An unmanned aerial vehicle propeller fault online monitoring and early warning system

Abstract

The application discloses an unmanned aerial vehicle propeller fault online monitoring and early warning system and relates to the technical field of unmanned aerial vehicle health monitoring.The system comprises a benchmark matrix construction module, a data acquisition module, a benchmark matching and deviation calculation module, a dynamic threshold adjustment module and a hierarchical early warning execution module.The application constructs a propeller health benchmark matrix covering multiple types of environment and multiple types of working condition combinations, provides a unified reference benchmark for fault determination in different flight scenarios, and based on the benchmark matching logic of the current environment and working condition characteristics, directly calls the corresponding health state characteristic vector as the benchmark when there is a completely matched environment and working condition combination, selects the characteristic vector corresponding to the environment and working condition combination with the smallest deviation as the benchmark when there is no completely matched combination, realizes accurate adaptation of the determination benchmark in different flight scenarios, and eliminates the interference of normal characteristic fluctuations of the propeller in different environments and working conditions on the fault determination result.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) health monitoring technology, specifically to an online monitoring and early warning system for UAV propeller malfunctions. Background Technology

[0002] As a core piece of equipment in the low-altitude economy, drones have been widely used in various fields such as industrial inspection, logistics and transportation, geographic surveying and mapping, and emergency rescue. Their operation scenarios cover a variety of complex environments, including plains, mountains, deserts, and high altitudes. The requirements for long endurance and high reliability in these operations are constantly increasing. As the core power actuator of drones, the propeller directly determines the flight stability and safety of drones. During operation, propellers are subjected to multiple influences such as alternating loads, changes in environmental temperature and humidity, and wind and sand erosion. They are a core component with a high incidence of drone failures. If propeller failures are not detected and dealt with in a timely manner, they can easily lead to crashes, causing equipment and property damage, and even posing a serious threat to the safety of personnel and facilities on the ground. With the continuous expansion of industrial-grade drone operation scenarios, the industry's demand for online monitoring and early warning of propeller operation status is becoming increasingly urgent.

[0003] However, existing UAV propeller fault monitoring technologies mostly employ feature comparison and judgment methods with fixed thresholds, failing to adapt to real-time changes in environmental conditions and operating conditions during flight. This results in a high probability of false alarms and missed alarms in complex environments and under varying load conditions. Furthermore, most solutions' health benchmarks do not cover a full range of environmental and operating condition combinations, making them unable to adapt to the characteristic fluctuation patterns of normal propeller operation under different flight scenarios. Some solutions lack clear rules for benchmark matching logic, failing to accurately retrieve the most suitable benchmark when no perfectly matching environmental and operating condition combination exists. Simultaneously, existing solutions lack a linkage mechanism between scenario risk levels and early warning thresholds. This prevents tightening of early warning standards in high-risk flight scenarios and reduces the probability of false alarms in low-risk scenarios. It also fails to achieve graded early warning and coordinated response from the flight control unit, making it difficult to meet the high-reliability monitoring requirements of complex industrial-grade UAV operation scenarios. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an online monitoring and early warning system for UAV propeller faults. This invention constructs a propeller health benchmark matrix covering multiple environmental and operational condition combinations to provide a unified reference benchmark for fault determination under different flight scenarios. Based on the benchmark matching logic of the current environmental and operational condition characteristics, when a perfectly matching environmental and operational condition combination exists, the corresponding health status feature vector is directly retrieved as the benchmark. When no perfectly matching combination exists, the feature vector corresponding to the environmental and operational condition combination with the smallest deviation is selected as the benchmark. This achieves accurate adaptation of the judgment benchmark under different flight scenarios, eliminates the interference of normal propeller characteristic fluctuations under different environments and operational conditions on the fault determination results, and ensures that the fault determination benchmark completely corresponds to the current flight state.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an online monitoring and early warning system for unmanned aerial vehicle (UAV) propeller faults, the system comprising: The benchmark matrix construction module: Based on the historical flight data and bench calibration data of the UAV, a propeller health benchmark matrix covering multiple environments and multiple operating conditions is constructed, including the health status feature vector corresponding to each combination of environmental conditions and operating conditions. After the matrix is ​​constructed, it is stored in the airborne storage unit. Data acquisition module: During the flight of the UAV, it continuously collects real-time data in environmental, working condition and status dimensions, and performs preprocessing and feature extraction to obtain multi-dimensional real-time feature values; Benchmark matching and deviation calculation module: Based on the real-time feature values ​​of the current environment dimension and operating condition dimension in the multi-dimensional real-time feature values, the module matches the health status feature vector of the corresponding environmental and operating condition combination from the propeller health benchmark matrix, compares the real-time feature value of the status dimension with the matched health status feature vector, and calculates the comprehensive deviation index. Dynamic threshold adjustment module: Calculates the scenario risk coefficient based on the real-time feature values ​​of the current environment dimension and working condition dimension in the multi-dimensional real-time feature values, and dynamically adjusts the warning judgment threshold according to the scenario risk coefficient; The graded early warning execution module compares the comprehensive deviation index with the dynamically adjusted early warning judgment threshold, triggers the corresponding early warning signal based on the comparison result, and coordinates with the flight control unit to execute the corresponding flight strategy.

