Overhead line inspection unmanned aerial vehicle wing and control method
Through the layered composite structure and high-precision sensor combined with neural network control, the problems of electronic circuit breakdown, electrostatic accumulation and airflow control lag in the high-voltage electric field are solved, and high-rootability and high-precision drone inspection are achieved.
Patent Information
- Application Number
- CN202510775884.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-11
- Publication Date
- 2025-07-22
AI Technical Summary
Existing drones have risks of electronic circuit breakdown, serious interference from electrostatic accumulation and complex airflow control lag under 25kV high-voltage electric field, affecting safety, accuracy and reliability.
It adopts a layered composite structural design, including the carbon fiber skeleton core layer, anti-high voltage coating and anti-static coating, combined with high-precision sensors and neural network control, to achieve adaptive wind-resistant control.
It significantly improves the safety, stability and reliability of the drone in high-voltage environments, reduces the risk of sensor false alarms and hardware damage, improves control accuracy and environmental adaptability, and reduces operation and maintenance costs.
Smart Images
Figure CN120348505A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of inspection of electric catenaries, and particularly to a wing for an unmanned aerial vehicle (UAV) for catenary inspection and a control method therefor. Background Art
[0002] In the field of intelligent inspection of railway catenaries, UAV technology has gradually replaced manual inspection due to its advantages of high efficiency and flexibility.
[0003] However, the existing technologies still have significant defects:
[0004] 1. Insufficient adaptability to high-voltage electric fields: The 25 kV high-voltage environment of the catenary is likely to cause breakdown of the electronic circuits of traditional UAVs. Although multi-rotor UAVs have the ability to hover, the insulation grade of their polymer wings is insufficient (breakdown voltage < 20 kV / mm), presenting a risk of short circuit.
[0005] 2. Lag in complex airflow control: Fixed-wing UAVs rely on PID control algorithms and are unable to dynamically respond to disturbances such as gusts and crosswinds.
[0006] 3. Severe interference caused by static electricity accumulation: The static charges (up to 10 kV) accumulated on the rigid wings in the high-voltage electric field trigger false alarms of sensors (such as a drift of barometric pressure data of ±5 Pa), and there is a lack of an effective dissipation mechanism, resulting in an increased hardware damage rate.
[0007] The above defects severely restrict the safety, accuracy, and reliability of UAVs in high-voltage catenary scenarios. Therefore, a systematic solution integrating high-voltage protection, intelligent wind resistance, and static electricity management is urgently needed. Summary of the Invention
[0008] In order to enable a UAV to safely inspect under a 25 kV electric field, and to perform adaptive wind resistance control and reduce the attitude error of the UAV, the present application provides a wing for an unmanned aerial vehicle for catenary inspection and a control method therefor.
[0009] The first invention object of the present application is achieved through the following technical solutions:
[0010] A contact network inspection drone wing comprises a wing body, a motor and an electric regulator, wherein the wing body is symmetrically arranged on the motor, the motor is connected to the electric regulator, and the electric regulator is connected to the drone, and a layered composite structure is arranged on the wing body, the layered composite structure comprises a core layer, a functional layer and a sensing layer, the core layer is composed of a carbon fiber skeleton for providing aerodynamic shape support for the wing, the functional layer covers the surface of the core layer, the functional layer comprises an anti-high voltage coating and an anti-static coating, the anti-high voltage coating is a multi-layer composite insulating coating, the anti-static coating is arranged in the edge area of the wing body, the anti-static coating comprises a grid structure formed by a conductive material, the sensing layer is embedded in the surface of the wing body, the sensing layer comprises at least one micro-electromechanical system air pressure sensor, and the arrangement area of the sensing layer avoids the coating coverage area of the functional layer to ensure the sensitivity of the sensor.
[0011] By adopting the above technical solutions, the drone wing has significantly improved its safety, stability and reliability in the high-voltage railway contact network environment through an innovative layered composite structure design (core layer, functional layer, sensor layer). The core layer adopts a carbon fiber skeleton to ensure aerodynamic shape support while achieving lightweight, reducing flight energy consumption and extending flight time. The anti-high-voltage coating of the functional layer is a three-layer composite, which increases the breakdown voltage to ≥30kV / mm, effectively resisting the interference of the 25kV high-voltage electric field and avoiding the risk of circuit breakdown; the anti-static coating sprays a carbon nanotube (CNT) grid (ground resistance <0.1Ω) on the edge of the wing, which can quickly dissipate the accumulated charge and eliminate the problem of sensor false alarms or hardware damage caused by static electricity. The MEMS pressure sensor of the sensing layer is arranged away from the coating area to ensure the sensitivity of the pressure data and provide accurate input for attitude control, so that the drone can patrol safely under the 25kV electric field, and can adaptively control wind resistance to reduce the attitude error of the drone.
