A method for planning the grinding path of a rolling mill groove

Abstract

This invention discloses a method for planning grinding paths in rolling mill grooves, relating to the field of automated grinding technology for steel rolling equipment. The method comprises the following steps: S1, acquiring the original point cloud data of the rolling mill surface and establishing a robot coordinate system and a local point cloud coordinate system for the contacting rolling mill; S2, distinguishing the cross-sectional fitting plane and axial direction according to preset groove type parameters; S3, calculating the center coordinates and fitting radius of the groove cross-section in the local coordinate system; S4, filtering the point cloud data participating in the binning process using a quantile range clipping strategy based on groove type and surface type; S5, calculating adaptive binning parameters based on the filtered point cloud data; S6, determining the axial center and radius function fitting of the groove; S7, symmetrically arranging grinding points to generate a grinding trajectory. This invention achieves high-precision automated grinding of rolling mill grooves, improving the efficiency and surface quality of rolling mill grinding operations, and reducing manual labor and safety risks.

Description

Technical Field

[0001] This invention relates to the field of automated grinding technology for steel rolling equipment, and in particular to a method for planning grinding paths for roll grooves. Background Technology

[0002] In bar and wire rod rolling production, the surface condition of the roll grooves directly determines the stability of the wire rod and the quality of the finished product. Due to the long-term exposure to high temperature and humidity, an iron oxide layer is easily formed on the surface of the roll grooves, leading to production accidents such as unstable steel biting, wire rod slippage, and steel piling. Therefore, when replacing the roll grooves, grinding and maintenance must be completed quickly and with high quality within the limited maintenance time.

[0003] Currently, the grinding of rolls relies on manual methods, but traditional manual grinding has significant limitations: high labor intensity, high risk of occupational diseases, and prominent safety hazards; the grinding effect depends on the operator's experience, and problems such as uneven grinding, missed grinding, and over-grinding are prone to occur, affecting the surface quality of the rolling groove and the quality of the finished product. As the steel industry upgrades towards intelligence, digitalization, and automation, roll grinding robots have become the core equipment to replace manual labor, and grinding path planning is the key technology to achieve automated and high-precision robotic grinding.

[0004] Existing grinding path planning methods have several shortcomings: they struggle to handle issues such as point cloud noise, small arc segments, and localized defects in the complex environment of steel mills, resulting in low groove shape recognition accuracy; they have poor adaptability to different types of rolling grooves, such as vertical and horizontal grooves, and lack versatility; the path planning does not fully consider the motion constraints of the robotic arm and the requirements of the grinding process, easily leading to collision risks or poor grinding consistency. Therefore, there is an urgent need for a high-precision, adaptive, and universal rolling groove grinding path planning method that can stably achieve high accuracy under complex working conditions. Summary of the Invention

[0005] The purpose of this invention is to propose a method for grinding path planning of roll grooves. With three-dimensional point cloud as the core, it achieves high-precision grinding path planning for rolls with different groove shapes under complex working conditions through multi-step collaborative processing, thereby improving grinding consistency and automation level.

[0006] To achieve the above objectives, this invention proposes a method for planning the grinding path of a roll groove, the specific steps of which are as follows:

[0007] Step S1: Use a 3D camera to acquire the original point cloud data of the roll surface, and based on the pre-completed hand-eye calibration, obtain the relative positional relationship between the camera and the robot, and establish the robot coordinate system and the local point cloud coordinate system of the roll.

[0008] Step S2: Cut out the sub-point cloud containing the target groove region from the point cloud data in the local point cloud coordinate system, and distinguish the cross-sectional fitting plane and the axial direction according to the preset groove type parameters;

[0009] Step S3: Calculate the center coordinates and fitting radius of the groove cross section in the local coordinate system, specifically including: decentering, improved Pratt linear circle fitting, angle coverage degradation detection and damping enhancement, and robust LM fine-tuning based on geometric distance residuals.

[0010] Step S4: Construct a bucketed point cloud mask, and use a quantile range clipping strategy to filter the point cloud data participating in bucketing according to the slot type and surface type;

[0011] Step S5: Calculate adaptive bucketing parameters based on the filtered point cloud data, statistically obtain radius data by bucketing along the axial direction, and form an axial radius profile after smoothing.

[0012] Step S6: Select either symmetrical quadratic fitting or flat segment search strategy according to the surface type to determine the axial center of the groove and complete the radius function fitting;

[0013] Step S7: Arrange three grinding points symmetrically near the center of the grinding groove, calculate the coordinates of the grinding contact points and the tangent direction, transform the three contact points in the local coordinate system to the robot coordinate system, and generate the grinding trajectory.

[0014] Preferably, in step S1, the method for constructing the local point cloud coordinate system is to obtain the main distribution direction of the point cloud on the roll surface based on principal component analysis, move the origin of the point cloud coordinates to the geometric center of the roll, and align the coordinate axis directions with the axial and radial directions of the roll.

[0015] Preferably, in step S2, the cross-sectional fitting plane and axial direction are distinguished according to the preset groove type parameters, specifically as follows:

[0016] When the groove type is vertical, the cylindrical axis is along the X-axis in the global coordinate system, and the cross-section is located in the YZ plane.