[0006] Furthermore, in the benchmark matrix construction module, the historical flight data is the fault-free historical operation data of the UAV with a cumulative flight time of more than 1000 hours in all scenarios, and the bench calibration data is the propeller ground bench calibration benchmark data covering the entire environment and all operating conditions. The historical flight data and bench calibration data are filtered and denoised and outlier removed. Multiple environmental types include ambient temperature, relative humidity, wind and dust concentration, and atmospheric pressure. Multiple operating conditions include rotor speed, maneuvering overload, flight attitude angle, and continuous flight time. ≥100 sets of normal operation samples are extracted under each combination of environmental conditions to generate corresponding health status feature vectors. The health status feature vectors corresponding to all combinations of environmental conditions are integrated to construct the propeller health benchmark matrix. The health status feature vectors include the blade root vibration spectrum benchmark, the motor bus current harmonic benchmark, the propeller tip passing sound frequency feature benchmark, and the propeller surface micro-strain distribution benchmark.

[0007] Furthermore, the data acquisition module is equipped with an environmental sensor group, an operating condition sensor group, and a state sensor group. The environmental sensor group collects real-time environmental data including ambient temperature, relative humidity, dust concentration, and atmospheric pressure. The operating condition sensor group collects real-time operating condition data including rotor speed, maneuvering overload, flight attitude angle, and continuous flight duration. The state sensor group collects real-time state data including blade root vibration amplitude and spectrum, motor bus current harmonic distortion rate, blade tip passing sound frequency characteristics, and blade surface micro-strain mean. All collected data are filtered, denoised, and normalized preprocessed. Then, corresponding parameters are extracted from the preprocessed environmental, operating condition, and state dimension data as real-time feature values, thereby obtaining multi-dimensional real-time feature values.

[0008] Furthermore, in the benchmark matching and deviation calculation module, real-time feature values ​​of the environmental dimension and operating condition dimension are extracted from the multi-dimensional real-time feature values. All environmental and operating condition combinations in the propeller health benchmark matrix are traversed. When there is a combination that completely matches the real-time feature values ​​of the current environmental and operating conditions dimensions, the health status feature vector corresponding to that combination is directly retrieved as the matching benchmark. When there is no completely matching combination, the environmental and operating condition combination with the smallest deviation from the real-time feature values ​​of the current environmental and operating conditions dimensions is selected, and the health status feature vector corresponding to that combination is retrieved as the matching benchmark. After benchmark matching is completed, the real-time feature values ​​of each item in the state dimension are extracted. The real-time feature values ​​of the state dimension are compared item by item with the corresponding benchmarks in the matched health status feature vectors. Then, the comprehensive deviation index is calculated using the comprehensive deviation index calculation formula.

