Method and system for maintaining an onboard inference engine to diagnose faults on an aircraft

Abstract

A method and system for maintaining an onboard inference machine to diagnose faults on an aircraft, the aircraft including aircraft systems configured to report faults to the onboard inference machine. The method includes accessing diagnostic data received from an onboard computer of the aircraft including the onboard inference machine. The off-board inference machine establishes an off-board diagnostic causal model that describes causal relationships between failed tests and diagnosed fault modes. The off-board diagnostic causal model is compared to the diagnostic data. Based on the comparison, differences between a graph of the off-board diagnostic causal model and another graph of the onboard diagnostic causal model are identified to determine new causal relationships relative to known causal relationships. The onboard diagnostic causal model is updated to further describe the new causal relationships, including generating an updated model, and the updated model is uploaded to the onboard computer.

Description

TECHNICAL FIELD

[0001] The present disclosure relates generally to aircraft maintenance, and more particularly to on-board diagnosis of aircraft system failure modes and related aircraft maintenance. BACKGROUND

[0002] Complex systems, such as machines including vehicles such as aircraft, spacecraft, watercraft, motor vehicles, rail vehicles, typically include some type of performance monitoring system that records data about the performance of the machine, including the performance of various systems (and subsystems) of the machine. The data includes a record of certain performance events that occur during operation of the machine. The performance monitoring system typically collects data and reports all of the collected data to a user. The user can then utilize the data to determine the type of maintenance or repair, if any, that the machine can need. For example, if the data indicates that a particular mechanical or electromechanical system of the machine has failed or that the performance of one or more mechanical or electromechanical systems can lead to future machine failures, the user can perform appropriate repairs to the machine at the next opportunity.

[0003] While current systems for machine performance and failure monitoring provide the necessary data to the user to make appropriate repair decisions, it is still necessary for the user to sort through all of the data to determine the most appropriate repair action to address the failure mode. Thus, the user must sort through and interpret the data based on the user’s knowledge of the particular machine. This can be time consuming and does not always result in the repair action that is performed first being the most appropriate repair action, especially for complex machines such as aircraft and other vehicles. For many types of machines, especially commercial vehicles, the amount of time that the vehicle is out of service is expensive for the vehicle owner. As such, the longer it takes to perform the most appropriate repair action for a given failure mode, the longer the vehicle will be out of service, which can be expensive for the vehicle owner if the vehicle would otherwise have been in service.

[0004] Accordingly, it would be desirable to have a system and method that addresses at least some of the above issues, as well as other possible issues. SUMMARY

[0005] Example implementations of the present disclosure relate to improved techniques for on-board diagnosis and correlation of failure data to maintenance actions, and closed loop diagnostic model shaping for complex systems. Example implementations provide an on-board reasoner on an aircraft and a process for diagnosing failures on the aircraft. Other example implementations provide an off-board reasoner and a process for closed loop diagnostic model shaping to maintain the on-board reasoner to diagnose failures on the aircraft.

[0006] Accordingly, the present disclosure includes, but is not limited to, the following example implementations.

[0007] Some example implementations provide a method of diagnosing a fault on an aircraft, the aircraft including aircraft systems configured to report faults to an onboard inference machine, the method comprising: receiving, at an onboard computer of the aircraft including the onboard inference machine, a fault report from one of the aircraft systems, the fault report indicating a failed test reported by the aircraft system; accessing, by the onboard inference machine, an onboard diagnostic causal model, the onboard diagnostic causal model represented by a graph that describes known causal relationships between possible failed tests reported by respective ones of the aircraft systems and possible fault modes of the respective ones of the aircraft systems; diagnosing, by the onboard inference machine, a fault mode of the one of the aircraft systems or another of the aircraft systems according to the failed test and using the graph theory algorithm and the onboard diagnostic causal model; determining a maintenance action for the fault mode; and generating, by the onboard computer, a maintenance message including at least the maintenance action.

[0008] In some example implementations of the method of any of the preceding example implementations, or any combination of the preceding example implementations, the onboard diagnostic causal model is represented by a graph including nodes connected by edges, the nodes representing possible failed tests and possible fault modes, and the edges indicating known causal relationships between the possible failed tests and the possible fault modes.

[0009] In some example implementations of the method of any of the preceding example implementations, or any combination of the preceding example implementations, the onboard diagnostic causal model is represented by a graph that is a collection of graphs for respective ones of the aircraft systems, the graphs reflecting fault propagation behavior within the respective ones of the aircraft systems, and the collection reflecting fault propagation behavior across connected ones of the aircraft systems.

[0010] In some example implementations of the method of any of the preceding example implementations, or any combination of the preceding example implementations, the graph theory algorithm is an optimal solution set (OSS) algorithm.

[0011] In some example implementations of the method of any of the preceding example implementations, or any combination of the preceding example implementations, the method further comprises sending the maintenance message to a display device on the aircraft or a display device on a maintenance component, the display device configured to establish a connection with the onboard computer to receive the maintenance message.

[0012] In some example implementations of the method of any of the preceding example implementations, or any combination of the preceding example implementations, the display of the maintenance message references instructions for performing the maintenance action to address the fault mode diagnosed by the onboard inference machine.

[0013] In some example implementations of the method of any of the foregoing example implementations, or any combination of the foregoing example implementations, the fault report is one of a plurality of fault reports of failed tests reported by respective ones of the aircraft systems, the fault pattern is one of a plurality of diagnosed fault patterns of at least some of the aircraft systems that caused the failed tests, and the method further comprises: accessing diagnostic data received from the onboard computer, the diagnostic data including the plurality of fault reports and the plurality of diagnosed fault patterns; accessing, by the off-board reasoning machine, an off-board diagnostic causal model, the off-board diagnostic causal model describing causal relationships between the failed tests and the plurality of diagnosed fault patterns, the off-board diagnostic causal model constructed utilizing a graph theory machine learning algorithm trained using historical diagnostic data; comparing the diagnostic data to the off-board diagnostic causal model; and based on the comparison, determining a new one of the causal relationships described by the off-board diagnostic causal model, the new causal relationship being new with respect to known causal relationships; and updating the onboard diagnostic causal model to further describe the new causal relationship, including producing an updated onboard diagnostic causal model, and uploading the updated onboard diagnostic causal model to the onboard computer using the loadable software aircraft component upload toolset.

[0014] In some example implementations of the method of any of the foregoing example implementations, or any combination of the foregoing example implementations, the diagnostic data further includes maintenance messages having maintenance actions determined by the onboard reasoning machine for respective ones of the plurality of diagnosed fault patterns, and the onboard diagnostic causal model further describes relationships between possible fault patterns and respective ones of the maintenance actions, and wherein, for a particular maintenance message having a particular maintenance action for a respective diagnosed fault pattern, the method further comprises: accessing a maintenance record having a performed maintenance action for the respective diagnosed fault pattern and thereby having a new maintenance action; and identifying a maintenance action difference between the particular maintenance action and the performed (new) maintenance action to determine a new relationship between the respective diagnosed fault pattern and the performed maintenance action, wherein updating the onboard diagnostic causal model further includes updating the onboard diagnostic causal model to describe the new relationship between the respective diagnosed fault pattern and the performed (new) maintenance action.

[0015] In some example implementations of the method of any of the foregoing example implementations, or any combination of the foregoing example implementations, the graph theory machine learning algorithm is a graph theory algorithm for finding maximal cliques.

[0016] Some example implementations provide an onboard computer for diagnosing faults on an aircraft, the aircraft including aircraft systems configured to report faults to an onboard inference machine, the onboard computer comprising: a memory configured to store computer-readable program code comprising the onboard inference machine; and a processing circuit configured to access the memory and execute the computer-readable program code to cause the device to perform at least the method of any preceding example implementation or any combination of any preceding example implementation.

[0017] Some example implementations provide a system for diagnosing faults on an aircraft, the aircraft including aircraft systems configured to report faults to an onboard inference machine, the system comprising: an onboard computer comprising the onboard inference machine, the onboard computer configured to perform the method of any preceding example implementation or any combination of any preceding example implementation.