[0012] Optionally, the conductive material grid structure in the antistatic coating is a carbon nanotube grid formed by spraying, and the grid has a grounding path with a resistance less than 0.1 ohm.
[0013] By adopting the above technical solution, the anti-static coating adopts a carbon nanotube grid structure formed by spraying. The high conductivity of the carbon nanotubes combined with the low-resistance grounding path can dissipate the accumulated charge within 0.1 seconds, effectively eliminating the electrostatic interference caused by the high-voltage electric field (25kV), and avoiding the risk of sensor data drift and electronic component breakdown. The grid structure is sprayed on the edge of the wing to specifically block the charge conduction path, while avoiding covering the sensor area to ensure the accuracy of the sensor data.
[0014] In the second aspect, the above invention objective of the present application is achieved through the following technical solutions:
[0015] A control method for the wings of an unmanned aerial vehicle (UAV) for catenary inspection, the control method for the wings of the UAV for catenary inspection comprising the steps of:
[0016] Synchronously collecting UAV flight environment data based on a sensing layer, the UAV flight environment data including three-dimensional wind field vector, air pressure difference, angular velocity, and ground relative velocity data;
[0017] Marking time stamps for the collected UAV flight environment data, and synchronizing and aligning multi-source data through time stamps to generate synchronized time-series multi-source data;
[0018] Inputting the synchronized time-series multi-source data into a preset spatio-temporal fusion neural network to generate original control commands including motor speed and wing tilt angle;
[0019] Invoking a preset dynamic compensation model based on the real-time wind field vector, correcting the original control commands, and outputting a final execution signal to the motor drive module and wing servo mechanism of the UAV.
[0020] By adopting the above technical solution, high-precision sensors integrated in the sensing layer comprehensively capture three-dimensional wind field vector, air pressure difference, angular velocity, and ground relative velocity data during the flight of the UAV, accurately quantify airflow disturbances (such as gusts and crosswinds) and spatial electric field interference, provide a data basis for wind resistance control and insulation protection, mark time stamps for the collected UAV flight environment data, and synchronize and align multi-source data through time stamps to solve the phase error caused by the difference in sensor sampling rates, avoid lag in control commands, generate high-consistency time-series data packets, provide high-quality input for the neural network, reduce the UAV wing attitude estimation error, input the synchronized time-series multi-source data into a preset spatio-temporal fusion neural network, the spatio-temporal fusion neural network captures long-time series dependencies, combines the attention mechanism to focus on key wind speed mutation moments, fuses bio-inspired features, generates original control commands including motor speed and wing tilt angle, maintains the stability of the wing flapping trajectory in complex airflow, invokes a preset dynamic compensation model based on the real-time wind field vector, corrects the original control commands, and outputs a final execution signal to the motor drive module and wing servo mechanism of the UAV, enabling the UAV to adaptively perform wind resistance control, reduce the UAV attitude error, provide a highly robust and high-precision technical guarantee for the intelligent inspection of railway catenaries, and greatly reduce the risk of manual inspection and operation and maintenance costs.
[0021] In a preferred example of the present application, it can be further configured that: the marking time stamps for the collected UAV flight environment data, and synchronizing and aligning multi-source data through time stamps to generate synchronized time-series multi-source data specifically includes:
[0022] Add a millisecond-level hardware timestamp to each sensor in the sensing layer, and use a time synchronization algorithm based on the PTP protocol to align the multi-source data collected by each sensor to form an initial multi-source data packet;
[0023] Verify the timing consistency of the initial multi-source data packet within the time window, identify abnormal data points, and output the synchronized timing multi-source data after alignment.
[0024] By adopting the above technical solution, a millisecond-level hardware timestamp is added to each sensor, and combined with the PTP synchronization algorithm, the multi-source data is forced to be aligned within a ±0.5 ms time window, eliminating the phase error caused by traditional asynchronous acquisition. For example, the timing misalignment between the angular velocity and air pressure data is compressed from an average of 20 ms to within 0.5 ms, avoiding attitude misjudgment caused by data mismatch during sudden air flow changes. Based on a sliding time window (such as a 100 ms window), the logical relevance of each sensor in the data packet is verified in real time (such as the air pressure difference needs to respond synchronously when the wind speed suddenly changes), automatically identifying and removing abnormal points caused by electromagnetic interference and sensor transient failures, preventing incorrect data from contaminating control instructions. Through hardware-level timestamp marking and strict timing verification, the control accuracy and system reliability of the UAV during high-voltage catenary inspection are significantly improved.
