Detection of electrical arcs in an aeronautical system based on electrical current and voltage measurements

Abstract

The invention relates to a detection device (4) for detecting the presence of an electrical arc on an electrical line (3) of an electrical circuit of an aeronautical system, comprising at least: - a measurement module adapted to obtain measurements of at least one current on said electrical line (3) and of an electrical voltage relative to said electrical line; - a computation module adapted to extract descriptors from said measurements; and the detection of the presence of an electrical arc, by applying a decision function, previously provided by a learning device (4'), to said descriptors.

Description

DESCRIPTION TITLE: Detection of electrical arcs in an aeronautical system by measuring electrical current and voltage TECHNICAL FIELD AND TECHNOLOGICAL BACKGROUND

[0001] The technical field is that of detecting faulty electrical arcs on an electrical line and applies in particular to the field of aeronautics and embedded systems.

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

[0003] Two types of electrical arcs can occur in an aeronautical environment: breaking arcs, present in contactors when a circuit is opened or closed, and fault arcs, which can appear unexpectedly on any component of the electrical system. The latter can cause severe damage to equipment, systems, and even the aircraft structure. This is therefore a critical 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] In current aircraft systems, these problems are partly resolved, or mitigated, by several factors: The distributed voltage levels are generally at most 230 V to 400 V, which reduces the impact of electric arcs; The distribution of an alternating (AC) waveform also helps to promote the self-extinction of the arc at each half-period, when the voltage passes through 0; In continuous mode, voltage levels are generally lower, on the order of 28 V; Passive protection measures can be used to limit the consequences of potential fault arcs: selection of materials resistant to these phenomena, spacing of various components, etc. Finally, active protection measures can also be implemented, such as arc detection mechanisms and line opening mechanisms that activate upon arc detection. Current active protection systems can detect most parallel arcs but not series arcs, for which no active protection is currently deployed.

[0006] However, recent work on aircraft electrification proposes a distribution of direct current (DC) waveforms with high voltage levels (so-called HVDC networks, for "High Voltage Direct Current"). Voltage levels can reach the kilovolt range. This paradigm shift challenges the passive protection strategy mentioned earlier.

[0007] Indeed, climate change is a major concern for many legislative and regulatory bodies worldwide. Various restrictions on carbon emissions have been, are being, or will be adopted by different countries. In particular, an ambitious standard applies to both new types of aircraft and those already in service, requiring the implementation of technological solutions to bring them into compliance with current regulations. Civil aviation has been actively working for several years now to contribute to the fight against climate change.

[0008] Technological research efforts have already led to significant improvements in the environmental performance of aircraft. The Applicant takes into account factors impacting all phases of design and development to obtain aeronautical components and products that are less energy-intensive, more environmentally friendly, and whose integration and use in civil aviation have moderate environmental impacts, with the aim of improving aircraft energy efficiency. Consequently, the Applicant is continuously working to reduce its climate impact by employing methods and operating virtuous development and manufacturing processes that minimize greenhouse gas emissions in order to reduce the environmental footprint of its business.

[0009] This research and development work therefore focuses on new generations of aircraft engines, the weight reduction of aircraft, in particular through the materials used and lighter on-board equipment, aviation biofuels and the development of the use of electrical technologies to provide propulsion.

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

[0011] There are several proposals for detecting an electrical arc on a power line in order to isolate it (by cutting the circuit). However, these proposals must comply with certain constraints: Reliability: Dangerous electrical phenomena must be detectable systematically and under all circumstances. In other words, a high true positive rate, close to 100%, is required. 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 in terms of reliability and / or robustness.

[0013] Machine learning methods can be subdivided into unsupervised learning, supervised learning, semi-supervised learning, reinforcement learning...

[0014] In the current state of the art, the most widely applied branch for detecting electrical arc faults is supervised learning. This type of learning is a group of methods designed to fine-tune a model, or decision function, capable of linking a set of inputs to a set of outputs. Using a training database in which these links are established (between inputs and desired outputs, typically called "labels"), the model is iteratively refined until it converges to a stable state that allows the prediction of a consistent output from new inputs not belonging to the training set.

