Stainless steel weld adaptive polishing path planning and force control method

Abstract

The application discloses a kind of stainless steel weld seam self-adapting polishing path planning and force control method, obtains weld seam surface three-dimensional data, reconstructs weld seam three-dimensional profile by feature extraction, and accordingly self-adapting polishing path is planned and tool posture is adjusted in real time;Subsequently, a contact mechanics model is established, and an impedance control algorithm is designed;Real-time acquisition of force data during polishing, calculate the position correction amount using fuzzy PID algorithm, dynamically adjust the end normal position offset of the robot arm, flexibly compensate the weld seam excess height change and clamping error, and realize constant force polishing;Iterative optimization is carried out while monitoring process signals.The application effectively solves the problems of poor track adaptability, large contact force fluctuation and easy overcutting and missing polishing in traditional polishing technology, significantly improves the automation level, processing efficiency and surface quality consistency of stainless steel weld seam polishing.

Description

Technical Field

[0001] This invention relates to the field of industrial grinding technology, and in particular to an adaptive grinding path planning and force control method for stainless steel welds. Background Technology

[0002] Currently, industrial robotic arm grinding is increasingly widely used in the grinding field, but most still employ traditional "teach-and-playback" or offline programming modes. This method relies on operators pre-setting fixed motion trajectories, requiring extremely high workpiece consistency and positioning accuracy. However, in actual production, due to fluctuations in welding processes, thermal deformation, and assembly errors, the width, height, and orientation of stainless steel welds often exhibit irregular variations. When the weld deviates or bends, the rigid trajectory cannot automatically follow, easily causing the grinding path to deviate from the weld center. For welds with high reinforcement height, fixed trajectories can easily cause "over-grinding," damaging the stainless steel base material and destroying the anti-corrosion layer; for areas with low reinforcement height or depressions, "missed grinding" occurs, leaving residual welds and making it difficult to meet surface flatness requirements. Stainless steel has poor thermal conductivity, a large coefficient of linear expansion, and a strong tendency for work hardening, making its grinding process extremely sensitive to contact force. Existing automated grinding equipment mostly uses pure position control, lacking force sensing and control capabilities, facing severe challenges: workpiece clamping and positioning errors, weld thermal deformation, and microscopic unevenness of the workpiece surface are unavoidable. Under rigid position control, these errors directly translate into drastic fluctuations in normal contact force. Excessive contact force leads to a sharp increase in grinding heat, causing burns, blackening, and even cracks on the stainless steel surface; insufficient contact force results in low grinding efficiency and unstable surface quality. The lack of a "flexible buffer" mechanism between the grinding tool and the workpiece prevents real-time pressure adjustment based on feel, unlike manual grinding, easily causing belt breakage or robot arm overload alarms. The grinding process is dynamic; belt wear and changes in material removal both affect the final result. Most existing technologies are "open-loop" systems, unable to monitor grinding current, vibration, and force feedback signals in real time, and unable to adjust the number of grinding passes or feed rate online based on the actual grinding effect. Once uneven grinding or quality defects occur, manual rework is the only remedial method, resulting in low production efficiency and difficulty meeting the stringent surface quality requirements of high-end stainless steel products. Summary of the Invention

[0003] An adaptive grinding path planning and force control method for stainless steel welds includes the following steps: S1. Acquisition and preprocessing of 3D data of weld surface: The stainless steel weld area is scanned using a line laser sensor or structured light vision sensor to acquire the original 3D point cloud data of the weld surface. The acquired point cloud data is denoised, filtered and simplified, and the vision sensor coordinate system and the robot arm base coordinate system are calibrated and unified through coordinate transformation. S2. Weld Feature Extraction and Contour Reconstruction: Based on the preprocessed point cloud data, feature extraction algorithms (such as the section method or normal vector estimation algorithm) are used to identify the centerline position, width and height of the weld. Based on the extracted feature points, the three-dimensional spatial contour curve of the weld is reconstructed to determine the geometric boundary and reinforcement distribution of the weld. S3. Adaptive Grinding Path Planning: Based on the reconstructed 3D contour of the weld, an initial grinding trajectory is generated. According to the changes in the width and height of the weld, the posture of the grinding tool (such as keeping the grinding head always perpendicular to the tangential direction of the weld) and the grinding spacing are adaptively adjusted. A reciprocating grinding path is planned using a variable step distance or equal residual height strategy, and a robot arm motion program containing pose information is generated. S4. Force control strategy model establishment: Establish a contact mechanics model between the grinding tool and the stainless steel workpiece surface. Based on the hardness of the stainless steel material and the grinding process requirements (such as surface roughness and removal amount), calculate the theoretical normal grinding force, design impedance control or admittance control algorithm, switch the position control mode to force-position hybrid control mode, and set the target contact force threshold and damping parameters. S5. Force-position coordinated control and real-time compensation: During the grinding process, the end contact force data is collected in real time by force / torque sensor. The actual contact force is compared with the target set force. The position correction amount is calculated by using PID or fuzzy PID algorithm. By adjusting the position offset of the end effector of the robot arm in the normal direction in real time, flexible compensation for the change of weld height and workpiece clamping error is achieved, ensuring constant force grinding. S6. Process monitoring and iterative optimization: Real-time monitoring of current, vibration and force feedback signals during the grinding process. When uneven weld removal or surface quality failure is detected, the system adaptively adjusts the grinding speed, feed rate or number of grinding passes, and performs iterative grinding until the surface flatness and smoothness of the weld meet the stainless steel process standards.

