Intelligent welding robot control system

Abstract

The invention provides an intelligent welding robot control system, belongs to the field of artificial intelligence, and is suitable for welding process control in industrial automatic production. Through integration of a multi-modal data fusion technology, a complex deep learning algorithm and a real-time path optimization and parameter adaptive control technology, all-directional intelligent monitoring and optimization control of the welding process are realized. The method is particularly suitable for the intelligent production environment needing high-precision and high-efficiency welding operation, the automation level and the product quality of a production line can be remarkably improved, and the production efficiency and the competitiveness of the intelligent manufacturing field are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to related technologies such as machine vision, computer vision, intelligent video analysis, and more particularly to an intelligent welding robot control system. Background Art

[0002] With the continuous acceleration of urbanization, the application of intelligent production and automation technology in the industrial field is becoming more and more extensive. In the modern industrial production process, welding, as a key processing technology, is widely used in various fields of manufacturing. However, due to the lack of intelligent monitoring means, traditional welding methods often face problems such as insufficient welding accuracy, untimely defect detection and inflexible welding parameter adjustment, which makes it difficult to meet the requirements of efficient and high-quality production.

[0003] In the context of intelligent production, intelligent welding robots have gradually become the main equipment in industrial welding due to their high efficiency, precision and flexibility. However, the current intelligent welding technology mainly relies on a single algorithm model for welding control and defect detection, which is difficult to adapt to different welding conditions and complex production environments. These technologies have the following main problems in practical applications:

[0004] Insufficient welding defect detection: Traditional welding defect detection methods mainly rely on single sensor data or visual inspection technology, which makes it difficult to fully monitor defects in the welding process. Especially when facing different materials and complex weld structures, the detection accuracy and reliability need to be improved.

[0005] Limited welding path planning: Existing welding path planning technologies mostly use fixed path planning algorithms, which lack the ability to respond to dynamic changes in the environment. Especially in complex production environments, traditional algorithms are difficult to achieve real-time path optimization and obstacle avoidance, affecting welding efficiency and quality.

[0006] Inflexible welding parameter adjustment: Traditional welding parameter control methods rely on preset parameter adjustments and lack the ability to respond to real-time data during the welding process. During the welding process, if parameter fluctuations or working conditions change, it is difficult to achieve adaptive adjustment of parameters, resulting in unstable welding quality. Summary of the invention

[0007] In order to solve the above technical problems, the present invention provides an intelligent welding robot control system, which uses multimodal data fusion technology, complex deep learning algorithms and adaptive fuzzy control technology to solve the problems of insufficient welding accuracy, incomplete defect detection and inflexible parameter adjustment in the prior art. The system realizes comprehensive detection of welding defects, dynamic optimization of welding paths and adaptive adjustment of welding parameters by real-time monitoring and intelligent analysis of data in the welding process. By integrating multimodal data and complex deep learning algorithms, it realizes comprehensive monitoring and optimization of the welding process, improves the robustness and adaptability of the system, overcomes many deficiencies in the prior art, and significantly improves the welding efficiency and quality in the intelligent production environment.

[0008] The technical solution of the present invention is:

[0009] An intelligent welding robot control system includes a multimodal data fusion module, a complex deep learning algorithm module, a real-time path optimization module, an adaptive fuzzy control module and a feedback regulation module.

[0010] in,

[0011] The multimodal data fusion module includes an optical image processing unit, an infrared thermal imaging analysis unit and an ultrasonic data processing unit, and can fuse and extract features from data from different sensors to form a unified feature map.

[0012] The complex deep learning algorithm module adopts the visual transformer (ViT) model, which is trained with multimodal data sets and can extract and fuse key features from multiple sensor data to generate a comprehensive feature map for defect detection.

[0013] The non-maximum suppression algorithm screens and suppresses candidate frames of welding defects according to the fused comprehensive feature map, and uses weighted parameters to correct the confidence of the candidate frames, thereby improving the accuracy and reliability of defect detection.

[0014] The real-time path optimization module performs multi-objective optimization through genetic algorithms, comprehensively considers welding quality, path length and energy consumption, generates the optimal welding path, adapts to complex production environments, and ensures the continuity and safety of the welding path through dynamic monitoring and real-time adjustment.

[0015] The adaptive fuzzy control module monitors and dynamically adjusts key parameters in the welding process in real time based on fuzzy logic rules, and realizes adaptive control of welding current, voltage, welding speed and welding time through online learning and adjustment.

