A robot grinding method and system for local defects of blades

Abstract

The present invention belongs to the field of robot defect grinding, and discloses a robot automatic grinding method and system for local defects of blades. The method includes: calibrating the hand-eye relationship of the line laser measuring instrument; converting the measuring point cloud coordinate system by combining the hand-eye relationship calibration result, the measurement data of the line laser measuring instrument and the robot pose data; processing the measurement data to obtain feature information such as the defect position, depth, normal direction, etc.; and planning the robot grinding tool path and grinding contact force according to this information and the material removal model. The present invention uses the line laser measuring instrument and the robot to work together, combines the measurement data with the grinding process, improves the grinding efficiency and accuracy of local defects of the blades, and realizes the smooth transition between the grinding area and the surrounding area.

Description

Technical Field

[0001] The present invention belongs to, but is not limited to, the technical field of robot defect grinding, and particularly relates to a robot automatic grinding method and system for local defects of blades. Background Art

[0002] With the continuous development of aero-engine technology, the production and processing requirements of blades are gradually increasing. The blade surface structure is complex, the manufacturing cost is high, and it is in an environment of strong corrosion and high dynamic load for a long time, making it easy to generate various local defects. These defects are small in size but great in harm, and the grinding of local defects of blades is an urgent problem to be solved.

[0003] Currently, the grinding method of aero-engine blades is mainly completed by manual grinding. Manual grinding has a large labor intensity, low processing efficiency, a harsh working environment, and at the same time, the grinding accuracy is greatly affected by the technical ability of workers, and the material removal consistency is poor, seriously affecting the performance and life of the ground blades. Compared with the traditional grinding method, the robot grinding system has the advantages of good flexibility, high efficiency, and easy expansion. Therefore, it is necessary to study robot grinding technology to achieve efficient and controllable defect grinding.

[0004] In view of the above analysis, the technical problems that urgently need to be solved in the existing technology are as follows:

[0005] Traditional defect measurement methods mostly use point cloud registration methods to obtain the defective parts, which have problems of large computational amount and low accuracy. In the existing technology, the grinding of local defects of blades has not been well solved. Summary of the Invention

[0006] In view of the problems existing in the prior art, the present invention provides a robot automatic grinding method and system for local defects of blades.

[0007] The present invention is implemented as follows. A robot automatic grinding method for local defects of blades is characterized in that the method specifically includes:

[0008] S1: Use a robot to hold a line laser measuring instrument to scan the blade to be ground, and synchronously collect the robot pose data and the line laser three-dimensional point cloud data;

[0009] S2: Calibrate the hand-eye relationship of the line laser measuring instrument by using the method of multi-pose scanning of a standard sphere by the line laser measuring instrument, and convert the three-dimensional point cloud coordinate data collected by the line laser into the base coordinate system;

[0010] S3: Preprocess the measured point cloud by using voxel grid downsampling and statistical filtering methods to obtain a measured point cloud with uniform spatial distribution;

[0011] S4: Divide and analyze the point cloud of the blade body, implement adaptive threshold segmentation through the Otsu method, and obtain the point cloud of the defective part;

[0012] S5: Conduct a neighborhood search on the point cloud of the blade body at the defective point, obtain the neighborhood point cloud within a certain radius of the defective point, perform RANSAC plane fitting on the neighborhood point cloud, and calculate the depth of the defective point according to the distance between the point cloud at the defective point position and the fitting plane;

[0013] S6: Analyze the contact area of the spherical grinding head according to the Hertz contact theory and Preston equation, establish a grinding material removal model, and calibrate the material removal coefficient through experiments;

[0014] S7: Plan the spiral grinding trajectory and grinding contact force based on the contact area analysis, material removal model, and the measured depth of the defective point;

[0015] S8: Use the machine force control actuator and force control algorithm to achieve the planned contact force.

