A robot polishing path adaptive control method and system

Abstract

The present application relates to the technical field of industrial robot control, and discloses a robot polishing path adaptive control method and system, the method comprises the following steps: obtaining force value data and deformation data of multiple contact points between the robot end and the workpiece surface, and performing fitting operation to obtain a stiffness estimation value; dividing the workpiece surface according to the stiffness estimation value to obtain a regional difference boundary; processing the force distribution data in the regional difference boundary to obtain a stiffness change trend; adjusting the current contact force according to the stiffness change trend to obtain an optimized contact force; correcting the initial polishing path according to the optimized contact force and calculating a force value distribution result; simulating and verifying the force value distribution result and judging the stability to obtain an effective distribution sequence; and controlling the robot to perform polishing according to the effective distribution sequence. The present application realizes real-time adaptive adjustment of the contact force parameter and the polishing path.

Description

Technical Field

[0001] This invention relates to the field of industrial robot control technology, and in particular to an adaptive control method and system for robot grinding paths. Background Technology

[0002] Currently, robotic grinding technology is widely used in the processing of large curved surface workpieces as an important means to improve the processing quality and efficiency of complex workpieces. During robotic grinding, the distribution of contact force between the robot's end effector and the workpiece, as well as the planning of the grinding path, determine the surface accuracy of the workpiece and affect subsequent assembly and performance.

[0003] In existing technologies, robotic grinding control methods typically employ a pre-set fixed grinding path and fixed contact force parameters to grind the workpiece. During grinding, the robot processes the workpiece surface according to the preset path and parameters, without sensing the surface stiffness distribution or adjusting the contact force parameters and grinding path based on changes in force distribution during processing. While some grinding control systems have developed adaptive capabilities with the advancement of machine intelligence technology, they still fail to effectively solve the problem of processing consistency caused by uneven surface stiffness distribution. The stiffness characteristics of curved workpieces often exhibit significant regional differences; convex areas typically have higher local stiffness, while concave areas typically have lower local stiffness. These regional stiffness differences prevent the robot from applying contact forces that match the stiffness characteristics of different locations during grinding, resulting in a deviation between the planned grinding path and the actual mechanical response during processing.

[0004] In summary, existing technologies have the problem of not being able to adaptively adjust the contact force parameters and grinding path in real time based on the local stiffness distribution of the workpiece surface. Summary of the Invention

[0005] This invention provides a robot grinding path adaptive control method and system to solve the problem in the prior art that the contact force parameters and grinding path cannot be adaptively adjusted in real time according to the local stiffness distribution of the workpiece surface, which makes it difficult to guarantee the consistency of processing.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides a robot grinding path adaptive control method, comprising: Force and deformation data at multiple contact points between the robot end effector and the workpiece surface are acquired, and the force and deformation data are fitted to obtain a stiffness estimate. The workpiece surface is divided into regions based on the stiffness estimate to obtain stiffness sub-regions. Stiffness difference boundaries between adjacent stiffness sub-regions are identified to obtain region difference boundaries. Force distribution data within the boundary of the region difference is obtained, the force distribution data is filtered to obtain smooth force distribution data, regression fitting is performed on the smooth force distribution data to obtain a fitting curve, and derivative calculation is performed on the fitting curve to obtain the stiffness change trend. The current contact force of the robot control system is obtained, and the current contact force is adjusted according to the stiffness change trend to obtain an optimized contact force; The initial grinding path of the workpiece is obtained, and the initial grinding path is corrected according to the optimized contact force to obtain the corrected path. The corrected path is discretized to obtain multiple control points. The force value of each control point is calculated according to the optimized contact force to obtain the force value distribution result. The force value allocation result is simulated to obtain the simulation result. Invalid allocations are removed from the force value allocation result based on the simulation result. The stability of the removed force value allocation result is judged to obtain the valid allocation sequence. The robot is controlled to perform polishing according to the effective allocation sequence, and the polishing process is monitored in real time to complete the polishing operation.

[0007] Secondly, the present invention provides a robot grinding path adaptive control system, comprising: The stiffness estimation module is used to acquire force and deformation data at multiple contact points between the robot end effector and the workpiece surface, and to perform fitting calculations on the force and deformation data to obtain stiffness estimates. The region division module is used to divide the workpiece surface into regions based on the stiffness estimate to obtain stiffness sub-regions, and to identify stiffness difference boundaries between adjacent stiffness sub-regions to obtain region difference boundaries. The trend perception module is used to acquire force distribution data within the boundary of the regional difference, filter the force distribution data to obtain smooth force distribution data, perform regression fitting on the smooth force distribution data to obtain a fitting curve, and perform derivative calculation on the fitting curve to obtain the stiffness change trend. The contact force optimization module is used to obtain the current contact force of the robot control system, and adjust the current contact force according to the stiffness change trend to obtain the optimized contact force; The path correction module is used to obtain the initial grinding path of the workpiece, correct the initial grinding path according to the optimized contact force to obtain the corrected path, discretize the corrected path to obtain multiple control points, and calculate the force value of each control point according to the optimized contact force to obtain the force value allocation result. The simulation filtering module is used to simulate the force value allocation result, obtain the simulation result, remove invalid allocations from the force value allocation result based on the simulation result, and perform stability judgment on the removed force value allocation result to obtain the valid allocation sequence. The execution monitoring module is used to control the robot to perform grinding according to the effective allocation sequence, monitor the grinding process in real time, and complete the grinding operation.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention obtains stiffness estimates by fitting force and deformation data of multiple contact points on the workpiece surface, and divides the workpiece surface into regions accordingly, identifies the boundaries of regional differences, and realizes quantitative perception, precise partitioning and automatic boundary recognition of local stiffness distribution on the workpiece surface. This solves the problem that the existing technology cannot perceive the differences in stiffness distribution on the workpiece surface and cannot distinguish between high and low stiffness regions, and significantly improves the robot's adaptive perception ability of stiffness characteristics of complex curved workpieces.

[0009] (2) This invention obtains the accurate stiffness change trend by performing Kalman filtering noise reduction, polynomial regression fitting and derivative calculation on the force distribution data within the boundary of regional differences, and adjusts the current contact force in real time according to the stiffness change trend to obtain the optimized contact force. This realizes the dynamic matching and real-time adaptive adjustment of the contact force parameters with the change of workpiece surface stiffness, solves the problems of fixed contact force parameters, inability to adapt to the regional differences of workpiece stiffness, and easy over-grinding or under-grinding in the prior art, and greatly improves the force control stability of the grinding process.

[0010] (3) This invention modifies the initial grinding path based on the optimized contact force, obtains an effective allocation sequence by combining finite element simulation verification, invalid force value elimination and force change rate stability judgment, and collects, analyzes and corrects the force distribution data in real time during the grinding operation. This realizes dynamic matching and closed-loop control between grinding path planning and actual mechanical response, solves the problems of disconnect between path planning and force distribution change in the processing process, difficulty in ensuring processing consistency and uneven surface accuracy in the prior art, and significantly improves the grinding quality and processing consistency of the workpiece.