[0025] In a preferred example, the present application can be further configured as: inputting the synchronized timing multi-source data into a preset spatio-temporal fusion neural network to generate an original control instruction including the motor speed and wing inclination angle, specifically including:
[0026] Input the synchronized timing multi-source data into a three-layer spatio-temporal fusion neural network to extract timing features;
[0027] Based on the attention mechanism layer, weight the hidden states of key time steps to generate a context vector, input the context vector into the fully connected layer of the spatio-temporal fusion neural network, and output a multi-dimensional control quantity;
[0028] Normalize the multi-dimensional control quantity to generate an original control instruction including the motor speed and wing inclination angle.
[0029] By adopting the above technical solution, the three-layer LSTM network deeply captures the dynamic evolution laws of multi-dimensional data such as wind speed, air pressure, and angular velocity (such as the correlation between gust duration and wing torque). The attention mechanism layer dynamically allocates weights to enhance the response priority to key events such as airflow mutations and electric field pulses, avoiding the oscillation risk caused by attitude control lag, fusing the weighted hidden states into a context vector, accurately coupling physical sensing data (such as IMU angular velocity) with biological excitation features (avian wing flapping patterns), improving the control generalization ability in complex environments, outputting multi-dimensional control quantities, normalizing the multi-dimensional control quantities through the Sigmoid function, generating original control instructions including motor speed and wing tilt angle, ensuring that the motor speed instruction and wing tilt angle instruction are strictly limited within the hardware safety threshold, and preventing overload damage. Through the collaborative optimization of the three-layer spatio-temporal fusion network and the attention mechanism, the dynamic response accuracy and environmental adaptability of the UAV in high-voltage catenary inspection are significantly improved.
[0030] In a preferred example of the present application, it can be further configured that: the preset dynamic compensation model is called based on the real-time wind field vector to correct the original control instruction and output the final execution signal to the motor drive module and wing servo mechanism of the UAV, specifically including:
[0031] Call the pre-trained wind disturbance resistance model, input the real-time wind field vector into the wind disturbance resistance model, and generate a feedforward compensation amount;
[0032] Obtain a deviation feedback compensation amount based on the feedforward compensation amount and the preset target attitude threshold;
[0033] Superimpose the feedforward compensation amount and the deviation feedback compensation amount according to the preset weight ratio to generate a correction instruction, and send the correction instruction to the brushless motor drive module and the wing servo actuator.
[0034] By adopting the above technical solution, call the pre-trained wind disturbance resistance model, input the real-time wind field vector into the model to generate a feedforward compensation amount, anticipate the dynamic effects of gusts and crosswinds in advance, generate a feedback compensation amount based on the deviation between the feedforward compensation amount and the target attitude threshold, correct the attitude drift in real time, and fuse the feedforward and feedback compensation amounts through the weight ratio to generate a correction instruction, and send the correction instruction to the brushless motor drive module and the wing servo actuator. Through the feedforward-feedback composite control and weight optimization strategy, the anti-interference ability and control accuracy of the UAV in the complex airflow environment of the high-voltage catenary are significantly improved.
[0035] In a preferred example of the present application, it can be further configured that: after calling the pre-trained wind disturbance resistance model, inputting the real-time wind field vector into the wind disturbance resistance model, and generating a feedforward compensation amount, the wing control method of the catenary inspection UAV further includes:
[0036] Collect the rotational speed feedback, attitude sensor data, and control error of the wing in real time, and calculate the anti-wind disturbance model error based on a composite function;
[0037] Update the anti-wind disturbance model parameters according to the anti-wind disturbance model error to adapt to real-time environmental changes.
[0038] By adopting the above technical solution, synchronously obtain the rotational speed feedback, attitude sensor data, and control error of the wing, construct a high-confidence optimization data set, calculate the anti-wind disturbance model error based on a composite function, balance accuracy and stability, and according to the anti-wind disturbance model error, use an incremental learning strategy to fine-tune the anti-wind disturbance model. Through real-time data closed-loop and dynamic parameter update, the environmental adaptability and control accuracy of the UAV in long-term inspection are significantly improved.
[0039] In a preferred example of the present application, it can be further configured as follows: after calling a preset dynamic compensation model based on the real-time wind field vector, correcting the original control instruction, and outputting the final execution signal to the motor drive module and wing servo mechanism of the UAV, the wing control method of the catenary inspection UAV further includes:
[0040] Divide the wind field into levels according to the real-time wind field vector, and generate a hierarchical anti-wind response strategy based on the wind field level;
[0041] Enable the PID fine-tuning strategy under a light wind field and pre-start the standby motor;
[0042] Activate the spatio-temporal fusion neural network reconstruction control law under a moderate wind field and deploy the anti-wind ailerons of the wing;
[0043] Switch to the fault-tolerant control mode under a severe wind field and activate the mooring device to lock the position of the UAV.