[0016] Furthermore, in S2, the hand-eye relationship of the line laser measuring instrument is calibrated by the method of multi-position scanning of the standard sphere using a line laser measuring instrument, an overdetermined linear equation set is established, and the hand-eye relationship matrix is solved using the least squares method:

[0017]

[0018] where s X, s Y, s Z respectively represent the coordinates of the center of the standard sphere in the line laser measurement coordinate system, R and t respectively represent the rotation matrix part and translation matrix part in the homogeneous coordinate transformation matrix, the left superscripts a, b respectively correspond to the robot end coordinate system and the robot base coordinate system, and the subscripts correspond to the serial numbers of the first laser measurement poses, Finally, it is combined into the hand-eye relationship matrix of the line laser measuring instrument;

[0019] The hand-eye relationship matrix is solved using the least squares method for the equation set, and finally, according to the tool coordinate system of the robot program The homogeneous transformation matrix calculated from the collected robot pose data and the hand-eye relationship matrix of the line laser instrument obtained through the hand-eye calibration experiment Convert the three-dimensional point cloud coordinate data collected by the line laser into the base coordinate system:

[0020]

[0021] Furthermore, in S3, the initial measurement point cloud is evenly divided into multiple voxel grids according to the spatial coordinates. Let the initial point cloud set be P = {p 1 , p2 ,..., p n}, divide the initial point cloud set into m voxel grids according to the spatial coordinates, and calculate the weighted average coordinates of the measurement point cloud in each voxel grid:

[0022]

[0023] Replace all the measurement point clouds in this voxel grid with the weighted average coordinates, and finally obtain the downsampled point cloud Q = {q 1 , q 2 ,..., q m}, downsample the initial measurement point cloud with the number of point clouds n to a uniform measurement point cloud with the number of point clouds m; perform neighborhood search on each measurement point and calculate the average distance from this measurement point to other points in its neighborhood:

[0024]

[0025] Assume that the obtained distance set D = {d 1 , d 2 , …, d m} is a Gaussian distribution, then its mean and standard deviation can be calculated:

[0026]

[0027] According to the need, a confidence interval (μ - λσ, μ + λσ) can be set. The measurement points falling within the confidence interval are valid points, and the measurement points outside the confidence interval are removed as outlier noise points.

[0028] Furthermore, the S4 includes:

[0029] (1) According to the spatial arrangement information contained in the row and column of the arrival point cloud measured by the line laser, perform grid segmentation on the blade body measurement point cloud. For each region obtained by the segmentation, calculate the diagonal center distance and use it as an evaluation index to identify the defective region. For the four corner points of each grid region:

[0030] A(x a , y a , z a ), B(x b , y b , z b ), C(x c , y c , z c ), D(x d , y d , z d )

[0031] Due to the spatial characteristics of the grid region, it can be considered that:

[0032] x a = x d , x b = x c , y a = y b , y c = y d ,

[0033] Therefore, taking the center distance of the regional diagonal as the evaluation index, we have

[0034] The center points of the diagonals are approximately coincident, V≈0. While in the defect area, the corner points of the area falling into the defect pit will have a drastic change in coordinates, and the center points of the diagonals cannot coincide, V≠0. Therefore, the defect position can be judged based on this;

[0035] (2) Use the Otsu method for adaptive threshold segmentation to divide the defect part and the non-defect part. Let the threshold t traverse all possible values. Through the threshold t, the point cloud is divided into the foreground point cloud family (V>t) and the background point cloud family (V≤t). Calculate the mean value μ 1 , μ 2 and the weight w 1 , w 2 , where the weight is the proportion of the number of point clouds in this point cloud family to the total number of point clouds. Calculate the between-class variance of the point cloud family:

[0036] σ 2 (t) = w 1 w 2 (μ 1 - μ 2 ) 2

[0037] The threshold t traverses all possible values, and the corresponding threshold t when the between-class variance is the largest is obtained max , which is the finally selected threshold. The area where the center distance of the regional diagonal is higher than the threshold is considered the defect area. The defect position point cloud is extracted through the point cloud index of the defect area. The center of the defect point cloud family can be considered the position of the defect point.