[0011] (4) By simulating and optimizing the force distribution results, this invention effectively avoids problems such as excessive contact stress, sudden force changes, and path jitter during the grinding process, protects the workpiece surface from damage, improves the stability of robot operation, extends the service life of end effector, and reduces the scrap rate. Attached Figure Description

[0012] Figure 1This is a schematic diagram of the robot grinding path adaptive control method provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the robot grinding path adaptive control system provided in the second embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Reference Figure 1 The first embodiment of the present invention provides an adaptive control method for robot grinding path, comprising the following steps: S11, acquire force value data and deformation data of multiple contact points between the robot end effector and the workpiece surface, and perform fitting calculation on the force value data and the deformation data to obtain a stiffness estimate; S12, the workpiece surface is divided into regions according to the stiffness estimate to obtain stiffness sub-regions, and the stiffness difference boundary between adjacent stiffness sub-regions is identified to obtain the region difference boundary. S13, acquire force distribution data within the boundary of the regional difference, filter the force distribution data to obtain smooth force distribution data, perform regression fitting on the smooth force distribution data to obtain a fitting curve, and perform derivative calculation on the fitting curve to obtain the stiffness change trend. S14, Obtain the current contact force of the robot control system, and adjust the current contact force according to the stiffness change trend to obtain an optimized contact force; S15, obtain the initial grinding path of the workpiece, correct the initial grinding path according to the optimized contact force to obtain the corrected path, discretize the corrected path to obtain multiple control points, calculate the force value of each control point according to the optimized contact force to obtain the force value distribution result; S16, Simulate the force value allocation result to obtain simulation result, remove invalid allocations from the force value allocation result based on the simulation result, and determine the stability of the removed force value allocation result to obtain an effective allocation sequence; S17, the robot is controlled to perform polishing according to the effective allocation sequence, the polishing process is monitored in real time, and the polishing operation is completed.

[0015] In step S11, force and deformation data at multiple contact points between the robot end effector and the workpiece surface are acquired. Fitting calculations are then performed on the force and deformation data to obtain a stiffness estimate, including: Force and displacement sensors installed at the end of the robot are used to collect force and deformation data at each contact point. A linear regression fit is performed on the force value data and the deformation data of each contact point to obtain the stiffness slope value corresponding to each contact point. Obtain the historical average stiffness value, and perform a weighted calculation on the stiffness slope value corresponding to each contact point and the historical average stiffness value to obtain the stiffness estimate value corresponding to each contact point.

[0016] In one implementation, force data refers to the contact force measurement value collected by a force sensor when the robot end effector contacts the workpiece, and deformation data refers to the local deformation measurement value collected by a displacement sensor under the action of the contact force. The force sensor and displacement sensor are mounted on the robot end effector and sense each contact point on the workpiece surface in real time as the robot end effector moves. The two types of sensors synchronously collect data at a fixed sampling frequency, forming a time-indexed force value sequence and deformation value sequence. The sampling frequency is determined based on the workpiece processing speed and sensor response characteristics. Those skilled in the art will understand that the sampling frequency can be selected in the range of 50 to 200 times per second; for example, in this embodiment, the sampling frequency is set to 100 times per second.

[0017] It should be noted that the stiffness slope value is a regression coefficient obtained by linearly regressing the force and deformation data at each contact point. Specifically, a linear regression equation is established with deformation data as the independent variable and force data as the dependent variable. The regression coefficient and intercept that minimize the sum of squared residuals are obtained using the least squares method. The slope obtained is used as the stiffness slope value corresponding to that contact point, which characterizes the local stiffness characteristics of the workpiece at that contact point. The stiffness slope value is then weighted and calculated with the historical stiffness mean to obtain a stiffness estimate.

[0018] It should be noted that the historical stiffness mean is a reference value for workpiece stiffness obtained from statistical analysis of historical grinding operation data. It is stored in the system database and retrieved by the system before each grinding operation begins. Specifically, the arithmetic mean of all stiffness slope values ​​collected from the same type of workpiece under similar processing conditions in historical grinding operations is used to introduce historical prior information in the weighted calculation. The stiffness estimate is obtained by weighting the stiffness slope value and the historical stiffness mean according to preset weights. Specifically, the stiffness slope value is multiplied by its corresponding weight, and the historical stiffness mean is multiplied by its corresponding weight. The two products are then added to obtain the stiffness estimate, where the sum of the weights of the stiffness slope value and the historical stiffness mean is 1. The weights are determined through a grid search on the historical dataset. Stiffness estimates are calculated under multiple candidate weight combinations and used for subsequent region division. Finally, the weight combination that best matches the manually labeled stiffness distribution in the region division is selected as the preset weight. The search space is defined as follows: the weights of stiffness slope values ​​range from 0.5 to 0.9, with a step size of 0.05; the weight of the historical stiffness mean is 1 minus the former, with the same step size, generating a total of 9 candidate weight combinations. The evaluation dataset comes from a pre-collected labeled dataset containing 50 samples. This dataset covers workpiece surfaces in high, medium, and low stiffness regions. Each sample includes the calculated stiffness slope value, the historical stiffness mean, and the true stiffness distribution value jointly labeled by 3 engineers with more than five years of work experience (label consistency requires an intra-group correlation coefficient greater than 0.85). During the grid search, the stiffness estimate of each sample is calculated using the 9 candidate weight combinations, and the region is divided accordingly. The division results are compared with the manually labeled true stiffness distribution values, and the optimal weight combination is selected based on the highest division accuracy. After search verification, the highest region division accuracy (92%) was achieved on the evaluation dataset when the weight of the stiffness slope value is 0.7 and the weight of the historical stiffness mean is 0.3. Therefore, this was used as the preset weight. Those skilled in the art will understand that the weight of the stiffness slope value is usually between 0.6 and 0.8, and the weight of the historical stiffness mean is correspondingly between 0.2 and 0.4. For scenarios with high real-time requirements, the weight of the stiffness slope value can be appropriately increased, and for scenarios with high stability requirements, the weight of the historical stiffness mean can be appropriately increased.

[0019] In step S12, the workpiece surface is divided into regions based on the stiffness estimate to obtain stiffness sub-regions. Stiffness difference boundaries between adjacent stiffness sub-regions are identified to obtain region difference boundaries, including: The workpiece surface is uniformly divided at a preset interval to obtain multiple grid units. The force and deformation data at each grid unit are fitted to obtain the stiffness estimate corresponding to each grid unit. The stiffness estimates of all the mesh elements are iteratively clustered using the k-means clustering algorithm to obtain multiple stiffness sub-regions; The stiffness estimates of adjacent grid cells are interpolated to obtain adjacent stiffness differences. The adjacent stiffness differences are compared with a preset difference threshold. Grid cells that exceed the preset difference threshold are marked as boundary points. All boundary points are connected to obtain regional difference boundaries.

