Demonstration-free welding robot dynamic energy-saving path planning method and system

Abstract

The invention provides a teaching-free welding robot dynamic energy-saving path planning method and system, and relates to the technical field of industrial robos.The method comprises the steps that 1, a workpiece is scanned through a laser vision device, weld joint coordinates, obstacle information and workpiece thermal deformation data are obtained, weld joint feature points are extracted, and the weld joint feature points are obtained; establishing a mapping relation between a workpiece coordinate system and a robot base coordinate system; step 2, based on the mapping relationship, in combination with a thermal deformation prediction mechanism and sensor feedback data, implanting a dynamic response node in the initial path, calculating path point offset by using the feedback data, and generating an anti-offset path; and step 3, aiming at the anti-offset path, constructing a multi-dimensional collaborative evaluation mechanism, and carrying out synchronous adjustment and comprehensive balance on the path length, the energy consumption of the whole welding process and the welding quality parameters so as to obtain a multi-dimensional collaborative evaluation result. Weld joint deformation, clamp offset and obstacle displacement are sensed in real time, the path is dynamically planned, and the flexibility of path planning is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial robots, and in particular to a dynamic energy-saving path planning method and system for a teaching-free welding robot. Background Art

[0002] In the welding of pressure vessel cylinders, traditional welding robot path planning technology has some shortcomings. Most pressure vessel cylinders are large in size and thick in wall thickness. During the welding process, the cylinder will produce certain deformation due to continuous heat input. For example, when welding the circumferential weld of the cylinder, as the welding progresses, the local temperature of the cylinder rises, and slight radial contraction or axial bending may occur. Traditional path planning mostly pre-sets a fixed trajectory. If this dynamic deformation is not taken into account, when the cylinder produces a radial offset of 1-2mm due to thermal deformation, the robot still welds according to the initial path, which may cause the weld to deviate from the correct position, affecting the sealing and structural strength of the welding. In addition, in order to correct the deviation, the robot may frequently adjust its posture, which indirectly increases energy consumption.

[0003] In addition, traditional technologies perform poorly in multi-variable coordinated adjustment. When welding the connecting welds between the cylinder and the head, if the robot adopts a fast broken-line motion trajectory in order to shorten the path length, it may cause the welding arc to be unstable and defects such as porosity and slag inclusions to appear. If the pursuit of welding quality is too slow and a welding speed is too slow, it will not only extend the welding time, but also cause the robot to consume more energy during long-term work, making it difficult to achieve a good balance between energy saving and quality in the dynamic welding process. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a dynamic energy-saving path planning method and system for a teaching-free welding robot, so as to realize dynamic energy-saving path planning, reduce energy consumption, and improve welding efficiency and quality.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows: In a first aspect, a dynamic energy-saving path planning method for a teaching-free welding robot is provided, the method comprising: Step 1: Scan the workpiece with a laser vision device to obtain weld coordinates, obstacle information, and workpiece thermal deformation data, extract weld feature points, and establish a mapping relationship between the workpiece coordinate system and the robot base coordinate system; Step 2: Based on the mapping relationship, combined with the thermal deformation prediction mechanism and sensor feedback data, dynamic response nodes are implanted in the initial path and the feedback data is used to calculate the path point offset to generate an anti-drift path; Step 3: Build a multi-dimensional collaborative evaluation mechanism for the anti-drift path, synchronously adjust and comprehensively balance the path length, energy consumption of the entire welding process, and welding quality parameters to obtain a multi-dimensional collaborative evaluation result; Step 4: Based on the multi-dimensional collaborative evaluation results, a global path search is performed using a bidirectional tree expansion. When weld deformation, fixture offset, and obstacle displacement exceed the set threshold, a local re-path planning mechanism is triggered. Step 5: Based on the re-planning mechanism, the dynamic deformation characteristics of the weld area are collected in real time through a multi-source deformation monitoring device to generate deformation-related data. The dynamic compensation amount is determined based on the distribution characteristics of the deformation-related data, and the multi-dimensional collaborative evaluation parameters are updated in real time to drive the dynamic energy-saving control of the welding parameters and motion trajectory.

[0006] Furthermore, based on the mapping relationship, combined with the thermal deformation prediction mechanism and sensor feedback data, dynamic response nodes are implanted in the initial path and the feedback data is used to calculate the path point offset to generate an anti-drift path, including: Based on the mapping relationship, combined with the thermal expansion characteristics of the material and the heat input characteristics during welding, the expected deformation of the characteristic points on the initial weld path is calculated; Based on the expected deformation, the coordinates of the feature points of the initial path are reversely superimposed on the direction opposite to the deformation to generate a thermal deformation pre-compensation path; Based on the thermal deformation pre-compensation path, the laser vision sensor collects the actual deformation data of the workpiece during welding online, and performs a real-time spatial error comparison between the actual deformation data and the expected deformation to obtain the error comparison result; Based on the error comparison results, the curvature mutation area and the error exceeding limit point are marked in the pre-compensation path, and the dynamic response node is implanted at the marked position to calculate the three-dimensional spatial position adjustment amount at the dynamic response node; Based on the three-dimensional spatial position adjustment amount, the spatial coordinates of the corresponding nodes in the pre-compensation path are corrected in real time to generate a thermal deformation-resistant offset path.

[0007] Furthermore, based on the error comparison results, the curvature mutation area and the error exceeding limit point are marked in the pre-compensation path, and a dynamic response node is implanted at the marked position. The three-dimensional spatial position adjustment amount at the dynamic response node is calculated, including: For the marked deformation error exceeding limit point, the position compensation weighting coefficient of the deformation error exceeding limit point is calculated based on the curvature change rate of the adjacent path segments and the thermal deformation gradient at the deformation error exceeding limit point; For the marked curvature mutation area, the normal vector direction of the weld feature point in the area is extracted, and the material shrinkage prediction relationship included in the thermal deformation prediction mechanism is integrated to calculate the normal compensation component of the regional feature point; The position compensation weighted coefficient is integrated with the normal compensation component to generate the three-dimensional spatial position adjustment of the corresponding dynamic response node.

[0008] Furthermore, a multi-dimensional collaborative evaluation mechanism is constructed for the anti-drift path to synchronously adjust and comprehensively balance the path length, energy consumption of the entire welding process, and welding quality parameters to obtain multi-dimensional collaborative evaluation results, including: Based on the anti-drift path, an evaluation system is established that includes the impact degree of three aspects: path length, energy consumption, and welding quality; Taking the evaluation system as the target, the spatial coordinate sequence of the path points of the anti-drift path and the robot motion speed parameters are adjusted simultaneously through the gradient descent method to generate the initial improved path; When the welding quality parameter in the initial improvement path is less than the set threshold, the path point spacing distribution of the path is adjusted to generate a path with enhanced penetration; when the energy consumption parameter in the initial improvement path is greater than the limit, the robot motion acceleration is reduced and the path turning curvature is smoothed to generate an energy consumption improvement path; The penetration enhancement path and the energy consumption improvement path are collaboratively balanced to generate the adjusted path parameters and motion parameter sets, i.e., the multi-dimensional collaborative evaluation results.

[0009] Furthermore, based on the results of multi-dimensional collaborative evaluation, a bidirectional tree expansion is used to perform global path search. When weld deformation, fixture offset, and obstacle displacement exceed the set threshold, a local re-path planning mechanism is triggered, including: Based on the path parameters and motion parameter sets, a bidirectional fast-expanding random tree global path search is performed starting from the planned path start position and end position synchronously; In the global path search, the real-time deformation monitoring data of the weld area is acquired and compared with the preset deformation safety threshold; the real-time displacement sensor data on the workpiece fixture is compared with the preset fixture offset safety threshold; the real-time contour change data of obstacles in the workspace is compared with the preset obstacle displacement safety threshold; When any one of the weld deformation data, fixture displacement data, and obstacle contour change data exceeds the corresponding preset safety threshold, the current bidirectional rapid expansion random tree global search process is immediately interrupted; After the interruption is triggered, the actual spatial position and joint state of the robot end effector at the moment of interruption are used as the new planning starting point, and the local path replanning mechanism is immediately started to generate a local path segment that adapts to the current environmental changes.

[0010] Furthermore, based on the re-planning mechanism, the dynamic deformation characteristics of the weld area are collected in real time through a multi-source deformation monitoring device to generate deformation correlation data, including: By deploying a multi-source deformation monitoring device in the welding area, the weld width shrinkage data, temperature gradient distribution data, and fixture stress deformation data are simultaneously acquired and time-stamped. The weld width shrinkage data, temperature gradient distribution data, and fixture stress and deformation data are numerically fused and calculated according to the corresponding time and space positions to generate a dynamic deformation feature vector containing multi-dimensional information. Based on the dynamic deformation feature vector, a three-dimensional spatial distribution expression of the dynamic deformation field in the weld area is constructed; The three-dimensional spatial distribution expression is associated and integrated with the dynamic deformation feature vector to form a deformation association dataset including dynamic deformation features and spatial distribution information.

[0011] Furthermore, the dynamic compensation amount is determined based on the distribution characteristics of the deformation-related data, and the multi-dimensional collaborative evaluation parameters are updated in real time to drive the dynamic energy-saving control of welding parameters and motion trajectories, including: Gaussian distribution statistical analysis is performed on the deformation correlation data set to calculate and extract the deformation mean parameter representing the overall deformation characteristics of the weld area and the deformation variance parameter representing the degree of deformation dispersion as dynamic compensation control quantities; The dynamic compensation control quantity is input into the multi-dimensional collaborative evaluation system, and the weight ratio parameters and energy consumption limit condition parameters related to the thermal deformation effect in the evaluation system are corrected in real time based on the dynamic compensation control quantity; Based on the revised multi-dimensional collaborative evaluation system, the welding current setting value in the welding process parameters and the motion trajectory parameters of each joint of the robot are reversely adjusted; By coordinating the welding current setting value and the robot joint motion trajectory parameters, the actual welding movement speed of the robot end effector can adaptively match the rate change trend of the current dynamic deformation in the weld area, thereby achieving dynamic energy-saving control during the welding process.