[0015] Among the proposals that address arc fault detection using supervised machine learning methods, examples include patent applications WO2021 / 212891 and WO2021 / 43027, as well as articles by Q. Lu et al., "A DC Series Arc Fault Detection Method Using Line Current and Supply Voltage," in IEEE Access, vol. 8, pp. 10134-10146, 2020; Navalpakkam Ananthan et al., "Voltage Differential Protection for Series Arc Fault Detection in Low-Voltage DC Systems," Inventions, 2021; and JC Kim et al., "Series AC Arc Fault Detection Using Only Voltage Waveforms," ​​2019 IEEE Applied Power Electronics Conference and Exposition (APEC), Anaheim, CA, USA, 2019.

[0016] However, in general, these machine learning-based proposals sometimes lack representativeness. Furthermore, they do not prove capable of ensuring high levels of robustness and reliability.

[0017] Moreover, in general, detection methods only use the measurement of the electrical current at the input of the learning models.

[0018] Patent EP4081811B1, entitled "Device for detecting a fault with a common-mode voltage measuring element in an electrical network and power grid," describes a mechanism for detecting and locating the location of an arc fault. However, this mechanism can only function for arc occurrences in specific regions of the power line. DESCRIPTION OF THE INVENTION

[0019] The invention aims to improve the state of the art. In particular, a method is proposed to improve the robustness and reliability rates of machine learning methods for detecting electric arcs.

[0020] In particular, it allows these electric arcs to be distinguished from transient phenomena related to the load, and therefore significantly reduces the false positive rate. In other words, the invention makes it possible to greatly increase the true positive rate and also greatly decrease the false positive rate.

[0021] To achieve this, it is proposed to use both electric current and voltage measurements, both in the learning phase and in the electric arc detection phase.

[0022] More specifically, a method for detecting electric arcs in an electrical circuit of an aeronautical system is proposed, the method being implemented by computer and comprising: - a first phase of learning a decision function on a training power line of a training power circuit approximating said power circuit, said first phase comprising: - a step of obtaining initial measurements of at least one intensity on said training power line and of an electrical voltage relative to said training power line; - a step of extracting first descriptors from said first measurements of electrical intensity and voltage; - a step of determining said decision function from at least some of said first descriptors; - a second phase of detecting the presence of an electric arc on an electrical line of said electrical circuit, comprising - a step of obtaining second measurements of at least one intensity on said power line and of an electrical voltage relative to said power line; - a step of extracting second descriptors from said second measurements; - a step of detecting the presence of an electric arc at the level of said power line by applying said decision function to said second descriptors

[0023] Thus, the invention advantageously allows the use of both source voltage measurements and electric current measurements to implement the detection of electric arcs.

[0024] 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: said first descriptors comprise at least one descriptor function of a combination of said current intensity and voltage measurements; - said descriptor is a temporal correlation between current intensity and voltage measurements, over a time window; - said electrical voltage is measured across the terminals of a source of said electrical circuits; - said learning phase further includes a step of labeling said first measurements from an arc voltage measurement, said first descriptors being extracted according to the labels determined for said first measurements

[0025] Another object relates to a device for detecting the presence of an electric arc on an electrical line of an electrical circuit in an aeronautical system, comprising at least - a measurement module adapted for obtaining measurements of at least one intensity on said power line and of an electrical voltage relative to said power line; - a suitable calculation module to implement an extraction of descriptors from said measurements; and a detection of the presence of an electric arc by applying a decision function previously provided by a learning device on said descriptors.

[0026] Another object concerns a learning device for a decision function for detecting the presence of an electric arc on a power line of an electrical circuit in an aeronautical system, comprising at least: - a measurement module adapted for obtaining measurements of at least one current on a training power line of a training power circuit approximating said power circuit, and of an electrical voltage relative to said training power line; and - a suitable calculation module to implement an extraction of descriptors from said electrical intensity and voltage measurements and a determination of said decision function from at least a part of said descriptors.