[0004] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In step S2, a feature extraction algorithm is used to identify the centerline position, width, and height of the weld. Based on the extracted feature points, the specific steps are as follows; S21. Point cloud slicing and section acquisition: Since welds are usually long and narrow, the "section method" can be used to process them. That is, along the extension direction of the weld, the point cloud data is sliced ​​vertically at a certain step size to obtain a series of two-dimensional section point sets. S22. Feature point recognition algorithm: In each two-dimensional cross-sectional point set, the boundary points (weld toes) and peak points of the weld are identified based on the rate of curvature change or the sudden change of the normal vector. The curvature value of each point is calculated, and a threshold is set to filter out the point with the largest curvature as a candidate feature point. Combined with the reflective characteristics of stainless steel, noise caused by reflection needs to be removed. The cross-sectional contour curve is fitted by the least squares method to smooth out abnormal protrusions. S23. Calculation of weld centerline and width: Connect the crests of each section to form the weld centerline, calculate the distance between the left and right boundary points (weld toes) to obtain the weld width, and calculate the distance from the crest to the base material surface to obtain the weld reinforcement. S24. Contour Mathematical Modeling: Based on the extracted centerline coordinates, a smooth three-dimensional spatial curve equation of the weld is fitted using B-spline or cubic spline interpolation methods, providing a continuous navigation reference for subsequent path planning.

[0005] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In specific step S22, the boundary points, weld toes, and wave crests of the weld are identified based on the abrupt change in the rate of curvature change, as follows; Calculate the curvature value at each point in the weld section point cloud. Set the weld boundary identification conditions as follows The rate of change of curvature threshold, and the angle between the normal vector at that point and the normal vector of the parent material plane. Only points that simultaneously meet the above conditions are identified as weld boundary points, thus eliminating interference from surface scratches.

[0006] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In step S3, the posture of the grinding tool is adaptively adjusted according to the changes in the width and height of the weld. The specific steps are as follows; S31. Adaptive tool posture adjustment (normal following): The grinding quality largely depends on the contact angle between the grinding head and the workpiece surface. Based on the reconstructed weld surface model, the normal vector and tangent vector at the grinding point are calculated, and the posture of the end effector is adjusted so that the contact wheel axis of the abrasive belt or grinding wheel is perpendicular to the tangent vector. At the same time, the grinding pressure direction is parallel to the normal vector and forms a specific process angle to ensure uniform contact. S32. Variable step pitch trajectory generation strategy: For stainless steel welds with varying widths, a strategy of equal residual height and variable step pitch is adopted to plan the reciprocating grinding path. In the wider area of ​​the weld, the grinding row spacing (step pitch) is automatically reduced to increase the grinding overlap rate and avoid missed grinding. In the narrower area of ​​the weld or the base material area, the step pitch is appropriately increased to improve efficiency and generate a series of discrete path points containing position and attitude information. S33. Speed ​​planning: Dynamically allocate the feed speed v according to the weld height H. In areas with high weld height (large amount of material removed), reduce the robot arm feed speed to ensure controllable single cutting depth; in areas with low weld height, appropriately increase the speed to prevent over-wearing.

[0007] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In specific step S32, a strategy of equal residual height and variable step distance is used to plan the reciprocating grinding path, as follows; Variable step distance logic: Introduce chord height error constraints to perform adaptive interpolation of path points, and calculate the distance from the midpoint of the line connecting adjacent path points to the actual weld curve: chord height h; Set the maximum allowable chord height error h max ; When h>h max Then, a new path point is inserted between the two points until all path segments meet the error requirements. The path points are dense in areas with large weld curvature and sparse in flat areas, balancing accuracy and efficiency.

