Unmanned aerial vehicle fault diagnosis method
By building and training a fault occurrence judgment model and combining drone sensor data and log data, accurate, efficient and real-time diagnosis of drone faults is achieved, solving the problems of low accuracy and lack of real-time performance in existing technologies.
Patent Information
- Application Number
- CN202510802805.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-16
- Publication Date
- 2025-09-23
AI Technical Summary
Existing drone fault diagnosis methods have low accuracy, low efficiency, and are not real-time, making it difficult to diagnose various types of faults.
By obtaining drone sensor data, pre-processing it and then inputting it into a pre-trained fault occurrence judgment model, fault judgment is performed in combination with the long short-term memory neural network LSTM, and the server is used to analyze the real-time uploaded log data to analyze the cause of the fault.
It achieves accurate, efficient and real-time diagnosis of UAV faults, improves the accuracy and real-time performance of fault diagnosis, and contributes to the safe flight of UAVs.
Smart Images
Figure CN120686781A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a UAV, and in particular to a UAV fault diagnosis method. Background Art
[0002] Drone systems are complex, primarily composed of subsystems such as flight control systems, power systems, sensor systems, and navigation systems. These systems are susceptible to complex environmental factors during operation, and various failures are inevitable. These failures can cause unstable flight, prevent the drone from following its intended route, or even lead to a crash, resulting in significant economic losses and potentially endangering personnel safety. For example, sensor failures (such as those in gyroscopes, accelerometers, and barometers) can result in inaccurate flight status information, compromising flight control.
[0003] At present, the traditional UAV fault diagnosis method mainly relies on data collection and expert experience to diagnose UAV faults based on the collected data. This method has many shortcomings:
[0004] 1) Low accuracy: Expert experience has certain limitations and cannot cover all possible fault conditions. In addition, experts may make misjudgments of some complex fault modes, resulting in inaccurate fault diagnosis results.
[0005] 2) Low efficiency: Relying on expert experience for fault diagnosis consumes a lot of time and manpower, especially when faced with massive amounts of drone operation data, resulting in low diagnostic efficiency.
[0006] 3) Lack of real-time diagnosis: Expert diagnosis usually needs to be performed after the flight, and it is impossible to detect and diagnose faults in time during the flight, increasing the risk of drone accidents.
[0007] 4) Difficulty in diagnosing multiple types of faults: Traditional methods based on expert experience can often only diagnose specific types of faults. When multiple types of faults occur at the same time, it is difficult to accurately judge and distinguish them. Summary of the Invention
[0008] (1) Technical problems solved
[0009] In view of the above shortcomings of the prior art, the present invention provides a UAV fault diagnosis method, which can effectively overcome the defects of the prior art such as low fault diagnosis accuracy and lack of real-time performance.
[0010] (2) Technical solution
[0011] To achieve the above objectives, the present invention is implemented through the following technical solutions:
[0012] A method for diagnosing a drone fault comprises the following steps:
[0013] S1. Obtain the drone's sensor data including timestamps and determine whether the sensor data exceeds the corresponding state threshold range;
[0014] S2. When the sensor data exceeds the corresponding state threshold range, determine whether the duration of the sensor data exceeding the state threshold range is greater than a preset time threshold;
[0015] S3. When the duration of the sensor data exceeding the state threshold range is greater than a preset time threshold, obtaining flight characterization data of the UAV within the duration and preprocessing the flight characterization data;
[0016] S4. Input the pre-processed flight characterization data into the pre-trained fault occurrence judgment model to obtain the fault occurrence probability, and judge whether the UAV has a fault based on the fault occurrence probability;
[0017] S5. When a UAV fails, an early warning message is sent to the UAV management terminal. After receiving the early warning message, the UAV management terminal sends a fault diagnosis request to the server;
[0018] S6. After receiving the fault diagnosis request, the server filters the target log data within the duration from the log data uploaded by the drone in real time to analyze the cause of the fault, and feeds back the fault diagnosis results to the drone management terminal.
[0019] Preferably, the sensor data includes measurement data, temperature data and noise of sensors, and the sensors include gyroscopes, accelerometers, magnetometers, global positioning system sensors, barometers, vision sensors, temperature sensors and humidity sensors.
