Equipment welding method and system in space environment, equipment and medium

Abstract

The invention discloses an equipment welding method and system in a space environment, equipment and a medium, relates to the technical field of space welding, and solves the problem of camera overexposure caused by strong light reflection by setting an exposure mode of an HDR camera into a short exposure mode to shoot an image on a sunny side. Meanwhile, welding arc light and metal surface reflection light are extracted by using a multispectral sensor, so that the metal reflection light and the welding arc light are distinguished; the exposure mode of the multi-angle camera set is set to be a long exposure mode on the shady face, infrared laser is matched to irradiate the welding position to shoot images, the image brightness is continuously enhanced, meanwhile, a thermal imaging sensor is used for recognizing the images, and therefore the problem that equipment generates low-temperature deformation during low illumination is solved; therefore, the welding seam position of the sunny side can be identified according to the position of the welding arc light, the welding seam position of the shady side can be identified through brightness compensation on the shady side, meanwhile, the welding problem of the sunny side and the shady side is solved, and the welding precision in the space environment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of space welding, and particularly relates to a device welding method, system, device and medium in a space environment. Background Art

[0002] With the continuous deepening of space exploration, space welding technology plays an increasingly important role in tasks such as space station maintenance and satellite repair; space welding can not only repair and maintain space equipment, but also play a key role in space manufacturing, providing technical support for long-term space missions; however, the extreme conditions of the space environment pose severe challenges to welding operations, especially the impact of the sunlit side and the shaded side on the robot vision system.

[0003] On the sunlit side, when welding parameters are set for the robot to weld along the set weld seam, due to the extremely fast movement speed of equipment in the space environment, it will cause deviation from the set route during welding. Therefore, when the robot welds the weld seam within the welding position on the sunlit side, it identifies the position of the subsequent weld seam while welding. However, on the sunlit side, due to direct sunlight, the light intensity is extremely high, exceeding 10 5 lux. This strong light reflection will cause overexposure of the camera of the robot vision system and loss of image details, resulting in the inability to accurately identify the welding arc light and the characteristics of the metal surface, and thus unable to distinguish the position of the subsequent weld seam; on the shaded side, the light intensity is extremely low, lower than 1 lux. This low light condition makes it difficult for the camera to capture clear images, and the signal-to-noise ratio (SNR) of the images is lower than 10 dB, seriously affecting the performance of the vision system; at the same time, the temperature on the shaded side can be as low as -180 °C, and the metal material will undergo shrinkage deformation under such extreme low temperatures, resulting in a decrease in welding accuracy.

[0004] The current space welding vision systems mainly rely on a single visible light camera, and have the following limitations: on the sunlit side, strong light reflection causes overexposure of the camera, and it is difficult to distinguish the metal reflection from the welding arc light under strong light, resulting in difficulty in identifying the position of the subsequent weld seam during welding, seriously affecting welding accuracy; on the shaded side, under low light conditions, the images captured by the camera are blurred, and due to low-temperature deformation of the equipment under low light, it is difficult to accurately identify the welding features in the images.

[0005] Therefore, the current welding systems are difficult to simultaneously solve the welding problems existing on the sunlit side and the shaded side, resulting in low accuracy in space welding and difficulty in being applied to complex space tasks such as space station maintenance and satellite repair. Summary of the Invention

[0006] An embodiment of the present invention provides a method, system, device and medium for welding equipment in a space environment, which can solve the problem in the prior art that the current welding system is difficult to simultaneously solve the welding problems on the sunny side and the shady side, resulting in low precision in welding in a space environment.

[0007] An embodiment of the present invention provides a method for welding equipment in a space environment, comprising the following steps: When the light intensity is higher than a preset threshold, set the exposure mode of the HDR camera to the short exposure mode. During the welding of the weld in the welding position, obtain the HDR image of the welding position through the HDR camera, and use the multispectral sensor to extract the spectral intensity values of the welding arc light and the metal surface reflection in the welding position; wherein, the short exposure mode means that the exposure time of the HDR camera is less than or equal to 0.1 ms; Map the spectral intensity values of the welding arc light and the metal surface reflection to the luminance channel of the HDR image, and through spatial coordinate alignment, make the spectral luminance correspond one-to-one with the pixels of the HDR image and fuse them to obtain the first welding image, identify the welding arc light and the metal surface reflection in the first welding image to distinguish the subsequent weld positions in the welding position, and weld the target equipment; When the light intensity is lower than the preset threshold, set the exposure mode of the multi-angle camera group to the short exposure mode. During the welding of the weld in the welding position, obtain the welding image of the welding position through the multi-angle camera group, and use the infrared laser to irradiate the welding position to obtain the infrared image and convert it into a visible light image; wherein, the high exposure mode means that the exposure time of the cameras of the multi-angle camera group is greater than or equal to 50 ms; Fuse the welding image and the visible light image through spatial coordinate alignment to obtain the second welding image, identify the weld position in the second welding image, and weld the target equipment.