[0008] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In step S4, a contact mechanics model between the grinding tool and the stainless steel workpiece surface is established. The specific steps are as follows: S41. Simplification of Geometric Contact Model: First, the complex physical contact scenario needs to be abstracted into a simple geometric model: the grinding tool (such as an elastic contact wheel) is regarded as an elastic cylinder, and the stainless steel weld surface is regarded as a rigid plane or curved surface. It is assumed that under the action of normal pressure, the contact area between the two forms a rectangular surface. The size of this surface depends on the magnitude of the pressure and the hardness of the contact wheel. The core variables in the model are defined as: normal grinding force (pressure), contact wheel hardness (elastic modulus), contact wheel radius, and grinding speed. S42. Contact Deformation and Pressure Distribution Analysis: When the robotic arm applies a normal force, the soft contact wheel will undergo compressive deformation. According to Hertz's contact theory (a classic elastic contact theory), the greater the pressure, the larger the contact area, and the pressure is greatest at the center of the contact area and least at the edge. Determining the maximum pressure value on the contact surface is crucial, as this value determines whether the abrasive grains can penetrate the stainless steel surface and is a key parameter for determining grinding efficiency. Establishing the correspondence between normal force and tool indentation (deformation) is equivalent to determining the spring stiffness of the contact wheel, telling the robotic arm: how much force needs to be applied to indent to how deep; S43. Material Removal Rate Model Construction: Establish the relationship between force and grinding effect (removal amount), and establish a basic law - the thickness of metal removed per unit time mainly depends on the product of contact pressure and grinding speed, that is: the greater the pressure and the faster the grinding, the more material is removed. Considering the stickiness and severe work hardening of stainless steel, a "grinding factor" is introduced. This factor is not fixed; it is related to the sharpness of the abrasive belt, the grit size, and the type of stainless steel. S44. Tangential Grinding Resistance Analysis: In addition to the vertically downward normal force, it is also necessary to analyze the resistance in the horizontal direction. During the grinding process, the rotation of the grinding head will generate tangential resistance. According to the principle of tribology, the tangential resistance is usually proportional to the normal force. Condition monitoring correlation: Establish the relationship between tangential force and grinding condition. When the tangential force suddenly increases, the model should be able to determine that it may be due to excessive feed speed or severe abrasive belt wear, thus providing a basis for subsequent force control adjustments. S45. Experimental Calibration of Model Parameters: After the theoretical model is established, the unknown parameters are "calibrated" through experiments, including the following steps: Specimen grinding experiment: A fixed-point grinding experiment was carried out on a standard stainless steel specimen block, with different pressure levels and grinding speeds set. Data measurement: Measure the material removal depth after each grinding process and record the actual force signal data; Parameter fitting: Substitute the experimental data into the previous theoretical model to calculate the "grinding coefficient" and "pressure index" that best match the actual situation. In this way, the model is transformed from a "theoretical formula" into a "practical tool" that can accurately predict the actual processing effect.

[0009] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In step S5, the positional offset of the end effector of the robotic arm in the normal direction is adjusted in real time to achieve flexible compensation for changes in weld reinforcement height and workpiece clamping errors, as detailed below: Closed-loop force error correction strategy: Real-time acquisition of end force sensor data, calculation of the difference between the actual contact force and the target set force, and proportional calculation of the position adjustment based on the magnitude of the force difference. The larger the force difference, the larger the position offset; conversely, when the actual force is greater than the target force, the end is controlled to retreat along the normal direction; when the actual force is less than the target force, the end is controlled to feed along the normal direction. Weld height adaptive strategy: When grinding to the high point of the weld, the contact force increases instantaneously. The system quickly instructs the end of the robot arm to retreat along the normal direction to reduce the amount of material removed and prevent over-cutting. When grinding to the concave part of the weld, the contact force decreases. The system instructs the end of the robot arm to advance along the normal direction to increase the amount of material removed and avoid missing areas. Damping control is introduced during the adjustment process to prevent the robot arm from stiffening or oscillating due to fluctuations in the force signal, and to ensure that the grinding head smoothly crosses the peaks and troughs of the weld. Safety limiting and anti-shaking strategy: Set the maximum allowable range of normal position offset. When the calculated adjustment amount exceeds this range, the system immediately locks the position to prevent the robot arm from colliding with the workpiece and fixture. Set the maximum step size of a single adjustment. When encountering sudden force changes caused by weld beads or hard points, limit the single adjustment amplitude to avoid the robot arm from shaking violently due to overreaction.

[0010] The beneficial effects of this invention are as follows: By acquiring three-dimensional weld data and reconstructing the contour, it breaks the dependence on fixed trajectories in the traditional "teach-and-playback" mode. Employing a variable step size strategy and adaptive attitude adjustment, it can automatically plan the path based on real-time changes in weld width, height, and orientation. This not only effectively solves the problem of "over-grinding" damaging the base material or "under-grinding" leaving residual height due to weld irregularities, but also ensures that the grinding head is always in the optimal grinding posture, significantly improving grinding coverage and surface smoothness. Flexible constant-force grinding is achieved, greatly improving surface quality consistency. Addressing the sensitivity of stainless steel to grinding force and its susceptibility to burns, this invention establishes a contact mechanics model and employs an impedance control algorithm. By adjusting the normal position offset in real time using a fuzzy PID algorithm, a "force-position" coordinated closed-loop control system is constructed. This method can flexibly compensate for workpiece clamping errors, thermal deformation, and weld height fluctuations, stably controlling the contact force within the target range, effectively avoiding grinding vibrations, surface burns, and abrasive belt breakage caused by rigid contact, and ensuring the consistency of weld surface finish. This invention enhances the robustness and intelligence of the process system by introducing a process monitoring and iterative optimization mechanism to monitor current, vibration, and force feedback signals in real time. The system can adaptively adjust the speed, feed rate, and number of grinding passes online according to the grinding status, forming an intelligent closed loop with self-decision-making capabilities. Attached Figure Description