[0020] Preferably, S3 obtains flight characterization data of the UAV within the duration and preprocesses the flight characterization data, including:
[0021] The flight characterization data of the UAV within the duration is obtained, and the Kalman filter algorithm is used to process the flight characterization data to filter out noise. The PCA algorithm is used to reduce the dimension of the flight characterization data after noise filtering, and the range transformation method is used to normalize the flight characterization data after dimensionality reduction.
[0022] Preferably, the flight characterization data includes measurement data of a gyroscope, an accelerometer, a magnetometer, a global positioning system sensor, a barometer, a vision sensor, a temperature sensor, and a humidity sensor.
[0023] Preferably, before inputting the pre-processed flight characterization data into the pre-trained fault occurrence judgment model to obtain the fault occurrence probability in S4, the following steps are included:
[0024] S41, dividing the historical data set into a training set, a validation set, and a test set according to a preset ratio;
[0025] S42. Setting a loss function and an optimizer for a fault occurrence judgment model;
[0026] S43, inputting the training set into the fault occurrence judgment model for model training;
[0027] S44, calculating a loss value based on the loss function, and the optimizer updating the model parameters according to the loss value and the network gradient information;
[0028] S45. If the loss value is less than the preset threshold, the model training ends and the current fault occurrence judgment model is the trained fault occurrence judgment model. Otherwise, the process returns to S43 and continues to use the training set for model training.
[0029] S46. Input the validation set into the trained fault occurrence judgment model, evaluate the generalization ability of the model by observing its performance on the validation set, and tune the model's hyperparameters and structure.
[0030] S47. Input the test set into the optimized fault occurrence judgment model to evaluate the model performance.
[0031] Among them, the fault occurrence judgment model is built based on the long short-term memory neural network LSTM.
[0032] Preferably, before dividing the historical data set into a training set, a validation set, and a test set according to a preset ratio in S41, the following steps are included:
[0033] Collect historical flight characterization data of the UAV within a historical period and pre-process the historical flight characterization data;
[0034] The preprocessed historical flight characterization data are assigned corresponding fault labels to construct a historical dataset.
[0035] Preferably, collecting historical flight characterization data of the UAV within a historical period and preprocessing the historical flight characterization data includes:
[0036] Collect the flight characterization data of the UAV during each continuous flight time at different locations during the historical period, and remove the flight characterization data of the UAV during the take-off and landing phases to obtain the historical flight characterization data of the UAV during the historical period;
[0037] The Kalman filter algorithm is used to process the historical flight characterization data to filter out noise, the PCA algorithm is used to reduce the dimension of the historical flight characterization data after noise filtering, and the range transformation method is used to normalize the historical flight characterization data after dimensionality reduction.
[0038] Preferably, after receiving the fault diagnosis request in S6, the server filters the target log data within the duration from the log data uploaded by the drone in real time to analyze the cause of the fault, and feeds back the fault diagnosis result to the drone management terminal, including:
[0039] S61. After receiving the fault diagnosis request, the server determines whether the drone management terminal has the authority to query log data;
[0040] S62: When the drone management terminal has the authority to query log data, the server selects target log data within the duration from the log data uploaded in real time by the corresponding drone according to the identity identifier in the fault diagnosis request to perform fault cause analysis;
[0041] S63. The server feeds back the analyzed fault cause and target log data within the duration to the drone management terminal;
[0042] Among them, if the drone management terminal does not have the authority to query log data, the server will prompt the drone management terminal that the authority is insufficient.
[0043] Preferably, the log data includes the drone's flight attitude information, motor status information, battery status information, working status information of each electronic component, and flight log information. The flight log information includes the operation log of the drone control system, the operation instructions issued to each electronic component, and the response results of each electronic component to the operation instructions.
[0044] (3) Beneficial effects
[0045] Compared with the existing technology, the drone fault diagnosis method provided by the present invention has the following beneficial effects:
[0046] 1) Obtaining sensor data containing timestamps from the drone, determining whether the sensor data exceeds a corresponding state threshold range, and when the sensor data exceeds the corresponding state threshold range, determining whether the duration for which the sensor data exceeds the state threshold range is greater than a preset time threshold. When the duration for which the sensor data exceeds the state threshold range is greater than the preset time threshold, obtaining flight characterization data of the drone during the duration, preprocessing the flight characterization data, and inputting the preprocessed flight characterization data into a pre-trained fault occurrence judgment model to obtain a fault occurrence probability, and determining whether the drone has a fault based on the fault occurrence probability. By constructing and training the fault occurrence judgment model and combining the drone's sensor data and flight characterization data, it is possible to accurately determine whether the drone has a fault.