[0008] Preferably, the acquisition of the first welding image includes: Install the HDR camera and the multispectral sensor at different positions of the welding robotic arm of the welding robot, and a polarization filter is provided at the end of the HDR camera; When the welding position of the target equipment is in the sunny side and the light intensity of the sunny side > 10^5 lux, during the welding of the weld in the welding position, set the exposure mode of the HDR camera to the short exposure mode, and cooperate with the polarization filter to take the image of the welding position. At the same time, use the multispectral sensor with a wavelength band of 400 nm - 1700 nm to extract the welding arc light and the metal surface reflection during the welding process to distinguish the subsequent weld positions in the welding position, and obtain the first welding image; Among them, the short exposure mode means that the exposure time of the HDR camera is less than or equal to 0.1 ms; the wavelength range of the welding arc light is 840 nm - 850 nm, and the wavelength range of the metal surface reflection is 540 nm - 560 nm.

[0009] Preferably, the acquisition of the second welding image includes: Install multi-angle camera groups, infrared lasers and thermal imaging sensors at different positions of the welding robot's welding manipulator respectively; When the welding position of the target device is in the shaded area and the light intensity of the shaded area < 10^2 lux, during the welding of the weld in the welding position, set the exposure mode of the multi-angle camera group to the high exposure mode, and cooperate with the infrared laser to irradiate the welding position to capture an image of the welding position, and at the same time use the thermal imaging sensor to identify the low-temperature deformation of the metal at the welding position to obtain the second welding image; Among them, the high exposure mode means that the exposure time of the cameras of the multi-angle camera group is greater than or equal to 50 ms.

[0010] Preferably, when welding the target device, synchronously planning the welding path of the welding manipulator includes: After obtaining the weld position on the sunny side or the shaded side, and based on the light intensity and temperature on the sunny side and the shaded side, use the PPO reinforcement learning algorithm to plan the welding path of the welding manipulator, and plan the welding path of the welding manipulator when welding on the sunny side and the shaded side; At the same time, adjust the welding torch offset in real time according to the temperature feedback during the welding process, and the offset equation is: coefficient of thermal expansion, Δx = α ×ΔT×L; Where: α represents the coefficient of thermal expansion, α =23×10 -6 / ℃; ΔT represents the real-time temperature feedback during the welding process; L represents the length of the welding path.

[0011] The embodiment of the present invention also provides a device welding system in a space environment, including: A sunny side recognition module, configured to set the exposure mode of the HDR camera to the short exposure mode when the light intensity is higher than a preset threshold. During the welding of the weld in the welding position, obtain the HDR image of the welding position through the HDR camera, and use the multispectral sensor to extract the spectral intensity values of the welding arc light and the metal surface reflection in the welding position; among them, the short exposure mode means that the exposure time of the HDR camera is less than or equal to 0.1 ms; Sunny-side welding module, which is used to map the spectral intensity values of welding arc light and metal surface reflection light to the luminance channel of the HDR image, and through spatial coordinate alignment, make the spectral luminance correspond one by one with the pixels of the HDR image and fuse them to obtain the first welding image, identify the welding arc light and metal surface reflection light in the first welding image to distinguish the subsequent weld positions within the welding position, and weld the target device; Shady-side recognition module, which is used to set the exposure mode of the multi-angle camera group to the short exposure mode when the light intensity is lower than the preset threshold. During the welding of the weld within the welding position, obtain the welding image of the welding position through the multi-angle camera group, and use the infrared laser to irradiate the welding position to obtain an infrared image and convert it into a visible light image; wherein, the high exposure mode means that the exposure time of the cameras in the multi-angle camera group is greater than or equal to 50 ms; Shady-side welding module, which is used to fuse the welding image and the visible light image through spatial coordinate alignment to obtain the second welding image, identify the weld position in the second welding image, and weld the target device.