[0018] Some example implementations provide a method of maintaining an onboard inference machine to diagnose faults on an aircraft, the aircraft including aircraft systems configured to report faults to the onboard inference machine, the method comprising: accessing diagnostic data received from an onboard computer of the aircraft comprising the onboard inference machine, the diagnostic data including a plurality of fault reports of failed tests reported by respective ones of the aircraft systems and a plurality of diagnosed fault patterns of at least some of the aircraft systems that caused the failed tests; building an off-board diagnostic causal model using an off-board inference machine, the off-board diagnostic causal model describing causal relationships between the failed tests and the plurality of diagnosed fault patterns, the off-board diagnostic causal model built using a graph theory machine learning algorithm trained using historical diagnostic data; comparing the diagnostic data to the off-board diagnostic causal model; and based on the comparison, determining a new causal relationship in the causal relationships described by the off-board diagnostic causal model, the new causal relationship being new relative to known causal relationships; and updating the onboard diagnostic causal model to further describe the new causal relationship, including producing an updated onboard diagnostic causal model, and uploading the updated onboard diagnostic causal model to the onboard computer using a loadable software aircraft component upload toolset.

[0019] In some example implementations of the method of any preceding example implementation or any combination of any preceding example implementation, the off-board diagnostic causal model is represented by a graph comprising nodes connected by edges, the nodes representing the failed tests and the plurality of diagnosed fault patterns, and the edges indicating known causal relationships between the failed tests and the plurality of diagnosed fault patterns.

[0020] In some example implementations of the method of any of the foregoing example implementations, or any combination of the foregoing example implementations, the comparison of the diagnostic data to the off-board diagnostic causal model further includes a comparison to actual fault data, diagnostic data, and maintenance action data collected from the on-board aircraft system that reflect actual causal relationships.

[0021] In some example implementations of the method of any of the foregoing example implementations, or any combination of the foregoing example implementations, the graph theory machine learning algorithm is a graph theory algorithm for finding a maximum clique.

[0022] In some example implementations of the method of any of the foregoing example implementations, or any combination of the foregoing example implementations, the plurality of fault reports includes fault reports indicating those failed tests caused by one of the plurality of diagnosed fault modes, and the method further includes: diagnosing, by the off-board reasoner, a corresponding fault mode of the aircraft system from those failed tests caused by the diagnosed fault mode and using the graph theory machine learning algorithm and graph, and reporting any discrepancies between the corresponding fault mode diagnosed by the off-board reasoner and the one of the plurality of diagnosed fault modes from the diagnostic data reports received from the on-board computer.

[0023] In some example implementations of the method of any of the foregoing example implementations, or any combination of the foregoing example implementations, the diagnostic data further includes maintenance messages having maintenance actions determined by the on-board reasoner for corresponding ones of the plurality of diagnosed fault modes, and the on-board diagnostic causal model further describes relationships between possible fault modes and corresponding ones of the maintenance actions, and wherein for a particular maintenance message having a particular maintenance action for a corresponding diagnosed fault mode, the method further includes: accessing a maintenance record having a performed maintenance action for the corresponding diagnosed fault mode and thereby having a new maintenance action; and identifying a maintenance action discrepancy between the particular maintenance action and the performed maintenance action to determine a new relationship between the corresponding diagnosed fault mode and the performed maintenance action, wherein updating the on-board diagnostic causal model further includes updating the on-board diagnostic causal model to describe the new relationship between the corresponding diagnosed fault mode and the performed maintenance action.

[0024] In some example implementations of the method of any of the foregoing example implementations, or any combination of the foregoing example implementations, the method further includes receiving, at an onboard computer of the aircraft that includes the onboard inferential machine, a fault report from one of the aircraft systems, the fault report indicating a failed test reported by the aircraft system; accessing, by the onboard inferential machine, the updated onboard diagnostic causal model; diagnosing, by the onboard inferential machine, a fault mode of the one of the aircraft systems or another of the aircraft systems according to the failed test and using the graph theory algorithm and the updated onboard diagnostic causal model; determining a maintenance action for the fault mode; and generating, by the onboard computer, a maintenance message that includes at least the maintenance action.

[0025] In some example implementations of the method of any of the foregoing example implementations, or any combination of the foregoing example implementations, the graph theory algorithm is an optimal solution set (OSS) algorithm.

[0026] Some example implementations provide a non-onboard computer for maintaining an onboard inferential machine to diagnose faults on an aircraft, the aircraft including aircraft systems configured to report faults to the onboard inferential machine, the non-onboard computer comprising: a memory configured to store computer readable program code including a non-onboard inferential machine; and a processing circuit configured to access the memory and execute the computer readable program code to cause the non-onboard computer to perform at least the method of any of the foregoing example implementations, or any combination of the foregoing example implementations.

[0027] Some example implementations provide a system for maintaining an onboard inferential machine to diagnose faults on an aircraft, the aircraft including aircraft systems configured to report faults to the onboard inferential machine, the system including an onboard computer that includes the onboard inferential machine and a non-onboard computer that includes a non-onboard inferential machine, the system components configured to perform the method of any of the foregoing example implementations, or any combination of the foregoing example implementations.

[0028] These and other features, aspects, and advantages of the present disclosure will become evident to those skilled in the art from a reading of the following detailed description, together with the accompanying drawings. The present disclosure includes any combination of two, three, four, or more of the features or elements set forth in the present disclosure, whether explicitly described or not, in any of the example implementations described herein. The present disclosure is intended to be read generically, such that any separable feature or element of the present disclosure in any of its aspects and example implementations should be considered to be combinable unless the context of the present disclosure explicitly indicates otherwise.

[0029] Accordingly, it is to be understood that the present description has been presented for the purpose of summarizing some example implementations only, and that it does not limit the scope of the disclosure or the protection sought. Additional example implementations, aspects, and advantages will become apparent to those of ordinary skill in the art upon reading the following detailed description, taken in conjunction with the accompanying drawings, which illustrate some of the described example implementations by way of example. BRIEF DESCRIPTION OF DRAWINGS

[0030] Accordingly, having generally described example implementations of the disclosure, reference will now be made to the drawings, which are not necessarily drawn to scale, and wherein:

[0031] Figure 1 An aircraft is shown in accordance with some example implementations of the disclosure;

[0032] Figure 2A A system for aircraft maintenance is shown in accordance with example implementations of the disclosure;

[0033] Figure 2B A block diagram of an onboard computer with an onboard inference machine functionally connected to an offboard computer with an offboard inference machine is shown in accordance with example implementations of the disclosure;

[0034] Figure 3 A graph-based representation of an onboard diagnostic causal model is shown in accordance with example implementations of the disclosure;

[0035] Figure 4 A subgraph representing a maximal clique is shown in accordance with example implementations;

[0036] Figure 5A , Figure 5B and Figure 5C are flowcharts illustrating various steps in a method of diagnosing a fault on an aircraft, the aircraft including aircraft systems configured to report faults to an onboard inference machine, in accordance with example implementations;

[0037] Figure 6A , Figure 6B and Figure 6C are flowcharts illustrating various steps in a method of maintaining an onboard inference machine to diagnose a fault on an aircraft, the aircraft including aircraft systems configured to report faults to the onboard inference machine, in accordance with example implementations; and

[0038] Figure 7 An apparatus is shown in accordance with some example implementations. DETAILED DESCRIPTION

[0039] Some embodiments of the present disclosure will now be described more fully with reference to the accompanying drawings. The present disclosure should not, however, be construed as limited to such embodiments. Rather, they should be understood to cover all alternatives, modifications, and equivalents. For example, unless otherwise specified, references to a thing should not be construed as implying a particular order unless otherwise specified; and similarly, a thing described as on top of another thing can instead be on the bottom, and vice versa; and similarly, a thing described as to the left of another thing can instead be to the right, and vice versa. Throughout, like reference numerals refer to like elements.

[0040] Example implementations of the present disclosure relate to improved techniques for correlation of onboard diagnostic and fault data with maintenance actions and closed loop diagnostic model shaping for complex systems. Example implementations provide an onboard reasoner on an aircraft and a process for diagnosing faults on the aircraft. Additionally or alternatively, some example implementations provide an off-board reasoner and a process for closed loop diagnostic model shaping to maintain the onboard reasoner to diagnose faults on the aircraft.

[0041] Figure 1 An aircraft 100 of the type that can benefit from example implementations of the present disclosure is shown. As shown, the aircraft includes a fuselage 102 having a body 104, wings 106, and a tail 108. The aircraft also includes a number of high level systems 110, such as a propulsion system. In the example shown, the propulsion system includes two engines 112 mounted on the wings. In other embodiments, the propulsion system can include other arrangements, such as engines carried by other portions of the aircraft including the body and / or the tail. The high level systems can also include an electrical system 114, a hydraulic system 116, and / or an environmental system 118. Any number of other systems can be included. Figure 1 In the example shown, the propulsion system includes two engines 112 mounted on the wings. In other embodiments, the propulsion system can include other arrangements, such as engines carried by other portions of the aircraft including the body and / or the tail. The high level systems can also include an electrical system 114, a hydraulic system 116, and / or an environmental system 118. Any number of other systems can be included.

