Method and system for determining the tripping range of an electrical circuit switching device

Abstract

Method and system for determining a tripping range of an electrical circuit switching device The invention relates to a method and system for determining a tripping range of an electrical circuit switching device (4) adapted to supply an electrical installation (12).The system (2) includes an electronic computing module configured (6) to receive data from at least one vibration sensor (16) integrated into said switching device (4) and configured to select, from spectrograms calculated from acquired time-domain vibration signals, a predetermined subset of operational characteristics; apply a parameterized prediction model using supervised machine learning to the values ​​of the operational characteristics to obtain an estimated breaking current value; and determine, based on the estimated breaking current value, a tripping range for the switching device from among: normal tripping, overload tripping, and short-circuit tripping. Figure 1 for the abbreviation.

Description

Title of the invention: Method and system for determining the tripping range of an electrical circuit switching device

[0001] The present invention relates to a method and a system for determining a triggering range of an electrical circuit switching device.

[0002] The invention lies in the field of electrical installation safety, and in particular of the interruption of electrical supply, commonly called disjunction, in the event of the presence of an electrical fault in an installation.

[0003] As is known, various defects can occur in an electrical installation, supplied by an electricity supply network, and jeopardize the safety of the equipment connected to the installation, or even in some cases present a risk of fire or a physical danger to an operator.

[0004] To ensure the safety of electrical installations, switching devices, commonly called circuit breakers, are used, configured to perform a very rapid power cut following the detection of a current exceeding a nominal current threshold.

[0005] Conventionally, a switching device comprises fixed and moving contacts. The moving contact is adapted to move between an open position, in which it is separated from the fixed contact and thus the flow of electric current is interrupted, and a closed position, in which the moving contact is pressed against the fixed contact and the flow of electric current is possible. The switching device also includes a triggering mechanism that causes the contacts of the switching device to open, as well as a reset lever.

[0006] The reset lever is generally accessible to an operator on one side of the housing containing the switching device and can be moved between three positions: an "open" position (corresponding to the open position of the contacts), a "tripped" position (corresponding to the open position of the contacts following a circuit breaker trip), and a "closed" position (corresponding to the closed position of the contacts). Moving the lever from the "open" position to the "closed" position and vice versa is done manually by an operator, and moving it from the "closed" position to the "tripped" position is triggered automatically by the activation of a power outage.

[0007] Compact molded-case circuit breakers are known in particular, also These devices, called MCCBs (for "Molded Case Circuit Breaker"), include a magnetothermal module. The thermal protection cuts off the power supply in case of an electrical circuit overload. The magnetic protection rapidly cuts off the electrical circuit in case of a fault on a load that results in a short-circuit current.

[0008] For such switching devices, one of the problems that arises is to know the cause of the electrical power outage.

[0009] In particular, it is useful to know, when the lever is in the "triggered" position, whether the cause of the triggering is an overload fault or a short circuit fault, because the corrective measures to be taken, and the safety risks for the operator differ depending on the situation.

[0010] Most switching devices based on a magneto-thermal protection module do not provide information on the cause of the tripping.

[0011] There are electronic triggers which make it possible to identify the causes of triggering, however there is a need to propose a less expensive means of identifying the causes of triggering by proposing an additional function to the magneto-thermal trigger.

[0012] There is therefore a need to determine the tripping range of a switching device, in order to be able to indicate it in a way that improves the safety of the electrical installation and of the operator who intervenes following a tripping.

[0013] The triggering domains comprise respectively:

[0014] - a normal triggering, on command from an operator, for example by actuation of the lever to switch from closed position to open position;

[0015] -an overload triggering, due to the presence of overconsumption of electricity by at least one of the loads supplied by the electrical installation;

[0016] -a fault triggering, due to the presence of an electrical fault, for example a short circuit, in the switching device.

[0017] The tripping domains are associated with breaking current domains, defined according to the nominal current threshold.

[0018] The invention aims to determine the triggering range of the switching device and to indicate it to the operator.