[0012] An embodiment of the present invention further provides an electronic device, including a memory and a processor; The memory is used to store a computer program; The processor, when executing the computer program stored in the memory, realizes the steps of a device welding method in a space environment as described above.

[0013] An embodiment of the present invention further provides a computer-readable storage medium, which is used to store a computer program, and when the computer program is executed by a processor, it realizes the steps of a device welding method in a space environment as described above.

[0014] An embodiment of the present invention provides a device welding method, system, device and medium in a space environment. Compared with the prior art, its beneficial effects are as follows: In the present invention, by considering the welding classification of the sunny side and the shady side separately, a sunny side sensor including an HDR camera, a polarization filter, and a multispectral sensor is set up, and at the same time, a shady side sensor including an infrared laser, a thermal imaging sensor, and a multi-angle camera group is set up. When the light intensity on the sunny side exceeds the threshold, by setting the exposure mode of the HDR camera to the short exposure mode and cooperating with the polarization filter to take images, the problem of overexposure of the camera caused by strong light reflection is solved. At the same time, the multispectral sensor with a wavelength band of 400 - 1700 nm is used to extract the welding arc light with a wavelength band of 840 - 860 nm and the metal surface reflection light with a wavelength of 540 - 560 nm, so as to distinguish the metal reflection light and the welding arc light under strong light. When the light intensity on the shady side is lower than the threshold, by setting the exposure mode of the multi-angle camera group to the long exposure mode, cooperating with the infrared laser to irradiate the welding position and take images, the image brightness can be continuously enhanced for shooting, and at the same time, the thermal imaging sensor is used to identify the image to solve the problem of low-temperature deformation of the device in low light. Thus, the weld position on the sunny side can be identified according to the position of the welding arc light, and the weld position on the shady side can be identified by brightness compensation on the shady side, while solving the welding problems existing on the sunny side and the shady side, and improving the welding accuracy in the space environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the system flow architecture of a device welding method in a space environment provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of the sensor layout of a device welding method in a space environment provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of the data processing flow of a device welding method in a space environment provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention will be given in conjunction with the accompanying drawings. Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0017] See Figure 1, the embodiments of the present invention provide a method for welding equipment in a space environment. The present invention aims to solve the limitations of existing space welding vision systems on the sunny side and the shady side. By introducing multi-modal sensors such as high-dynamic range (HDR) cameras, polarization filters, multi-spectral sensors, infrared lasers, thermal imaging sensors, and lidar, it can adapt to different lighting and temperature conditions. At the same time, adaptive exposure and filtering algorithms are developed, such as a reflection light separation model based on the U-Net architecture and a low-light enhancement algorithm based on CycleGAN, to improve image quality and the accuracy of weld feature extraction. In addition, reinforcement learning algorithms such as the PPO algorithm are used to adjust the welding path and parameters in real time according to environmental perception data to ensure welding accuracy and efficiency. The present invention can not only cope with the extreme conditions of the space environment, but also significantly improve the efficiency and quality of welding work, and is applicable to complex space tasks such as space station maintenance and satellite repair.

[0018] Specifically: I. Dynamic multi-modal sensor fusion.

[0019] 1. Sensor selection and configuration.

[0020] Sunny side (light intensity > 5×10 4 lux): HDR camera: Use FLIR Blackfly S BFS-U3-51S5M-C, with a dynamic range of 120dB, which can effectively suppress strong light reflection and direct sunlight interference. The camera can still maintain high resolution and high clarity in high light intensity environments to ensure image quality.

[0021] Polarization filter: The extinction ratio is 100:1, which can effectively reduce specular reflection and further improve image quality. The use of a polarization filter can significantly reduce the interference of strong light on the camera and ensure clear imaging of the welding arc.

[0022] Multi-spectral sensor: The wavelength band is 400 - 1700nm, which can distinguish the welding arc (peak wavelength 850nm) from metal reflection (peak wavelength 550nm). The use of a multi-spectral sensor can effectively extract welding features and improve the recognizability of images.

[0023] Shady side (light intensity < 10 2 lux): Active light source: 940nm infrared laser, with a power of 50mW, meeting the Class 1 safety standard, which can compensate for insufficient light in low-light environments to ensure image clarity and recognizability. The use of an infrared laser can significantly improve the brightness and contrast of images on the shady side.