[0012] In a second aspect, a method and system for dynamic energy-saving path planning of a teaching-free welding robot is provided, comprising: The scanning and mapping module is used to scan the workpiece through a laser vision device to obtain the weld coordinates, obstacle information and workpiece thermal deformation data, extract weld feature points, and establish a mapping relationship between the workpiece coordinate system and the robot base coordinate system; The path generation module is used to generate an anti-drift path by implanting dynamic response nodes into the initial path based on the mapping relationship, combining the thermal deformation prediction mechanism with sensor feedback data, and calculating the path point offset using the feedback data; The collaborative evaluation module is used to build a multi-dimensional collaborative evaluation mechanism for the anti-drift path, synchronously adjust and comprehensively balance the path length, energy consumption of the entire welding process, and welding quality parameters to obtain multi-dimensional collaborative evaluation results; The search and planning module is used to perform global path search based on the results of multi-dimensional collaborative evaluation, using a bidirectional tree expansion. When weld deformation, fixture offset, and obstacle displacement exceeding the set threshold are detected, a local re-path planning mechanism is triggered; The dynamic energy-saving module is used to collect the dynamic deformation characteristics of the weld area in real time through a multi-source deformation monitoring device based on a re-planning mechanism, generate deformation-related data, determine the dynamic compensation amount based on the distribution characteristics of the deformation-related data, update the multi-dimensional collaborative evaluation parameters in real time, and drive the dynamic energy-saving control of welding parameters and motion trajectories.

[0013] According to a third aspect, a computing device includes: one or more processors; The storage device is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method.

[0014] In a fourth aspect, a computer-readable storage medium stores a program, which implements the method when executed by a processor.

[0015] The above solution of the present invention includes at least the following beneficial effects: There is no need for manual pre-teaching of the weld path. The workpiece can be directly scanned by the laser vision device to obtain weld information and establish coordinate mapping, which reduces the pre-preparation time. It is especially suitable for welding small batches and multiple varieties of workpieces, and improves the rapid response capability and flexibility of the production line. Combining the thermal deformation prediction mechanism with the real-time feedback data of the sensor, dynamic response nodes are implanted in the path and the offset adjustment amount is calculated, which can offset the path deviation caused by factors such as thermal deformation of the workpiece and fixture offset. Through three-dimensional space position adjustment and normal compensation, it is ensured that the robot end effector always accurately tracks the weld, reduces defects such as welding deviation and lack of fusion caused by offset, and improves the consistency of welding quality; builds a collaborative evaluation mechanism of path length, energy consumption, and welding quality, and dynamically adjusts the motion parameters and path parameters to ensure that quality indicators such as penetration depth meet the standards. Under the premise of reducing the acceleration of robot motion, smoothing the path curvature, reducing unnecessary energy consumption, using bidirectional tree expansion for global path search, and setting a local re-planning trigger mechanism. When sudden situations such as weld deformation and obstacle displacement are detected, it can quickly generate local path segments adapted to the current environment, avoiding downtime adjustments caused by environmental changes, and ensuring the continuous and stable welding process. It is especially suitable for industrial scenarios with dynamic interference. Dynamic features are collected and associated data is generated through multi-source deformation monitoring devices. The compensation amount is determined based on its distribution characteristics, and the evaluation parameters and welding parameters are updated in real time, so that the robot can adapt to changes in weld deformation rate. It not only improves the fault tolerance to material properties and working condition fluctuations, but also can further optimize energy consumption and trajectory accuracy through coordinated parameter adjustment, thereby extending the service life of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1The present invention provides a flow chart of a method for dynamic energy-saving path planning of a teaching-free welding robot.

[0017] Figure 2 Schematic diagram of a dynamic energy-saving path planning system for a teaching-free welding robot provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0019] like Figure 1 As shown, an embodiment of the present invention provides a dynamic energy-saving path planning method for a teaching-free welding robot, the method comprising the following steps: Step 1: Scan the workpiece with a laser vision device to obtain weld coordinates, obstacle information, and workpiece thermal deformation data, extract weld feature points, and establish a mapping relationship between the workpiece coordinate system and the robot base coordinate system; Step 2: Based on the mapping relationship, combined with the thermal deformation prediction mechanism and sensor feedback data, dynamic response nodes are implanted in the initial path and the feedback data is used to calculate the path point offset to generate an anti-drift path; Step 3: Build a multi-dimensional collaborative evaluation mechanism for the anti-drift path, synchronously adjust and comprehensively balance the path length, energy consumption of the entire welding process, and welding quality parameters to obtain a multi-dimensional collaborative evaluation result; Step 4: Based on the multi-dimensional collaborative evaluation results, a global path search is performed using a bidirectional tree expansion. When weld deformation, fixture offset, and obstacle displacement exceed the set threshold, a local re-path planning mechanism is triggered. Step 5: Based on the re-planning mechanism, the dynamic deformation characteristics of the weld area are collected in real time through a multi-source deformation monitoring device to generate deformation-related data. The dynamic compensation amount is determined based on the distribution characteristics of the deformation-related data, and the multi-dimensional collaborative evaluation parameters are updated in real time to drive the dynamic energy-saving control of the welding parameters and motion trajectory.

[0020] In the embodiment of the present invention, the workpiece is directly scanned and coordinate mapping is established with the help of a laser vision device. Without manual pre-teaching, key information such as welds, obstacles and thermal deformation can be quickly obtained, which not only simplifies the operation process and reduces manual intervention, but also can flexibly adapt to workpieces of different specifications, improves the efficiency of early preparation, and combines thermal deformation prediction with sensor feedback to generate an anti-offset path. By implanting dynamic response nodes and calculating the offset in real time, it can effectively offset the impact of workpiece deformation during welding, ensure that the robot always accurately aligns with the weld, reduce welding defects caused by trajectory deviation, and improve the stability of welding quality. The multi-dimensional collaborative evaluation mechanism synchronously adjusts and balances the path length, energy consumption and welding quality to ensure that the welding quality meets the standards. On this basis, it can optimize the path layout to shorten the stroke and reduce energy consumption by adjusting the motion parameters, thus achieving the dual goals of efficient operation and energy saving and consumption reduction; it uses a bidirectional tree expansion for global path search and sets local replanning trigger conditions. It can not only quickly plan the global final path, but also start local adjustments in time when encountering emergencies such as weld deformation, fixture offset or obstacle displacement, avoiding interruption of the welding process and ensuring the continuity and stability of the operation. Through multi-source monitoring, dynamic deformation characteristics are collected in real time and the compensation amount is determined. The evaluation parameters are updated in real time and the welding parameters and motion trajectory are adjusted. It can adapt to the dynamically changing welding environment, further optimize the energy-saving effect, and ensure that stable welding quality can be maintained in complex dynamic scenes.

[0021] In a preferred embodiment of the present invention, the above step 1, scanning the workpiece by a laser vision device, obtaining weld coordinates, obstacle information, and workpiece thermal deformation data, extracting weld feature points, and establishing a mapping relationship between the workpiece coordinate system and the robot base coordinate system, may include: In an embodiment of the present invention, after the laser vision device is activated, the internal laser emitter is first activated to emit a preset laser pattern. If a line laser is used, a continuous laser line will be formed and projected onto the workpiece surface; if a dot matrix laser is used, a regularly arranged laser dot matrix will be formed. The angle and range of the laser projection are pre-adjusted according to the size of the workpiece to ensure that the entire area to be welded and the surrounding area where obstacles may exist can be covered. At the same time, the laser emitter operates continuously at a fixed frequency, forming a synchronous scan with the shooting frame rate of the vision sensor. The vision sensor (usually an industrial camera) maintains a fixed relative position with the laser emitter and starts to capture the image of the workpiece surface at the same time as the laser is projected. To ensure image clarity, the camera automatically adjusts the exposure parameters according to the reflective characteristics of the workpiece surface. For highly reflective materials (such as stainless steel), the exposure intensity is reduced to avoid overexposure of the light spot; for matte materials, the exposure intensity is increased to enhance the recognition of the laser pattern. The captured original image contains the bright spots or bright lines formed by the laser on the workpiece surface, as well as the contour information of the workpiece itself.

[0022] Next, the original image is preprocessed. The first step is noise reduction. Random noise points in the image (such as those caused by dust reflection) are removed through smoothing, and the continuous pattern formed by the laser is retained. The second step is contrast enhancement. By adjusting the grayscale range of the image, the light and dark difference between the laser pattern and the workpiece surface is made more significant, which is convenient for feature recognition. When extracting the weld coordinates, the distorted area of ​​the laser pattern is first identified based on the preprocessed image. Due to the presence of grooves or protrusions at the weld, the arrangement of the laser line or laser point will be locally bent or displaced. By scanning the image line by line, the continuous trajectory of the laser pattern is tracked. When regularity is detected in the trajectory, the laser pattern is detected. When there is a bend (such as a V-groove causing the laser line to split into two symmetrical branches), the area is determined to be a weld. Subsequently, a point is marked at fixed intervals (such as 0.5 mm) along the direction of the weld. The pixel coordinates of these points are converted into actual space coordinates through the conversion ratio between image pixels and actual physical dimensions (this ratio is pre-calibrated by the camera focal length and shooting distance) to form a continuous coordinate sequence of the weld. At the same time, the starting point of the weld (the location where the laser pattern first becomes distorted) and the end point (the location where the laser pattern returns to its normal trajectory) are recorded, and the points where the degree of trajectory bending suddenly changes (such as the weld corners) are marked as key feature points.

[0023] When identifying obstacles, the image is analyzed for areas where the laser pattern is blocked or completely missing. Obstacles block laser projection, resulting in neither laser reflections nor the normal texture of the workpiece surface in the image at the corresponding position. The edge detection algorithm is used to outline the contours of these areas, determine whether their shapes are regular (for example, fixtures are usually rectangular or cylindrical), and measure the maximum length and width of the contours to exclude small noise areas (such as impurities with a diameter of less than 2 mm). For confirmed obstacles, the pixel-to-physical size conversion is also used to obtain the spatial coordinates of each point on its contour boundary to determine its position and distribution range around the workpiece.

[0024] Multiple rounds of scanning and comparison are required to obtain the thermal deformation data of the workpiece. The first scan is performed at room temperature before welding, and the laser pattern trajectory and corresponding coordinates of the entire workpiece surface (especially the area near the weld) are recorded; the second and third scans are performed during the welding process (such as when the welding is 50% complete) and after the welding is completed, respectively, to obtain the laser trajectory data of the same area. The trajectory of the subsequent scan is compared point by point with the reference trajectory of the first scan, and the coordinate deviation of the corresponding position is calculated. If the deviation of a point in the X-axis direction exceeds 0.1 mm, or the deviation in the Y-axis and Z-axis directions exceeds 0.05 mm, and multiple adjacent points show the same direction of deviation (such as shrinkage along the length of the weld), it is determined to be a displacement caused by thermal deformation. These deviation data are summarized to form a distribution map of the workpiece thermal deformation, marking the area with the most significant deformation (usually within 20 mm on both sides of the weld).

[0025] When establishing the workpiece coordinate system, select a fixed feature on the workpiece that does not change with the welding process as the reference - give priority to the positioning features during design, such as the right-angle vertex of the edge of the workpiece (as the origin). The direction of the vertex extending along the long side of the workpiece is the X-axis, the direction extending along the short side is the Y-axis, and the upward direction perpendicular to the workpiece surface is the Z-axis. Measure the distance of all weld feature points (starting point, end point, corner point) relative to the origin to determine their coordinate values ​​in the workpiece coordinate system.

