Robot system for steel structure welding

Abstract

The invention discloses a robot system for steel structure welding, relates to the field of steel structure machining equipment, and aims at solving the problem that a steel structure can be welded by collecting robot track data and a welding seam image in the welding process, constructing a track continuity evaluation model, identifying a path change trend and fusing an image structure and track change information. The position and confidence of a track interruption section are recognized, so that the high-precision recognition and automatic recovery capability after path interruption is improved; a smooth compensation track is generated by adopting third-order Bessel interpolation based on starting and ending points of an interruption section, curvature and posture consistency registration with an original track is completed, and continuity and stability of a welding path are ensured; multi-modal data such as current, voltage, temperature, acoustics and welding seam images are collected through multiple channels, welding abnormity is recognized, state feature vectors are formed, and the problems such as splashing abnormity and welding seam deviation can be found in real time; through multi-dimensional analysis and dynamic adjustment of the whole welding process, the stability, intelligence and adaptability of the welding system are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of steel structure processing equipment, and in particular to a robot system for steel structure welding. Background Art

[0002] As industrial production transitions toward automation and flexible manufacturing, robotic welding is increasingly being used in steel-related construction projects, bridge construction, heavy machinery, and other fields. Steel structures are characterized by high material strength, large workpiece dimensions, and stringent weld joint requirements. To ensure welding quality, dynamic adjustment of welding parameters based on path changes, workpiece posture deviations, and multi-source status data is required to enhance the stability and adaptability of the welding system.

[0003] Some robotic systems suitable for steel structure welding integrate offline databases or manually set process parameter libraries, enabling them to retrieve preset parameters based on the input steel material type and thickness. However, existing systems offer limited support for high-precision recognition and automatic recovery after welding path interruptions. Steel structure welds are typically long and require extremely high continuity, a drawback that can easily lead to weld discontinuities and increased rework rates. Furthermore, the ability to monitor the multi-source quality status during welding is insufficient. Steel structure welding is susceptible to current and voltage fluctuations, ambient temperature, and other factors, making it difficult to detect spatter anomalies and weld offset in real time. Furthermore, process parameter control relies on fixed value matching. Steel structure welding requires adapting to different groove structures and welding positions, and the lack of a real-time feedback closed-loop adjustment mechanism compromises welding quality stability. Due to their large size and weight, steel structure workpieces are susceptible to spatial posture changes caused by assembly errors or fixture offsets. Existing systems fail to effectively compensate for these changes, leading to trajectory deviations and quality fluctuations. Finally, the lack of a knowledge strategy library with self-learning and adaptive capabilities limits the system's ability to intelligently evolve under complex steel structure welding conditions. These issues impact the efficiency and reliability of overall control of steel structure welding. Summary of the Invention

[0004] The main purpose of the present invention is to provide a robot system for steel structure welding, which can effectively solve the technical problems in related technologies such as the inability of welding systems to accurately identify and automatically recover after path interruption, the lack of multi-source quality status monitoring, the reliance on fixed value matching for process parameter control, the failure to effectively compensate for workpiece posture changes, and the lack of self-learning ability.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] The present invention provides a robot system suitable for steel structure welding, the system comprising:

[0007] The path perception module is used to obtain robot trajectory data and weld seam images during the welding process, build a trajectory continuity evaluation model, and identify path change trends;

[0008] The interruption recognition module is used to fuse image structure and trajectory change information, and identify the location and confidence of trajectory interruption segments through image segmentation model and sliding window curvature analysis;

[0009] The recovery trajectory planning module is used to generate a smooth compensation trajectory based on the start and end points of the interrupted segment using third-order Bezier interpolation, and complete the curvature and posture consistency registration with the original trajectory;

[0010] The welding status monitoring module is used to collect current, voltage, temperature, acoustic signals and weld image data through multiple channels, identify welding anomalies through principal component analysis and normalization model, and form a status feature vector;

[0011] The process parameter control module is used to generate parameter adjustment strategies based on the state feature vector through the fuzzy controller and the time series prediction model to achieve dynamic optimization of current, voltage and wire feeding speed;

[0012] The posture recognition and adaptation module is used to collect the point cloud of the workpiece and fixture, calculate the posture offset through the point cloud matching algorithm, and apply the offset matrix to the trajectory compensation segment to achieve spatial transformation;

[0013] The knowledge base module is used to store welding parameter sample data by material type, thickness, and groove structure dimensions, optimize the parameter adjustment model based on supervised learning, and match the optimal initial parameter combination according to the task characteristics in new tasks;

[0014] The control execution module is used to fuse the trajectory recovery results, posture compensation information and parameter adjustment results to generate execution instructions, convert them into a format recognizable by the robot control system, and monitor the execution effect.