[0038] Furthermore, in step S5, according to the size of the defect points to be detected, extract the neighborhood point cloud within a suitable radius at the defect point position; perform RANSAC plane fitting on the neighborhood point cloud, and the normal vector of the fitted plane is considered the leaf normal direction at the defect point; calculate the distance from each point in the defect point cloud family to the fitted plane, and the longest distance is considered the depth of the defect point.

[0039] Furthermore, step S7 includes:

[0040] (1) Contact area analysis: According to Hertz contact theory, analyze the contact area and the distribution of contact stress within the contact area during the grinding head dressing process. The contact area can be calculated using the following formula:

[0041] S = πa s b s

[0042] a s ,b s are the major semi-axis length and minor semi-axis length of the elliptical contact area respectively, which can be calculated based on parameters such as contact pressure. Usually, in the elliptical contact area, the maximum principal stress occurs at the center of the ellipse. The maximum principal stress can be calculated using the following formula:

[0043]

[0044] k is the ratio of the major and minor semi-axis lengths of the elliptical contact area, k' is the reciprocal of k, Δ is calculated from the elastic moduli and Poisson's ratios of the two contact surfaces respectively, Z represents the distance of the stress position from the contact plane (along the normal direction of the contact plane), E is the first kind of elliptic integral of the elliptical contact area. Within the entire elliptical contact area, the contact stress distribution can be calculated using the following formula:

[0045]

[0046] (2) Material removal analysis: Establish a material removal model based on Hertz contact theory and Preston equation. The material removal model is:

[0047]

[0048] where Μ is the material removal amount over the entire contact area, k w is a coefficient, affected by Preston coefficient, surface hardness, and material density. For a specific dressing tool and blade, the coefficient k w is a constant, v is the sliding speed, t is the dwell time, F is the dressing pressure. By calibrating the coefficient K through experiments, the material removal equation can be established:

[0049]

[0050] (3) Plan the dressing contact force: Plan a spiral dressing tool path at the defect point, add variable force control of the dressing contact force. As the spiral trajectory expands from the inside outwards, control the dressing pressure to gradually decrease, so as to obtain an increasingly smaller material removal amount to ensure a smooth transition of the surface curvature of the blade after dressing.

[0051] Another object of the present invention is to provide a robot automatic dressing system for local defects of blades. This system specifically includes:

[0052] An input module for collecting robot pose data and line laser three-dimensional point cloud data;

[0053] A calibration module for calibrating the hand-eye relationship of the line laser measuring instrument;

[0054] A point cloud processing module for preprocessing the measured point cloud;

[0055] A defect detection module for extracting defect position point clouds;

[0056] A material removal module for establishing a grinding material removal model and calibrating the material removal coefficient;

[0057] A contact force planning module for planning a helical grinding trajectory and a grinding contact force.

[0058] Combining the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by the present invention are as follows:

[0059] First, in view of the technical problems existing in the above-mentioned prior art and the difficulty of solving the problems, closely combining the technical solution to be protected by the present invention and the results and data in the R & D process, etc., analyze in detail and profoundly how the technical solution of the present invention solves the technical problems and the creative technical effects brought after solving the problems. The specific description is as follows:

[0060] The defect point measurement and grinding methods provided by the present invention both work by using the method of clamping the end effector by the robot, without the need to replace the part clamping, which is beneficial to improving the efficiency and realizing automation. Through the calibration of the hand-eye relationship of the line laser measuring instrument and the coordinate transformation of the measured point cloud combined with the robot pose, the cumbersome robot tool coordinate and workpiece coordinate calibration work are omitted, with high automation and meeting the requirements of high efficiency and high adaptability.

