Electric power unmanned aerial vehicle self-inspection system

Abstract

The invention, which relates to the technical field of the electric unmanned aerial vehicle, discloses an electric unmanned aerial vehicle self-inspection system comprising the following components: a terahertz transmitting module, a terahertz receiving module, an image acquisition and processing module, a data analysis and fault diagnosis module, and a display and alarm module. By adopting the terahertz transmitting and receiving module and combining with an image acquisition and processing algorithm, accurate detection of the internal structure of the unmanned aerial vehicle can be realized, and terahertz waves have unique penetrability and sensitivity to material structures, so that the system can accurately identify the states of electronic components and circuit board components, and the detection accuracy is improved. According to the method, the defects of welding spot looseness and tiny line fracture are detected, meanwhile, through a deep neural network algorithm fusing multiple features, the system can intelligently analyze terahertz images, the accuracy of fault diagnosis is further improved, in addition, the image reconstruction process is accelerated through an improved iterative reconstruction algorithm and a self-adaptive step length adjustment mechanism, and the detection efficiency is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power drones, and particularly to a power drone self-check system. Background Art

[0002] With the wide application of power drones in the fields of power inspection and line maintenance, their operation safety and reliability have become the focus of the industry. When a power drone performs tasks, the states of its internal electronic components, circuit board components and key components directly determine the performance and safety of the drone. However, since drones often need to operate in complex and changeable environments, their internal structures are prone to failures due to various factors, such as loose solder joints and fine circuit breaks. These minor defects are often difficult to detect in time by the naked eye or conventional detection means, posing potential hazards to the safe flight of the drone.

[0003] Traditional technologies have deficiencies. On the one hand, the emission and reception of terahertz waves require high-precision equipment and complex processing technologies, which pose high requirements for the stability and reliability of the system. On the other hand, the internal structure of the drone is complex, and different materials and components have large differences in the response to terahertz waves, which increases the difficulty of image reconstruction and fault diagnosis.

[0004] In summary, although the existing non-destructive self-check methods for the internal structure of power drones based on terahertz technology have certain advantages, there are still many deficiencies in practical applications. Therefore, it is particularly important to develop a power drone self-check system. Summary of the Invention

[0005] The purpose of the present invention is to make up for the deficiencies of the existing technology, and provide a power drone self-check system. It can achieve precise detection of the internal structure of the drone by adopting a terahertz emission and reception module and combining an image acquisition and processing algorithm. At the same time, through a deep neural network algorithm that fuses multiple features, the system can intelligently analyze terahertz images and further improve the accuracy of fault diagnosis.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: A power drone self-check system, which includes the following components: a terahertz emission module, a terahertz reception module, an image acquisition and processing module, a data analysis and fault diagnosis module, and a display and alarm module;

[0007] The terahertz emission module: generates terahertz waves with a frequency range between 0.1 THz and 10 THz. This module is designed based on a quantum cascade structure, and generates terahertz waves with specific frequencies by precisely controlling the energy level transitions of electrons in the quantum well. The output frequency f of the terahertz wave is determined by the following formula:

[0008]

[0009] Among them, ΔE is the energy level difference of electron transition in the quantum well, which is jointly determined by the material properties of the quantum well, the well width w, the well depth d, and the externally applied electric field strength E ext The specific relationship is ΔE = α·w + β·d + γ·E ext , where α, β, and γ are coefficients obtained through a combination of experiments and theoretical calculations based on the material properties of the quantum well, and h is Planck's constant. By adjusting the values of w, d, and E ext , precise control of the terahertz wave frequency can be achieved;

[0010] The terahertz receiving module: is located at multiple specific positions inside the drone, and is used to receive the terahertz wave that has penetrated the internal structure of the drone and convert it into an electrical signal for amplification and preprocessing. This module adopts a detector structure based on a new semiconductor material, and its responsivity R is determined by the following formula:

[0011]

[0012] Among them, I ph is the photocurrent, which is related to the number of terahertz photons N ph absorbed by the detector, the quantum efficiency η of the detector, and the electron charge e, that is, I ph = η·N ph ·e; P thz is the power of the incident terahertz wave. By optimizing the material structure and process parameters of the detector, the quantum efficiency η is improved, thereby increasing the responsivity of the receiving module;