[0017] When the groove type is a transverse groove, the cylindrical axis is along the Z-axis in the global coordinate system, and the cross-section is located in the XY plane.

[0018] Preferably, in step S3, the specific calculation method is as follows:

[0019] Step S31, Decentralization Processing:

[0020] For point cloud sets in local coordinate system Calculate the centroid The cross-sectional plane is selected according to the type of the rolling groove, as shown in the following formula:

[0021] ;

[0022] The vertical groove is fitted to a circle using the YZ plane as the plane, as shown in the following formula:

[0023] ;

[0024] The horizontal groove is fitted to a circle using the XY plane as the plane, as shown in the following formula:

[0025] ;

[0026] in, N The total number of point clouds in the rolling mill groove. For the point cloud of the rolling groove in the local coordinate system X Mean along the axis For the point cloud of the rolling groove in the local coordinate system Y Mean along the axis For the point cloud of the rolling groove in the local coordinate system Z Mean along the axis , The coordinates are the decentralized ones. , , The first i The point cloud of the rolling groove in the local coordinate system X axis, Y axis, Z Coordinates along the axis;

[0027] The improved Pratt linear circle fitting is as follows:

[0028] Second- and third-order statistics are constructed based on two-dimensional point sets, as shown in the following formulas:

[0029] ;

[0030] ;

[0031] in, , , The set composed of the point clouds of the rolling groove is in u Second-order cumulant in the direction, v Second-order cumulant in the direction, UV The correlation cumulative amount between the two axes, , , , The set composed of the point clouds of the rolling groove is in u Third-order cumulants in the direction, in v Third-order cumulants in the direction u and v 2 Mixed third-order cumulants, u 2 and v The mixed third-order cumulants;

[0032] A system of linear equations can be established, as shown in the following formula:

[0033] ;

[0034] ;

[0035] ;

[0036] in, A The coefficient matrix, b To solve for the right-hand vector of the equation of the center of the circle, , These are the x and y coordinates of the initial value of the center of the cross-section, respectively.

[0037] Step S32: Introduce adaptive damping The damping strength is automatically adjusted based on the matrix condition number to obtain the initial value of the center of the circle, as shown in the following formula:

[0038] ;

[0039] The initial value of the radius is calculated using the mean square distance, as follows:

[0040] ;

[0041] in, This is the initial estimate of the circle's radius;

[0042] Step S33: Through angle coverage degradation detection and damping enhancement, calculate the polar angle of each point with the initial value center, using the following formula:

[0043] ;

[0044] in, For the first i The polar angle of each point relative to the center of the initial value circle;

[0045] Step S34: Robust LM fine-tuning based on geometric distance residuals improves the robustness of the circle center and radius, as shown in the following formula:

[0046] ;

[0047] in, For the first i The Euclidean distance from each point to the current center of the circle. For the first i The geometric distance residuals of each point relative to the current circle. r This is the radius estimate for the current iteration;

[0048] Step S35: Output the fitting radius of the groove cross-section and the coordinates of the center of the circle in the local coordinate system, as follows:

[0049] The center of the two-dimensional fitting circle The inverse transformation back to a three-dimensional coordinate system is as follows:

[0050] When fitting the YZ plane:

[0051] ;

[0052] in,( () represents the coordinates of the center of the circle in the YZ plane;

[0053] When fitting to the XY plane:

[0054] ;

[0055] in,( () represents the coordinates of the center of the circle in the XY plane;

[0056] The fitting radius is calculated using the average distance under the final circle center, as shown in the following formula:

[0057] ;

[0058] Where R is the fitted radius of the groove cross section.

[0059] Preferably, in step S34, a robust estimation strategy is used for iterative optimization, specifically including:

[0060] Truncation and outlier removal: Calculate residual magnitude using initial values The sample with the largest residual is removed, and the trusted sample is retained to participate in the iteration.

[0061] Dual-weighted function weighting: Weights are constructed based on residuals and scaling estimates. When the residuals are too large, the weights automatically decay to zero.

[0062] For parameter vectors Construct the Jacobian matrix, establish the weighted normal equation, and the... i The Jacobian formula for a row is as follows:

[0063] ;

[0064] All of Stacking them together yields the overall Jacobian matrix:

[0065] ;

[0066] ;

[0067] in, For Jacobian matrices,W It is a diagonal weight matrix. e For the geometric distance residual set, T This is the matrix transpose; during the iterative initialization phase, all sample points participating in the fitting are assigned consistent initial weights (x). 𝑖 =1), and then the geometric residuals φ at each point are calculated based on the initial values ​​of the circular parameters obtained from the Pratt linear solution. 𝑖 The Tukey two-weight robust function is used to map the residuals to weights 𝑤. 𝑖 ∈[0,1], construct a diagonal weight matrix from the weights:

[0068] ;

[0069] In each subsequent iteration, the residuals are recalculated and updated using the current circle parameters. W This is to reduce the impact of outliers on the fitting results.

[0070] Solve for increment and update The iteration stops when the increment norm is less than the threshold or the maximum number of iterations is reached.

