A high-precision UAV state judgment method and system based on fault search tree

Abstract

The present invention discloses a high-precision UAV state judgment method and system based on a fault search tree, including: calculating a fault search tree activation value by the posture of the UAV IMU and the posture of the GPS, judging whether the fault search tree activation value is within the fault diagnosis range; if the fault search tree activation value is within the fault diagnosis range, monitoring the engine module and building a health state model; if the fault search tree activation value is not within the fault diagnosis range, building a first-level fault search tree model to determine the first-level fault module; building a second-level fault search tree model to determine the final fault module; when building the first-level fault search tree model and the second-level fault search tree model, updating the fault diagnosis range. The present invention determines whether the UAV posture is abnormal by calculating the fault search tree activation value, and jointly completes the UAV state judgment by the health state model, the first-level fault search tree model and the second-level fault search tree model, which can take into account accuracy while ensuring real-time performance.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a high-precision UAV state judgment method and system based on a fault search tree. Background Art

[0002] With the development of science and technology, the tasks and scenarios faced by social production and life are becoming more and more abundant. As an emerging type of convenient mobile robot, drones have played an increasingly important role in surveying and mapping, firefighting and rescue, agricultural disinfection and other fields with their advantages of high sensitivity and strong detection capabilities.

[0003] In the process of drones performing tasks, it is inevitable to encounter various emergencies. In order to deal with the impact of various emergencies on drones, it is necessary to grasp the status of the drone in time. However, in the process of drone flight, on the one hand, there are many parameters involved. If all parameters are directly calculated, it is very time-consuming and difficult to achieve the effect of real-time status judgment. On the other hand, in the process of drone flight, airflow can easily interfere with the drone state. It is difficult to identify the impact of airflow. Therefore, it is necessary to propose a new high-precision drone state judgment method to realize the judgment of drone state. Summary of the invention

[0004] The present invention provides a high-precision UAV state judgment method and system based on a fault search tree to solve the current technical problems of low real-time performance and inaccurate state judgment of UAV states.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] On the one hand, the present invention provides a high-precision UAV state judgment method based on a fault search tree, and the high-precision UAV state judgment method based on a fault search tree includes:

[0007] The fault search tree activation value is calculated based on the position and posture of the drone's IMU and GPS. If the fault search tree activation value is within the fault diagnosis range, the engine module is monitored. If the fault search tree activation value is not within the fault diagnosis range, the power module, engine module and radio communication module are monitored.

[0008] If the fault search tree activation value is within the fault diagnosis range, the engine module is monitored, a health status model is constructed, and the health status of the drone is determined by the consistency of the average posture change rate of the drone and the engine speed;

[0009] If the fault search tree activation value is not within the fault diagnosis range, the power module, engine module and radio communication module are monitored, a first-level fault search tree model is constructed, and a first-level fault module is determined;

[0010] Construct a secondary fault search tree model, use the secondary fault search tree model to verify the primary fault module and determine the final fault module;

[0011] When constructing the first-level fault search tree model and the second-level fault search tree model, the position and posture of the drone's IMU and GPS are uploaded in real time to update the fault diagnosis range.

[0012] Further, the fault search tree activation value is calculated by the posture of the drone IMU and the posture of the GPS, and if the fault search tree activation value is within the fault diagnosis range, the engine module is monitored; if the fault search tree activation value is not within the fault diagnosis range, the power module, the engine module and the radio communication module are monitored, including:

[0013] The IMU is located at the center of the drone. The pose of the IMU is the pose of the drone. The pose P of the IMU is obtained. imu , the specific expression is as follows:

[0014] P imu ={x imu ,y imu , z imu}

[0015] Among them, x imu Indicates the coordinate value of IMU on the x-axis, y imu Indicates the coordinate value of IMU on the y-axis, z imu Indicates the coordinate value of IMU on the z-axis;

[0016] The position P of the UAV GPS is obtained GPS , the specific expression is as follows:

[0017] P GPS ={x GPS ,y GPS , Z GPS}

[0018] Among them, x GPS Indicates the x-axis coordinate value of the drone obtained by GPS, y GPS Indicates the y-axis coordinate value of the drone obtained by GPS, z GPS Indicates the coordinate value of the drone on the z-axis obtained by GPS;

[0019] Calculate the fault search tree activation value J at the current time j j , the specific calculation formula is shown as follows:

[0020]

[0021] in, Indicates the coordinate value of the IMU on the x-axis at the current time j, Indicates the coordinate value of the IMU on the y-axis at the current time j, Indicates the coordinate value of the IMU on the z-axis at the current time j, represents the x-axis coordinate value of the drone obtained by GPS at the current time j, represents the y-axis coordinate value of the drone obtained by GPS at the current time j, Indicates the z-axis coordinate value of the drone obtained by GPS at the current time j;

[0022] Determine whether the fault search tree activation value is within the fault diagnosis range;

[0023] Initialization fault diagnosis range Φ=[α 1 , α 2 ], where α 1 represents the lower limit of the fault diagnosis range, α 2 Indicates the upper limit of the fault diagnosis range;

[0024] Determine whether the fault search tree activation value at the current time j is within the fault diagnosis range. The judgment formula is as follows:

[0025]

[0026] Among them, FLAG represents the judgment flag bit. If the fault search tree activation value at the current time j is within the fault diagnosis range, the judgment flag bit is set to 1, and the engine module needs to be monitored;

[0027] If the fault search tree activation value at the current moment j is not within the fault diagnosis range, the judgment flag is set to 0, and the power module, engine module and radio communication module need to be monitored;

[0028] The setting of the fault search tree activation value and the fault diagnosis range can avoid the full monitoring of all data when judging the status of the drone, which will cause a waste of computing resources. In addition, real-time monitoring of the posture can also improve the command center's control over the drone, promptly discover the degree of deviation of the drone's posture, intervene in the drone, avoid excessive deviation of the drone's posture trajectory during the mission, and improve the reliability of mission completion.