[0026] When establishing a mapping relationship with the robot base coordinate system, first determine the position of the laser vision device itself in the robot base coordinate system. By measuring the relative distance between the device mounting base and the robot base (such as the horizontal distance and vertical height between the device origin and the robot base origin), as well as the installation angle of the device (such as the angle with the robot spindle), determine the spatial position relationship between the device coordinate system and the robot base coordinate system. Subsequently, the feature point coordinates in the workpiece coordinate system are converted to the device coordinate system (adjust the direction and offset of the coordinates according to the viewing angle of the device when shooting), and then combine the position of the device and the robot base to determine the coordinate system of the robot base. The robot coordinate system is converted to the robot base coordinate system based on the device origin's X coordinate in the robot base coordinate system. For example, if the device origin's X coordinate in the robot base coordinate system is 500 mm, and the X coordinate of a weld feature point in the device coordinate system is 100 mm, then the X coordinate of that point in the robot base coordinate system is 500 plus 100 mm (the sign needs to be adjusted according to the direction). Finally, calibration is performed using the corresponding coordinates of at least three non-collinear feature points in the two coordinate systems to correct for conversion deviations caused by installation errors and ensure the accuracy of the mapping relationship. This allows the robot to directly obtain the position of the weld in its own coordinate system through this mapping.

[0027] In a preferred embodiment of the present invention, the above step 2, based on the mapping relationship, combined with the thermal deformation prediction mechanism and sensor feedback data, implanting dynamic response nodes in the initial path and using the feedback data to calculate the path point offset to generate the anti-drift path, may include: Step 220 , based on the mapping relationship, combined with the thermal expansion characteristics of the material and the heat input characteristics during welding, calculate the expected deformation of the characteristic points on the initial weld path; Step 221 , based on the expected deformation amount, reversely superimpose the coordinates of the feature points of the initial path in the opposite direction of the deformation direction to generate a thermal deformation pre-compensation path; Step 222 , based on the thermal deformation pre-compensation path, the actual deformation data of the workpiece during welding is collected online by a laser vision sensor, and a real-time spatial error comparison is performed between the actual deformation data and the expected deformation to obtain an error comparison result; Step 223: Based on the error comparison results, curvature mutation areas and error exceeding points are marked in the pre-compensation path, and dynamic response nodes are implanted at the marked locations. The three-dimensional spatial position adjustment amount at the dynamic response node is calculated. Specifically, the following steps are performed: for the marked deformation error exceeding point, a position compensation weighted coefficient is calculated based on the curvature change rate of the adjacent path segments and the thermal deformation gradient at the deformation error exceeding point; for the marked curvature mutation area, the normal vector direction of the weld feature point in the area is extracted, and the material shrinkage prediction relationship included in the thermal deformation prediction mechanism is integrated to calculate the normal compensation component of the regional feature point; and the position compensation weighted coefficient is integrated with the normal compensation component to generate the three-dimensional spatial position adjustment amount of the corresponding dynamic response node. In step 224 , based on the three-dimensional spatial position adjustment amount, the spatial coordinates of the corresponding nodes in the pre-compensation path are corrected in real time to generate a thermal deformation resistant offset path.

[0028] In the embodiment of the present invention, the specific coordinates of all feature points on the initial weld path in the robot coordinate system are determined based on the mapping relationship between the workpiece and the robot base coordinate system, such as the starting point at 100 mm on the X axis, 50 mm on the Y axis, and 10 mm on the Z axis. Then, the thermal expansion characteristics of the workpiece material are collected. For example, the linear expansion coefficient of a certain steel is 12×10 -6 / ℃, that is, when the temperature rises by 1℃, the length of each millimeter will stretch by 12×10 -6 millimeters. At the same time, it is known that this material will begin to deform significantly when the temperature reaches 300°C. The heat input characteristics of the welding process are recorded. The welding current is set to 200A, the arc voltage is 25V, and the welding speed is 500mm per minute. The time the heat source acts on each point during welding is about 0.1 seconds. When estimating the temperature field distribution in the weld area, the center of the weld will reach 800°C due to direct arc heating. The temperature at a distance of 1mm from the center of the weld is about 600°C, the temperature at a distance of 5mm is about 300°C, and the temperature at a distance of 10mm and above is basically within 100°C, with little change.

[0029] Calculate the expected deformation of each feature point. Take the feature point 5mm away from the center of the weld as an example. The temperature rises from room temperature (25℃) to 300℃, and the temperature change is 275℃. The original length from this point to the center of the weld is 5mm. According to the linear expansion coefficient, the elongation is 5mm×275℃×12×10 -6 / ℃=0.0165mm. Considering the impact of material cooling shrinkage, the high-temperature area will shrink slightly more than the expansion during cooling. For example, the shrinkage of the weld center after cooling will increase by 10% on the basis of the expansion. Therefore, the previously calculated expansion is corrected to obtain the expected deformation of each feature point in the X, Y, and Z directions.

[0030] Analyze the deformation direction of each feature point. For points closer to the center of the weld, the overall deformation direction is toward the center of the weld because the material expands due to heat and contracts due to cooling during welding. For example, a feature point is expected to shrink 0.02mm toward the center of the weld along the positive direction of the X-axis. According to the principle of the opposite direction of the deformation, the coordinates of the feature points of the initial path are adjusted. The original X coordinate of the feature point is 100mm, so it is adjusted to 100mm+0.02mm=100.02mm. After completing this coordinate correction for all feature points, these adjusted points are connected with a smooth curve to form a continuous thermal deformation pre-compensation path. This path takes into account possible subsequent deformation in advance and reserves appropriate space for robot movement.

[0031] During the welding process, the laser vision sensor scans the weld area 20 times per second. By comparing the images before and during welding, the actual displacement of each feature point is identified, and the number of displacement pixels of the feature point in the image is converted into the actual spatial deformation according to the pre-calibrated ratio. For example, if a feature point is displaced by 2 pixels in the image, and it is known that 1 pixel corresponds to 0.01mm in reality, then the actual deformation of the point is 0.02mm. At the same time, the deformation direction is determined to be the positive direction of the X-axis, and the actual deformation is compared point by point with the expected deformation. The expected deformation of the feature point is 0.015mm in the positive direction of the X-axis, and the actual deformation is 0.02mm. Therefore, the error in the X direction is 0.02mm-0.015mm=0.005mm; the error threshold is set to 0.01mm, and all points with an absolute error value exceeding 0.01mm are marked. At the same time, it is observed that the errors at the weld corners are generally larger than those at the straight line segments, and this information is compiled into the error comparison results.

[0032] According to the error comparison results, the curvature mutation area is marked, such as the 90-degree corner of the weld, where the curvature of this path changes from 0 to a larger curvature value within a very short distance; at the same time, the error exceeding limit points are marked, that is, those feature points whose absolute error value exceeds 0.01mm, and dynamic response nodes are implanted at these marked positions. For the error exceeding limit points, the position compensation weighted coefficient is calculated. First, the curvature change rate of the adjacent previous path is calculated to be 0.3 / mm, and the curvature change rate of the adjacent next path is 0.5 / mm. The average value of the two is 0.4 / mm. Then, the thermal deformation gradient of the point is calculated to be 0.2℃ / mm, that is, the temperature decreases by 0.2℃ for every 1mm away from the center of the weld. According to the ratio of curvature change rate accounting for 60% and thermal deformation gradient accounting for 40%, the position compensation weighted coefficient is calculated, that is, 0.4×0.6+0.2×0.4=0.24+0.08=0.32.

[0033] For the area with sudden curvature change, the normal vector direction of each feature point in the area is extracted. The direction is perpendicular to the weld surface and upward. Combined with the material shrinkage prediction relationship, it is known that the transverse shrinkage rate during cooling is 1.5 times that of the longitudinal shrinkage rate, and the longitudinal shrinkage rate is 0.01mm / mm. Therefore, the transverse shrinkage rate is 0.015mm / mm. The compensation component of each point in the normal direction is calculated to be 0.008mm. Finally, the position compensation weighting coefficient is 0.32. Integrating with the normal compensation component of 0.008mm and considering the errors of the point in the X, Y, and Z directions, the three-dimensional spatial position adjustment amounts of +0.006mm in the X direction, -0.002mm in the Y direction, and +0.004mm in the Z direction are obtained. For each dynamic response node, the coordinates of the corresponding node in the pre-compensation path are corrected in real time according to the calculated three-dimensional spatial position adjustment amounts. The current X coordinate of a node is 100.02mm, the adjustment amount is +0.006mm, and the X coordinate becomes 100.026mm after correction. After the correction is completed, check whether the path between adjacent nodes is smooth. If sharp corners or mutations occur, fine-tune the connection curves between the nodes to ensure a continuous and smooth path. After such processing, the resulting thermal deformation-resistant offset path can well adapt to the dynamic deformation during welding.

[0034] By analyzing the thermal expansion characteristics of the material and the heat input characteristics of welding in detail, the expected deformation calculated is closer to the actual deformation of the workpiece, which improves the accuracy of pre-compensation and reduces the blindness of the initial path design. The laser vision sensor collects data in real time and compares it with the expected value, which can detect deformation errors at the first time. Then, the adjustment amount is calculated in time through the dynamic response node, so that the path can quickly adapt to the actual deformation, so that the robot can always accurately align with the weld and avoid the accumulation of welding deviations caused by deformation. It specifically processes areas with sudden changes in curvature and points where errors exceed the limit, taking into account the stress concentration and uneven temperature distribution in these areas. Through unique weighting coefficients and compensation component calculations, it can accurately respond to local severe deformation and ensure the welding quality of these complex areas. The dual mechanism of pre-compensation + real-time correction ensures that the arc always acts stably on the center of the weld, keeps the weld penetration depth and width consistent, reduces the probability of defects such as weld deviation and undercut, and improves the stability of welding quality. The dynamic adjustment mechanism makes the robot's motion trajectory more in line with actual needs, avoids unnecessary adjustment actions caused by deformation, reduces ineffective motion, thereby reducing energy consumption, while also reducing wear on various robot components and extending the service life of the equipment.

[0035] In a preferred embodiment of the present invention, in step 3, a multi-dimensional collaborative evaluation mechanism is constructed for the anti-drift path to synchronously adjust and comprehensively balance the path length, energy consumption of the entire welding process, and welding quality parameters to obtain a multi-dimensional collaborative evaluation result, which may include: Step 330: Based on the anti-drift path, an evaluation system is established that includes the impact degree of three aspects: path length, energy consumption, and welding quality; Step 331 , using the evaluation system as a target, the spatial coordinate sequence of the path points of the anti-drift path and the robot motion speed parameters are adjusted simultaneously by the gradient descent method to generate an initial improved path; Step 332: When the welding quality parameter in the initial improved path is less than a set threshold, the path point spacing distribution of the path is adjusted to generate a path with enhanced penetration; when the energy consumption parameter in the initial improved path is greater than a limit, the robot motion acceleration is reduced and the path curvature is smoothed to generate an energy consumption improved path; In step 333 , the penetration enhancement path and the energy consumption improvement path are collaboratively balanced to generate an adjusted path parameter and motion parameter set, i.e., a multi-dimensional collaborative evaluation result.