[0044] By adopting the above technical solution, based on the three-dimensional dynamic modeling of the real-time wind field vector (such as turbulence intensity and gust frequency), divide the wind field into levels, including light wind field, moderate wind field, and severe wind field, generate a hierarchical anti-wind response strategy according to different wind field levels, enable the PID fine-tuning strategy under a light wind field and pre-start the standby motor, activate the spatio-temporal fusion neural network to generate anti-wind instructions in real time under a moderate wind field, deploy the anti-wind ailerons of the wing, increase the lift area, and offset the gust torque. Under a severe wind field, switch to the fault-tolerant control mode, activate the mooring device to lock the position of the UAV, and forcefully lock the position of the UAV to completely avoid the risk of hitting the catenary. Through the coordination of dynamic wind field level division and targeted control strategies, the safety, stability, and energy efficiency of the UAV in the complex wind field environment of high-voltage catenary are significantly improved.
[0045] In summary, the present application includes at least one of the following beneficial technical effects:
[0046] 1. The drone wing, through an innovative hierarchical composite structure design (core layer, functional layer, sensing layer), significantly enhances its safety, stability, and reliability in the high-voltage railway catenary environment. The core layer uses a carbon fiber skeleton, which achieves lightweight while ensuring the support of the aerodynamic shape, reduces flight energy consumption, and extends the endurance time. The anti-high-voltage coating of the functional layer is composed of three layers in composite, which raises the breakdown voltage to ≥30 kV / mm, effectively resists the interference of the 25 kV high-voltage electric field, and avoids the risk of circuit breakdown. The anti-static coating sprays a carbon nanotube (CNT) grid (grounding resistance < 0.1 Ω) on the edge of the wing, which can quickly dissipate the accumulated charge and eliminate the problem of sensor false alarms or hardware damage caused by static electricity. The MEMS pressure sensor of the sensing layer is arranged by avoiding the coating area to ensure the sensitivity of the air pressure data, providing accurate input for attitude control, enabling the drone to conduct safe inspections under the 25 kV electric field, and being able to adaptively resist wind control, reducing the attitude error of the drone;
[0047] 2. The anti-static coating adopts a carbon nanotube grid structure formed by spraying. The high conductivity of the carbon nanotubes combined with the low-resistance grounding path can dissipate the accumulated charge within 0.1 second, effectively eliminating the static interference caused by the high-voltage electric field (25 kV), and avoiding the risks of sensor data drift and electronic component breakdown. The grid structure is sprayed on the edge of the wing, specifically blocking the charge conduction path while avoiding covering the sensor area to ensure the accuracy of the sensing data;
[0048] 3. Through the high-precision sensors integrated in the sensing layer, three-dimensional wind field vectors, air pressure differences, angular velocities, and ground relative velocity data during the flight of the drone are comprehensively captured, accurately quantifying the airflow disturbances (such as gusts and crosswinds) and spatial electric field interference, providing a data basis for wind resistance control and insulation protection. Timestamp tags are added to the collected drone flight environment data, and multi-source data are aligned through timestamp synchronization to solve the phase error caused by the difference in sensor sampling rates, avoid control instruction lag, generate high-consistency time-series data packets, provide high-quality input for the neural network, reduce the drone wing attitude estimation error. The synchronized time-series multi-source data are input into a preset spatio-temporal fusion neural network. The spatio-temporal fusion neural network captures long-term time dependencies, combines the attention mechanism to focus on key wind speed mutation moments, fuses bio-inspired features, generates original control instructions including motor speed and wing inclination angle, maintains the stability of the wing flapping trajectory in complex airflow, calls a preset dynamic compensation model based on the real-time wind field vector to correct the original control instructions, and outputs the final execution signal to the motor drive module and wing servo mechanism of the drone, enabling the drone to adaptively resist wind control, reducing the attitude error of the drone, providing a highly robust and accurate technical guarantee for the intelligent inspection of railway catenaries, and significantly reducing the risks of manual inspection and operation and maintenance costs;
[0049] 4. Synchronously obtain the wing rotation speed feedback, attitude sensor data, and control error, construct a high-confidence optimization data set, calculate the anti-wind disturbance model error based on a composite function, balance accuracy and stability, and fine-tune the anti-wind disturbance model according to the anti-wind disturbance model error. Through real-time data closed-loop and dynamic parameter update, the environmental adaptability and control accuracy of the UAV during long-term inspection are significantly improved. Brief Description of the Drawings
[0050] Figure 1 is the wing structure diagram of the catenary inspection UAV in an embodiment of the present application;
[0051] Figure 2 is the implementation flowchart of the wing control method of the catenary inspection UAV in an embodiment of the present application;
[0052] Figure 3 is the implementation flowchart of step S20 in the wing control method of the catenary inspection UAV in an embodiment of the present application;
[0053] Figure 4 is the implementation flowchart of step S30 in the wing control method of the catenary inspection UAV in an embodiment of the present application;
[0054] Figure 5 is the implementation flowchart of step S40 in the wing control method of the catenary inspection UAV in an embodiment of the present application;
[0055] Figure 6 is another implementation flowchart of the wing control method of the catenary inspection UAV in an embodiment of the present application;
[0056] Figure 7 is another implementation flowchart of the wing control method of the catenary inspection UAV in an embodiment of the present application.