[0020] In one implementation, a grid cell is the smallest spatial unit obtained by uniformly dividing the workpiece surface at a preset interval, with each grid cell corresponding to a local region of the workpiece surface. The preset interval is determined based on the workpiece surface size and the required stiffness distribution resolution. The smaller the interval, the more grid cells there are, and the finer the stiffness distribution is depicted, but the corresponding computational overhead is also greater. Those skilled in the art will understand that the preset interval can be selected within the range of 5 mm to 50 mm based on the workpiece size. For example, the workpiece surface is divided into 100 grid cells. The linear regression fitting and weighted operation described in step S11 are performed on the force value data and deformation data at each grid cell to obtain the stiffness estimate corresponding to that grid cell, thereby discretizing the stiffness distribution of the workpiece surface into a set of stiffness estimates with grid cells as the basic unit, which is used for subsequent clustering.

[0021] It should be noted that the k-means clustering algorithm is an iterative algorithm that divides a dataset into k categories based on similarity. The specific execution process of the k-means clustering algorithm includes: first, an initialization phase, randomly selecting k values ​​from the stiffness estimates of all grid cells as initial cluster centers; second, an allocation phase, calculating the Euclidean distance between the stiffness estimate of each grid cell and each cluster center, and assigning the grid cell to the category corresponding to the nearest cluster center; third, an update phase, taking the arithmetic mean of the stiffness estimates of all grid cells in each category, and using this as the new cluster center for that category; fourth, an iteration termination judgment, repeating steps two and three until the change in cluster centers is less than a preset convergence threshold or a preset maximum number of iterations is reached, resulting in the final k stiffness sub-regions. For example, in this embodiment, the preset convergence threshold is set to 0.01 (Newtons / meter), and the preset maximum number of iterations is set to 100. The value of k is determined based on the actual complexity of the workpiece surface stiffness distribution. Those skilled in the art will understand that the value of k can be selected in the range of 2 to 5. A smaller k value can be selected for workpieces with relatively simple stiffness distribution, while a larger k value can be selected for workpieces with relatively complex stiffness distribution. For example, in this embodiment, the k value is set to 3, dividing the workpiece surface into three stiffness sub-regions: high stiffness, medium stiffness, and low stiffness.

[0022] It should be noted that the adjacent stiffness difference is obtained by interpolating the stiffness estimates of adjacent mesh elements, and is used to measure the magnitude of stiffness variation between adjacent mesh elements. The adjacent stiffness difference is compared with a preset difference threshold. Mesh elements exceeding the preset difference threshold are marked as boundary points. All boundary points are connected to form continuous regional difference boundaries, which are used to determine the collection range of force distribution data in step S13. The preset difference threshold is determined by statistical analysis of the stiffness differences of adjacent mesh elements in historical grinding operation data. Specifically, the mean and standard deviation of all adjacent stiffness differences in historical data are calculated, and the preset difference threshold is set as the mean plus a certain multiple of the standard deviation. In this embodiment, the multiple is 1.5. This value is based on the fact that a multiple that is too low (e.g., 1x) will misjudge stiffness fluctuations caused by normal processing errors as boundaries, resulting in oversegmentation; a multiple that is too high (e.g., 3x) will ignore the actual stiffness variation boundaries, resulting in undersegmentation. 1.5x strikes a balance between preserving the true boundaries and suppressing noise, and is suitable for most metal workpiece surfaces. Those skilled in the art will understand that the preset difference threshold can be adjusted within the range of 50 N / m to 200 N / m depending on the type of workpiece. For example, the preset difference threshold is set to 100 N / m, and when the stiffness difference between adjacent mesh cells exceeds 100 N / m, the corresponding mesh cell is marked as a boundary point.

[0023] In step S13, force distribution data within the boundary of the regional difference is acquired, the force distribution data is filtered to obtain smooth force distribution data, regression fitting is performed on the smooth force distribution data to obtain a fitted curve, and derivative calculation is performed on the fitted curve to obtain the stiffness change trend, including: The force distribution data is processed using a Kalman filter, and the state estimation is iteratively updated based on the preset process noise covariance and the preset measurement noise covariance to obtain smooth force distribution data. The smooth force distribution data is fitted with a second-order polynomial regression to obtain a fitted curve. The derivative of the fitted curve is then calculated to obtain the stiffness variation trend.

[0024] In one implementation, the force distribution data refers to the set of contact force measurements collected by the force sensor at each grid cell within the regional difference boundary determined in step S12, reflecting the spatial distribution of forces on the workpiece surface within that boundary region. Since the sensor is affected by factors such as mechanical vibration and electrical interference during actual data acquisition, the original force distribution data contains a certain degree of noise. Before performing regression fitting, the force distribution data needs to be filtered to obtain smooth force distribution data.

[0025] It should be noted that the Kalman filter is used to filter force distribution data, and its state-space model is defined as a state vector. Defined as a two-dimensional vector ,in This is the estimated contact force at the k-th sampling point. Let be the rate of change of the contact force along the path direction s, with dimensions 2×1. The state transition matrix F is set as follows: Δs is the path step size (in meters) between adjacent sampling points. This matrix describes the recursive relationship of the state variables from the k-th sampling point to the (k+1)-th sampling point. The observation matrix H is set to [1,0], indicating that the observed values ​​directly reflect the contact force. The rate of change is not directly observed. Initial state estimate. Set as ,in The measured force value at the first sampling point, with the rate of change initially set to 0. Initial error covariance. Set as ,in Take the initial variance of the measured force value (for example, set it to 0.01). The initial uncertainty of the rate of change is set (exemplarily, 0.01). The preset process noise covariance Q is set to [[0.05,0],[0,0.01]], representing the uncertainty in the state transition process; the preset measurement noise covariance R is set to [0.1] (scalar), representing the variance of sensor measurement noise. The Kalman filter performs optimal state estimation on noisy observation data through two alternating stages of prediction and update. The above process is executed sequentially for each data point in the force distribution data to gradually filter out random noise and obtain smooth force distribution data.

[0026] It should be noted that the preset process noise covariance is used to describe the uncertainty in the system state transition process, and the preset measurement noise covariance is used to describe the degree of deviation between the sensor measurement value and the true value. The preset process noise covariance and the preset measurement noise covariance are determined through statistical analysis of data collected by the sensor under static conditions. Specifically, the sensor is placed in a static environment without external force, and a certain number of measurement values ​​are continuously collected. The variance of the collected data is used as the basis for determining the preset measurement noise covariance. The preset process noise covariance is pre-calibrated based on the force change characteristics during the robot's end effector movement. Those skilled in the art will understand that the preset process noise covariance is typically between 0.01 and 0.1, and the preset measurement noise covariance is typically between 0.05 and 0.5. For example, the process noise covariance is set to 0.05, and the measurement noise covariance is set to 0.1.