[0029] Furthermore, if the fault search tree activation value is within the fault diagnosis range, the engine module is monitored, a health status model is constructed, and the health status of the drone is judged by the consistency of the average posture change rate of the drone and the engine speed, including:

[0030] If the fault search tree activation value is within the fault diagnosis range, monitor the engine module;

[0031] Assume the engine speed is n, the initialization time sliding window is R, and R is a positive integer;

[0032] Calculate the average pose of the drone Where i represents a positive integer, Indicates the coordinate value of the IMU on the x-axis at the current time i, Indicates the coordinate value of the IMU on the y-axis at the current time i, Indicates the coordinate value of the IMU on the z-axis at the current time i, It represents the x-axis coordinate value of the drone obtained by GPS at the current time i. It represents the y-axis coordinate value of the drone obtained by GPS at the current time i. It represents the z-axis coordinate value of the drone obtained by GPS at the current time i. Represents the average posture at the current time i. For example, R can be 10, and the average posture of the drone 10 seconds before the current time is calculated;

[0033] Calculate the average posture change rate of the drone in

[0034] Get engine speed n = {n 1 ,...,n i ,...,n R-1}, where the engine speed only takes the values ​​of the first R-1 speeds in the time sliding window R;

[0035] Initialize the consistency threshold β, where β is the ratio of the rotation speed to the change rate of the posture in the absence of wind;

[0036] Calculate the consistency parameter L. The specific expression is as follows:

[0037]

[0038] Where L = {L 1 , ..., L i , ..., L R-1};

[0039] Sort L from large to small and get the maximum value L i , if |L i -β| / β is greater than α 3 , indicating that the average posture change rate of the drone is inconsistent with the engine speed of the drone, indicating that the drone encounters interference from other external environments such as airflow, and is in a disturbed state, where α 3 is a positive real number;

[0040] If if |L i -β| / β is less than or equal to α 3 , indicating that the average posture change rate of the UAV is consistent with the engine speed of the UAV, indicating that the UAV is operating normally and is in a healthy state;

[0041] During the normal flight of the UAV, the changes in the engine speed and the UAV posture are proportional. Therefore, under normal circumstances, the ratio will not change significantly. However, when the UAV encounters airflow, the ratio will increase. Therefore, by setting a regular threshold, the change in the ratio can be used to determine whether the UAV encounters external interference such as airflow. This solution is easy to calculate during implementation and can accurately determine the situation of encountering airflow.

[0042] Furthermore, if the fault search tree activation value is not within the fault diagnosis range, the power module, the engine module and the radio communication module are monitored, a primary fault search tree model is constructed, and a primary fault module is determined, including:

[0043] If the fault search tree activation value is not within the fault diagnosis range, monitor the power module, engine module and radio communication module;

[0044] The power supply monitoring module determines whether a power supply failure occurs by monitoring the change rate ΔU of the voltage U;

[0045] The engine monitoring module determines whether the engine fails by monitoring the engine speed n;

[0046] The radio communication monitoring module determines whether a radio communication failure occurs by monitoring the carrier-to-noise ratio S;

[0047] Calculate the rate of change of voltage U in the C+1 moments before the current moment, ΔU, where ΔU = {ΔU 1 , ..., ΔU i , ..., ΔU C}, ΔU i =|U i+1 -U i |, C is a positive integer;

[0048] In the C+1 moments before the current moment, obtain the engine speed n in the previous C moments = {n 1 , ..., n i , ..., n C};

[0049] In the C+1 moments before the current moment, obtain the carrier-to-noise ratio S in the previous C moments = {S 1 , ..., S i , ..., S C};

[0050] Construct a first-level fault search tree model. The specific steps are as follows:

[0051] (1) Construct a matrix X with C rows and 3 columns based on the voltage change rate ΔU, engine speed n, and carrier-to-noise ratio S;

[0052] (2) Normalize the matrix X, calculate the mean E(X) and variance e(X) of the matrix X, and calculate the normalized matrix The specific calculation formula is as follows:

[0053]

[0054] (3) Calculation matrix The covariance matrix V of in Represents the matrix The transpose of

[0055] (4) Calculate the eigenvalues ​​and eigenvectors of the covariance matrix V;

[0056] (5) Initialize the contribution threshold ∈ and calculate the cumulative contribution of the eigenvalues ​​of the covariance matrix V. The specific expression is as follows:

[0057]

[0058] If the cumulative contribution of the eigenvalues ​​of the covariance matrix V is greater than or equal to the contribution threshold ∈, the cumulative contribution of the eigenvalues ​​is considered qualified, and the number of eigenvalues ​​that meet the requirements q is obtained;

[0059] (6) Arrange the eigenvectors in rows according to the size of the eigenvalues ​​and construct the dimension reduction matrix D;

[0060] When the UAV performs a mission, there are many parameters to monitor. This solution only takes some parameters as examples. If all the data are calculated and monitored, the calculation complexity is very high. Therefore, it is necessary to reduce the dimension of the data in order to improve the real-time calculation.

