IMPROVED METHOD FOR FAULT DETECTION IN AN AIRCRAFT AND SYSTEM CONFIGURED TO EXECUTE THE METHOD.

Abstract

The invention relates to a method for detecting and locating a fault in an input sensor (10a) of a flight control computer (1000) of an aircraft. The method comprises a step of comparing a first piece of information representing a probability of occurrence of said fault, provided by a neural network, with a second piece of information representing a probability of occurrence of said fault, provided by a classifier trained to detect and locate said fault. It is thus advantageously possible to optimize fault diagnosis and fault localization of an input sensor useful for determining flight controls. Fig. 2

Description

Title of the invention: IMPROVED METHOD FOR DETECTING FAILURES IN AN AIRCRAFT AND CONFIGURED SYSTEM TO CARRY OUT THE PROCESS. Technical field

[0001] The present invention relates to a method for detecting a fault in a system of an aircraft. The invention relates more specifically to a method for detecting and locating a fault in a sensor connected to an input of a flight control computer of an aircraft. STATE OF THE PRIOR ART

[0002] Current aircraft include numerous on-board devices and systems, including flight control computers. A flight control computer of an aircraft controls actuators of the aircraft, such as, for example, the aircraft's control surfaces (ailerons, elevator, rudder, for example). The flight commands issued by such a computer are defined from control instructions but also from information provided by sensors associated with the controlled elements. For example, a flight control computer controls a control surface actuator by taking into account the detected position of the control surface, which position is obtained via a position sensor of this control surface. In order to guarantee the operational safety of the aircraft, a flight control computer monitors the various sensors so as to detect and signal any failures of said sensors.Since a flight control computer must have a very high level of operational safety, as well as a very high level of availability, it is customary to dedicate the computing capacity of a flight control computer processor to the sole control of the aircraft. Consequently, the monitoring of sensors by a flight control computer is limited to the detection and reporting of faults, without including a location or identification of the detected faults. A maintenance operator must then carry out a search for the fault, which may for example relate to the sensor itself or to wiring between the sensor and the flight control computer. This search can be all the more time-consuming since the sensors used are generally passive sensors which are not designed to communicate information on possible faults.It would be desirable to be able to effectively identify and locate a fault in order to optimize the performance of maintenance operations and reduce the downtime of an aircraft undergoing maintenance. Statement of the invention

[0003] An object of the present invention is to provide an improved method for detecting and locating failures of sensors providing information signals to a flight control computer of an aircraft.

[0004] To this end, a method for detecting and locating a fault in an input sensor of a flight control computer of an aircraft is proposed, the method being executed by a fault detection and location module comprising electronic circuitry configured to execute the steps of said method, which comprises:

[0005] - shaping at least a first signal delivered to said sensor by said cal flight control unit and a shaping of at least a second signal delivered to said computer by said sensor, according to a predetermined format,

[0006] - a step of analyzing said first and second signals shaped by a neural network trained to carry out detection and localization of said fault from said first and second signals representative of said shaped signals delivered by said flight control computer,

[0007] - a step of comparing a first piece of information representative of a pro probability of occurrence of said fault, delivered by said neural network and, with a second piece of information representative of a probability of occurrence of said fault, delivered by a classifier trained to carry out detection and localization of said fault, from signals representative of said shaped signals, then,

[0008] - a provision of a third piece of information representative of a probability occurrence of said failure from said first and second information.

[0009] It is thus advantageously possible to optimize a fault diagnosis and a location or estimation of location of a fault of an input sensor useful for determining flight controls.

[0010] Furthermore, when the fault detection and location module comprising said electronic circuitry is independent of the main processor or processors of the flight control computer, the computing capacity of these processors remains dedicated to the determination of flight control signals and the performance and safety of the flight controls are increased.

[0011] The method for detecting and locating faults according to the invention may further comprise the following characteristics, considered in isolation or in combination:

[0012] - The method further comprises a comparison of the third information by relative to at least one predetermined threshold value.

[0013] - The method further comprises filtering the third information and a provision of a fourth piece of information representative of a probability of occurrence of said failure based on the third filtered piece of information.