[0071] Preferably, in step S4, a bucketed point cloud mask is constructed, and the specific filtering strategy is as follows:

[0072] When it is a vertical groove, use 5% to 95% of the cutting position in the Z direction;

[0073] For the axial X direction:

[0074] Concave surfaces: cut at 5% to 95% of the surface area;

[0075] Flat curved surfaces: use 20% to 80% segmentation for cutting;

[0076] When it is a horizontal groove, the X direction is cut at 5% to 95% intervals;

[0077] For the axial direction Z:

[0078] Concave surfaces: cut at 5% to 95% of the surface area;

[0079] Flat curved surfaces: use 20% to 80% segmentation for cutting;

[0080] The filtered point cloud used for binning is obtained by using the binning point cloud mask, as shown in the following formula:

[0081] ;

[0082] in, For the set of slotted point clouds participating in the barrel splitting, For the first i There are 1 set of groove point clouds, where P is the original set of groove point clouds.

[0083] Preferably, in step S5, the specific steps are as follows:

[0084] Step S51: Adaptive bucketing parameter calculation, based on the filtered parameters. Calculate the physical dimensions of the groove in the axial direction and determine the number of barrels based on the preset resolution. N bucket And apply boundary constraints; at the same time, set the minimum number of points per bucket and set a smoothing window according to the surface type;

[0085] Step S52: Construct the cross-sectional radius of the bucket. Using the determined center of the cross-section as a reference, calculate the cross-sectional radius from each point to the center of the circle.

[0086] When the vertical groove is in the YZ plane, the X-axis is used as the axial coordinate of the dividing barrel:

[0087] ;

[0088] When the horizontal groove is in the XY plane, the Z-axis is used as the axial coordinate of the dividing barrel:

[0089] ;

[0090] in, For the first i The distance from the point cloud of each participating bucket to the center of the cross-section, and the coordinates of the center of the cross-section of the vertical groove are... and The coordinates of the center of the cross-section of the transverse groove are and .

[0091] Preferably, in step S6, the axial center of the rolling groove is determined, and the radius function is fitted. The specific steps are as follows:

[0092] Step S61: When the surface of the rolling groove is a concave curved surface, the radius profile exhibits a valley bottom feature near the center of the groove. A symmetrical quadratic fitting is used to determine the axial center and output the quadratic function coefficients, as follows:

[0093] Step S611: Sort the centers of each bucket by axial coordinate to obtain x, corresponding to the radius y. Select the position with the smallest radius within the middle 2 / 5 to 3 / 5 interval as the initial center value. c And calculate the span;

[0094] Step S612, construct with the center c Reference Gaussian distance kernel weight k ( x ), and then exponentially pressurize the number of buckets. The shrinkage is multiplied by the distance kernel to form the counting weight, which is then proportionally multiplied by the uniform weight. Convex combination, while clipping the upper and lower bounds of the weights;

[0095] Step S613: Use weighted quadratic fitting Get vertex position As a new center candidate c new And limited to the middle range [ x lo ,x hi Within, through the step size coefficient Update the center, iterate until convergence; where, , , These are the coefficients of the quadratic term, the coefficients of the linear term, and the constant of the fitted general quadratic function, respectively.

[0096] Step S614, at the final center c The fitted symmetric model is given by the following formula:

[0097] ;

[0098] in, , These are the coefficients and constants of the fitted vertex-form quadratic function, respectively;

[0099] Step S62: When the groove surface is planar, the radius profile does not have obvious valley characteristics. The axial center is determined using the flat segment center strategy, and the radius trend of the flat curved surface is linearly fitted.

[0100] Step S621: Calculate the reference mean radius in the middle 2 / 5 to 3 / 5 interval. And construct the residual sequence:

[0101] ;

[0102] in, R i For each bucket r i The polymerization radius obtained after polymerization;

[0103] Step S622: Within the candidate center index range, construct a sliding window with a length of 2 / 3 of the effective buckets, and calculate the residual sum within the window as the cost function; for cases where the window is insufficient due to proximity to the edge, use the maximum residual for penalty filling; select the axial coordinate corresponding to the position with the minimum cost function. c new As the center of the flat curved surface;

[0104] Step S623, to c new Write back the axial components of the 3D circle center, and output them simultaneously. As a reference mean radius;

[0105] Step S624: Filter the point cloud The bucketing method is executed again, and a linear fit is performed on the relationship between the reference mean radius and the axial coordinate, as shown in the following formula:

[0106] ;

[0107] in, This represents the radius corresponding to the axial coordinate x.

[0108] Preferably, in step S7, the specific steps are as follows:

[0109] Step S71: Three-point axial arrangement. Based on the current groove width parameters, calculate the half-width and use the center coordinates (X... C ,Y C Z C Based on the reference point, three target circles are arranged symmetrically along the axis: the center point is offset relative to the left and right points or the top and bottom points by half the width; according to the pressing depth of the grinding head... r d and the angle of the grinding head The coordinates of the three grinding contact points are obtained using the following formula:

[0110] When the vertical groove is in the YZ plane:

[0111] ;

[0112] When the horizontal groove is in the XY plane, the Z-axis is used as the axial coordinate of the dividing barrel:

[0113] ;

[0114] in, , axial coordinates x The distance from the corresponding grinding point to the center coordinates;

[0115] Step S72, Calculation of tangent direction, as follows:

[0116] When creating a vertical groove, the circle lies in the Y-Z plane, and the tangent should be perpendicular to the normal and still lie in the Y-Z plane, constructed from the normal component.