[0015] Furthermore, when the path perception module acquires robot trajectory data and weld seam images during welding, builds a trajectory continuity evaluation model, and identifies path change trends, it is specifically used to:

[0016] Real-time recording of welding trajectory points based on robot controller feedback;

[0017] Collect video streams or image frames of the welding area;

[0018] A continuity judgment model is constructed based on the trajectory point sequence to identify trajectory mutation areas;

[0019] Analyze the continuity of the weld line in the image to determine whether there is occlusion, offset or interruption;

[0020] The image information and trajectory change information are integrated to output the start and end boundaries of the interruption area and the confidence score.

[0021] Furthermore, when the interruption identification module fuses the image structure and trajectory change information and identifies the location and confidence of the trajectory interruption segment through the image segmentation model and sliding window curvature analysis, it is specifically used to:

[0022] Analyze the weld image based on the image segmentation model to identify the continuity of the weld line;

[0023] Combine the trajectory point sequence to build a continuity judgment model and locate the trajectory mutation area;

[0024] The location of the trajectory interruption segment is determined by the sliding window curvature analysis method;

[0025] Generate confidence scores for trajectory break segments and output break boundary information.

[0026] Furthermore, when the recovery trajectory planning module generates a smooth compensation trajectory based on the start and end points of the interrupted segment using third-order Bezier interpolation and completes the curvature and posture consistency registration with the original trajectory, it is specifically used to:

[0027] Locate the valid welding trajectory points adjacent to both ends of the interrupted segment;

[0028] Generate a transition curve based on the spatial relationship between the starting and ending points to ensure continuity in curvature and posture with the trajectory before and after the interruption;

[0029] The initial values ​​of welding parameters for the compensation segment are matched with those for the original trajectory segment.

[0030] Furthermore, when the welding state monitoring module performs multi-channel acquisition of current, voltage, temperature, acoustic signals and weld image data, identifies welding anomalies through principal component analysis and normalization model, and forms a state feature vector, it is specifically used to:

[0031] Collect current, voltage, temperature, penetration and weld image data;

[0032] Determine whether there is arc instability, excessive spatter, or abnormal weld offset based on the preset model;

[0033] The monitoring data is converted into quality indicator values ​​and used for subsequent feedback control.

[0034] Furthermore, when forming the state feature vector, the welding state monitoring module is specifically used to:

[0035] Perform principal component analysis on the collected current, voltage, temperature, acoustic signals and weld image data to extract key features;

[0036] Standardize key features through standardization models;

[0037] The normalized key features are integrated into the state feature vector.

[0038] Furthermore, when the process parameter control module generates a parameter adjustment strategy based on the state feature vector by combining the fuzzy controller with the time series prediction model to achieve dynamic optimization of current, voltage, and wire feeding speed, it is specifically used to:

[0039] Identify abnormal patterns corresponding to different welding state abnormalities;

[0040] According to the recognition results, the corresponding strategy is called to adjust the current, voltage, wire feeding speed and welding gun movement speed;

[0041] Update parameter settings in real time without interrupting the welding process.

[0042] Furthermore, when the posture recognition and adaptation module collects the workpiece and fixture point clouds, calculates the posture offset through the point cloud matching algorithm, and applies the offset matrix to the trajectory compensation segment to achieve spatial transformation, it is specifically used to:

[0043] Identify the spatial posture of the workpiece based on laser scanning or structured light;

[0044] Compare the recognized actual posture with the preset standard posture and calculate the six-degree-of-freedom offset;

[0045] Adjust the welding trajectory or welding gun posture according to the offset to ensure that the weld path accurately corresponds to the workpiece seam.

[0046] Furthermore, the knowledge base module is used to store welding parameter sample data according to material type, thickness, and groove structure dimensions, optimize the parameter adjustment model based on supervised learning, and match the optimal initial parameter combination according to task characteristics in a new task.

