A distributed airborne fault diagnosis reasoning method

Abstract

The application discloses a kind of distributed airborne fault diagnosis reasoning methods, comprising: according to the test point and the failure mode of analysis in the testability report of avionics system, construct the system fault model of the form of "evidence-failure bidirectional graph", system fault model is combined by three layers of fault model of member system level, regional level and aircraft level;Member system level is according to sensor threshold range to judge the "normal / abnormal" result of measured value or receiver test result to judge test "pass / fail" result, both are regarded as binary evidence, and the probability of fault assumption in the fault fuzzy group contained by fault condition is calculated;Regional level is divided according to function domain, jurisdiction multiple member systems, collect the fault condition reported, realize fault isolation and cascade elimination;All regional level reported fault conditions of aircraft level comprehensive avionics system are integrated in time dimension on time fusion.The application realizes airborne fault diagnosis and improves aircraft safety.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computer software, in particular to a distributed airborne fault diagnosis reasoning method for aircraft avionics system faults. BACKGROUND

[0002] In recent years, with the rapid progress of communication, computer and other related technologies, avionics system has developed rapidly, and the level of integration, intelligence and modularization has been continuously improved, and has become an indispensable part of the aircraft. Therefore, the risk and harm of avionics system faults have also increased day by day, which has brought great difficulties to the normal use and maintenance support of the aircraft. It is estimated that in the whole life cycle, the fault of avionics system accounts for more than 40% of the total fault of the aircraft, and the maintenance cost accounts for more than 70% of the total cost of the aircraft. In order to ensure the safety of aircraft flight and task, it is an important challenge to correctly and quickly diagnose the faults of avionics system components and subsystems,

[0003] Fault diagnosis involves two processes of fault isolation and fault identification, aiming to determine the location and mechanism of potential failure causes, and to provide correct guidance for fault mitigation and maintenance support. Fault diagnosis relies on the monitoring information provided by the state monitoring and anomaly detection functions, but when a set of specific monitoring information corresponds to multiple components or systems with possible fault modes, and it is impossible to determine which one contains the mechanism causing the fault. Fault diagnosis reasoning integrates the monitoring information of components or systems, extracts the potential causal relationship between components, systems and harmful events affecting safety by using data mining technology, and isolates and identifies the anomalies and faults of aircraft avionics system by using causal reasoning algorithm.

[0004] Typical faults of avionics system often have the characteristics of cascading and false alarm, which will greatly affect the correctness of fault diagnosis. Fault cascading refers to the failure of one or more aircraft subsystems, which may affect other subsystems, and these faults may affect each other. False alarm is often caused by sensor data deviation or unreasonable threshold setting, etc., resulting in false alarm of in-flight test. SUMMARY

[0005] The purpose of the present application is to provide a distributed airborne fault diagnosis reasoning method to realize airborne fault diagnosis and improve the safety of aircraft.

[0006] In order to achieve the above-mentioned task, the present application adopts the following technical scheme:

[0007] A distributed airborne fault diagnosis reasoning method, comprising the following steps:

[0008] According to the test points set in the avionics system testability report and the analyzed fault modes, a system fault model in the form of "evidence-fault bidirectional graph" is constructed, the system fault model is composed of three layers of fault models at the member system level, the region level and the aircraft level, wherein the evidence represents the monitor results of the test points, the fault represents the fault modes induced by the test analysis, and the bidirectional graph represents the information relationship between the test points and the fault modes;

[0009] The member system level judges the "normal / abnormal" result of the measured value according to the sensor threshold range or judges the test "pass / fail" result according to the receiver test result, both of which are regarded as binary evidence, "1" representing the monitor activation and "0" representing the monitor inactivation; the member system level forms a fault condition by using the monitor result in combination with the system fault model, and calculates the probability of the fault hypothesis in the fault fuzzy group contained in the fault condition;

[0010] The region level is divided according to the functional domain, governs multiple member systems, and collects the fault conditions reported by all the member systems; the region level further diagnoses the fault according to the collected fault conditions, realizes fault isolation and cascade elimination;

[0011] The aircraft level integrates all the fault conditions reported by the region level, performs time fusion in a long time dimension, isolates the fault into an evolving fault with increasing severity over time and an intermittent fault with repetitive frequency over time, and suppresses the fault false alarm caused by sensor false alarm and the like.