[0057] Reference Signs: 1, wing body; 2, motor; 3, electronic speed controller; 4, connecting wire; 5, high-voltage protection coating; 6, anti-static coating; 7, servo steering gear. Detailed Description of the Embodiment
[0058] The following further elaborates on the present application with reference to the accompanying drawings.
[0059] As Figure 1As shown, the present application discloses a contact network inspection drone wing, including a wing body 1, a motor 2 and an electric regulator 3, the wing body 1 is symmetrically arranged on the motor 2, the motor 2 is connected to the electric regulator 3, the electric regulator 3 is connected to the drone, a layered composite structure and a servo actuator 7 are arranged on the wing body 1, a connecting line 4 is arranged under the wing body 1, the layered composite structure includes a core layer, a functional layer and a sensor layer, the core layer is composed of a carbon fiber skeleton, which is used to provide aerodynamic shape support for the wing, the functional layer is covered on the surface of the core layer, the functional layer includes an anti-high voltage coating 5 and an anti-static coating 6, the anti-high voltage coating 5 is a multi-layer composite insulating coating, the anti-static coating 6 is arranged in the edge area of the wing body 1, the anti-static coating 6 includes a grid structure formed by a conductive material, the conductive material grid structure in the anti-static coating 6 is a carbon nanotube grid formed by spraying, the grid has a grounding path, and its resistance is less than 0.1 ohm, the sensor layer is embedded in the surface of the wing body 1, the sensor layer includes at least one micro-electromechanical system air pressure sensor, and the arrangement area of the sensor layer avoids the coating coverage area of the functional layer to ensure the sensitivity of the sensor.
[0060] In one embodiment, if Figure 2 As shown, the present application discloses a method for controlling the wing of a contact network inspection drone, which specifically includes the following steps:
[0061] S10: Synchronously collect UAV flight environment data based on the sensor layer, wherein the UAV flight environment data includes three-dimensional wind field vector, air pressure difference, angular velocity and ground relative velocity data.
[0062] Specifically, through the high-precision sensors integrated in the sensing layer, the three-dimensional wind field vector, air pressure difference, angular velocity and ground relative speed data of the UAV during flight are fully captured, and the airflow disturbance (such as gusts and crosswinds) and spatial electric field interference are accurately quantified, providing a data basis for wind resistance control and insulation protection.
[0063] S20: Timestamp the collected UAV flight environment data, and synchronize and align multi-source data through timestamps to generate synchronized time series multi-source data.
[0064] Specifically, the collected UAV flight environment data is timestamped, and multi-source data is aligned through timestamp synchronization to resolve phase errors caused by differences in sensor sampling rates, avoid control command lags, generate highly consistent timing data packets, provide high-quality input for the neural network, and reduce the error in UAV wing attitude estimation.
[0065] S30: Inputting the synchronized time-series multi-source data into a preset spatiotemporal fusion neural network to generate original control instructions including motor speed and wing inclination angle.
[0066] Specifically, the synchronized time-series multi-source data is input into a preset spatio-temporal fusion neural network. The spatio-temporal fusion neural network captures long-term time-series dependencies, combines the attention mechanism to focus on key wind speed mutation moments, and fuses bio-inspired features to generate an original control command containing the motor speed and wing inclination angle, so as to maintain the stability of the wing flapping trajectory in complex airflows.
[0067] S40: Based on the real-time wind field vector, call a preset dynamic compensation model to correct the original control command, and output the final execution signal to the motor drive module and wing servo mechanism of the drone.
[0068] Specifically, based on the real-time wind field vector, call a preset dynamic compensation model to correct the original control command, and output the final execution signal to the motor drive module and wing servo mechanism of the drone, so that the drone can perform adaptive anti-wind control and reduce the attitude error of the drone.