[0014] The invention also relates to a device for assisting in the detection and localization of a fault in an input sensor of a flight control computer of an aircraft, the device comprising a fault detection and localization module comprising electronic circuitry configured to operate: - a shaping of at least a first signal delivered to said sensor by said flight control computer and a shaping of at least a second signal delivered to said computer by said sensor, according to a predetermined format, - a step of analyzing said first and second signals shaped by a neural network trained to perform detection and localization of said fault from said first and second signals representative of said shaped signals delivered by said flight control computer, - a step of comparing a first piece of information representative of a probability of occurrence of said breakdown, delivered by said neural network with a second piece of information representative of a probability of occurrence of said breakdown, delivered by a classifier trained to carry out a detection and a localization of said breakdown, from signals representative of said formatted signals, then, - a provision of at least a third piece of information representative of a probability of occurrence of said failure based on said first and second pieces of information.

[0015] The fault detection and location device according to the invention may further comprise the following characteristics, considered in isolation or in combination: - The device further comprises electronic circuitry configured to carry out a comparison of the third information with respect to at least one predetermined threshold value. - The device further comprises electronic circuitry configured to filter said third information and to provide a fourth information representative of a probability of occurrence of said fault from the third filtered information.

[0016] The invention also relates to a flight control calculator device comprising a first, main processor, configured to determine flight control signals and a device for assisting in the detection and localization of a fault as previously described comprising a detection and localization module fault correction, which module comprises said electronic circuitry, which circuitry being independent of said first processor.

[0017] The invention also relates to an aircraft comprising a device for assisting in the detection and localization of failure or a flight control computer as previously described.

[0018] The invention finally relates to a computer program product comprising program code instructions for executing the steps of a method for detecting and locating faults as previously described as well as a storage medium comprising a computer program product as mentioned above. Brief description of the drawings

[0019] The above-mentioned features of the invention, as well as others, will appear more clearly on reading the following description of an exemplary embodiment, said description being made in relation to the attached drawings:

[0020] [Fig-1] illustrates an aircraft comprising a device for assisting in the detection and locating a fault in an input sensor of a flight control computer according to one embodiment of the invention;

[0021] [Fig.2] schematically illustrates a device for assisting in the detection and loca detection of a failure of an input sensor of a flight control computer according to an embodiment of the invention;

[0022] [Fig.3] schematically illustrates a method for assisting in the detection and localization of a fault in an input sensor of a flight control computer executed by the device for assisting and localizing a fault already illustrated in [Fig.2];

[0023] [Fig.4] schematically illustrates an internal architecture of the assistance and fault location device already illustrated in [Fig.2].

[0024] [Fig.5] schematically illustrates internal details of the assistance and fault location device already illustrated in [Fig.2], comprising a post-processing unit; and,

[0025] [Fig.6] schematically illustrates details of the post-processing unit already illustrated in [Fig.5] of the fault assistance and location device.

[0026] DETAILED DESCRIPTION OF EMBODIMENTS

[0027] [Fig.l] schematically and symbolically represents an aircraft 1 comprising a device (not detailed in [Fig.l]) for assisting in the detection and localization of failures of input sensors of a flight control computer of the aircraft 1. Advantageously, such a device on board the aircraft 1 makes it possible to optimize the detection and localization of a determined failure of a sensor connected to a flight control computer of the aircraft 1. More generally, the device for assisting in the detection and location of a fault 10 is configured to detect a plurality of types of faults and a plurality of faults, as well as the at least approximate location, if not more precise, of each of the detectable and detected faults for the set of sensors connected to one or more inputs of a flight control computer of the aircraft 1.

[0028] In the remainder of the description, a connection link carrying a signal, this signal carried by said connection link and the information coded by this signal carried by this connection link are referred to interchangeably. For example, a reference “P” designates both an electrical signal P and a connection link P which carries the signal P, just as information P carried (or coded) by the signal P on the connection link which carries the signal P.

[0029] [Fig. 2] schematically and symbolically represents a flight control computer 1000 of the aircraft 1 whose inputs are connected to sensors, for example control surface position sensors of the aircraft 1 and whose outputs are connected to actuators, for example control surface actuators of the aircraft 1. For the purposes of simplification, a single sensor 10a is shown in [Fig. 2] although many sensors of different types can be connected to the input of the flight control computer 1000. For the same reasons aimed at simplifying the description of the invention, a single flight control actuator control signal 1001 is illustrated in [Fig. 2], although many actuators of different types can be connected to the output of the flight control computer 1000.