[0011] Figure 1 This is a flowchart of an adaptive grinding path planning and force control method for stainless steel welds. Detailed Implementation

[0012] A method for adaptive grinding path planning and force control of stainless steel welds, the process is as follows: Figure 1 The process, as shown, includes the following steps: S1. Acquisition and preprocessing of 3D data of weld surface: The stainless steel weld area is scanned using a line laser sensor or structured light vision sensor to acquire the original 3D point cloud data of the weld surface. The acquired point cloud data is denoised, filtered and simplified, and the vision sensor coordinate system and the robot arm base coordinate system are calibrated and unified through coordinate transformation. S2. Weld Feature Extraction and Contour Reconstruction: Based on the preprocessed point cloud data, feature extraction algorithms (such as the section method or normal vector estimation algorithm) are used to identify the centerline position, width and height of the weld. Based on the extracted feature points, the three-dimensional spatial contour curve of the weld is reconstructed to determine the geometric boundary and reinforcement distribution of the weld. S3. Adaptive Grinding Path Planning: Based on the reconstructed 3D contour of the weld, an initial grinding trajectory is generated. According to the changes in the width and height of the weld, the posture of the grinding tool (such as keeping the grinding head always perpendicular to the tangential direction of the weld) and the grinding spacing are adaptively adjusted. A reciprocating grinding path is planned using a variable step distance or equal residual height strategy, and a robot arm motion program containing pose information is generated. S4. Force control strategy model establishment: Establish a contact mechanics model between the grinding tool and the stainless steel workpiece surface. Based on the hardness of the stainless steel material and the grinding process requirements (such as surface roughness and removal amount), calculate the theoretical normal grinding force, design impedance control or admittance control algorithm, switch the position control mode to force-position hybrid control mode, and set the target contact force threshold and damping parameters. S5. Force-position coordinated control and real-time compensation: During the grinding process, the end contact force data is collected in real time by force / torque sensor. The actual contact force is compared with the target set force. The position correction amount is calculated by using PID or fuzzy PID algorithm. By adjusting the position offset of the end effector of the robot arm in the normal direction in real time, flexible compensation for the change of weld height and workpiece clamping error is achieved, ensuring constant force grinding. S6. Process monitoring and iterative optimization: Real-time monitoring of current, vibration and force feedback signals during the grinding process. When uneven weld removal or surface quality failure is detected, the system adaptively adjusts the grinding speed, feed rate or number of grinding passes, and performs iterative grinding until the surface flatness and smoothness of the weld meet the stainless steel process standards.

[0013] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In step S2, a feature extraction algorithm is used to identify the centerline position, width, and height of the weld. Based on the extracted feature points, the specific steps are as follows; S21. Point Cloud Slicing and Section Acquisition: Since welds are usually long and narrow, the "section method" can be used for processing. That is, along the extension direction of the weld, the point cloud data is sliced ​​vertically at a certain step size to obtain a series of two-dimensional cross-sectional point sets; S22. Feature point recognition algorithm: In each two-dimensional cross-sectional point set, the boundary points (weld toes) and peak points of the weld are identified based on the rate of curvature change or the sudden change of the normal vector. The curvature value of each point is calculated, and a threshold is set to filter out the point with the largest curvature as a candidate feature point. Combined with the reflective characteristics of stainless steel, noise caused by reflection needs to be removed. The cross-sectional contour curve is fitted by the least squares method to smooth out abnormal protrusions. S23. Calculation of weld centerline and width: Connect the crests of each section to form the weld centerline, calculate the distance between the left and right boundary points (weld toes) to obtain the weld width, and calculate the distance from the crest to the base material surface to obtain the weld reinforcement. S24. Contour Mathematical Modeling: Based on the extracted centerline coordinates, a smooth three-dimensional spatial curve equation of the weld is fitted using B-spline or cubic spline interpolation methods, providing a continuous navigation reference for subsequent path planning.

[0014] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In specific step S22, the boundary points, weld toes, and wave crests of the weld are identified based on the abrupt change in the rate of curvature change, as follows; Calculate the curvature value at each point in the weld section point cloud. Set the weld boundary identification conditions as follows The rate of change of curvature threshold, and the angle between the normal vector at that point and the normal vector of the parent material plane. Only points that simultaneously meet the above conditions are identified as weld boundary points, thus eliminating interference from surface scratches.