[0019] To this end, the invention proposes, according to one aspect, a method for determining a tripping range of an electrical circuit switching device adapted to supply an electrical installation comprising at least one load, the switching device having an open position in which the flow of an electric current in said electrical circuit is interrupted, and a closed position in which an electric current flows in said electrical circuit, the tripping of the transition from the closed position to the open position being effected either following

[0020]

[0021]

[0022]

[0023]

[0024]

[0025] a command, either following the detection of a breaking current flowing in said switching device, the breaking current having a level exceeding a nominal current threshold. This process is implemented by an electronic computing module configured to receive data from at least one vibration sensor integrated into said switching device, and it includes, in an operational phase, the steps of: - obtaining at least one time-domain vibration signal, over a chosen acquisition period, from data from said at least one vibration sensor, with a given sampling frequency, following passage through said opening position, - for each of the time-domain vibration signals, transformation of the time-domain vibration signal into a spectrogram represented in a time-frequency representation space, - selection of a predetermined subset of spectrogram characteristics, called the operational characteristics subset, - application of a parameterized prediction model using supervised machine learning, with the values ​​of said subset of operational characteristics being provided as input to said parameterized prediction model, to obtain a class corresponding to an estimated value of the cutoff current, - determination, based on the estimated breaking current value and said nominal current threshold, of a tripping range for the switching device among: a normal tripping by command, an overload tripping following an overconsumption of electricity by at least one load of the electrical installation and a tripping following a short circuit. Advantageously, the proposed method makes it possible to estimate a cutting current from signals obtained by a vibration sensor integrated into the switching device, which makes it possible to deduce the triggering range of the switching device. The method for determining a triggering domain according to the invention may have one or more of the characteristics below, taken independently or according to all acceptable combinations. Determining the tripping range of the switching device involves comparing the estimated breaking current value to the nominal current threshold, and -when the estimated breaking current value is lower than the nominal current threshold, the tripping is a normal tripping, - when the said estimated breaking current value is greater than the threshold of If the nominal current is less than the nominal current threshold multiplied by a predetermined multiplication factor, the tripping is an overload tripping.

[0026] - when said estimated breaking current value is greater than the threshold of Nominal current multiplied by a multiplication factor, the tripping is a fault tripping.

[0027] The method further includes a display of an indication relating to the determined triggering range on the human-machine interface of the switching device.

[0028] Obtaining at least one time-domain vibration signal involves applying a time delay of predetermined duration after the switching device is triggered.

[0029] The method includes a learning phase prior to said operational phase, comprising

[0030] - an acquisition of a plurality of time-domain vibration signals, each time-domain vibration signal being associated with a known cutoff current value,

[0031] -a transformation of each time-domain vibration signal into a spectrogram and a storage of said spectrograms and associated cutoff current values ​​in a library of learning spectrograms.

[0032] The learning phase further includes a recursive feature elimination step, comprising a recursive application of steps of:

[0033] - a) learning the parameters of said prediction model taking as input feature values ​​of a subset of training features of the calculated spectrograms, said prediction model being parameterized to provide, from the feature values ​​of said subset of training features, a classification into a class associated with an estimated current value,

[0034] - b) evaluation of an importance score, in relation to the result of the model of prediction, of each characteristic of said subset of learning characteristics,

[0035] -c) elimination of the least important scoring characteristic of said sub- set of learning characteristics,

[0036] -d) updating the subset of learning features and repetition of steps a) to d) until the subset of learning features is empty.

[0037] The method further comprises a step of obtaining a ranking of the features by importance based on the order of elimination of the features of said subset of training features and selecting a predetermined number N of the highest importance rank features for form the subset of operational characteristics.

[0038] The method further includes an optimization of the sampling frequency, the acquisition time and a time offset to be applied when obtaining the time-domain vibration signals to be processed.

[0039] According to another aspect, the invention relates to a computer program comprising software instructions which, when executed by a programmable electronic computing module, implement a method for determining a triggering range of an electrical circuit switching device as briefly described above.

[0040] According to another aspect, the invention relates to a system for determining a triggering domain comprising a switching device including at least one vibration sensor and an electronic computing module configured to implement a method for determining a triggering domain as briefly described above.

[0041] Advantageously, the system for determining a tripping domain is configured to implement a method for determining a tripping domain of an electrical circuit switching device as briefly described above, according to all the variants envisaged.