[0016] The feedback regulation module continuously optimizes the welding parameters based on real-time monitoring data through a closed-loop control mechanism to ensure the stability of the welding process and improve the welding quality.

[0017] Furthermore,

[0018] (1) Obtain multimodal data, run the multimodal data fusion module, combine optical images, infrared thermal imaging and ultrasonic detection data, and perform data fusion and feature extraction.

[0019] (2) Through the complex deep learning algorithm module, the fused feature data is analyzed to detect defects in the welding process, and the visual transformer (ViT) model is used to perform multi-level feature extraction to generate a comprehensive feature map.

[0020] (3) Using the comprehensive feature map obtained in step 2, a non-maximum suppression algorithm is applied to screen and locate the detected defects to determine the location and type of welding defects.

[0021] (4) Obtain the three-dimensional model data of the welding workpiece, use the real-time path optimization module, and use the genetic algorithm to perform multi-objective optimization to generate the optimal welding path. Dynamically monitor environmental changes during the welding process and adjust the welding path.

[0022] (5) Based on the optimal path data obtained in step 4, the adaptive fuzzy control module is run to monitor and dynamically adjust the key parameters of the welding process such as current, voltage, welding speed, and welding time in real time.

[0023] (6) Through the feedback adjustment module, the welding parameters are adjusted in a closed loop based on real-time data to optimize the welding process and ensure welding quality and stability.

[0024] in,

[0025] The non-maximum suppression algorithm comprises the following steps:

[0026] (1) Construct an initial candidate box set H, which contains all candidate boxes; construct a set M to store the optimal box and initialize it to an empty set;

[0027] (2) Search all candidate boxes and find a candidate box m with the highest correction confidence. Move it from set H to set M, traverse all candidate boxes in H, and calculate the overlap rate (IoU) with m. If it exceeds the threshold, it is considered to overlap with m and this box is removed from H.

[0028] (3) Repeat step 2 until the set H is empty, and the candidate boxes in the set M are the final returned targets.

[0029] Furthermore,

[0030] During the modified confidence calculation process, the weight of each algorithm module is calculated and adjusted through the respective detection confidence to improve the comprehensive reliability and accuracy of defect detection.

[0031] Improve the comprehensiveness and accuracy of welding defect detection: The present invention integrates multiple sensor data such as optical images, infrared thermal imaging and ultrasonic detection, and performs multi-level feature extraction and analysis through multimodal data fusion technology based on visual transformer (ViT), which solves the problems of incomplete defect detection and high missed detection rate in the prior art, and significantly improves the accuracy and sensitivity of welding defect detection.

[0032] Optimize the flexibility and adaptability of welding path planning: The present invention adopts a multi-objective optimization algorithm, performs path optimization through a genetic algorithm, and combines it with dynamic obstacle avoidance technology to achieve real-time planning and adjustment of the welding path, solving the problems of limited path planning and difficulty in coping with complex production environments in the prior art, ensuring the continuity and safety of the welding process and improving production efficiency.

[0033] Realize adaptive adjustment of welding parameters: The present invention uses an adaptive fuzzy control algorithm to dynamically adjust the current, voltage, welding speed and welding time in the welding process, and adjusts the control parameters online according to the real-time monitored welding data, thereby solving the problem of inflexible welding parameter adjustment in the prior art and ensuring the stability and consistency of welding quality.

[0034] Improve the robustness and adaptability of the system: Based on multimodal data and complex deep learning algorithms, the present invention integrates models through the idea of ​​non-maximum suppression (NMS), optimizes the adaptability of the welding robot under different welding conditions and materials, and solves the problems of poor model adaptability and difficulty in coping with complex welding scenarios in the prior art, significantly improving the robustness and adaptability of the system.

[0035] The present invention constructs an intelligent welding robot control system that can efficiently adapt to complex production environments and realize all-round intelligent monitoring and optimization control of the welding process, meeting the needs of modern industrial production for high-precision and high-quality welding operations, and providing a set of efficient and reliable solutions for the field of intelligent manufacturing.

[0036] The beneficial effects of the present invention are

[0037] The solution of the present invention can calculate the shooting angle and shooting distance of the camera by adopting a human feature detection algorithm, and perform model fusion according to the shooting angle and distance. The obtained integrated model has better compatibility and greater advantages in accuracy.