[0027] It should be noted that second-order polynomial regression fitting refers to establishing a second-order polynomial regression equation with spatial location coordinates as the independent variable and smoothed force distribution data as the dependent variable. The regression coefficients that minimize the sum of squared residuals are then solved using the least squares method to obtain a fitted curve describing the relationship between force distribution and location. The second-order polynomial regression equation consists of a quadratic term, a linear term, and a constant term. The coefficients before the quadratic term are the quadratic coefficients, the coefficients before the linear term are the linear coefficients, and the constant term is the intercept. These three coefficients are determined by the fitting operation. The reason for choosing a second-order polynomial over a higher-order polynomial is that it can characterize the nonlinear variation trend of the force distribution data while avoiding the overfitting problem caused by higher-order polynomials. For example, following the high-stiffness sub-region defined in step S12, the smooth force distribution data collected within the range of 0 meters to 1 meter along the grinding path in the sub-region is subjected to a second-order polynomial regression fitting, resulting in a quadratic term coefficient of -30, a linear term coefficient of 80, and an intercept of 100. The corresponding force distribution changes from 100 Newtons to 150 Newtons within this range, reflecting the nonlinear trend of force distribution with position in this region.

[0028] It should be noted that the stiffness variation trend is obtained by performing derivative calculations on the fitted curve. Specifically, the first derivative of the fitted curve with respect to the position coordinates is calculated, and the resulting derivative value is used as the rate of change of force distribution data with position, i.e., the stiffness variation trend. The derivative values ​​at all positions within the boundary of regional differences are calculated to obtain the complete stiffness variation trend, which is used to adjust the current contact force in step S14.

[0029] In step S14, the current contact force of the robot control system is obtained, and the current contact force is adjusted according to the stiffness change trend to obtain an optimized contact force, including: The stiffness change trend is compared with a preset trend threshold. If the stiffness change trend exceeds the preset trend threshold, the contact force deviation between the current contact force and the preset target contact force is calculated. The current contact force is iteratively adjusted based on the contact force deviation to obtain the adjusted contact force. The difference between the adjusted contact force and the current contact force is calculated along the X-axis, Y-axis and Z-axis to obtain the force compensation vector; The force compensation vector is superimposed on the current contact force to obtain the optimized contact force.

[0030] In one implementation, the current contact force refers to the set contact force applied to the workpiece by the robot end effector by the robot control system at the current moment, which is read in real time by the robot control system. The preset trend threshold is a stiffness change trend judgment benchmark preset according to the workpiece material characteristics and processing requirements, used to determine whether the stiffness change amplitude in the current area needs to trigger contact force adjustment. Specifically, the derivative values ​​of 100 sets of workpiece surface stiffness change trends in historical grinding operations are collected as statistical samples, the 95th percentile of the sample is calculated, and the preset trend threshold is set to this percentile. Statistically, the derivative values ​​of the 100 sets of samples range from 0.3 to 2.8, with a mean of 1.2, a standard deviation of 0.6, and a 95th percentile of 1.5. Those skilled in the art will understand that the preset trend threshold can be selected in the range of 0.5 to 3.0. For example, based on the above statistical analysis results, the preset trend threshold is set to 1.5. When the derivative value of the stiffness change trend exceeds 1.5, the contact force adjustment mechanism is triggered.

[0031] It should be noted that the contact force deviation is the difference between the current contact force and the preset target contact force, reflecting the degree to which the current contact force deviates from the target value. The preset target contact force is a desired contact force value pre-set based on the workpiece material characteristics and machining accuracy requirements, and is stored in the robot control system. For example, the current contact force is 150.2 Newtons, the preset target contact force is 160.0 Newtons, and the contact force deviation is 9.8 Newtons. The contact force deviation is used for subsequent iterative adjustments to the current contact force to obtain the adjusted contact force.

[0032] It should be noted that the adjusted contact force is obtained by iteratively adjusting the current contact force based on the contact force deviation. Specifically, a proportional-integral-derivative (PID) control algorithm is used to perform multiple iterative calculations on the current contact force. In each iteration, a proportional term, an integral term, and a derivative term are calculated based on the current contact force deviation. The proportional term is the product of the current contact force deviation and the proportional coefficient. The integral term is the product of the cumulative sum of the contact force deviation during the iteration process and the integral coefficient. The derivative term is the product of the change in contact force deviation between two adjacent iterations and the derivative coefficient. The proportional term, the integral term, and the derivative term are added together to obtain the control output for this iteration. This control output is then superimposed on the current contact force to obtain the contact force after this iteration. The contact force deviation is recalculated, and the above process is repeated until the contact force deviation converges to a preset convergence threshold range, at which point the iteration ends. The proportional coefficient, integral coefficient, and derivative coefficient are determined by parameter tuning based on historical grinding operation data. This embodiment uses the critical proportionality method for tuning, and the specific steps are as follows: First, set the integral coefficient Ki and the derivative coefficient Kd to 0, retaining only the proportional control; then, gradually increase the proportional coefficient Kp and observe the system response. When constant amplitude oscillation occurs, record the critical proportional coefficient Ku = 1.2 and the critical oscillation period Tu = 0.8 seconds; according to the Ziegler-Nichols rule, calculate Kp = 0.6 × Ku = 0.72, Ki = 1.2 × Ku / Tu = 0.108, and Kd = 0.075 × Ku × Tu = 0.072; finally, combining the workpiece material characteristics (aluminum alloy, elastic modulus 70 GPa) and the grinding process requirements (target contact force 160 Newtons), fine-tune the above coefficients to obtain the final tuning result. For example, the proportional coefficient is set to 0.6, the integral coefficient is set to 0.1, and the differential coefficient is set to 0.05. After iterative adjustment, the adjusted contact force is 158.7 Newtons, and the contact force deviation converges from the initial 9.8 Newtons to 1.3 Newtons.

[0033] It should be noted that the force compensation vector is a three-dimensional vector obtained by calculating the difference between the adjusted contact force and the current contact force along the X, Y, and Z axes. Specifically, the difference between the adjusted contact force and the current contact force is calculated, and based on the direction cosine matrix corresponding to the current pose of the robot's end effector, the difference is decomposed into the X, Y, and Z axes to obtain the force compensation components in each direction, which are then combined to form the force compensation vector. For example, when the adjusted contact force is 158.7 Newtons and the current contact force is 150.2 Newtons, the difference is 8.5 Newtons. Projecting this value according to the direction cosine of the current pose of the robot's end effector yields a force compensation vector with a component of 5.1 Newtons in the X-axis direction, 4.8 Newtons in the Y-axis direction, and 4.7 Newtons in the Z-axis direction. This force compensation vector is then superimposed onto the current contact force along the corresponding coordinate axes to obtain the optimized contact force.