[0061] The technical solution of the present invention solves the problems of low measurement and positioning accuracy of micro defect points and difficult accurate repair in the prior art in industrial applications. The discovery and grinding of defects on the blade surface are usually completed manually by workers, with low efficiency, high cost, poor accuracy and consistency. Traditional defect measurement and positioning methods are usually realized through point cloud analysis and registration methods, which not only have poor accuracy when measuring and positioning micro defect points, but also are difficult to adapt to different types of blades and cannot meet the precise positioning and feature extraction requirements for subsequent grinding.

[0062] The present invention improves the efficiency and accuracy of defect location detection through the evaluation index of the center distance of the regional diagonal line and the adaptive threshold segmentation method of the maximum inter-class variance, combined with the flexible characteristics of the robot's motion space and the characteristics of the point cloud measured by the line laser measuring instrument. Since the robot's repeated positioning accuracy is higher than the absolute positioning accuracy, the robot successively holds different end effectors to measure and grind the blade, improving the grinding accuracy of the defect points.

[0063] The present invention combines the analysis of the grinding contact area and the analysis of material removal, and plans the grinding tool path and the changing grinding contact force according to the position, normal line, curvature and depth information measured at the defect points, effectively ensuring the smooth transition between the defect area and the surrounding area after grinding.

[0064] Second, as the creative auxiliary evidence of the claims of the present invention, it is also reflected in the following important aspects:

[0065] (1) The expected benefits and commercial values after the transformation of the technical solution of the present invention are:

[0066] Improve the efficiency and accuracy of the repair of local defects on the blade, meet the different defect repair requirements of various blades of aero-engines, and thus bring higher production benefits to relevant enterprises. Promote the combination of robot grinding and polishing technology and optical measurement technology, give full play to the advantages of the flexible movement of the robot, and open up a broader market. The method for repairing local defects on the blade based on the present invention can reduce costs, improve efficiency, enhance product competitiveness, and bring considerable economic benefits to enterprises.

[0067] (2) The technical solution of the present invention fills the technical gaps in the domestic and international industries:

[0068] The existing blade repair methods mostly weld the blade defects, and the measurement data is mainly used to guide the blade profile after welding, while the fixed-point grinding for local defects is rarely. The present invention designs a repair process combining measurement and grinding for local defects on the blade, filling this technical gap and providing a new robot grinding method for the measurement and repair of local defects on the blade.

[0069] (3) The technical solution of the present invention solves the technical problems that people have been eager to solve but have never succeeded in:

[0070] Traditional defect measurement methods mostly use point cloud registration methods to obtain the defective part, which has the problems of large computational amount and low accuracy. While this patent uses the evaluation index of the center distance of the regional diagonal line and the adaptive threshold segmentation method to detect the defect position, and uses the method of combining the line laser measuring instrument and the robot pose data to obtain the blade point cloud in the robot base coordinate system, improving the computational efficiency and measurement accuracy, and solving this technical problem.

[0071] In the prior art, the grinding of local defects on the blade has not been well solved. By combining defect measurement and grinding, and planning the grinding tool path and grinding contact force according to the measurement data, the present invention realizes a smooth transition between the ground defect area and the surrounding area, thus solving this technical problem.

[0072] (4) The technical solution of the present invention overcomes the technical prejudice:

[0073] In the previous technical concepts, more attention was often paid to the overall registration of the blade measurement point cloud, while ignoring the characteristics of the line laser measurement point cloud itself and the advantages of the combination of the robot and the measurement; the blade measurement and grinding were often considered separately, ignoring the direct guiding role of the line laser measurement point cloud in blade grinding. The present invention overcomes this technical prejudice, gives full play to the advantages of the combination of the flexible movement ability of the robot and the accurate measurement data of the line laser measuring instrument, measures the blade body by the line laser, calculates the position and characteristic information of the defective part, and plans the grinding parameters accordingly, realizing the automatic, efficient and high-precision grinding of local defects on the blade.