[0042] Other features and advantages of the invention will become apparent from the description given below, by way of example and not limitation, with reference to the accompanying figures, among which:

[0043] [Fig-1] [Fig.1] is a schematic representation of the functional modules of a system for determining a triggering domain according to an embodiment;

[0044] [Fig.2] [Fig.2] illustrates examples of vibration recordings in the time domain for a normal tripping and for a fault tripping of the switching device;

[0045] [Fig.3] [Fig.3] illustrates examples of spectrograms corresponding to the enre deposits of the [Fig.2];

[0046] [Fig.4] [Fig.4] is a synoptic diagram of the main steps in a determination process mination of a triggering domain in an operational phase;

[0047] [Fig. 5] [Fig. 5] is a synoptic diagram of the main steps in a determination process. mination of a triggering domain in a learning phase prior to the operational phase.

[0048] Fig. 1 schematically illustrates the main functional blocks of a system 2 for determining a triggering domain according to one embodiment.

[0049] System 2 includes an electrical circuit switching device 4, also called circuit breaker thereafter, and an electronic computing module 6.

[0050] The circuit breaker 4 comprises a thermal-magnetic module 8 configured to trip a circuit interruption of the electrical power supply, schematically represented by a wire 10, of an electrical installation 12, the electrical installation 12 comprising at least one load 14, the tripping occurring in the event of an overload or electrical fault. The electrical power source is, for example, a three-phase or single-phase power supply network, not shown.

[0051] Various embodiments of magnetothermic module 8 are known in the field of electrical current interruption devices (in English, 'circuit breakers').

[0052] In one embodiment, the switching device 4 is a compact MCCB (for "Molded Case Circuit Breaker") type device, adapted to operate for currents ranging from 10 Amperes to 630 Amperes.

[0053] The circuit breaker 4 includes, in addition to the magnetothermal module 8, one or more vibration sensors 16.

[0054] Preferably, the vibration sensor or each vibration sensor 16 is a 3-axis accelerometer.

[0055] Advantageously, the vibration sensor 16 is suitable for capturing a signal representative of the vibrations of the switching device. Indeed, in the event of an opening being triggered, for a transition from the closed position to the "open" position (e.g., interruption of the flow of electric current in the electrical supply circuit of the electrical installation), whether by command or by tripping, a mechanical shock occurs which excites the parts of the switching device.

[0056] In the event of the opening of the switching device 4 being triggered by an overload or fault, a shock wave also occurs due to the intensity of the breaking current. This shock wave induces additional vibrations, which are superimposed on the mechanical vibrations of the switching device.

[0057] The vibration sensor(s) 16 are preferably positioned close to the magnetothermal module 8.

[0058] Preferably, each vibration sensor 16 is positioned so as to be easily mounted and removed on a new or existing product. Advantageously, this allows for the improvement of existing circuit breaker products.

[0059] For example, in the case of a three-phase current, the switching device 4 includes a vibration sensor 16 per phase.

[0060] By way of example, [Fig.2] illustrates two graphs G1 and G2, corresponding respectively to the case of opening in normal tripping (graph Gl) or in fault tripping (graph G2), where the breaking current is on the order of 12xln, L being the nominal current threshold.

[0061] Each graph Gl, G2 illustrates the evolution curve, in the time domain, of the time-domain vibration signal captured by a vibration sensor, the x-axis representing time expressed in samples (the numbers representing the measurement sample numbers) and the y-axis representing a measured acceleration value expressed in g.

[0062] These time signals are transformed by applying a discrete Fourier transform to obtain spectrograms, in the spectral time-frequency domain.

[0063] As can be seen in [Fig.2], the two graphs Gl, G2 have similar profiles.

[0064] By way of example, [Fig. 3] illustrates two graphs G3, G4; comprising spec trograms, represented in the time domain (abscissa) and frequency domain (ordinate) corresponding respectively to the curves of graphs Gl and G2 of [Fig.2],

[0065] The proposed method implements tools developed to discriminate the triggering domains, and to estimate the cutoff current from the spectrograms of the acquired time-domain vibration signals.

[0066] The circuit breaker 4 further preferably includes a human-machine interface 15, for example a screen or one or more indicator lights, allowing the determined tripping range to be displayed. For example, if the human-machine interface 15 includes one or more indicator lights, a different colored light is provided for the tripping range to be illuminated, for example green for a normal trip, orange for an overload trip, and red for a fault trip.