[0047] Classify and store welding parameter sample data by material type, thickness, and groove structure;

[0048] Train and optimize the parameter adjustment model through supervised learning;

[0049] Receive the task features of the new task and match the corresponding optimal initial parameter combination from the stored sample data.

[0050] Furthermore, the control execution module generates execution instructions by fusing the trajectory recovery results, posture compensation information, and parameter adjustment results, converts them into a format recognizable by the robot control system, and monitors the execution effect. Specifically, it is used to:

[0051] Receive trajectory recovery results, attitude compensation information and parameter adjustment results;

[0052] Fusing and processing the received information to generate execution instructions;

[0053] Convert the execution instructions into a format recognizable by the robot control system and send it;

[0054] Monitor the robot's execution effect on instructions in real time and provide feedback on the monitoring results.

[0055] The present invention has the following beneficial effects: the embodiment of the present invention collects robot trajectory data and weld images during the welding process, constructs a trajectory continuity evaluation model, identifies path change trends, and fuses image structure and trajectory change information to identify the position and confidence of trajectory interruption segments, thereby improving high-precision recognition and automatic recovery capabilities after path interruption; a smooth compensation trajectory is generated based on the start and end points of the interruption segment using third-order Bessel interpolation, and the curvature and posture consistency are aligned with the original trajectory to ensure the continuity and stability of the welding path; multi-channel acquisition of multi-modal data such as current, voltage, temperature, acoustics and weld images, and identification of welding anomalies through PCA and Z-score models to form a state feature vector, which can detect spatter anomalies, weld offset and other problems in real time; based on the state feature vector, through the fuzzy controller and the prediction The measurement model jointly generates a parameter adjustment strategy to achieve dynamic optimization of current, voltage, and wire feeding speed, solving the problem that process parameter control relies on fixed value matching; collects workpiece and fixture point clouds, calculates posture offset through the ICP algorithm, and applies the offset matrix to the trajectory compensation segment to achieve spatial transformation, effectively compensating for the spatial posture changes of the workpiece caused by assembly error or fixture offset; stores welding parameter sample data according to dimensions such as material type, thickness, and groove structure, optimizes the parameter adjustment model based on supervised learning, and matches the optimal initial parameter combination according to the task characteristics in the new task, thereby improving the self-learning and self-adaptation capabilities of the system; integrates trajectory recovery results, posture compensation information, and parameter adjustment results to generate execution instructions, converts them into a format recognizable by the robot control system, and monitors the execution effect, thus achieving closed-loop control of the welding process. The present invention significantly improves the stability, intelligence, and adaptability of the welding system through multi-dimensional analysis and dynamic adjustment of the entire welding process. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a schematic diagram of the overall framework of the welding system provided by an embodiment of the present invention.

[0057] Figure 2 Flowchart of the trajectory interruption identification and compensation process in an embodiment of the present invention.

[0058] Figure 3 Schematic diagram of the structure of multimodal data acquisition and welding status monitoring in an embodiment of the present invention

[0059] Figure 4 Schematic diagram of the principle of workpiece posture offset calculation and compensation in an embodiment of the present invention.

[0060] Figure 5 Flowchart of dynamic optimization and execution control of welding parameters in an embodiment of the present invention.

[0061] The accompanying drawings are numbered as follows:

[0062] 1. Path perception module; 2. Interruption recognition module; 3. Recovery trajectory planning module; 4. Welding status monitoring module; 5. Process parameter control module; 6. Posture recognition and adaptation module; 7. Knowledge base module; 8. Control execution module. DETAILED DESCRIPTION

[0063] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0064] The present invention provides a robot system suitable for steel structure welding, such as Figures 1 to 5 As shown in Figure 1, the system includes a path perception module 1, an interruption identification module 2, a recovery trajectory planning module 3, a welding state monitoring module 4, a process parameter control module 5, a posture recognition and adaptation module 6, a knowledge base module 7, and a control execution module 8. These modules work closely together through data flow and signal interaction to complete the entire process of welding tasks. The following detailed description of the specific implementation of the system is provided in conjunction with the accompanying drawings.