[0009] Furthermore, in the benchmark matching and deviation calculation module, the formula for calculating the comprehensive deviation index is as follows: ,in, This is a comprehensive deviation index, with a value range of 0-1. The total number of feature dimensions for real-time feature values ​​in the state dimension. Let i be the i-th real-time feature value in the state dimension. The benchmark is the corresponding feature of the i-th feature in the matched health status feature vector. The weight coefficient of the i-th feature is determined by the statistical results of the contribution of each state feature to the occurrence of the fault in the historical fault data of the UAV propeller, and the sum of all weight coefficients is 1.

[0010] Furthermore, in the dynamic threshold adjustment module, real-time feature values ​​of the environmental dimension and the working condition dimension are extracted from multi-dimensional real-time feature values. The scenario risk coefficient is calculated using the scenario risk coefficient calculation formula. Based on the scenario risk coefficient, the warning judgment threshold is dynamically adjusted using the warning judgment threshold dynamic adjustment formula. The warning judgment threshold includes a low threshold and a high threshold.

[0011] Furthermore, in the dynamic threshold adjustment module, the formula for calculating the scene risk coefficient is: ,in, This is the scenario risk coefficient, with a value ranging from 0 to 1. This is the environmental comprehensive risk weighting coefficient. This is the comprehensive risk weighting coefficient for the working condition, and , Let m be the real-time feature value of the environmental dimension. The maximum allowable value for the m-th real-time feature value in the environmental dimension is... Let be the weight coefficient of the m-th real-time feature value in the environmental dimension, and , For the nth real-time feature value in the working condition dimension, The maximum allowable value for the nth real-time feature value in the operating condition dimension. Let be the weight coefficient of the nth real-time feature value in the working condition dimension, and All weight coefficients were determined by statistical results of the contribution of each parameter to the occurrence of the fault in the historical fault data of the UAV propeller.

[0012] Furthermore, in the dynamic threshold adjustment module, the calculation formula for the dynamic adjustment of the early warning judgment threshold is as follows: ,in, The lower threshold is dynamically adjusted. The high threshold is dynamically adjusted. The baseline low threshold is determined by the statistical results of the cumulative fault-free flight data of the UAV. The baseline high threshold is determined by statistical results from historical samples of drone propeller failures. The threshold adjustment coefficient is determined by statistical results of historical flight data and historical fault data corresponding to all UAV flight scenarios in all environments and under all operating conditions. This represents the scenario risk coefficient.

[0013] Furthermore, in the graded early warning execution module, the comprehensive deviation index is compared with the dynamically adjusted low threshold and the dynamically adjusted high threshold, respectively. When the comprehensive deviation index is less than the dynamically adjusted low threshold, the propeller is determined to be in a healthy operating state, and no early warning signal is triggered. When the comprehensive deviation index is greater than or equal to the dynamically adjusted low threshold and less than the average of the dynamically adjusted low threshold and the dynamically adjusted high threshold, a first-level early warning is triggered, sending a yellow early warning signal to the ground control terminal, and the flight control unit maintains the current flight state and increases the frequency of propeller operation data acquisition and monitoring. When the comprehensive deviation index is greater than or equal to the average of the dynamically adjusted low threshold and the dynamically adjusted high threshold and less than the dynamically adjusted high threshold, a second-level early warning is triggered, sending an orange early warning signal to the ground control terminal, and the flight control unit adjusts the flight attitude to reduce the propeller operating load. When the comprehensive deviation index is greater than or equal to the dynamically adjusted high threshold, a third-level mandatory early warning is triggered, sending a red early warning signal to the ground control terminal, and the flight control unit executes an autonomous return-to-home procedure or an autonomous forced landing procedure at the nearest site.

