects normal behavior associated with the load connected to the wire as an arc. Consequently, the false positive rate is high.

[0017] This also applies to patent application WO202143027, which describes a machine learning-based method. Summary of the Invention

[0018] Therefore, current proposals in related technologies need to be improved.

[0019] This invention specifically addresses the improvement in the detection performance of arc faults on electrical wires. In particular, this invention enables the differentiation between these arcs and load-related transient phenomena, and thus significantly improves the false alarm rate. In other words, this invention allows for a substantial increase in the true positive rate and an equally substantial decrease in the false positive rate.

[0020] In a first aspect, the present invention can be implemented using a method for detecting electric arcs in the circuits of an aviation system, the method comprising: - In the first stage, the recurring patterns in the behavior of the current in the wires of the circuit are identified, and the recurring patterns are stored in a pattern database. - The second stage, which is a diagnostic stage, includes comparing the behavior of the current on the wire with recurring patterns stored in the pattern database, and detecting an arc when the behavior does not correspond to a recurring pattern stored in the pattern database.

[0021] According to a preferred embodiment, the present invention includes one or more of the following features, which may be used alone, in partial or complete combination with each other: - A time threshold triggers the switch from the first stage to the second stage; - The time threshold corresponds to a period of time since the first start-up of the aviation system, the period of time being greater than or equal to 15 hours and less than or equal to 35 hours, preferably 25 hours; - The first stage includes the steps of acquiring measurement data related to the current, storing the measurement data to form a time series, and identifying recurring patterns within the time series of the measurements; - The identification step of the recurring pattern uses the k-motiflet method; - The second stage includes the steps of searching for transient phenomena in the current and comparing the transient phenomena with stored recurring patterns; - Extract the time window corresponding to the transient phenomenon, calculate the distance between the time window and each of the recurring patterns in the stored recurring patterns, and then detect the presence of the electric arc based on the set of distances. The calculated distance is preferably the z-normalized Euclidean distance.

[0022] Another aspect of the invention relates to a computer program comprising instructions that, when executed on a data processing platform, implement methods such as those previously described.

[0023] Another aspect of the present invention relates to a detection device for detecting electric arcs in the circuits of an aviation system, comprising: - A function discovery module, adapted to discover recurring patterns in the behavior of current in a wire, and adapted to store said recurring patterns in a pattern database, and - A functional diagnostic module adapted to compare the behavior of the current on the wire with recurring patterns stored in the pattern database, and to detect an arc when the behavior does not correspond to a recurring pattern stored in the pattern database.

[0024] Another aspect of the present invention relates to a monitoring device, particularly a monitoring device of the semiconductor power controller type, which includes the detection device as described above.

[0025] Another aspect of the present invention relates to an aircraft comprising at least one detection device as described above.

[0026] According to a preferred embodiment, the apparatus may include one or more previously mentioned features related to the method, which can be used individually or in partial or complete combination with each other.

[0027] Other features and advantages of the invention will become apparent from the following description of preferred embodiments of the invention given as examples and with reference to the accompanying drawings. Attached Figure Description

[0028] The accompanying drawings illustrate an example of the invention: Figure 1 An example illustrating the temporal sequence implementation of the detection method in the first and second stages in one embodiment of the present invention is shown.

[0029] Figure 2a and Figure 2b The functional architecture that enables the use of detection devices according to two embodiments of the present invention is illustrated schematically.

[0030] Figure 3 The functional architecture of a detection device according to an embodiment of the present invention is illustrated schematically.

[0031] Figure 4 A flowchart of a detection method according to an embodiment of the present invention is shown schematically.

[0032] Figure 5a and Figure 5b Two examples of the behavior of current in a monitored wire according to an implementation of the present invention are shown.

[0033] Figure 6a and Figure 6bTwo examples of transient phenomena caused by electric arcs according to an implementation method of the present invention are shown. Detailed Implementation

[0034] The proposed method is based on two separate stages arranged in chronological order, such as Figure 1 As shown.

[0035] The principle itself is based on the concept that, theoretically, an aircraft's aviation systems operate correctly during the first few hours of flight, and that failures are statistically more likely to occur over time.

[0036] Therefore, it can be assumed that the current behavior on the monitored wire is normal during the time period [t0, t1] corresponding to several hours prior to the flight, i.e., there is no arcing fault. Arcing faults can only occur during the second time period starting from time t1, thus forming a time threshold.