[0027] Another aspect of the invention relates to an aerial vehicle, such as an aircraft, comprising at least one detection device as previously described.

[0028] Another aspect of the invention relates to a computer program comprising instructions for implementing a process as previously described when executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Other aspects, objectives, advantages, and features of the invention will become clearer upon reading the following detailed description of preferred embodiments thereof, given by way of non-limiting example, and made with reference to the accompanying drawings in which: - Figure 1 schematically represents a flowchart of a process according to one embodiment; - Figure 2 schematically illustrates a functional architecture allowing the implementation of a detection device according to one embodiment; - Figures 3a and 3b illustrate examples of the behavior of the electric current intensity and arc voltage; - Figure 4 illustrates a possible system for implementing the proposed process, according to one embodiment; - Figure 5 illustrates examples of the behavior of electric current intensity and arc voltage. DETAILED DESCRIPTION OF SPECIFIC METHODS OF IMPLEMENTATION

[0030] The proposed method aims at detecting electrical arcs on an electrical line of a circuit in an aeronautical system. This aeronautical system includes various components on board an aeronautical vehicle, or aircraft, such as an airplane, a drone or UAV (for "Unmanned Aerial Vehicle"), a helicopter, etc.

[0031] As illustrated in Figure 1, the proposed process is based on two distinct phases: a PI learning phase (or training phase, the two terms being equivalent) and a P2 detection phase for the presence of an electric arc. This P2 phase can also be called the prediction or exploitation phase since it uses the model, or the decision function determined during the PI learning phase.

[0032] The PI learning phase aims to determine a numerical model capable of distinguishing between arcing regimes and nominal regimes, based on electrical measurements of current and voltage (measured across the power supply terminals). This nominal behavior excludes any possible failure.

[0033] This model constitutes a decision function because it allows us to determine whether the data provided as input is indicative of the presence of an arc or not.

[0034] The purpose of the P2 diagnostic phase is to classify the input data into arcs and nominal regimes.

[0035] In particular, it allows us to classify normal electrical artifacts corresponding to normal variations in electrical charges from artifacts corresponding to faulty electrical arcs that are problematic for the condition of the electrical circuit, the components connected to it, or even for the aircraft.

[0036] This allows us to improve the detection rate of fault arcs (true positive rate) but also to considerably reduce the false positive rate.

[0037] These two phases, Pl, P2 can follow each other in time, the exploitation and diagnostic phase following the learning phase.

[0038] It is proposed that the PI learning phase be preferably carried out in a laboratory, with a data acquisition chain as close as possible to the reality of the actual system to be monitored. In other words, for the PI learning phase, a power line of a training circuit is considered, approximating the electrical circuit to be monitored. This training circuit must therefore have characteristics as close as possible to the circuit onboard the aircraft, particularly in terms of power supply, current waveform, electrical charge, etc.

[0039] The P2 diagnostic phase can be implemented by a device onboard an aircraft. This detection device can be integrated into the monitoring systems typically found in the aircraft's electrical core, such as solid-state power controllers (SSPCs). An SSPC consists of one or more solid-state switching devices and associated solid-state circuits for protection, control signal actuation, and status information provision.

[0040] Figure 2 illustrates one embodiment among several possible ones.

[0041] In the embodiment of Figure 2, the detection device 4 is positioned upstream of the power line 3, close to the current source 1.

[0042] Preferably, the detection device 4 is designed to measure an electrical current and voltage across the terminals of the source.

[0043] It has been observed that when an electric arc occurs in series with the load, it causes a drop in the electric current and a rise in the voltage measured across its terminals. If the voltage exceeds the critical value of approximately 15V, the presence of a fault arc is certain. Therefore, it is both possible and advantageous to measure the arc voltage across the source terminals, since the arc voltage cannot be measured during operation. During training, the voltage across the source terminals can also be considered to create an effective training dataset (even though the arc voltage could be measured in the laboratory), allowing for the best possible approximation of flight situations within the P2 diagnostic phase.