[0030] According to the example described, the sensor 10a is connected to the flight control computer 1000 by a first connection carrying a first signal 10b delivered by the flight control computer 1000 to the sensor 10a and by a second connection carrying a second signal 10c delivered to the flight control computer 1000 by the sensor 10a when the first signal 10b has a predetermined shape. The signal 10b is for example an electrical power supply signal for the sensor 10a, or an electrical excitation signal for the sensor 10a, useful for its nominal operation. Such an excitation signal makes it possible, for example, to power a mobile element whose movement causes an output signal from the sensor 10a to vary, such as the signal 10c, for example. Obviously, and depending on the nature and type of sensor, additional connections may exist and several signals of one or more types may be delivered by the computer 1000 to the sensor 10a, and vice versa.

[0031] Advantageously, the flight control computer 1000 comprises a device 10 for assisting in the detection and localization of faults for, in particular, the detection and localization of faults of the sensor 10a. For this purpose, the first and second signals 10b and 10c are connected to the device 10 for assisting in the detection and fault location, inside the flight control computer 10. In other words, the connections between the flight control computer 10 and the sensor 10a pass through the fault detection and location assistance device 10. The same applies more broadly to all the connections between the flight control computer 1000 and the various sensors connected to it, at the input of the flight control computer 1000. Thus, given that the fault detection and location assistance device 10 is included in the flight control computer 1000, its operation does not require the installation of specific wiring.On the one hand, this makes it possible to preserve operational safety; on the other hand, this even allows the implementation of such a device 10 for assisting in the detection and localization of faults in a flight control computer of an existing aircraft by simply replacing said flight control computer. The flight control computer 1000 further comprises a flight control control unit 100, also referred to here as the “main processor 100” of the flight control computer. The main processor 100 is configured to determine the state of the flight control outputs to the flight control actuators, based on the setpoint inputs and the information and signals from the various sensors connected to the flight control computer.The instruction signals applied to the flight control computer may come from instruction devices (for example a control stick or steering wheel in the cockpit of the aircraft 1), and / or from a pilot assistance device (for example, a flight control correction system or an automated air navigation aid system).

[0032] According to one embodiment, the fault detection and location assistance device 10 delivers to the main processor 100 an input information in the form of a signal 12, which is representative of the state of the signal 10c provided by the sensor 10a at the input of the flight control computer 1000, as well as the information 11 (also in the form of a signal) representative of a probability of failure of a predefined type of the sensor 10a at a predefined location of the on-board systems of the aircraft 1. According to one embodiment, the information 11, carried by a coded signal, is a composite digital information, that is to say that it comprises a plurality of fields each representative of an information element relating to the probability of a failure in relation to a type of failure of the sensor 10a.For example, one information field encodes a fault type from a list of predetermined fault types, another information field encodes a fault from a list of predetermined faults, and yet another field encodes an area or location of the fault in the aircraft systems from a list of predetermined areas or locations. According to one embodiment, the processor 100 generates a signal. information 1002 on at least one of its outputs, from the internal signal 12, aimed at notifying a third-party system of the presence of a failure of a flight control sensor, for example an internal alert device in the aircraft 1 or even a ground station configured to supervise the state of the aircraft 1 and any maintenance operations to be carried out.

[0033] According to an alternative embodiment, the information 11 is an analog or digital signal associated with two predetermined states, one state of which codes a fault detection and one state of which codes an absence of fault. According to this example, the same is then true for the signal 1002.

[0034] Cleverly and advantageously, the device for assisting with the detection and localization of faults 10 of the flight control computer 1000 comprises a neural network NN and a classifier CLASS as well as a shaping module PRE configured to operate a normalization of the signals transmitted from the flight control computer 1000 to the sensors connected thereto and from these sensors to the flight control computer 1000.