[0013] The image acquisition and processing module: is connected to the terahertz receiving module, receives the preprocessed electrical signal, and reconstructs the terahertz image of the internal structure of the drone through an iterative reconstruction algorithm. This algorithm is based on the principle of maximum likelihood estimation and approximates the real image by iteratively updating the pixel values of the image. The specific iterative formula is as follows:

[0014]

[0015] Among them, is the value of the i-th pixel at the k-th iteration, P j is the j-th measurement data received by the terahertz receiving module, H ij is the measurement matrix, which represents the contribution weight of the i-th pixel to the j-th measurement data. This weight is determined by the geometric shape of the internal structure of the drone, the terahertz wave propagation path, and the detector position factors. Through multiple iterations, the error of the reconstructed image is gradually reduced, and finally a clear terahertz image is obtained, and the image is enhanced and noise-reduced based on adaptive filtering;

[0016] The data analysis and fault diagnosis module: Analyzes the reconstructed terahertz image, uses a deep neural network algorithm that fuses multiple features to identify the states of electronic components and circuit board components in the image, and detects whether there are minor defects such as loose solder joints and fine circuit breaks. This neural network model consists of an input layer, multiple hidden layers, and an output layer. The input layer receives terahertz image data. After feature extraction and non-linear transformation by the hidden layers, the fault diagnosis result is output at the output layer. The feature extraction of the hidden layer is based on the following convolution operation:

[0017]

[0018] Among them, is the feature map element at the l-th layer, m-th row, and n-th column, is the convolution kernel weight of the l-th layer, which is obtained by training a large number of normal and faulty terahertz image samples through the backpropagation algorithm, is the input feature map element of the (l - 1)-th layer, F is the size of the convolution kernel, and b l is the bias term of the l-th layer;

[0019] The display and alarm module: Displays the detection results of the data analysis and fault diagnosis module. This module is connected to the on-board display of the drone or the communication interface of the ground control station to display the detection results in an intuitive graphical interface. When a fault is detected inside the drone, an audible and visual alarm signal is triggered through pre-set fault codes and alarm rules.

[0020] Furthermore, the material of the quantum well in the terahertz emission module adopts a multi-layer heterostructure based on III-V group compound semiconductors. The well width w and well depth d of the quantum well are precisely controlled through molecular beam epitaxy technology. During the growth process, through real-time monitoring and feedback control technology, the accuracy of the well width w is ensured to reach the sub-nanometer level, and the accuracy of the well depth d reaches within 1%. At the same time, the externally applied electric field strength E ext is regulated through the power supply module. The output accuracy of the power supply module is ±0.1V. Through these precise material preparation and parameter control technologies, the stability and controllability of the terahertz wave frequency are further improved, enabling the terahertz wave to penetrate different materials inside the drone more accurately and providing a more reliable signal source for subsequent detection.

[0021] Furthermore, the novel semiconductor material in the terahertz receiving module adopts a structure based on the composite of two-dimensional materials and traditional semiconductors. This composite structure grows two-dimensional materials with atomic-level thickness on the surface of traditional semiconductors, and utilizes the unique optoelectronic properties of two-dimensional materials to improve the quantum efficiency η of the detector. The selection of two-dimensional materials is based on their band structures and carrier mobilities. Through a method combining theoretical calculations and experimental tests, the optimal combination of two-dimensional materials and traditional semiconductors is determined. During the preparation process, chemical vapor deposition technology is used to uniformly grow two-dimensional materials on the surface of traditional semiconductors, and the thickness and quality of two-dimensional materials are precisely controlled by controlling deposition temperature and gas flow parameters. Through this composite structure, the quantum efficiency η of the detector is increased by at least 30% compared with traditional detectors, thereby significantly improving the responsivity and sensitivity of the terahertz receiving module, enabling it to receive weak terahertz signals more accurately and providing richer information for subsequent image reconstruction.

