A fixed-wing unmanned aerial vehicle level flight anomaly detection method

Abstract

The application discloses a fixed-wing unmanned aerial vehicle (UAV) abnormality detection method, and belongs to the technical field of UAV fault diagnosis. The method comprises the following steps: analyzing historical flight parameter data, extracting characteristic parameters in the flat flight stage, combining a flight mechanics model and a statistical method, establishing a normal threshold range of a residual error of a theoretical normal acceleration and a measured normal acceleration, and thus realizing real-time monitoring and abnormality detection on the flat flight state of the UAV.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) fault diagnosis technology, specifically relating to a method for detecting abnormalities in the level flight of a fixed-wing UAV. Background Technology

[0002] With the development of the global drone industry, drones are widely used in military and civilian fields, and flight safety issues are becoming increasingly prominent. Among them, level flight is an important flight state of drones in the mission area, accounting for a large proportion of the entire flight process. Therefore, conducting anomaly detection during the level flight phase of drones has important engineering application value.

[0003] Traditional anomaly detection methods for drones mainly rely on experience-based judgment or simple threshold comparisons, resulting in low detection efficiency, high false alarm rates, and difficulty adapting to complex flight environments. While some intelligent anomaly detection methods based on flight parameters exist, these are purely data-driven and lack robustness. Therefore, there is an urgent need to find a robust, efficient, and low-false-alarm-rate anomaly detection method capable of accurately identifying abnormal behavior during the level flight phase of drones.

[0004] Existing patent CN117708741A, "A Method for Anomaly Detection in Multidimensional Flight Data of UAVs for Parameter Selection," discloses a method for anomaly detection in multidimensional flight data of UAVs for parameter selection. This patent uses a PS-AE-LSTM model for anomaly detection, combined with the Laida criterion. However, this patent suffers from low accuracy in detecting anomalies during level flight. Summary of the Invention

[0005] The purpose of this invention is to solve the problems of the prior art and provide a method for detecting anomalies in the level flight of a fixed-wing UAV. By analyzing historical flight parameter data, characteristic parameters of the level flight phase are extracted. Combined with flight mechanics models and statistical methods, a normal threshold range of the residual between theoretical normal acceleration and measured normal acceleration is established, thereby realizing real-time monitoring and anomaly detection of the level flight state of the UAV.

[0006] This invention is achieved through the following technical solution: A method for detecting anomalies in the level flight of a fixed-wing UAV is proposed. By analyzing historical flight parameter data, characteristic parameters of the level flight phase are extracted. Combined with flight mechanics models and statistical methods, a normal threshold range for the residual between theoretical and measured normal acceleration is established, thereby realizing real-time monitoring and anomaly detection of the UAV's level flight status.

[0007] Preferably, it includes the following steps: Step 1: Data Acquisition: Collect historical flight parameter data from multiple drone sorties; Step 2: Level Flight Data Extraction: Extract level flight data based on the variation characteristics of flight parameters; Step 3: Data Preprocessing: Preprocess the extracted flight parameter data; Step 4: Calculation of normal acceleration: Based on the flight mechanics model, calculate the theoretical normal acceleration of the UAV using the roll angle; Step 5: Residual Calculation: Compare the theoretical normal acceleration with the measured normal acceleration and calculate the residual; Step 6: Threshold Establishment: Based on the Laida criterion, perform statistical analysis on the residual data to establish the residual threshold range under normal flight conditions; Step 7, Anomaly Detection: Compare the residual of the flight parameter to be tested with the threshold range. If it exceeds the threshold range, it is determined to be an anomaly.

[0008] Preferably, in step one, the historical flight parameter data includes flight phase status words, flight time, roll angle, and normal acceleration.

[0009] Preferably, in step two, the level flight data is extracted: first, the data of each sortie is divided into flight stages according to the flight stage status word, and the data of the level flight segment is extracted as the level flight data.

[0010] Preferably, in step three, the preprocessing of the extracted flight parameter data includes time-stamp alignment and noise removal.

[0011] Preferably, the time-stamp alignment is as follows: First, the data is time-stamped, then the same time interval is used as the data sampling interval, and all parameters use the most recent data as the data at the current time. The time-stamped data is then concatenated, and the data at different times after concatenation can be represented by formula (1): (1); in: Represents the roll angle array. For parameters The elements in the table represent the roll angle data at different times; Represents the array of normal accelerations. for The elements in the table represent the normal acceleration data at different times.

[0012] Preferably, the noise removal involves using a moving average filtering method to denoise all data, selecting an appropriate sliding window length based on the data characteristics. The moving average filtering method is shown in formula (2). (2).