[0069] In this embodiment, through the high-precision sensors integrated in the sensing layer, three-dimensional wind field vectors, air pressure differences, angular velocities, and ground relative velocity data during the flight of the drone are comprehensively captured, and air flow disturbances (such as gusts and crosswinds) and space electric field interference are accurately quantified, providing a data basis for anti-wind control and insulation protection. Timestamp tags are added to the collected drone flight environment data, and multi-source data is aligned through timestamp synchronization to solve the phase error caused by different sensor sampling rates, avoid control command lag, generate high-consistency time-series data packets, provide high-quality inputs for the neural network, reduce the wing attitude estimation error of the drone, input the synchronized time-series multi-source data into a preset spatio-temporal fusion neural network, the spatio-temporal fusion neural network captures long-term time-series dependencies, combines the attention mechanism to focus on key wind speed mutation moments, and fuses bio-inspired features to generate an original control command containing the motor speed and wing inclination angle, so as to maintain the stability of the wing flapping trajectory in complex airflows, based on the real-time wind field vector, call a preset dynamic compensation model to correct the original control command, and output the final execution signal to the motor drive module and wing servo mechanism of the drone, so that the drone can perform adaptive anti-wind control and reduce the attitude error of the drone, providing a highly robust and accurate technical guarantee for the intelligent inspection of railway catenaries, and greatly reducing the risks and operation and maintenance costs of manual inspections.
[0070] In one embodiment, as Figure 2 shown, in step S20, that is, timestamp tags are added to the collected drone flight environment data, and multi-source data is aligned through timestamp synchronization to generate synchronized time-series multi-source data, which specifically includes:
[0071] S21: Add millisecond-level hardware timestamp tags to each sensor in the sensing layer, and use a time synchronization algorithm based on the PTP protocol to align the multi-source data collected by each sensor to form an initial multi-source data packet.
[0072] Specifically, add millisecond-level hardware timestamps to each sensor, and combine with the PTP synchronization algorithm to force multi-source data to align within a ±0.5 ms time window, eliminating the phase error caused by traditional asynchronous acquisition. For example, the timing misalignment between angular velocity and barometric pressure data is compressed from an average of 20 ms to within 0.5 ms, avoiding attitude misjudgment caused by data mismatch during sudden airflow changes.
[0073] S22: Verify the timing consistency of the initial multi-source data packet within the time window, identify abnormal data points, and output the synchronized timing multi-source data after alignment.
[0074] Specifically, based on a sliding time window (such as a 100 ms window), real-time verify the logical relevance of each sensor within the data packet (such as the barometric pressure difference needs to respond synchronously when the wind speed changes suddenly), automatically identify and eliminate abnormal points caused by electromagnetic interference and sensor transient failures, prevent incorrect data from contaminating control commands, and significantly improve the control accuracy and system reliability of the drone during high-voltage catenary inspection through hardware-level timestamp marking and strict timing verification.
[0075] In one embodiment, as Figure 3 shown, in step S30, input the synchronized timing multi-source data into a preset spatio-temporal fusion neural network to generate an original control command including motor speed and wing inclination angle, specifically including:
[0076] S31: Input the synchronized timing multi-source data into a three-layer spatio-temporal fusion neural network to extract timing features.
[0077] S32: Based on the attention mechanism layer, weight the hidden states of key time steps to generate a context vector, input the context vector into the fully connected layer of the spatio-temporal fusion neural network, and output a multi-dimensional control quantity.
[0078] S33: Normalize the multi-dimensional control quantity to generate an original control command including motor speed and wing inclination angle.
[0079] Specifically, the three-layer LSTM network deeply captures the dynamic evolution laws of multi-dimensional data such as wind speed, air pressure, and angular velocity (such as the correlation between gust duration and wing torque). The attention mechanism layer dynamically allocates weights to enhance the response priority of key events such as airflow mutations and electric field pulses, avoiding the oscillation risk caused by delayed attitude control. It fuses the weighted hidden states into a context vector, precisely couples physical sensing data (such as IMU angular velocity) with biological excitation features (avian wing flapping patterns), improves the control generalization ability in complex environments, outputs multi-dimensional control quantities, normalizes the multi-dimensional control quantities through the Sigmoid function, generates original control instructions including motor speed and wing inclination, ensures that the motor speed instruction and wing inclination instruction are strictly limited within the hardware safety threshold, and prevents overload damage. Through the collaborative optimization of the three-layer spatio-temporal fusion network and the attention mechanism, the dynamic response accuracy and environmental adaptability of the UAV in high-voltage catenary inspection are significantly improved.