[0042] The above high-level system can include a plurality of sensors and subsystems that provide fault and sensor data that is communicated to an aircraft condition monitoring system (ACMS) via an aircraft data communication bus network and / or an onboard network system (ONS). The ACMS can collect, monitor, log, and report real-time aircraft system data, which can include error messages from a flight deck effects (FDE) system, system test reports, fault reports, and other information. Data collected by the ACMS is used, among many other aircraft performance functions, to perform cabin pressure and temperature monitoring, hard landing detection, crew monitoring, and engine monitoring, for example. The received data is then utilized to analyze aircraft performance, log significant flight events, report aircraft system test reports and fault reports, and troubleshoot faults.

[0043] The ACMS can be in communication with an onboard component / computer 120, which can also be referred to as a central maintenance computer (CMC), an aircraft health management or maintenance management system, and a diagnostic maintenance computing function (DMCF) can reside on the central maintenance computer. The onboard computer 120, including the DMCF, can receive aircraft system test reports and fault reports, and can also include an onboard diagnostic model. The DMCF can provide data collection for the onboard diagnostic model, which receives test report and fault report data.

[0044] Figure 2A A system 200 for maintaining an aircraft 202, which can correspond to the aircraft 100, is shown in accordance with some example implementations. As described in greater detail below, the system can include any of a number of different subsystems (each of which is a separate system) for performing one or more functions or operations. The subsystems can be co-located or directly coupled to one another, or in some examples, different ones of the subsystems can be in communication with one another over one or more computer networks 204. Further, although shown as part of the system, it should be appreciated that any one or more of the subsystems can function or operate as a separate system without regard to any of the other subsystems. It should also be appreciated that the system can include one or more additional or optional subsystems in addition to the subsystems shown. Figure 2A The subsystems shown can be in communication with one another via the one or more computer networks 204.

[0045] As shown, in some examples, the system 200 includes an onboard computer 206, which can correspond to the onboard component / computer 120. The onboard computer includes an onboard inference engine 208, and in some examples, the onboard computer is configured to diagnose faults on an aircraft that includes aircraft systems configured to report faults to the onboard inference engine. Examples of suitable aircraft systems include the propulsion system 110, the electrical system 114, the hydraulic system 116, and / or the environmental system 118, at least when the aircraft 202 corresponds to the aircraft 100. Any number of other systems can be included.

[0046] In some examples, the onboard computer 206 is configured to receive a fault report from one of the aircraft systems, the fault report indicating a failed test reported by the aircraft system. The onboard computer is configured to access, by the onboard reasoner 208, an onboard diagnostic causal model that can be represented by a graph that describes known causal relationships between possible failed tests reported by a respective one of the aircraft systems and possible fault modes of the respective one of the aircraft systems. In some more particular examples, the onboard diagnostic causal model is represented by a graph that includes nodes connected by edges, where the nodes represent possible failed tests and possible fault modes, and the edges indicate known causal relationships between the possible failed tests and the possible fault modes. Additionally or alternatively, in some examples, the onboard diagnostic causal model is represented by a graph that is a collection of graphs for respective ones of the aircraft systems, where the graphs reflect fault propagation behavior within the respective ones of the aircraft systems, and the collection reflects fault propagation behavior across connected ones of the aircraft systems.

[0047] With reference to Figure 2B , the onboard computer includes an onboard reasoner 208 that is configured to diagnose faults on the aircraft that includes aircraft systems configured to report faults to the onboard reasoner. The onboard computer 206 is configured to diagnose, by the onboard reasoner 208, fault modes of one or another of the aircraft systems from failed tests and using the onboard diagnostic causal model 210 and a graph theory algorithm, such as an optimal solution set (OSS) algorithm. The onboard computer is further configured to generate a fault report associated with the diagnosed fault modes, from which a maintenance message indicating a maintenance action can be generated. As Figure 2BAs shown, the off-board computer 218 is configured to maintain an on-board inference engine 208. In some examples, the off-board computer is configured to access diagnostic data received directly or indirectly from the on-board computer. The diagnostic data includes a plurality of fault reports of failed tests reported by respective ones of the aircraft systems and a plurality of diagnosed fault patterns of at least some of the aircraft systems that caused the failed tests. The off-board computer 218 includes an off-board inference engine 224 configured to determine causal relationships between the failed tests and the plurality of diagnosed fault patterns. The off-board diagnostic causal model utilizes a machine learning algorithm trained using historical diagnostic data, which can be a machine learning algorithm such as a graph theory algorithm for finding maximal cliques, such as the Bron-Kerbosch algorithm. The machine learning algorithm for finding maximal cliques uses diagnostic data to determine known causal relationships between fault patterns and failed tests, which includes historical data relating maintenance actions to maintenance messages or fault patterns (obtained from a database of actual aircraft diagnostic data, which includes different fault patterns; relationships between test reports, fault reports, and components; maintenance action data, etc.). The off-board inference engine employing the machine learning algorithm is configured to use actual fault data, diagnostic data, and / or maintenance action data received from the on-board aircraft systems (from which the relationships between fault patterns and failed tests are determined), which is compared to diagnostic data accessible from the on-board diagnostic system of the aircraft (on-board computer 206). The off-board inference engine can determine differences based on the comparison, which can then update and improve the off-board diagnostic causal model. The identification of differences can include identifying new maintenance actions from historical diagnostic data and maintenance action data collected from the on-board system of the aircraft, where the difference identification identifies new maintenance actions associated with fault patterns.

[0048] Figure 3An example graph-based representation of an onboard diagnostic causal model 300 suitable as the onboard diagnostic causal model 210 is shown in accordance with some example implementations. As shown, a fault report 302 from an aircraft brake system anti-skid wheel speed transducer 304 can indicate a failed landing gear retraction brake test 306A, 306B. As also shown, the onboard diagnostic causal model can include nodes 308A-308C connected by edges 310. The nodes can include nodes 308A, 308B representing the failed landing gear retraction brake test and a node 308C representing a fault mode having a known causal relationship with the failed test, as indicated by the edges connecting the nodes. In this regard, the failed test shares a common fault mode of misoperation of the aircraft brake system anti-skid wheel speed transducer. This fault mode can then be diagnosed from the failed test and using the onboard causal model. A maintenance action can be implemented for the fault mode, and a maintenance message 312 referencing instructions 314 for performing the maintenance action can be generated. Such maintenance action data can be collected from the aircraft condition monitoring system or other systems as historical maintenance action data and / or diagnostic data, where actual historical diagnostic data and maintenance action data can be used to determine actual causal relationships.

[0049] Returning to Figure 2A , the onboard computer 206 is configured to diagnose a fault mode of one or another of the aircraft systems from the failed test and using a graph theory algorithm and the onboard diagnostic causal model 210 by the onboard reasoner 208. One example of a suitable graph theory algorithm is an optimal solution set (OSS) algorithm. The OSS algorithm can utilize submodular minimization of data to provide an efficient distributed optimization algorithm, where data about various failed test reports can be associated with elements or categories of test reports or fault modes, and only data containing a subset of a particular failed test report or fault mode itself is considered, while all other data is ignored. The optimization algorithm can also create a graph of nodes for graph theory analysis, and can use adjacent node and / or edge probability analysis to determine a diagnosed fault mode that will describe a diagnostic causal relationship between the failed test reported by the aircraft and a possible fault mode of the reported aircraft system. The onboard computer is also configured to determine a maintenance action for the fault mode, and to generate a maintenance message 212 including at least the maintenance action.

[0050] In some examples, the onboard computer 206 is also configured to send the maintenance message 212 to a display device 214 on the aircraft or a display device on a maintenance component configured to establish a connection with the onboard computer to receive and display the maintenance message 212. In some of these examples, the display of the maintenance message references instructions 216 for performing the maintenance action to address the fault mode diagnosed by the onboard reasoner 208.

[0051] For example Figure 2A As shown, in some examples, system 200 includes an off-board computer 218 configured to maintain an onboard inference engine 208. In some examples, the off-board computer is configured to access diagnostic data received directly or indirectly from the onboard computer. This diagnostic data includes multiple failure reports of failed tests reported by corresponding aircraft systems in the aircraft system, and multiple diagnosed failure modes of at least some of the aircraft systems that caused the failed tests.