[0035] Advantageously, the neural network NN is configured to be able to provide a probability of occurrence of sensor failure, and in particular, among all the sensors which are connected to the input of the flight control computer 1000, to provide information representative of a probability of occurrence of failures of the sensor 10a. A failure of the sensor 10a may be a power supply fault of the sensor 10a, due to a fault in the connection carrying the signal 10b, or a fault in the output connection of the sensor 10a, that is to say a fault in the link carrying the signal 10c, or even a fault internal to the sensor 10a, for example a fault in an internal transducer or any other internal element or component of the sensor 10a.In order to be able to provide information on the probability of occurrence of failures, the NN neural network is previously trained from predetermined signals corresponding to a plurality of scenarios or configurations relating to normal operating states and relating to abnormal states, i.e. malfunction states. This corresponds, for example, to supervised learning. According to one embodiment, the NN neural network is trained before its insertion into the aircraft systems for operation. According to a variant, the NN neural network is used for the detection and localization of failures during certain periods of flight and taxiing of the aircraft 1 while still being used in learning mode during other periods of flight and taxiing of the aircraft 1.

[0036] The operations of normalization of the signals applied at the input of the NN neural network make it possible to guarantee the conformity of the format of these signals with the signals previously applied during a learning phase (in particular supervised learning) of the NN neural network prior to its exploitation at on board the aircraft. The classifier CLASS of the fault detection and localization assistance device 10 of the flight control computer 1000 is implemented in a unit (or module) called “post-processing” POST, which makes it possible, from the same signals as those applied to the input of the neural network NN, to obtain a second piece of information on the probability of failure of the sensor 10a, as well as of the other sensors connected to the input of the flight control computer 1000 not shown in the figures.

[0037] Advantageously and cleverly, it is then possible to verify the consistency of two pieces of information representative of the probability of a breakdown obtained by two different means which are on the one hand a neural network, for example a FCNN type neural network (“fully connected neural network”, from the English, and which means “fully connected neural network”) and a classifier connected to a database of breakdown information, then to provide a third piece of information of probability of occurrence of a breakdown of a sensor connected to the input of the flight control computer 1000. This makes it possible to achieve a desired performance of the breakdown detection and localization assistance device 10 by using a less complex neural network than if the neural network were used alone (without a classifier).This also makes it easier to train the neural network, for example by using a smaller training dataset, allowing the classifier to cover unlikely scenarios for which the neural network would not have been trained.

[0038] It should be noted that the learning phase(s) of the neural network NN as well as the structure of the classifier CLASS associated with a database DB are such that information on the probability of a failure comprises or is associated with information on the location of said failure. For example, the scenarios chosen to simulate or reproduce failures of the input sensors of the flight control computer 1000 or of the associated circuits useful for the operation of the sensors, comprise reproductions or simulations of failures of the same type at different locations. For example, in the case of a leakage current, the fault inducing such a leakage current is produced at several locations of the same electrical equipotential connection, so as to obtain diverse and varied configurations of the other signals linked to the same sensor as that for which a failure is simulated.

[0039] According to one embodiment, the device for assisting in the detection and localization of a fault 10 of a sensor is independent of the main processor 100 in the flight control computer 1000. In other words, this means that apart from the signals delivered by the device 10 to the main processor 100, these two elements do not share their respective computing powers to carry out processing. In other words again, the detection and / or localization of faults is carried out by the device 10 and notified to the main processor 100 which can take it into account for determining the flight controls, but the computing power of the main processor 100 is not made available to the method for detecting and locating faults of an input sensor of the flight control computer 1000, in order to optimize safety in determining the flight controls of the aircraft 1.

[0040] [Fig.3] is a flowchart illustrating steps of a method for providing information representative of a probability of occurrence of a failure of the sensor 10a connected to the input of the flight control computer 1000, from two other intermediate pieces of information representative of a probability of occurrence of a failure of the sensor 10a respectively obtained using a network of the neural network NN of the device for assisting with the detection and localization of a failure 10 and the classifier CLASS of this same device for assisting with the detection and localization of a failure 10.

[0041] An initial step S0 is a step of initializing all of the systems described at the end of which all of the systems of the aircraft are correctly initialized, configured and able to operate nominally. During a step S1, the device for assisting with the detection and localization of a failure 10 carries out a shaping, also called normalization or conformity of the signals in relation to the sensor 10a, whether they are signals coming from the sensor 10a or signals intended for the sensor 10a. The signals thus shaped are applied as input to the neural network NN on the one hand and as input to the classifier CLASS on the other hand. Then, during a step S2, the neural network NN determines and provides a first piece of information PI representative of a probability of a failure from an analysis of the signals 10b and 10c shaped and applied as input to the neural network NN.