[0044] Figures 3a and 3b illustrate examples of the behavior of electric current intensity and arc voltage (which can only be measured in the laboratory under intentionally induced arcs for experimental studies). It can be seen that the voltage rises corresponding to electric arcs are correlated over time with drops in current intensity.

[0045] This voltage, known as arc voltage, is impossible to measure in a real system because the location of the arc fault is uncertain. Therefore, most state-of-the-art arc fault detection algorithms rely solely on measuring the electrical current to identify arc occurrence. The goal is then to detect drops in electrical current along the monitored line.

[0046] However, the inventors observed during extensive experimental studies that the voltage measurement across an unregulated source was disrupted by the presence of a fault series arc. Specifically, this disruption resulted in a rise in the measured voltage, while the direction of current variation dropped immediately after the arc appeared.

[0047] In addition to these variations, high-frequency disturbances can also be observed. These disturbances are likely related to the dynamic variation of the series arc impedance.

[0048] It is therefore proposed to take into account a voltage measurement relative to the observed line, in addition to the current measurement. Combining these two measurements improves the understanding of the phenomena involved and, consequently, enhances the robustness and reliability of arc flash detection.

[0049] In particular, according to one embodiment, it is proposed to detect a series arc from measurements of the current and electrical voltage measured across the terminals of the source.

[0050] This arrangement, corresponding to that described in relation to Figure 2, is advantageous because current and voltage measurements across the source terminals are already commonly implemented in avionics systems for regulation or protection purposes. Furthermore, this technical solution could also allow for arc fault localization for maintenance purposes.

[0051] As previously indicated and illustrated in Figure 1, the proposed process is based on: a first phase, PI, of learning a decision function; and, a second phase, P2, of detecting the presence of an electric arc.

[0052] Figure 5 also illustrates the behavior of current intensity (top curves) and voltage (bottom curves) during the occurrence of an electric arc (dotted vertical line).

[0053] We can see the disturbance of voltage signals but also of current intensity signals.

[0054] Figure 4 illustrates a possible system for implementing the proposed method. It consists of a first subsystem, SI, corresponding to a system present in the laboratory in which the PI learning phase can be implemented, and a second subsystem S2 corresponding to the aerial vehicle (such as an aircraft) carrying the electrical circuit whose electrical line 3 is to be monitored.

[0055] In order for the PI learning phase to enable the determination of a decision function 5 which is relevant for the detection of electrical arcs on the electrical line 3 of the aerial vehicle, it is important that the laboratory system SI imitates as closely as possible the electrical characteristics of the second on-board system S2 (as previously described).

[0056] The PI training phase aims to collect measurements via a training device 4' on a training power line 3' of a training power circuit (approximately representing the power circuit to be monitored). It may include a source 1' identical or similar to the one 1 used on board the aircraft. Similarly, a load 2' simulates the load actually connected to load 2 in the onboard system S2.

[0057] In an Eli step, the learning device 4' allows obtaining measurements of at least one intensity on the learning power line and an electrical voltage relative to this power line, for example at the terminals of the source l'.

[0058] This measurement module of the learning device 4' can for example include one or more Hall effect type sensors, for example with a bandwidth of 500kHz.

[0059] In one embodiment, an arc voltage measurement is also recorded. This arc voltage allows for the unambiguous determination of the presence of an electric arc.

[0060] It is planned that measurements will be taken to cover various electrical phenomena, including the presence and absence of electrical arcs. Indeed, proper learning can only enable the determination of a decision function if the training set contains different situations, at a minimum belonging to the two classes we wish to distinguish (presence and absence of arcs). It is also beneficial to include other types of electrical disturbances to allow for the determination of a decision function that discriminates between electrical arcs and other electrical phenomena.

[0061] One process may involve acquiring measurements over a sufficiently long period, during which various electrical disturbances, including electrical arcs, are induced on the learning line 3'. These measurements can be performed over several time windows distributed throughout time.