[0013] Preferably, in step four, the normal acceleration calculation involves calculating the theoretical normal acceleration of the UAV using the roll angle based on the flight mechanics model. The specific calculation formula is formula (3): (3); in: g The acceleration due to gravity is typically taken as 9.8 m / s². 2 .

[0014] Preferably, in step five, the residual calculation involves comparing the theoretical normal acceleration with the measured normal acceleration to calculate the residual. The calculation method is as shown in formula (4): (4).

[0015] Preferably, in step six, the threshold is established as follows: based on the Laida criterion, statistical analysis is performed on the residual data, and the mean and standard deviation of the residuals are calculated according to formula (5) to establish the residual threshold range under normal flight conditions. The range is set to be within the normal range of the residual between the theoretical normal acceleration and the measured normal acceleration. (5).

[0016] Preferably, in step seven, the residuals of the flight parameters to be tested are compared with the threshold range. To avoid false alarms, if the residuals of multiple consecutive test data exceed the threshold range, it is determined to be abnormal and an alarm signal is issued.

[0017] Compared with the prior art, the present invention has the following advantages and beneficial effects: I. The present invention provides a method for detecting anomalies in level flight of a fixed-wing UAV, (1) accurately extracting flight data in the level flight phase by analyzing flight data characteristics; (2) calculating theoretical normal acceleration by combining flight mechanics model, thereby improving the accuracy and reliability of anomaly detection; (3) effectively distinguishing between normal flight state and abnormal flight state by establishing a residual threshold range based on the Laida criterion; (4) determining anomalies by setting multiple consecutive residuals of test data exceeding the threshold range, thereby effectively avoiding false alarms; (5) this method is applicable to a variety of fixed-wing UAVs and has strong versatility.

[0018] II. The present invention provides a method for detecting level flight anomalies of fixed-wing UAVs, (1) combining flight mechanics models with statistical methods for the first time for level flight anomaly detection; (2) by setting the residual of multiple consecutive test data to exceed the threshold range, it is determined to be an anomaly, effectively avoiding false alarms. Attached Figure Description

[0019] Figure 1 This is a schematic block diagram of level flight anomaly detection in this invention. Detailed Implementation

[0020] The present invention will be further described in detail below with reference to embodiments, but the implementation of the present invention is not limited thereto.

[0021] Example 1 A method for detecting anomalies in the level flight of a fixed-wing UAV is proposed. By analyzing historical flight parameter data, characteristic parameters of the level flight phase are extracted. Combined with flight mechanics models and statistical methods, a normal threshold range for the residual between theoretical and measured normal acceleration is established, thereby realizing real-time monitoring and anomaly detection of the UAV's level flight status.

[0022] Example 2 like Figure 1 As shown, a method for detecting anomalies in the level flight of a fixed-wing UAV is proposed. By analyzing historical flight parameter data, characteristic parameters of the level flight phase are extracted. Combined with flight mechanics models and statistical methods, a normal threshold range for the residual between theoretical normal acceleration and measured normal acceleration is established, thereby realizing real-time monitoring and anomaly detection of the UAV's level flight status.

[0023] This includes the following steps: Step 1: Data Acquisition: Collect historical flight parameter data from multiple drone sorties; Step 2: Level Flight Data Extraction: Extract level flight data based on the variation characteristics of flight parameters; Step 3: Data Preprocessing: Preprocess the extracted flight parameter data; Step 4: Calculation of normal acceleration: Based on the flight mechanics model, calculate the theoretical normal acceleration of the UAV using the roll angle; Step 5: Residual Calculation: Compare the theoretical normal acceleration with the measured normal acceleration and calculate the residual; Step 6: Threshold Establishment: Based on the Laida criterion, perform statistical analysis on the residual data to establish the residual threshold range under normal flight conditions; Step 7, Anomaly Detection: Compare the residual of the flight parameter to be tested with the threshold range. If it exceeds the threshold range, it is determined to be an anomaly.

[0024] In step one, the historical flight parameter data includes the flight phase status word, flight time, roll angle, and normal acceleration.

[0025] In step two, the level flight data is extracted: First, the data of each flight is divided into flight stages according to the flight stage status word, and the level flight segment data is extracted as the level flight data.

[0026] In step three, the preprocessing of the extracted flight parameter data includes time-stamp alignment and noise removal.

[0027] The time-stamp alignment involves first aligning the data with time stamps, then using the same time interval as the data sampling interval, with all parameters using the most recent data as the current data, concatenating the first and second parts of the time-stamped data, and then using formula (1) to represent the concatenated data from different times. (1); in: Represents the roll angle array. For parameters The elements in the table represent the roll angle data at different times; Represents the array of normal accelerations. for The elements in the table represent the normal acceleration data at different times.