[0015] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In step S3, the posture of the grinding tool is adaptively adjusted according to the changes in the width and height of the weld. The specific steps are as follows; S31. Adaptive tool posture adjustment (normal following): The grinding quality largely depends on the contact angle between the grinding head and the workpiece surface. Based on the reconstructed weld surface model, the normal vector and tangent vector at the grinding point are calculated, and the posture of the end effector is adjusted so that the contact wheel axis of the abrasive belt or grinding wheel is perpendicular to the tangent vector. At the same time, the grinding pressure direction is parallel to the normal vector and forms a specific process angle to ensure uniform contact. S32. Variable step pitch trajectory generation strategy: For stainless steel welds with varying widths, a strategy of equal residual height and variable step pitch is adopted to plan the reciprocating grinding path. In the wider area of ​​the weld, the grinding row spacing (step pitch) is automatically reduced to increase the grinding overlap rate and avoid missed grinding. In the narrower area of ​​the weld or the base material area, the step pitch is appropriately increased to improve efficiency and generate a series of discrete path points containing position and attitude information. S33. Speed ​​planning: Dynamically allocate the feed speed v according to the weld height H. In areas with high weld height (large amount of material removed), reduce the robot arm feed speed to ensure controllable single cutting depth; in areas with low weld height, appropriately increase the speed to prevent over-wearing.

[0016] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In specific step S32, a strategy of equal residual height and variable step distance is used to plan the reciprocating grinding path, as follows; Variable step distance logic: Introduce chord height error constraints to perform adaptive interpolation of path points, and calculate the distance from the midpoint of the line connecting adjacent path points to the actual weld curve: chord height h; Set the maximum allowable chord height error h max ; When h>h max Then, a new path point is inserted between the two points until all path segments meet the error requirements. The path points are dense in areas with large weld curvature and sparse in flat areas, balancing accuracy and efficiency.

[0017] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In step S4, a contact mechanics model between the grinding tool and the stainless steel workpiece surface is established. The specific steps are as follows: S41. Simplification of Geometric Contact Model: First, the complex physical contact scenario needs to be abstracted into a simple geometric model: the grinding tool (such as an elastic contact wheel) is regarded as an elastic cylinder, and the stainless steel weld surface is regarded as a rigid plane or curved surface. It is assumed that under the action of normal pressure, the contact area between the two forms a rectangular surface. The size of this surface depends on the magnitude of the pressure and the hardness of the contact wheel. The core variables in the model are defined as: normal grinding force (pressure), contact wheel hardness (elastic modulus), contact wheel radius, and grinding speed. S42. Contact Deformation and Pressure Distribution Analysis: When the robotic arm applies a normal force, the soft contact wheel will undergo compressive deformation. According to Hertz's contact theory (a classic elastic contact theory), the greater the pressure, the larger the contact area, and the pressure is greatest at the center of the contact area and least at the edge. Determining the maximum pressure value on the contact surface is crucial, as this value determines whether the abrasive grains can penetrate the stainless steel surface and is a key parameter for determining grinding efficiency. Establishing the correspondence between normal force and tool indentation (deformation) is equivalent to determining the spring stiffness of the contact wheel, telling the robotic arm: how much force needs to be applied to indent to how deep; S43. Material Removal Rate Model Construction: Establish the relationship between force and grinding effect (removal amount), and establish a basic law - the thickness of metal removed per unit time mainly depends on the product of contact pressure and grinding speed, that is: the greater the pressure and the faster the grinding, the more material is removed. Considering the stickiness and severe work hardening of stainless steel, a "grinding factor" is introduced. This factor is not fixed; it is related to the sharpness of the abrasive belt, the grit size, and the type of stainless steel. S44. Tangential Grinding Resistance Analysis: In addition to the vertically downward normal force, it is also necessary to analyze the resistance in the horizontal direction. During the grinding process, the rotation of the grinding head will generate tangential resistance. According to the principle of tribology, the tangential resistance is usually proportional to the normal force. Condition monitoring correlation: Establish the relationship between tangential force and grinding condition. When the tangential force suddenly increases, the model should be able to determine that it may be due to excessive feed speed or severe abrasive belt wear, thus providing a basis for subsequent force control adjustments. S45. Experimental Calibration of Model Parameters: After the theoretical model is established, the unknown parameters are "calibrated" through experiments, including the following steps: Specimen grinding experiment: A fixed-point grinding experiment was carried out on a standard stainless steel specimen block, with different pressure levels and grinding speeds set. Data measurement: Measure the material removal depth after each grinding process and record the actual force signal data; Parameter fitting: Substitute the experimental data into the previous theoretical model to calculate the "grinding coefficient" and "pressure index" that best match the actual situation. In this way, the model is transformed from a "theoretical formula" into a "practical tool" that can accurately predict the actual processing effect.