[0061] The statistical method is used to calculate the detection abnormal value and then determine the first-level fault module;

[0062] Calculate the statistic H 2 , the specific expression is as follows:

[0063]

[0064] Where Λ represents the diagonal matrix consisting of q eigenvalues;

[0065] Calculate the statistical threshold The specific expression is as follows:

[0066]

[0067] Among them, F α (q,Cq) represents the F distribution with the test level α and the degrees of freedom q and Cq;

[0068] If H 2 Greater than or equal to This indicates that the drone's power module, engine module, or radio module is faulty, or other modules may be faulty and need to be reported to the command center for further monitoring;

[0069] In order to further determine the location of the faulty module, it is necessary to calculate the contribution of the voltage change rate ΔU, engine speed n and carrier-to-noise ratio S. The module with the largest contribution is the primary faulty module.

[0070] In actual situations, there may be multiple faults, but in this solution, only the impact of the faulty module with the greatest impact on the drone is considered. For multiple faults, further analysis and processing can be carried out after eliminating or solving the most serious fault.

[0071] Furthermore, the constructing of the secondary fault search tree model, using the secondary fault search tree model to verify the primary fault module, and determining the final fault module, includes:

[0072] Construct a secondary fault search tree model to further verify the primary fault module;

[0073] Initialize the variance threshold σ ΔU , σ n , σ S ;

[0074] Calculate the variance e(ΔU) of the rate of change ΔU of the voltage U in the C+1 moments before the current moment;

[0075] Calculate the variance e(n) of the engine speed n in the C+1 moments before the current moment;

[0076] Calculate the variance e(S) of the carrier-to-noise ratio S in the C+1 moments before the current moment;

[0077] If the primary fault module is the power module, compare e(ΔU) and σ ΔU If e(ΔU) is less than or equal to σ ΔU , indicating that the result of the first-level fault search tree is correct, and the final faulty module is determined to be the power module. Otherwise, it needs to be recalculated and reported to the command center as an error;

[0078] If the primary fault module is the engine module, compare e(n) and σ nIf e(n) is less than or equal to σ n , indicating that the result of the first-level fault search tree is correct, and the final faulty module is determined to be the engine module. Otherwise, it needs to be recalculated and reported to the command center as an error;

[0079] If the primary fault module is the radio communication module, compare e(S) and σ S If e(S) is less than or equal to σ S , indicating that the result of the first-level fault search tree is correct, and the final faulty module is determined to be the radio communication module. Otherwise, it needs to be recalculated and reported to the command center as an error.

[0080] Furthermore, when constructing the primary fault search tree model and the secondary fault search tree model, the position and posture of the drone IMU and the GPS are uploaded in real time to update the fault diagnosis scope, including:

[0081] When constructing the first-level fault search tree model and the second-level fault search tree model, the position and posture of the drone IMU and the GPS are uploaded in real time. Every 30 minutes, the error average value I of the position and posture of the drone IMU and the GPS is recalculated. The specific calculation formula is shown in the following formula:

[0082]

[0083] Among them, m represents the total number of postures within 30 minutes, x imu (i) represents the coordinate value of the i-th x-axis of the IMU, y imu (i) represents the coordinate value of the i-th y-axis of the IMU, z imu (i) represents the coordinate value of the i-th z-axis of the IMU, x GPS (i) represents the x-axis coordinate value of the i-th drone obtained by GPS, and y GPS (i) represents the coordinate value of the y-axis of the i-th drone obtained by GPS, z GPS (i) represents the coordinate value of the z-axis of the i-th drone obtained by GPS;

[0084] Update the fault diagnosis range. The specific calculation formula is as follows:

[0085]

[0086] The fault activation value is re-determined using the updated fault diagnosis range.

[0087] On the other hand, the present invention also provides a high-precision UAV state judgment system based on a fault search tree, and the high-precision UAV state judgment system based on a fault search tree includes the following modules:

[0088] Activation module: Calculate the fault search tree activation value based on the position and posture of the drone IMU and GPS. If the fault search tree activation value is within the fault diagnosis range, monitor the engine module. If the fault search tree activation value is not within the fault diagnosis range, monitor the power module, engine module and radio communication module.

[0089] Health status judgment module: If the fault search tree activation value is within the fault diagnosis range, the engine module is monitored, a health status model is constructed, and the health status of the drone is judged by the consistency of the average posture change rate of the drone and the engine speed;

[0090] Fault search tree model building module: If the fault search tree activation value is not within the fault diagnosis range, monitor the power module, engine module and radio communication module, build a first-level fault search tree model, determine the first-level fault module, build a second-level fault search tree model, and use the second-level fault search tree model to verify the first-level fault module to determine the final fault module;

[0091] Fault diagnosis range update module: When constructing the first-level fault search tree model and the second-level fault search tree model, the position and posture of the drone IMU and GPS are uploaded in real time to update the fault diagnosis range.

[0092] The beneficial effects brought about by the technical solution provided by the present invention include at least:

[0093] 1. The present invention determines whether to monitor the power module, engine module and radio communication module of the UAV in real time by calculating the fault search tree activation value. The fault search tree activation value is calculated only by the posture of the UAV IMU and the posture of the GPS. Not only is the calculation convenient and fast, but the flight trajectory of the UAV can be observed in time. Compared with the full-process monitoring of all data, the present invention can avoid the consumption of computing power for monitoring the power module, engine module and radio communication module due to the large deviation of the UAV posture, further reduce the hardware cost of the UAV, and improve the lightweight level of the UAV.

[0094] 2. Due to the complex state of the drone when performing tasks, this solution sets the drone's state to healthy state, interfered state and fault state. When making state judgments, the drone's IMU posture and GPS posture are used, and information such as the voltage change rate of the power module, the engine speed of the engine module and the carrier-to-noise ratio of the radio communication module are used. Compared with the method of judging the drone state by only using a certain indicator, the present invention uses more indicators and more comprehensive monitoring content. It adopts a combination of rapid judgment and verification of the fault module, which further improves accuracy and efficiency while ensuring real-time performance.