[0052] As shown in the figures, in some examples, system 200 includes at least one data source 220. In some examples, the source includes storage that can be located at a single source or distributed across multiple sources. Data can be stored in a variety of different ways, such as as a database or flat file of any of a variety of different types or formats. In some of these examples, the Aircraft Condition Monitoring System (ACMS) on aircraft 202 can collect, monitor, record, and report diagnostic data. At least some of the diagnostic data can be accessed from reports generated by the ACMS and can be transmitted directly or wirelessly via satellite 222 or network 204 to a specific data source shown and sometimes referred to as the Aircraft Health Management (AHM) system 220A. In other examples of these examples, diagnostic data can be transmitted via a wired connection or a portable data storage device (e.g., flash memory, thumb drive).

[0053] In some examples, the off-board computer 218 includes an off-board inference engine 224. The off-board computer is configured to use the off-board inference engine to construct an off-board diagnostic causal model 226, which can describe the causal relationship between failed tests and multiple diagnosed failure modes. Here, the off-board diagnostic causal model is constructed using a graph-based machine learning algorithm trained using historical diagnostic data. And in some examples, the off-board diagnostic causal model is represented by a graph including nodes connected by edges, where nodes represent failed tests and multiple diagnosed failure modes, and edges indicate known causal relationships between failed tests and multiple diagnosed failure modes. The off-board diagnostic causal model 226 employs a machine learning algorithm that compares fault data, historical diagnostic data, and / or maintenance action data received from the onboard aircraft system with diagnostic data accessible from the aircraft's onboard diagnostic system (on-board computer 206). The off-board diagnostic causal model 226 can determine discrepancies based on its comparison with the diagnostic data. Figure 3In the example on-board diagnostic causal model of FIG. 3, a graph-based representation of a fault report 302 from an aircraft brake system anti-skid wheel speed transducer 304 indicates failed landing gear extension and retraction brake tests 306A, 306B. The off-board diagnostic causal model 226 can identify differences based on collected maintenance action data, such as wheel speed transducer maintenance instructions (MM_32_30710) to define new relationships in the diagnostic causal model related to the fault to incorporate into the on-board diagnostic reasoning machine model of the aircraft.

[0054] One example of a suitable graph theory machine learning algorithm is a graph theory algorithm for finding maximal cliques, such as the Bron-Kerbosch algorithm. The graph theory machine learning algorithm for finding maximal cliques can use diagnostic data to determine known causal relationships between fault modes and failed tests. Figure 4 Diagnostic data is shown, and the relationships can be converted into a graph of maximal cliques 400 with nodes 402, 404 and edges 406 as described below, each of which is called out in the figure. Here, the nodes 402 represent possible failed tests Tl, T2, T3, and T4, and the nodes 404 represent possible fault modes F2, F4, F5, F8, and FlO. The edges indicate known causal relationships between the possible failed tests and the possible fault modes.

[0055] The maximal cliques 400 can be determined from the graph based on the diagnostic data, which in this example includes all of the nodes 402, 404 and edges 406 in the graph. In other examples, the maximal cliques can include fewer nodes and edges than all of the nodes and edges. From the maximal cliques, specific fault modes associated with failed tests can be isolated.

[0056] For example, the edges 406 can be examined for active (powered) fault modes 404 associated with failed tests reported at the nodes 402. If a clique has more than one node for a failed test, the off-board reasoning machine 224 can determine that the clique is a maximal clique (e.g., maximal clique 400). Further, if all nodes for a failed test have one or more shared fault modes determined from the edges connecting the nodes, the maximal clique can be identified as a monochromatic clique. The shared fault modes become fault modes of interest, and the monochromatic clique can then be used to identify fault components associated with the fault modes of interest. In Figure 4 In the example of FIG. 4, the maximal clique 400 is a monochromatic clique with the shared fault mode F8, as identified from the edges 406 connecting the nodes 402 representing failed tests with the node 404 representing the shared fault mode F8.

[0057] Returning again to the example of FIG. 3, the maximal clique 400 is a monochromatic clique with the shared fault mode F8, as identified from the edges 406 connecting the nodes 402 representing failed tests with the node 404 representing the shared fault mode F8. Figure 2AThe off-board diagnostic causal model 226 can be shaped by discovering any new relationships between failure modes and failure tests that are not present in the on-board diagnostic causal model 210 or that are results of emergency behavior in any aircraft system. This can be mined from historical data relating maintenance actions to maintenance messages 212 or failure modes, which can then update and improve the off-board diagnostic causal model.

[0058] The off-board diagnostic causal model 226 can have a particular signature that defines the expected propagation of a failure mode. In operation, the off-board reasoner 224 can record portions of the signature that it encounters in performing diagnostics. The off-board diagnostic causal model can be improved from this historical data (using historical signatures), which can be compared to previous off-board diagnostic causal models, such that any new relationships can be considered for addition to an updated off-board diagnostic causal model.

[0059] The off-board computer 218 is configured to compare diagnostic data from the on-board computer 206 to the off-board diagnostic causal model 226 to diagnose a failure mode of an aircraft system. In this regard, the diagnostic data can describe known causal relationships between possible failure tests reported by a respective aircraft system in the aircraft system and possible failure modes of the respective aircraft system in the aircraft system.

[0060] Based on the above, in these examples, the off-board computer 218 is configured to determine a new causal relationship in the causal relationships described by the off-board diagnostic causal model 226 that is new with respect to the known causal relationships. The off-board computer is also configured to update the on-board diagnostic causal model to further describe the new causal relationship. This includes the off-board computer being configured to produce an updated on-board diagnostic causal model and upload the updated on-board diagnostic causal model to the on-board computer 206 as a loadable software aircraft part (LSAP) via the LSAP upload toolset 228.

[0061] In some examples, the plurality of failure reports includes failure reports that indicate those failure tests caused by one of the plurality of diagnosed failure modes. In some of these examples, the off-board computer 218 is also configured to diagnose, by the off-board reasoner 224, a respective failure mode of the aircraft system from those failure tests caused by the diagnosed failure mode and using a graph theory machine learning algorithm and graph. The off-board computer is then also configured to report any differences between the respective failure mode diagnosed by the off-board reasoner and the one of the plurality of diagnosed failure modes from the diagnostic data received from the on-board computer 206.

[0062] In some examples, the diagnostic data accessed by the off-board computer 218 also includes maintenance messages 212 having maintenance actions determined by the on-board reasoner 208 for respective diagnosed failure modes of the plurality of diagnosed failure modes. In some of these examples, the on-board diagnostic causal model 210 also describes relationships between possible failure modes and respective ones of the maintenance actions. Then, for a particular maintenance message having a particular maintenance action for a respective diagnosed failure mode, the off-board computer 218 is also configured to access a maintenance record having a performed maintenance action for the respective diagnosed failure mode and thereby having a new maintenance action. The off-board computer is also configured to identify a maintenance action difference between the particular maintenance action and the performed maintenance action to determine a new relationship between the respective diagnosed failure mode and the performed maintenance action. The new relationship can be a new maintenance action identified from the historical diagnostic data and from maintenance action data collected from the aircraft, where the new maintenance action is associated with the failure mode. And the off-board computer is configured to update the on-board diagnostic causal model to describe the new relationship between the respective diagnosed failure mode and the performed maintenance action. For example, the updated on-board diagnostic causal model can also include a new maintenance action identified from the collected historical maintenance action data that is associated with the failure mode.

[0063] Figure 5A , Figure 5B and Figure 5C is a flowchart illustrating various steps in a method 500 of diagnosing a failure on an aircraft 202, the aircraft including aircraft systems configured to report failures to an on-board reasoner 208, in accordance with example implementations of the present disclosure.

[0064] As shown in Figure 5A , at block 502, the method 500 includes receiving, at an on-board computer 206 of the aircraft 202 including the on-board reasoner 208, a failure report from one of the aircraft systems, the failure report indicating a failed test reported by the aircraft system. As shown in block 504, the method includes accessing, by the on-board reasoner, an on-board diagnostic causal model 210 represented by a graph that describes known causal relationships between possible failed tests reported by respective ones of the aircraft systems and possible failure modes of the respective ones of the aircraft systems.

[0065] As shown in block 506, the method includes diagnosing, by the onboard reasoning machine 208, a failure mode of one of the aircraft systems or another of the aircraft systems in accordance with the failed tests and using a graph theory algorithm (e.g., the OSS algorithm) and the onboard diagnostic causal model 210. As shown in block 508, the method includes determining a maintenance action for the failure mode. And as shown in block 510, the method includes generating, by the onboard computer 206, a maintenance message 212 including at least the maintenance action.

[0066] In some examples, the method 500 further includes sending the maintenance message 212 to a display device 214 on the aircraft 202 or a display device on a maintenance component configured to establish a connection with the onboard computer to receive the maintenance message, as shown in block 512.