[0018] Furthermore, an adaptive grinding path planning and force control method for stainless steel welds is proposed. In step S5, the positional offset of the end effector of the robotic arm in the normal direction is adjusted in real time to achieve flexible compensation for changes in weld reinforcement height and workpiece clamping errors, as detailed below: Closed-loop force error correction strategy: Real-time acquisition of end force sensor data, calculation of the difference between the actual contact force and the target set force, and proportional calculation of the position adjustment based on the magnitude of the force difference. The larger the force difference, the larger the position offset; conversely, when the actual force is greater than the target force, the end is controlled to retreat along the normal direction; when the actual force is less than the target force, the end is controlled to feed along the normal direction. Weld height adaptive strategy: When grinding to the high point of the weld, the contact force increases instantaneously. The system quickly instructs the end of the robot arm to retreat along the normal direction to reduce the amount of material removed and prevent over-cutting. When grinding to the concave part of the weld, the contact force decreases. The system instructs the end of the robot arm to advance along the normal direction to increase the amount of material removed and avoid missing areas. Damping control is introduced during the adjustment process to prevent the robot arm from stiffening or oscillating due to fluctuations in the force signal, and to ensure that the grinding head smoothly crosses the peaks and troughs of the weld. Safety limiting and anti-shaking strategy: Set the maximum allowable range of normal position offset. When the calculated adjustment amount exceeds this range, the system immediately locks the position to prevent the robot arm from colliding with the workpiece and fixture. Set the maximum step size of a single adjustment. When encountering sudden force changes caused by weld beads or hard points, limit the single adjustment amplitude to avoid the robot arm from shaking violently due to overreaction.

[0019] Example 2 This embodiment is applied to the automated grinding operation of the circumferential seam of a certain type of stainless steel pressure vessel. The specific implementation process is as follows: S1. Acquisition and Preprocessing of 3D Data of Weld Surface: A line laser sensor (model: Keyence LJ-V7000) is installed on the side of the flange at the end of the robotic arm. The robotic arm scans the weld area of ​​the stainless steel cylinder at a uniform speed of 50 mm / s. The system acquires the original 3D point cloud data of the weld surface in real time. Statistical filtering algorithms are used to remove isolated noise points caused by the mirror reflection of the stainless steel surface. A voxelized mesh method is used to simplify the point cloud and improve processing efficiency. Finally, the point cloud coordinates are transformed from the sensor coordinate system to the robotic arm base coordinate system using a hand-eye calibration matrix. S2. Weld Feature Extraction and Contour Reconstruction: In the preprocessed point cloud data, the feature extraction algorithm based on the abrupt change in cross-sectional curvature is used to identify the left and right boundary points and peak points of the weld. The weld centerline is fitted by the least squares method, and the average width of the circumferential weld is measured to be 18 mm, and the maximum excess height is 2.5 mm. Based on this, the system reconstructs the three-dimensional spatial contour curve of the weld and generates the weld boundary contour envelope map. S3. Adaptive Grinding Path Planning: Based on the reconstructed contour, a reciprocating grinding trajectory along the tangential direction of the weld is generated, and the grinding tool posture adjustment rules are set: the normal vector of the trajectory point is calculated in real time, and the axis of the sanding belt contact wheel is always perpendicular to the tangential direction of the weld. For the change of weld width, a variable step distance strategy is adopted. In the wider area of ​​the weld (>20mm), the grinding row spacing is set to 2mm, and in the narrower area (<15mm), the spacing is set to 3mm. The robot arm motion program code containing X, Y, Z coordinates and posture angles is generated. S4. Force control strategy model establishment: Based on the hardness characteristics of stainless steel (304 material) and the Preston equation, a contact mechanics model is established, the theoretical normal grinding force is set to 35N, the feed speed is 100mm / s, the impedance control algorithm is designed, the upper limit of the target contact force threshold is set to 45N (to prevent overcutting), the lower limit is set to 25N (to prevent false grinding), and the control mode is switched to force-position hybrid control mode; S5. Force-position coordinated control and real-time compensation: After grinding begins, the six-dimensional force sensor collects the end contact force in real time at a frequency of 100Hz. When the actual contact force is detected to be 30N (less than the target value), the fuzzy PID controller outputs a positive compensation amount, controlling the end of the robotic arm to feed 0.5mm along the normal direction. When the contact force suddenly increases to 42N after passing the high point of the weld seam, the controller quickly outputs a negative compensation amount, controlling the end to retract 1.0mm. This process corrects the normal position offset in real time, effectively compensating for the ellipticity clamping error of the cylinder and the local residual height fluctuation, keeping the contact force stable within the range of 35N±2N. S6. Process Monitoring and Iterative Optimization: The system monitors the spindle motor current and vibration signal in real time. After the first rough grinding, the monitoring data shows that the surface texture in some areas is too deep and the current fluctuation is too large. The system determines that the surface quality does not meet the standard and automatically starts the iterative optimization strategy: reduce the grinding feed speed to 80mm / s and add a fine grinding process. Finally, after two rounds of adaptive grinding, the surface flatness of the weld reaches within 0.1mm and the surface roughness reaches Ra 3.2, which meets the stainless steel process standard, and the grinding operation ends.