[0034] In step S15, the initial grinding path of the workpiece is obtained, and the initial grinding path is corrected according to the optimized contact force to obtain a corrected path. The corrected path is discretized to obtain multiple control points. The force value of each control point is calculated according to the optimized contact force to obtain a force value allocation result, including: A three-dimensional surface model of the workpiece is performed to obtain the curvature values ​​at various locations on the workpiece surface; The path correction coefficient is obtained by calculating the curvature value and the optimized contact force. The initial polishing path is adjusted according to the path correction coefficient to obtain the corrected path; The corrected path is discretized and sampled to obtain multiple control points; The curvature value is obtained based on the position coordinates of each control point. The force value of each control point is interpolated based on the curvature value and the optimized contact force to obtain the force value allocation result.

[0035] In one implementation, the initial grinding path is a robot motion trajectory pre-planned based on the workpiece geometry before the grinding operation begins, and stored in the robot control system. Since the initial grinding path does not consider the differences in workpiece surface stiffness distribution during the planning stage, its path parameters deviate from the mechanical response during actual processing, requiring correction based on the optimized contact force obtained in step S14. A three-dimensional surface model is performed on the workpiece to obtain curvature values ​​at various locations on the workpiece surface. These curvature values ​​describe the degree of bending of the workpiece surface at each location, reflecting the geometric characteristics of the workpiece surface, and are used for calculating subsequent path correction coefficients.

[0036] It should be noted that the path correction coefficient is calculated based on the curvature values ​​and optimized contact forces at various locations on the workpiece surface, and is used to quantitatively correct the trajectory parameters of the initial grinding path. Specifically, the path correction coefficient is calculated as follows: first, the curvature values ​​and optimized contact forces at various locations on the workpiece surface are normalized. In this embodiment, the min-max normalization method is used to linearly map the curvature values ​​and optimized contact forces to the [0,1] interval. The normalized curvature values ​​and normalized optimized contact forces are then weighted and summed according to preset weights. The summation result is added to the baseline coefficient 1 to obtain the path correction coefficient, which is a dimensionless value greater than 1. The baseline coefficient 1 represents the path state without correction, and the weighted summation result reflects the combined influence of curvature and contact force on path correction. In this embodiment, the weight corresponding to the curvature value is set to 0.5, the weight corresponding to the optimized contact force is set to 0.5, and the sum of the weights corresponding to the curvature value and the optimized contact force is 1. The physical meaning of the aforementioned weight values ​​is that curvature and contact force contribute equally to path correction; that is, the degree of geometric curvature of the workpiece surface and the magnitude of the contact force applied by the robot are equally important in affecting the offset of the grinding path. This value was verified through linear regression analysis of historical grinding operation data. Using the mean surface roughness after grinding as the dependent variable and the normalized curvature value and normalized optimized contact force as independent variables, the curvature value weight was fitted on 50 sets of sample data, resulting in a range of 0.45 to 0.55. The median value of 0.5 was ultimately chosen as the preset weight. Those skilled in the art will understand that the weight corresponding to the curvature value is typically between 0.4 and 0.6, and the corresponding weight for the optimized contact force is also between 0.4 and 0.6. The trajectory parameters of the initial grinding path are adjusted according to the path correction coefficient. Specifically, the deviation of the initial grinding path at that position is multiplied by the path correction coefficient to obtain the corrected deviation, and thus the corrected path.

[0037] It should be noted that discretization sampling refers to dividing the correction path into a finite number of discrete location points according to a preset sampling interval, with each location point being a control point. Each control point contains its spatial coordinate information, used for subsequent force value allocation calculations. The preset sampling interval is determined based on the total length of the correction path and the required control accuracy. Those skilled in the art will understand that for correction paths with lengths on the order of meters, the preset sampling interval can be selected within the range of 1 centimeter to 10 centimeters. For example, a correction path with a total length of approximately 10 meters is discretized and sampled at 10-centimeter intervals, resulting in 100 control points, each corresponding to a spatial location on the correction path.

[0038] It should be noted that the force value allocation result is obtained by interpolating the force value of each control point based on the curvature value corresponding to the position coordinates of each control point and the optimized contact force. Specifically, several key control points are first selected as interpolation reference points on the correction path. The selection criteria for key control points are: calculating the second derivative of the curvature at each position on the path, and selecting the positions where the absolute value of the second derivative of curvature exceeds the preset curvature change threshold as key control points. In this embodiment, the preset curvature change threshold is set to 0.2 / m². This threshold is chosen because when it is less than 0.2 / m², the curvature change is gradual, and linear interpolation can meet the accuracy requirements; when it is greater than or equal to 0.2 / m², the curvature change is significant, and key control points need to be added to ensure interpolation accuracy. Based on the position coordinates of the key control points, the curvature values ​​corresponding to the key control points are obtained from the curvature values ​​at each position on the workpiece surface. Combined with the optimized contact force, the initial force value of each key control point is calculated according to the preset mapping relationship. The preset mapping relationship is defined using a linear weighted method, specifically: the initial force value of the key control point equals the curvature value multiplied by the curvature coefficient, plus the optimized contact force multiplied by the contact force coefficient, plus a reference bias term. In this embodiment, the curvature coefficient is set to 200 N·m, indicating that an additional 200 N of contact force is required per unit curvature (per meter); the contact force coefficient is set to 0.8, indicating that 80% of the optimized contact force contributes to the initial force value; the reference bias term is set to 10 N, used to ensure that the initial force value is not lower than the minimum effective grinding force. The physical meaning of this mapping relationship is: the greater the curvature, the more severe the bending of the workpiece surface, and the greater the required contact force; the greater the optimized contact force, the greater the initial force value; the reference bias term ensures that even when the curvature and optimized contact force are both small, the initial force value can still meet the basic grinding requirements. Then, a third-order Bézier curve interpolation algorithm is used, with the initial force value of the key control point as the reference point, to interpolate the force values ​​of the remaining control points on the correction path.

[0039] Specifically, the basic principle of third-order Bézier curve interpolation is to introduce two intermediate control points between two adjacent critical control points, in addition to the two endpoints themselves. These four points together determine the shape of the force value variation curve for that segment. The rule for generating the two intermediate control points is as follows: an offset is applied in the vertical direction of the line connecting the force values ​​of adjacent critical control points. The offset amount is proportional to the curvature change within that path segment; the greater the curvature change, the greater the offset, thus ensuring that the force value variation curve adapts to the curvature characteristics of the workpiece surface. The above process is repeated for all control points on the correction path to obtain the force value corresponding to each control point, forming the force value allocation result.

[0040] In step S16, the force allocation result is simulated to obtain simulation results. Invalid allocations are then removed based on the simulation results. The stability of the removed force allocation results is then assessed to obtain a valid allocation sequence, including: Finite element simulation was performed on each control point in the force distribution result to obtain the peak contact stress corresponding to each control point, and the results were summarized into simulation results. The control points in the simulation results whose contact stress peak exceeds the preset stress threshold are removed from the force value allocation results to obtain the removed force value allocation results. The rate of change of force values ​​between adjacent control points in the eliminated force value allocation result is calculated, and the rate of change of force values ​​is compared with a preset rate of change threshold. The force values ​​of control points that exceed the preset rate of change threshold are locally adjusted so that the rate of change of force values ​​does not exceed the preset rate of change threshold, thereby obtaining an effective allocation sequence.