[0012] Further, the member system level contains monitors, fault modes and detection edges, the region level contains fault modes, symptom cascades, fault cascades and detection edges, and the aircraft level contains fault modes, symptom cascades, fault cascades and detection edges.

[0013] Further, the detection edge is a directed edge from the monitor to the fault mode, the symptom cascade is that a fault mode in one subsystem can trigger a monitor in another subsystem, and the fault cascade is that a fault mode in one subsystem can cause a fault mode in another subsystem.

[0014] Further, the fault condition is composed of a group of fault modes and a group of corresponding monitors, wherein it is assumed that there is only one fault mode in the group of fault modes and the fault mode is occurring, and thus the group of corresponding monitors can be triggered.

[0015] Further, the member system level reports the processed fault condition for further fault diagnosis, regardless of whether the fault isolation is completed.

[0016] Further, according to the fault hypothesis, the member system level traverses the set of fault modes in the fault condition, generates all possible fault hypotheses, and calculates the occurrence probability of all fault hypotheses in combination with the binary evidence of the corresponding set of monitors; if there is a fault hypothesis whose occurrence probability is greater than a first preset threshold, the member system level fault isolation ends, and the fault hypothesis is the fault mode obtained by fault isolation; if there is a fault hypothesis whose occurrence probability is lower than a second preset threshold, the fault mode is deleted from the fault condition.

[0017] Further, the regional level fault isolation is a case where a single fault mode cannot be isolated in the fault condition reported by the member system level, and actively queries the monitor results in the fault condition reported by the member system level, supports on-demand calling of active in-machine test result query.

[0018] Further, the regional level cascade elimination is a case where, after the fault conditions reported by multiple member systems, the fault cascade and evidence cascade described in the regional level fault model are combined, the cascade faults of multiple systems in the same functional domain are eliminated, and the fault mode of the root cause is isolated.

[0019] Further, the aircraft level manages all member system level and regional level fault diagnosis results, performs fault suppression, selectively displays the most important information to the pilot and maintenance personnel, and supports the association of off-board ground detection equipment and fault diagnosis services.

[0020] A distributed airborne fault diagnosis reasoning system, comprising:

[0021] A fault model construction module is configured to construct a system fault model in the form of an "evidence-fault bidirectional graph" according to the test points set in the avionics system testability report and the analyzed fault modes, the system fault model being composed of three layers of member system level, regional level and aircraft level fault models, wherein the evidence represents the monitor results of the test points, the fault represents the fault modes induced by the test analysis, and the bidirectional graph represents the information relationship between the test points and the fault modes;

[0022] A fault condition generation module is configured to judge the "normal / abnormal" result of the measured value according to the sensor threshold range or receive the in-machine test result to judge the "pass / fail" result of the test, both of which are regarded as binary evidence, "1" representing the activation of the monitor and "0" representing the inactivation of the monitor; the member system level utilizes the monitor results to form a fault condition in combination with the system fault model, and calculates the probability of the fault hypothesis in the fault fuzzy set contained in the fault condition;

[0023] A fault condition collection module is used for dividing according to a function domain at a region level, governing a plurality of member systems, and collecting fault conditions reported by all member systems; the region level further diagnoses faults according to the collected fault conditions, and realizes fault isolation and cascade elimination;

[0024] A fault isolation module is used for fault conditions reported by all region levels of the integrated avionics system at the aircraft level, time fusion in a long time dimension, isolation of the fault into an evolving fault with increasing severity over time and an intermittent fault with repetitive frequency over time, and suppression of fault false alarms caused by sensor false alarms.

[0025] Compared with the prior art, the present application has the following technical features:

[0026] 1. The present application adopts a three-layer distributed fault diagnosis reasoning mechanism at the member system level, the region level and the aircraft level, the member system level completes simple fault isolation, the region level completes fault isolation and cascade elimination, and the aircraft level completes fault isolation and time fusion, and the calculation and communication requirements of fault reasoning are considered to balance the reasoning performance and resource occupation.