[0080] In one embodiment, as Figure 4 shown, in step S40, that is, based on the real-time wind field vector, a preset dynamic compensation model is called to correct the original control instruction and output the final execution signal to the motor drive module and wing servo mechanism of the UAV, which specifically includes:
[0081] S41: Call the pre-trained wind disturbance resistance model, input the real-time wind field vector into the wind disturbance resistance model, and generate a feedforward compensation amount.
[0082] S42: Obtain a deviation feedback compensation amount based on the feedforward compensation amount and the preset target attitude threshold.
[0083] S43: Superimpose the feedforward compensation amount and the deviation feedback compensation amount according to the preset weight ratio to generate a correction instruction, and send the correction instruction to the brushless motor drive module and wing servo actuator.
[0084] Specifically, call the pre-trained wind disturbance resistance model, input the real-time wind field vector into the model to generate a feedforward compensation amount, anticipate the dynamic effects of gusts and crosswinds in advance, generate a feedback compensation amount based on the deviation between the feedforward compensation amount and the target attitude threshold, correct the attitude drift in real time, and fuse the feedforward and feedback compensation amounts with the weight ratio to generate a correction instruction, and send the correction instruction to the brushless motor drive module and wing servo actuator. Through the feedforward-feedback composite control and weight optimization strategy, the anti-interference ability and control accuracy of the UAV in the complex airflow environment of high-voltage catenary are significantly improved.
[0085] In one embodiment, as Figure 5 shown, after step S41, the wing control method of the catenary inspection UAV further includes:
[0086] S401: Collect the rotational speed feedback of the wing, attitude sensor data, and control error in real time, and calculate the anti-wind disturbance model error based on a composite function.
[0087] S402: Update the anti-wind disturbance model parameters according to the anti-wind disturbance model error to adapt to real-time environmental changes.
[0088] Specifically, synchronously obtain the rotational speed feedback of the wing, attitude sensor data, and control error, construct a high-confidence optimization data set, calculate the anti-wind disturbance model error based on a composite function, balance accuracy and stability, and according to the anti-wind disturbance model error, use an incremental learning strategy to fine-tune the anti-wind disturbance model. Through real-time data closed-loop and dynamic parameter update, the environmental adaptability and control accuracy of the UAV in long-term inspection are significantly improved.
[0089] In one embodiment, as Figure 6 shown, after step S40, that is, after calling a preset dynamic compensation model based on the real-time wind field vector, correcting the original control instruction, and outputting the final execution signal to the motor drive module and wing servo mechanism of the UAV, the wing control method of the catenary inspection UAV further includes:
[0090] S50: Divide the wind field into levels according to the real-time wind field vector, and generate a hierarchical anti-wind response strategy based on the wind field levels.
[0091] Specifically, based on the three-dimensional dynamic modeling of the real-time wind field vector (such as turbulence intensity and gust frequency), divide the wind field into levels, including a light wind field, a moderate wind field, and a severe wind field, and generate a hierarchical anti-wind response strategy according to different wind field levels.
[0092] S60: Enable the PID fine-tuning strategy in the light wind field and pre-start the standby motor.
[0093] S70: Activate the spatio-temporal fusion neural network reconstruction control law in the moderate wind field and deploy the wing anti-wind ailerons.
[0094] S80: Switch to the fault-tolerant control mode in the severe wind field and activate the tethering device to lock the UAV position.
[0095] Specifically, enable the PID fine-tuning strategy in the light wind field and pre-start the standby motor. In the moderate wind field, activate the spatio-temporal fusion neural network to generate anti-wind instructions in real time, deploy the wing anti-wind ailerons, increase the lift area, and offset the gust torque. In the severe wind field, switch to the fault-tolerant control mode, activate the tethering device to lock the UAV position, and forcefully lock the UAV position to completely avoid the risk of hitting the catenary. Through the coordination of dynamic wind field level division and targeted control strategies, the safety, stability, and energy efficiency of the UAV in the complex wind field environment of high-voltage catenaries are significantly improved.
[0096] It should be understood that the sequence numbers of the steps in the above embodiments do not imply the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.
[0097] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.
Claims
1. An inspection unmanned aerial vehicle wing for an overhead contact line, characterized in that, The invention comprises a wing body (1), a motor (2) and an electric regulator (3), wherein the wing body (1) is symmetrically arranged on the motor (2), the motor (2) is connected to the electric regulator (3), and the electric regulator (3) is connected to a drone, and a layered composite structure is arranged on the wing body (1), wherein the layered composite structure comprises a core layer, a functional layer and a sensor layer, wherein the core layer is composed of a carbon fiber skeleton and is used to provide aerodynamic shape support for the wing, the functional layer is covered on the surface of the core layer, and the functional layer comprises an anti-high voltage coating (5) and an anti-static coating (6), wherein the anti-high voltage coating (5) is a multi-layer composite insulating coating, the anti-static coating (6) is arranged on the edge area of the wing body (1), and the anti-static coating (6) comprises a grid structure formed by a conductive material, the sensor layer is embedded in the surface of the wing body (1), the sensor layer comprises at least one micro-electromechanical system air pressure sensor, and the arrangement area of the sensor layer avoids the coating coverage area of the functional layer to ensure the sensitivity of the sensor.