[0041] In one implementation, the finite element analysis algorithm is a numerical calculation method that discretizes a continuum into a finite number of elements for mechanical analysis. By establishing a finite element model of the workpiece, the mechanical response of the workpiece under force values ​​at each control point is simulated, and the contact stress distribution on the workpiece surface at each control point is calculated. In this embodiment, the workpiece material is aluminum alloy 6061-T6 (elastic modulus 70 GPa, Poisson's ratio 0.33, yield strength 280 MPa), using tetrahedral solid elements with a mesh size of 2 mm (refined to 0.5 mm in the contact area). The bottom surface of the workpiece is completely fixed, and the contact is achieved using a penalty function algorithm (friction coefficient 0.2). The force distribution result obtained in step S15 is used as the load input for the finite element simulation. Simulation calculations are performed on each control point in the force distribution result to obtain the peak contact stress corresponding to each control point. The peak contact stress values ​​of all control points are summarized to form the simulation result. The simulation result is used for subsequent invalid allocation elimination to identify control points in the force distribution result that have a risk of stress exceeding the limit.

[0042] It should be noted that the peak contact stress is the maximum contact stress value borne by the workpiece surface at a certain control point, obtained from finite element simulation calculations. It reflects the actual load effect of the force distribution at that control point on the workpiece surface. The preset stress threshold is a pre-set safe upper limit for contact stress based on the mechanical properties of the workpiece material. It is used to determine whether the force distribution at a certain control point will cause damage to the workpiece surface. The peak contact stress corresponding to each control point in the simulation results is compared with the preset stress threshold. Control points whose peak contact stress exceeds the preset stress threshold are removed from the force distribution results, resulting in the removed force distribution results. For example, for workpieces with a soft coating (such as a 0.5 mm thick polyurethane elastomer coating), the preset stress threshold is set to... Pa, when the peak contact stress at a certain control point reaches If the stress exceeds the preset stress threshold, the control point is removed from the force distribution results. For uncoated metal workpieces (such as aluminum alloys), the preset stress threshold should be calculated based on the material's yield strength with a safety factor of 0.5, and is exemplarily set as follows: Pa.

[0043] It should be noted that the force value change rate is calculated by subtracting the force value difference between adjacent control points and the distance between control points in the force value allocation result after rejection, reflecting the magnitude of force value change between adjacent control points. The preset change rate threshold is an upper limit of the force value change rate set in advance according to the processing requirements, used to determine whether the force value allocation between adjacent control points is stable. The force value change rate is compared with the preset change rate threshold, and the force value of control points exceeding the preset change rate threshold is locally adjusted. Specifically, the force values ​​of the control point and its adjacent control points are smoothed so that the force value change rate between adjacent control points after adjustment does not exceed the preset change rate threshold. The above process is repeated for all adjacent control points in the force value allocation result after rejection to obtain an effective allocation sequence. For example, for the fine grinding stage of aluminum alloy workpieces, the preset change rate threshold is set to 0.6 Newtons per meter. When the force value change rate between adjacent control points in a certain path reaches 0.8 Newtons per meter, exceeding the preset change rate threshold, the force value of the control points in that path is locally adjusted to reduce the force value change rate to 0.55 Newtons per meter. For the rough grinding stage, this threshold can be relaxed to 1.2 Newtons per meter; for steel workpieces, this threshold can be relaxed to 1.0 Newtons per meter.

[0044] In step S17, the robot is controlled to perform polishing according to the effective allocation sequence, the polishing process is monitored in real time, and the polishing operation is completed, including: The effective allocation sequence is transmitted to the robot execution unit, and the robot is controlled to perform the polishing operation according to the effective allocation sequence. Force distribution data is collected during the grinding process, and the force distribution data is subjected to fast Fourier transform to obtain the frequency domain characteristics of force fluctuations. The deviation between the force value data exceeding the preset upper limit in the force fluctuation frequency domain feature and the force value of the corresponding control point in the effective allocation sequence is calculated to obtain the deviation correction amount, and the contact force of the robot execution unit is adjusted according to the deviation correction amount. Surface roughness data during the grinding process is acquired, and statistical calculations are performed on the force distribution data and the surface roughness data to obtain a processing consistency index.

[0045] In one implementation, the robot execution unit is an execution control module responsible for receiving control commands and driving the robot end effector to perform grinding actions according to specified parameters. The effective allocation sequence obtained in step S16 is transmitted to the robot execution unit. The robot execution unit, according to the position coordinates and stress values ​​of each control point in the effective allocation sequence, sequentially controls the robot end effector to move along the corrected path and apply the corresponding contact force to complete the grinding operation on the workpiece surface.

[0046] It should be noted that the force distribution data is a set of contact force measurements collected in real time by sensors installed at the robot's end effector during the grinding process. This data reflects the actual force distribution applied to the workpiece by the robot's end effector during grinding. The Fast Fourier Transform (FFT) is a mathematical algorithm that converts a time-domain signal into a frequency-domain signal. By performing an FFT on the force distribution data, the time-domain force fluctuation signal is converted into a frequency-domain representation, yielding the force fluctuation frequency domain characteristics. These characteristics describe the amplitude distribution of the force fluctuation signal at each frequency component during the grinding process. In this embodiment, the sampling frequency of the force sensor is set to 100 Hz, consistent with the sampling frequency in step S11. The number of points for the FFT is set to 1024 to obtain sufficient frequency resolution. Before the transformation, a Hanning window is used to window the force distribution data to suppress spectral leakage. The Hanning window is chosen because its sidelobe attenuation performance is superior to that of a rectangular window, effectively reducing spectral interference caused by the non-periodic truncation of the force signal.

[0047] It should be noted that the preset upper limit is a safety upper limit for the force fluctuation amplitude pre-set according to the grinding process requirements. It is used to determine whether the amplitude of each frequency component in the force fluctuation frequency domain feature exceeds the normal range. The force value data exceeding the preset upper limit in the force fluctuation frequency domain feature is compared with the force value of the corresponding control point in the effective allocation sequence to calculate the deviation correction amount. The deviation correction amount reflects the magnitude of the deviation between the current actual contact force and the set force value in the effective allocation sequence. The contact force of the robot execution unit is adjusted according to the deviation correction amount so that the actual applied contact force returns to the range set by the effective allocation sequence, ensuring the stability of the contact force during the grinding operation. For example, when the force value at a certain control point suddenly changes to 125.8 Newtons, exceeding the preset upper limit of 123.0 Newtons, the deviation correction amount is calculated to be 2.8 Newtons. Based on this, the contact force of the robot execution unit is adjusted to adjust the actual contact force to 122.5 Newtons.