[0014] Compared with existing technologies, this online monitoring and early warning system for UAV propeller faults has the following advantages: I. This invention constructs a propeller health benchmark matrix covering multiple environmental and operating condition combinations, providing a unified reference benchmark for fault determination in different flight scenarios. Based on the benchmark matching logic of the current environmental and operating condition characteristics, when a perfectly matching environmental and operating condition combination exists, the corresponding health status feature vector is directly retrieved as the benchmark. When no perfectly matching combination exists, the feature vector corresponding to the environmental and operating condition combination with the smallest deviation is selected as the benchmark. This achieves accurate adaptation of the judgment benchmark under different flight scenarios, eliminates the interference of normal propeller characteristic fluctuations under different environments and operating conditions on the fault determination results, and ensures that the fault determination benchmark completely corresponds to the current flight state.

[0015] Second, this invention calculates the scenario risk coefficient based on the current environment and operating conditions, dynamically adjusts the warning judgment threshold based on the scenario risk coefficient, establishes a linkage between the scenario risk level and the warning judgment threshold, realizes the adaptation and tightening of the warning standard in high-risk scenarios and the reasonable relaxation of the warning standard in low-risk scenarios, realizes graded warning by comparing the interval of the comprehensive deviation index and the dynamically adjusted warning threshold, and links the flight control unit to execute the corresponding flight strategy to realize the layered response of different fault levels, ensuring the accuracy of the warning judgment and the rationality of the fault response in different flight scenarios.

[0016] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0018] Figure 1 A flowchart of an online monitoring and early warning system for unmanned aerial vehicle (UAV) propeller faults; Figure 2 This is a framework diagram of a benchmark matching and deviation calculation module in an online monitoring and early warning system for UAV propeller faults. Figure 3 This is a framework diagram of a dynamic threshold adjustment module in an online monitoring and early warning system for UAV propeller faults. Detailed Implementation

[0019] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below. Example

[0020] In the scenario of industrial-grade hexacoach drones inspecting power lines in mountainous areas, the operating area has a large altitude range, significant temperature difference between day and night, high air humidity, and abundant vegetation and dust along the line. The drones need to frequently perform actions such as hovering, changing altitude, and turning maneuvers. The continuous operation time of a single drone generally exceeds 2 hours. The propellers are subjected to alternating loads and environmental erosion for a long time, resulting in a high risk of hidden failures.

[0021] The benchmark matrix construction module first collects historical flight data of the target model hexacoach UAV with a cumulative flight time of 1200 hours in the entire scenario of mountain power inspection without any faults. This includes all normal flight samples of different altitudes, seasons, and operating times in the inspection area to ensure the scenario adaptability of the data. At the same time, it collects bench calibration data of the propeller on the ground to simulate the environmental parameter range and operating condition range of the mountain power inspection scenario. During the bench calibration process, it reproduces the temperature, humidity, dust, and air pressure range of the inspection scenario, as well as the entire normal operating range of rotor speed, maneuvering overload, and flight attitude to obtain standard operating data of the propeller under ideal healthy conditions. Historical flight data and bench calibration data undergo unified preprocessing, specifically filtering and outlier removal: filtering and denoising employs a common method combining moving average filtering and wavelet denoising to eliminate invalid signals such as electromagnetic interference and mechanical vibration noise introduced during data acquisition; outlier removal uses the 3σ criterion to delete invalid samples that deviate from the normal range due to instantaneous fluctuations in sensors and packet loss in data transmission, ensuring the accuracy of the baseline data. When constructing a propeller health benchmark matrix covering multiple environments and operating conditions, the multiple environments include ambient temperature, relative humidity, wind and dust concentration, and atmospheric pressure, while the multiple operating conditions include rotor speed, maneuvering overload, flight attitude angle, and continuous flight duration. The environmental and operating condition parameters are discretized and combined to form multiple independent environmental and operating condition combinations. No less than 100 normal operation samples are extracted under each environmental and operating condition combination. The samples within the same group are statistically averaged to generate the health state feature vector corresponding to the combination. The health state feature vector specifically includes the blade root vibration spectrum benchmark, the motor bus current harmonic benchmark, the propeller tip passing sound frequency feature benchmark, and the propeller surface micro-strain distribution benchmark. The health status feature vectors corresponding to all environmental conditions are integrated to construct a propeller health benchmark matrix covering the entire scenario of power line inspection in mountainous areas. After construction, the matrix is ​​stored in the UAV's onboard storage unit. Figure 1 As shown, this provides a fixed and directly accessible reference benchmark for subsequent real-time monitoring.