[0024] Thermal imaging sensor: FLIR A65, with an accuracy of ±2°C, capable of monitoring the low-temperature shrinkage deformation of metals to ensure the welding seam positioning accuracy; the use of the thermal imaging sensor can effectively detect metal deformation and improve welding accuracy.

[0025] LiDAR: Velodyne VLP-16, with a scanning frequency of 20Hz, generating 3D point cloud data, and a resolution of ±3mm; the use of LiDAR can provide high-precision 3D environmental information and enhance the robustness of the vision system.

[0026] Multi-angle camera group: with a field of view of 120°×90°, covering the blind area of the welding area; the use of the multi-angle camera group can ensure the full coverage of the vision system in complex environments and improve the flexibility and accuracy of welding operations.

[0027] 2. Adaptive exposure and filtering algorithms.

[0028] Dynamic exposure logic: When the light intensity > 10^5 lux, switch to the short exposure mode (exposure time 0.1ms) and trigger HDR synthesis (3-frame fusion) to reduce overexposure and ensure clear image details.

[0029] When the light intensity < 10^2 lux, enable the active light source and extend the exposure to 50ms to enhance the image brightness and improve the image quality.

[0030] Image enhancement algorithm: On the sunny side, a reflected light separation model based on the U-Net architecture, with a training dataset containing 100,000 simulated space strong light images, can effectively separate the reflected light on the metal surface from the welding arc light and improve the clarity and accuracy of the image.

[0031] On the shady side, CycleGAN is used for low-light enhancement. The generative adversarial network contains 4 residual blocks, and the loss function is SSIM+L1, which can significantly improve the brightness and contrast of the image and enhance the visibility and recognizability of the image.

[0032] II. Image processing and feature extraction for enhanced illumination robustness.

[0033] 1. Multi-modal data fusion processing.

[0034] Feature-level fusion: Input visible light images (resolution 1920×1080), infrared images (resolution 640×512), and LiDAR point cloud data (density 500 points / cm²) into an FPGA accelerator (Xilinx Zynq UltraScale+), and output a composite feature map through a weighted fusion algorithm; on the sunny side, the LiDAR weight is increased to 70% to avoid strong light interference; on the shady side, the weight of the infrared image is increased to 60% to enhance edge detection; this fusion technology can extract accurate weld features under complex lighting conditions, improving the accuracy and reliability of welding operations.

[0035] Decision logic: Dynamically adjust the weights of the data of each sensor according to the environmental lighting conditions to ensure that accurate weld features can be extracted under different lighting conditions; this dynamic adjustment mechanism can effectively cope with the lighting changes in the space environment, improving the adaptability and robustness of the system.

[0036] 2. Adaptive feature matching model.

[0037] Transfer learning framework: Based on the pre-trained ResNet-50 model, fine-tune it using a ground simulation dataset (including metal deformation data with a temperature gradient from -180°C to +150°C), and output the coordinates of the weld center line (error ±0.05mm); the use of transfer learning can significantly improve the generalization ability and adaptability of the model, ensuring high-precision welding operations in the space environment.

[0038] Real-time online learning: Deploy a lightweight PyTorch model, and update the convolution kernel parameters every 10 welding operations (learning rate 0.001, momentum 0.9) to adapt to environmental changes; the real-time online learning mechanism can ensure that the system is continuously optimized during long-term operation, improving the accuracy and reliability of welding operations.

[0039] III. Dynamic path planning and real-time regulation of welding parameters.

[0040] 1. Path optimization based on environmental perception.

[0041] Reinforcement learning strategy: Adopt the PPO reinforcement learning algorithm, the input states include light intensity, temperature, and point cloud data, and the reward function is path length × 0.3 + energy consumption × 0.2 + welding quality × 0.5; the algorithm can autonomously optimize the welding path in a complex environment, enabling the robot to select the optimal welding angle and movement speed according to the lighting conditions; the use of reinforcement learning can significantly improve the autonomy and adaptability of the system, ensuring high-precision welding operations under different lighting conditions.

[0042] Thermal Expansion Compensation: The torch offset is adjusted in real time according to the temperature feedback. The compensation formula is Δx = α × ΔT × L, where α is the coefficient of thermal expansion (23×10^-6 / °C), ΔT is the temperature change, and L is the welding path length. The thermal expansion compensation technology can effectively cope with the temperature changes in the space environment and ensure welding accuracy.