[0074] Through these improvements, the present invention significantly improves the grinding accuracy and efficiency of local defects on the blade, realizes a higher degree of automation in defect repair, and solves the problems of low efficiency and accuracy in traditional repair methods. In addition, by combining defect point measurement and grinding, the grinding accuracy is effectively improved and a smooth transition between the ground area and the surrounding area is realized, thus promoting the technological progress of advanced manufacturing technology in the field of blade repair.

[0075] Third, the technical solution of the present invention solves several key problems existing in the prior art in industrial applications and has made remarkable technical progress:

[0076] 1) Existing technical problems: The accuracy and consistency of blade defect repair are insufficient.

[0077] Traditional blade repair methods usually rely on manual operations, and the repair accuracy is limited by the technical level and experience of the operator. At the same time, it is difficult to ensure the consistency of the repair by manual operation, especially in mass production, and it is impossible to ensure that each blade can achieve the same repair effect. This inconsistency will affect the overall performance and service life of the blade.

[0078] The present invention introduces robot automatic grinding technology, accurately collects the three-dimensional point cloud data of the blade surface by using a line laser measuring instrument, and performs defect identification and repair path planning through precise algorithms. The robot executes force-controlled grinding operations, realizing high-precision repair of surface defects on the blade. This automation technology ensures the consistency of the repair and significantly improves the quality and reliability of blade repair.

[0079] 2) Existing technical problems: The processing efficiency of complex surface point cloud data is low.

[0080] On the surface of complex blades, traditional point cloud data processing methods are often inefficient, prone to generating noise and redundant data, which affect the accuracy of defect detection. This inefficient data processing method makes the repair process time-consuming and laborious, and it is difficult to meet the high-efficiency requirements of industrial production.

[0081] The present invention adopts advanced voxel grid downsampling and statistical filtering methods to efficiently preprocess the collected point cloud data, significantly reducing data redundancy and improving the processing efficiency of the point cloud. Through region division and adaptive threshold segmentation, the defect area is accurately identified, and the RANSAC plane fitting technology is used to accurately calculate the defect depth. These technological innovations improve the speed and accuracy of point cloud data processing, providing reliable data support for the repair operation.

[0082] 3) Problems of the prior art: The difficult problem of contact force control during the grinding process.

[0083] During the blade grinding process, precise contact force control is the key to ensuring uniform material removal and repair quality. However, traditional methods are difficult to achieve precise control of the contact force, which easily leads to excessive or insufficient material removal, thereby affecting the performance and lifespan of the blade.

[0084] The present invention realizes precise control of the contact force during the grinding process through a material removal model based on Hertz contact theory and Preston equation, combined with a force control actuator of the robot. Through precise planning of the grinding trajectory and contact force, the uniformity of material removal during the grinding process is ensured. This advancement greatly improves the reliability of the repair process and guarantees the repair effect and long-term use performance of the blade.

[0085] 4) Problems of the prior art: Low automation level and insufficient production efficiency.

[0086] In the traditional blade repair process, the automation level is low and it relies on manual operation, which is not only inefficient but also prone to human errors, and cannot meet the requirements of modern industry for high-efficiency and intelligent production.

[0087] The present invention realizes the whole process of blade defect repair through comprehensive automation technology. The robot system can independently complete defect detection, repair path planning, and grinding operations, greatly improving production efficiency, reducing manual intervention, and reducing operation errors. This high level of automation technology not only improves the efficiency of the production line but also ensures the consistency of product quality, providing strong technical support for industrial applications.