[0067] Advantageously, the indication of the triggering domain makes it possible to guide maintenance work, for example a rebalancing of loads in case of overload or an intervention on the electrical installation in case of short circuit.

[0068] To carry out the determination of the triggering domains, the electronic computing module 6 includes an electronic memory unit 18 and a computing processor 20, configured to implement software 22 comprising program instructions which, when executed by the computing processor, perform a method for determining a triggering domain as described below.

[0069] Alternatively, the electronic computing module 6 is a dedicated module implemented as a programmable logic component such as an FPGA (Field Programmable Gate Array) or as a dedicated integrated circuit, such as an ASIC (Application Specified Integrated Circuit).

[0070] The electronic memory 18 is adapted to store, in particular, parameters 24 of a prediction model parameterized by machine learning, adapted to provide, from a subset of spectral characteristics of a vibration signal, a estimation of a cutoff current value.

[0071] In one embodiment, the electronic calculation module 6 is integrated in the same housing as the circuit breaker 4, thus forming an electrical circuit switching device configured to self-determine an electrical circuit opening tripping range, between a normal trip, an overload trip or an electrical fault trip.

[0072] In an unrepresented variant, the electronic calculation module 6 configured to run the software 22 for determining the tripping domain of the circuit breaker 4 is remote from the circuit breaker 4. In this variant, each of the devices 4, 6 further includes a communication interface, adapted to communicate according to a given communication protocol, so as to transmit the vibration signals acquired by the vibration sensor to the electronic calculation module, the electronic calculation module 6 transmitting information relating to the determined tripping domain for display on the human-machine interface 15.

[0073] The method of determining a triggering domain of a switching device comprises two phases, which are respectively a preliminary learning phase and an operational phase of determining the triggering domain in real time.

[0074] For example, the preliminary learning phase is carried out for a given category of circuit breakers, before they are put into operation in an electrical circuit.

[0075] Thus, the preliminary learning phase is carried out by category of circuit breakers, the categories being defined by manufacturing specifications.

[0076] Fig. 4 is a synoptic diagram of the main steps of a method for determining a triggering range of an electrical circuit switching device, in the learning phase.

[0077] The method includes an acquisition 30 of the time-domain vibration signals, from one or more circuit breakers 4, for example from a plurality of circuit breakers 4 belonging to a given category of circuit breakers.

[0078] In the case where the vibration sensor is a 3-axis accelerometer, each signal consists of a series of discrete acceleration values ​​recorded at a predetermined sampling frequency. For example, the sampling frequency Fe is between 6.4 kHz and 25.6 kHz, for example, equal to 25.6 kHz. During the learning phase, the breaking current associated with each recorded vibration time signal is known, as different fault and overload current values ​​are injected into the circuit breaker for learning purposes.

[0079] Each time-domain vibration signal is acquired over a given acquisition time, and comprises a number of samples depending on the acquisition time and the sampling frequency.

[0080] The acquisition 30 is followed by a step 32 of transformation of each of the time-wave vibration signals into a spectrogram represented in a time-frequency representation space.

[0081] Transformation step 32 implements, for example, a discrete Fourier transform, preferably a Fast Fourier Transform (FFT), using the Hann window. Of course, this is just an example; other windowing variants are possible.

[0082] Each column of the spectrogram corresponds to a given time window.

[0083] In one embodiment, the method also optionally includes a normalization step 34 in which each column of the spectrogram is normalized with respect to the sum of the column values, and then the normalized spectrogram is converted to the logarithmic scale in the scaling step 36.

[0084] Steps 30 to 36 are repeated for a plurality of given cutoff current values, resulting in a library of training spectrograms, each training spectrogram being associated with a cutoff current value. The library of training programs is stored.

[0085] Each spectrogram is represented as a matrix of size LxM, comprising values, each value being relative to a time index i and a frequency index j. An element of the spectrogram identified by indices (i,j) is also called a characteristic, each characteristic having an associated value for a given spectrogram.

[0086] The method then involves implementing a step 40 of recursive feature elimination, allowing the determination of a subset of operational features, which are the features considered most relevant for carrying out the estimation of the cutoff current, by applying a chosen machine learning parameterized prediction model.