[0095] 3. Update the fault diagnosis range in time to avoid fault diagnosis errors caused by inaccurate initial value of the fault diagnosis range, and at the same time improve the accuracy of the fault diagnosis range to further reduce the waste of computing power. In addition, in view of the influence of the external environment, the influence of airflow is monitored to avoid misdiagnosis of faults caused by abnormal changes in engine speed, thereby improving the comprehensiveness of state judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0097] Figure 1 The present invention is a flow chart of a high-precision UAV state judgment method based on a fault search tree. DETAILED DESCRIPTION

[0098] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0099] Example 1

[0100] This embodiment will describe in more detail a high-precision UAV state judgment method based on a fault search tree of the present invention in conjunction with the corresponding drawings. Figure 1 As shown, the method is mainly divided into five steps. First, the fault search tree activation value is calculated by the posture of the UAV IMU and the posture of the GPS. If the fault search tree activation value is within the fault diagnosis range, the engine module is monitored. If the fault search tree activation value is not within the fault diagnosis range, the power module, the engine module and the radio communication module are monitored; second, if the fault search tree activation value is within the fault diagnosis range, the engine module is monitored, and a health status model is constructed. The health status of the UAV is judged by the consistency of the average posture change rate of the UAV and the engine speed; third, if the fault search tree activation value is not within the fault diagnosis range, the power module, the engine module and the radio communication module are monitored, and a first-level fault search tree model is constructed to determine the first-level fault module; fourth, a second-level fault search tree model is constructed, and the first-level fault module is verified by the second-level fault search tree model to determine the final fault module; fifth, when constructing the first-level fault search tree model and the second-level fault search tree model, the posture of the UAV IMU and the posture of the GPS are uploaded in real time to update the fault diagnosis range.

[0101] Specifically, the method of this embodiment includes the following steps:

[0102] Step S1: The fault search tree activation value is calculated by the posture of the drone IMU and the posture of the GPS. If the fault search tree activation value is within the fault diagnosis range, the engine module is monitored. If the fault search tree activation value is not within the fault diagnosis range, the power module, the engine module and the radio communication module are monitored, including:

[0103] The IMU is located at the center of the drone. The pose of the IMU is the pose of the drone. The pose P of the IMU is obtained. imu , the specific expression is as follows:

[0104] P imu ={x imu ,y imu ,z imu}

[0105] Among them, x imu Indicates the coordinate value of IMU on the x-axis, y imu Indicates the coordinate value of IMU on the y-axis, z imu Indicates the coordinate value of IMU on the z-axis;

[0106] The position P of the UAV GPS is obtained GPS , the specific expression is as follows:

[0107] P GPS ={x GPS ,y GPS ,z GPS}

[0108] Among them, x GPS Indicates the x-axis coordinate value of the drone obtained by GPS, y GPS Indicates the y-axis coordinate value of the drone obtained by GPS, z GPS Indicates the coordinate value of the drone on the z-axis obtained by GPS;

[0109] Calculate the fault search tree activation value J at the current time j j , the specific calculation formula is shown as follows:

[0110]

[0111] in, Indicates the coordinate value of the IMU on the x-axis at the current time j, Indicates the coordinate value of the IMU on the y-axis at the current time j, Indicates the coordinate value of the IMU on the z-axis at the current time j, represents the x-axis coordinate value of the drone obtained by GPS at the current time j, represents the y-axis coordinate value of the drone obtained by GPS at the current time j, Indicates the z-axis coordinate value of the drone obtained by GPS at the current time j;

[0112] Determine whether the fault search tree activation value is within the fault diagnosis range;

[0113] Initialization fault diagnosis range Φ=[α 1 ,α 2 ], where α 1 represents the lower limit of the fault diagnosis range, α 2 Indicates the upper limit of the fault diagnosis range;

[0114] Determine whether the fault search tree activation value at the current time j is within the fault diagnosis range. The judgment formula is as follows:

[0115]

[0116] Among them, FLAG represents the judgment flag bit. If the fault search tree activation value at the current time j is within the fault diagnosis range, the judgment flag bit is set to 1, and the engine module needs to be monitored;

[0117] If the fault search tree activation value at the current moment j is not within the fault diagnosis range, the judgment flag is set to 0, and the power module, engine module and radio communication module need to be monitored;

[0118] When performing tasks, drones need to monitor the tasks to be performed and their own parameters in real time. However, if all parameters are monitored before status judgment, it will cause a waste of computing power, thereby increasing the hardware cost of the drone. Therefore, this solution starts from the posture of the drone. Once the posture change of the drone exceeds expectations, that is, the calculated fault search tree activation value is not within the fault diagnosis range, indicating that the state of the drone is not within the ideal range, and then monitor its own important modules and parameters. It should be noted that only some important parameters are monitored in this solution. In the actual operation of the drone, many parameters are involved, so the computing power saved by using this solution will be more obvious.