[0027] 2. The present application uses a system fault model of "evidence-fault bidirectional graph" as basic information, and combines the actual results of the monitor as supplementary information to realize fault diagnosis reasoning. The system fault model and the monitor can be continuously optimized according to the actual operation state of the system and fault case analysis to improve the correctness of the system fault model description and the accuracy of the fault diagnosis reasoning. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 is a schematic diagram of a distributed airborne fault diagnosis reasoning system;

[0029] Figure 2 is a schematic diagram of a member system level fault model;

[0030] Figure 3 is a schematic diagram of a fault condition. DETAILED DESCRIPTION

[0031] The present application proposes a distributed airborne fault diagnosis reasoning method, which can realize fault isolation, cascade elimination, time fusion and other diagnosis functions, and a method of cooperating with fault diagnosis according to a member system level manager, a region level manager and an aircraft level manager.

[0032] Referring to the drawings, the distributed airborne fault diagnosis reasoning method of the present application comprises the following steps:

[0033] Step 1, according to the test points set in the avionics system testability report and the failure modes analyzed, a system failure model in the form of "evidence-failure bidirectional graph" is constructed; here, evidence represents the monitor results of test points, failure represents the failure modes induced by test analysis, and bidirectional graph represents the information relationship between test points and failure modes; the system failure model is composed of three layers of failure models, i.e. member system level, region level and aircraft level.

[0034] Among them, the member system level contains monitors, failure modes and detection edges, the region level includes failure modes, symptom cascades, failure cascades and detection edges, and the aircraft level includes failure modes, symptom cascades, failure cascades and detection edges.

[0035] Among them, the detection edge refers to the directed edge from the monitor to the failure mode, the symptom cascade refers to the failure mode in one subsystem that can trigger the monitor in another subsystem; the failure cascade refers to the failure mode in one subsystem that can cause the failure mode in another subsystem.

[0036] Step 2, the member system level judges the "normal / abnormal" result of the measured value according to the sensor threshold range or the "pass / fail" result of the receiver test, both of which can be regarded as binary evidence, "1" representing monitor activation and "0" representing monitor inactivation; the member system level uses the monitor results to form a failure condition in combination with the system failure model, and calculates the probability of the failure hypothesis in the failure fuzzy group contained in the failure condition. Among them, the failure condition is composed of a group of failure modes and a group of corresponding monitors, and it is assumed that there is only one failure mode in the group of failure modes and it is occurring, so it may trigger the group of corresponding monitors.

[0037] According to the above failure hypothesis, the member system level traverses the group of failure modes in the failure condition to generate all possible failure hypotheses, and calculates the occurrence probability of all failure hypotheses in combination with the binary evidence of the group of corresponding monitors.

[0038] If the occurrence probability of a certain failure hypothesis is too high, far exceeding the probabilities of other failure hypotheses, the member system level fault isolation ends, and the failure hypothesis is the failure mode obtained by fault isolation; if the occurrence probability of a certain failure hypothesis is too low, far lower than the probabilities of other failure hypotheses, the failure mode is deleted from the failure condition. Whether the fault isolation is completed or not, the member system level will report the processed failure condition for further fault diagnosis.

[0039] Step 3, the region level is divided according to the functional domain, governs multiple member systems, and collects the failure conditions reported by all member systems. The region level further diagnoses the fault according to the collected failure conditions, and realizes the functions of fault isolation and cascade elimination.

[0040] Regional level fault isolation is to actively query the monitor results of the member system level reported fault conditions when a single fault mode cannot be isolated in the member system level reported fault conditions. The active in-machine test result query is supported according to the needs.

[0041] Regional level cascade elimination is to eliminate the cascade faults of multiple systems in the same function domain after the fault conditions reported by multiple member systems, in combination with the fault cascades and evidence cascades described in the regional level fault model, and isolate the fault mode to the root cause.

[0042] Step 4, the fault conditions of all function domain systems of the aircraft level integrated avionics system reported at the regional level are time fused in a long time dimension, the fault is isolated into an evolving fault with increasing severity over time and an intermittent fault with a frequency of occurrence repeated over time, and the fault false alarm caused by sensor false alarm is suppressed.