[0022] Furthermore, in the improved iterative reconstruction algorithm of the image acquisition and processing module, in order to accelerate the iterative convergence speed, an adaptive step size adjustment mechanism is introduced. During each iteration process, according to the error and convergence trend of the current image reconstruction, the iterative step size λ is dynamically adjusted. The error calculation formula is:

[0023]

[0024] where, is the data actually measured by the terahertz receiving module, is the corresponding data obtained from the current iterative reconstruction, N is the total number of measurement data. According to the magnitude of the error E, the step size λ is adjusted through the following formula:

[0025] λ = λ 0 ·exp(-α·E)

[0026] where, λ 0 is the initial step size, α is the step size adjustment coefficient determined according to experiments. When the error E is large, the step size λ is increased to accelerate the convergence speed. When the error E is small, the step size λ is decreased to improve the reconstruction accuracy. Through this adaptive step size adjustment mechanism, on the premise of ensuring the quality of the reconstructed image, the iterative convergence speed is increased by at least 50%, greatly shortening the image reconstruction time and improving the real-time performance of the system.

[0027] Furthermore, in the deep neural network algorithm with multi-feature fusion of the data analysis and fault diagnosis module, in order to better fuse the texture, shape, and grayscale multi-feature information of the image, a feature fusion layer is introduced into the network. The feature fusion layer adopts a fusion method based on the attention mechanism. By learning the importance weights of different features, different features are dynamically fused. For the input texture feature T, shape feature S, and grayscale feature G, they are first mapped to the same dimension through a linear transformation to obtain T ′ 、S ′ and G ′ , and then the attention weights of each feature are calculated:

[0028]

[0029] where ω T 、ω S and ω G are weight parameters obtained through network training. Finally, the fused feature F is:

[0030] F = A T ·T′ + A S ·S′ + A G ·G′

[0031] Through this feature fusion method based on the attention mechanism, the importance of different features in fault diagnosis can be more effectively highlighted, improving the accuracy and robustness of fault diagnosis.

[0032] Furthermore, in the display and alarm module, in order to achieve a more intuitive and clear display of detection results, augmented reality (AR) technology is adopted. On the on-board display screen or the display of the ground control station, the terahertz image reconstruction result is fused with the 3D model of the UAV through AR technology. The attitude and position information of the UAV are obtained through the UAV's inertial measurement unit IMU and global positioning system GPS. Then, the fault position information obtained by reconstructing the terahertz image is mapped onto the 3D model of the UAV. When displaying, the AR engine is used to fuse the 3D model with the fault label in real time with the actual scene. The operator can operate the display content through gesture interaction or voice commands. At the same time, for the alarm signal, a combination of voice and vision is adopted, and the type and position of the fault are prompted to the operator in a prominent color and animation effect in the AR scene, enabling the operator to more quickly and accurately understand the fault situation of the UAV and improving the efficiency of fault handling.

[0033] Furthermore, high-speed and low-latency fiber-optic communication technology is used for data transmission between the terahertz emission module, terahertz reception module, image acquisition and processing module, data analysis and fault diagnosis module, and display and alarm module. The fiber-optic communication system adopts wavelength-division multiplexing technology to modulate the signals transmitted between different modules to different wavelengths λ sig . Multiple signals are transmitted simultaneously in one fiber, greatly improving the bandwidth and efficiency of data transmission. To ensure the accuracy and reliability of data transmission, forward error correction (FEC) coding technology is adopted to encode the transmitted data. At the receiving end, decoding algorithms are used to correct possible errors during the transmission process. Specifically, Reed-Solomon code is used for FEC coding. By adding redundant check symbols to the original data, the anti-interference ability of the data is improved. In the design of the fiber-optic communication system, the electromagnetic environment and space limitations inside the UAV are considered, and special fiber-optic cabling and protection measures are adopted to ensure that the fiber-optic communication system can work stably and reliably in the complex internal environment of the UAV, avoid data loss and errors during data transmission, and ensure the normal operation of the entire self-check system.