[0022] Data acquisition module: Three independent sensor groups are mounted on the drone body: the environmental sensor group is installed on the top of the drone body in an unobstructed position, with a sampling frequency of 1Hz; the operating condition sensor group is integrated inside the drone flight control unit, with a sampling frequency of 10Hz; and the status sensor group is installed at the root of the propeller motor and the root of the propeller blade, with a sampling frequency of 1kHz. The three sets of sensors collect real-time data in their respective dimensions: the environmental sensor set collects real-time environmental data including ambient temperature, relative humidity, dust concentration, and atmospheric pressure; the operating condition sensor set collects real-time operating condition data including rotor speed, maneuvering overload, flight attitude angle, and continuous flight duration; and the state sensor set collects real-time state dimension data including blade root vibration amplitude and spectrum, motor bus current harmonic distortion rate, blade tip passing sound frequency characteristics, and average micro-strain of the blade surface. During the drone's power line inspection flight mission, three sets of sensors continuously collect real-time data in the corresponding dimensions. All collected data are uniformly filtered, denoised, and normalized preprocessed. The normalization adopts the min-max normalization method to convert data with different dimensions and different numerical ranges into standardized data in the 0-1 range, eliminating differences in data magnitude. For the preprocessed environmental, operating condition, and state data, the corresponding operating parameters are extracted as real-time feature values, and finally integrated to obtain multi-dimensional real-time feature values.

[0023] The benchmark matching and deviation calculation module extracts real-time feature values ​​of the environmental and operational conditions from multi-dimensional real-time feature values, traverses all environmental and operational condition combinations within the propeller health benchmark matrix in the onboard storage unit, and executes benchmark matching logic. If there is an environmental condition combination in the propeller health baseline matrix that perfectly matches the real-time feature values ​​of the current environmental dimension and operating condition dimension, the health status feature vector corresponding to that combination is directly retrieved as the matching baseline for this monitoring. If there is no perfectly matching combination in the propeller health baseline matrix, the Euclidean distance between the real-time feature values ​​of the current environmental dimension and operating condition dimension and the feature vectors of each environmental condition combination in the matrix is ​​calculated. The environmental condition combination with the smallest Euclidean distance deviation is selected, and the health status feature vector corresponding to that combination is retrieved as the matching baseline to ensure the adaptability of the matching baseline to the current flight scenario. After benchmark matching is completed, real-time feature values ​​of each item in the state dimension are extracted. These real-time feature values ​​are then compared item by item with the corresponding benchmark values ​​in the matched health state feature vector. The comprehensive deviation index is then quantified and calculated using the comprehensive deviation index calculation formula. Figure 2 As shown; the formula for calculating the comprehensive deviation index is: ,in, This is a comprehensive deviation index, with a value range of 0-1. The total number of feature dimensions for real-time feature values ​​in the state dimension. Let i be the i-th real-time feature value in the state dimension. The benchmark is the corresponding feature of the i-th feature in the matched health status feature vector. The weight coefficient of the i-th feature is determined by the statistical results of the contribution of each state feature to the occurrence of the fault in the historical fault data of the UAV propeller, and the sum of all weight coefficients is 1.