[0119] Step S2: If the fault search tree activation value is within the fault diagnosis range, monitor the engine module, build a health status model, and judge the health status of the drone by the consistency of the average posture change rate of the drone and the engine speed, including:

[0120] If the fault search tree activation value is within the fault diagnosis range, monitor the engine module;

[0121] Assume the engine speed is n, the initialization time sliding window is R, and R is a positive integer;

[0122] Calculate the average pose of the drone Where i represents a positive integer, Indicates the coordinate value of the IMU on the x-axis at the current time i, Indicates the coordinate value of the IMU on the y-axis at the current time i, Indicates the coordinate value of the IMU on the z-axis at the current time i, It represents the x-axis coordinate value of the drone obtained by GPS at the current time i. It represents the y-axis coordinate value of the drone obtained by GPS at the current time i. It represents the z-axis coordinate value of the drone obtained by GPS at the current time i. represents the average posture at the current time i;

[0123] Calculate the average posture change rate of the drone in

[0124] Get engine speed n = {n 1 ,...,n i ,...,n R-1}, where the engine speed only takes the values ​​of the first R-1 speeds in the time sliding window R;

[0125] Initialize the consistency threshold β, where β is the ratio of the rotation speed to the change rate of the posture in the absence of wind;

[0126] Calculate the consistency parameter L. The specific expression is as follows:

[0127]

[0128] Where L = {L 1 ,...,L i ,...,L R-1};

[0129] Sort L from large to small and get the maximum value L i , if |L i -β| / β is greater than α 3 , indicating that the average posture change rate of the drone is inconsistent with the engine speed of the drone, indicating that the drone encounters interference from other external environments such as airflow, and is in a disturbed state, where α 3 is the consistency verification value, which is a positive real number;

[0130] In actual situations, α 3and β are determined according to the function and size of the drone. For example, for a small drone, in the case of uniform flight, the rotation speed is 100 rpm and the speed of the drone is 10 m / s, that is, the rate of change of the posture is 10 m / s. Therefore, the consistency threshold β is normally 10. When the drone encounters airflow, if the rotation speed is 100 rpm and the speed of the drone is 14 m / s, the consistency parameter is 7.14. If the speed of the drone is 8 m / s, the consistency parameter is 12.5. Therefore, α can be 3 Set it to 0.15 to distinguish the drone from encountering air currents;

[0131] When the drone encounters airflow, there will be differences in acceleration, deceleration or uniform motion. This solution is only for the uniform motion stage. Of course, the appropriate α 3 and β can also be applied to the flow detection of UAVs during the whole mission execution;

[0132] If if |L i -β| / β is less than or equal to α 3 , indicating that the average posture change rate of the UAV is consistent with the engine speed of the UAV, indicating that the UAV is operating normally and is in a healthy state.

[0133] Step S3: If the fault search tree activation value is not within the fault diagnosis range, monitor the power module, the engine module and the radio communication module, build a first-level fault search tree model, and determine the first-level fault module, including:

[0134] If the fault search tree activation value is not within the fault diagnosis range, monitor the power module, engine module and radio communication module;

[0135] The power supply monitoring module determines whether a power supply failure occurs by monitoring the change rate ΔU of the voltage U;

[0136] The engine monitoring module determines whether the engine fails by monitoring the engine speed n;

[0137] The radio communication monitoring module determines whether a radio communication failure occurs by monitoring the carrier-to-noise ratio S;

[0138] Calculate the rate of change of voltage U in the C+1 moments before the current moment, ΔU, where ΔU = {ΔU 1 ,...,ΔU i ,...,ΔU C}, ΔU i =|U i+1 -U i |, C is a positive integer;

[0139] In the C+1 moments before the current moment, obtain the engine speed n in the previous C moments = {n 1,...,n i ,...,n C};

[0140] In the C+1 moments before the current moment, obtain the carrier-to-noise ratio S in the previous C moments = {S 1 ,...,S i ,...,S C};

[0141] Construct a first-level fault search tree model. The specific steps are as follows:

[0142] (1) Construct a matrix X with C rows and 3 columns based on the voltage change rate ΔU, engine speed n, and carrier-to-noise ratio S;

[0143] (2) Normalize the matrix X, calculate the mean E(X) and variance e(X) of the matrix X, and calculate the normalized matrix The specific calculation formula is as follows:

[0144]

[0145] (3) Calculation matrix The covariance matrix V of in Represents the matrix The transpose of

[0146] (4) Calculate the eigenvalues ​​and eigenvectors of the covariance matrix V;

[0147] (5) Initialize the contribution threshold ∈ and calculate the cumulative contribution χ(q) of the eigenvalues ​​of the covariance matrix V. The specific expression is as follows:

[0148]

[0149] Among them, λ i represents the i-th eigenvalue. If the cumulative contribution of the eigenvalues ​​of the covariance matrix V is greater than or equal to the contribution threshold ∈, the cumulative contribution of the eigenvalue is considered qualified, and the number of eigenvalues ​​that meet the requirements is obtained;

[0150] (6) Arrange the eigenvectors in rows according to the size of the eigenvalues ​​and construct the dimension reduction matrix D;

[0151] The dimensionality reduction depends on the size of the eigenvalue, that is, the severity of the fault. For faults with greater severity, the eigenvalue is larger, and the number of eigenvalues ​​that meet the requirements is smaller, so the matrix dimension will be smaller, and the real-time performance will be higher. For faults with less severity, traditional numerical analysis is difficult to find the fault point. This solution can still make a correct diagnosis through the statistical threshold and find the module with greater impact through a fault tree search model. Compared with searching all data one by one, this solution also has better real-time performance, so it also has accuracy.