[0117] When the groove is horizontal, the circle is located in the X-Y plane, the tangent should be perpendicular to the normal and located in the XY plane, and the direction of the normal should be selected according to the actual interference situation.

[0118] Step S73: Target point coordinate transformation. The three contact points in the local coordinate system are transformed to the robot coordinate system. After obtaining the three grinding points and the surface type of the groove, the robot automatically executes linear Cartesian motion or arc Cartesian motion to carry out the grinding task.

[0119] Therefore, this invention proposes a method for planning the grinding path of a roll groove, which has the following advantages:

[0120] (1) The present invention adopts a multi-step collaborative point cloud processing strategy. Through improved Pratt linear circle fitting, robust LM fine-tuning and other algorithms, it effectively copes with complex working conditions such as point cloud noise, small arc segments and local defects, significantly improves the recognition accuracy of the center and radius of the groove cross section, and lays a reliable foundation for subsequent path planning.

[0121] (2) The present invention can automatically distinguish between vertical grooves and horizontal grooves, and uses symmetrical quadratic fitting and flat segment search strategies to determine the axial center for concave curved surfaces and flat curved surfaces respectively. It can meet the grinding needs of rolls of different models and different wear states, and has wide applicability.

[0122] (3) The present invention adopts a three-point grinding point layout, combined with tangential direction optimization and coordinate transformation. The generated grinding trajectory not only covers the key grinding area, but also avoids over-processing of non-target areas. At the same time, it fully considers the motion constraints and process requirements of the robotic arm, reduces the risk of collision, and improves grinding consistency and surface quality.

[0123] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0124] Figure 1 This is a flowchart of a method for planning the grinding path of a rolling mill groove according to the present invention. Detailed Implementation

[0125] To make the technical solutions, advantages, and objectives of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below. The described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0126] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0127] Example

[0128] like Figure 1 As shown, the present invention provides a method for planning the grinding path of a rolling mill groove, the specific steps of which are as follows:

[0129] Step S1: Use a 3D camera to acquire the original point cloud data of the roll surface, and based on the pre-completed hand-eye calibration, obtain the relative positional relationship between the camera and the robot, and establish the robot coordinate system and the local point cloud coordinate system of the roll.

[0130] The method for constructing the local point cloud coordinate system is to obtain the main distribution direction of the point cloud on the roll surface based on principal component analysis, move the origin of the point cloud coordinates to the geometric center of the roll, and align the coordinate axis directions with the axial and radial directions of the roll.

[0131] Step S2: Extract the sub-point cloud containing the target groove region from the point cloud data in the local point cloud coordinate system, and distinguish the cross-sectional fitting plane and axial direction according to the preset groove type parameters, as follows:

[0132] When the groove type is vertical, the cylindrical axis is along the X-axis in the global coordinate system, and the cross-section is located in the YZ plane.

[0133] When the groove type is a transverse groove, the cylindrical axis is along the Z-axis in the global coordinate system, and the cross-section is located in the XY plane.

[0134] Step S3: Calculate the coordinates of the center of the groove cross-section and the fitted radius in the local coordinate system. The specific calculation is as follows:

[0135] Step S31, Decentralization Processing:

[0136] For point cloud sets in local coordinate system Calculate the centroid The cross-sectional plane is selected according to the type of the rolling groove, as shown in the following formula:

[0137] ;

[0138] The vertical groove is fitted to a circle using the YZ plane as the plane, as shown in the following formula:

[0139] ;

[0140] The horizontal groove is fitted to a circle using the XY plane as the plane, as shown in the following formula:

[0141] ;

[0142] in, N The total number of point clouds in the rolling mill groove. For the point cloud of the rolling groove in the local coordinate system X Mean along the axis For the point cloud of the rolling groove in the local coordinate system Y Mean along the axis For the point cloud of the rolling groove in the local coordinate system ZMean along the axis , The coordinates are the decentralized ones. , , The first i The point cloud of the rolling groove in the local coordinate system X axis, Y axis, Z Coordinates along the axis;

[0143] The improved Pratt linear circle fitting is as follows:

[0144] Second- and third-order statistics are constructed based on two-dimensional point sets, as shown in the following formulas:

[0145] ;

[0146] ;

[0147] in, , , The set composed of the point clouds of the rolling groove is in u Second-order cumulant in the direction, v Second-order cumulant in the direction, UV The correlation cumulative amount between the two axes, , , , The set composed of the point clouds of the rolling groove is in u Third-order cumulants in the direction, in v Third-order cumulants in the direction u and v 2 Mixed third-order cumulants, u 2 and v The mixed third-order cumulants;

[0148] A system of linear equations can be established, as shown in the following formula:

[0149] ;

[0150] ;

[0151] ;

[0152] in, A The coefficient matrix, b To solve for the right-hand vector of the equation of the center of the circle, , These are the x and y coordinates of the initial value of the center of the cross-section, respectively.