[0043] The aircraft level manages the fault diagnosis results of the member system level and the regional level, executes fault suppression, selectively displays the most important information to the pilot and the maintenance personnel, and supports the correlation of off-aircraft ground detection equipment and fault diagnosis services.

[0044] Embodiment:

[0045] Suppose that the fault diagnosis reasoning system of an avionics system is as shown in the figure, including eight monitors, three member system levels, two regional levels and one aircraft level. The on-board fault diagnosis reasoning machine designed by the application is distributed in the member system level, the regional level and the aircraft level, and generates the current fault diagnosis result according to the evidence of the monitor. The reasoning machine provides the most reasonable explanation for all the evidence shown by all the monitors, creates a new fault condition for new monitor evidence, deletes a fault hypothesis that may be insufficient or has no evidence, and realizes the fault diagnosis results of fault isolation, cascade elimination and time fusion.

[0046] The technical solution of the application specifically implements the following steps:

[0047] According to the testability report, failure mode and effects analysis report and fault case of the avionics system, a "evidence-fault bidirectional graph" system fault model is constructed. Figure 2 A member system level schematic diagram of the system fault model is shown, the left side represents four monitor evidences, and the right side represents three fault modes of the member system 1.

[0048] The member system level fault diagnosis process is as follows:

[0049] a. When one of the four monitors triggers an evidence E1, a fault condition is created as shown in Figure 3 , which has two fault modes and three monitor evidences;

[0050] b. According to the fault mode group in the fault condition, three fault hypotheses are generated, respectively as follows:

[0051] h p = {fm1=1}

[0052] h p = {fm2=1}

[0053] h p = {fm1=1, fm2=1}

[0054] c. Uncertainty reasoning is performed using a probability method, and the problem is expressed by a probability formula, wherein h p is the hypothesis of the state of the fault mode, Tr represents all the evidence equal to 1, and Q represents all the evidence equal to 0. It is equivalent to that the evidence and the fault are random variables, and the knowledge about the problem is expressed as a joint probability, i.e. the joint probability of the evidence and the fault:

[0055]

[0056] d. The joint probability of all fault hypotheses is calculated, and fault isolation and fault splitting operations are performed according to the probability value, δ1 and δ2 are preset values, and are as follows:

[0057] Fault isolation:

[0058]

[0059] Fault splitting:

[0060] P(fm i =1, fm j =1, Tr, Q) > δ2 + P(fm i =1, Tr, Q)

[0061] The above examples are only used to illustrate the technical solutions of the present application, but not to limit the same; although the present application has been described in detail with reference to the foregoing examples, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing examples can be modified, or some technical features thereof can be replaced by equivalent ones; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A distributed onboard fault diagnosis reasoning method, characterized in that, The method comprises the following steps: According to the test points set in the avionics system testability report and the analyzed fault modes, a system fault model in the form of "evidence-fault bidirectional graph" is constructed, the system fault model is composed of three layers of fault models at the member system level, the region level and the aircraft level, wherein the evidence represents the monitor results of the test points, the fault represents the fault modes induced by the test analysis, and the bidirectional graph represents the information relationship between the test points and the fault modes; The member system level judges the "normal / abnormal" result of the measured value according to the sensor threshold range or judges the test "pass / fail" result according to the receiver internal test result, both of which are regarded as binary evidence, "1" representing the activation of the monitor and "0" representing the inactivation of the monitor; the member system level forms a fault condition by using the monitor result in combination with the system fault model and calculates the probability of the fault hypothesis in the fault fuzzy group contained in the fault condition; The region level is divided according to the functional domain, governs a plurality of member systems and collects the fault conditions reported by all the member systems; the region level further diagnoses the fault according to the collected fault conditions, realizes fault isolation and cascade elimination; The aircraft level integrates all the fault conditions reported by the region level and performs time fusion in the time dimension, isolates the fault into an evolving fault with increasing severity over time and an intermittent fault with repetitive frequency over time, and suppresses the fault false alarm caused by sensor false reporting. 2.The distributed onboard fault diagnosis reasoning method according to claim 1, characterized in that, The member system level contains monitors, fault modes and detection edges, the region level contains fault modes, symptom cascades, fault cascades and detection edges, and the aircraft level contains fault modes, symptom cascades, fault cascades and detection edges. 3.The distributed onboard fault diagnosis reasoning method according to claim 2, characterized in that, The detection edge is a directed edge from the monitor to the fault mode, the symptom cascade is that a fault mode in one subsystem can trigger a monitor in another subsystem, and the fault cascade is that a fault mode in one subsystem can cause a fault mode in another subsystem. 4.The distributed onboard fault diagnosis reasoning method according to claim 1, characterized in that, The fault condition is composed of a group of fault modes and a group of corresponding monitors, wherein it is assumed that there is only one fault mode occurring in the group of fault modes, which can trigger the corresponding monitor.