[0034] Furthermore, to improve the reliability and fault tolerance of the system, a distributed redundancy design is adopted. In the terahertz emission module, terahertz reception module, image acquisition and processing module, data analysis and fault diagnosis module, and display and alarm module, redundant backups are set for key components and circuits. In the terahertz emission module, two independent quantum cascade lasers are set as terahertz wave sources. When one laser fails, the other laser can automatically take over the work to ensure the continuous emission of terahertz waves. In the terahertz reception module, each detector is equipped with a backup detector. When the main detector fails, the backup detector can be immediately activated to continue receiving terahertz signals. In the image acquisition and processing module and data analysis and fault diagnosis module, a multi-processor parallel processing architecture is adopted. When a certain processor fails, other processors can share its work tasks to ensure that the processing ability of the system is not affected. In the display and alarm module, multiple display and alarm devices are set. When one device fails, other devices can work normally to ensure that operators can obtain detection results and alarm information in a timely manner. Through this distributed redundancy design, the reliability and fault tolerance of the system are greatly improved. Even when some components fail, the system can still continue to operate, providing a more reliable guarantee for the safe flight of the power UAV.

[0035] Compared with the prior art, the power UAV self-check system has the following beneficial effects:

[0036] I. By adopting a terahertz emission and reception module and combining image acquisition and processing algorithms, this system can achieve precise detection of the internal structure of drones. The terahertz wave has unique penetrability and sensitivity to the structure of substances, enabling the system to accurately identify the status of electronic components and circuit board parts, detect minor defects such as loose solder joints and fine circuit fractures. At the same time, through the deep neural network algorithm that fuses multiple features, the system can intelligently analyze terahertz images, further improving the accuracy of fault diagnosis. In addition, the improved iterative reconstruction algorithm and adaptive step size adjustment mechanism accelerate the image reconstruction process, significantly improving the detection efficiency.

[0037] II. By adopting a distributed redundant design, this system sets redundant backups in key components and circuits, such as dual quantum cascade lasers as terahertz wave sources and backup detectors, ensuring that the system can continue to operate even when some components fail. In addition, the multi-processor parallel processing architecture and the setting of multiple display and alarm devices further improve the processing ability of the system and the reliability of fault alarms. This design enables the power drone to detect and respond to potential faults in a timely manner during the self-check process, providing a more reliable guarantee for the safe flight of the drone. At the same time, the application of high-speed and low-latency optical fiber communication technology and forward error correction FEC coding technology also enhances the accuracy and reliability of data transmission, avoiding data loss and errors during the data transmission process and ensuring the normal operation of the entire self-check system.

[0038] Other advantages, objectives, and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0040] Figure 1 It is a flowchart of the operation of a power drone self-check system. DETAILED DESCRIPTION OF THE INVENTION

[0041] To further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention objective, the following will, in conjunction with the drawings and preferred embodiments, describe in detail the specific embodiments, structures, features, and effects of the present invention as follows.

[0042] Embodiment 1

[0043] In this embodiment, during the inspection of high-voltage transmission lines, a power drone needs to comprehensively detect the lines, towers and ancillary equipment. The drone flies to the target area and uses a terahertz self-inspection system to detect its internal structure to ensure flight safety and the accuracy of detection data. At the same time, it conducts an external inspection of the transmission line to check whether there are defects in the line.

[0044] A multi-layer heterostructure quantum well based on group III-V compound semiconductors is adopted. The well width w and well depth d are controlled by molecular beam epitaxy technology, and the external electric field strength E ext is regulated by a power supply module. For example, the well width w = 5 nm (with an accuracy reaching the sub-nanometer level), the well depth d = 200 meV (with an accuracy within 1%), and the external electric field strength E ext = 5 V / cm. According to the formula ΔE = α·w + β·d + γ·E ext (assuming α = 10 meV / nm, β = 2 meV / meV, γ = 0.5 meV / (V / cm)), the energy level difference ΔE is calculated, and then (h is Planck's constant) is used to determine the terahertz wave output frequency f, generating terahertz waves with a frequency of 3 THz to scan the inside of the drone.

[0045] Based on a detector with a composite structure of two-dimensional materials and traditional semiconductors, two-dimensional materials (such as graphene) are grown on the surface of traditional semiconductors by chemical vapor deposition technology. The deposition temperature is controlled at 800 °C and the gas flow rate is 50 sccm to precisely control the thickness and quality of the two-dimensional materials, improving the quantum efficiency η. The detector receives the terahertz waves after penetrating the inside of the drone, converts them into electrical signals, and uses the formula (where I ph = η·N ph ·e) to calculate the responsivity and amplify and preprocess it.