[0048] It should be noted that the surface roughness data is the measured value of the workpiece surface roughness collected by a roughness detection device during the grinding operation. The processing consistency index is a comprehensive evaluation index obtained by statistically calculating the force distribution data and surface roughness data. Specifically, the processing consistency index is calculated as follows: first, the ratio of the standard deviation to the mean of the force distribution data and the ratio of the standard deviation to the mean of the surface roughness data are calculated separately; then, these two ratios are subtracted from 1 to obtain two stability components; finally, the two stability components are multiplied by their respective weight coefficients and summed to obtain the processing consistency index, where the sum of the two weight coefficients is 1. The weight coefficients are determined by performing a grid search on historical grinding operation data. The processing consistency index is calculated under multiple candidate weight combinations, and finally, the weight combination that has the highest correlation between the processing consistency index and the manually evaluated processing quality level is selected as the preset weight coefficient. Those skilled in the art will understand that the weight coefficient corresponding to the force distribution data is usually between 0.4 and 0.6. For example, the weighting coefficient corresponding to the force distribution data is set to 0.5, and the weighting coefficient corresponding to the surface roughness data is set to 0.5. When the mean of the force distribution data is 120 Newtons and the standard deviation is 2.3 Newtons, and the mean of the surface roughness data is 0.45 micrometers and the standard deviation is 0.05 micrometers, the processing consistency index is 0.93 obtained by statistical calculation.

[0049] In another implementation, before the simulation verification in step S16, a preliminary screening of the force allocation results can be performed, directly eliminating control points that exceed the preset force value range. The preset force value range is a pre-defined range of force values ​​based on the rated output force of the robot's end effector and the mechanical properties of the workpiece material. The lower limit of the preset force value range is determined by the minimum contact force required to complete effective grinding, and the upper limit is determined by the smaller of the rated output force of the robot's end effector and the bearing capacity of the workpiece material. Those skilled in the art will understand that the preset force value range can be adjusted within the range of 50 Newtons to 300 Newtons depending on the robot model and the workpiece material. This method ensures that subsequent simulation verification only calculates for control points with force values ​​within a reasonable range, thereby reducing the computational load of the simulation.

[0050] In summary, this invention discloses an adaptive control method for a robot grinding path, comprising: acquiring force and deformation data at multiple contact points between the robot end effector and the workpiece surface; performing fitting operations on the force and deformation data to obtain a stiffness estimate; dividing the workpiece surface into regions based on the stiffness estimate to obtain stiffness sub-regions; identifying stiffness difference boundaries between adjacent stiffness sub-regions to obtain region difference boundaries; acquiring force distribution data within the region difference boundaries; filtering the force distribution data to obtain smoothed force distribution data; performing regression fitting on the smoothed force distribution data to obtain a fitting curve; performing derivative calculations on the fitting curve to obtain the stiffness change trend; and acquiring the current... The invention employs a method of obtaining an optimized contact force by adjusting the current contact force based on the stiffness variation trend. It then acquires the initial grinding path of the workpiece, corrects it according to the optimized contact force, and discretizes the corrected path to obtain multiple control points. The force value at each control point is calculated based on the optimized contact force to obtain a force allocation result. The force allocation result is simulated to obtain simulation results. Invalid allocations are eliminated based on the simulation results, and the stability of the eliminated force allocation results is assessed to obtain an effective allocation sequence. The robot is then controlled to perform grinding according to the effective allocation sequence, and the grinding process is monitored in real time to complete the grinding operation. This invention solves the problem in existing technologies where real-time adaptive adjustment of contact force parameters and grinding paths based on the local stiffness distribution of the workpiece surface, leading to difficulties in ensuring processing consistency, is impossible due to the inability to do so.

[0051] Reference Figure 2 The second embodiment of the present invention provides a robot grinding path adaptive control system, comprising: The stiffness estimation module is used to acquire force and deformation data at multiple contact points between the robot end effector and the workpiece surface, and to perform fitting calculations on the force and deformation data to obtain stiffness estimates. The region division module is used to divide the workpiece surface into regions based on the stiffness estimate to obtain stiffness sub-regions, and to identify stiffness difference boundaries between adjacent stiffness sub-regions to obtain region difference boundaries. The trend perception module is used to acquire force distribution data within the boundary of the regional difference, filter the force distribution data to obtain smooth force distribution data, perform regression fitting on the smooth force distribution data to obtain a fitting curve, and perform derivative calculation on the fitting curve to obtain the stiffness change trend. The contact force optimization module is used to obtain the current contact force of the robot control system, and adjust the current contact force according to the stiffness change trend to obtain the optimized contact force; The path correction module is used to obtain the initial grinding path of the workpiece, correct the initial grinding path according to the optimized contact force to obtain the corrected path, discretize the corrected path to obtain multiple control points, and calculate the force value of each control point according to the optimized contact force to obtain the force value allocation result. The simulation filtering module is used to simulate the force value allocation result, obtain the simulation result, remove invalid allocations from the force value allocation result based on the simulation result, and perform stability judgment on the removed force value allocation result to obtain the valid allocation sequence. The execution monitoring module is used to control the robot to perform grinding according to the effective allocation sequence, monitor the grinding process in real time, and complete the grinding operation.

[0052] It should be noted that the robot grinding path adaptive control system provided in this embodiment of the invention is used to execute all the process steps of the robot grinding path adaptive control method in the above embodiment. The working principle and beneficial effect of the two are one-to-one, so they will not be described again.

[0053] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0054] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An adaptive control method for robot grinding path, characterized in that, include: Force and deformation data at multiple contact points between the robot end effector and the workpiece surface are acquired, and the force and deformation data are fitted to obtain a stiffness estimate. The workpiece surface is divided into regions based on the stiffness estimate to obtain stiffness sub-regions. Stiffness difference boundaries between adjacent stiffness sub-regions are identified to obtain region difference boundaries. Force distribution data within the boundary of the region difference is obtained, the force distribution data is filtered to obtain smooth force distribution data, regression fitting is performed on the smooth force distribution data to obtain a fitting curve, and derivative calculation is performed on the fitting curve to obtain the stiffness change trend. The current contact force of the robot control system is obtained, and the current contact force is adjusted according to the stiffness change trend to obtain an optimized contact force; The initial grinding path of the workpiece is obtained, and the initial grinding path is corrected according to the optimized contact force to obtain the corrected path. The corrected path is discretized to obtain multiple control points. The force value of each control point is calculated according to the optimized contact force to obtain the force value distribution result. The force value allocation result is simulated to obtain the simulation result. Invalid allocations are removed from the force value allocation result based on the simulation result. The stability of the removed force value allocation result is judged to obtain the valid allocation sequence. The robot is controlled to perform polishing according to the effective allocation sequence, and the polishing process is monitored in real time to complete the polishing operation.