[0152] The statistical method is used to calculate the detection abnormal value and then determine the first-level fault module;

[0153] Calculate the statistic H 2 , the specific expression is as follows:

[0154]

[0155] Wherein, Λ represents a diagonal matrix composed of q eigenvalues. It should be noted that the size of the eigenvalues ​​in the diagonal matrix increases from top to bottom;

[0156] Calculate the statistical threshold The specific expression is as follows:

[0157]

[0158] Among them, F α (q,Cq) represents the F distribution with the test level α and the degrees of freedom q and Cq;

[0159] If H 2 Greater than or equal to This indicates that the drone's power module, engine module, or radio module is faulty, or other modules may be faulty and need to be reported to the command center for further monitoring;

[0160] In order to further determine the location of the faulty module, it is necessary to calculate the contribution of the change rate ΔU, engine speed n and carrier-to-noise ratio S. The contribution of the i-th variable The calculation formula is as follows:

[0161]

[0162] Where w represents an integer, represents the i-th variable, represents the i-th statistic, and the module with the largest contribution is the first-level fault module;

[0163] The contribution diagram method is mainly used to find the specific fault point. The contribution rate of each variable to the fault is compared to further determine the specific location of the fault point. The variable with the largest contribution rate is the variable monitored by the first-level fault module.

[0164] Step S4: constructing a secondary fault search tree model, verifying the primary fault module by the secondary fault search tree model, and determining the final fault module, including:

[0165] Construct a secondary fault search tree model to further verify the primary fault module;

[0166] Initialize the variance threshold σ ΔU , σ n , σ S ;

[0167] Calculate the variance e(ΔU) of the rate of change ΔU of the voltage U in the C+1 moments before the current moment;

[0168] Calculate the variance e(n) of the engine speed n in the C+1 moments before the current moment;

[0169] Calculate the variance e(S) of the carrier-to-noise ratio S in the C+1 moments before the current moment;

[0170] If the primary fault module is the power module, compare e(ΔU) and σ ΔU If e(ΔU) is less than or equal to σ ΔU , indicating that the result of the first-level fault search tree is correct, and the final faulty module is determined to be the power module, indicating that the drone is in a faulty state. Otherwise, it needs to be recalculated and reported to the command center;

[0171] If the primary fault module is the engine module, compare e(n) and σ n If e(n) is less than or equal to σ n , indicating that the result of the first-level fault search tree is correct, and the final fault module is determined to be the engine module, indicating that the drone is in a fault state, otherwise it needs to be recalculated and reported to the command center;

[0172] If the primary fault module is the radio communication module, compare e(S) and σ S If e(S) is less than or equal to σ S , indicating that the result of the first-level fault search tree is correct, and the final fault module is determined to be the radio communication module, indicating that the drone is in a fault state, otherwise it needs to be recalculated and reported to the command center;

[0173] If the variance is calculated for all data, the amount of calculation will be huge. Therefore, the first-level fault module is obtained through the first-level fault tree search tree model, and then the fault module is verified. This can not only make up for the lack of data loss after the dimensionality reduction processing of the first-level fault search tree model, but also avoid misjudgment through verification, so that the command center can grasp the status of the drone in time.

[0174] Step S5: When constructing the primary fault search tree model and the secondary fault search tree model, the position and posture of the drone IMU and the GPS are uploaded in real time to update the fault diagnosis range, including:

[0175] When constructing the first-level fault search tree model and the second-level fault search tree model, the position and posture of the drone IMU and the GPS are uploaded in real time. Every 30 minutes, the error average value I of the position and posture of the drone IMU and the GPS is recalculated. The specific calculation formula is shown in the following formula:

[0176]

[0177] Among them, m represents the total number of postures within 30 minutes, x imu (i) represents the coordinate value of the i-th x-axis of the IMU, y imu (i) represents the coordinate value of the i-th y-axis of the IMU, z imu (i) represents the coordinate value of the i-th z-axis of the IMU, x GPS (i) represents the x-axis coordinate value of the i-th drone obtained by GPS, and y GPS (i) represents the coordinate value of the y-axis of the i-th drone obtained by GPS, z GPS (i) represents the coordinate value of the z-axis of the i-th drone obtained by GPS;

[0178] Update the fault diagnosis range. The specific calculation formula is as follows:

[0179]

[0180] Re-determine the fault activation value using the updated fault diagnosis range;

[0181] Since the operation of drones in different time periods is different, compared with the fixed range solution, the fault activation value will deviate when judging. Therefore, this solution regularly updates the fault diagnosis range to further improve the accuracy of drone status judgment.

[0182] Example 2

[0183] This embodiment provides a high-precision drone status judgment system based on a fault search tree:

[0184] Activation module: Calculate the fault search tree activation value based on the position and posture of the drone IMU and GPS. If the fault search tree activation value is within the fault diagnosis range, monitor the engine module. If the fault search tree activation value is not within the fault diagnosis range, monitor the power module, engine module and radio communication module.

[0185] Health status judgment module: If the fault search tree activation value is within the fault diagnosis range, the engine module is monitored, a health status model is constructed, and the health status of the drone is judged by the consistency of the average posture change rate of the drone and the engine speed;

[0186] Fault search tree model building module: If the fault search tree activation value is not within the fault diagnosis range, monitor the power module, engine module and radio communication module, build a first-level fault search tree model, determine the first-level fault module, build a second-level fault search tree model, and use the second-level fault search tree model to verify the first-level fault module to determine the final fault module;

[0187] Fault diagnosis range update module: When constructing the first-level fault search tree model and the second-level fault search tree model, the position and posture of the drone IMU and GPS are uploaded in real time to update the fault diagnosis range.

[0188] The high-precision UAV state judgment system based on fault search tree of the present embodiment corresponds to the high-precision UAV state judgment method based on fault search tree of the above embodiment; wherein, the functions implemented by each functional module in the high-precision UAV state judgment system based on fault search tree of the present embodiment correspond one by one to each process step in the high-precision UAV state judgment method based on fault search tree of the above embodiment; therefore, they will not be repeated here.