[0067] As Figure 5B shown in some examples, the failure report is one of a plurality of failure reports of the failed tests reported by the respective ones of the aircraft systems, and the failure mode is one of a plurality of diagnosed failure modes of at least some of the aircraft systems that caused the failed tests. As shown in block 514, the method 500 further includes accessing the diagnostic data received from the onboard computer 206, and the diagnostic data includes the plurality of failure reports and the plurality of diagnosed failure modes.

[0068] As shown in block 516, the method 500 further includes accessing, by the non- onboard reasoning machine 224, a non-onboard diagnostic causal model 226 that describes causal relationships between the failed tests and the plurality of diagnosed failure modes. The non- onboard diagnostic causal model is constructed with a graph theory machine learning algorithm trained using historical diagnostic data. The method further includes comparing the diagnostic data to the non-onboard diagnostic causal model, as shown in block 518.

[0069] Based on the comparison at block 518, the method 500 further includes identifying a difference between the diagnostic data and the non-onboard diagnostic causal model 226 to determine a new causal relationship in the causal relationships described by the non- onboard diagnostic causal model 226 that is new with respect to the known causal relationships, as shown in block 520. As shown in block 522, the method further includes updating the onboard diagnostic causal model to further describe the new causal relationship, including producing an updated onboard diagnostic causal model, and uploading the updated onboard diagnostic causal model to the onboard components using the LSAP upload toolset 228.

[0070] As Figure 5CAs shown, in some examples, the diagnostic data also includes maintenance messages 212 having maintenance actions determined by the onboard reasoning machine 208 for respective diagnosed fault modes of the plurality of diagnosed fault modes, and the onboard diagnostic causal model 210 also describes relationships between possible fault modes and respective ones of the maintenance actions. For a particular maintenance message having a particular maintenance action for a respective diagnosed fault mode, the method 500 also includes accessing a maintenance record having a performed maintenance action for the respective diagnosed fault mode and thereby having a new maintenance action, as shown in block 524. As shown in block 526, the method also includes identifying a maintenance action difference between the particular maintenance action and the performed maintenance action to determine a new relationship between the respective diagnosed fault mode and the performed maintenance action. The step of identifying the maintenance action difference can also include identifying the new maintenance action from historical diagnostic data and from maintenance action data collected from the aircraft state monitoring system, where the difference identification identifies the new maintenance action associated with the fault mode. Updating the onboard diagnostic causal model then also includes updating the onboard diagnostic causal model to describe the new relationship between the respective diagnosed fault mode and the performed maintenance action, as shown in block 522A.

[0071] Figure 6A , Figure 6B and Figure 6C is a flowchart illustrating various steps in a method 600 of maintaining an onboard reasoning machine 208 to diagnose faults on an aircraft 202, the aircraft including aircraft systems configured to report faults to the onboard reasoning machine. As shown in block 602, the method includes accessing diagnostic data received from onboard computers 206 of the aircraft including the onboard reasoning machine, the diagnostic data including a plurality of fault reports of failed tests reported by respective ones of the aircraft systems and a plurality of diagnosed fault modes of at least some of the aircraft systems that caused the failed tests. Figure 6A

[0072] As shown in block 604, the method 600 includes constructing an offboard diagnostic causal model 226 using the offboard reasoning machine 224, the offboard diagnostic causal model describing causal relationships between failed tests and the plurality of diagnosed fault modes. The offboard diagnostic causal model is constructed with a graph theory machine learning algorithm trained using historical diagnostic data.

[0073] ​As shown in block 606, the method 600 includes comparing the diagnostic data to the off-board diagnostic causal model 226. Based on the comparison at block 606, the method 600 includes determining a new causal relationship in the causal relationships described by the off-board diagnostic causal model 226 that is new with respect to the known causal relationships, as shown in block 608. As shown in block 610, the method includes updating the on-board diagnostic causal model to further describe the new causal relationship, including generating an updated on-board diagnostic causal model, and uploading the updated on-board diagnostic causal model to the on-board computer using the LSAP upload toolset 228.

[0074] In some examples, the plurality of failure reports includes failure reports indicating those failed tests caused by one of the plurality of diagnosed fault modes, and as shown in block 612, the method 600 further includes diagnosing, by the off-board reasoning machine 224, a corresponding fault mode of the aircraft system from those failed tests caused by the diagnosed fault mode and using a graph theory machine learning algorithm and graph. As shown in block 614, the method further includes reporting any differences between the corresponding fault mode diagnosed by the off-board reasoning machine and the one of the plurality of diagnosed fault modes according to the diagnostic data received from the on-board computer 206. Figure 6B

[0075] In some examples, the diagnostic data further includes maintenance messages 212 having a maintenance action determined by the on-board reasoning machine 208 for a corresponding one of the plurality of diagnosed fault modes, and the on-board diagnostic causal model 210 further describes relationships between possible fault modes and corresponding ones of the maintenance actions. For a particular maintenance message having a particular maintenance action for a corresponding diagnosed fault mode, as shown in block 616, the method 600 further includes accessing a maintenance record having a performed maintenance action for the corresponding diagnosed fault mode and thereby having a new maintenance action. The method further includes identifying a maintenance action difference between the particular maintenance action and the performed maintenance action to determine a new relationship between the corresponding diagnosed fault mode and the performed maintenance action, as shown in block 618. Updating the on-board diagnostic causal model then further includes updating the on-board diagnostic causal model to describe the new relationship between the corresponding diagnosed fault mode and the performed maintenance action, as shown in block 610A. Figure 6C

[0076] ​​In some examples, the method 600 can incorporate the method 500, where the updated onboard diagnostic causal model can be used to diagnose a failure mode of an aircraft system. This can include receiving a failure report at the onboard computer 206 of the aircraft 202 including the onboard reasoner 208 (see block 502). The onboard reasoner can access the updated onboard diagnostic causal model 210 and diagnose a failure mode of one of the aircraft systems or another aircraft system according to the failed test and using the graph theory algorithm and the updated onboard diagnostic causal model (see blocks 504 and 506). A maintenance action for the failure mode can be determined, a maintenance message 212 including at least the maintenance action can be generated (see blocks 508 and 510).

[0077] According to example implementations of the present disclosure, the system 200 and its subsystems can be implemented by various means. Means for implementing the system and its subsystems can include hardware alone or in combination with one or more computer programs from a computer-readable storage medium. In some examples, one or more devices can be configured to function as or otherwise implement the system and its subsystems shown and described herein. In examples involving more than one device, the respective devices can be connected to each other or otherwise in communication with each other in a number of different ways, such as directly or indirectly via wired or wireless networks, etc.

[0078] Figure 7 Devices 700 of the onboard computer 206 and / or the non-onboard computer 218 shown are illustrated as being configured to implement the system 200 and its subsystems in various example implementations. Figure 2A Generally, devices of example implementations of the present disclosure can comprise, include, or be embodied in one or more fixed or portable electronic devices. Examples of suitable electronic devices include smartphones, tablet computers, laptop computers, desktop computers, workstation computers, server computers, etc. The devices can include one or more of each of a number of components, such as, for example, a processing circuit 702 (e.g., a processor unit) connected to a memory 704 (e.g., a storage device).

[0079] The processing circuit 702 can consist of one or more processors alone or in combination with one or more memories. Processing circuitry is generally any computer hardware capable of processing information such as, for example, data, computer programs, and / or other suitable electronic information. Processing circuitry consists of a collection of electronic circuits some of which can be packaged as integrated circuits or multiple interconnected integrated circuits (integrated circuits are sometimes more generally referred to as “chips”). The processing circuitry can be configured to execute a computer program that can be stored on the processing circuitry or otherwise stored in the memory 704 (of the same or another device).

[0080] According to particular implementations, the processing circuit 702 can be a plurality of processors, a multi-core processor, or some other type of processor. Further, the processing circuit can be implemented with a plurality of heterogeneous processor systems in which a host processor is present on a single chip with one or more secondary processors. As another illustrative example, the processing circuit can be a symmetric multi-processor system containing a plurality of processors of the same type. In yet another example, the processing circuit can be implemented as or otherwise include one or more ASICs, FPGAs, etc. Thus, while the processing circuit is capable of executing a computer program to perform one or more functions, the processing circuit of various examples is capable of performing the function or functions without the aid of a computer program. In any instance, the processing circuit can be suitably programmed to perform the functions or operations according to example implementations of the present disclosure.