[0088] In summary, through technological innovation, the present invention has made remarkable progress in aspects such as the precision, consistency, data processing efficiency, and automation level of blade defect repair. Its successful implementation in industrial applications not only solves many problems in the prior art but also provides an efficient and reliable repair solution for modern manufacturing, greatly improving the quality and production efficiency of blade products. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 is a flowchart of the robot automatic grinding method for local blade defects provided by an embodiment of the present invention;

[0090] Figure 2 is a detailed flowchart of the robot automatic grinding method for local blade defects provided by an embodiment of the present invention;

[0091] Figure 3 is a flowchart of the measurement of blade defect points provided by an embodiment of the present invention;

[0092] Figure 4 is a flowchart of the grinding of blade defect points provided by an embodiment of the present invention;

[0093] Figure 5 is a schematic diagram of the coordinate system relationship between the robot and the measuring tool provided by an embodiment of the present invention;

[0094] Figure 6 is a schematic diagram of the adaptive threshold segmentation of the maximum inter-class variance provided by an embodiment of the present invention;

[0095] Figure 7 is a module diagram of the robot automatic grinding system for local blade defects provided by an embodiment of the present invention;

[0096] Figure 8 is a schematic diagram of the material removal simulation of the spiral grinding trajectory provided by an embodiment of the present invention;

[0097] Figure 9 is a schematic diagram of the surface topography of the blade body in the defect area before and after grinding provided by an embodiment of the present invention;

[0098] Figure 10 is a schematic diagram of the change amount of the surface normal of the blade body in the defect area before and after grinding provided by an embodiment of the present invention;

[0099] Figure 11 is a comparison diagram of the surface of the blade body in the defect area before and after grinding provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0100] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0101] As Figure 1 shown, an embodiment of the present invention provides a robot automatic grinding method for local defects of blades, and the method specifically includes:

[0102] S1: Use a robot to hold a line laser measuring instrument to scan the blade to be ground, and synchronously collect the pose data of the robot and the three-dimensional point cloud data of the line laser;

[0103] S2: Calibrate the hand-eye relationship of the line laser measuring instrument by using the method of multi-pose scanning of a standard sphere with the line laser measuring instrument, and convert the three-dimensional point cloud coordinate data collected by the line laser into the base coordinate system;

[0104] S3: Preprocess the measured point cloud by using voxel grid downsampling and statistical filtering methods to obtain a spatially uniformly distributed measured point cloud;

[0105] S4: Divide and analyze the point cloud of the blade body, and achieve adaptive threshold segmentation through the maximum inter-class variance method to obtain the point cloud of the defective part;

[0106] S5: Perform a neighborhood search on the point cloud of the blade body at the defective point to obtain the neighborhood point cloud within a certain radius of the defective point, perform RANSAC plane fitting on the neighborhood point cloud, and calculate the depth of the defective point according to the distance between the point cloud at the defective point position and the fitting plane;

[0107] S6: Analyze the contact area of the spherical grinding head according to Hertz contact theory and Preston equation, establish a grinding material removal model, and calibrate the material removal coefficient through experiments;

[0108] S7: Plan the spiral grinding trajectory and grinding contact force based on the contact area analysis, material removal model and the measured depth of the defective point;

[0109] S8: Use a robot force control actuator and a force control algorithm to achieve the planned contact force.

[0110] The robot automatic grinding method for local defects of blades provided by the embodiment of the present invention mainly realizes the precise repair of the blade surface through multiple steps. First, in step S1, the robot holds a line laser measuring instrument to scan the blade to be ground. During this process, the system synchronously collects the pose data of the robot and the three-dimensional point cloud data generated by the line laser to obtain the precise topography of the blade surface. These data provide high-precision basic information for subsequent grinding operations.

[0111] In step S2, the hand-eye relationship of the line laser measuring instrument is calibrated by using the multi-position scanning method of the standard sphere. This calibration process converts the three-dimensional point cloud coordinate data collected by the line laser measuring instrument into the base coordinate system, making the point cloud data consistent with the coordinate system of the robot operating system. This conversion ensures data accuracy and operation precision in subsequent operations.