Claims

1. A method for adaptive grinding path planning and force control of stainless steel welds, characterized in that, Includes the following steps: S1. Acquisition and preprocessing of 3D data of weld surface: The stainless steel weld area is scanned using a line laser sensor to acquire the original 3D point cloud data of the weld surface. The acquired point cloud data is denoised, filtered and simplified. The coordinate system of the vision sensor and the base coordinate system of the robot arm are calibrated and unified through coordinate transformation. S2. Weld Feature Extraction and Contour Reconstruction: Based on the preprocessed point cloud data, a feature extraction algorithm is used to identify the centerline position, width, and height of the weld. Based on the extracted feature points, the three-dimensional spatial contour curve of the weld is reconstructed to determine the geometric boundary and excess height distribution of the weld. S3. Adaptive Grinding Path Planning: Based on the reconstructed 3D contour of the weld, an initial grinding trajectory is generated. According to the changes in the width and height of the weld, the posture of the grinding tool is adaptively adjusted: the grinding head is always perpendicular to the tangential direction of the weld and the grinding spacing. A variable step size strategy is used to plan the reciprocating grinding path and generate a robot arm motion program containing pose information. S4. Force control strategy model establishment: Establish a contact mechanics model between the grinding tool and the surface of the stainless steel workpiece. Based on the hardness of the stainless steel material and the grinding process requirements, calculate the theoretical normal grinding force, design the impedance control algorithm, switch the position control mode to the force-position hybrid control mode, and set the target contact force threshold and damping parameters. S5. Force-position coordinated control and real-time compensation: During the grinding process, the end contact force data is collected in real time by force / torque sensor. The actual contact force is compared with the target set force. The position correction amount is calculated by using fuzzy PID algorithm. By adjusting the position offset of the end effector of the robot arm in the normal direction in real time, flexible compensation for the change of weld height and workpiece clamping error is achieved, ensuring constant force grinding. S6. Process monitoring and iterative optimization: Real-time monitoring of current, vibration and force feedback signals during the grinding process. When uneven weld removal or failure to meet surface quality standards is detected, the system adaptively adjusts the grinding speed, feed rate and number of grinding passes, and performs iterative grinding until the weld surface flatness and smoothness meet the stainless steel process standards.

2. The method for adaptive grinding path planning and force control of stainless steel welds as described in claim 1, characterized in that, In step S2, a feature extraction algorithm is used to identify the centerline position, width, and height of the weld. Based on the extracted feature points, the specific steps are as follows; S21. Point Cloud Slicing and Section Acquisition: Since welds are usually long and narrow, the section method can be used for processing. That is, the point cloud data is sliced ​​vertically along the extension direction of the weld at a certain step size to obtain a series of two-dimensional cross-sectional point sets; S22. Feature point recognition algorithm: In each two-dimensional cross-sectional point set, the boundary points of the weld are identified based on the sudden change in the rate of curvature: weld toe and peak point. The curvature value of each point is calculated, and a threshold is set to filter out the point with the largest curvature as a candidate feature point. Combined with the reflective properties of stainless steel, noise caused by reflection needs to be removed. The cross-sectional contour curve is fitted by the least squares method to smooth abnormal protrusions. S23. Calculation of weld centerline and width: Connect the crests of each section to form the weld centerline, calculate the distance between the left and right boundary points to obtain the weld width, and calculate the distance from the crest to the base material surface to obtain the weld reinforcement. S24. Contour Mathematical Modeling: Based on the extracted centerline coordinates, the three-dimensional spatial curve equation of the smooth weld is fitted using the cubic spline interpolation method, providing a continuous navigation reference for subsequent path planning.

3. The method for adaptive grinding path planning and force control of stainless steel welds as described in claim 2, characterized in that, In specific step S22, the boundary points, weld toes, and wave crests of the weld are identified based on the abrupt change in the rate of curvature change, as follows; Calculate the curvature value at each point in the weld section point cloud. Set the weld boundary identification conditions as follows The rate of change of curvature threshold, and the angle between the normal vector at that point and the normal vector of the parent material plane. Only points that simultaneously meet the above conditions are identified as weld boundary points, thus eliminating interference from surface scratches.

4. The method for adaptive grinding path planning and force control of stainless steel welds as described in claim 1, characterized in that, In step S3, the posture of the grinding tool is adaptively adjusted according to the changes in the width and height of the weld. The specific steps are as follows; S31. Adaptive adjustment of tool posture: The grinding quality largely depends on the contact angle between the grinding head and the workpiece surface. Based on the reconstructed weld surface model, the normal vector and tangent vector at the grinding point are calculated, and the posture of the end effector is adjusted so that the contact wheel axis of the sanding belt and grinding wheel is perpendicular to the tangent vector, while ensuring that the grinding pressure direction is parallel to the normal vector and forms a specific process angle. S32. Variable step pitch trajectory generation strategy: For stainless steel welds with varying widths, a strategy of equal residual height and variable step pitch is adopted to plan the reciprocating grinding path. In the wider weld area, the grinding row spacing is automatically reduced to increase the grinding overlap rate and avoid missed grinding. In the narrower weld area and base material area, the step pitch is appropriately increased to improve efficiency and generate a series of discrete path points containing position and attitude information. S33. Speed ​​planning: Dynamically allocate the feed speed v according to the weld height H. In areas with high weld height, reduce the robot arm feed speed to ensure controllable single cutting depth; in areas with low weld height, appropriately increase the speed to prevent over-grinding.