[0046] The preprocessed electrical signals are received, and the terahertz image of the drone's internal structure is reconstructed through an improved iterative reconstruction algorithm (introducing an adaptive step size adjustment mechanism). For example, in a certain iteration, the total number of measurement data N = 100, and the calculation error assuming the initial error E = 0.5 and the initial step size λ 0 = 0.1, and the step size adjustment coefficient α = 2. The step size is adjusted according to λ = λ 0 ·exp(-α·E), and the iterative formula is used for multiple iterations to enhance and denoise the image.

[0047] A deep neural network algorithm that fuses multiple features is adopted. The network inputs terahertz image data and performs convolution operations Feature extraction and non - linear transformation are carried out. Through the feature fusion layer (fusing texture, shape, and grayscale features based on the attention mechanism), the importance of different features is highlighted, and the fault diagnosis result is output at the output layer to detect the states of internal electronic components and circuit board components of the UAV, such as minor defects like loose solder joints and fine circuit breaks.

[0048] Embodiment 2

[0049] This embodiment describes that the substation environment is complex and has strong electromagnetic interference. The UAV performs regular self - inspections before entering the substation to execute the inspection task and during the task execution, promptly discovers internal faults that may be caused by electromagnetic interference factors, and ensures the smooth progress of the inspection task.

[0050] Terahertz waves are generated in the substation. The quantum well material of it controls the precision through molecular beam epitaxy technology to ensure the stable output of terahertz waves. The emission module is connected to other modules through optical fiber communication. The optical fiber adopts special wiring and protection measures to resist electromagnetic interference. At the same time, the redundant design of the double - quantum - cascade laser in the emission module ensures the stability of the terahertz wave source.

[0051] The detector adopts a special composite structure to improve the quantum efficiency in an electromagnetic interference environment. Its backup detector is on standby at any time to ensure the normal reception of terahertz signals. The reception module transmits the received signals to the image acquisition and processing module after processing.

[0052] The iterative reconstruction algorithm is used to reconstruct the terahertz image. During the iteration process, the measurement matrix Hij is optimized according to the characteristics of the substation environment, and an adaptive step - size adjustment mechanism is introduced to adapt to data changes in a complex environment, obtaining a clear image of the UAV's internal structure and transmitting it to the data analysis and fault diagnosis module.

[0053] The neural network algorithm is trained through a large number of samples collected in the substation environment, enabling the network to accurately identify fault characteristics that may be generated due to electromagnetic interference. The feature fusion layer effectively fuses various feature information, accurately diagnoses faults, and transmits the results to the display and alarm module through optical fiber.

[0054] The AR technology is used to display the fused image at the ground control station. The operator can intuitively view the internal condition of the UAV through interactive operations. The alarm information prompts the fault in a prominent way, and the redundant design of multiple display and alarm devices ensures the normal transmission of information.