[0036] In an embodiment of the present invention, the specific spatial coordinates of all path points on the anti-drift path are first obtained. For example, the coordinates of the first path point are (100, 50, 10), the second is (102, 51, 10), etc., and the X, Y, and Z values ​​of each point are recorded in sequence. When calculating the path length, the straight-line distance between two adjacent path points is calculated one by one. The calculation method is to first calculate the distance difference in the X, Y, and Z directions based on the coordinate difference of the two points, and then obtain the straight-line distance between the two points through the geometric relationship. For example, the first point (100, 50, 1 The first point (0) and the second point (102, 51, 10) have a difference of 2mm in the X direction, 1mm in the Y direction, and 0mm in the Z direction. The distance between the two points is the square root of the sum of the squares of these three differences, which is approximately 2.24mm. The distances of all adjacent points are added together to obtain the total length of the anti-drift path, assuming it is 500mm. The theoretical shortest path length is set to 480mm, and the ratio of the actual path length to the theoretical shortest path length is calculated, that is, 500 / 480≈1.04. This ratio is an indicator of the degree of influence of the path length dimension.

[0037] For the energy consumption dimension, first count the energy consumption of each joint movement of the robot. According to the movement speed, acceleration and load of each joint, combined with the running time, estimate the energy consumption of each joint, and then add them up to get the total energy consumption of the robot movement. For example, joint 1 consumes 200J during movement, joint 2 consumes 150J, and so on, the total is 600J. Then count the energy consumption of the welding equipment. According to the current, voltage and welding time during the welding process, calculate the energy consumption of the welding equipment. Assuming it is 500J, then the total energy consumption of the entire welding process is 600+500=1100J. Set the average energy consumption upper limit of similar welding tasks to 1200J, and calculate the ratio of actual energy consumption to the upper limit value, that is, 1100 / 1200≈0.92, as the impact degree indicator of the energy consumption dimension.

[0038] In the welding quality dimension, the penetration, width and reinforcement of the welded joint are measured by special detection equipment. Assuming that the measured penetration is 2.5 mm, the width is 5 mm, and the reinforcement is 1 mm, the standard values of the penetration, width and reinforcement are 2.5 mm, 5 mm and 1 mm respectively. The deviation rate of each parameter is calculated as (actual value-standard value) / standard value. The deviation rates of the three parameters are all 0. According to the importance of the three parameters, the penetration is given a weight of 0.4, the width is given a weight of 0.3, and the reinforcement is given a weight of 0.3. The deviation rates of each parameter are multiplied by the corresponding weight and added together to obtain the influence degree index of the welding quality dimension, which is 0. Finally, the path length, energy consumption and welding quality dimensions are respectively given weights of 0.2, 0.3 and 0.5. The influence degree index of each dimension is multiplied by the corresponding weight and summed, that is, 1.04*0.2+0.92*0.3+0*0.5≈0.208+0.276+0=0.484. This comprehensive evaluation score constitutes the core content of the evaluation system.

[0039] The evaluation system established in step 330 is the target, that is, to pursue a lower comprehensive evaluation score (because the better the dimension index is, the lower the score is). When adjusting the path point space coordinate sequence, starting from the first path point, the X coordinate is first slightly adjusted, such as from 100 mm to 100.05 mm, while the Y and Z coordinates remain unchanged. Then the total length, energy consumption and welding quality parameters of the adjusted path are recalculated to obtain a new comprehensive evaluation score. If the new score is lower than the original 0.484, the adjustment is retained; if the score is higher, the X coordinate is restored to 100 mm. Then the Y coordinate and the Z coordinate are adjusted in the same way, and the other path points are adjusted in turn. In the adjustment process, the adjustment amplitude is controlled within 0.1 mm to avoid excessive influence on the overall shape of the path. Repeat the operation until the small adjustment of the path point coordinates cannot further reduce the comprehensive score.

[0040] When adjusting the robot motion speed parameter, the path is divided into straight line segments and turning segments. The original motion speed of the straight line segment is set to 8 mm / s. First, try to adjust the speed to 7 mm / s and calculate the comprehensive evaluation score. Then adjust the speed to 9 mm / s and calculate the score. After comparison, select the speed that can make the score lower. For the turning segment, the original speed is 4 mm / s. Similarly, try speeds of 3 mm / s and 5 mm / s and select the final value. In this way, appropriate motion speeds are determined for different segments. The adjusted path point coordinate sequence and the motion speed parameter are combined to form the initial improved path.

[0041] Assume that the penetration of the welding quality parameters in the initial improvement path is measured to be 1.8mm on a certain path, and the set threshold is 2mm. At this time, it is necessary to generate a penetration enhancement path. First, check the path point spacing of this path. It was originally 1mm, that is, a path point was set every 1mm. In order to increase the penetration, the path point spacing of this path is adjusted to 0.8mm. In this way, the robot will pass through more path points within the same length, the movement time will be increased, and the welding arc action time will be longer, thereby increasing the penetration. In this way, the path point spacing is adjusted for all sections with a penetration depth of less than 2mm, and the original spacing of other sections remains unchanged. A penetration enhancement path is generated to ensure that the penetration depth on this path reaches 2mm or above. If the total energy consumption of the initial improvement path is 1300J, and the set energy consumption limit is 1200J, it is necessary to generate an energy consumption improvement path. First, reduce the robot's motion acceleration. The original acceleration of the robot when accelerating from standstill to 8mm / s was 0.5mm / s², which is now adjusted to 0.3mm / s 2 , making the robot's acceleration process smoother and reducing energy consumption. Then check the turning curvature of the path. At a certain turn, the original curvature radius was 5mm, which was large, and the robot consumed more energy when turning. Adjust the curvature radius to 8mm to make the turn smoother and reduce energy consumption when turning. Make such adjustments to all acceleration stages and turns with higher energy consumption to generate an energy-saving path.

[0042] First, the comprehensive evaluation scores of the penetration enhancement path and the energy consumption improvement path are calculated respectively. Assuming that the score of the penetration enhancement path is 0.46 and the score of the energy consumption improvement path is 0.44, since the score of the energy consumption improvement path is lower, a coordinated balance is performed based on the energy consumption improvement path. The penetration depth on the energy consumption improvement path is checked, and it is found that the penetration depth of a section of the path is 2.1mm, which is close to the threshold of 2mm and there is a potential risk. Therefore, referring to the path point spacing of the corresponding section in the penetration enhancement path, the path point spacing of this section of the path is adjusted from 1mm to 0.9mm, which ensures that the penetration depth is stable above the threshold without increasing energy consumption too much. Then By checking the energy consumption in the penetration enhancement path, it was found that the spacing between the path points of a certain straight section was 0.8mm, and the energy consumption was slightly higher. In the energy consumption improvement path, the spacing between the sections was 1mm, the energy consumption was better, and the penetration depth could reach 2.2mm. Therefore, the spacing of this section of the path was adjusted to 1mm, and such fine-tuning was repeated. The comprehensive evaluation score was recalculated after each adjustment until a parameter combination with the lowest score was found. The final adjusted path parameters included the path point spacing of each section, the curvature radius at the turn, etc. The motion parameter set included the motion speed and acceleration of each section. These parameters together constituted the multi-dimensional collaborative evaluation results.

[0043] The specific parameters and calculation methods of path length, energy consumption, and welding quality are considered in detail. Various indicators and weights are set in combination with actual welding scenarios, so that the evaluation system can accurately reflect the impact of different factors on the welding process and avoid the optimization direction deviation caused by vague evaluation criteria. The path point coordinates and movement speed are carefully fine-tuned through the gradient descent method, and the final parameters are found point by point and segment by segment. Compared with the overall adjustment, it can find more subtle optimization space, so that the generated initial improvement path can improve the overall performance while maintaining the anti-drift ability. When the welding quality or energy consumption does not meet the standards, the path point spacing, acceleration, curvature and other parameters are adjusted in a targeted manner to directly act on the key links where the problem lies. For example, by reducing the path point spacing, the penetration depth can be accurately increased, and reducing the acceleration and increasing the curvature radius can effectively reduce energy consumption, making the optimization of a single indicator more efficient and thorough. The penetration enhancement path and the energy consumption improvement path are carefully compared and integrated. Instead of simply choosing one path, the strengths of both are taken and the weaknesses of both are compensated. While ensuring welding quality, energy consumption is controlled and the rationality of path length is taken into account, avoiding the situation of optimizing a single indicator at the expense of other indicators. The detailed calculation and adjustment process makes the final path parameters and motion parameter sets more in line with actual production needs, which can not only ensure the quality of welding products, but also reduce energy consumption and path length, reduce production costs, and improve production efficiency, making this path planning method more practical and competitive in actual industrial applications.

[0044] In a preferred embodiment of the present invention, step 4 above uses a bidirectional tree expansion to perform a global path search based on the multi-dimensional collaborative evaluation results. When weld deformation, fixture offset, and obstacle displacement are detected to exceed a set threshold, a local re-path planning mechanism is triggered, which may include: Step 440 , based on the path parameters and the motion parameter set, starting from the planned path start position and end position synchronously, perform a bidirectional fast expanding random tree global path search; Step 441, in the global path search, real-time deformation monitoring data of the weld area is obtained and compared with a preset deformation safety threshold; real-time displacement sensor data on the workpiece fixture is compared with a preset fixture offset safety threshold; real-time contour change data of obstacles in the workspace is compared with a preset obstacle displacement safety threshold; Step 442: When any one of the weld deformation data, fixture displacement data, and obstacle profile change data exceeds the corresponding preset safety threshold, the current bidirectional rapid expansion random tree global search process is immediately interrupted; Step 443: After the interrupt is triggered, the actual spatial position and joint state of the robot end effector at the moment of interruption are used as the new planning starting point, and the local path replanning mechanism is immediately started to generate a local path segment that adapts to the current environmental changes.