[0153] Step S32: Introduce adaptive damping The damping strength is automatically adjusted based on the matrix condition number to obtain the initial value of the center of the circle, as shown in the following formula:

[0154] ;

[0155] The initial value of the radius is calculated using the mean square distance, as follows:

[0156] ;

[0157] in, This is the initial estimate of the circle's radius;

[0158] Step S33: Through angle coverage degradation detection and damping enhancement, calculate the polar angle of each point with the initial value center, using the following formula:

[0159] ;

[0160] in, For the first i The polar angle of each point relative to the center of the initial value circle;

[0161] Step S34: Robust LM fine-tuning based on geometric distance residuals improves the robustness of the circle center and radius, as shown in the following formula:

[0162] ;

[0163] in, For the first i The Euclidean distance from each point to the current center of the circle. For the first i The geometric distance residuals of each point relative to the current circle. r This is the radius estimate for the current iteration;

[0164] Iterative optimization using a robust estimation strategy includes:

[0165] Truncation and outlier removal: Calculate residual magnitude using initial values The sample with the largest residual is removed, and the trusted sample is retained to participate in the iteration.

[0166] Dual-weighted function weighting: Weights are constructed based on residuals and scaling estimates. When the residuals are too large, the weights automatically decay to zero.

[0167] For parameter vectors Construct the Jacobian matrix, establish the weighted normal equation, and the... i The Jacobian formula for a row is as follows:

[0168] ;

[0169] All of Stacking them together yields the overall Jacobian matrix:

[0170] ;

[0171] ;

[0172] in, For Jacobian matrices, W It is a diagonal weight matrix. e For the geometric distance residual set, T This is the matrix transpose.

[0173] Solve for increment and update The iteration stops when the increment norm is less than the threshold or the maximum number of iterations is reached.

[0174] Step S35: Output the fitting radius of the groove cross-section and the coordinates of the center of the circle in the local coordinate system, as follows:

[0175] The center of the two-dimensional fitting circle The inverse transformation back to a three-dimensional coordinate system is as follows:

[0176] When fitting the YZ plane:

[0177] ;

[0178] in,( () represents the coordinates of the center of the circle in the YZ plane;

[0179] When fitting to the XY plane:

[0180] ;

[0181] in,( () represents the coordinates of the center of the circle in the XY plane;

[0182] The fitting radius is calculated using the average distance under the final circle center, as shown in the following formula:

[0183] ;

[0184] Where R is the fitted radius of the groove cross section.

[0185] Step S4: Construct a bucketed point cloud mask, and use a quantile range pruning strategy to filter the point cloud data participating in bucketing according to the slot type and surface type.

[0186] The specific screening strategy is as follows:

[0187] When it is a vertical groove, use 5% to 95% of the cutting position in the Z direction;

[0188] For the axial X direction:

[0189] Concave surfaces: cut at 5% to 95% of the surface area;

[0190] Flat curved surfaces: use 20% to 80% segmentation for cutting;

[0191] When it is a horizontal groove, the X direction is cut at 5% to 95% intervals;

[0192] For the axial Z direction:

[0193] Concave surfaces: cut at 5% to 95% of the surface area;

[0194] Flat curved surfaces: use 20% to 80% segmentation for cutting;

[0195] The filtered point cloud used for binning is obtained by using the binning point cloud mask, as shown in the following formula:

[0196] ;

[0197] in, For the set of slotted point clouds participating in the barrel splitting, For the first i There are 1 set of groove point clouds, where P is the original set of groove point clouds.

[0198] Step S5: Calculate adaptive bucketing parameters based on the filtered point cloud data, statistically obtain radius data by bucketing along the axial direction, and form an axial radius profile after smoothing. The specific steps are as follows:

[0199] Step S51: Adaptive bucketing parameter calculation, based on the filtered parameters. Calculate the physical dimensions of the groove in the axial direction and determine the number of barrels based on the preset resolution. N bucket And apply boundary constraints; at the same time, set the minimum number of points per bucket and set a smoothing window according to the surface type;

[0200] Step S52: Construct the cross-sectional radius of the bucket. Using the determined center of the cross-section as a reference, calculate the cross-sectional radius from each point to the center of the circle.

[0201] When the vertical groove is in the YZ plane, the X-axis is used as the axial coordinate of the dividing barrel:

[0202] ;

[0203] When the horizontal groove is in the XY plane, the Z-axis is used as the axial coordinate of the dividing barrel:

[0204] ;

[0205] in, For the first iThe distance from the point cloud of each participating bucket to the center of the cross-section, and the coordinates of the center of the cross-section of the vertical groove are... and The coordinates of the center of the cross-section of the transverse groove are and .

[0206] Step S6: Select either symmetrical quadratic fitting or flat segment search strategy according to the surface type to determine the axial center of the groove and complete the radius function fitting;

[0207] The specific steps are as follows:

[0208] Step S61: When the surface of the rolling groove is a concave curved surface, the radius profile exhibits a valley bottom feature near the center of the groove. A symmetrical quadratic fitting is used to determine the axial center and output the quadratic function coefficients, as follows:

[0209] Step S611: Sort the centers of each bucket by axial coordinate to obtain x, corresponding to the radius y. Select the position with the smallest radius within the middle 2 / 5 to 3 / 5 interval as the initial center value. c And calculate the span;

[0210] Step S612, construct with the center c Reference Gaussian distance kernel weight k ( x ), and then exponentially pressurize the number of buckets. The shrinkage is multiplied by the distance kernel to form the counting weight, which is then proportionally multiplied by the uniform weight. Convex combination, while clipping the upper and lower bounds of the weights;

[0211] Step S613: Use weighted quadratic fitting Get vertex position As a new center candidate c new And limited to the middle range [ x lo ,x hi Within, through the step size coefficient Update the center, iterate until convergence; where, , , These are the coefficients of the quadratic term, the coefficients of the linear term, and the constant of the fitted general quadratic function, respectively.