[0065] In practical applications, the path perception module 11 primarily acquires data related to the welding process and analyzes path changes. Based on feedback from the robot controller, it records welding trajectory points in real time. These points represent the spatial coordinates of the points the robot actually passes through during welding. For example, when welding a 10-meter-long straight weld on a steel structure, a point is recorded every 0.01 meter, resulting in a trajectory point sequence consisting of 1,000 points. Simultaneously, it captures a video stream or image frames of the welding area, typically using a high-definition industrial camera mounted near the robot's welding torch. The camera's frame rate is set to 30 frames per second, enabling clear images of the weld and surrounding area. A continuity assessment model is then constructed based on the trajectory point sequence. This model calculates the distance and angle changes between adjacent trajectory points to identify areas with sudden trajectory changes. For example, a sudden increase in the distance between two adjacent points exceeding a preset threshold of 0.05 meters, or a change in angle exceeding 5 degrees, is identified as a trajectory abrupt change. The continuity of the weld line in the image is also analyzed to determine whether there are any occlusions, offsets, or interruptions. For example, a noticeable break in the weld line or a deviation from the expected position exceeding 0.03 meters is flagged as an anomaly. Finally, the image information and trajectory change information are fused to output the start and end boundaries and confidence level of the interruption area. The higher the confidence level, the higher the accuracy of the judgment. For example, a confidence level of 0.9 means that there is a 90% certainty that the area is an interruption area.

[0066] Discontinuity Identification Module 2 integrates image structure and trajectory change information to identify trajectory discontinuities. It first analyzes the weld image based on an image segmentation model. This model can separate the weld area from the image, precisely outlining the weld contour in an image containing both the weld and the rest of the steel structure, thereby identifying the continuity of the weld line. Simultaneously, a continuity judgment model constructed using the trajectory point sequence is used to locate areas of trajectory mutation. The location of the trajectory discontinuity is then determined using sliding window curvature analysis. This involves setting a fixed-length window, such as a window containing five trajectory points, and sliding the window across the trajectory point sequence to calculate the curvature of the trajectory within each window. Curvature is a quantity that describes the degree of curvature of the trajectory. When the curvature within the window suddenly increases beyond a preset value, a trajectory discontinuity may exist. In this way, combined with the weld continuity information obtained by image segmentation, the location and confidence of the trajectory interruption segment can be identified. For example, in a trajectory, through analysis, it is found that the curvature between the 500th point and the 520th point is abnormal and the weld line in the image is broken. This interval will be determined as a discontinuity segment with a confidence level of 0.85.

[0067] The recovery trajectory planning module 3 is used to generate the compensation trajectory and connect it with the original trajectory. It will first locate the valid welding trajectory points adjacent to each other at the two ends of the interruption segment, that is, the last normal point before the start of the interruption segment and the first normal point after the end of the interruption segment. Assuming that the coordinates of the start and end points of the interruption segment are (x1, y1, z1) and (x2, y2, z2) respectively, a smooth compensation trajectory is generated based on these two points using third-order Bezier interpolation. The formula of the third-order Bezier curve is B(t) = P0*(1-t) 3 +3P1t*(1-t) 2 +3P2t 2 *(1-t)+P3*t 3 , where t ranges from 0 to 1, P0 is the starting point of the interruption segment, P3 is the end point of the interruption segment, and P1 and P2 are control points. These two control points can be determined based on the tangent direction of the original trajectory at the starting and end points. For example, if the tangent direction of the original trajectory at the starting point is along the positive x-axis, P1 can be set to a point 0.05 meters away from the starting point along the tangent direction. Similarly, P2 is set based on the tangent direction at the end point. Using this formula, when t changes from 0 to 1, a smooth curve is generated from the starting point to the end point. After generating the compensation trajectory, the curvature and posture consistency with the original trajectory must be achieved. That is, the curvature and posture of the compensation trajectory at the starting point must be the same as those of the original trajectory before the interruption, and the curvature and posture of the compensation trajectory at the end point must be the same as those of the original trajectory after the interruption. This ensures a smooth transition of the welding trajectory. Furthermore, the initial welding parameters of the compensation segment are matched to those of the original trajectory. For example, if the current of the original trajectory is 280A and the voltage is 32V, the compensation segment will also initially use similar parameters.