[0189] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0190] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0191] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A high-precision UAV state judgment method based on fault search tree, characterized in that: The following steps are involved: S1: Calculate the fault search tree activation value based on the position and posture of the drone IMU and the GPS. If the fault search tree activation value is within the fault diagnosis range, monitor the engine module. If the fault search tree activation value is not within the fault diagnosis range, monitor the power module, engine module and radio communication module. S2: If the fault search tree activation value is within the fault diagnosis range, the engine module is monitored, a health status model is constructed, and the health status of the UAV is determined by the consistency of the average posture change rate of the UAV and the engine speed; S3: If the fault search tree activation value is not within the fault diagnosis range, the power module, the engine module and the radio communication module are monitored, and a first-level fault search tree model is constructed to determine the first-level fault module; S4: construct a secondary fault search tree model, and use the secondary fault search tree model to verify the primary fault module to determine the final fault module; S5: When constructing the first-level fault search tree model and the second-level fault search tree model, the position and posture of the drone IMU and the GPS are uploaded in real time to update the fault diagnosis range.

2. The high-precision UAV state judgment method based on fault search tree according to claim 1 is characterized in that: The step S1 comprises: The IMU is located at the center of the drone. The pose of the IMU is the pose of the drone. The pose P of the IMU is obtained. imu , the specific expression is as follows: P imu ={x imu ,y imu ,z imu } Among them, x imu Indicates the coordinate value of IMU on the x-axis, y imu Indicates the coordinate value of IMU on the y-axis, z imu Indicates the coordinate value of IMU on the z-axis; Get the GPS position P of the drone GPS , the specific expression is as follows: P GPS ={x GPS ,y GPS ,z GPS } Among them, x GPS Indicates the x-axis coordinate value of the drone obtained by GPS, y GPS Indicates the y-axis coordinate value of the drone obtained by GPS, z GPS Indicates the z-axis coordinate value of the drone obtained by GPS; Calculate the fault search tree activation value J at the current time j j , the specific calculation formula is shown as follows: in, Indicates the coordinate value of the IMU on the x-axis at the current time j, Indicates the coordinate value of the IMU on the y-axis at the current time j, Indicates the coordinate value of the IMU on the z-axis at the current time j, represents the x-axis coordinate value of the drone obtained by GPS at the current time j, represents the y-axis coordinate value of the drone obtained by GPS at the current time j, Indicates the z-axis coordinate value of the drone obtained by GPS at the current time j; Determine whether the fault search tree activation value is within the fault diagnosis range.

3. According to the high-precision UAV state judgment method based on fault search tree in claim 2, the method of judging whether the fault search tree activation value is within the fault diagnosis range comprises the following steps: Initialize the fault diagnosis range Φ = [α1, α2], where α1 represents the lower limit of the fault diagnosis range, and α2 represents the upper limit of the fault diagnosis range; Determine whether the fault search tree activation value at the current time j is within the fault diagnosis range. The judgment formula is as follows: Among them, FLAG represents the judgment flag bit. If the fault search tree activation value at the current time j is within the fault diagnosis range, the judgment flag bit is set to 1, and the engine module needs to be monitored; If the fault search tree activation value at the current moment j is not within the fault diagnosis range, the judgment flag is set to 0, and the power module, engine module and radio communication module need to be monitored.

4. The high-precision UAV state judgment method based on fault search tree according to claim 1 is characterized in that: The step S2 comprises: If the fault search tree activation value is within the fault diagnosis range, the engine module is monitored; Assume the engine speed is n, the initialization time sliding window is R, and R is a positive integer; Calculate the average pose of the drone Where i represents a positive integer, Indicates the coordinate value of the IMU on the x-axis at the current time i, Indicates the coordinate value of the IMU on the y-axis at the current time i, Indicates the coordinate value of the IMU on the z-axis at the current time i, It represents the x-axis coordinate value of the drone obtained by GPS at the current time i. It represents the y-axis coordinate value of the drone obtained by GPS at the current time i. It represents the coordinate value of the UAV on the z-axis obtained by GPS at the current time i. represents the average posture at the current time i; Calculate the average posture change rate of the drone in Get engine speed n = {n 1 ,...,n i ,...,n R-1 }, where the engine speed only takes the values ​​of the first R-1 speeds in the time sliding window R; Initialize the consistency threshold β, where β is the ratio of the rotation speed to the change rate of the posture in the absence of wind; Calculate the consistency parameter L. The specific expression is as follows: Where L = {L 1 ,...,L i ,...,L R-1 }; Sort L from large to small and get the maximum value L i , if |L i -β| / β is greater than α3, indicating that the average posture change rate of the UAV is inconsistent with the engine speed of the UAV, indicating that the UAV encounters interference from other external environments such as airflow, and is in a disturbed state, where α3 is a positive real number; If if |L i -β| / β is less than or equal to α3, indicating that the average posture change rate of the UAV is consistent with the engine speed of the UAV, indicating that the UAV is operating normally and is in a healthy state.