[0212] Step S614, at the final center c The fitted symmetric model is given by the following formula:

[0213] ;

[0214] in, , These are the coefficients and constants of the fitted vertex-form quadratic function, respectively;

[0215] Step S62: When the groove surface is planar, the radius profile does not have obvious valley characteristics. The axial center is determined using the flat segment center strategy, and the radius trend of the flat curved surface is linearly fitted.

[0216] Step S621: Calculate the reference mean radius in the middle 2 / 5 to 3 / 5 interval. And construct the residual sequence:

[0217] ;

[0218] in, R i For each bucket r i The polymerization radius obtained after polymerization;

[0219] Step S622: Within the candidate center index range, construct a sliding window with a length of 2 / 3 of the effective buckets, and calculate the residual sum within the window as the cost function; for cases where the window is insufficient due to proximity to the edge, use the maximum residual for penalty filling; select the axial coordinate corresponding to the position with the minimum cost function. c new As the center of the flat curved surface;

[0220] Step S623, to c new Write back the axial components of the 3D circle center, and output them simultaneously. As a reference mean radius;

[0221] Step S624: Filter the point cloud The bucketing method is executed again, and a linear fit is performed on the relationship between the reference mean radius and the axial coordinate, as shown in the following formula:

[0222] ;

[0223] in, This represents the radius corresponding to the axial coordinate x.

[0224] Step S7: Arrange three grinding points symmetrically near the center of the grinding groove, calculate the coordinates of the grinding contact points and the tangent direction, transform the three contact points in the local coordinate system to the robot coordinate system, and generate the grinding trajectory. The specific steps are as follows:

[0225] Step S71: Three-point axial arrangement. Based on the current groove width parameters, calculate the half-width and use the center coordinates (X... C ,Y C Z CBased on the reference point, three target circles are arranged symmetrically along the axis: the center point is offset relative to the left and right points or the top and bottom points by half the width; according to the pressing depth of the grinding head... r d and the angle of the grinding head The coordinates of the three grinding contact points are obtained using the following formula:

[0226] When the vertical groove is in the YZ plane:

[0227] ;

[0228] When the horizontal groove is in the XY plane, the Z-axis is used as the axial coordinate of the dividing barrel:

[0229] ;

[0230] in, , axial coordinates x The distance from the corresponding grinding point to the center coordinates;

[0231] Step S72, Calculation of tangent direction, as follows:

[0232] When creating a vertical groove, the circle lies in the Y-Z plane, and the tangent should be perpendicular to the normal and still lie in the Y-Z plane, constructed from the normal component.

[0233] When the groove is horizontal, the circle is located in the X-Y plane, the tangent should be perpendicular to the normal and located in the XY plane, and the direction of the normal should be selected according to the actual interference situation.

[0234] Step S73: Target point coordinate transformation. The three contact points in the local coordinate system are transformed to the robot coordinate system. After obtaining the three grinding points and the surface type of the groove, the robot automatically executes linear Cartesian motion or arc Cartesian motion to carry out the grinding task.

[0235] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.

[0236] Therefore, this invention provides a method for planning the grinding path of roll grooves. Through multi-step collaborative point cloud processing and adaptive path planning strategies, it can effectively cope with complex working conditions, adapt to different groove types and curved surface types, realize high-precision and high-consistency automated grinding of roll grooves, improve work efficiency and surface quality, reduce manual load and safety risks, and provide technical support for the intelligent upgrading of the steel industry.

[0237] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for planning a grinding path for a roll groove, characterized in that The specific steps are as follows: Step S1, acquiring original point cloud data of the roll surface by using a three-dimensional camera, obtaining the relative position relationship between the camera and the robot according to the pre-completed hand-eye calibration, and establishing a robot coordinate system and a local point cloud coordinate system fitted to the roll; Step S2, cropping the sub-point cloud containing the target roll groove region from the point cloud data in the local point cloud coordinate system, and distinguishing the cross-sectional fitting plane and the axial direction according to the preset roll groove type parameters; Step S3, calculating the roll groove cross-sectional center coordinates and fitting radius in the local coordinate system, specifically including: decentralization processing, improved Pratt linear circle fitting, angle coverage degradation detection and damping enhancement, and robust LM fine tuning based on geometric distance residual; Step S4, constructing a bucketed point cloud mask, and screening the point cloud data participating in the bucketing according to the quantile range cutting strategy according to the groove type and the curved surface type; Step S5, calculating adaptive bucketing parameters based on the filtered point cloud data, and obtaining radius data by axial bucketing statistics, and forming an axial radius profile after smoothing processing; Step S6, determining the roll groove axial center according to the curved surface type selection symmetric quadratic fitting or flat section search strategy to complete the radius function fitting; Step S7, symmetrically arranging three polishing points near the roll groove center, calculating the polishing contact point coordinates and tangent direction, converting the three-point contact points in the local coordinate system to the robot coordinate system, and generating a polishing trajectory.