2. The adaptive control method for robot grinding path according to claim 1, characterized in that, The process of acquiring force and deformation data at multiple contact points between the robot end effector and the workpiece surface, and performing fitting calculations on the force and deformation data to obtain a stiffness estimate includes: Force and displacement sensors installed at the end of the robot are used to collect force and deformation data at each contact point. A linear regression fit is performed on the force value data and the deformation data of each contact point to obtain the stiffness slope value corresponding to each contact point. Obtain the historical average stiffness value, and perform a weighted calculation on the stiffness slope value corresponding to each contact point and the historical average stiffness value to obtain the stiffness estimate value corresponding to each contact point.

3. The adaptive control method for robot grinding path according to claim 1, characterized in that, The step of dividing the workpiece surface into regions based on the stiffness estimate to obtain stiffness sub-regions, and identifying stiffness difference boundaries between adjacent stiffness sub-regions to obtain region difference boundaries includes: The workpiece surface is uniformly divided at a preset interval to obtain multiple grid units. The force and deformation data at each grid unit are fitted to obtain the stiffness estimate corresponding to each grid unit. The stiffness estimates of all the mesh elements are iteratively clustered using the k-means clustering algorithm to obtain multiple stiffness sub-regions; The stiffness estimates of adjacent grid cells are interpolated to obtain adjacent stiffness differences. The adjacent stiffness differences are compared with a preset difference threshold. Grid cells that exceed the preset difference threshold are marked as boundary points. All boundary points are connected to obtain regional difference boundaries.

4. The robot grinding path adaptive control method according to claim 1, characterized in that, The process of acquiring force distribution data within the boundary of the regional difference, filtering the force distribution data to obtain smooth force distribution data, performing regression fitting on the smooth force distribution data to obtain a fitted curve, and performing derivative calculation on the fitted curve to obtain the stiffness change trend includes: The force distribution data is processed using a Kalman filter, and the state estimation is iteratively updated based on the preset process noise covariance and the preset measurement noise covariance to obtain smooth force distribution data. The smooth force distribution data is fitted with a second-order polynomial regression to obtain a fitted curve. The derivative of the fitted curve is then calculated to obtain the stiffness variation trend.

5. The robot grinding path adaptive control method according to claim 1, characterized in that, The process of acquiring the current contact force of the robot control system and adjusting the current contact force according to the stiffness change trend to obtain an optimized contact force includes: The stiffness change trend is compared with a preset trend threshold. If the stiffness change trend exceeds the preset trend threshold, the contact force deviation between the current contact force and the preset target contact force is calculated. The current contact force is iteratively adjusted based on the contact force deviation to obtain the adjusted contact force. The difference between the adjusted contact force and the current contact force is calculated along the X-axis, Y-axis and Z-axis to obtain the force compensation vector; The force compensation vector is superimposed on the current contact force to obtain the optimized contact force.

6. The robot grinding path adaptive control method according to claim 1, characterized in that, The process involves obtaining an initial grinding path for the workpiece, correcting the initial grinding path based on the optimized contact force to obtain a corrected path, discretizing the corrected path to obtain multiple control points, and calculating the force value for each control point based on the optimized contact force to obtain a force value allocation result, including: A three-dimensional surface model of the workpiece is performed to obtain the curvature values ​​at various locations on the workpiece surface; The path correction coefficient is obtained by calculating the curvature value and the optimized contact force. The initial polishing path is adjusted according to the path correction coefficient to obtain the corrected path; The corrected path is discretized and sampled to obtain multiple control points; The curvature value is obtained based on the position coordinates of each control point. The force value of each control point is interpolated based on the curvature value and the optimized contact force to obtain the force value allocation result.

7. The adaptive control method for robot grinding path according to claim 1, characterized in that, The process involves simulating the force allocation results to obtain simulation results, removing invalid allocations based on the simulation results, and then performing a stability assessment on the removed force allocation results to obtain a valid allocation sequence, including: Finite element simulation was performed on each control point in the force distribution result to obtain the peak contact stress corresponding to each control point, and the results were summarized into simulation results. The control points in the simulation results whose contact stress peak exceeds the preset stress threshold are removed from the force value allocation results to obtain the removed force value allocation results. The rate of change of force values ​​between adjacent control points in the eliminated force value allocation result is calculated, and the rate of change of force values ​​is compared with a preset rate of change threshold. The force values ​​of control points that exceed the preset rate of change threshold are locally adjusted so that the rate of change of force values ​​does not exceed the preset rate of change threshold, thereby obtaining an effective allocation sequence.

8. The adaptive control method for robot grinding path according to claim 1, characterized in that, The step of controlling the robot to perform polishing according to the effective allocation sequence, monitoring the polishing process in real time, and completing the polishing operation includes: The effective allocation sequence is transmitted to the robot execution unit, and the robot is controlled to perform the polishing operation according to the effective allocation sequence. Force distribution data is collected during the grinding process, and the force distribution data is subjected to fast Fourier transform to obtain the frequency domain characteristics of force fluctuations. The deviation between the force value data exceeding the preset upper limit in the force fluctuation frequency domain feature and the force value of the corresponding control point in the effective allocation sequence is calculated to obtain the deviation correction amount, and the contact force of the robot execution unit is adjusted according to the deviation correction amount. Surface roughness data during the grinding process is acquired, and statistical calculations are performed on the force distribution data and the surface roughness data to obtain a processing consistency index.

9. A robot grinding path adaptive control system, characterized in that, include: The stiffness estimation module is used to acquire force and deformation data at multiple contact points between the robot end effector and the workpiece surface, and to perform fitting calculations on the force and deformation data to obtain stiffness estimates. The region division module is used to divide the workpiece surface into regions based on the stiffness estimate to obtain stiffness sub-regions, and to identify stiffness difference boundaries between adjacent stiffness sub-regions to obtain region difference boundaries. The trend perception module is used to acquire force distribution data within the boundary of the regional difference, filter the force distribution data to obtain smooth force distribution data, perform regression fitting on the smooth force distribution data to obtain a fitting curve, and perform derivative calculation on the fitting curve to obtain the stiffness change trend. The contact force optimization module is used to obtain the current contact force of the robot control system, and adjust the current contact force according to the stiffness change trend to obtain the optimized contact force; The path correction module is used to obtain the initial grinding path of the workpiece, correct the initial grinding path according to the optimized contact force to obtain the corrected path, discretize the corrected path to obtain multiple control points, and calculate the force value of each control point according to the optimized contact force to obtain the force value allocation result. The simulation filtering module is used to simulate the force value allocation result, obtain the simulation result, remove invalid allocations from the force value allocation result based on the simulation result, and perform stability judgment on the removed force value allocation result to obtain the valid allocation sequence. The execution monitoring module is used to control the robot to perform grinding according to the effective allocation sequence, monitor the grinding process in real time, and complete the grinding operation.

Images