[0112] Next, in steps S3 to S5, the collected three-dimensional point cloud data is preprocessed. First, the voxel grid downsampling and statistical filtering methods are used to process the measured point cloud data to obtain a spatially uniformly distributed point cloud. Then, through the regional division and analysis of the blade point cloud, the Otsu method (maximum inter-class variance method) is used to achieve adaptive threshold segmentation, thereby identifying and extracting the defect area on the blade surface. Finally, a neighborhood search is performed on the point cloud of the defect area, and through the RANSAC plane fitting technique, the distance from the defect point to the fitted plane is calculated to accurately measure the defect depth.

[0113] In steps S6 to S8, based on the Hertz contact theory and the Preston equation, the material removal characteristics of the contact area between the spherical grinding head and the blade are analyzed, and a grinding material removal model is established. The material removal coefficient is calibrated through experiments, and combined with the measured defect depth, the spiral trajectory of grinding and the corresponding grinding contact force are planned. Finally, the robot uses a force control actuator and an accurate force control algorithm to perform the grinding operation according to the planned contact force, ensuring stable material removal during grinding and successfully repairing the local defects on the blade surface. Through this series of operations, the system can efficiently and accurately repair the blade surface defects under automated conditions, improving the service life and performance of the blade.

[0114] As Figure 2 shown, the method includes two major parts: defect measurement and process parameter planning. The flowcharts of blade defect point measurement and blade defect point grinding are respectively as Figure 3 、 Figure 4 shown.

[0115] In S2, the valid data in the line laser scan data is extracted, that is, the point cloud data belonging to the standard sphere. A certain circular cross-section of the standard sphere is fitted according to the valid data, and based on the relationship between the diameter of the cross-section circle and the diameter of the standard sphere, the coordinates of the center of the standard sphere relative to the line laser measurement coordinate system are obtained.

[0116] Figure 5 represents the coordinate system relationship between the robot and the measuring tool during the line laser hand-eye relationship calibration in the present invention. The robot base coordinate system O b -X b Y b Z bLocated at the robot base, the Z-axis is vertically upward along the first axis of the robot. The position and direction of the coordinate system do not change with the robot's pose and can be considered as an absolute coordinate system; the coordinate system at the end of the robot is O a -X a Y a Z a Located at the center of the flange at the end of the robot, the Z-axis is perpendicular to the flange plane and is the installation coordinate system of the line laser measuring instrument; the line laser measuring coordinate system is O s -X s Y s Z s Located at the origin of the measuring plane of the line laser measuring instrument, the Z-axis is along the direction of the line laser irradiation.

[0117] Calibrate the hand-eye relationship of the line laser measuring instrument by using the method of multi-pose scanning of the standard sphere with the line laser measuring instrument, establish an overdetermined linear equation system, and solve the hand-eye relationship matrix using the least squares method:

[0118]

[0119] where s X, s Y, s Z respectively represent the coordinates of the center of the standard sphere in the line laser measuring coordinate system, R and t respectively represent the rotation matrix part and the translation matrix part in the homogeneous coordinate transformation matrix, the left subscripts a, b respectively correspond to the coordinate system at the end of the robot and the base coordinate system of the robot, and the subscript at the lower right corresponds to the sequence number of the pose of the first laser measurement. Finally, it is combined into the hand-eye relationship matrix of the line laser measuring instrument.

[0120] Solve the equation system using the least squares method to obtain the hand-eye relationship matrix. Finally, according to the tool coordinate system of the robot program The homogeneous transformation matrix calculated from the pose data of the robot collected and the hand-eye relationship matrix of the line laser instrument obtained through the hand-eye calibration experiment Convert the three-dimensional point cloud coordinate data collected by the line laser into the base coordinate system:

[0121]

[0122] In step S3, through the voxel grid downsampling and statistical filtering methods, make the distribution of the measurement point cloud uniform and the density appropriate, and remove the edge noise points that affect the subsequent defect point judgment.