[0042] The signals 10b and 10c being further applied as input to the classifier CLASS, a second piece of information P2 representative of a probability of a breakdown is determined by the classifier CLASS connected to its database DB, from the signals 10b and 10c formatted and then applied as input to the classifier CLASS, so as to then compare, during a step S3, the first piece of information PI representative of a probability of a breakdown and the second piece of information P2 representative of a probability of a breakdown thus obtained.A third piece of information P3 representative of a probability of occurrence of a failure of the sensor 10a, resulting from the comparison between the first piece of information PI representative of a probability of a failure and the second piece of information P2 representative of a probability of a failure is then determined during step S3, so as to implement a classification step, then provided at the output of the device for assisting in the detection and localization of a failure during a step S4.

[0043] According to one embodiment, the information signal 11 delivered by the device for assisting with the detection and localization of a fault 10 is a signal carrying or encoding the information P3. According to one embodiment, an additional step of comparing the information P3 with one or more threshold values ​​T is carried out so as to implement an improved decision step for detecting a fault of the sensor 10a. According to one embodiment, a final temporal filtering is carried out at the output of the device for assisting with the detection and localization of a fault, so as to eliminate accidental transients and to confirm over time the state of the final information representative of a probability of occurrence of a fault, provided at the output of the device 10.

[0044] [Fig. 5] schematically illustrates internal details of the fault detection and location assistance device 10 connected to the sensor 10a via the connection links carrying the signals 10b and 10c. The pre-processing module PRE is configured to operate the shaping or normalization of the signals 10b and 10c and to deliver on its output as many normalized signals as input signals applied to it as input. All of the output signals applied as input to the neural network NN is called I and is further applied as input to the post-processing module POST carrying out a processing subsequent to the determination of the information PI, namely the operations carried out during steps S3 and S4 of the method described in relation to [Fig. 3]. The information signals 11 and 12 delivered at the output of the device 10 come from the post-processing module POST.

[0045] [Fig.6] schematically and symbolically illustrates an internal electronic circuit of the post-processing module POST configured to operate the aforementioned steps S3 and S4 during which the information PI and P2 are compared by a comparator circuit C1 to generate an information signal P3 as well as a comparator circuit C2 used to compare the information P3 with one or more threshold values ​​T, optional, and a time filtering circuit F, also optional, to operate a time filtering of the output signal, if necessary.

[0046] Thus, the information PI corresponds to a probability of occurrence of a failure of the sensor 10a as provided by the neural network NN, the classifier CLASS determines a statistical membership P2 of a failure to a type of failure among predefined types of failure and stored in the database DB of predefined failures, and the comparison stage C2 and the filter F jointly or separately define a final decision module, when they are implemented.

[0047] [Fig. 4] is a schematic representation of an example of the internal architecture of the device for assisting with the detection and localization of a fault 10. Let us consider for the purposes of illustration that [Fig. 4] illustrates an internal arrangement of the device for assisting with the detection and localization of a fault 10 as installed in the aircraft 1. note that [Fig.4] could also schematically illustrate an example of hardware architecture of the internal fault detection and localization module of the fault detection and localization assistance device 10.

[0048] According to the example of hardware architecture shown in [Fig. 4], the fault detection and location device 10 then comprises, connected by a communication bus 10-9: a processor or CPU (“Central Processing Unit” in English) 10-1; a RAM (“Random Access Memory” in English) 10-2; a ROM (“Read Only Memory” in English) 10-3; a storage unit such as a hard disk (or a storage media reader, such as an SD (“Secure Digital” in English) card reader) 10-4; a communication interface module 10-5 allowing the system of the fault detection and location assistance device 10 to communicate with remote devices, such as other on-board systems of the aircraft 1, among which are elements related to the CLASS classifier and the NN neural network used.

[0049] The processor 10-1 of the fault detection and localization assistance device 10 is capable of executing instructions loaded into the RAM 10-2 from the ROM 10-3, from an external memory (not shown), from a storage medium (such as an SD card), or from a communication network. When the fault detection and localization assistance device 10 is powered on, the processor 10-1 is capable of reading instructions from the RAM 10-2 and executing them. These instructions form a computer program causing the processor 10-1 to implement all or part of a fault detection and localization assistance method described in relation to [Fig. 3] or described variants of this method.