[0024] Dynamic threshold adjustment module: Extracts real-time feature values ​​of environmental and working condition dimensions from multi-dimensional real-time feature values, and calculates the scene risk coefficient of the current inspection scene using the scene risk coefficient calculation formula. The scene risk coefficient calculation formula is as follows: ,in, This is the scenario risk coefficient, with a value ranging from 0 to 1. This is the environmental comprehensive risk weighting coefficient. This is the comprehensive risk weighting coefficient for the working condition, and , Let m be the real-time feature value of the environmental dimension. The maximum allowable value for the m-th real-time feature value in the environmental dimension is... Let be the weight coefficient of the m-th real-time feature value in the environmental dimension, and , For the nth real-time feature value in the working condition dimension, The maximum allowable value for the nth real-time feature value in the operating condition dimension. Let be the weight coefficient of the nth real-time feature value in the working condition dimension, and All weight coefficients were determined by statistical results of the contribution of each parameter to the occurrence of the fault in the historical fault data of the UAV propeller; After calculating the scenario risk coefficient, the warning judgment thresholds, including low and high thresholds, are dynamically adjusted using a dynamic adjustment formula for the warning judgment threshold. Figure 3 As shown; the calculation formula for the dynamic adjustment of the early warning judgment threshold is: ,in, The lower threshold is dynamically adjusted. The high threshold is dynamically adjusted. The baseline low threshold is determined by the statistical results of the cumulative fault-free flight data of the UAV. The baseline high threshold is determined by statistical results from historical samples of drone propeller failures. The threshold adjustment coefficient is determined by statistical results of historical flight data and historical fault data corresponding to all UAV flight scenarios in all environments and under all operating conditions. This represents the scenario risk coefficient.

[0025] The graded early warning execution module, based on the comprehensive deviation index and dynamically adjusted low and high thresholds, compares the comprehensive deviation index with each threshold within a range, and performs graded judgment and response: When the comprehensive deviation index is less than the dynamically adjusted low threshold, the propeller is determined to be in a healthy operating state, no early warning signal is triggered, and only the current operating data is stored; when the comprehensive deviation index is greater than or equal to the dynamically adjusted low threshold and less than the average of the dynamically adjusted low and high thresholds, a level one early warning is triggered, sending a yellow early warning signal to the ground power inspection control terminal, while the flight control unit maintains the current flight state and increases the frequency of propeller operation data collection and monitoring to strengthen status tracking; when the comprehensive deviation index is greater than or equal to the average of the dynamically adjusted low and high thresholds and less than the dynamically adjusted high threshold, a level two early warning is triggered, sending an orange early warning signal to the ground control terminal, and the flight control unit adjusts to a level flight cruise attitude, reducing the amplitude of maneuvers and propeller operating load to slow down the development of the fault; when the comprehensive deviation index is greater than or equal to the dynamically adjusted high threshold, a level three mandatory early warning is triggered, sending a red early warning signal to the ground control terminal, and the flight control unit executes an autonomous return-to-home procedure or an autonomous forced landing procedure at the nearest site.

[0026] In summary, for power line inspection applications in mountainous areas, a benchmark matrix construction module establishes a comprehensive health reference benchmark adapted to the inspection scenario; a data acquisition module enables standardized collection and feature extraction of multi-dimensional data; a benchmark matching and deviation calculation module completes accurate benchmark matching and quantitative calculation of fault deviation; a dynamic threshold adjustment module enables scenario-adaptive adjustment of early warning thresholds; and a graded early warning execution module completes fault classification judgment and flight control linkage response. These modules are closely integrated and form a closed-loop logic, effectively solving the problems of high false alarm and missed alarm rates and untimely response in propeller fault monitoring under complex scenarios.