[0043] 2. Adaptive adjustment of welding parameters.

[0044] Closed-loop Control System: According to the weld characteristics and temperature data feedback by the vision system, the welding parameters such as current, voltage, and wire feeding speed are dynamically adjusted. On the shaded side, the preheating current increases by 20% (reference value 150A → 180A), and the wire feeding speed decreases to 0.8m / min. On the sunny side, the arc voltage decreases by 15% (reference value 28V → 23.8V). The use of the closed-loop control system can ensure the real-time adjustment of welding parameters and improve the accuracy and reliability of welding operations.

[0045] Predictive Maintenance: Early warning is given by monitoring abnormal heat distribution (such as local temperature difference > 50°C), and the fault detection rate is ≥ 95%. The predictive maintenance mechanism can effectively improve the reliability and safety of the system and ensure long-term stable operation.

[0046] IV. System Verification and Space Environment Simulation Test.

[0047] 1. Ground simulation experiment.

[0048] Test environment: Sunny side simulation chamber: The xenon lamp array light intensity is 1.2×10^5 lux ± 5%, and the temperature cycle range is from -170°C to +130°C, which can simulate the high light intensity and high temperature environment of the space sunny side.

[0049] Shaded side simulation chamber: The background radiation of the blackbody radiation source is <0.01W / m², and the vacuum degree is <10^-5 Pa, which can simulate the low light and low temperature environment of the space shaded side. Through these simulation experiments, the performance of the system in extreme environments is verified.

[0050] Experimental results: Positioning error: 0.08mm on the sunny side and 0.12mm on the shaded side, which is more than 6 times higher than the traditional technology, indicating the high-precision welding ability of the system under different lighting conditions.

[0051] Image signal-to-noise ratio: 28dB on the shaded side, which is 19dB higher than the traditional technology, indicating the high image quality of the system in low light environments.

[0052] Welding path deviation rate: <5%, which is 7% lower than the traditional technology, indicating the high-precision path planning ability of the system in complex environments.

[0053] 2. On-orbit real-time calibration.

[0054] Calibration mechanism: The homography matrix is corrected every 24 hours through a carbon fiber calibration plate (checkerboard accuracy ±0.01 mm), and the radiation drift compensation rate ≥ 90%; this calibration mechanism can effectively cope with the impact of space radiation on the sensor performance, ensuring the long-term stability and reliability of the system.

[0055] V. Structural design and radiation hardening.

[0056] 1. Modular design.

[0057] Encapsulation with radiation-resistant materials: The camera and laser are encapsulated with hafnium-doped silica glass, with a total dose tolerance > 100 krad, and can maintain high performance in the space radiation environment; this encapsulation material can effectively resist the damage of radiation to electronic components, ensuring the long-term reliability of the system.

[0058] Quick plug-and-play replacement: Support quick plug-and-play replacement, with an operation time < 5 minutes, to meet the needs of long-term space missions; this design improves the maintainability and scalability of the system, ensuring flexibility and reliability in long-term missions.

[0059] 2. Manipulator layout.

[0060] Six-degree-of-freedom manipulator: The repeat positioning accuracy is ±0.02 mm, equipped with a telescopic pan-tilt head, and the telescopic range is ±30 cm, avoiding the occlusion of the spacecraft structure; this layout ensures that the field of view of the sensor is not blocked, improving the flexibility and accuracy of welding operations.

[0061] Specific welding implementation: I. Welding operation on the sunny side.

[0062] When the robot vision system detects that the current environment is the sunny side (light intensity > 5×10 4 lux), the system automatically switches to the sunny-side welding mode; specifically including: 1. Sensor activation: Activate the high-dynamic-range (HDR) camera (FLIR Blackfly S BFS-U3-51S5M-C), which has a dynamic range of 120 dB and can effectively suppress strong light reflection; at the same time, cooperate with a polarization filter (extinction ratio 100:1) to further reduce specular reflection and ensure clear images.

[0063] 2. Multi-spectral sensor startup: Start the multi-spectral sensor (band 400 - 1700 nm), and extract the welding arc light (threshold: 850 nm ± 10 nm) through the band separation algorithm to distinguish metal reflection from welding arc light; this step ensures that the characteristics of the welding arc light can be accurately identified in a strong light environment.