[0037] Time t0 corresponds to the first start-up of the aviation system (which can be all or part of the aircraft). It can be assumed that t0 = 0.

[0038] Therefore, in this embodiment, the time threshold t1 corresponds to the time elapsed since the launch time t0 of the aircraft.

[0039] The value of time t1 can depend on the type of aircraft and can be parameterized. One possible value could be greater than or equal to 15 hours and less than or equal to 35 hours, and could, for example, be 25 hours corresponding to the time during which preventative maintenance of the aircraft is required according to some regulations.

[0040] The method includes a first phase P1, which may correspond to a first time period, during which recurring patterns are searched and stored in a pattern database. Assuming the aircraft's behavior is fault-free, these recurring patterns correspond to the nominal function of the monitored wiring, i.e., the electrical behavior of the load (e.g., engine starting, etc.).

[0041] The period [t0; t1] must be long enough to detect multiple recurring patterns: since some of these patterns may occur infrequently, it is necessary to examine the electrical behavior over a sufficiently long period to identify these patterns.

[0042] It should also be noted that, according to the embodiments, pattern recognition does not occur on a single event, but rather on multiple events representing non-accidental electrical phenomena that may recur in the future. For this reason, the period [t0; t1] must be long enough to allow for the identification of multiple occurrences of the same pattern.

[0043] Therefore, this method learns the behavior of wires without any pre-defined assumptions. It requires identifying all recurring patterns appearing on the wires over time, allowing consideration of the specifications of the line (especially the electrical load). Thus, it is not necessary to provide specific identification mechanisms or parameters based on line type or load type. In particular, the proposed method avoids laboratory learning for each individual load and / or modeling signals associated with complex combinations of loads in the aircraft's electrical network under nominal behavior (which may also vary over time, e.g., depending on flight phase). Instead, the method is based on a self-discovery mechanism.

[0044] The second stage, P2, can correspond to the time period starting at time t1. This second stage involves detecting abnormal electrical behavior, particularly arcing faults. This detection is effective for both series arcs (current decrease) and parallel arcs (current increase).

[0045] The diagnostic phase P2 involves comparing the behavior of the current on the wire with recurring patterns stored in a pattern database. If the behavior does not match a pattern stored in the database, an arc fault is detected.

[0046] Therefore, this method enables the learning of nominal behavior (in stage P1), making it possible (in stage P2) to distinguish between normal electrical artifacts that are not features of an arc and artifacts that do indeed correspond to an arc.

[0047] This can improve the detection rate (true positive rate) of arc faults, but also significantly reduce the false positive rate.

[0048] Phases P1 and P2 of this method can be implemented using a device that enables the acquisition of a measurement stream of the current transmitted on the line to be monitored. These measurements can be measurements of current intensity.

[0049] Figure 2a and Figure 2b Two of several possible embodiments are shown.

[0050] exist Figure 2a In the embodiment, the measuring device 4 is located upstream of the wire 3, close to the current source 1. Figure 2b In one embodiment, the measuring device 4 is located downstream of the load 2 on the wire 3.

[0051] Typically, the proposed method can be integrated into monitoring devices commonly found in the electrical core of an aircraft, such as a Solid State Power Controller (SSPC). An SSPC consists of one or more semiconductor switching devices and associated semiconductor circuitry for protection, control signal activation, and status reporting.

[0052] Figure 3 Shows the high-level functional architecture of the detection device. As functional elements, the different functional modules shown can be arbitrarily combined or subdivided for technical implementation.

[0053] Will be combined with Figure 4 to explain the functions of this architecture, Figure 4 Illustrates an example of the flowchart of the method. Here, the correspondence between these steps and functional modules can vary according to the actual implementation of the principles of the proposed method and device.

[0054] In one embodiment of the detection device 4, a functional measurement module 41 is provided to perform measurements on the wire 3, such as measuring the current intensity.

[0055] Depending on the stage of the method, a functional routing module 42 can be provided to route these measurement results to a functional discovery module 43 or a functional search module 45. This determination can be based on the current time t, which can be compared with a time threshold t1 between the discovery stage P1 and the diagnostic stage P2.

[0056] Therefore: - If t < t1, the functional discovery module 43 is active and the functional search module 45 is inactive, and - If t > t1, the functional search module 45 is active and the functional discovery module 43 is inactive.