[0038] The present invention adopts the weighted non-maximum suppression method for model integration to solve the problem that a single model is easily unstable due to changes in shooting angle and distance, and can meet the needs of regional population counting in various scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the workflow of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0041] The present invention proposes a system and method based on intelligent welding robot control, which realizes all-round intelligent monitoring and optimization control of the welding process by integrating multimodal data fusion technology, complex deep learning algorithm and adaptive fuzzy control technology. The technical solution mainly includes three core parts: welding defect detection based on multimodal data fusion, welding path control based on multi-objective optimization, and welding parameter adjustment based on adaptive fuzzy control.

[0042] Welding defect detection based on multimodal data fusion:

[0043] The present invention utilizes multiple sensor data such as optical images, infrared thermal imaging and ultrasonic detection, combined with complex deep learning algorithms based on Vision Transformer (ViT) to perform multi-level feature extraction to achieve high-precision detection of welding defects. The process includes the following key steps:

[0044] First, the multimodal data of the welding process is obtained through optical cameras, infrared sensors and ultrasonic probes. After the data is normalized by the preprocessing module, it forms a standardized input format. Then, the ViT model is used to process the multimodal data. The model uses the self-attention mechanism to perform global correlation analysis on the features of each modal data, extract key feature points, and generate a comprehensive feature map.

[0045] During the feature extraction and fusion process, the ViT model can capture the subtle differences and associations between different modal data, and the generated comprehensive feature map can fully reflect the surface and internal conditions of the weld. Finally, based on the non-maximum suppression (NMS) algorithm, the candidate welding defects are suppressed and screened to accurately locate the location and type of welding defects. By integrating the advantages of multimodal data, the algorithm significantly improves the accuracy and comprehensiveness of welding defect detection, and can monitor the defects that may occur in the welding process in real time, reducing false detection and missed detection problems.

[0046] Multi-objective optimization of welding path control:

[0047] The present invention proposes a welding path control system based on a multi-objective optimization algorithm, which aims to optimize the motion path of the welding robot arm and improve welding efficiency and quality. The implementation steps of the path control system are as follows:

[0048] First, the system acquires the 3D model data of the welding workpiece, and preliminarily plans the welding path through the path planning module to generate an initial welding path. The initial path determines the starting and ending points of the welding by simulating the complexity of the welding process by computer, and forms a preliminary motion path planning.

[0049] In the path optimization stage, the genetic algorithm (GA) is used to perform multi-objective optimization of the welding path, taking into account factors such as welding quality, path length and energy consumption. The genetic algorithm iteratively optimizes the welding path through selection, crossover and mutation operations to achieve a balance between various optimization objectives. This process combines the welding process requirements and the complexity of the production environment to generate the optimal welding path.

[0050] During the actual welding process, the welding environment is dynamically monitored through the real-time monitoring module, and the sensor detects environmental changes and potential obstacles in the welding area. Based on dynamic obstacle avoidance technology, the system can update the welding path in real time to avoid obstacles affecting the welding process and ensure the continuity and safety of the welding process. Finally, after multi-objective optimization and dynamic adjustment, the welding path is fully optimized, which improves welding efficiency, reduces energy consumption, and ensures the stability of welding quality.

[0051] Adaptive fuzzy control of welding parameters adjustment:

[0052] The present invention adopts an adaptive fuzzy control algorithm to adjust key parameters in the welding process in real time to ensure the stability and quality of the welding process.

[0053] The technical implementation of adaptive fuzzy control includes the following steps:

[0054] First, sensors are used to monitor the changes in key parameters during welding, including welding current, voltage, welding speed, welding time, etc. The acquired real-time parameter data is processed by the data preprocessing module to form standardized input data, which is then passed to the fuzzy controller.

[0055] In the fuzzy control stage, the adaptive fuzzy control algorithm is used to analyze the real-time parameter data. The fuzzy controller generates a preliminary control strategy based on fuzzy logic rules and dynamically adjusts the welding parameters. The adaptive control algorithm can learn and adjust the control rules online according to the real-time data changes during the welding process and optimize the output of the fuzzy controller.

[0056] Finally, the system uses the feedback control module to perform closed-loop adjustment of welding parameters and optimizes parameter settings in real time during the welding process. The adaptive fuzzy control algorithm can quickly respond to parameter fluctuations during the welding process and dynamically adjust key parameters such as welding current and voltage to ensure the stability and consistency of the welding process. This method improves welding quality and reduces welding defects through online monitoring and real-time adjustment, and adapts to the needs of different materials and complex welding conditions.