[0064] 3. LiDAR Data Fusion: The LiDAR (Velodyne VLP-16) generates 3D point cloud data with a resolution of ±3 mm. During the welding process on the sunny side, the weight of the LiDAR is increased to 70% to avoid strong light interference. By fusing the 3D point cloud data and 2D visual data, an anti-interference welding path is generated to ensure welding accuracy.

[0065] 4. Image Processing: A reflected light separation model based on the U-Net architecture is used to process the collected images to separate the metal surface reflection and welding arc light. The model is trained with 100,000 simulated space strong light images, which can effectively improve the clarity and accuracy of the images.

[0066] 5. Path Planning and Parameter Regulation: According to the environmental perception data, the PPO reinforcement learning algorithm is started for path optimization. The algorithm generates the optimal welding path based on inputs such as light intensity, temperature, and point cloud data. At the same time, the torch offset is adjusted in real-time according to the temperature feedback (compensation formula: Δx = α × ΔT × L, α = 23 × 10^-6 / °C) to ensure welding accuracy. In terms of welding parameters, the arc voltage is reduced by 15% (reference value 28 V → 23.8 V) to avoid overheating.

[0067] 6. System Verification: Experimental verification is carried out in the sunny side simulation chamber (xenon lamp array, light intensity 1.2 × 10 5 lux ± 5%, temperature cycle -170°C to +130°C). The experimental results show that the system positioning error is 0.08 mm, which is more than 6 times better than traditional technologies, the image signal-to-noise ratio reaches 28 dB, and the welding path deviation rate is controlled within <5%.

[0068] II. Welding Operation on the Shaded Side.

[0069] When the robot vision system detects that the current environment is the shaded side (light intensity < 10 2 lux), the system automatically switches to the shaded side welding mode; specifically including: 1. Active Light Source Compensation: Enable a 940 nm infrared laser (power 50 mW, compliant with Class 1 safety standards) to compensate for the low light environment; the wavelength selection of the infrared laser avoids interference with the astronaut's line of sight while ensuring sufficient illumination intensity.

[0070] 2. Thermal Imaging Sensor Monitoring: Start the thermal imaging sensor (FLIR A65, accuracy ±2°C) to monitor the low-temperature shrinkage deformation of the metal; the sensor can maintain high accuracy in a low-temperature environment (-180°C) to ensure the weld seam positioning accuracy.

[0071] 3. Multimodal data fusion: Input visible light images (resolution 1920×1080), infrared images (resolution 640×512), and lidar point cloud data (density 500 points / cm²) into an FPGA accelerator (Xilinx Zynq UltraScale+), and output a composite feature map through a weighted fusion algorithm; during the welding process on the shaded side, the weight of the infrared image is increased to 60% to enhance edge detection and improve the accuracy of weld feature extraction.

[0072] 4. Image enhancement: Use a low-light enhancement algorithm based on CycleGAN to enhance the collected low-light images; this algorithm contains 4 residual blocks, and the loss function is SSIM+L1, which can significantly improve the brightness and contrast of the images, and enhance the visibility and recognizability of the images.

[0073] 5. Path planning and parameter regulation: According to the environmental perception data, start the PPO reinforcement learning algorithm for path optimization; the algorithm generates the optimal welding path based on inputs such as light intensity, temperature, and point cloud data; at the same time, adjust the torch offset in real time according to the temperature feedback (compensation formula: Δx = α×ΔT×L, α = 23×10-6 / ℃) to ensure welding accuracy; in terms of welding parameters, the preheating current is increased by 20% (benchmark value 150A → 180A), and the wire feeding speed is reduced to 0.8m / min to meet the welding requirements in a low-temperature environment.

[0074] 6. System verification: Conduct experimental verification in a shaded-side simulation chamber (blackbody radiation source, background radiation <0.01W / m², vacuum degree <10-5Pa); the experimental results show that the system positioning error is 0.12mm, which is more than 6 times better than traditional technologies, the image signal-to-noise ratio reaches 28dB, and the welding path deviation rate is controlled within <5%.

[0075] III. Online learning mechanism.

[0076] During the welding process, the system collects image data in real time and continuously optimizes the model through the online learning mechanism; specifically including: 1. Image acquisition: Collect 100 frames of images in real time during the welding process and input them into the PyTorch model; these images include visible light images, infrared images, and lidar point cloud data, covering multimodal information of the welding area.