5. The high-precision UAV state judgment method based on fault search tree according to claim 1 is characterized in that: The step S3 comprises: If the fault search tree activation value is not within the fault diagnosis range, monitor the power module, engine module and radio communication module; The power supply monitoring module determines whether a power supply failure occurs by monitoring the change rate ΔU of the voltage U; The engine monitoring module determines whether the engine fails by monitoring the engine speed n; The radio communication monitoring module determines whether a radio communication failure occurs by monitoring the carrier-to-noise ratio S; Calculate the rate of change of voltage U in the C+1 moments before the current moment, ΔU, where ΔU = {ΔU 1 ,...,ΔU i ,...,ΔU C }, ΔU i =|U i+1 -U i |, C is a positive integer; In the C+1 moments before the current moment, obtain the engine speed n in the previous C moments = {n 1 ,...,n i ,...,n C }; In the C+1 moments before the current moment, obtain the carrier-to-noise ratio S in the previous C moments = {S 1 ,...,S i ,...,S C }; Construct a first-level fault search tree model. The specific steps are as follows: (1) Construct a matrix X with C rows and 3 columns based on the voltage change rate ΔU, engine speed n, and carrier-to-noise ratio S; (2) Normalize the matrix X, calculate the mean E(X) and variance e(X) of the matrix X, and calculate the normalized matrix The specific calculation formula is as follows: (3) Calculation matrix The covariance matrix V of in Represents the matrix The transpose of (4) Calculate the eigenvalues ​​and eigenvectors of the covariance matrix V; (5) Initialize the contribution threshold ∈, calculate the cumulative contribution of the eigenvalues ​​of the covariance matrix V, if the cumulative contribution of the eigenvalues ​​of the covariance matrix V is greater than or equal to the contribution threshold ∈, then the cumulative contribution of the eigenvalue is considered qualified, and the number of eigenvalues ​​q is obtained; (6) Arrange the eigenvectors in rows according to the size of the eigenvalues ​​and construct the dimension reduction matrix D; The statistical method is used to calculate the detection anomaly value and then determine the first-level fault module.

6. The high-precision UAV state judgment method based on fault search tree according to claim 5 is characterized in that: The statistical method is used to calculate the detection abnormal value and determine the first-level fault module, which includes the following steps: Calculate the statistic H 2 , the specific expression is as follows: Where Λ represents the diagonal matrix consisting of q eigenvalues; Calculate the statistical threshold The specific expression is as follows: Among them, F α (q,Cq) represents the Snyder Kerr-Pearson distribution with a test level of α and degrees of freedom of q and Cq; If H 2 Greater than or equal to This indicates that the drone's power module, engine module, or radio module is faulty; otherwise, other modules are faulty and need to be reported to the command center for further monitoring; In order to further determine the location of the faulty module, it is necessary to calculate the contribution of the voltage change rate ΔU, engine speed n and carrier-to-noise ratio S, among which the module with the largest contribution is the primary faulty module.

7. The high-precision UAV state judgment method based on fault search tree according to claim 6 is characterized in that: The step S4 comprises: Construct a secondary fault search tree model to further verify the primary fault module; Initialize the variance threshold σ ΔU , σ n , σ S ; Calculate the variance e(ΔU) of the rate of change ΔU of the voltage U in the C+1 moments before the current moment; Calculate the variance e(n) of the engine speed n in the C+1 moments before the current moment; Calculate the variance e(S) of the carrier-to-noise ratio S in the C+1 moments before the current moment; If the primary fault module is the power module, compare e(ΔU) and σ ΔU If e(ΔU) is less than or equal to σ ΔU , indicating that the result of the first-level fault search tree is correct, and the final faulty module is determined to be the power module. Otherwise, it needs to be recalculated and reported to the command center as an error; If the primary fault module is the engine module, compare e(n) and σ n If e(n) is less than or equal to σ n , indicating that the result of the first-level fault search tree is correct, and the final faulty module is determined to be the engine module. Otherwise, it needs to be recalculated and reported to the command center as an error; If the primary fault module is the radio communication module, compare e(S) and σ S If e(S) is less than or equal to σ S , indicating that the result of the first-level fault search tree is correct, and the final faulty module is determined to be the radio communication module. Otherwise, it needs to be recalculated and reported to the command center as an error.

8. The high-precision UAV state judgment method based on fault search tree according to claim 1 is characterized in that: The step S5 comprises: When constructing the first-level fault search tree model and the second-level fault search tree model, the position and posture of the drone IMU and the GPS are uploaded in real time. Every 30 minutes, the error average value I of the position and posture of the drone IMU and the GPS is recalculated. The specific calculation formula is shown in the following formula: Among them, m represents the total number of postures within 30 minutes, x imu (i) represents the coordinate value of the i-th x-axis of the IMU, y imu (i) represents the coordinate value of the i-th y-axis of the IMU, z imu (i) represents the coordinate value of the i-th z-axis of the IMU, x GPS (i) represents the x-axis coordinate value of the i-th drone obtained by GPS, and y GPS (i) represents the coordinate value of the y-axis of the i-th drone obtained by GPS, z GPS (i) represents the coordinate value of the z-axis of the i-th drone obtained by GPS; Update the fault diagnosis range. The specific calculation formula is as follows: The fault activation value is re-determined using the updated fault diagnosis range.

9. A high-precision UAV state judgment system based on a fault search tree, characterized in that: Includes the following modules: Activation module: The fault search tree activation value is calculated based on the position and posture of the drone IMU and the GPS. If the fault search tree activation value is within the fault diagnosis range, the engine module is monitored; If the fault search tree activation value is not within the fault diagnosis range, monitor the power module, engine module and radio communication module; Health status judgment module: If the fault search tree activation value is within the fault diagnosis range, the engine module is monitored, a health status model is constructed, and the health status of the drone is judged by the consistency of the average posture change rate of the drone and the engine speed; Fault search tree model building module: If the fault search tree activation value is not within the fault diagnosis range, the power module, engine module and radio communication module are monitored, and a first-level fault search tree model is built to determine the first-level fault module, and a second-level fault search tree model is built. The second-level fault search tree model verifies the first-level fault module and determines the final fault module; Fault diagnosis range update module: when constructing the first-level fault search tree model and the second-level fault search tree model, the position and posture of the drone's IMU and GPS are uploaded in real time to update the fault diagnosis range; To realize a high-precision UAV state judgment method based on a fault search tree as described in any one of claims 1-8.

Images