[0068] The welding status monitoring module 4 is responsible for collecting various data during the welding process and identifying anomalies. It collects current, voltage, temperature, acoustic signals, and weld image data. Current and voltage are acquired by sensors in the welding power supply at a sampling frequency of 1000 Hz, providing real-time current and voltage values ​​of the welding arc. Temperature is acquired by an infrared thermometer with a measurement range of 0 to 500 degrees Celsius, monitoring the temperature of the weld area. Acoustic signals are collected by a microphone installed near the welding table, capturing arc sounds and other acoustic signals during welding. Weld image data is also acquired by an industrial camera. Principal component analysis (PCA) is then performed on these collected data. PCA is a data dimensionality reduction method that extracts a small number of uncorrelated principal components from multiple correlated variables. These principal components can reflect the majority of the original data information. For example, for the three variables of current, voltage, and temperature, PCA may reveal a single principal component that explains over 85% of the variance in these three variables. The key features are then normalized using a standardization model. The standardization formula is Z = (X - μ) / σ, where X is the raw data, μ is the mean of the raw data, and σ is the standard deviation of the raw data. This results in a mean of 0 and a standard deviation of 1 for the processed data, facilitating subsequent analysis and comparison. For example, if a set of current data has a mean of 280A and a standard deviation of 10A, and one current value is 300A, the normalized value is (300 - 280) / 10 = 2. Finally, the normalized key features are integrated into a state feature vector, which contains key information that reflects the welding state. This vector can be used to identify welding anomalies, such as arc instability, excessive spatter, and weld offset.

[0069] The process parameter control module 5 adjusts welding parameters based on the welding state feature vector. It first identifies abnormal patterns corresponding to different welding state anomalies. For example, when the normalized current value in the state feature vector fluctuates significantly and the acoustic signal's characteristic value is abnormal, this corresponds to an abnormal pattern of arc instability. Based on the identification results, it then invokes corresponding strategies to adjust the current, voltage, wire feed speed, and welding gun movement speed. A fuzzy controller and a time series prediction model are used to jointly generate parameter adjustment strategies. The fuzzy controller transforms expert experience into control rules, while the time series prediction model predicts future welding conditions based on historical data. For example, if arc instability is identified, the fuzzy controller will initially determine based on preset rules that the current needs to be reduced by 5% to 10%. The time series prediction model, based on the current trend over the past 10 seconds, predicts that an 8% current reduction will stabilize the arc within the next 5 seconds. Ultimately, the current is adjusted from 300A to 276A. Furthermore, parameter settings are updated in real time without interrupting the welding process, ensuring the continuity of the welding process.

[0070] Posture recognition and adaptation module 6 is used to adjust trajectory deviations caused by workpiece or fixture posture deviations. It uses laser scanning or structured light to identify the workpiece's spatial posture. The laser scanner's scanning range can cover the entire workpiece, with a scanning accuracy of 0.1 mm, and can acquire a large amount of point cloud data from the workpiece surface. Structured light projects a specific pattern of light onto the workpiece, then uses a camera to capture the deformation of the pattern to acquire point cloud data. After acquiring the point cloud data, the identified actual posture is compared with a preset standard posture. A point cloud matching algorithm is used to calculate the six-degree-of-freedom offset, which includes translational offsets along the x, y, and z axes and rotational offsets around them. For example, if the actual workpiece is translated by 2 mm along the x-axis and rotated by 0.5 degrees around the y-axis, this is the calculated offset. The welding trajectory or welding gun posture is then adjusted based on the offset. The offset matrix is ​​applied to the trajectory compensation segment to achieve spatial transformation, ensuring that the weld path accurately corresponds to the workpiece seam.

[0071] The knowledge base module 7 is used to store and optimize welding parameters. It stores welding parameter sample data according to the dimensions of material type, thickness, and groove structure. For example, the material type is divided into Q345 steel, Q235 steel, etc., with thickness ranging from 5 mm to 50 mm, and groove structures including V-type, U-type, I-type, etc. Each combination corresponds to welding parameters that have been verified in practice, such as current, voltage, wire feed speed, etc. The parameter adjustment model is optimized based on supervised learning. Supervised learning is a machine learning method that uses labeled training data to train the model. The training data here is the welding parameters and corresponding welding quality results under different working conditions. Through training, the model can predict better welding parameters according to the working conditions. In a new task, according to the task characteristics, such as the material is Q345 steel, the thickness is 20 mm, and the V-groove is matched from the knowledge base with the optimal initial parameter combination, such as current 290A, voltage 34V, and wire feed speed 8 meters per minute.