2. The roll groove polishing path planning method of claim 1, wherein, In step S1, the construction method of the local point cloud coordinate system is to obtain the main distribution direction of the roll surface point cloud based on the principal component analysis method, move the coordinate origin of the point cloud to the geometric center of the roll, and align the coordinate axis direction with the axial direction and the radial direction of the roll.

3. The roll groove polishing path planning method of claim 2, wherein, In step S2, the cross-sectional fitting plane and the axial direction are distinguished according to the preset roll groove type parameters, specifically: When the roll groove type is vertical groove, the cylindrical axis is along the X-axis in the global coordinate system, and the cross section is in the Y-Z plane; When the roll groove type is horizontal groove, the cylindrical axis is along the Z-axis in the global coordinate system, and the cross section is in the X-Y plane.

4. The roll groove polishing path planning method of claim 3, wherein, In step S3, the specific calculation method is as follows: Step S31, decentralization processing: A point cloud set under a local coordinate system Computing the centroid And according to the type of groove, the cross-section plane is selected, and the formula is as follows: ； For vertical groove, the Y-Z plane is used as the circle fitting plane, and the formula is as follows: ； For horizontal groove, the X-Y plane is used as the circle fitting plane, and the formula is as follows: ； wherein, N is the total number of point clouds of the rolling groove, is the mean value of the rolling groove point cloud in the local coordinate system X in the axial direction, is the mean value of the rolling groove point cloud in the local coordinate system Y in the axial direction, is the mean value of the rolling groove point cloud in the local coordinate system Z in the axial direction, , is the coordinate after decentralization, , , is the coordinate of the point cloud of the first i rolling groove in the local coordinate system X axis, Y axis, Z axis direction, respectively; Improved Pratt linear circle fitting, specifically as follows: Constructing second-order and third-order statistics based on two-dimensional point set, the formula is as follows: ； ； wherein, , , are the second order cumulants of the set of groove point clouds in the u direction, in the v direction, and the cross-cumulants of the two axes, Establishing a linear equation set, the formula is as follows: , , , are the third order cumulants of the set of groove point clouds in the u direction, in the v direction, and the cross-cumulants of the two axes, u and v 2 the mixed third order cumulants of u 2 and v the mixed third order cumulants of​ Calculating the radius initial value by mean square distance, the formula is as follows: ； ； ； wherein, A is the coefficient matrix, b is the right-hand side vector for solving the circle center equation, , are the horizontal and vertical coordinates of the solved cross-sectional circle center initial value, respectively; Step S32, introducing adaptive damping And according to the matrix condition number to automatically adjust the damping strength, the initial value of the center is obtained, and the formula is as follows: ； Step S33, calculating the polar angle of each point by angle coverage degradation detection and damping enhancement based on the initial center, the formula is as follows: ； wherein is an initial estimate of the radius of the circle; Step S34, robust LM fine tuning based on geometric distance residual to improve the robustness of the center and the radius, the formula is as follows: ； wherein is the first i polar angle of the i-th point relative to the center of the initial circle; Step S35, outputting the fitting radius of the roll groove cross section and the center coordinates in the local coordinate system, specifically as follows: ； wherein, is the Euclidean distance of the i-th point to the current circle center, i is the Euclidean distance of the i-th point to the current circle center, is the geometric distance residual of the i-th point with respect to the current circle, i is the geometric distance residual of the i-th point with respect to the current circle, r is the radius estimate of the current iteration; When fitting in the Y-Z plane: Fitting the circle center in two dimensions Back-transforming to the three-dimensional coordinate system, the formula is as follows: When fitting in the X-Y plane: ； wherein, ) is the center coordinate of the circle under the Y-Z plane; The fitting radius is calculated by the average distance under the final center, the formula is as follows: ； wherein, ) is the center coordinate of the circle in the X-Y plane; Wherein, R is the fitting radius of the roll groove cross section. ； In step S34, iterative optimization is performed by using a robust estimation strategy, specifically including:

5. The roll groove polishing path planning method of claim 4, wherein, ​ Truncation: remove outliers based on initial residual magnitude , remove samples with largest residual, keep trusted samples for iteration Double weight function weighting: weight is constructed according to residual and scale estimation, and weight is automatically attenuated to zero when residual is too large; The parameter vector The Jacobian matrix is constructed, the weighted normal equations are established, and the i The Jacobian formula of the line is as follows: ； Stack all Stack to get the overall Jacobian matrix: ； ； wherein, is a Jacobian matrix, W is a diagonal weight matrix, e is a set of geometric distance residuals, T is a matrix transpose; Solve the increment and update Stop when the increment norm is less than a threshold or a maximum number of iterations is reached.