[0055] The above are only the preferred embodiments of the present invention and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with the preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments by using the above-disclosed technical content within the scope of the technical solution of the present invention. However, as long as it does not depart from the content of the technical solution of the present invention, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A self-test system for electric drones, characterized in that: The system includes the following components: terahertz transmitting module, terahertz receiving module, image acquisition and processing module, data analysis and fault diagnosis module, and display and alarm module; The terahertz emission module generates a large terahertz wave with a frequency range of 0.1THz-10THz. The module is designed based on a quantum cascade structure and generates a terahertz wave of a specific frequency by precisely controlling the energy level transition of electrons in the quantum well. The output frequency f of the terahertz wave is determined by the following formula: Among them, ΔE is the energy level difference of electron transition in the quantum well, which is determined by the material properties of the quantum well, the well width w, the well depth d, and the externally applied electric field strength E ext The specific relationship is ΔE=α·w+β·d+γ·E ext , α, β, and γ are coefficients obtained by combining experiments and theoretical calculations based on the properties of quantum well materials, and h is Planck's constant. ext The value of can achieve precise control of the terahertz wave frequency; The terahertz receiving module is located at multiple specific locations inside the drone and is used to receive the terahertz waves that have penetrated the internal structure of the drone and convert them into electrical signals for amplification and preprocessing. The module adopts a detector structure based on new semiconductor materials, and its responsivity R is determined by the following formula: Among them, I ph is the photocurrent, which is related to the number of high-Hertz photons absorbed by the detector N ph , the quantum efficiency η of the detector and the electron charge e, that is, I ph =η·N ph ·e;P thz is the incident terahertz wave power. By optimizing the material structure and process parameters of the detector, the quantum efficiency η is improved, thereby improving the responsiveness of the receiving module. Image acquisition and processing module: connected to the terahertz receiving module, receives the pre-processed electrical signal, and reconstructs the terahertz image of the internal structure of the drone through an iterative reconstruction algorithm. The algorithm is based on the principle of maximum likelihood estimation and approximates the real image by iteratively updating the image pixel value. The specific iterative formula is as follows: in, is the value of the i-th pixel at the k-th iteration, P j is the jth measurement data received by the terahertz receiving module, H ij is the measurement matrix, which represents the contribution weight of the i-th pixel to the j-th measurement data. The weight is determined by the geometry of the internal structure of the UAV, the propagation path of the terahertz wave, and the position of the detector. Through multiple iterations, the error of the reconstructed image is gradually reduced, and finally a clear terahertz image is obtained, and the image is enhanced and denoised based on adaptive filtering; The data analysis and fault diagnosis module: analyzes the reconstructed terahertz image, uses a deep neural network algorithm that integrates multiple features to identify the status of electronic components and circuit board components in the image, and detects whether there are loose solder joints and minor circuit fractures. The neural network model consists of an input layer, multiple hidden layers, and an output layer. The input layer receives terahertz image data, and after feature extraction and nonlinear transformation of the hidden layer, the output layer outputs the fault diagnosis result. The feature extraction of the hidden layer is based on the following convolution operation: in, is the feature map element of the mth row and nth column of the lth layer, is the convolution kernel weight of the lth layer, which is obtained by training a large number of normal and faulty high-Hz image samples through the back-propagation algorithm. is the input feature map element of the l-1th layer, F is the size of the convolution kernel, b l is the bias term of the lth layer; The display and alarm module displays the detection results of the data analysis and fault diagnosis module. The module is connected to the UAV's onboard display screen or the communication interface of the ground control station to display the detection results in an intuitive graphical interface. When a fault is detected inside the UAV, an audible and visual alarm signal is triggered through a pre-set fault code and alarm rules.

2. The self-test system for electric drone according to claim 1, characterized in that: The material of the quantum well in the terahertz emission module adopts a multilayer heterostructure based on 111-V group compound semiconductors. The well width w and well depth d of the quantum well are precisely controlled by molecular beam epitaxy technology. During the growth process, the precision of the well width w is ensured to reach the sub-nanometer level and the precision of the well depth d is ensured to reach within 1% through real-time monitoring and feedback control technology. At the same time, the externally applied electric field strength E ext It is regulated by the power module, and the output accuracy of the power module is ±0.1V.

3. The self-test system for electric drone according to claim 1, characterized in that: The novel semiconductor material in the terahertz receiving module adopts a structure based on a composite of two-dimensional materials and traditional semiconductors. This composite structure grows two-dimensional materials with atomic-level thickness on the surface of traditional semiconductors and utilizes the unique photoelectric properties of two-dimensional materials to improve the quantum efficiency η of the detector. The selection of two-dimensional materials is based on their energy band structure and carrier mobility. The optimal combination of two-dimensional materials and traditional semiconductors is determined by combining theoretical calculations with experimental tests. During the preparation process, chemical vapor deposition technology is used to uniformly grow two-dimensional materials on the surface of traditional semiconductors. The thickness and quality of the two-dimensional materials are precisely controlled by controlling the deposition temperature and gas flow parameters.