[0081] The memory 704 is generally any computer hardware able to store information such as data, computer program (e.g., computer-readable program code 706), and / or other suitable information, on a temporary or permanent basis. The memory can include volatile and / or non-volatile memory and can be fixed or removable. Examples of suitable memory include random access memory (RAM), read only memory (ROM), hard drives, flash memory, thumb drives, removable computer disks, optical disks, magnetic tapes, or some combination of the aforementioned. The optical disks can include compact discs - read only memories (CD-ROMs), compact discs - read / write (CD-R / W), DVDs, etc. In various instances, the memory can be referred to as a computer-readable storage medium. A computer-readable storage medium is a non-transitory device able to store information and distinguish from a computer-readable transmission medium such as an electrical, optical, or electromagnetic transitory signal capable of carrying information from one location to another. A computer-readable medium as described herein can generally refer to a computer-readable storage medium or a computer-readable transmission medium.

[0082] In addition to the memory 704, the processing circuit 702 can be connected to one or more interfaces for displaying, transmitting and / or receiving information. The interfaces can include a communication interface 708 (e.g., a communication unit) and / or one or more user interfaces. The communication interface can be configured to transmit and / or receive information, such as to and / or from other devices, networks, etc. The communication interface can be configured to transmit and / or receive information over physical (wired) and / or wireless communication links. Examples of suitable communication interfaces include network interface controllers (NICs), wireless NICs (WNICs), etc.

[0083] The user interface can include a display device 710 and / or one or more user input interfaces 712 (e.g., input / output units). The display device can be configured to present or otherwise display information to a user, suitable examples of which include a liquid crystal display (LCD), a light-emitting diode display (LED), a plasma display panel (PDP), etc. The user input interface can be wired or wireless, and can be configured to receive information from a user into the device, such as for processing, storage, and / or display. Suitable examples of user input interfaces include a microphone, an image or video capture device, a keyboard or keypad, a joystick, a touch-sensitive surface (separate from or integrated into a touchscreen), a biometric sensor, etc. The user interface can also include one or more interfaces for communicating with peripheral devices such as printers, scanners, etc.

[0084] As indicated above, program code instructions can be stored in memory and executed by processing circuitry programmed thereby to implement the functions of the systems, subsystems, tools, and their respective elements described herein. As will be appreciated, any suitable program code instructions can be loaded onto a computer or other programmable apparatus to produce a particular machine, such that the particular machine becomes a means for implementing the functions specified herein. These program code instructions can also be stored in a computer-readable storage medium that can direct a computer, a processing circuitry, or other programmable apparatus to function in a particular manner, such that the particular machine or particular article of manufacture becomes a means for implementing the functions specified herein. The instructions stored in the computer-readable storage medium can produce an article of manufacture, where the article of manufacture becomes a means for implementing the functions specified herein. The program code instructions can be retrieved from the computer-readable storage medium and loaded into a computer, processing circuitry, or other programmable apparatus to configure the computer, processing circuitry, or other programmable apparatus to execute operations to implement the functions specified in the computer, processing circuitry, or other programmable apparatus, or to operate in conjunction with the computer, processing circuitry, or other programmable apparatus.

[0085] The retrieval, loading, and execution of the program code instructions can be performed in sequence such that one instruction is retrieved, loaded, and executed at a time. In some example implementations, the retrieval, loading, and / or execution can be performed in parallel such that multiple instructions are retrieved, loaded, and / or executed together. The execution of the program code instructions can result in a computer-implemented process such that the instructions executed by the computer, processing circuitry, or other programmable apparatus provide operations for implementing the functions described herein.

[0086] Execution of instructions by processing circuitry or storage of instructions in a computer-readable storage medium supports combinations of operations for performing the specified functions. In this manner, the device 700 can include processing circuitry 702 and a computer-readable storage medium or memory 704 coupled to the processing circuitry, where the processing circuitry is configured to execute computer-readable program code 706 stored in the memory. It will also be appreciated that one or more functions and combinations of functions can be implemented by special purpose hardware-based computer systems which perform some or all of the operations described, and / or by a combination of dedicated hardware and program code instructions.

[0087] Further, the present disclosure includes implementations in accordance with the following:

[0088] Item 1. A non-onboard computer (218, 700) for maintaining an onboard inference machine (208) to diagnose faults on an airplane (202), the airplane including airplane systems configured to report faults to the onboard inference machine, the non-onboard computer comprising:

[0089] a memory (704) configured to store computer-readable program code (706) including the onboard inference machine (224); and

[0090] processing circuitry (702) configured to access the memory and execute the computer-readable program code to cause the non-onboard computer to at least:

[0091] access (602) diagnostic data received from an onboard computer (206, 700) of the airplane including the onboard inference machine, the diagnostic data including a plurality of fault reports (302) of failed tests (306A, 306B) reported by respective ones of the airplane systems and a plurality of diagnosed fault patterns (308C) of at least some of the airplane systems that caused the failed tests;

[0092] construct (604), using the non-onboard inference machine (224), a non-onboard diagnostic causal model (226) that describes causal relationships between the failed tests (402) and the plurality of diagnosed fault patterns (404), the non-onboard diagnostic causal model constructed using a graph theory machine learning algorithm trained using historical diagnostic data;

[0093] compare (606) the diagnostic data to the non-onboard diagnostic causal model; and based on the comparison,

[0094] determine (608) a new one of the causal relationships described by the non-onboard diagnostic causal model, the new causal relationship being new with respect to known causal relationships; and

[0095] updating (610) the on-board diagnostic causal model to further describe new causal relationships, including causing the off-board computer to generate an updated on-board diagnostic causal model, and uploading the updated on-board diagnostic causal model to the on-board computer using the loadable software airplane component upload toolset (228).

[0096] Clause 2. The off-board computer (218, 700) of Clause 1, wherein the off-board diagnostic causal model (226) is represented by a graph (400) comprising nodes (402, 404) connected by edges (406), the nodes representing failed tests (402) and a plurality of diagnosed fault modes (404), and the edges indicating known causal relationships between the failed tests and the plurality of diagnosed fault modes.

[0097] Clause 3. The off-board computer (218, 700) of Clause 1, wherein the graph theory machine learning algorithm is a graph theory algorithm for finding a maximum clique.

[0098] Clause 4. The off-board computer (218, 700) of Clause 1, wherein the plurality of fault reports includes fault reports (302) indicating those failed tests (306A, 306B) caused by one of the plurality of diagnosed fault modes (308C), and further causing the off-board computer to:

[0099] diagnose (612), by the off-board reasoning machine (224), a respective fault mode of the airplane system from those failed tests (402) caused by the diagnosed fault modes (404) and using the graph theory machine learning algorithm and the graph (400); and

[0100] report (614), to the on-board computer (206, 700), any discrepancies between the respective fault mode and one of the plurality of diagnosed fault modes.

[0101] Clause 5. The off-board computer (218, 700) of Clause 1, wherein the diagnostic data further includes maintenance messages (212, 312) having maintenance actions determined by the on-board reasoning machine (208) for respective ones of the plurality of diagnosed fault modes (308C), and the on-board diagnostic causal model (210, 300) further describes relationships between possible fault modes and respective ones of the maintenance actions,

[0102] wherein, for a particular maintenance message having a particular maintenance action for a respective diagnosed fault mode, the processing circuitry (702) is configured to execute the computer-readable program code (706) to cause the off-board computer to further at least:

[0103] accessing (616) maintenance records having performed maintenance actions for the respective diagnosed fault modes; and

[0104] identifying (618) maintenance action differences between the particular maintenance action and the performed maintenance actions to determine new relationships between the respective diagnosed fault modes and the performed maintenance actions, and

[0105] wherein the non-onboard computer that causes the updating of the onboard diagnostic causal model further comprises a non-onboard computer that causes updating (610A) of the onboard diagnostic causal model to describe the new relationships between the respective diagnosed fault modes and the performed maintenance actions.

[0106] Item 6. A method (600) of maintaining an onboard reasoner (208) to diagnose faults on an aircraft (202), the aircraft comprising aircraft systems configured to report faults to the onboard reasoner, the method comprising:

[0107] accessing (602) diagnostic data received from onboard computers (206) of the aircraft including the onboard reasoner, the diagnostic data comprising a plurality of fault reports (302) of failed tests (306A, 306B) reported by respective ones of the aircraft systems and a plurality of diagnosed fault modes (308C) of at least some of the aircraft systems that caused the failed tests;

[0108] constructing (604), using a non-onboard reasoner (224), a non-onboard diagnostic causal model (226) that describes causal relationships between failed tests (402) and a plurality of diagnosed fault modes (404), the non-onboard diagnostic causal model being constructed using a graph theory machine learning algorithm trained using historical diagnostic data;

[0109] comparing (606) the diagnostic data to the non-onboard diagnostic causal model; and based on the comparison,

[0110] determining (608) a new one of the causal relationships described by the non-onboard diagnostic causal model, the new causal relationship being new with respect to known causal relationships; and

[0111] updating (610) the onboard diagnostic causal model to further describe the new causal relationship, including producing an updated onboard diagnostic causal model, and uploading the updated onboard diagnostic causal model to the onboard computer using a loadable software aircraft component upload toolset (228).