[0057] In other words, the time threshold t1 triggers the switch from the first stage to the second stage.

[0058] Therefore, a functional discovery module 43 can be provided to discover recurring patterns in the current behavior measured on the wire. Then the discovered recurring patterns can be stored in the pattern database 44.

[0059] More specifically, in the Figure 4 example, the proposed method can include an acquisition step S1 of measurement data (derived from the functional measurement module 41). These measurement data are stored in a memory not shown.

[0060] Therefore, these measurement data form a time series.

[0061] A time series or time sequence is a sequence of numerical values representing the variation of a specific quantity over time.

[0062] In step S2, the switching condition to stage P2 is tested.

[0063] As long as the switching condition is not effective (i.e., the time threshold t1 has not been reached), new measurement results are acquired, and thus the method loops back to step S1.

[0064] When the switching condition (t>t1) is met, step S3 can be performed to process the stream of stored measurement results before switching to the diagnostic phase P2.

[0065] Step S3 aims to identify recurring patterns in the time series of measurement results.

[0066] A pattern or motif can be defined as a time series that repeats (with possible error rates) over a longer time series.

[0067] The concept of motif discovery in time series was proposed by P. Patel, E. Keogh, J. Lin and S. Lonardi in their paper “Mining motifs in massive time series database”, Proceedings of the 2002 IEEE International Conference on Data Mining, pp. 370-377.

[0068] A time series T=(t1, t2,…tn) of length n can be defined as an ordered sequence of n real values.

[0069] It is also possible to subsequence S of time series T i,l (where 1≤i≤n and 1≤i+l≤n) is defined as a time series of length l, which consists of l consecutive real values ​​belonging to T and starting from index i: S i,l =(t i , t i+1 , ..., t i+l-1 ).

[0070] Then, motif discovery is equivalent to searching for longer subsequences and verifying the minimum distance between them. The distance can be Euclidean distance, or more specifically, z-normalized Euclidean distance.

[0071] We can define the z-normalized Euclidean distance between two subsequences such that:

[0072] For each time series, the average value μ is also defined. (1) and μ (2) and standard deviation δ (1) and δ (2) .

[0073] The z-normalized Euclidean distance ZED can then be written as:

[0074] However, other distances can also be used.

[0075] Then, the concepts of matching and trivial matching between two subsequences can be defined based on the distance measurement results.

[0076] If and only if: ,and S i,l S j,l No trivial matching At that time, the two subsequences S i,l S j,l A match is formed.

[0077] The radius r is a parameter that defines the closeness of two subsequences to be considered a "match". Therefore, the term "R-match" is used to define this match.

[0078] A trivial match is a match formed by any subsequence and a sequence that is slightly shifted in time. These matches are excluded in pattern search.

[0079] One possible definition of trivial matching is: two subsequences S of the same length l and in the same time series. i,l S j,l A trivial match is formed if and only if they share at least l / 2 common indices of the time series T: (l / 2) common indices share l / 2.

[0080] The article above proposes the first method called "k-motif".

[0081] Given a time series T, a subsequence length n, and a radius R, the most important motif in T is the subsequence C1 with the largest number of nontrivial matches. A K-motif is the largest set of subsequences of length l, where each subsequence matches all other subsequences in the set (R-matches).

[0082] Many other methods have been proposed to allow the discovery of motifs in time series based on the nature or type of the motif to be discovered.

[0083] For example, the article by M. Linardi, Y. Zhu, T. Papanas, and E. Keogh in the proceedings of the 2018 International Conference on Data Management, pages 1757-1760. Valmod: A suite for easy and exact detection of variable length motifs in data series The method described in the document requires first determining a pair of subsequences that form a match, and then iteratively determining new subsequences that form a match with elements of a set whose size gradually increases.

[0084] In one embodiment, the k-motiflet method is used. This method is described in the article " Motiflets – Fast and Accurate detection of Motifs in Time Series " by Patrick Schäfer and Ulf Leser in Woodstock'18: ACM Workshop on Neural Gaze Detection (June 3 - 5, 2018, Woodstock, New York, doi.org / 10.1145 / 1122445.1122456).

[0085] The concept of the extent of a set S of motifs is defined as the maximum of all Euclidean distances between each pair of motifs taken from the set. In other words:

[0086] Then, the best k-motiflet can be defined as the set of subsequences S of cardinality k and length I, where: - All subsequences of S form pairwise matches.