[0077] 2. Loss function calculation: Calculate the loss function (cross entropy + mean square error) to evaluate the difference between the model prediction and the actual weld position; by comparing the weld center line predicted by the model with the actual weld center line, calculate the error value to provide feedback for model optimization.

[0078] 3. Parameter Update: Update the convolution kernel parameters through the SGD optimizer (learning rate 0.001, momentum 0.9) to optimize the model performance; after each update, the prediction accuracy of the model will be improved, gradually adapting to the complexity of the space welding environment.

[0079] 4. Model Deployment: The updated model is immediately deployed to the FPGA inference engine for real-time feature extraction and path planning of subsequent welding operations; this online learning mechanism ensures that the system can continuously adapt to environmental changes during long-term operation, maintaining high accuracy and high reliability.

[0080] IV. System Verification and Calibration.

[0081] To ensure the long-term stability and reliability of the system, system verification and calibration are carried out regularly; specifically including: 1. Ground Simulation Experiment: Conduct experimental verification in the sunny-side simulation chamber and the shady-side simulation chamber; the light intensity of the xenon lamp array in the sunny-side simulation chamber is 1.2×10 5 lux ± 5%, and the temperature cycle range is from -170°C to +130°C; the background radiation of the blackbody radiation source in the shady-side simulation chamber is <0.01 W / m², and the vacuum degree is <10 -5 Pa; through these simulation experiments, verify the performance of the system in extreme environments.

[0082] 2. Analysis of Experimental Results: The positioning error on the sunny side is 0.08 mm, the positioning error on the shady side is 0.12 mm, the image signal-to-noise ratio reaches 28 dB, and the welding path deviation rate is controlled within <5%; these results indicate that the system has high accuracy and high reliability on both the sunny side and the shady side.

[0083] 3. On-Orbit Real-Time Calibration: Perform homography matrix correction every 24 hours through a carbon fiber calibration plate (checkerboard accuracy ±0.01 mm), and the radiation drift compensation rate ≥90%; this calibration mechanism ensures the stability of the system during long-term operation and can effectively cope with the impact of space radiation on the performance of sensors.

[0084] Through technical solutions such as dynamic multi-modal sensor fusion technology, image processing and feature extraction technology with enhanced light robustness, dynamic path planning and real-time regulation of welding parameters, system verification and space environment simulation testing, and structural design and anti-radiation reinforcement, the present invention solves the problems of overexposure of strong light on the sunny side and low light and low-temperature deformation on the shady side in space welding; the present invention is applicable to complex space missions such as space station maintenance and satellite repair, has significant technical advantages and innovation, and can effectively improve the accuracy and reliability of space welding operations; at the same time, the welding positioning accuracy ≤0.1 mm and the image signal-to-noise ratio ≥25 dB (shady side) are achieved.

[0085] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the appended claims.

Claims

1. A method for welding equipment in a space environment, characterized in that, The method includes the following steps: When the light intensity is higher than a preset threshold, set the exposure mode of the HDR camera to the short exposure mode. During the welding process of the weld seam within the welding position, obtain the HDR image of the welding position through the HDR camera, and use the multispectral sensor to extract the spectral intensity values of the welding arc light and the metal surface reflection within the welding position. Among them, the short exposure mode means that the exposure time of the HDR camera is less than or equal to 0.1 ms; Map the spectral intensity values of the welding arc light and the metal surface reflection to the luminance channel of the HDR image, and through spatial coordinate alignment, make the spectral luminance correspond one-to-one with the pixels of the HDR image and fuse them to obtain the first welding image. Identify the welding arc light and the metal surface reflection in the first welding image to distinguish the subsequent weld seam positions within the welding position, and perform welding on the target device; When the light intensity is lower than the preset threshold, set the exposure mode of the multi-angle camera group to the short exposure mode. During the welding process of the weld seam within the welding position, obtain the welding image of the welding position through the multi-angle camera group, and use the infrared laser to irradiate the welding position to obtain the infrared image and convert it into a visible light image. Among them, the high exposure mode means that the exposure time of the cameras of the multi-angle camera group is greater than or equal to 50 ms; Fuse the welding image and the visible light image through spatial coordinate alignment to obtain the second welding image. Identify the weld seam position in the second welding image and perform welding on the target device.