[0050] All or part of the method described in relation to [Fig. 4] or its described variants may be implemented in software form by executing a set of instructions by a programmable machine, for example a DSP ("Digital Signal Processor" in English) or a microcontroller, or be implemented in hardware form by a machine or a dedicated component, for example an FPGA ("Field-Programmable Gate Array" in English) or an ASIC ("Application-Specific Integrated Circuit" in English). In general, the system for assisting in the detection and localization of a fault 10 of a sensor comprises electronic circuitry configured to implement the method described in relation to itself.Obviously, the fault detection and location assistance device 10 further comprises all the elements usually present in a system comprising a control unit and its peripherals, such as a power supply circuit, a power supply supervision circuit, one or more clock circuits, a reset circuit, input-output ports, interrupt inputs, bus drivers, this list being non-exhaustive. exhaustive.

[0051] According to one embodiment, the signals processed by the neural network correspond to analog signals. According to another embodiment, the signals processed by the neural network correspond to digital signals. According to yet another embodiment, the signals processed by the neural network correspond to a combination of analog signals and digital signals.

Claims

Claims

1. Method for detecting and locating a fault in an input sensor (10a) of a flight control computer (1000) of an aircraft (1), the method being executed by a fault detection and locating module (10) comprising electronic circuitry configured to execute the steps of said method, which comprises: - shaping (SI) at least one first signal (10b) delivered to said sensor (10a) by said flight control computer (1000) and shaping at least one second signal (10c) delivered to said computer (1000) by said sensor (10a), according to a predetermined format, - an analysis (S2) of said first (10b) and second (10c) shaped signals, by a neural network trained to carry out a detection and a localization of said fault from said first (10a) and second (10b) signals representative of said shaped signals form issued by said flight control computer (1000),- a comparison (S3) of a first piece of information representative of a probability of occurrence of said fault, delivered by said neural network and, with a second piece of information representative of a probability of occurrence of said fault, delivered by a classifier trained to carry out a detection and a localization of said fault, from signals representative of said signals (10b, 10c) shaped, then, - a supply (S4) of at least a third piece of information (11) representative of a probability of occurrence of said fault from said first and second pieces of information.,

2. Method for detecting and locating a fault according to claim 1, further comprising a comparison of said third information (11) with respect to at least one predetermined threshold value (T).

3. Method for detecting and locating a fault according to claim 1 or 2, further comprising filtering (S4a) said third information and providing fourth information (11a) representing indicating a probability of occurrence of said failure from the third filtered information.

4. Device for assisting in the detection and localization of a fault (10) of an input sensor of a flight control computer of an aircraft, the device comprising a fault detection and localization module comprising electronic circuitry configured to operate: - a shaping of at least a first signal delivered to said sensor by said flight control computer and a shaping of at least a second signal delivered to said computer by said sensor, according to a predetermined format, - an analysis of said first and second shaped signals, by a neural network trained to operate a detection and localization of said fault from said first and second signals representative of said shaped signals delivered by said or said flight control computer, - a comparison of a first information of a probability of occurrence of said fault,delivered by said neural network, with a second piece of information representative of a probability of occurrence of said breakdown, delivered by a classifier trained to carry out detection and localization of said breakdown, from signals representative of said shaped signals, then, - a supply of at least a third piece of information representative of a probability of occurrence of said breakdown from said first and second pieces of information.,

5. Device for assisting in the detection and localization of a fault (10) according to claim 4, further comprising electronic circuitry configured to operate a filtering of said third information and to operate a supply of a fourth information representative of a probability of occurrence of said fault from the third filtered information.

6. Flight control calculator device (1000) comprising a first, main processor (100), configured to determine flight control signals and a detection assistance device and to the location of a fault (10) according to one of claims 4 to 5 comprising a fault detection and location module comprising said electronic circuitry, which is independent of said first processor.

7. Aircraft comprising a device for assisting with the detection and location of failure according to one of claims 4 to 5 or a flight control calculator device according to claim 6.

8. Computer program product comprising program code instructions for executing the steps of a method according to one of claims 1 to 3 when said instructions are executed by a processor of a device for assisting in the detection and localization of faults.

9. A storage medium comprising a computer program product according to claim 8.

Images