[0045] In an embodiment of the present invention, first, the path parameters are comprehensively collected, including the specific coordinate value of the starting point in the three-dimensional coordinate system, the three-dimensional coordinate value of the end point, the precise coordinates of several key nodes that the path must pass through, the maximum curvature allowed by the path (that is, the upper limit of the angle change within each meter of path length), the maximum value that the total length of the path cannot exceed, etc.; at the same time, a set of motion parameters is collected, covering the maximum movement speed that each joint of the robot can achieve (such as the maximum speed of joint 1 is 30 degrees per second), the maximum acceleration of each joint (such as the maximum acceleration of joint 2 is 15 degrees per second), the range of motion angles of each joint (such as joint 3 can only move between -90 degrees and 90 degrees), the allowable error range of the position of the end effector during movement (such as not exceeding ±0.5 mm), etc.

[0046] The initial node of the first rapidly expanding random tree is set to the starting point of the planned path, and the 3D coordinates of this starting point and the corresponding robot joint angles are stored as attributes of the initial node. The initial node of the second rapidly expanding random tree is set to the end point of the planned path, and its 3D coordinates and corresponding joint angles are also stored. The two random trees are then synchronized and expanded. During each expansion cycle, the system randomly generates a 3D sampling point within the robot workspace (its X, Y, and Z coordinates are all within the boundaries of the workspace). For each random tree, the straight-line distance between all existing nodes in the tree and the sampling point is calculated to find the closest node. Starting from this closest node, a new potential node is generated by moving a certain distance (set according to the path accuracy requirements, such as 0.3 mm per movement) towards the sampling point, in accordance with the kinematic constraints of each robot joint (such as joint angles cannot exceed the range and movement speed does not exceed the maximum value). A collision detection algorithm is used to check whether the straight line segment from the closest node to the new potential node overlaps with fixed obstacles in the workspace. If no collision occurs, the new node is added to the corresponding random tree, and the connection relationship between the node and the closest node is recorded.

[0047] The above expansion process is repeated continuously. After each expansion, the mutual distance between all nodes in the two random trees is calculated. When two nodes from different trees appear and the straight-line distance between them is less than the preset connection threshold (this threshold is determined based on the diameter of the end effector and the motion safety margin, such as 5 mm), the expansion process is stopped; these two nodes are connected, and the node connection relationship of each tree is traced back in turn to form a complete global path from the starting point through the intermediate node to the end point, and the coordinates of each node on the path and the corresponding robot joint status are recorded.

[0048] At each time interval (e.g., every 0.1 second) during the global path search using the bidirectionally rapidly expanding random tree, multiple laser displacement sensors installed at different locations in the weld area simultaneously collect distance data between each monitoring point on the weld surface and the sensor. These distance data are then subtracted from the original distance data when the weld is not deformed to obtain the deformation variable of each monitoring point (e.g., if the original distance of a monitoring point is 100 mm and the current distance is 102 mm, then the deformation variable is 2 mm). The deformation variables of all monitoring points are then averaged as the real-time deformation monitoring data of the weld area. At the same time, the grating rulers installed on the workpiece fixture in the X, Y, and Z directions collect the position data of the fixture in each direction in real time, compare the current position data with the standard data of the initial installation position of the fixture, and calculate the displacement in each direction (e.g., if the current position in the X direction is 50.2 mm and the standard position is 50.0 mm, then the displacement in the X direction is 0.2 mm). The displacements in the three directions are combined to form the real-time displacement sensor data of the fixture.

[0049] In addition, a 3D camera within the workspace captures an image containing the obstacle every 0.2 seconds. An image recognition algorithm is used to extract the obstacle's contour feature points (e.g., 20 key contour points representing the obstacle's shape) and obtain the three-dimensional coordinates of each feature point. These coordinates are compared with the coordinates of the obstacle's contour feature points captured at the previous moment, and the coordinate change of each feature point in the X, Y, and Z directions is calculated. The average of the coordinate changes of all feature points is then calculated as the real-time contour change data of the obstacle. The calculated real-time weld deformation monitoring data is numerically compared with a preset deformation safety threshold (e.g., set to 1.5 mm based on welding process requirements) to determine whether the real-time data exceeds the threshold. The displacement of the fixture in each of the three directions is compared with the preset fixture offset safety threshold (e.g., 0.3 mm for the X and Y directions and 0.2 mm for the Z direction) to determine whether the displacement in any direction exceeds the corresponding threshold. The real-time contour change data of the obstacle is compared with the preset obstacle displacement safety threshold (e.g., set to 2 mm) to determine whether it exceeds the threshold.

[0050] The three comparison results in step 441 are continuously monitored, and the comparison status is checked every 0.05 seconds. If it is found that the real-time deformation monitoring data of the weld (such as 1.6 mm calculated) is greater than the preset deformation safety threshold (1.5 mm), or the displacement of the fixture in a certain direction (such as 0.4 mm displacement in the X direction) is greater than the fixture offset safety threshold (0.3 mm) in that direction, or the real-time contour change data of the obstacle (such as 2.1 mm) is greater than the preset obstacle displacement safety threshold (2 mm), as long as any of these situations occurs, an interrupt command is immediately sent to the bidirectional fast expansion random tree global search. Upon receiving the command, all ongoing search operations such as random tree node sampling, nearest node search, new node generation, and tree node connection are immediately stopped. At the same time, the current node information of the two random trees at the time of interruption and the generated partial path data are saved, but this data is no longer processed further, and the global path search process is completely terminated.

[0051] After the interrupt is triggered, the high-precision encoder installed on each joint of the robot reads the actual angle value of each joint at the time of interruption (such as 35.2 degrees for joint 1 and -15.7 degrees for joint 2), and records these angle values ​​as the current joint state data. At the same time, the laser positioning system installed on the end effector accurately measures its actual coordinate position in three-dimensional space (such as 120.5 mm for X, 80.3 mm for Y, and 50.1 mm for Z) and the posture angle of the end effector (such as 3 degrees rotation around the X axis, 2 degrees rotation around the Y axis, and 5 degrees rotation around the Z axis). These data are recorded as the actual spatial position of the end effector, and the actual spatial position of the end effector and the actual angle values ​​of each joint obtained above are set as the new starting point for local path replanning, replacing the original global path starting point, and immediately starting the local path replanning mechanism. The mechanism first determines the local path replanning based on the environmental change data that exceeds the threshold (such as the weld deformation is 1.6 mm, The fixture is offset in the X direction by 0.4 mm or the obstacle is displaced by 2.1 mm). The range of the area where the environmental change affects the robot's motion is determined (such as the area formed by extending 100 mm around the change point). Then, a path search is performed within the local space around the new starting point (such as a cubic space with the new starting point as the center and extending 200 mm in the X, Y, and Z directions). Sampling points within this range are randomly generated. The path from the new starting point to each sampling point is calculated to see whether it avoids the dangerous areas caused by environmental changes (such as the weld area after deformation, the fixture position after offset, and the obstacle outline after movement). At the same time, the path is checked to see whether it meets the constraints such as the robot's joint motion range and speed limit. From all qualified paths, the path with the shortest length and least curvature is selected as the local path segment. The starting point of this path segment is the new starting point at the interruption moment, and the end point is the position that can avoid the current environmental changes and smoothly connect with the rest of the original global path.

[0052] The bidirectional rapid expansion random tree expands from the starting point and the end point at the same time, which can cover the workspace in a shorter time, reduce the invalid search area, and make the global path found closer to the final solution. The detailed application of path parameters and motion parameters ensures that the path is consistent with the actual motion capability of the robot, avoiding planning of unexecutable paths, and real-time monitoring of the status data of welds, fixtures and obstacles. It can judge whether it exceeds the safety threshold through precise numerical comparison, and can find problems at the first time when abnormal changes occur in the environment, avoiding collisions or operational errors when the robot moves along the original path due to delayed detection. When changes exceeding the threshold are detected, the global search is immediately interrupted to prevent the system from continuing to plan paths based on erroneous environmental information. ; Local replanning with the current actual state as the new starting point can ensure that the new path is fully adapted to the changed environment, avoid collisions between the robot and obstacles or offset fixtures, and ensure that the welding quality of the weld is not affected by deformation. Local replanning only adjusts the path for the affected local area, and there is no need to replan the entire global path, which shortens the planning time and enables the robot to quickly resume operations, reducing downtime caused by environmental changes and improving overall operation efficiency. Through the combination of real-time monitoring and local replanning, the robot can adapt to the dynamically changing working environment. Even in sudden situations such as weld deformation, fixture offset or obstacle movement, it can flexibly adjust the path to ensure the smooth completion of the task.

[0053] In a preferred embodiment of the present invention, the above step 5, based on the re-planning mechanism, collects the dynamic deformation characteristics of the weld area in real time by a multi-source deformation monitoring device to generate deformation correlation data, which may include: Step 550 , using a multi-source deformation monitoring device deployed in the welding area, synchronously acquire weld width shrinkage data, temperature gradient distribution data, and fixture stress deformation data, and perform unified time stamping. Step 551 , performing numerical fusion calculation on the weld width shrinkage data, the temperature gradient distribution data, and the fixture stress and deformation data according to the corresponding time and space positions to generate a dynamic deformation feature vector including multi-dimensional information; Step 552: constructing a three-dimensional spatial distribution expression of the dynamic deformation field of the weld region based on the dynamic deformation feature vector; In step 553 , the three-dimensional spatial distribution expression is associated and integrated with the dynamic deformation feature vector to form a deformation association data set including dynamic deformation features and spatial distribution information.

[0054] In an embodiment of the present invention, multiple groups of different types of deformation monitoring devices are deployed at key locations in the welding area. Laser width sensors are installed every 5 centimeters along the length of the weld to monitor changes in weld width. An infrared thermal imager is placed on each side of the weld, with the lens focal length adjusted to just cover the entire weld area, to collect temperature distribution. Two strain gauges are installed at each of the four corners where the fixture contacts the workpiece, with the sensitive axes aligned in the horizontal and vertical directions, respectively, to detect stress and deformation in different directions of the fixture. The acquisition frequency of all monitoring devices is set to 20 times per second. A unified synchronous trigger signal is used to ensure that all groups of devices begin data acquisition at the same time. During each scan, the laser width sensor emits a laser beam across the weld width, receives reflected light, and calculates the spot position to obtain the current weld width value (e.g., a measured width of 6.2 mm at a certain moment). This value is then subtracted from the initial width value recorded before welding (e.g., an initial width of 7.0 mm) to obtain the width shrinkage at that location (6.2 - 7.0 = -0.8 mm, with the negative sign indicating shrinkage). Each sensor generates shrinkage data at five evenly distributed points during each acquisition.

[0055] The infrared thermal imager generates a 1024×768 pixel temperature image each time it collects data. Each pixel corresponds to an area of ​​0.1 mm×0.1 mm in the welding area. The pixel coordinates are converted into actual spatial coordinates through image calibration. The temperature value of each coordinate point is then read (for example, the temperature of a certain point is 285°C), and the ratio of the temperature difference between adjacent pixel points to the spatial distance is calculated to obtain the temperature gradient data (for example, if the temperature difference between adjacent points in the horizontal direction is 5°C and the distance is 0.1 mm, the horizontal temperature gradient is 50°C / mm).