[0112] Item 7. The method (600) of item 6, wherein the off-board diagnostic causal model (226) is represented by a graph (400) comprising nodes (402, 404) connected by edges (406), the nodes representing failed tests (402) and a plurality of diagnosed fault modes (404), and the edges indicating known causal relationships between the failed tests and the plurality of diagnosed fault modes.

[0113] Item 8. The method (600) of item 6, wherein the comparison of the diagnostic data to the off-board diagnostic causal model further comprises a comparison to actual fault data collected from the on-board aircraft system reflecting actual causal relationships, diagnostic data, and maintenance action data.

[0114] Item 9. The method (600) of item 6, wherein the graph theory machine learning algorithm is a graph theory algorithm for finding a maximum clique.

[0115] Item 10. The method (600) of item 6, wherein the plurality of fault reports includes fault reports (302) indicating those failed tests (306A, 306B) caused by one of the plurality of diagnosed fault modes (308C), and the method further comprises:

[0116] diagnosing (612), by the off-board reasoning machine (224), a corresponding fault mode of the aircraft system from those failed tests (402) caused by the diagnosed fault mode (404) and using the graph theory machine learning algorithm and the graph (400); and

[0117] reporting (614), from the diagnostic data received from the on-board computer (206, 700), any discrepancies between the corresponding fault mode and one of the plurality of diagnosed fault modes.

[0118] Item 11. The method (600) of item 6, wherein the diagnostic data further includes maintenance messages (212, 212) having maintenance actions determined by the on-board reasoning machine (208) for a corresponding one of the plurality of diagnosed fault modes (308C), and the on-board diagnostic causal model (210, 300) further describes relationships between possible fault modes and corresponding ones of the maintenance actions, and

[0119] wherein, for a particular maintenance message having a particular maintenance action for a corresponding diagnosed fault mode, the method further comprises:

[0120] accessing (616) a maintenance record having the performed maintenance action for the corresponding diagnosed fault mode; and

[0121] identifying (618) a maintenance action difference between the particular maintenance action and the performed maintenance action to determine a new relationship between the respective diagnosed fault mode and the performed maintenance action,

[0122] wherein updating the on-board diagnostic causal model further comprises updating (610A) the on-board diagnostic causal model to describe the new relationship between the respective diagnosed fault mode and the performed maintenance action.

[0123] Item 12. The method (600, 500) of item 6, further comprising:

[0124] receiving (502), at an on-board computer (206, 700) of the aircraft (202) comprising an on-board reasoner (208), a fault report (302) from one of the aircraft systems, the fault report indicating a failed test (306A, 306B) reported by the aircraft system;

[0125] accessing (504), by the on-board reasoner, the updated on-board diagnostic causal model (210, 300);

[0126] diagnosing (506), by the on-board reasoner, a fault mode (308C) of one of the aircraft systems or another aircraft system from the failed test (308A, 308B) and using a graph theory algorithm and the updated on-board diagnostic causal model;

[0127] determining (508) a maintenance action for the fault mode; and

[0128] generating (510), by the on-board computer, a maintenance message (212, 312) comprising at least the maintenance action.

[0129] Item 13. The method (600) of item 12, wherein the graph theory algorithm is an optimal solution set (OSS) algorithm.

[0130] Item 14. A system (200) for maintaining an on-board reasoner (208) to diagnose faults on an aircraft (202), the aircraft comprising aircraft systems configured to report faults to the on-board reasoner, the system comprising:

[0131] an on-board computer (206, 700) comprising the on-board reasoner, the on-board computer configured to send diagnostic data comprising a plurality of fault reports (302) of failed tests (306A, 306B) reported by respective ones of the aircraft systems and a plurality of diagnosed fault modes (308C) of at least some of the aircraft systems that caused the failed tests; and

[0132] an off-board computer (218, 700) comprising an off-board reasoner (224), the off-board computer configured to:

[0133] accessing (602) diagnostic data received from the onboard computer;

[0134] constructing (604) an off-board diagnostic causal model (226) that describes causal relationships between failed tests (402) and a plurality of diagnosed fault modes (404), the off-board diagnostic causal model constructed using a graph theory machine learning algorithm trained using historical diagnostic data;

[0135] comparing (606) the diagnostic data to the off-board diagnostic causal model; and based on the comparison,

[0136] determining (608) a new causal relationship in the causal relationships described by the off-board diagnostic causal model, the new causal relationship being new with respect to known causal relationships; and

[0137] updating (610) the onboard diagnostic causal model to further describe the new causal relationship, including the off-board computer being configured to produce an updated onboard diagnostic causal model, and uploading the updated onboard diagnostic causal model to the onboard computer using a loadable software airplane component upload toolset (228).

[0138] Item 15. The system (200) of item 14, wherein the off-board diagnostic causal model (226) is represented by a graph (400) comprising nodes (402, 404) connected by edges (406), the nodes representing failed tests (402) and a plurality of diagnosed fault modes (404), and the edges indicating known causal relationships between failed tests and the plurality of diagnosed fault modes.

[0139] Item 16. The system (200) of item 14, wherein the comparison of the diagnostic data to the off-board diagnostic causal model further comprises a comparison to actual fault data, diagnostic data, and maintenance action data collected from the onboard airplane system that reflect actual causal relationships.

[0140] Item 17. The system (200) of item 14, wherein the graph theory machine learning algorithm is a graph theory algorithm for finding a maximum clique.

[0141] Item 18. The system (200) of item 14, wherein the plurality of fault reports includes fault reports (302) that indicate those failed tests (306A, 306B) caused by a diagnosed fault mode (308C) of the plurality of diagnosed fault modes, and the off-board computer (218, 700) is further configured to:

[0142] diagnosing (612), by the off-board reasoner (224), a respective failure mode of the aircraft system from those failed tests (402) caused by the diagnosed failure mode (404) and using the graph theory machine learning algorithm and graph (400); and

[0143] reporting (614) any discrepancies between the respective failure mode and one of the plurality of diagnosed failure modes from the diagnostic data received from the on-board computer (206, 700).

[0144] Item 19. The system (200) of item 14, wherein the diagnostic data further comprises maintenance messages (212, 312) having maintenance actions determined by the on-board reasoner (208) for respective ones of the plurality of diagnosed failure modes (308C), and the on-board diagnostic causal model (210, 300) further describes relationships between possible failure modes and respective ones of the maintenance actions,

[0145] wherein, for a particular maintenance message having a particular maintenance action for a respective diagnosed failure mode, the off-board computer (218, 700) is configured to:

[0146] access (616) a maintenance record having the performed maintenance action for the respective diagnosed failure mode; and

[0147] identify (618) a maintenance action discrepancy between the particular maintenance action and the performed maintenance action to determine a new relationship between the respective diagnosed failure mode and the performed maintenance action, and

[0148] wherein the off-board computer configured to update the on-board diagnostic causal model further comprises the off-board computer configured to update (610A) the on-board diagnostic causal model to describe the new relationship between the respective diagnosed failure mode and the performed maintenance action.

[0149] Item 20. The system (200) of item 14, wherein the on-board computer (206) is further configured to:

[0150] receive (502), from one of the aircraft systems, a failure report (302) of the aircraft (202) indicating failed tests (306A, 306B) reported by the aircraft system;

[0151] access (504), by the on-board reasoner (208), the updated on-board diagnostic causal model (210, 300);

[0152] diagnose (506), by the on-board reasoner, a failure mode of one of the aircraft systems or another aircraft system in the aircraft systems according to the failed tests (308A, 308B) and using a graph theory algorithm and the updated on-board diagnostic causal model (308C);

[0153] determine (508) a maintenance action for the failure mode; and

[0154] generate (510) a maintenance message (212, 312) including at least the maintenance action.

[0155] Item 21. The system (200) of item 20, wherein the graph theory algorithm is an optimal solution set (OSS) algorithm.