[0047] 2) When a UAV malfunctions, an early warning message is sent to the UAV management terminal. After receiving the early warning message, the UAV management terminal sends a fault diagnosis request to the server. After receiving the fault diagnosis request, the server determines whether the UAV management terminal has the authority to query the log data. When the UAV management terminal has the authority to query the log data, the server filters the target log data within the duration from the log data uploaded by the corresponding UAV in real time according to the identity in the fault diagnosis request to analyze the cause of the fault. The server feeds back the analyzed fault cause and the target log data within the duration to the UAV management terminal, so that accurate and efficient fault cause analysis can be performed through the log data uploaded by the UAV in real time, ensuring the accuracy of the fault diagnosis results. At the same time, the real-time nature of the fault diagnosis is improved, which helps the UAV to better perform flight missions. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.
[0049] Figure 1 It is a schematic diagram of the process of the present invention;
[0050] Figure 2 The figure is a flow chart of the process of training and optimizing the fault occurrence judgment model in the present invention. DETAILED DESCRIPTION
[0051] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.
[0052] A method for diagnosing UAV faults, such as Figure 1 As shown, S1, obtains the sensor data including the timestamp of the drone, and determines whether the sensor data exceeds the corresponding state threshold range.
[0053] In the technical solution of the present application, the sensor data includes the measurement data, temperature data and noise of the sensor, and the sensor includes a gyroscope, an accelerometer, a magnetometer, a global positioning system sensor, a barometer, a visual sensor, a temperature sensor and a humidity sensor.
[0054] S2. When the sensor data exceeds the corresponding state threshold range, determine whether the duration for which the sensor data exceeds the state threshold range is greater than a preset time threshold.
[0055] S3. When the duration of the sensor data exceeding the state threshold range is greater than a preset time threshold, flight characterization data of the UAV within the duration is obtained and the flight characterization data is preprocessed.
[0056] Specifically, the flight characterization data of the UAV within the duration is obtained and the flight characterization data is preprocessed, including:
[0057] The flight characterization data of the UAV within the duration is obtained, and the Kalman filter algorithm is used to process the flight characterization data to filter out noise. The PCA algorithm is used to reduce the dimension of the flight characterization data after noise filtering, and the range transformation method is used to normalize the flight characterization data after dimensionality reduction.
[0058] In the technical solution of the present application, the flight characterization data includes measurement data of a gyroscope, an accelerometer, a magnetometer, a global positioning system sensor, a barometer, a visual sensor, a temperature sensor, and a humidity sensor.
[0059] S4. Input the pre-processed flight characterization data into the pre-trained fault occurrence judgment model to obtain the fault occurrence probability, and judge whether the UAV has a fault based on the fault occurrence probability.
[0060] In S4, the pre-processed flight characterization data is input into the pre-trained fault occurrence judgment model to obtain the fault occurrence probability, such as Figure 2 As shown, including:
[0061] S41, dividing the historical data set into a training set, a validation set, and a test set according to a preset ratio;
[0062] S42. Setting a loss function and an optimizer for a fault occurrence judgment model;
[0063] S43, inputting the training set into the fault occurrence judgment model for model training;
[0064] S44, calculating a loss value based on the loss function, and the optimizer updating the model parameters according to the loss value and the network gradient information;
[0065] S45. If the loss value is less than the preset threshold, the model training ends and the current fault occurrence judgment model is the trained fault occurrence judgment model. Otherwise, the process returns to S43 and continues to use the training set for model training.
[0066] S46. Input the validation set into the trained fault occurrence judgment model, evaluate the generalization ability of the model by observing its performance on the validation set, and tune the model's hyperparameters and structure.
[0067] S47. Input the test set into the optimized fault occurrence judgment model to evaluate the model performance.
[0068] Among them, the fault occurrence judgment model is built based on the long short-term memory neural network LSTM.
[0069] Specifically, before dividing the historical data set into training set, validation set and test set according to the preset ratio, Figure 2 Shown, including:
[0070] Collect historical flight characterization data of the UAV within a historical period and pre-process the historical flight characterization data;
[0071] The preprocessed historical flight characterization data are assigned corresponding fault labels to construct a historical dataset.