[0062] The different measurements (intensity, voltage and arc voltage) are stored while preserving the temporal correlations (i.e., by dating each measurement).

[0063] These measurements can, for example, be sampled and digitized to form a time series. This time series associates, at each instant t, at least one current measurement, one voltage measurement, and (optionally) one arc voltage measurement.

[0064] According to one embodiment, the process includes a step E12 for labeling the measurements. This step consists of assigning a class, or label, to the measurements taken (or at least to a portion of these measurements). This label can be binary: presence or absence of an electric arc.

[0065] This step can be performed manually, with a human operator indicating for each current / voltage measurement pair whether it corresponds to an arc or not. It can also be deduced from the times associated with each measurement, since the arcs are artificially induced, allowing the precise moments of impact to be determined.

[0066] According to one embodiment, an original labeling mechanism is proposed, based on the physics of electric arcs and the measurement of arc voltage also collected during the Eli measurement step.

[0067] We use the arc voltage measurement as a reference to automatically label the arc current measurement. We can then automatically label the current and voltage measurements corresponding to the same time instant.

[0068] To achieve this, according to one embodiment, it is possible to scan all the measurements of the plurality of measurements in order to precisely identify the arc start and end information within the arc voltage measurements. The relevant portion of each measurement can thus be selected, along with the corresponding current and voltage.

[0069] It is considered that, in the case of electrical signatures of fault arcs, the only information that can guarantee the presence of an electric arc is the measurement of the voltage across its terminals (the arc voltage), since this value will be zero in the absence of discharge and greater than a certain value only in the presence of an arc. As a reminder, this arc voltage information is not measured (because it is unavailable) during the P2 detection phase. It is only accessible and used during the PI design phase, in the laboratory.

[0070] Arc voltage measurement is used exclusively to label current and voltage measurements across the terminals of the source l'.

[0071] Furthermore, during this E12 labeling step, the characteristics of the phenomenon to be detected must be clearly defined. They represent the specific nature of the defect that we are trying to detect.

[0072] In an example implementation, the characteristics of the arcs to be detected are as follows: - an arc duration > 1 ms; - an increase in arc voltage from 0 to a minimum of 10 V.

[0073] In an example implementation, the characteristics of the windows labeled as arcs are as follows: - a window duration of between 1 and 10 ms; - presence of a negative time lag with a duration proportional to the length of the window; - absence of arc termination signature (correlated with a drop in arc voltage towards 0 V).

[0074] Of course, these values ​​are provided as examples only, and other values ​​may be used in other implementation examples.

[0075] In the current state of the art, labeling is a delicate and time-consuming operation. However, in the described process, this step E12 can be automated, ensuring the same result as manual implementation while optimizing the time spent on it.

[0076] To achieve this, in this embodiment, the labeling step E12 includes, for each measurement and for each time interval t of the measurement considered, the following sub-steps: - when the measured arc voltage U ar c exceeds a predetermined threshold value Us for the first time (that is, at the first occurrence of the following relation: U arc (t > U s and U arc t — 1) < U s , where t is the interval (or sample) considered, a recording step of the interval considered as the starting point of the electric arc (we then define td = t); and, - when the measured arc voltage U arc, falls below said threshold value U for the first time s predetermined (that is, at the first occurrence of the following relation: U arc (t < U s and U arc t — 1) > U s ; - a recording step of the interval preceding the considered interval (i.e., t-1) as the end point of the electric arc (t is then defined as e = t - 1).

[0077] The measured arc voltage U arc is compared to the threshold arc voltage U sIf the measured arc voltage is higher than the threshold arc voltage, previous arc voltage values ​​are checked. If none of these previous values ​​were above the threshold value, this means that the arc voltage has crossed above the threshold value for the first time, signaling the start of the arc. Otherwise, no action is taken. Similarly, if the measured arc voltage is below the threshold value and the threshold value was above the threshold value, this signals that the arc voltage has fallen below the threshold value, signaling the end of the arc.