[0087] - There does not exist a set S' with extent(S') < extent(S) that also satisfies these constraints.

[0088] The above article also describes an algorithm that enables the determination of k - motiflets within a time series.

[0089] One of the advantages of the implementation based on the k-motiflet method is that it does not depend on the radius r required by the Valmod and k-motif methods. Therefore, the analysis of the current can be independent of the electrical monitoring system.

[0090] A possible method for automatically finding the values of k and I is described in the article. This method requires recording the value of each extent(S<000​​​​​​​​​​​​​​​This behavior is captured by measurements of current density, for example, expressed in amperes (A in the figure). Thus, all these measurements represent a time series with a corresponding current value A at each time t.

[0093] In this example, the discovery phase revealed a pattern appearing at locations T1, T2, and T3. This recurring pattern corresponds to transient current phenomena caused by loads on the power grid. For example, it could involve the periodic or irregular starting of an engine or other equipment, such as in an aircraft. In all cases, it corresponds to its normal operation.

[0094] Figure 5b Another example of discovery pattern T4 is shown.

[0095] Then, the patterns discovered by analyzing the stored time series are stored in the pattern database 44.

[0096] Therefore, upon completion of the discovery phase, the pattern database 44 contains a dictionary of patterns corresponding to transient phases of the current and normal operation of the circuit. As previously mentioned, a time threshold t1 is selected such that the dictionary includes all possible normal transients of the current throughout the entire operation of the spacecraft.

[0097] The discovery phase can be interrupted when the diagnostic phase is triggered.

[0098] Assume that there is no longer a guarantee that the electrical system will not exhibit arcing faults, and the goal is to use the previously constructed pattern dictionary to obtain the detection of arcing faults with the highest possible accuracy.

[0099] When the switching condition is met, that is, when, for example, a time threshold t1 has elapsed, the function routing module 42 routes the measurement data obtained by the function measurement module 41 to the function search module 45.

[0100] The function search module 45 can perform the search step S4 for transient phenomena in the current.

[0101] It should be noted that when an arc fault occurs, it also generates a transient current phenomenon. This applies to both series arcs (current decrease) and parallel arcs (current increase).

[0102] Figure 6a and Figure 6b Two examples of the transient phenomenon are shown. Figure 6a Two transients, T5 and T6 (intensity A decrease), associated with the series arc are shown. Figure 6b Three transients T7, T8, and T9 (intensity A increases) associated with the parallel arc are shown.

[0103] Therefore, in a continuous manner (such as Figure 4(As shown in the loop at step S4), transient phenomena are searched within the time series formed by the acquired measurement data.

[0104] Therefore, a sliding window can be defined to analyze time series and current thresholds.

[0105] For example, if the increase or decrease in current exceeds a given threshold (e.g., 5%), a transient phenomenon can be considered detected. A sliding window located near this crossover point can be extracted for subsequent comparison step S5 between the detected phenomenon and patterns stored in the pattern database 44. This step S5 can be performed by the functional diagnostic module 46.

[0106] Step S5 aims to determine whether the detected transient phenomenon corresponds to normal behavior associated with the load connected to the wire, or to an arc fault.

[0107] To do this, the distance between the data contained in the sliding window and the data contained in each pattern in the pattern database can be measured.

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

[0109] Dynamic Time Warping (DTW) is an algorithm that enables the measurement of the similarity between two sequences, which can change over time.

[0110] Time window y and pattern x contained in the pattern database i The normalized Euclidean distance between them can be described as:

[0111] In this expression, the tilde (~) symbol indicates amplitude normalization. For example:

[0112] In these equations: -N represents the total number of points in the pattern (i.e., the length of the subsequence that is identified as the pattern in the discovery phase P1); -x i This represents pattern i in pattern database 44. Therefore, for each pattern x in the pattern database... i Estimate the same distance; -y indicates the analysis window extracted in step S4. It is also a subsequence of the time series; -µ x µ y respectively, mode x i And the mean of y in the analysis window; -δx δ y They represent pattern x respectively i And the standard deviation of the analysis window y.

[0113] Euclidean distance is sensitive to time offset, expansion / contraction (of the model relative to the analysis window), outliers, and delay. However, normalization increases sensitivity to amplitude variations.