5. The distributed onboard fault diagnosis reasoning method according to claim 1, characterized in that, Whether the fault isolation is completed or not, the member system level will report the processed fault condition for further fault diagnosis.

6. The distributed onboard fault diagnosis reasoning method according to claim 4, characterized in that, According to the fault hypothesis, the member system level traverses the fault modes in the fault condition, generates all possible fault hypotheses and calculates the occurrence probability of all the fault hypotheses in combination with the binary evidence of the corresponding monitor; if the occurrence probability of a certain fault hypothesis is greater than a first preset threshold, the member system level ends the fault isolation, and the fault hypothesis is the fault mode obtained by the fault isolation; if the occurrence probability of a certain fault hypothesis is lower than a second preset threshold, the fault mode is deleted from the fault condition.

7. The distributed onboard fault diagnosis reasoning method according to claim 1, characterized in that, The region level fault isolation is to actively query the monitor results in the fault condition reported by the member system level in the case that a single fault mode cannot be isolated in the fault condition reported by the member system level, thereby supporting the on-demand calling of the active internal test result query.

8. The distributed onboard fault diagnosis reasoning method according to claim 1, characterized in that, The regional cascade elimination is to eliminate the cascade failure of multiple systems in the same function domain after the failure conditions reported by multiple member systems, combined with the failure cascade and evidence cascade described in the regional failure model.

9. The distributed onboard fault diagnosis reasoning method according to claim 1, characterized in that, The aircraft level manages the failure diagnosis results of all member system level and regional level, executes failure suppression, selectively displays the most important information to the pilot and maintenance personnel, and supports the correlation of off-aircraft ground detection equipment and failure diagnosis service.

10. A distributed onboard fault diagnosis reasoning system, characterized by, The aircraft level manages the failure diagnosis results of all member system level and regional level, executes failure suppression, selectively displays the most important information to the pilot and maintenance personnel, and supports the correlation of off-aircraft ground detection equipment and failure diagnosis service. The failure model construction module is used to construct a system failure model in the form of "evidence-failure bidirectional graph" according to the test points and analyzed failure modes set in the avionics system testability report, and the system failure model is composed of three layers of member system level, regional level and aircraft level, wherein the evidence represents the monitor results of the test points, the failure represents the failure modes induced by the test analysis, and the bidirectional graph represents the information relationship between the test points and the failure modes; The failure condition generation module is used to judge the "normal / abnormal" result of the measured value according to the sensor threshold range or the "pass / fail" result of the receiver test result at the member system level, both of which are regarded as binary evidence, "1" representing the monitor activation and "0" representing the monitor inactivation; the member system level utilizes the monitor results to form a failure condition in combination with the system failure model, and calculates the probability of the failure hypothesis in the failure fuzzy set contained in the failure condition; The failure condition collection module is used to divide multiple member systems according to the function domain at the regional level, and collect the failure conditions reported by all member systems; the regional level further diagnoses the failure according to the collected failure conditions, realizes failure isolation and cascade elimination; The failure isolation module is used to integrate the failure conditions reported by all regional levels of the comprehensive avionics system at the aircraft level, perform time fusion in a long time dimension, isolate the failure into an evolving failure with increasing severity over time and an intermittent failure with repetitive frequency over time, and suppress the false alarm of the failure caused by sensor false reporting.

Images