2. The wing of an overhead line inspection drone according to claim 1, wherein, The conductive material grid structure in the antistatic coating (6) is a carbon nanotube grid formed by spraying, and the grid has a grounding path with a resistance of less than 0.1 ohm.
3. A control method for the wings of an overhead catenary inspection drone, characterized in that, The contact network inspection drone wing control method comprises the following steps: Based on the sensor layer, the UAV flight environment data is synchronously collected, and the UAV flight environment data includes three-dimensional wind field vector, air pressure difference, angular velocity and ground relative velocity data; The collected UAV flight environment data is timestamped, and multi-source data is aligned through timestamp synchronization to generate synchronized time-series multi-source data; Inputting the synchronized time-series multi-source data into a preset spatiotemporal fusion neural network to generate original control instructions including motor speed and wing inclination angle; The preset dynamic compensation model is called based on the real-time wind field vector, the original control instructions are corrected, and the final execution signal is output to the motor drive module and wing servo mechanism of the UAV.
4. The catenary inspection UAV wing control method according to claim 3, wherein The collected UAV flight environment data is timestamped, and multi-source data is aligned by timestamping to generate synchronized time series multi-source data, specifically including: Add millisecond-level hardware timestamps to each sensor in the sensing layer, and use a time synchronization algorithm based on the PTP protocol to align the multi-source data collected by each sensor to form an initial multi-source data packet; Verify the timing consistency of the initial multi-source data packet within the time window, propose abnormal data points, and output aligned synchronous timing multi-source data.
5. The method for controlling the wings of an OHL inspection UAV according to claim 3, characterized in that, The synchronized time-series multi-source data is input into a preset spatiotemporal fusion neural network to generate original control instructions including motor speed and wing inclination angle, specifically including: The synchronized time series multi-source data is input into the three-layer spatiotemporal fusion neural network to extract the time series features; Based on the hidden state of the weighted key time step of the attention mechanism layer, a context vector is generated, the context vector is input into the fully connected layer of the spatiotemporal fusion neural network, and a multi-dimensional control quantity is output; The multi-dimensional control quantity is normalized to generate an original control instruction including the motor speed and the wing inclination angle.
6. The method for controlling the wing of an overhead line inspection drone according to claim 3, wherein The above-mentioned method calls a preset dynamic compensation model based on the real-time wind field vector, corrects the original control command, and outputs the final execution signal to the motor drive module and wing servo mechanism of the drone, which specifically includes: Call a pre-trained wind disturbance resistance model, input the real-time wind field vector into the wind disturbance resistance model, and generate a feedforward compensation amount; Obtain a deviation feedback compensation amount based on the feedforward compensation amount and a preset target attitude threshold; Superimpose the feedforward compensation amount and the deviation feedback compensation amount according to a preset weight ratio to generate a correction command, and send the correction command to the brushless motor drive module and the wing servo actuator.
7. The catenary inspection drone wing control method according to claim 6, characterized in that After calling the pre-trained wind disturbance resistance model, inputting the real-time wind field vector into the wind disturbance resistance model, and generating a feedforward compensation amount, the wing control method of the catenary inspection drone further includes: Real-time collect the rotational speed feedback of the wing, attitude sensor data and control error, and calculate the wind disturbance resistance model error based on a composite function; Update the parameters of the wind disturbance resistance model according to the wind disturbance resistance model error to adapt to the real-time environmental changes.
8. The method for controlling the wing of an OHL inspection UAV according to claim 3, characterized in that, After calling the preset dynamic compensation model based on the real-time wind field vector, correcting the original control command, and outputting the final execution signal to the motor drive module and wing servo mechanism of the drone, the wing control method of the catenary inspection drone further includes: Divide the wind field level according to the real-time wind field vector, and generate a hierarchical wind resistance response strategy based on the wind field level; Enable the PID fine-tuning strategy under a light wind field and pre-start the standby motor; Activate the spatio-temporal fusion neural network reconstruction control law under a moderate wind field and deploy the wind resistance ailerons of the wing; Switch to the fault-tolerant control mode under a severe wind field and start the tethering device to lock the position of the drone.