2. The device welding method in a space environment according to claim 1, wherein The obtaining of the first welding image includes: Install the HDR camera and the multispectral sensor at different positions of the welding robot's welding manipulator respectively. A polarization filter is provided at the end of the HDR camera; When the welding position of the target device is within the sunny side and the light intensity of the sunny side > 10^5 lux, during the welding process of the weld seam within the welding position, set the exposure mode of the HDR camera to the short exposure mode, and cooperate with the polarization filter to take the image of the welding position. At the same time, use the multispectral sensor with a wavelength range of 400 nm - 1700 nm to extract the welding arc light and the metal surface reflection during the welding process to distinguish the subsequent weld seam positions within the welding position and obtain the first welding image; Among them, the short exposure mode means that the exposure time of the HDR camera is less than or equal to 0.1 ms; the wavelength range of the welding arc light is 840 nm - 850 nm, and the wavelength range of the metal surface reflection is 540 nm - 560 nm.

3. A method for welding equipment in a space environment according to claim 1, characterized in that, The obtaining of the second welding image includes: Install the multi-angle camera group, the infrared laser and the thermal imaging sensor at different positions of the welding robot's welding manipulator respectively; When the welding position of the target device is within the shady side and the light intensity of the shady side < 10^2 lux, during the welding process of the weld seam within the welding position, set the exposure mode of the multi-angle camera group to the high exposure mode, and cooperate with the infrared laser to irradiate the welding position to take the image of the welding position. At the same time, use the thermal imaging sensor to identify the low-temperature deformation of the metal at the welding position and obtain the second welding image; Among them, the high-exposure mode means that the exposure time of the cameras in the multi-angle camera group is greater than or equal to 50 ms.

4. The welding method of a device in a space environment according to claim 1, characterized in that When welding the target device, synchronously plan the welding path of the welding robot arm, including: After obtaining the weld position on the sunny side or the shady side, and based on the light intensity and temperature on the sunny side and the shady side, use the PPO reinforcement learning algorithm to plan the welding path of the welding robot arm, and plan the welding path of the welding robot arm when welding on the sunny side and the shady side; At the same time, adjust the torch offset in real time according to the temperature feedback during the welding process. The offset equation is: coefficient of thermal expansion, Δx = α ×ΔT×L; Wherein: α represents the coefficient of thermal expansion, α = 23×10 -6 / °C; ΔT represents the real-time temperature feedback during the welding process; L represents the welding path length.

5. A device welding system in a space environment, characterized in that, including: Sunny-side identification module, used to set the exposure mode of the HDR camera to the short-exposure mode when the light intensity is higher than the preset threshold. During the welding process of the weld in the welding position, obtain the HDR image of the welding position through the HDR camera, and use the multispectral sensor to extract the spectral intensity values of the welding arc light and the metal surface reflection in the welding position; among them, the short-exposure mode means that the exposure time of the HDR camera is less than or equal to 0.1 ms; Sunny-side welding module, used to map the spectral intensity values of the welding arc light and the metal surface reflection to the luminance channel of the HDR image, and through spatial coordinate alignment, make the spectral luminance correspond one-to-one with the pixels of the HDR image and fuse them to obtain the first welding image, identify the welding arc light and the metal surface reflection in the first welding image to distinguish the subsequent weld positions in the welding position, and weld the target device; Shady-side identification module, used to set the exposure mode of the multi-angle camera group to the short-exposure mode when the light intensity is lower than the preset threshold. During the welding process of the weld in the welding position, obtain the welding image of the welding position through the multi-angle camera group, and use the infrared laser to irradiate the welding position to obtain the infrared image and convert it into a visible light image; among them, the high-exposure mode means that the exposure time of the cameras in the multi-angle camera group is greater than or equal to 50 ms; Shady-side welding module, used to fuse the welding image and the visible light image through spatial coordinate alignment to obtain the second welding image, identify the weld position in the second welding image, and weld the target device.

6. An electronic device, characterized in that, including: A memory and a processor; The memory is used to store computer programs; The processor, when executing the computer program stored in the memory, implements the steps of a method for welding a device in a space environment as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, For storing a computer program, the computer program, when executed by a processor, implements the steps of a method for welding a device in a space environment as described in any one of claims 1 to 4.

Images