[0056] The strain gauge on the fixture converts the tiny deformation caused by stress into a change in resistance. The resistance change is converted into a voltage signal through a signal amplifier. The actual strain is then calculated based on the strain gauge sensitivity coefficient (e.g., 2.0 mV / V) (e.g., if the output voltage change of a strain gauge is 0.5 mV and the excitation voltage is 5 V, the strain is 0.5 / (5×2.0)=0.05). The stress value is then calculated based on the elastic modulus of the fixture material (e.g., 200 GPa). Finally, the stress and deformation data of the fixture are obtained based on the relationship between stress and deformation (e.g., the stress and deformation at a certain point is 0.12 mm). After each acquisition is completed, the same timestamp is automatically added to the laser width shrinkage data, infrared temperature gradient distribution data, and fixture stress and deformation data. The timestamp is accurate to microseconds (e.g., 2024-06-10 09:45:12.345678) and stored in chronological order to ensure that the three sets of data are completely corresponding in the time dimension.

[0057] A three-dimensional spatial coordinate system for the welding area is established, with the weld starting point as the origin, the weld length direction as the X-axis, the direction perpendicular to the weld as the Y-axis, and the direction perpendicular to the workpiece surface as the Z-axis. All monitoring data are mapped to this coordinate system, and the welding area is divided into three-dimensional grid units of 0.5 mm × 0.5 mm × 0.5 mm. Each grid unit has a unique coordinate identifier (such as X = 10.0 mm, Y = 2.5 mm, Z = 0.5 mm). For each grid unit, three types of data at the corresponding position are extracted, and the laser width sensor is used to measure the three types of data. From the sensor data, the width shrinkage of the grid in the Y-axis direction is calculated by interpolation (for example, the shrinkage of a grid unit is -0.75 mm). From the infrared thermal imager data, the temperature gradient values ​​of the grid unit in the X, Y, and Z directions are extracted (for example, 30°C / mm in the X direction, 25°C / mm in the Y direction, and 5°C / mm in the Z direction). From the fixture stress and deformation data, the stress and deformation variables of the fixture in the X and Y directions corresponding to the grid unit are calculated through spatial mapping (for example, 0.1 mm in the X direction and 0.08 mm in the Y direction).

[0058] Each type of data is standardized. For width shrinkage, the maximum shrinkage during the entire welding process is determined (e.g., -1.0 mm), and the shrinkage of each grid is divided by the maximum value (e.g., -0.75 / -1.0=0.75). For temperature gradient, the maximum absolute value of the gradient in the three directions (e.g., 30°C / mm) is taken as the representative value of the temperature gradient of the grid, and then divided by the preset maximum temperature gradient threshold (e.g., 50°C / mm) to obtain the standardized value (30 / 50=0.6). For fixture stress deformation, the square root of the sum of the squares of the deformation in the X and Y directions is calculated, and then divided by the maximum value. The maximum allowable stress deformation is obtained by normalizing the value. Each grid cell forms a sub-vector containing 5 normalized values, [normalized value of width shrinkage, normalized value of temperature gradient, normalized value of stress deformation in the X direction, normalized value of stress deformation in the Y direction, normalized value of stress-deformation sum]. The grid cells are arranged in the order of the X, Y, and Z axes, and the sub-vectors of all grid cells are connected end to end to form a one-dimensional dynamic deformation feature vector containing all grid information. The vector length is the total number of grid cells multiplied by 5 (for example, if there are 10,000 grid cells, the vector length is 50,000).

[0059] The original data of each grid cell (unstandardized width shrinkage, temperature gradient, stress deformation value) and the corresponding spatial coordinates are extracted from the dynamic deformation feature vector. Taking the three-dimensional coordinate system as the framework, each grid cell is regarded as a point in space, and its coordinates are the center coordinates of the cell (such as X=10.25 mm, Y=2.75 mm, Z=0.75 mm). For the weld width shrinkage, the inverse distance weighted interpolation method is used to fill the blank area between the grid cells. For example, there are 4 known grid cells around a blank point, with distances of 0.3, 0.5, 0.7, and 0.9 mm, respectively, and the shrinkage is -0.7, -0.8, -0.6, and -0.5 mm, respectively. The shrinkage of the blank point is calculated as ((-0.7 / 0.3)+(-0.8 / 0.5)+(-0.6 / 0.7)+(-0.5 / 0.9))÷(1 / 0.3+1 / 0.5+1 / 0.7+1 / 0.9)≈-0.68 mm; for the temperature gradient, the Kriging interpolation method is used for space filling, taking into account the temperature change trend in different directions, so that the interpolation result is more consistent with the actual temperature field distribution. For the stress deformation of the fixture, the linear interpolation method is used to calculate the deformation value of the non-monitoring point according to the rigid structural characteristics of the fixture to ensure that the deformation distribution conforms to the law of mechanical transfer. After completing the data filling of all blank areas, the complete deformation data of the spatial points every 0.1 mm in the welding area are obtained. Using 3D visualization software, the width shrinkage is represented by different colors (for example, the greater the shrinkage, the darker the color), the temperature gradient is represented by a directional arrow (the length of the arrow represents the gradient size, and the direction represents the gradient direction), and the stress deformation is represented by the degree of grid deformation (the greater the deformation, the more obvious the grid stretching), finally forming an intuitive three-dimensional spatial distribution image of the dynamic deformation field of the weld area.

[0060] Extract the complete information of each spatial point from the three-dimensional spatial distribution expression, including the three-dimensional coordinates (X, Y, Z accurate to 0.01 mm), the original value and change rate of the width shrinkage (the difference from the previous moment divided by the time interval), the temperature gradient value and the absolute value of the temperature in three directions, the stress deformation value and stress magnitude of the fixture in two directions, and extract the standardized data of the corresponding spatial point from the dynamic deformation feature vector, including the standardized values ​​of each dimension, the timestamp of data acquisition, and the confidence of the data (calculated according to the sensor accuracy, such as the confidence of laser sensor data is 0.95, and that of infrared data is 0.90). Establish the spatial coordinate and feature vector index. Mapping relationship, the three-dimensional coordinates of each spatial point correspond to the unique index position in the feature vector. The corresponding standardized data can be quickly found through coordinate calculation (for example, the coordinate (X=10.0, Y=2.5, Z=0.5) corresponds to the 1256th index). The three-dimensional distribution data and feature vector data of the same spatial point are merged into one record, which contains 15 items of information including timestamp, three-dimensional coordinates, original deformation data, standardized feature data, and data confidence. The records of all spatial points are arranged in timestamp order to form a structured data set. Each time point contains approximately 50,000 spatial point records, which is a complete deformation association data set.

[0061] Through synchronous acquisition and unified timestamp marking, the time difference of different sensor data is eliminated. Combined with spatial coordinate mapping, the width, temperature, stress and other data are accurately corresponded in time and space. Through multi-dimensional data fusion, the originally independent physical quantities are integrated into a unified feature vector, which not only retains the original characteristics of each parameter, but also achieves the comparability of data of different dimensions through standardization processing, and can more comprehensively reflect the complex characteristics of weld deformation. The three-dimensional spatial distribution expression converts abstract deformation data into visual images, so that operators can directly observe the distribution law, strength change and development trend of deformation in space, which is convenient for quickly locating key deformation areas and deformation. The associated dataset integrates raw data, feature data, and spatial information, which can not only provide accurate environmental parameters for path replanning, but also provide a multi-dimensional basis for welding quality assessment, thereby improving data reuse and decision-making support capabilities. Through fine grid division and interpolation processing, discrete sensor data is expanded into a continuous spatial deformation field, which can capture tiny deformation changes and local anomalies, and improve the perception accuracy of dynamic deformation of welds. Due to the high frequency of data acquisition and the coherent processing flow, the generated deformation associated dataset can reflect the latest status of the weld in real time, providing a data basis for the robot control system to quickly respond to deformation changes and adjust the welding strategy in a timely manner.

[0062] In another preferred embodiment of the present invention, the above step 5, in addition to determining the dynamic compensation amount based on the distribution characteristics of the deformation correlation data, updating the multi-dimensional collaborative evaluation parameters in real time, and driving the dynamic energy-saving control of the welding parameters and motion trajectory, may include: Step 554 , performing Gaussian distribution statistical analysis on the deformation correlation data set, calculating and extracting the deformation mean parameter representing the overall deformation characteristics of the weld area and the deformation variance parameter representing the degree of deformation dispersion, as dynamic compensation control variables; Step 555: Input the dynamic compensation control amount into the multi-dimensional collaborative evaluation system, and modify the weight ratio parameters and energy consumption restriction condition parameters related to the thermal deformation effect in the evaluation system in real time based on the dynamic compensation control amount. Step 556 , based on the modified multi-dimensional collaborative evaluation system, reversely adjust the welding current setting value in the welding process parameters and the motion trajectory parameters of each joint of the robot; Step 557, by coordinating the welding current setting value and the robot joint motion trajectory parameters, the actual welding movement speed of the robot end effector can adaptively match the current dynamic deformation rate change trend of the weld area, thereby realizing dynamic energy-saving control during the welding process.

[0063] In an embodiment of the present invention, all valid data records are screened from the deformation-related data set, and abnormal values ​​caused by sensor failure or signal interference are excluded (such as values ​​that obviously exceed the physical reasonable range, such as a point where the width shrinkage suddenly appears +5 mm). The screened data are divided into three categories according to the physical parameter type: weld width shrinkage data set, temperature gradient data set, and fixture stress deformation data set. Taking the weld width shrinkage data set as an example, the shrinkage values ​​of all spatial points contained in the data set are counted. Assuming that there are 20,000 valid spatial points in the data set, the shrinkage values ​​of each point are counted. The shrinkage is recorded accurately to 0.01 mm (such as -0.45 mm, -0.62 mm, -0.58 mm, etc.). The deformation mean parameter is calculated, and the 20,000 shrinkage values ​​are added one by one to obtain the total (such as the total is -11200.50 mm). The total is then divided by the total number of data points 20,000 to obtain the mean of the width shrinkage (-11200.50 ÷ 20,000 ≈ -0.56 mm). This mean reflects the average shrinkage level of the entire weld area in the width direction. The larger the absolute value of the mean, the more obvious the overall shrinkage.