[0028] The noise removal process involves using a moving average filtering method to denoise all data. A suitable sliding window length is selected based on the data characteristics. The moving average filtering method is shown in formula (2). (2).

[0029] In step four, the calculation of normal acceleration involves using the roll angle to calculate the theoretical normal acceleration of the UAV based on the flight mechanics model. The specific calculation formula is formula (3): (3); in: g The acceleration due to gravity is typically taken as 9.8 m / s². 2 .

[0030] In step five, residual calculation involves comparing the theoretical normal acceleration with the measured normal acceleration to calculate the residual. The calculation method is as shown in formula (4): (4).

[0031] In step six, the threshold is established as follows: based on the Laida criterion, statistical analysis is performed on the residual data, and the mean and standard deviation of the residuals are calculated according to formula (5) to establish the residual threshold range under normal flight conditions. The range is set to be within the normal range of the residual between the theoretical normal acceleration and the measured normal acceleration. (5).

[0032] In step seven, the residuals of the flight parameters to be tested are compared with the threshold range. To avoid false alarms, if the residuals of multiple consecutive test data exceed the threshold range, it is determined to be abnormal and an alarm signal is issued.

[0033] Compared with the prior art, the present invention has the following advantages and beneficial effects: I. The present invention provides a method for detecting anomalies in level flight of a fixed-wing UAV, (1) accurately extracting flight data in the level flight phase by analyzing flight data characteristics; (2) calculating theoretical normal acceleration by combining flight mechanics model, thereby improving the accuracy and reliability of anomaly detection; (3) effectively distinguishing between normal flight state and abnormal flight state by establishing a residual threshold range based on the Laida criterion; (4) determining anomalies by setting multiple consecutive residuals of test data exceeding the threshold range, thereby effectively avoiding false alarms; (5) this method is applicable to a variety of fixed-wing UAVs and has strong versatility.

[0034] II. The present invention provides a method for detecting level flight anomalies of fixed-wing UAVs, (1) combining flight mechanics models with statistical methods for the first time for level flight anomaly detection; (2) by setting the residual of multiple consecutive test data to exceed the threshold range, it is determined to be an anomaly, effectively avoiding false alarms.

[0035] Example 3 A method for detecting level flight anomalies in a fixed-wing unmanned aerial vehicle (UAV) includes the following steps: (S1) Data Acquisition The system acquires historical flight parameter data from multiple UAV flight control systems, including key parameters such as flight phase status words, flight time, roll angle, and normal acceleration.

[0036] (S2) Level Flight Data Extraction Based on the changing characteristics of flight parameters, the level flight data is extracted. The specific method is as follows: First, the data of each sortie is divided into flight stages according to the flight stage status word, and the level flight segment data (in this implementation case, flight status word = 8 indicates the level flight stage of the UAV) is extracted as the level flight data.

[0037] (S3) Data Preprocessing The extracted data is preprocessed, including time-stamp alignment and noise removal.

[0038] ①Time stamp alignment: First, the data is time stamp aligned, that is, the data reading start time and interval are reset, the first frame of data is taken as the zero time, and then the same time interval (40ms in this case) is used as the data sampling interval. All parameters are based on the most recent data as the current time data. The time stamp aligned data is concatenated first and second, and the data at different times after splicing can be represented by formula (1).

[0039] (1); in: Represents the roll angle array. For parameters The elements in the table represent the roll angle data at different times; Represents the array of normal accelerations. for The elements in the table represent the normal acceleration data at different times.

[0040] ② Noise Removal: After data time-stamp alignment, due to electromagnetic radiation and other factors affecting the circuitry, the data will contain various types of noise. Therefore, noise removal is necessary. A moving average filtering method is used to remove noise from all data. An appropriate sliding window length is selected based on the data characteristics (10 data points in this case), using parameters... For example, the filtered parameters are expressed as follows: Other parameter filtering methods are similar, and the moving average filtering method is shown in formula (2).

[0041] (2).

[0042] (S4) Calculation of normal acceleration Based on the flight mechanics model, the theoretical normal acceleration of the UAV is calculated using the roll angle. The specific calculation formula is formula (3): (3); in: g The acceleration due to gravity is typically taken as 9.8 m / s². 2 .

[0043] (S5) Residual Calculation The theoretical normal acceleration is compared with the measured normal acceleration, and the residual is calculated. The calculation method is as shown in formula (4).

[0044] (4).