[0078] In one embodiment, for each measurement, the labeling is associated with a measurement window on which the learning is performed, said window being given by the interval:

[0079] With : Tmax, the duration of the maximum window, Tmin, the minimum window duration, - P g The "left step": its value is normalized for the duration of the studied window and is between 0 and 1; Pd, the "right step": we have Pd=lP g This is also a normalized value for the duration of the window studied and is between 0 and 1.

[0080] According to this embodiment, it is from the data of this interval that the descriptors are extracted in a step E13 of extraction of descriptors from the measurements of electrical intensity and voltage.

[0081] Descriptors are characteristics derived from measurements of current intensity and electrical voltage. Their purpose is to extract, or "capture," semantic content capable of differentiating situations involving electrical arcs from those without. Multiple descriptors are preferentially used to maximize the usable semantic content.

[0082] Several types of descriptors can be used.

[0083] It is possible to use an unsupervised learning method, such as those based on autoencoders.

[0084] It is also possible to use descriptors corresponding to physical or statistical quantities (which in themselves carry a physical meaning). These descriptors may relate to, in particular: - The frequency content, for example via a time-frequency transform of the Fourier transform type; - Statistical moments, such as the mean, standard deviation, skewness and kurtosis; - Temporal analyses, etc.

[0085] These descriptors can relate to electrical intensity and / or electrical voltage (for example, across the terminals of the source).

[0086] Also, according to one embodiment, the descriptors include at least one descriptor that is a function of a combination of intensity and voltage measurements.

[0087] For example, such a descriptor could be a temporal correlation between intensity measurements and voltage measurements, over a time window.

[0088] In one embodiment, the PI learning phase further includes a descriptor ranking and selection step. Indeed, not all extracted descriptors necessarily possess characteristics that allow for optimal separation of the two existing window classes ("arcs" and nominal or "non-arcs"). Thus, analyzing the importance and level of correlation of the descriptors with the label associated with each window can prove useful in improving the performance of the decision function.

[0089] An inappropriate descriptor can indeed disrupt prediction and reduce (sometimes significantly) the overall performance of the decision function. Therefore, descriptor selection can generally lead to better learning performance, notably through increased efficiency, reduced computational cost, and improved performance of the automatic arc detection model. This approach is particularly advantageous when dealing with a large number of descriptors.

[0090] The objective of the descriptor selection step is therefore to maximize relevance while minimizing information redundancy. In one example implementation, this step can be carried out using a supervised filtering method.

[0091] The ranking of descriptors can, for example, be performed using a Gram-Schmidt orthogonalization method, while the selection can be carried out using an incremental approach. Of course, other methods can also be used for ranking or selection.

[0092] Although optional, this step contributes to improving the decision function obtained at the end of the PI learning phase.

[0093] The PI learning phase also includes an E14 step of determining the decision function from at least some of the descriptors, for example after a prior sorting and selection step.

[0094] From the descriptors X extracted from the intensity and voltage measurements, step E14 aims to determine a decision function f(X) allowing to discriminate the signals corresponding to the presence of an electric arc, and those corresponding to the absence of an electric arc (or "nominal" case).

[0095] In one embodiment, this step E3 comprises a plurality of iterations of the following two substeps: - a design substep, by learning, of a classifier; - a sub-step of validating the classifier based on the labeled data.

[0096] These sub-steps are repeated until the most efficient classifier is obtained, that is, for example, the one that minimizes the classification error below a certain threshold on a database of untrained data, thus ensuring a generalization capacity of the classifier.

[0097] This database can be constructed in the same way as the training database used for step E13 of descriptor extraction, but was not used for this step E13. The aim is to test the ability of the chosen descriptors to classify untrained situations.

[0098] The decision function f( ) is then determined from this / these classifier(s).

[0099] Thus, the determination of a classifier allows, through learning (corresponding to the plurality of iterations of the sub-steps described previously), to obtain a high-performing model guaranteeing effective decision-making.