[0064] To calculate the deformation variance parameter, first calculate the difference between the shrinkage of each data point and the mean (for example, the shrinkage of a point is -0.45 mm, and the difference from the mean of -0.56 mm is 0.11 mm); then square each difference (0.11 mm squared is 0.0121 square mm); then add all the square values ​​to get the sum of squares (for example, the sum of all square values ​​is 32.80 square mm); finally, divide the sum of squares by the total number of data points 20,000 to get the variance of the width shrinkage (32.80 ÷ 20,000 = 0.001 64 square millimeters), the larger the variance value, the more significant the difference in shrinkage at different spatial points, and the more uneven the deformation distribution. For the temperature gradient data set (recording the temperature gradient values ​​of each point in the X, Y, and Z directions) and the fixture stress deformation data set (recording the stress deformation values ​​of each point in the horizontal and vertical directions), the above mean and variance calculation process is repeated to obtain the mean of the temperature gradient in the three directions (such as the mean in the X direction is 35℃ / mm, the mean in the Y direction is 28℃ / mm, and the mean in the Z direction is 8℃ / mm) and the variance (such as the variance in the X direction is 4.2℃² / mm) 2 ), as well as the mean (such as 0.12 mm in the horizontal direction and 0.09 mm in the vertical direction) and variance (such as 0.0025 square millimeters in the horizontal direction) of stress deformation in two directions. The three means (width shrinkage, temperature gradient comprehensive value, stress deformation comprehensive value) and three variances (corresponding to the discrete degree of the three types of parameters) obtained by the above calculations are integrated into a set of dynamic compensation control quantities, with a total of 6 specific values, which are used to correct the evaluation system.

[0065] The multi-dimensional collaborative evaluation system initially includes five core evaluation dimensions, namely welding quality stability (initial weight 30%), thermal deformation control effect (initial weight 20%), motion trajectory accuracy (initial weight 20%), energy consumption level (initial weight 20%), and equipment operation safety (initial weight 10%). The total weight of each dimension is 100%. At the same time, the initial energy consumption limit condition parameters are set, with an upper limit of 180A for welding current, an upper limit of 30V for welding voltage, an upper limit of 500W for robot joint motion power, and an upper limit of 50,000J for total energy consumption in a single welding process. The temperature gradient mean value in the dynamic compensation control quantity (such as the comprehensive mean value of 32°C / mm) and the preset The relationship between the actual mean value and the thermal deformation warning threshold (25°C / mm) is that since the actual mean value exceeds the threshold, it indicates that the impact of thermal deformation has intensified, and the weight ratio of the "thermal deformation control effect" dimension needs to be increased. According to the preset adjustment rule (the weight increases by 1% for every 1°C / mm the temperature gradient exceeds the threshold), the weight value to be increased is calculated (32-25=7°C / mm, corresponding to a 7% increase), and the weight of this dimension is revised from 20% to 27%. Correspondingly, the weights of other dimensions are reduced to maintain a total of 100%. Priority is given to reducing the weight of the "motion trajectory accuracy" dimension, which has a lower correlation with the current working conditions, from 20% to 13% (a 7% reduction) to ensure that the weight distribution is tilted towards thermal deformation control.

[0066] According to the comparison results of the width shrinkage variance (such as 0.00164 square millimeters) and the preset uniformity threshold (0.001 square millimeters), the actual variance exceeds the threshold, indicating that the deformation distribution unevenness has increased. It is necessary to increase the proportion of the "local quality deviation" sub-dimension in the "welding quality stability" (from the original 20% to 30%), so that the evaluation system pays more attention to the local abnormal deformation area. Combined with the relationship between the stress deformation mean (such as the comprehensive mean 0.11 mm) and the fixture safety stress threshold (0.10 mm), the actual mean slightly exceeds the threshold, indicating that the force on the fixture is close to the safety upper limit and the energy needs to be adjusted. The energy consumption limit condition parameters are adjusted, and the short-term current limit is appropriately relaxed to enhance welding stability (the current upper limit is raised from 180A to 185A), but the long-term energy consumption constraint is tightened (the total energy consumption upper limit is reduced from 50,000J to 48,000J) to avoid excessive equipment load due to excessive energy consumption. The corrected weight ratio parameters (such as thermal deformation control effect 27%, motion trajectory accuracy 13%, etc.) and energy consumption limit condition parameters (such as current upper limit 185A, total energy consumption 48,000J, etc.) are updated in real time to the multi-dimensional collaborative evaluation system to ensure that the system can accurately reflect the impact of the current deformation state on the welding process.

[0067] Key constraint indicators are extracted from the revised multi-dimensional collaborative evaluation system. Due to the increased weight of "thermal deformation control effect", the system's requirements for "reducing heat input to reduce deformation" are strengthened. At the same time, the "energy consumption level" dimension still maintains a 20% weight. It is necessary to find a balance between temperature control and energy saving. The welding current setting value is reversely calculated, referring to the average temperature gradient of 32°C / mm (exceeding the 25°C / mm threshold). According to the rule of "negative correlation between temperature gradient and welding current" in the evaluation system (for every 1°C / mm above the threshold, the current needs to be reduced by 2A), the current adjustment amount is calculated (32-25=7°C / mm, corresponding to a reduction of 14A). The original current setting value is 150A, and the revised current setting value is 150-14=136A to reduce heat input. At the same time, it is checked whether the current value complies with the revised energy consumption limit (136A < 185A upper limit). After confirming compliance, the setting is retained.

[0068] The spatial distribution characteristics of the width shrinkage amount are analyzed (combined with the dispersion of variance 0.00164 square millimeters), in the area with larger shrinkage amount (such as locally-0.8 millimeters, much higher than the average-0.56 millimeters), the smoothness parameter of the robot joint motion trajectory needs to be adjusted, the joint angle change rate in the original trajectory in this area is 30 degrees per second, and after correction, it is reduced to 20 degrees per second, at the same time, the path curvature radius is increased from 6 millimeters to 10 millimeters, by slowing down the motion speed and reducing the turning amplitude, the local deformation caused by mechanical vibration is avoided, for the area near the fixture where the stress deformation exceeds the standard (such as horizontal direction deformation 0.13 millimeters), the safety distance parameter of the trajectory is adjusted in reverse, the minimum distance between the original trajectory and the fixture is 1.0 millimeter, and after correction, it is increased to 1.5 millimeters, at the same time, the joint motion priority is adjusted (the shoulder joint is adjusted first, and then the elbow joint), the indirect force on the fixture during robot operation is reduced, and the stress deformation is reduced.

[0069] All the adjusted parameters are checked again to ensure that the power corresponding to the welding current 136A (combined with the voltage 25V, the power is 136*25=3400W) is lower than the upper limit of the equipment power, and after the joint motion parameter adjustment, the trajectory accuracy still meets the minimum requirement of the "motion trajectory accuracy" dimension in the evaluation system (deviation not more than ±0.1 millimeter), if not, fine tuning is performed (such as reducing the current by 2A to 134A), until all parameters meet the constraints of the modified system.

[0070] The rate change of the dynamic deformation of the weld area is calculated in real time, the width shrinkage amount data of two adjacent time stamps (interval 0.05 seconds) are selected, such as the average of the previous moment is-0.56 millimeters, and the average of the current moment is-0.59 millimeters, the difference between the two is-0.03 millimeters, and the average shrinkage rate is obtained by dividing the difference by the time interval (0.05 seconds), that is-0.03÷0.05=-0.6 millimeters / second, the negative sign indicates that the shrinkage rate is accelerating, according to the shrinkage rate, the end effector moving speed is adjusted, when the absolute value of the shrinkage rate (0.6 millimeters / second) is greater than the preset rate threshold (0.3 millimeters / second), it means that the deformation develops quickly, the welding moving speed needs to be reduced to match the deformation rhythm, the original moving speed is 8 millimeters / second, and after correction, it is reduced to 5 millimeters / second, so that the welding process has enough time to adapt to the deformation, and the poor weld joint caused by too fast speed is avoided.

[0071] The welding current and the moving speed are cooperatively adjusted to realize energy saving. After the moving speed is reduced (from 8 mm / s to 5 mm / s), in order to ensure that the heat input per unit length of the weld is stable (to avoid excessive heat input due to the reduction of the speed), the current is adjusted according to the principle that the speed is proportional to the current. The original speed of 8 mm / s corresponds to the current of 136 A, and the new speed of 5 mm / s corresponds to the current of 136 * (5 ÷ 8) = 85 A. The adjustment reduces the power consumption per unit time from 136 * 25 = 3400 W to 85 * 25 = 2125 W, thereby reducing unnecessary energy waste. For the area with a gentle deformation rate (for example, the shrinkage rate is -0.2 mm / s, and the absolute value is less than the threshold of 0.3 mm / s), the moving speed is increased to 10 mm / s, and the current is proportionally increased to 136 * (10 ÷ 8) = 170 A (which does not exceed the upper limit of 185 A). On the premise of ensuring the welding quality, the welding time is shortened by increasing the speed, and the standby energy consumption of the equipment is reduced (for example, from the original 125 seconds per meter to 100 seconds, reducing 25 seconds of standby energy consumption of the equipment). A dynamic adjustment cycle is established, and the process of “calculating the deformation rate, adjusting the moving speed, and matching the welding current” is repeated every 0.05 seconds, so that the moving speed of the end effector always keeps consistent with the current trend of the deformation rate (walk slowly when the deformation is fast, and walk fast when the deformation is slow), and the current and the speed are cooperatively changed to maintain a reasonable heat input. Through the dynamic matching, the energy consumption per unit length of the entire welding process is reduced from the original 300 J / mm to 220 J / mm, thereby realizing the energy saving effect.

[0072] The mean and variance parameters obtained by Gaussian distribution statistics can accurately reflect the overall deformation trend of the weld and capture local deformation differences, so that the dynamic compensation control quantity is more suitable for the actual working condition, and the one-sidedness of single parameter evaluation is avoided. The weight proportion and energy consumption parameter are real-time corrected according to the deformation state, so that the evaluation system always matches the current welding environment, and the evaluation of the welding quality, energy consumption and other dimensions is more accurate, thereby providing a reliable basis for parameter adjustment. The welding current and the motion trajectory can be targeted to deal with problems such as thermal deformation and stress concentration (for example, reducing the current in the high-temperature area and adjusting the trajectory in the high-stress area) through the reverse adjustment mechanism combined with the corrected evaluation system, thereby improving the stability of the welding quality. Through the cooperative adjustment of the speed and the current, the welding quality is adapted to the deformation rhythm while avoiding energy waste (reducing heat redundancy by reducing the current at low speed, and reasonably increasing the efficiency to shorten the time at high speed), thereby reducing the energy consumption per unit welding length.