6. The roll groove grinding path planning method of claim 5, wherein, In step S4, a bucketed point cloud mask is constructed, and the specific screening strategy is as follows: When it is a vertical groove, 5%-95% quantile clipping is adopted for the Z direction; For the axial X direction: Concave surface: 5%-95% quantile clipping is adopted; Flat surface: 20%-80% quantile clipping is adopted; When it is a horizontal groove, 5%-95% quantile clipping is adopted for the X direction; For the axial Z direction: Concave surface: 5%-95% quantile clipping is adopted; Flat surface: 20%-80% quantile clipping is adopted; Through the bucketed point cloud mask, a filtered point cloud for bucketing is obtained, and the formula is as follows: ； wherein, is a set of pass point clouds participating in the bucketing, is the i pass point cloud, and P is the set of original pass point clouds.

7. The roll groove grinding path planning method of claim 6, wherein, In step S5, the specific steps are as follows: Step S51, adaptive bucket parameter calculation, based on the filtered , calculate the physical size of the nip in the axial direction, and determine the number of buckets according to the preset resolution N bucket , and impose boundary constraints; at the same time, set the minimum number of points in a single bucket, and set a smoothing window according to the surface type; Step S52, bucket radius profile construction: taking the determined cross-sectional center as the reference, the cross-sectional radius of each point to the center is calculated: When it is a vertical groove Y-Z plane, the X axis is taken as the bucketing axial coordinate: ； When it is a horizontal groove X-Y plane, the Z axis is taken as the bucketing axial coordinate: ； wherein, is the distance from the point cloud to the cross-sectional center of the vertical slot, and the cross-sectional center coordinates of the vertical slot are i and is the distance from the point cloud to the cross-sectional center of the horizontal slot, and the cross-sectional center coordinates of the horizontal slot are and .​ 8. The roll groove grinding path planning method of claim 7, wherein, In step S6, the rolling groove axial center is determined, and the radius function fitting is completed, and the specific steps are as follows: Step S61, when the rolling groove surface is a concave surface, the radius profile has a valley bottom feature near the groove center, and a symmetric quadratic fitting is adopted to determine the axial center and output the quadratic function coefficients, and the specific steps are as follows: Step S611, sort the center of each barrel by axial coordinate to get x, and the corresponding radius is y, and the minimum position of the radius in the middle 2 / 5-3 / 5 interval is selected as the center initial value c and calculate the span; Step S612, construct with center c Gaussian distance kernel weight for reference k ( x ), and the number of bucket points is compressed by an exponential distance kernel to form a counting weight, and the uniform weight is proportionally convex combination, and the upper and lower bounds of the weight are clipped; Step S613, using weighted quadratic fitting Obtaining vertex position As a new center candidate c new And limited in the middle interval[ x lo ,x hi ] inside, through the step coefficient Update the center, iterate until convergence; wherein, 、 、 The quadratic term coefficient, the first term coefficient and the constant of the general formula quadratic function of fitting, respectively; Step S614, fitting a symmetric model at the final center c with the formula: ； wherein , are the coefficients and the constant of the fitted vertexed quadratic function, respectively; Step S62, when the rolling groove surface is a flat surface, the radius profile does not have obvious valley bottom feature, and a flat section center strategy is adopted to determine the axial center, and a linear fitting is performed on the flat surface radius trend: Step S621, calculate the reference mean radius in the middle 2 / 5-3 / 5 interval and construct the residual sequence: ； wherein, R i the radius of gyration after polymerization; and r i the radius of gyration after polymerization; and Step S622, within the candidate center index range, a sliding window with a length of 2 / 3 of the effective bucket number is constructed, the residual sum in the window is calculated as a cost function; for the case of edge leading to insufficient window, the maximum residual is used for penalty to make up; the axial coordinate corresponding to the minimum position of the cost function is selected c new as the center of the tangent plane Step S623, writing back the axial component of the three-dimensional circle center, while outputting c new writing back the axial component of the three-dimensional circle center, while outputting as a reference mean radius; Step S624, filtering the point cloud The above is again performed with the bucketing method and a linear fit is performed on the reference mean radius versus axial coordinate relationship, with the following formula: ； wherein is the radius size corresponding to the axial coordinate x.

9. The roll groove polishing path planning method of claim 8, wherein, In step S7, the specific steps are as follows: Step S71: Three-point axial arrangement. Based on the current groove width parameters, calculate the half-width and use the center coordinates (X... C ,Y C Z C Based on the reference point, three target circles are arranged symmetrically along the axis: the center point is offset relative to the left and right points or the top and bottom points by half the width; according to the pressing depth of the grinding head... r d and the angle of the grinding head The coordinates of the three grinding contact points are obtained using the following formula: When it is a vertical groove Y-Z plane: ； When it is a horizontal groove X-Y plane, the Z axis is taken as the bucketing axial coordinate: ； wherein , is the axial coordinate x the corresponding distance of the polishing point to the center of the circle Step S72, tangent direction calculation, and the specific steps are as follows: When it is a vertical groove, the circle is located in the Y-Z plane, the tangent should be perpendicular to the normal and still located in the Y-Z plane, which is constructed from the normal component; When it is a horizontal groove, the circle is located in the X-Y plane, the tangent should be perpendicular to the normal and located in the X-Y plane, and the normal direction is selected according to the actual interference; Step S73, target point coordinate conversion, the three-point contact points in the local coordinate system are converted to the robot coordinate system, and the robot automatically performs linear Cartesian motion or arc Cartesian motion to carry out the polishing task after obtaining the three-point polishing points and the rolling groove surface type.

Images