[0114] An analysis window with the same length as the recurring patterns stored in the pattern database can be provided. For implementations based on the k-motiflet method, this length l is determined when identifying recurring patterns.

[0115] For transient phases caused by loads longer than the time window, detection and identification based on sub-parts of the transient phase are still feasible.

[0116] Dynamic Time Warping (DTW) is more robust to expansion / contraction.

[0117] Different distances and different mechanisms and algorithms can be used to perform step S5.

[0118] Therefore, for each time window extracted in step S4, the distance is calculated using the pattern set stored in the pattern database 44.

[0119] Then, the decision function can be determined based on the set of distances.

[0120] In particular, the decision function f can be written such that:

[0121] in:

[0122] d(x i ,y) is in pattern x i The distance measured between the analysis window y and the analysis window y. For example, this distance could be the normalized Euclidean distance d described earlier. nEUC (x i ,y).

[0123] M is the number of recurring patterns stored in the pattern database 44.

[0124] S is the detection threshold. S can have a predefined value.

[0125] The value of the decision function f directly gives the desired result: - If f=0, then the analysis window compares the patterns x in those patterns stored in the pattern database. i Correspondingly, this is normal electrical behavior. Figure 4 In the example above, the method then returns to step S4 to search for new transient phenomena.

[0126] If f=1, the analysis window does not correspond to any of the patterns stored in the pattern database. Therefore, it involves anomalous electrical behavior, i.e., electric arcing.

[0127] If an electric arc is detected, step S6 of the detection processing can be triggered. This step S6 can be executed by the function processing module 47.

[0128] This step can include emergency responses such as shutting down and isolating the monitored wires: an arc indicates abnormal behavior that could generate other arcs in the future, each of which could cause localized damage to the system (including the aircraft), as previously observed. Therefore, once the first arc is detected, any potential damage must be prevented by cutting the wires.

[0129] In addition, an alarm can be triggered to alert operators who can remedy the failure (triggering possible backup systems after the failed system is shut down, actions to repair the failure, etc.).

[0130] Obviously, the present invention is not limited to the examples and embodiments described and shown, but is defined by the claims. In particular, the present invention can have various variations, which are available to those skilled in the art.

Claims

1. A method for detecting electric arcs in the circuits of an aviation system, comprising: - First stage (P1), the first stage discovers recurring patterns in the behavior of the current in the wires (3) of the circuit, and stores the recurring patterns in a pattern database (44), and - Second stage (P2), the second stage is a diagnostic stage, which includes comparing the behavior of the current on the wire (3) with recurring patterns stored in the pattern database (44), and detecting an arc when the behavior does not correspond to the recurring patterns stored in the pattern database (44).

2. The method according to the preceding claim, wherein, A time threshold (t1) triggers the switch from the first stage (P1) to the second stage (P2).

3. The method according to the preceding claim, wherein, The time threshold (t1) corresponds to a period of time since the first start-up of the aviation system, which is greater than or equal to 15 hours and less than or equal to 35 hours, preferably 25 hours.

4. The method according to any one of the preceding claims, wherein, The first stage (P1) includes the steps of acquiring measurement data related to the current, storing the measurement data to form a time series (S1), and identifying recurring patterns within the time series of the measurements (S3).

5. The method according to the preceding claim, wherein, The step (S3) of identifying recurring patterns uses the k-motiflet method.

6. The method according to any one of the preceding claims, wherein, The second stage (P2) includes the step of searching for transient phenomena in the current (S4) and the step of comparing the transient phenomena with stored recurring patterns (S5).

7. The method according to the preceding claim, wherein, Extract the time window corresponding to the transient phenomenon, calculate the distance between the time window and each of the recurring patterns in the stored recurring patterns, and then detect the presence of the electric arc based on the set of distances. The calculated distance is preferably the z-normalized Euclidean distance.

8. A computer program comprising instructions that, when executed on a data processing platform, implement the method according to any one of the preceding claims.

9. A detection device (4) for detecting electric arcs in the circuits of an aviation system, comprising: - Function discovery module (43), adapted to discover recurring patterns in the behavior of current on wire (3), and store said recurring patterns in pattern database (44), and - Functional diagnostic module (46) adapted to compare the behavior of the current on the wire with recurring patterns stored in the pattern database (44), and detect an arc when the behavior does not correspond to the recurring patterns stored in the pattern database (44).

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

Images