[0045] (S6) Threshold establishment Based on the Laida criterion (3σ principle), statistical analysis was performed on the residual data, and the mean of the residuals was calculated according to formula (5). (This implementation case calculates) ) and standard deviation (This implementation case calculates) Establish the residual threshold range under normal flight conditions, and based on the Laida criterion, [determine / set the range]. The normal range of the residual between the theoretical normal acceleration and the measured normal acceleration is set (the normal range calculated in this implementation case is -0.85028335~0.90848709).

[0046] (5).

[0047] (S7) Anomaly Detection The residuals of the flight parameters to be tested are compared with the threshold range. To avoid false alarms, if the residuals of multiple consecutive test data (10 consecutive points in this implementation case) exceed the threshold range, it is judged as abnormal and an alarm signal is issued.

[0048] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for detecting level flight anomalies in a fixed-wing unmanned aerial vehicle (UAV), characterized in that: By analyzing historical flight parameter data, characteristic parameters of the level flight phase are extracted. Combining flight mechanics models and statistical methods, a normal threshold range for the residual between theoretical and measured normal acceleration is established, thereby enabling real-time monitoring and anomaly detection of the UAV's level flight status. Includes the following steps: Step 1: Data Acquisition: Collect historical flight parameter data from multiple drone sorties; Step 2: Level Flight Data Extraction: Extract level flight data based on the variation characteristics of flight parameters; Step 3: Data Preprocessing: Preprocess the extracted flight parameter data; Step 4: Calculation of normal acceleration: Based on the flight mechanics model, calculate the theoretical normal acceleration of the UAV using the roll angle; Step 5: Residual Calculation: Compare the theoretical normal acceleration with the measured normal acceleration and calculate the residual; Step 6: Threshold Establishment: Based on the Laida criterion, perform statistical analysis on the residual data to establish the residual threshold range under normal flight conditions; Step 7, Anomaly Detection: Compare the residual of the flight parameter to be tested with the threshold range. If it exceeds the threshold range, it is judged as an anomaly. In step one, the historical flight parameter data includes flight phase status words, flight time, roll angle, and normal acceleration; In step three, the preprocessing of the extracted flight parameter data includes time-stamp alignment and noise removal; The time-stamp alignment is as follows: First, the data is time-stamped. Then, the same time interval is used as the data sampling interval. All parameters use the most recent data as the data at the current time. The time-stamped data is then concatenated. The data at different times after concatenation can be expressed by formula (1): （1）； in: Represents the roll angle array. For parameters The elements in the table represent the roll angle data at different times; Represents the array of normal accelerations. for The elements in the table represent normal acceleration data at different times; In step four, the calculation of normal acceleration involves calculating the theoretical normal acceleration of the UAV using the roll angle based on the flight mechanics model. The specific calculation formula is formula (3): （3）； in: g The acceleration due to gravity is taken as 9.8 m / s². 2 ; φ' i This represents the roll angle data at time i after sliding filtering.

2. The method for detecting level flight anomalies in a fixed-wing unmanned aerial vehicle according to claim 1, characterized in that: In step two, the level flight data is extracted: First, the data of each flight is divided into flight stages according to the flight stage status word, and the data of the level flight segment is extracted as the level flight data.

3. The method for detecting level flight anomalies in a fixed-wing unmanned aerial vehicle according to claim 1, characterized in that: The noise removal process involves using a moving average filtering method to denoise all data. A suitable sliding window length is selected based on the data characteristics. The moving average filtering method is shown in formula (2). （2）； in, i Indicates the index of the current processing time; j Indicates a loop counter; l Indicates the length of the sliding window; φ j This represents the roll angle data extracted at the j-th time moment; φ' i This represents the roll angle data at time i after sliding filtering.

4. The method for detecting level flight anomalies in a fixed-wing unmanned aerial vehicle according to claim 3, characterized in that: In step five, residual calculation involves comparing the theoretical normal acceleration with the measured normal acceleration to calculate the residual. The calculation method is as shown in formula (4): （4）； in, This is the theoretical normal acceleration; This represents the measured normal acceleration.

5. A method for detecting level flight anomalies in a fixed-wing unmanned aerial vehicle according to claim 4, characterized in that: In step six, threshold establishment: Based on the Laida criterion, statistical analysis is performed on the residual data, and the mean and standard deviation of the residuals are calculated according to formula (5) to establish the residual threshold range under normal flight conditions. The range is set to be within the normal range of the residual between the theoretical normal acceleration and the measured normal acceleration. （5）； in, n It is an integer.

6. The method for detecting level flight anomalies in a fixed-wing unmanned aerial vehicle according to claim 5, characterized in that: In step seven, the residuals of the flight parameters to be tested are compared with the threshold range. To avoid false alarms, if the residuals of multiple consecutive test data exceed the threshold range, it is determined to be abnormal and an alarm signal is issued.

Images