[0057] Further technical implementation details:

[0058] 1) Multimodal data fusion feature extraction:

[0059] The Vision Transformer (ViT) model is used to train with a self-built multimodal data set, which can simultaneously process optical images, infrared thermal imaging, and ultrasonic data. The trained ViT model extracts and fuses features from multimodal data to generate a comprehensive feature map, which significantly improves the accuracy of welding defect detection.

[0060] 2) Multi-objective optimization algorithm:

[0061] Genetic Algorithm (GA) is used for multi-objective optimization. The welding path is optimized through selection, crossover and mutation operations. Factors such as welding quality, path length and energy consumption are comprehensively considered to generate the optimal welding path.

[0062] 3) Adaptive fuzzy control:

[0063] The adaptive fuzzy controller dynamically optimizes the welding parameters through online learning and adjustment based on fuzzy logic rules. The controller adjusts the fuzzy control rules according to real-time monitoring data to ensure parameter stability and accuracy during the welding process.

[0064] 4) Improvement of non-maximum suppression algorithm:

[0065] The weighted non-maximum suppression (NMS) algorithm is introduced to correct the confidence of the candidate box. Combined with the comprehensive feature map after multimodal data fusion, welding defects are accurately detected, which optimizes the accuracy and robustness of the detection results.

[0066] The present invention realizes comprehensive intelligent monitoring and optimization control of the welding process by integrating multimodal data fusion technology, complex deep learning algorithm and adaptive fuzzy control technology, significantly improves the welding accuracy and quality, is suitable for intelligent welding application scenarios in modern industrial production, and provides an efficient and reliable intelligent welding solution.

[0067] The intelligent welding robot control system integrates three core algorithms, including a welding defect detection algorithm based on multimodal data fusion, a welding path control algorithm based on multi-objective optimization, and a welding parameter adjustment algorithm based on adaptive fuzzy control. It is suitable for comprehensive monitoring and optimization of the welding process, and can effectively improve welding accuracy and reduce welding defects.

[0068] The welding defect detection method uses multimodal data fusion technology and combines optical images, infrared thermal imaging and ultrasonic detection. Multi-level feature extraction is performed through the Vision Transformer (ViT) based on the attention mechanism, and low-level edge detection and texture analysis are further combined to achieve high-precision detection of surface and internal defects of welds. The ViT model uses the self-attention mechanism to capture image features globally, which can effectively extract and integrate complex feature relationships in multimodal data, significantly improve the accuracy and sensitivity of defect detection, and reduce the problems of false detection and missed detection.

[0069] The welding path control system adopts a multi-objective optimization algorithm, comprehensively considers factors such as welding quality, path length and energy consumption, and performs multi-objective optimization through a genetic algorithm to generate an optimal welding path. The system uses dynamic obstacle avoidance technology to achieve real-time detection and path adjustment of obstacles that may appear during the welding process, ensuring the continuity and safety of the welding process. This method improves welding efficiency, reduces energy consumption, and ensures the stability of product quality.

[0070] The welding parameter adjustment system adopts an adaptive fuzzy control algorithm to dynamically adjust the current, voltage, welding speed and welding time during the welding process. By real-time monitoring the changes of various parameters during the welding process, the fuzzy control rules are adaptively adjusted to achieve adaptive control of different welding conditions, improve the stability and flexibility of the welding process, and can quickly respond to fluctuations in parameters during the welding process, optimize welding effects, and reduce welding defects.

[0071] The present invention realizes all-round intelligent control and monitoring of the welding process by integrating the visual transformer (ViT) algorithm based on the attention mechanism, the multi-objective optimization algorithm and the adaptive fuzzy control algorithm, improves the robustness and welding accuracy of the system, and is suitable for a variety of complex welding environments and materials, providing an efficient and reliable solution for modern industrial welding. The integrated system is much better than the traditional single algorithm system in welding accuracy, defect detection and welding efficiency.