5. The method for adaptive grinding path planning and force control of stainless steel welds as described in claim 4, characterized in that, In specific step S32, a strategy of equal residual height and variable step distance is used to plan the reciprocating grinding path, as follows; Variable step distance logic: Introduce chord height error constraints to perform adaptive interpolation of path points, and calculate the distance from the midpoint of the line connecting adjacent path points to the actual weld curve: chord height h; Set the maximum allowable chord height error h max ; When h > h max Then, a new path point is inserted between the two points until all path segments meet the error requirements. The path points are dense in areas with large weld curvature and sparse in flat areas, balancing accuracy and efficiency.

6. The method for adaptive grinding path planning and force control of stainless steel welds as described in claim 1, characterized in that, In step S4, a contact mechanics model between the grinding tool and the stainless steel workpiece surface is established. The specific steps are as follows: S41. Simplification of Geometric Contact Model: First, the complex physical contact scenario needs to be abstracted into a simple geometric model: the grinding tool is regarded as an elastic cylinder, and the stainless steel weld surface is regarded as a rigid plane. Under the action of normal pressure, the contact area between the two forms a rectangular surface. The size of this surface depends on the magnitude of the pressure and the hardness of the contact wheel. The core variables in the model are defined as: normal grinding force, contact wheel hardness, contact wheel radius, and grinding speed. S42. Contact Deformation and Pressure Distribution Analysis: When the robotic arm applies a normal force, the soft contact wheel will undergo compressive deformation. According to Hertz's contact theory, the greater the pressure, the larger the contact area, and the pressure at the center of the contact area is the greatest, while the pressure at the edge is the least. Determine the maximum pressure value on the contact surface and establish the correspondence between the normal force and the tool indentation and deformation. S43. Material Removal Rate Model Construction: Establish the relationship between force and grinding effect: removal amount, and establish a basic rule - the thickness of metal removed per unit time mainly depends on the product of contact pressure and grinding speed; S44. Tangential Grinding Resistance Analysis: In addition to the vertically downward normal force, the horizontal resistance is analyzed. During the grinding process, the rotation of the grinding head generates tangential resistance. According to the tribological principle, the tangential resistance is proportional to the normal force. The relationship between the tangential force and the grinding state is established. When the tangential force suddenly increases, the model should be able to determine whether the feed rate is too fast or the abrasive belt is severely worn, thus providing a basis for subsequent force control adjustments. S45. Experimental Calibration of Model Parameters: After the theoretical model is established, the unknown parameters are calibrated experimentally, including the following steps: Specimen grinding experiment: A fixed-point grinding experiment was carried out on a standard stainless steel specimen block, with different pressure levels and grinding speeds set. Data measurement: Measure the material removal depth after each grinding process and record the actual force signal data; Parameter fitting: Substitute the experimental data into the previous theoretical model to calculate the grinding coefficient and pressure index that best match the actual situation.

7. The method for adaptive grinding path planning and force control of stainless steel welds as described in claim 1, characterized in that, In step S5, the positional offset of the end effector of the robotic arm in the normal direction is adjusted in real time to achieve flexible compensation for changes in weld reinforcement height and workpiece clamping errors, as detailed below: Closed-loop force error correction strategy: Real-time acquisition of end force sensor data, calculation of the difference between the actual contact force and the target set force, and proportional calculation of the position adjustment amount based on the magnitude of the force difference. When the actual force is greater than the target force, the end is controlled to retreat along the normal direction. When the actual force is less than the target force, control the end feed along the normal direction; Weld height adaptive strategy: When grinding to the high point of the weld, the contact force increases instantaneously. The system quickly instructs the end of the robot arm to retreat along the normal direction to reduce the amount of material removed and prevent over-cutting. When grinding to the concave part of the weld, the contact force decreases. The system instructs the end of the robot arm to advance along the normal direction to increase the amount of material removed and avoid missing areas. Damping control is introduced during the adjustment process to prevent the robot arm from stiffening due to fluctuations in the force signal and to ensure that the grinding head smoothly crosses the peaks and troughs of the weld. Safety limiting and anti-shaking strategy: Set the maximum allowable range of normal position offset. When the calculated adjustment amount exceeds this range, the system immediately locks the position to prevent the robot arm from colliding with the workpiece and fixture. Set the maximum step size of a single adjustment. When encountering sudden force changes caused by weld beads or hard points, limit the single adjustment amplitude to avoid the robot arm from shaking violently due to overreaction.

Images