[0156] Many modifications and other implementations of the present disclosure will occur to those of ordinary skill in the art upon reading the foregoing description and associated drawings in the present disclosure. Therefore, it is to be understood that the present disclosure is not to be limited to the particular implementations disclosed and that modifications and other implementations are intended to be included within the scope of the appended claims. Moreover, although the foregoing description and the associated drawings describe example implementations in the context of certain example combinations of elements and / or functions, it should be appreciated that different combinations of elements and / or functions can be provided by alternative implementations without departing from the scope of the appended claims. In this regard, for example, an element described herein functioning as a certain type of component in one implementation can be combined with an element that functions as a different type of component in another implementation. Likewise, given the typically endless number of possible combinations of elements and functions, different features can be combined using a plurality of other equivalent elements, thereby providing additional implementations of the present disclosure. Although the elements and / or functions of the present disclosure are primarily described in the singular and / or the negative, the singular is not excludes the plural and vice versa, unless the context clearly dictates otherwise.

Claims

1. A method for maintaining an airborne inference engine to diagnose faults on an aircraft, the aircraft including an aircraft system configured to report faults to the airborne inference engine, the method comprising: Access diagnostic data received from an onboard computer of the aircraft, including the onboard inference engine, the diagnostic data including multiple failure reports of failed tests reported by corresponding aircraft systems in the aircraft system and multiple diagnosed failure modes of at least some aircraft systems in the aircraft system that caused the failed tests. An off-board diagnostic causal model is constructed using an off-board inference engine of an off-board computer. The off-board diagnostic causal model describes the causal relationship between the failed test and the multiple diagnosed failure modes. The off-board diagnostic causal model is constructed using a graph theory machine learning algorithm trained with historical diagnostic data. The diagnostic data is compared with the off-aircraft diagnostic causal model using the off-aircraft inference engine; and based on the comparison... The off-board inference engine is used to determine new causal relationships within the causal relationships described by the off-board diagnostic causal model, these new causal relationships being novel relative to known causal relationships; and The onboard diagnostic causal model is updated using the off-board inference engine to further describe the new causal relationships, including generating an updated onboard diagnostic causal model and uploading the updated onboard diagnostic causal model to the onboard computer using a loadable software aircraft component upload toolset. The comparison of the diagnostic data with the off-aircraft diagnostic causal model further includes comparing the diagnostic data with actual fault data, diagnostic data, and maintenance action data collected from airborne aircraft systems that reflect the actual causal relationships.

2. The method according to claim 1, wherein, The off-board diagnostic causal model is represented by a graph comprising nodes connected by edges, the nodes representing the failed test and the plurality of diagnosed failure modes, and the edges indicating known causal relationships between the failed test and the plurality of diagnosed failure modes.

3. The method according to claim 1, wherein, The graph theory machine learning algorithm is a graph theory algorithm used to find maximal cliques.

4. The method according to claim 2, wherein, The plurality of fault reports include fault reports indicating those failed tests caused by a fault mode diagnosed by one of the plurality of diagnosed fault modes, and the method further includes: The off-board inference engine uses the graph theory machine learning algorithm and the graph to diagnose corresponding failure modes of the aircraft system based on the failure tests caused by the plurality of diagnosed failure modes; and The diagnostic data received from the onboard computer reports any differences between the corresponding fault mode and one of the plurality of diagnosed fault modes.

5. The method according to claim 1, wherein, The diagnostic data also includes maintenance messages with maintenance actions determined by the onboard inference engine for the corresponding diagnosed fault modes among the plurality of diagnosed fault modes, and the onboard diagnostic causal model further describes the relationship between possible fault modes and the corresponding maintenance actions among the maintenance actions. The method further includes, for a specific maintenance message having a specific maintenance action for the corresponding diagnosed fault mode, the following: The off-board inference engine is used to access maintenance records containing maintenance actions performed for the corresponding diagnosed fault modes; and The off-board inference engine is used to identify the maintenance action differences between the specific maintenance action and the performed maintenance action, in order to determine a new relationship between the corresponding diagnosed fault mode and the performed maintenance action. Updating the airborne diagnostic causal model using the off-board inference engine further includes updating the airborne diagnostic causal model to describe a new relationship between the corresponding diagnosed fault mode and the performed maintenance action.

6. The method according to claim 1, further comprising: The fault report is received from one of the aircraft systems at the onboard computer of the aircraft, which includes the onboard inference engine, and the fault report indicates a failed test reported by the one aircraft system. The updated airborne diagnostic causal model is accessed by the airborne inference engine; The onboard inference engine uses the graph theory machine learning algorithm and the updated onboard diagnostic causal model to diagnose the failure modes of one or more aircraft systems based on the failure tests. Determine the maintenance actions for the said failure mode; and The onboard computer generates a maintenance message that includes at least the maintenance actions.

7. The method according to claim 6, wherein, The graph theory machine learning algorithm is the optimal solution set (OSS) algorithm.

8. A system for maintaining an airborne inference engine to diagnose malfunctions on an aircraft, the aircraft including an aircraft system configured to report malfunctions to the airborne inference engine, the system comprising: An onboard computer including the airborne inference engine, the onboard computer being configured to transmit diagnostic data, the diagnostic data including multiple fault reports of failed tests reported by corresponding aircraft systems in the aircraft system, and multiple diagnosed fault modes of at least some of the aircraft systems that caused the failed tests; and An off-board computer including an off-board inference engine, the off-board computer being configured as follows: Access the diagnostic data received from the onboard computer; An off-airborne diagnostic causal model is constructed using the off-airborne inference engine. The off-airborne diagnostic causal model describes the causal relationship between the failed test and the multiple diagnosed failure modes. The off-airborne diagnostic causal model is constructed using a graph theory machine learning algorithm trained with historical diagnostic data. The diagnostic data is compared with the off-aircraft diagnostic causal model using the off-aircraft inference engine; and based on the comparison... The off-board inference engine is used to determine new causal relationships within the causal relationships described by the off-board diagnostic causal model, these new causal relationships being novel relative to known causal relationships; and Updating the onboard diagnostic causal model using the off-board inference engine to further describe the new causal relationships includes configuring the off-board computer to generate the updated onboard diagnostic causal model and uploading the updated onboard diagnostic causal model to the onboard computer using a loadable software aircraft component upload toolset. The comparison between the diagnostic data and the off-aircraft diagnostic causal model also includes a comparison with actual fault data, diagnostic data, and maintenance action data collected from the airborne aircraft system that reflect the actual causal relationship.

9. The system according to claim 8, wherein, The off-board diagnostic causal model is represented by a graph comprising nodes connected by edges, the nodes representing the failed test and the plurality of diagnosed failure modes, and the edges indicating known causal relationships between the failed test and the plurality of diagnosed failure modes.

10. The system according to claim 8, wherein, The graph theory machine learning algorithm is a graph theory algorithm used to find maximal cliques.

11. The system according to claim 9, wherein, The plurality of fault reports include fault reports indicating those failed tests caused by a fault mode diagnosed by one of the plurality of diagnosed fault modes, and the off-board computer is further configured to: The off-board inference engine uses the graph theory machine learning algorithm and the graph to diagnose corresponding failure modes of the aircraft system based on the failure tests caused by the plurality of diagnosed failure modes; and The diagnostic data received from the onboard computer reports any differences between the corresponding fault mode and one of the plurality of diagnosed fault modes.

12. The system according to claim 8, wherein, The diagnostic data also includes maintenance messages with maintenance actions determined by the onboard inference engine for the corresponding diagnosed fault modes among the plurality of diagnosed fault modes, and the onboard diagnostic causal model further describes the relationship between possible fault modes and corresponding maintenance actions among the maintenance actions. Specifically, for a specific maintenance message having a specific maintenance action for the corresponding diagnosed fault mode, the off-board inference engine is configured as follows: Access maintenance records containing the maintenance actions performed for the corresponding diagnosed fault modes; and Identify the differences between the specific maintenance action and the performed maintenance action to determine a new relationship between the corresponding diagnosed fault mode and the performed maintenance action, and The non-airborne computer configured to update the airborne diagnostic causal model also includes the non-airborne computer configured to update the airborne diagnostic causal model to describe a new relationship between the corresponding diagnosed fault mode and the maintenance action performed.

13. The system according to claim 8, wherein, The onboard computer is also configured to: Receive a fault report from one of the aircraft systems, the fault report indicating a failed test reported by the one aircraft system; The updated airborne diagnostic causal model is accessed by the airborne inference engine; The onboard inference engine uses the graph theory machine learning algorithm and the updated onboard diagnostic causal model to diagnose the failure modes of one or more aircraft systems based on the failure tests. Determine the maintenance actions for the said failure mode; and Generate a maintenance message that includes at least the maintenance action.

Images