[0087] A subset of features is understood to be a subset of points (k,l) of the matrix representing the spectrogram.

[0088] Step 40 of recursive feature elimination includes iterating the following steps, for a chosen prediction model.

[0089] The prediction model is a parameterized model whose parameters are learned by supervised machine learning, among: support vector classifiers (SVCs), decision tree classifiers, logistic regression, ridge classifiers, k-neighbors classifiers, bagging classifiers, and random forest classifiers. classifier”), gradient boosting classification tree, or suitable neural networks, such as a multilayer perceptron network.

[0090] In one embodiment, a linear kernel SVC classifier is implemented.

[0091] The recursive feature elimination step 40 includes a supervised training step 42 of the chosen prediction model, by providing as input a subset of spectrogram feature values, called the training feature subset. Initially, the training feature subset is equal to the set of spectrogram features, i.e., the LxM matrix representing the spectrogram.

[0092] The application of the prediction model provides, as output, a class corresponding to an estimate of breaking current, which makes it possible to determine a classification in a tripping domain among the three tripping domains: normal tripping, overload tripping, or fault tripping, due to the presence of a short circuit in the switching device.

[0093] For example, in one embodiment, the prediction model provides as output a classification into one class among C classes, for example C=3 corresponding respectively to the three triggering domains.

[0094] Alternatively, the number C of classes is greater than 3, several classes of the classification then corresponding to the same triggering domain among the three triggering domains listed.

[0095] Given that the nominal current threshold In of the implemented circuit breaker is known, the tripping domains are associated with the breaking current level le as follows: - normal tripping: breaking current less than In; - Overload triggering: breaking current between In and Qxln; - Fault tripping: breaking current greater than Qxln

[0096] where Q is a predetermined multiplicative factor, for example Q=10.

[0097] Since the cutoff current value le associated with each spectrogram is known in the learning phase, the model parameters are set so that the application of the model provides, for each spectrogram, an estimated cutoff current value that is closest, according to the chosen classification, to the initial cutoff current value le.

[0098] Step 42 is followed by a step 44 of evaluation of an importance score associated with each element of the subset of learning features provided as input with respect to the result of the application of the prediction model.

[0099] For example, the importance score is obtained by statistical correlation, linear model coefficients, decision trees or permutation scores depending on the prediction model applied.

[0100] Following the application of step 44 of the importance score evaluation, the lowest importance score characteristic is determined in step 46, and eliminated from the subset of characteristics whose values ​​are provided as input.

[0101] The feature subset is then updated in step 48.

[0102] Steps 42 to 48 are repeated as long as at least one feature remains in the updated feature subset (test 50). In particular, the training of the chosen prediction model is repeated from the updated feature subset.

[0103] When the updated feature subset is empty, test 50 is followed by a step 52 of obtaining the ranking of the spectrogram features by rank of importance.

[0104] The order of elimination of features provides a ranking of the spectrogram features, each feature having an associated rank, for example in descending order of importance, the feature of rank 1 being the most important feature, eliminated last, and the feature of highest rank being the least important feature, eliminated first.

[0105] Step 52 is followed by a step 54 of selecting a number N of the most important features for the success of the classification. The number N is chosen, preferably N is between 1 and a few thousand, for example between 1 and 4000, more particularly preferably N is equal to 200.

[0106] When N equals 200, the features with ranks from 1 to 200 are retained and form a subset of features called the operational features subset. In other words, the N spectrogram features eliminated last form the operational features subset, selected in selection step 54.

[0107] The subset of operational characteristics is stored, in step 54, for example in the form of a coordinate table (k,l) of each selected characteristic in the spectrogram matrix, or in any other suitable storage structure.

[0108] Advantageously, selecting a number N of features reduces computational complexity.

[0109] The method further includes, optionally, a performance evaluation step 56 consisting for example of calculating a confusion matrix on the training spectrogram library, by applying the prediction model trained with the N selected spectrogram characteristics, for the classification of spectrograms which have not been used for training.

[0110] In one embodiment, a first percentage, for example 80%, of the spectrograms from the spectrogram library are used for training and a second percentage, for example 20%, for performance evaluation.