[0027] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An online monitoring and early warning system for unmanned aerial vehicle (UAV) propeller faults, characterized in that, The system includes: The benchmark matrix construction module: Based on the historical flight data and bench calibration data of the UAV, a propeller health benchmark matrix covering multiple environments and multiple operating conditions is constructed, including the health status feature vector corresponding to each combination of environmental conditions and operating conditions. After the matrix is ​​constructed, it is stored in the airborne storage unit. Data acquisition module: During the flight of the UAV, it continuously collects real-time data in environmental, working condition and status dimensions, and performs preprocessing and feature extraction to obtain multi-dimensional real-time feature values; Benchmark matching and deviation calculation module: Based on the real-time feature values ​​of the current environment dimension and operating condition dimension in the multi-dimensional real-time feature values, the module matches the health status feature vector of the corresponding environmental and operating condition combination from the propeller health benchmark matrix, compares the real-time feature value of the status dimension with the matched health status feature vector, and calculates the comprehensive deviation index. Dynamic threshold adjustment module: Calculates the scenario risk coefficient based on the real-time feature values ​​of the current environment dimension and working condition dimension in the multi-dimensional real-time feature values, and dynamically adjusts the warning judgment threshold according to the scenario risk coefficient; The graded early warning execution module compares the comprehensive deviation index with the dynamically adjusted early warning judgment threshold, triggers the corresponding early warning signal based on the comparison result, and coordinates with the flight control unit to execute the corresponding flight strategy.

2. The online monitoring and early warning system for UAV propeller faults according to claim 1, characterized in that, In the benchmark matrix construction module, historical flight data consists of fault-free historical operation data of the UAV with a cumulative flight time of over 1000 hours across all scenarios. Bench calibration data consists of propeller ground bench calibration benchmark data covering the entire environmental and operating condition range. The historical flight data and bench calibration data are filtered, denoised, and outlier removed. Multiple environmental types include ambient temperature, relative humidity, dust concentration, and atmospheric pressure. Multiple operating conditions include rotor speed, maneuvering overload, flight attitude angle, and continuous flight time. ≥100 sets of normal operation samples are extracted for each environmental condition combination to generate corresponding health status feature vectors. The health status feature vectors corresponding to all environmental condition combinations are integrated to construct the propeller health benchmark matrix. The health status feature vectors include blade root vibration spectrum benchmark, motor bus current harmonic benchmark, blade tip passing sound frequency feature benchmark, and blade surface micro-strain distribution benchmark.

3. The online monitoring and early warning system for UAV propeller faults according to claim 1, characterized in that, The data acquisition module is equipped with an environmental sensor group, an operating condition sensor group, and a state sensor group. The environmental sensor group collects real-time environmental data including ambient temperature, relative humidity, dust concentration, and atmospheric pressure. The operating condition sensor group collects real-time operating condition data including rotor speed, maneuvering overload, flight attitude angle, and continuous flight duration. The state sensor group collects real-time state data including blade root vibration amplitude and spectrum, motor bus current harmonic distortion rate, blade tip passing sound frequency characteristics, and blade surface micro-strain mean. All collected data are filtered, denoised, and normalized preprocessed. Then, corresponding parameters are extracted from the preprocessed environmental, operating condition, and state dimension data as real-time feature values, thereby obtaining multi-dimensional real-time feature values.

4. The online monitoring and early warning system for UAV propeller faults according to claim 1, characterized in that, In the benchmark matching and deviation calculation module, real-time feature values ​​of the environmental dimension and operating condition dimension are extracted from the multi-dimensional real-time feature values. All environmental and operating condition combinations in the propeller health benchmark matrix are traversed. When there is a combination that completely matches the real-time feature values ​​of the current environmental and operating conditions dimensions, the health status feature vector corresponding to the combination is directly retrieved as the matching benchmark. When there is no completely matching combination, the environmental and operating condition combination with the smallest deviation from the real-time feature values ​​of the current environmental and operating conditions dimensions is selected, and the health status feature vector corresponding to the combination is retrieved as the matching benchmark. After benchmark matching is completed, the real-time feature values ​​of each item of the state dimension are extracted. The real-time feature values ​​of the state dimension are compared item by item with the corresponding benchmarks in the matched health status feature vectors. Then, the comprehensive deviation index is calculated using the comprehensive deviation index calculation formula.