[0072] Specifically, the historical flight characterization data of the UAV in the historical period is collected and preprocessed, such as Figure 2 Shown, including:
[0073] Collect the flight characterization data of the UAV during each continuous flight time at different locations during the historical period, and remove the flight characterization data of the UAV during the take-off and landing phases to obtain the historical flight characterization data of the UAV during the historical period;
[0074] The Kalman filter algorithm is used to process the historical flight characterization data to filter out noise, the PCA algorithm is used to reduce the dimension of the historical flight characterization data after noise filtering, and the range transformation method is used to normalize the historical flight characterization data after dimensionality reduction.
[0075] The above technical solution obtains sensor data containing timestamps from the drone, determines whether the sensor data exceeds the corresponding state threshold range, and when the sensor data exceeds the corresponding state threshold range, determines whether the duration of the sensor data exceeding the state threshold range is greater than a preset time threshold. When the duration of the sensor data exceeding the state threshold range is greater than the preset time threshold, the flight characterization data of the drone within the duration is obtained, and the flight characterization data is preprocessed. The preprocessed flight characterization data is input into a pre-trained fault occurrence judgment model to obtain the fault occurrence probability, and judge whether the drone has a fault based on the fault occurrence probability. By constructing and training the fault occurrence judgment model, combined with the drone's sensor data and flight characterization data, it is possible to accurately judge whether the drone has a fault.
[0076] like Figure 1As shown, S5, when a UAV fails, an early warning message is sent to the UAV management terminal. After receiving the early warning message, the UAV management terminal sends a fault diagnosis request to the server.
[0077] S6. After receiving the fault diagnosis request, the server selects the target log data within the duration from the real-time log data uploaded by the drone to analyze the cause of the fault and feeds back the fault diagnosis results to the drone management terminal, specifically including:
[0078] S61. After receiving the fault diagnosis request, the server determines whether the drone management terminal has the authority to query log data;
[0079] S62: When the drone management terminal has the authority to query log data, the server selects target log data within the duration from the log data uploaded in real time by the corresponding drone according to the identity identifier in the fault diagnosis request to perform fault cause analysis;
[0080] S63. The server feeds back the analyzed fault cause and target log data within the duration to the drone management terminal;
[0081] Among them, if the drone management terminal does not have the authority to query log data, the server will prompt the drone management terminal that it has insufficient authority.
[0082] In the technical solution of the present application, the log data includes the UAV's flight attitude information, motor status information, battery status information, working status information of each electronic device and flight log information. The flight log information includes the operation log of the UAV control system, the operation instructions issued to each electronic device and the response results of each electronic device to the operation instructions.
[0083] In the above technical solution, when a drone malfunctions, an early warning message is sent to the drone management terminal. After receiving the early warning message, the drone management terminal sends a fault diagnosis request to the server. After receiving the fault diagnosis request, the server determines whether the drone management terminal has the authority to query log data. When the drone management terminal has the authority to query log data, the server filters the target log data within the duration from the log data uploaded in real time by the corresponding drone based on the identity identifier in the fault diagnosis request to analyze the cause of the fault. The server feeds back the analyzed fault cause and the target log data within the duration to the drone management terminal, so that accurate and efficient fault cause analysis can be performed through the log data uploaded in real time by the drone, ensuring the accuracy of the fault diagnosis results, while improving the real-time nature of fault diagnosis, and helping the drone to better perform flight missions.
[0084] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.
Claims
1. A method for diagnosing a drone fault, characterized by: The following steps are involved: S1. Obtain the drone's sensor data including timestamps and determine whether the sensor data exceeds the corresponding state threshold range; S2. When the sensor data exceeds the corresponding state threshold range, determine whether the duration of the sensor data exceeding the state threshold range is greater than a preset time threshold; S3. When the duration of the sensor data exceeding the state threshold range is greater than a preset time threshold, obtaining flight characterization data of the UAV within the duration and preprocessing the flight characterization data; S4. Input the pre-processed flight characterization data into the pre-trained fault occurrence judgment model to obtain the fault occurrence probability, and judge whether the UAV has a fault based on the fault occurrence probability; S5. When a UAV fails, an early warning message is sent to the UAV management terminal. After receiving the early warning message, the UAV management terminal sends a fault diagnosis request to the server; S6. After receiving the fault diagnosis request, the server filters the target log data within the duration from the log data uploaded by the drone in real time to analyze the cause of the fault, and feeds back the fault diagnosis results to the drone management terminal.