[0111] According to one variant, all the spectrograms from the library obtained for several instances of a product belonging to the same range (i.e. category of circuit breakers) except one (instance) are used for training, the remaining spectrograms (corresponding to an instance not used in training) being used for evaluation, the operation being repeated for example on all possible combinations.

[0112] Optionally, the method further includes an optimization of the sampling frequency, the signal acquisition time and a time offset time after the switching device is in the open position, various parameters being tested and the performance being evaluated in the classification performance evaluation step 56.

[0113] Thus, the inventors highlighted the possibility of reducing the sampling frequency and the signal acquisition time, while maintaining good classification performance.

[0114] Moreover, advantageously, the application of a time offset A, of the order of 20ms after the tripping of the circuit breaker also makes it possible to obtain satisfactory classification performance.

[0115] The sampling frequency parameters Fe, time offset A and acquisition time Dt obtained by optimization are stored for use in the operational phase.

[0116] The [Fig.5] is a synoptic diagram of the main steps of a process for determining a triggering range of an electrical circuit switching device, in the operational phase.

[0117] The method includes a step 60 of acquiring time-domain vibration signals, at the sampling frequency Fe previously determined, following a vibratory shock of the switching device, corresponding to the passage from the closed position to the open position following a disjunction.

[0118] The acquisition 60 is followed by a step 62 of obtaining the time signals of vibrations to be processed, by applying the time offset A and acquisition duration Dt previously obtained.

[0119] In the case where the vibration sensor is a three-axis accelerometer, for each axis, the spectrogram of the signal is calculated in step 64, for example by applying a fast discrete Fourier transform FFT.

[0120] Step 64 is followed, for each calculated spectrogram, by the selection 66 of a subset of spectrogram characteristics, the characteristic subset being the previously defined and stored operational characteristic subset.

[0121] For each calculated spectrogram, the values ​​corresponding to said subset of operational characteristics are extracted in step 66 and provided as input to a step 68 of application of the prediction model parameterized by supervised machine learning, the parameters of which were set and stored during the prior learning step.

[0122] In one embodiment, the values ​​corresponding to said subset of operational characteristics are extracted from each spectrogram, and combined into an input structure (e.g. vector or matrix) on which the parameterized prediction model is applied.

[0123] At the end of step 68, a classification is obtained providing an estimate of the value of the breaking current, as a function of the values ​​of the subset of operational characteristics provided as input, and a tripping domain is then determined, in the tripping domain determination step 70, by comparing the estimated breaking current value with the nominal current threshold In.

[0124] More specifically, if the is less than In, it is a normal triggering by command; if the is between In and Qxln, it is an overload triggering; if the is greater than Qxln, it is a fault triggering, the factor Q being for example equal to 10.

[0125] Alternatively, the triggering domain is obtained in a single step, the result of the classification carried out by the parameterized prediction model being a classification in one of the predicted triggering domains.

[0126] The triggering domain determination step 70 is followed by a step 72 of displaying an indication relating to the determined triggering domain on the human-machine interface of the switching device, for example by lighting up a corresponding colour indicator light, according to a colour code previously chosen.

Claims

1. Demands Method for determining a tripping range of an electrical circuit switching device (4) adapted to supply an electrical installation (12) comprising at least one load (14), the switching device (4) having an open position in which the flow of an electric current in said electrical circuit is interrupted, and a closed position in which an electric current flows in said electrical circuit, the tripping of the transition from the closed position to the open position being effected either following a command or following the detection of a breaking current flowing in said switching device (4), the breaking current having a level exceeding a nominal current threshold, the method being implemented by an electronic computing module configured (6) to receive data from at least one vibration sensor (16) integrated in said switching device (4),the process being characterized in that it comprises, in an operational phase, the following steps: - obtaining (62) at least one time-domain vibration signal, over a chosen acquisition time, from data from said at least one vibration sensor, with a given sampling frequency, following passage through said opening position, - for each of the time-domain vibration signals, transformation (64) of the time-domain vibration signal into a spectrogram represented in a time-frequency representation space, - selection (66) of a predetermined subset of spectrogram characteristics, called the subset of operational characteristics, - application (68) of a parameterized prediction model by supervised machine learning, the values ​​of said subset of operational characteristics being provided as input to said parameterized prediction model, to obtain a class corresponding to an estimated value of cutoff current, - determination (70), depending on the current value of estimated cut-off and said nominal current threshold, of a switching device tripping range among: a normal trip by command, an overload trip following an overconsumption of electricity of at least one load of the electrical installation and a trip following a short circuit.