[0072] The above description is only a preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, and is not used to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. An intelligent welding robot control system, characterized in that: include: The multimodal data fusion module, including an optical image processing unit, an infrared thermal imaging analysis unit, and an ultrasonic data processing unit, fuses and extracts features from data from different sensors to form a unified feature map; Complex deep learning algorithm module, using the visual transformer ViT model, trained with multimodal data sets, extracts and fuses key features from multiple sensor data, and generates comprehensive feature maps for defect detection; The real-time path optimization module uses genetic algorithms to perform multi-objective optimization, comprehensively considers welding quality, path length and energy consumption, generates the optimal welding path, adapts to complex production environments, and ensures the continuity and safety of the welding path through dynamic monitoring and real-time adjustment; Adaptive fuzzy control module, which monitors and dynamically adjusts key parameters in the welding process in real time based on fuzzy logic rules, and realizes adaptive control of welding current, voltage, welding speed and welding time through online learning and adjustment; The feedback adjustment module continuously optimizes the welding parameters based on real-time monitoring data through a closed-loop control mechanism to ensure the stability of the welding process and improve the welding quality.

2. The system according to claim 1, characterized in that , The working process is as follows: (1) Obtain multimodal data, run the multimodal data fusion module, combine optical images, infrared thermal imaging and ultrasonic detection data, and perform data fusion and feature extraction; (2) Through the complex deep learning algorithm module, the fused feature data is analyzed to detect defects in the welding process, and the visual transformer (ViT) model is used to perform multi-level feature extraction to generate a comprehensive feature map; (3) Using the comprehensive feature map obtained in step (2), a non-maximum suppression algorithm is applied to screen and locate the detected defects to determine the location and type of the welding defects; (4) Obtain the three-dimensional model data of the welding workpiece, use the real-time path optimization module, use the genetic algorithm to perform multi-objective optimization, generate the optimal welding path, and dynamically monitor environmental changes during the welding process to adjust the welding path; (5) running an adaptive fuzzy control module based on the optimal path data obtained in step (4) to monitor and dynamically adjust the parameters during the welding process in real time; (6) Through the feedback adjustment module, the welding parameters are adjusted in a closed loop based on real-time data to optimize the welding process and ensure welding quality and stability.

3. The system according to claim 2, characterized in that In step (2), the data is normalized after preprocessing to form a standardized input format; then, the ViT model is used to process the multimodal data. The model uses the Self-Attention Mechanism to perform global correlation analysis on the features of each modal data, extract key feature points, and generate a comprehensive feature map.

4. The system according to claim 2, characterized in that During the actual welding process, the welding environment is dynamically monitored, and sensors detect environmental changes and potential obstacles in the welding area. Based on dynamic obstacle avoidance technology, the welding path is updated in real time. Finally, after multi-objective optimization and dynamic adjustment, the welding path is fully optimized.

5. The system according to claim 2, characterized in that Adaptive fuzzy control includes the following steps: First, the sensors are used to monitor the changes of key parameters in the welding process, including welding current, voltage, welding speed and welding time. The acquired real-time parameter data is processed by the data preprocessing module to form standardized input data, which is then passed to the fuzzy controller. In the fuzzy control stage, the adaptive fuzzy control algorithm is used to analyze the real-time parameter data; the fuzzy controller generates a preliminary control strategy based on fuzzy logic rules and dynamically adjusts the welding parameters; The adaptive control algorithm learns and adjusts the control rules online according to the real-time data changes during the welding process to optimize the output of the fuzzy controller; Finally, the welding parameters are adjusted in a closed loop through feedback control to optimize the parameter settings during the welding process in real time.

6. The system according to claim 2, characterized in that The non-maximum suppression algorithm screens and suppresses the candidate frames of welding defects according to the fused comprehensive feature map, and uses weighted parameters to correct the confidence of the candidate frames to improve the accuracy and reliability of defect detection.

7. The system according to claim 6, characterized in that The non-maximum suppression algorithm comprises the following steps: (1) Construct an initial candidate box set H, which contains all candidate boxes; construct a set M to store the optimal box and initialize it to an empty set; (2) Search all candidate boxes and find the candidate box m with the highest correction confidence. Move it from set H to set M, traverse all candidate boxes in H, and calculate the overlap rate IoU with m. If it exceeds the threshold, it is considered to overlap with m and this box is removed from H. (3) Repeat step (2) until the set H is empty, and the candidate boxes in the set M are the final returned targets.

8. The system according to claim 7, characterized in that During the modified confidence calculation process, the weight of each algorithm module is calculated and adjusted through the respective detection confidence to improve the comprehensive reliability and accuracy of defect detection.

Images