5. The online monitoring and early warning system for UAV propeller faults according to claim 4, characterized in that, In the benchmark matching and deviation calculation module, the formula for calculating the comprehensive deviation index is as follows: ,in, This is a comprehensive deviation index, with a value range of 0-1. The total number of feature dimensions for real-time feature values ​​in the state dimension. Let i be the i-th real-time feature value in the state dimension. The benchmark is the corresponding feature of the i-th feature in the matched health status feature vector. The weight coefficient of the i-th feature is determined by the statistical results of the contribution of each state feature to the occurrence of the fault in the historical fault data of the UAV propeller, and the sum of all weight coefficients is 1.

6. The online monitoring and early warning system for UAV propeller faults according to claim 1, characterized in that, In the dynamic threshold adjustment module, real-time feature values ​​of environmental dimension and working condition dimension are extracted from multi-dimensional real-time feature values. The scenario risk coefficient is calculated by the scenario risk coefficient calculation formula. Based on the scenario risk coefficient, the warning judgment threshold is dynamically adjusted by the warning judgment threshold dynamic adjustment formula. The warning judgment threshold includes low threshold and high threshold.

7. The online monitoring and early warning system for UAV propeller faults according to claim 6, characterized in that, In the dynamic threshold adjustment module, the formula for calculating the scene risk coefficient is: ,in, This is the scenario risk coefficient, with a value ranging from 0 to 1. This is the environmental comprehensive risk weighting coefficient. This is the comprehensive risk weighting coefficient for the working condition, and , Let m be the real-time feature value of the environmental dimension. The maximum allowable value for the m-th real-time feature value in the environmental dimension is... Let be the weight coefficient of the m-th real-time feature value in the environmental dimension, and , For the nth real-time feature value in the working condition dimension, The maximum allowable value for the nth real-time feature value in the operating condition dimension. Let be the weight coefficient of the nth real-time feature value in the working condition dimension, and .

8. The online monitoring and early warning system for UAV propeller faults according to claim 6, characterized in that, In the dynamic threshold adjustment module, the calculation formula for the dynamic adjustment of the early warning judgment threshold is as follows: ,in, The lower threshold is dynamically adjusted. The high threshold is dynamically adjusted. The baseline low threshold is determined by the statistical results of the cumulative fault-free flight data of the UAV. The baseline high threshold is determined by statistical results from historical samples of drone propeller failures. The threshold adjustment coefficient is determined by statistical results of historical flight data and historical fault data corresponding to all UAV flight scenarios in all environments and under all operating conditions. This represents the scenario risk coefficient.

9. The online monitoring and early warning system for UAV propeller faults according to claim 1, characterized in that, In the graded early warning execution module, the comprehensive deviation index is compared with the dynamically adjusted low threshold and the dynamically adjusted high threshold, respectively. When the comprehensive deviation index is less than the dynamically adjusted low threshold, the propeller is determined to be in a healthy operating state, and no early warning signal is triggered. When the comprehensive deviation index is greater than or equal to the dynamically adjusted low threshold and less than the average of the dynamically adjusted low threshold and the dynamically adjusted high threshold, a first-level early warning is triggered, a yellow early warning signal is sent to the ground control terminal, and the flight control unit is linked to maintain the current flight state and increase the frequency of propeller operation data acquisition and monitoring. When the comprehensive deviation index is greater than or equal to the average of the dynamically adjusted low threshold and the dynamically adjusted high threshold and less than the dynamically adjusted high threshold, a second-level early warning is triggered, an orange early warning signal is sent to the ground control terminal, and the flight control unit is linked to adjust the flight attitude to reduce the propeller operating load. When the comprehensive deviation index is greater than or equal to the dynamically adjusted high threshold, a level three mandatory early warning is triggered, a red warning signal is sent to the ground control terminal, and the flight control unit is linked to execute the autonomous return procedure or the autonomous emergency landing procedure at the nearest site.

Images