4. The self-checking system for electric drone according to claim 1, characterized in that: In the improved iterative reconstruction algorithm of the image acquisition and processing module, an adaptive step adjustment mechanism is introduced to accelerate the iterative convergence speed. In each iteration, the iterative step λ is dynamically adjusted according to the error and convergence trend of the current image reconstruction. The error calculation formula is: in, It is the data actually measured by the terahertz receiving module. is the corresponding data obtained by the current iterative reconstruction, N is the total number of measured data, and the step size λ is adjusted according to the size of the error E by the following formula: λ=λ0·exp(-α·E) Among them, λ0 is the initial step size, α is the step size adjustment coefficient determined according to the experiment. When the error E is large, the step size λ is increased to speed up the convergence speed. When the error E is small, the step size λ is reduced to improve the reconstruction accuracy.

5. The self-test system for electric drone according to claim 1, characterized in that: In the deep neural network algorithm for fusion of multiple features of the data analysis and fault diagnosis module, a feature fusion layer is introduced into the network in order to better fuse the texture, shape, and grayscale multi-feature information of the image. The feature fusion layer adopts a fusion method based on the attention mechanism. By learning the importance weights of different features, different features are dynamically fused. For the input texture feature T, shape feature S, and grayscale feature G, they are first mapped to the same dimension through linear transformation to obtain T′, S′, and G′, and then the attention weight of each feature is calculated: Among them, ω T ,ω S and ω G is the weight parameter learned through network training. Finally, the fused feature F is: F=A T ·T′+A S ·S′+A G ·G′ Through this feature fusion method based on the attention mechanism, the importance of different features in fault diagnosis can be more effectively highlighted.

6. The self-checking system for electric drone according to claim 1, characterized in that: In the display and alarm module, augmented reality (AR) technology is used. On the airborne display screen or the display of the ground control station, the terahertz image reconstruction result is fused with the three-dimensional model of the UAV through AR technology and displayed. The attitude and position information of the UAV is obtained through the inertial measurement unit IMU and the global positioning system GPS of the UAV, and then the fault position information obtained by the terahertz image reconstruction is mapped to the three-dimensional model of the UAV. When displaying, the three-dimensional model with the fault mark is fused with the actual scene in real time by using the AR engine. The operator can operate the displayed content through gesture interaction or voice command. At the same time, for the alarm signal, a combination of voice and vision is adopted to prompt the operator with the type and location of the fault in the AR scene with eye-catching colors and animation effects.

7. The self-test system for electric drone according to claim 1, characterized in that: The terahertz transmission module, the high-hertz receiving module, the image acquisition and processing module, the data analysis and fault diagnosis module, and the display and alarm module use high-speed, low-latency optical fiber communication technology for data transmission. The optical fiber communication system uses wavelength division multiplexing technology to modulate the signals transmitted between different modules to different wavelengths λ sig In the process of transmitting multiple signals simultaneously in one optical fiber, the bandwidth and efficiency of data transmission are greatly improved. Forward error correction FEC coding technology is adopted to encode the transmitted data, and the errors that may occur in the transmission process are corrected by decoding algorithm at the receiving end. The specific FEC coding adopts Reed-Solomon code, and the anti-interference ability of the data is improved by adding redundant check code elements to the original data. In the design of the optical fiber communication system, the electromagnetic environment and space limitations inside the drone are taken into consideration, and special optical fiber wiring and protection measures are adopted.

8. The self-checking system for electric drone according to claim 1, characterized in that: In order to improve the reliability and fault tolerance of the system, a distributed redundant design is adopted. In the terahertz transmitting module, the terahertz receiving module, the image acquisition and processing module, the data analysis and fault diagnosis module, and the display and alarm module, key components and circuits are set with redundant backup. Two independent quantum cascade lasers are set in the terahertz transmitting module as terahertz wave sources. When one laser fails, the other laser can automatically take over the work. In the terahertz receiving module, each detector is equipped with a backup detector. When the main detector fails, the backup detector can be started immediately to continue to receive terahertz signals. In the image acquisition and processing module and the data analysis and fault diagnosis module, a multi-processor parallel processing architecture is adopted. When a processor fails, other processors can share its work tasks. In the display and alarm module, multiple display and alarm devices are set. When one device fails, other devices can work normally.

Images