2. A method according to claim 1, wherein the determination (70) of a tripping range of the switching device comprises comparing the estimated breaking current value to the nominal current threshold, and - when said estimated breaking current value is less than the nominal current threshold, the tripping is a normal tripping, - when said estimated breaking current value is greater than the nominal current threshold and less than the nominal current threshold multiplied by a predetermined multiplication factor, the tripping is an overload tripping, and - when said estimated breaking current value is greater than the nominal current threshold multiplied by said multiplication factor, the tripping is a fault tripping.

3. Method according to claim 1 or 2, further comprising a display (72) of an indication relating to the determined tripping range on the human-machine interface of the switching device.

4. A method according to any one of claims 1 to 3, wherein obtaining (62) at least one time-domain vibration signal is carried out after applying a time offset of predetermined duration after triggering the switching device.

5. A method according to any one of claims 1 to 4, comprising a learning phase prior to said operational phase, comprising - an acquisition (30) of a plurality of time-domain vibration signals, each time-domain vibration signal being associated with a known cutoff current value, - a transformation (32) of each time-domain vibration signal into a spectrogram and a storage (32) of said spectrograms and the associated cutoff current values ​​in a library of learning spectrograms.

6. A method according to claim 5, wherein said learning phase further comprises a recursive feature elimination step (40), comprising a recursive application of steps of: - a) learning (42) the parameters of said prediction model taking as input feature values ​​from a subset of training features of the calculated spectrograms, said prediction model being parameterized to provide, from the feature values ​​of said subset of training features, a classification into a class associated with an estimated current value, - b) evaluating (44) an importance score, relative to the result of the prediction model, of each feature of said subset of training features, - c) eliminating (46) the lowest importance score feature of said subset of training features,-d) update (48) of the training feature subset and repeat steps a) to d) until the training feature subset is empty.

7. A method according to claim 6, further comprising a step of obtaining (52) a ranking of the features by rank of importance according to the order of elimination of the features of said subset of learning features and selection (54) of a predetermined number N of the highest rank features to form the subset of operational features.

8. A method according to any one of claims 5 to 7, further comprising an optimization (56) of the sampling frequency, the acquisition time and a time offset to be applied when obtaining the time-domain vibration signals to be processed.

9. A computer program comprising software instructions which, when executed by a programmable electronic computing module, implement a method for determining a triggering range of an electrical circuit switching device (4) according to claims 1 to 8, said programmable electronic computing module being further configured to receive data from at least one vibration sensor (16) integrated into said switching device (4).

10. A system (2) for determining a tripping range of an electrical circuit switching device adapted to supply an electrical installation (12) comprising at least one load, the determination system (2) comprising a switching device (4) having an open position in which the flow of an electric current in said electrical circuit is interrupted, and a closed position in which an electric current flows in said electrical circuit, the tripping of the transition from the closed position to the open position being effected either following a command or following the detection of a breaking current flowing in said switching device (4), the breaking current having a level exceeding a nominal current threshold, the switching device (4) comprising at least one integrated vibration sensor (16),the system (2) comprising an electronic computing module configured (6) to receive data from at least one vibration sensor (16) integrated into said switching device (4), and configured to implement, in an operational phase, modules of:, - obtaining at least one time-domain vibration signal, over a chosen acquisition period, from data from said at least one sensor, with a given sampling frequency, following passage through said opening position, - for each of the time-domain vibration signals, transformation of the time-domain vibration signal into a spectrogram represented in a time-frequency representation space, - selection of a predetermined subset of spectrogram characteristics, called the operational characteristics subset, - application of a parameterized prediction model using supervised machine learning, with the values ​​of said subset of operational characteristics being provided as input to said parameterized prediction model, to obtain a class corresponding to an estimated value of the cutoff current, - determination, based on the estimated breaking current value and said nominal current threshold, of a tripping range for the switching device from among: a normal tripping by command, an overload tripping following an overconsumption of electricity of at least one load of the electrical installation and a tripping following a short circuit.

Images