[0073] As shown in Figure 2 The embodiment of the present application also provides a dynamic energy-saving path planning system for a teaching-free welding robot, which comprises: A scanning mapping module is configured to scan a workpiece by a laser vision device, acquire weld coordinates, obstacle information and workpiece thermal deformation data, and extract weld feature points to establish a mapping relationship between a workpiece coordinate system and a robot base coordinate system. The path generation module is used to generate an anti-drift path by implanting dynamic response nodes into the initial path based on the mapping relationship, combining the thermal deformation prediction mechanism with sensor feedback data, and calculating the path point offset using the feedback data; The collaborative evaluation module is used to build a multi-dimensional collaborative evaluation mechanism for the anti-drift path, synchronously adjust and comprehensively balance the path length, energy consumption of the entire welding process, and welding quality parameters to obtain multi-dimensional collaborative evaluation results; The search and planning module is used to perform global path search based on the results of multi-dimensional collaborative evaluation, using a bidirectional tree expansion. When weld deformation, fixture offset, and obstacle displacement exceeding the set threshold are detected, a local re-path planning mechanism is triggered; The dynamic energy-saving module is used to collect the dynamic deformation characteristics of the weld area in real time through a multi-source deformation monitoring device based on a re-planning mechanism, generate deformation-related data, determine the dynamic compensation amount based on the distribution characteristics of the deformation-related data, update the multi-dimensional collaborative evaluation parameters in real time, and drive the dynamic energy-saving control of welding parameters and motion trajectories.

[0074] It should be noted that this system is a system corresponding to the above method, and all implementation methods in the above method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0075] An embodiment of the present invention further provides a computing device comprising: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the above-described method. All implementations in the above-described method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0076] The embodiment of the present invention further provides a computer-readable storage medium storing instructions, which, when executed on a computer, causes the computer to execute the above-described method. All implementations in the above-described method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0077] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A dynamic energy-saving path planning method for a teaching-free welding robot, characterized in that: The method comprises: Step 1: Scan the workpiece with a laser vision device to obtain weld coordinates, obstacle information, and workpiece thermal deformation data, extract weld feature points, and establish a mapping relationship between the workpiece coordinate system and the robot base coordinate system; Step 2: Based on the mapping relationship, combined with the thermal deformation prediction mechanism and sensor feedback data, dynamic response nodes are implanted in the initial path and the feedback data is used to calculate the path point offset to generate an anti-drift path; Step 3: Build a multi-dimensional collaborative evaluation mechanism for the anti-drift path, synchronously adjust and comprehensively balance the path length, energy consumption of the entire welding process, and welding quality parameters to obtain a multi-dimensional collaborative evaluation result; Step 4: Based on the multi-dimensional collaborative evaluation results, a global path search is performed using a bidirectional tree expansion. When weld deformation, fixture offset, and obstacle displacement exceed the set threshold, a local re-path planning mechanism is triggered. Step 5: Based on the re-planning mechanism, the dynamic deformation characteristics of the weld area are collected in real time through a multi-source deformation monitoring device to generate deformation-related data. The dynamic compensation amount is determined based on the distribution characteristics of the deformation-related data, and the multi-dimensional collaborative evaluation parameters are updated in real time to drive the dynamic energy-saving control of the welding parameters and motion trajectory.

2. The dynamic energy-saving path planning method for a teaching-free welding robot according to claim 1 is characterized in that: Based on the mapping relationship, combined with the thermal deformation prediction mechanism and sensor feedback data, dynamic response nodes are implanted in the initial path and the feedback data is used to calculate the path point offset to generate an anti-drift path, including: Based on the mapping relationship, combined with the thermal expansion characteristics of the material and the heat input characteristics during welding, the expected deformation of the characteristic points on the initial weld path is calculated; Based on the expected deformation, the coordinates of the feature points of the initial path are reversely superimposed on the direction opposite to the deformation to generate a thermal deformation pre-compensation path; Based on the thermal deformation pre-compensation path, the laser vision sensor collects the actual deformation data of the workpiece during welding online, and performs a real-time spatial error comparison between the actual deformation data and the expected deformation to obtain the error comparison result; Based on the error comparison results, the curvature mutation area and the error exceeding limit point are marked in the pre-compensation path, and the dynamic response node is implanted at the marked position to calculate the three-dimensional spatial position adjustment amount at the dynamic response node; Based on the three-dimensional spatial position adjustment amount, the spatial coordinates of the corresponding nodes in the pre-compensation path are corrected in real time to generate a thermal deformation-resistant offset path.

3. The dynamic energy-saving path planning method for a teaching-free welding robot according to claim 2 is characterized in that: Based on the error comparison results, the curvature mutation area and the error exceeding limit point are marked in the pre-compensation path, and the dynamic response node is implanted at the marked position. The three-dimensional spatial position adjustment amount at the dynamic response node is calculated, including: For the marked deformation error exceeding limit point, the position compensation weighting coefficient of the deformation error exceeding limit point is calculated based on the curvature change rate of the adjacent path segments and the thermal deformation gradient at the deformation error exceeding limit point; For the marked curvature mutation area, the normal vector direction of the weld feature point in the area is extracted, and the material shrinkage prediction relationship included in the thermal deformation prediction mechanism is integrated to calculate the normal compensation component of the regional feature point; The position compensation weighted coefficient is integrated with the normal compensation component to generate the three-dimensional spatial position adjustment of the corresponding dynamic response node.

4. The dynamic energy-saving path planning method for a teaching-free welding robot according to claim 3 is characterized in that: For the anti-drift path, a multi-dimensional collaborative evaluation mechanism is established to synchronously adjust and comprehensively balance the path length, energy consumption of the entire welding process, and welding quality parameters to obtain multi-dimensional collaborative evaluation results, including: Based on the anti-drift path, an evaluation system is established that includes the impact degree of three aspects: path length, energy consumption, and welding quality; Taking the evaluation system as the target, the spatial coordinate sequence of the path points of the anti-drift path and the robot motion speed parameters are adjusted simultaneously through the gradient descent method to generate the initial improved path; When the welding quality parameter in the initial improvement path is less than the set threshold, the path point spacing distribution of the path is adjusted to generate a path with enhanced penetration; when the energy consumption parameter in the initial improvement path is greater than the limit, the robot motion acceleration is reduced and the path turning curvature is smoothed to generate an energy consumption improvement path; The penetration enhancement path and the energy consumption improvement path are collaboratively balanced to generate the adjusted path parameters and motion parameter sets, i.e., the multi-dimensional collaborative evaluation results.

5. The dynamic energy-saving path planning method for a teaching-free welding robot according to claim 4 is characterized in that: Based on the results of multi-dimensional collaborative evaluation, a global path search is performed using a bidirectional tree expansion. When weld deformation, fixture offset, and obstacle displacement exceed the set threshold, a local re-path planning mechanism is triggered, including: Based on the path parameters and motion parameter sets, a bidirectional fast-expanding random tree global path search is performed starting from the planned path start position and end position synchronously; In the global path search, the real-time deformation monitoring data of the weld area is acquired and compared with the preset deformation safety threshold; the real-time displacement sensor data on the workpiece fixture is compared with the preset fixture offset safety threshold; the real-time contour change data of obstacles in the workspace is compared with the preset obstacle displacement safety threshold; When any one of the weld deformation data, fixture displacement data, and obstacle contour change data exceeds the corresponding preset safety threshold, the current bidirectional rapid expansion random tree global search process is immediately interrupted; After the interruption is triggered, the actual spatial position and joint state of the robot end effector at the moment of interruption are used as the new planning starting point, and the local path replanning mechanism is immediately started to generate a local path segment that adapts to the current environmental changes.

6. The method for dynamic energy-saving path planning of a teaching-free welding robot according to claim 5, characterized in that: Based on the re-planning mechanism, the dynamic deformation characteristics of the weld area are collected in real time through the multi-source deformation monitoring device to generate deformation correlation data, including: By deploying a multi-source deformation monitoring device in the welding area, the weld width shrinkage data, temperature gradient distribution data, and fixture stress deformation data are simultaneously acquired and time-stamped. The weld width shrinkage data, temperature gradient distribution data, and fixture stress and deformation data are numerically fused and calculated according to the corresponding time and space positions to generate a dynamic deformation feature vector containing multi-dimensional information. Based on the dynamic deformation feature vector, a three-dimensional spatial distribution expression of the dynamic deformation field in the weld area is constructed; The three-dimensional spatial distribution expression is associated and integrated with the dynamic deformation feature vector to form a deformation association dataset including dynamic deformation features and spatial distribution information.

7. The method for dynamic energy-saving path planning of a teaching-free welding robot according to claim 6, characterized in that: Based on the distribution characteristics of deformation-related data, the dynamic compensation amount is determined, multi-dimensional collaborative evaluation parameters are updated in real time, and dynamic energy-saving control of welding parameters and motion trajectories is driven, including: Gaussian distribution statistical analysis is performed on the deformation correlation data set to calculate and extract the deformation mean parameter representing the overall deformation characteristics of the weld area and the deformation variance parameter representing the degree of deformation dispersion as dynamic compensation control quantities; The dynamic compensation control quantity is input into the multi-dimensional collaborative evaluation system, and the weight ratio parameters and energy consumption limit condition parameters related to the thermal deformation effect in the evaluation system are corrected in real time based on the dynamic compensation control quantity; Based on the revised multi-dimensional collaborative evaluation system, the welding current setting value in the welding process parameters and the motion trajectory parameters of each joint of the robot are reversely adjusted; By coordinating the welding current setting value and the robot joint motion trajectory parameters, the actual welding movement speed of the robot end effector can adaptively match the rate change trend of the current dynamic deformation in the weld area, thereby achieving dynamic energy-saving control during the welding process.

8. A dynamic energy-saving path planning system for a teaching-free welding robot, the system implementing the method according to any one of claims 1 to 7, characterized in that: include: The scanning and mapping module is used to scan the workpiece through a laser vision device to obtain the weld coordinates, obstacle information and workpiece thermal deformation data, extract weld feature points, and establish a mapping relationship between the workpiece coordinate system and the robot base coordinate system; The path generation module is used to generate an anti-drift path by implanting dynamic response nodes into the initial path based on the mapping relationship, combining the thermal deformation prediction mechanism with sensor feedback data, and calculating the path point offset using the feedback data; The collaborative evaluation module is used to build a multi-dimensional collaborative evaluation mechanism for the anti-drift path, synchronously adjust and comprehensively balance the path length, energy consumption of the entire welding process, and welding quality parameters to obtain multi-dimensional collaborative evaluation results; The search and planning module is used to perform global path search based on the results of multi-dimensional collaborative evaluation, using a bidirectional tree expansion. When weld deformation, fixture offset, and obstacle displacement exceeding the set threshold are detected, a local re-path planning mechanism is triggered; The dynamic energy-saving module is used to collect the dynamic deformation characteristics of the weld area in real time through a multi-source deformation monitoring device based on a re-planning mechanism, generate deformation-related data, determine the dynamic compensation amount based on the distribution characteristics of the deformation-related data, update the multi-dimensional collaborative evaluation parameters in real time, and drive the dynamic energy-saving control of welding parameters and motion trajectories.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images