[0072] The control execution module 8 is the system's core execution engine. It integrates trajectory recovery results, posture compensation information, and parameter adjustment results to generate execution instructions. These instructions include information such as the robot's motion trajectory, welding parameters, and welding gun posture. These instructions are then converted into a format recognizable by the robot control system. Different robot control systems have different instruction formats, for example, one brand of robot uses specific pulse instructions to control movement. Simultaneously, it monitors execution performance, receiving actual motion and welding parameter data from the robot and comparing it with the expected data in the instructions. If any deviation exceeds the allowable range, feedback is promptly provided to other modules for adjustment.

[0073] The following example illustrates the application of this system using the welding of a steel box girder for a large steel bridge. The steel box girder is made of Q345 steel, 20 mm thick, with a V-groove angle of 60 degrees. The weld length is 8 meters. During the welding process, the path perception module 1 uses the robot controller to acquire trajectory points, recording one point every 0.01 meters, while an industrial camera captures weld images. If the trajectory is interrupted due to slight workpiece movement, the interruption recognition module 2 uses image segmentation to detect a break in the weld image 3 meters from the starting point. Combined with sliding window curvature analysis, it determines that the interruption segment extends from 3 meters to 3.05 meters, with a confidence level of 0.92. The recovery trajectory planning module 3 uses third-order Bezier interpolation based on the start and end points of the interruption segment to generate a compensation trajectory. Control points are set according to the tangent direction of the original trajectory. The resulting compensation trajectory maintains the same curvature and posture as the original trajectory. The welding state monitoring module 4 detects current fluctuations between 280A and 300A, voltage around 32V, and temperature around 350°C. Principal component analysis and normalization generate a state feature vector, identifying slightly large current fluctuations but no obvious abnormalities. The process parameter control module 5, based on this state feature vector and using a fuzzy controller and time series prediction model, determines that no significant parameter adjustments are required and simply adjusts the wire feed speed by 0.2 meters per minute. The posture recognition and adaptation module 6 detects a 1mm translational offset along the workpiece's z-axis through laser scanning. After calculating the offset matrix, it adjusts the trajectory accordingly. Based on the characteristics of Q345 steel, 20mm thickness, and V-groove, the knowledge base module 7 matches the initial current of 290A, voltage of 34V, and wire feed speed of 8 meters per minute. After system optimization, these parameters are applied to the welding process. The control execution module 8 integrates all this information to generate instructions, converts them into a format recognizable by the robot, controls the robot to complete the welding process, and monitors the welding process in real time to ensure that the welding quality meets requirements.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention. The basic principles and main features of the present invention and the advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited by the above-mentioned embodiments. The above-mentioned embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the invention to be protected. The scope of protection claimed in the present invention is defined by the attached claims and their equivalents.

Claims

1. A robot system suitable for steel structure welding, characterized in that: The system comprises: Path perception module (1), which obtains robot trajectory data and weld seam images during welding, builds a trajectory continuity evaluation model, and identifies path change trends; The interruption identification module (2) integrates the image structure and trajectory change information, and identifies the location and confidence of the trajectory interruption segment through image segmentation model and sliding window curvature analysis; The recovery trajectory planning module (3) generates a smooth compensation trajectory based on the start and end points of the interrupted segment using third-order Bezier interpolation, and completes the curvature and posture consistency registration with the original trajectory; The welding state monitoring module (4) collects current, voltage, temperature, acoustic signal and weld image data through multiple channels, identifies welding anomalies through principal component analysis and standardization model, and forms a state feature vector; The process parameter control module (5) generates a parameter adjustment strategy based on the state characteristic vector through a fuzzy controller and a time series prediction model to achieve dynamic optimization of current, voltage and wire feeding speed; The posture recognition and adaptation module (6) is used to collect the point cloud of the workpiece and the fixture, calculate the posture offset through the point cloud matching algorithm, and apply the offset matrix to the trajectory compensation segment to realize the spatial transformation; The knowledge base module (7) stores welding parameter sample data according to material type, thickness, and groove structure dimensions, optimizes the parameter adjustment model based on supervised learning, and matches the optimal initial parameter combination according to the task characteristics in the new task; The control execution module (8) generates an execution instruction by fusing the trajectory recovery result, the posture compensation information and the parameter adjustment result, converts it into a format that can be recognized by the robot control system, and monitors the execution effect.