2. The UAV fault diagnosis method according to claim 1, characterized in that: The sensor data includes measurement data, temperature data, and noise of sensors, including a gyroscope, an accelerometer, a magnetometer, a global positioning system sensor, a barometer, a vision sensor, a temperature sensor, and a humidity sensor.
3. The UAV fault diagnosis method according to claim 1, characterized in that: Obtain the flight characterization data of the drone within the duration in S3 and preprocess the flight characterization data, including: The flight characterization data of the UAV within the duration is obtained, and the Kalman filter algorithm is used to process the flight characterization data to filter out noise. The PCA algorithm is used to reduce the dimension of the flight characterization data after noise filtering, and the range transformation method is used to normalize the flight characterization data after dimensionality reduction.
4. The UAV fault diagnosis method according to claim 3, characterized in that: The flight characterization data includes measurement data from a gyroscope, an accelerometer, a magnetometer, a global positioning system sensor, a barometer, a vision sensor, a temperature sensor, and a humidity sensor.
5. The UAV fault diagnosis method according to claim 1, characterized in that: In S4, the pre-processed flight characterization data is input into the pre-trained fault occurrence judgment model to obtain the fault occurrence probability, including: S41, dividing the historical data set into a training set, a validation set, and a test set according to a preset ratio; S42. Setting a loss function and an optimizer for a fault occurrence judgment model; S43, inputting the training set into the fault occurrence judgment model for model training; S44, calculating a loss value based on the loss function, and the optimizer updating the model parameters according to the loss value and the network gradient information; S45. If the loss value is less than the preset threshold, the model training ends and the current fault occurrence judgment model is the trained fault occurrence judgment model. Otherwise, the process returns to S43 and continues to use the training set for model training. S46. Input the validation set into the trained fault occurrence judgment model, evaluate the generalization ability of the model by observing its performance on the validation set, and tune the model's hyperparameters and structure. S47. Input the test set into the optimized fault occurrence judgment model to evaluate the model performance. Among them, the fault occurrence judgment model is built based on the long short-term memory neural network LSTM.
6. The UAV fault diagnosis method according to claim 5, characterized in that: Before S41 divides the historical data set into training set, validation set, and test set according to the preset ratio, it includes: Collect historical flight characterization data of the UAV within a historical period and pre-process the historical flight characterization data; The preprocessed historical flight characterization data are assigned corresponding fault labels to construct a historical dataset.
7. The UAV fault diagnosis method according to claim 6, characterized in that: The collecting of historical flight characterization data of the UAV within a historical period and preprocessing of the historical flight characterization data include: Collect the flight characterization data of the UAV during each continuous flight time at different locations during the historical period, and remove the flight characterization data of the UAV during the take-off and landing phases to obtain the historical flight characterization data of the UAV during the historical period; The Kalman filter algorithm is used to process the historical flight characterization data to filter out noise, the PCA algorithm is used to reduce the dimension of the historical flight characterization data after noise filtering, and the range transformation method is used to normalize the historical flight characterization data after dimensionality reduction.
8. The UAV fault diagnosis method according to claim 1, characterized in that: After receiving the fault diagnosis request in S6, the server selects the target log data within the duration from the real-time log data uploaded by the drone to analyze the cause of the fault and feeds back the fault diagnosis results to the drone management terminal, including: S61. After receiving the fault diagnosis request, the server determines whether the drone management terminal has the authority to query log data; S62: When the drone management terminal has the authority to query log data, the server selects target log data within the duration from the log data uploaded in real time by the corresponding drone according to the identity identifier in the fault diagnosis request to perform fault cause analysis; S63. The server feeds back the analyzed fault cause and target log data within the duration to the drone management terminal; Among them, if the drone management terminal does not have the authority to query log data, the server will prompt the drone management terminal that the authority is insufficient.
9. The UAV fault diagnosis method according to claim 8, characterized in that: The log data includes the drone's flight attitude information, motor status information, battery status information, working status information of each electronic component, and flight log information. The flight log information includes the operation log of the drone control system, the operation instructions issued to each electronic component, and the response results of each electronic component to the operation instructions.