[0100] Several supervised classification methods can be used to implement this step E14. The choice of method may depend on the desired detection time: some methods, while more efficient, may have higher computational complexity. Preferably, the classification method adopted for this step is that of Support Vector Machines (SVMs), which is well-known to experts in the field and therefore not detailed here.

[0101] In an alternative embodiment, the classification method can be chosen from the following options: - Artificial neural networks (simple, convolutional, recurrent...); - Hidden Markov models; - K-nearest neighbors method; - Decision trees; - Ensemble learning (combination of two methods from, in particular, those previously listed).

[0102] The decision function f( ) can be stored and used for an S2 system onboard an aircraft.

[0103] As we have seen, the decision function is determined during the PI learning phase on a laboratory SI system that imitates and replaces the embedded SI system. It is then possible to determine a set of decision functions f() for different embedded systems S2 by reconfiguring the SI simulation system. The appropriate decision function f() is then selected to be transmitted to the embedded S2 system.

[0104] At the end of the PI learning phase, the decision function (and therefore the detection strategy) is fixed.

[0105] Thus, the descriptors selected during this PI learning phase are those that will be used during the P2 detection phase.

[0106] During the P2 detection phase, these descriptors are extracted from the measured signals (for example, via signal processing techniques) and used as inputs to the decision function. The latter will then assign the signal a category: either "arc" or "nominal regime" (i.e., absence of electric arc).

[0107] This P2 detection phase consists of detecting the presence of an electric arc on a monitored electrical line 3 of an electrical circuit on board an aerial vehicle.

[0108] This P2 detection phase implements "mirror" steps to the PI learning phase, but applied to the monitored line 3 (and not to the learning line 3').

[0109] In addition, unlike the PI learning phase where arc voltage is an available measurement, in real-world conditions (i.e., in an aerial vehicle and outside the laboratory), during the P2 detection phase, only current and voltage measurements across the source terminals are available.

[0110] In particular, the detection phase P2 includes a step E21 of obtaining measurements of at least one intensity on said power line 3 and of an electrical voltage relative to the power line, for example at the terminals of the on-board source 1.

[0111] This E21 step is preferably performed continuously and in real time to allow for rapid detection of the presence of an arc. Alternatively, it can be performed after triggering a "diagnostic" operating mode, which can be triggered periodically or manually.

[0112] The measurements can be obtained by a measurement module of a detection device 4, which can be very similar to that of the learning device, 4', used in the laboratory during the PI learning phase. Of course, it cannot, however, obtain the arc voltage.

[0113] Similar to the PI learning phase, the P2 detection phase also includes a step E22 for extracting descriptors from current and voltage measurements. This step can be implemented by a calculation module of the detection device 4.

[0114] As previously mentioned, the descriptors extracted during this E22 step are identical to those selected during the PI learning phase (the values ​​are, however, obviously different since these same descriptors are applied to different signals, intensity and voltage).

[0115] The descriptors can be extracted continuously, as current and voltage measurements are acquired.

[0116] The detection phase also includes a detection step E23 of the presence of an electric arc, at the level of the monitored power line, by applying the decision function, determined during the PI learning phase and previously provided by the learning module 4, to the descriptors extracted during step E22. This step can also be implemented by a calculation module of the measurement device 4.

[0117] During operation of the monitored line 3, the measuring device 4 can immediately obtain voltage and current measurements, allowing the extraction of selected descriptors and subsequent application of the decision function. Thus, in the presence of an electric arc, the decision function can immediately detect it. In other words, the proposed method can enable the detection of an electric arc in real time (within computation time).

[0118] The process may include a detection processing step when an electric arc is detected.

[0119] This step may include emergency responses such as shutting down and isolating the supervised power line: an electrical arc indicates abnormal behavior that can generate further electrical arcs in the future, each potentially damaging the system locally, including the aircraft, as previously discussed. It is therefore essential to prevent any potential damage by shutting down the power line as soon as the first electrical arc is detected.