2. A robot system suitable for steel structure welding according to claim 1, characterized in that: The path perception module (1) is specifically used to obtain robot trajectory data and weld seam images during welding, build a trajectory continuity evaluation model, and identify path change trends: Real-time recording of welding trajectory points based on robot controller feedback; Collect video streams or image frames of the welding area; A continuity judgment model is constructed based on the trajectory point sequence to identify trajectory mutation areas; Analyze the continuity of the weld line in the image to determine whether there is occlusion, offset or interruption; The image information and trajectory change information are integrated to output the start and end boundaries of the interruption area and the confidence score.

3. A robot system suitable for steel structure welding according to claim 1, characterized in that: The interruption identification module (2) is specifically used to: Analyze the weld image based on the image segmentation model to identify the continuity of the weld line; Combine the trajectory point sequence to build a continuity judgment model and locate the trajectory mutation area; The location of the trajectory interruption segment is determined by the sliding window curvature analysis method; Generate confidence scores for trajectory break segments and output break boundary information.

4. A robot system suitable for steel structure welding according to claim 1, characterized in that: The restoration trajectory planning module (3) is specifically used to generate a smooth compensation trajectory based on the start and end points of the interrupted segment using third-order Bezier interpolation and complete the curvature and posture consistency registration with the original trajectory: Locate the valid welding trajectory points adjacent to both ends of the interrupted segment; Generate a transition curve based on the spatial relationship between the starting and ending points to ensure continuity in curvature and posture with the trajectory before and after the interruption; The initial values ​​of welding parameters for the compensation segment are matched with those for the original trajectory segment.

5. A robot system suitable for steel structure welding according to claim 1, characterized in that: The welding state monitoring module (4) is specifically used to: when executing multi-channel acquisition of current, voltage, temperature, acoustic signal and weld image data, identifying welding anomalies through principal component analysis and standardization model, and forming a state feature vector; Collect current, voltage, temperature, penetration and weld image data; Determine whether there is arc instability, excessive spatter, or abnormal weld offset based on the preset model; The monitoring data is converted into quality indicator values ​​and used for subsequent feedback control.

6. A robot system suitable for steel structure welding according to claim 5, characterized in that: When forming a state feature vector, the welding state monitoring module (4) is specifically used to: Perform principal component analysis on the collected current, voltage, temperature, acoustic signals and weld image data to extract key features; Standardize key features through standardization models; The normalized key features are integrated into the state feature vector.

7. A robot system suitable for steel structure welding according to claim 1, characterized in that: The process parameter control module (5) is specifically used to: Identify abnormal patterns corresponding to different welding state abnormalities; According to the recognition results, the corresponding strategy is called to adjust the current, voltage, wire feeding speed and welding gun movement speed; Update parameter settings in real time without interrupting the welding process.

8. The robot system suitable for steel structure welding according to claim 1, characterized in that: The posture recognition and adaptation module (6) is specifically used for collecting the workpiece and fixture point clouds, calculating the posture offset through the point cloud matching algorithm, and applying the offset matrix to the trajectory compensation segment to realize the space transformation. Identify the spatial posture of the workpiece based on laser scanning or structured light; Compare the recognized actual posture with the preset standard posture and calculate the six-degree-of-freedom offset; Adjust the welding trajectory or welding gun posture according to the offset to ensure that the weld path accurately corresponds to the workpiece seam.

9. The robot system suitable for steel structure welding according to claim 1, characterized in that: The knowledge base module (7) is specifically used to store welding parameter sample data according to material type, thickness, and groove structure dimensions, optimize the parameter adjustment model based on supervised learning, and match the optimal initial parameter combination according to task characteristics in a new task: Classify and store welding parameter sample data by material type, thickness, and groove structure; Train and optimize the parameter adjustment model through supervised learning; Receive the task features of the new task and match the corresponding optimal initial parameter combination from the stored sample data.

10. The robot system suitable for steel structure welding according to claim 1, characterized in that: The control execution module (8) is specifically used to: generate an execution instruction by executing the fusion trajectory recovery result, the posture compensation information and the parameter adjustment result, convert it into a format recognizable by the robot control system, and monitor the execution effect; Receive trajectory recovery results, attitude compensation information and parameter adjustment results; Fusing and processing the received information to generate execution instructions; Convert the execution instructions into a format recognizable by the robot control system and send it; Monitor the robot's execution effect on instructions in real time and provide feedback on the monitoring results.

Images