[0120] Also, an alert can be triggered to notify a human operator who can compensate for this fault (activation of a possible redundant system following the shutdown of the faulty one, interventions to troubleshoot the fault, etc.)

[0121] In one embodiment, it is therefore possible to interrupt the current in the electrical line 3 on which the detection is performed using a current interruption device. The proposed method includes, when an electric arc is detected on an electrical line 3 during the detection step E23, a step of sending a current interruption instruction to the interruption device for the line 3 in question. This makes it possible to avoid, or at least limit, the damage associated with the formation of an electric arc on the current line 3 in question.

[0122] The quality of detection depends on the quality of the training, which itself depends on the quality of the data and its labeling. Labeling can be automated using a plurality of measurements (generally performed in the laboratory), each measurement comprising a time-correlated measurement of the arc voltage on the one hand, and of the current and voltage across the source on the other.

[0123] Another contribution is the inclusion of voltage measurement in addition to current measurement. Not only can these two measurements be used (to extract descriptors for each type of measurement), but they can also be combined to exploit semantic information related to this combination, particularly their correlation.

[0124] Of course, the present invention is not limited to the examples and embodiment described and illustrated. In particular, it is susceptible to numerous variations accessible to those skilled in the art.

Claims

DEMANDS 1. A method for detecting electric arcs in an electrical circuit of an aeronautical system, the method being implemented by computer and comprising: a first phase (PI) of learning a decision function on a training electrical line (3') of a training electrical circuit approximating said electrical circuit, said first phase (PI) comprising: a step (Eli) of obtaining initial measurements of at least one intensity on said training power line and of an electrical voltage relative to said training power line; a step (E13) of extracting first descriptors from said first measurements of electrical intensity and voltage; a step (E14) of determining said decision function from at least a part of said first descriptors; a second phase (P2) of detecting the presence of an electric arc on an electrical line (3) of said electrical circuit, comprising a step (E21) of obtaining second measurements of at least one intensity on said power line (3) and of an electrical voltage relative to said power line; a step (E22) of extracting second descriptors from said second measures; a detection step (E23) of the presence of an electric arc at the level of said power line by applying said decision function on said second descriptors.

2. A method according to claim 1, wherein said first descriptors comprise at least one descriptor that is a function of a combination of said current and voltage measurements.

3. A method according to the preceding claim, wherein said descriptor is a temporal correlation between current and voltage measurements, over a time window.

4. A method according to any one of the preceding claims, wherein said electrical voltage is measured across the terminals of a source of said electrical circuits.

5. A method according to any one of the preceding claims, wherein said learning phase further comprises a labeling step (E12) of said first measurements from an arc voltage measurement, said first descriptors being extracted according to the labels determined for said first measurements.

6. Computer program comprising instructions to implement a method according to one of the preceding claims when executed by processors of a detection device (4) and a learning device (4').

7. Device for detecting (4) the presence of an electric arc on an electrical line (3) of an electrical circuit of an aeronautical system, comprising at least a measurement module adapted for obtaining measurements of at least one intensity on said power line (3) and of an electrical voltage relative to said power line; a calculation module adapted to implement an extraction of descriptors from said measurements; and a detection of the presence of an electric arc by applying a decision function previously provided by a learning device (4') on said descriptors.

8. A learning device (4') for a decision function for detecting the presence of an electric arc on an electrical line (3) of an electrical circuit of an aeronautical system, comprising at least: a measurement module adapted for obtaining measurements of at least one intensity on a training electrical line (3') of a training electrical circuit approximating said electrical circuit, and of an electrical voltage relative to said training electrical line; and a calculation module adapted to implement an extraction of descriptors from said measurements of intensity and electrical voltage and a determination of said decision function from at least a part of said descriptors.

9. Aerial vehicle, such as an aircraft, comprising at least one detection device (4) according to claim 7.

10. System comprising a detection device according to claim 7 and a learning device (4') according to claim 8.

Images