[0123] Evenly divide the initial measurement point cloud into multiple voxel grids according to the spatial coordinates. Let the initial point cloud set be P = {p 1 , p 2 , …, p n}, the initial point cloud set is divided into m voxel grids according to spatial coordinates, and the weighted average coordinates of the measurement point cloud in each voxel grid are calculated:

[0124]

[0125] All the measurement point clouds within this voxel grid are replaced with the weighted average coordinates, and finally the downsampled point cloud Q = {q 1 , q 2 , …, q m} is obtained, and the initial measurement point cloud with the number of point clouds n is downsampled into a uniform measurement point cloud with the number of point clouds m.

[0126] Neighborhood search is performed for each measurement point, and the average distance from this measurement point to other points within its neighborhood is calculated:

[0127]

[0128] Assume that the obtained distance set D = {d 1 , d 2 , …, d m} is a Gaussian distribution, then its mean and standard deviation can be calculated:

[0129]

[0130] According to the need, a confidence interval (μ - λσ, μ + λσ) can be set. The measurement points falling within the confidence interval are valid points, and the measurement points outside the confidence interval are removed as outlier noise points.

[0131] The said S4 includes:

[0132] (1) Since the line laser measurement data is obtained by the measurement line sweeping across the blade body, the horizontal and vertical coordinate indices of the line laser measurement point cloud contain certain spatial information, that is, the horizontal coordinate index represents the measurement points arranged from left to right on the measurement line during a single measurement; while the vertical coordinate index represents different measurement times.

[0133] According to the spatial arrangement information contained in the row and column of the arrival point cloud of the line laser measurement, the blade body measurement point cloud is segmented into grids. In order to avoid the influence of point cloud tilt and the blade edge part during defect detection, for each region obtained by segmentation, the diagonal center distance is calculated and used as an evaluation index to identify the defect region.

[0134] For the four corner points of each grid region:

[0135] A(x a , y a , z a ), B(x b , y b , zb ), C(x c , y c , z c ), D(x d , y d , z d )

[0136] Due to the spatial characteristics of the grid area, it can be considered that:

[0137] x a = x d , x b = x c , y a = y b , y c = y d ,

[0138] Therefore, taking the center distance of the diagonal of the area as the evaluation index, there is

[0139] In the non-defective area, the curvature change of each area is small. The inclination of the blade surface will cause the coordinate changes of two corner points on one side of the area at the same time, which will not affect the center distance of the diagonal. Therefore, the center points of the diagonal are approximately coincident, V≈0; while in the defective area, the corner points of the area falling into the defective pit will have drastic coordinate changes, and the center points of the diagonal cannot coincide, V≠0. Therefore, the defective position can be judged based on this.

[0140] (2) Use the Otsu method for adaptive threshold segmentation to divide the defective part and the non-defective part.

[0141] The Otsu method is a commonly used means for image segmentation and can still be used in the work of threshold segmentation of point clouds. Let the threshold t traverse all possible values. The point cloud is divided into the foreground point cloud family (V>t) and the background point cloud family (V≤t) through the threshold t. Calculate the mean value μ 1 , μ 2 and the weight w 1 , w 2 , where the weight is the proportion of the number of point clouds in this point cloud family to the total number of point clouds. Calculate the between-class variance of the point cloud family:

[0142] σ 2 (t) = w 1 w 2 (μ 1 - μ 2 ) 2

[0143] The threshold t traverses all possible values, and the threshold t corresponding to the maximum between-class variance is obtained max , which is the final threshold taken.

[0144] Figure 6 Adaptive threshold segmentation representing the maximum between-class variance. The x-y coordinates represent the horizontal and vertical coordinate indices of the line laser measurement point cloud respectively, and the z coordinate represents the evaluation index of the center distance of the regional diagonal. It can be seen that the obvious three peaks are the defect regions.

[0145] Finally, the regions where the center distance of the regional diagonal is higher than the threshold are considered defect regions. The defect position point cloud is extracted through the point cloud index of the defect region, and the center of the defect point cloud cluster can be considered as the position of the defect point.

[0146] In step S5, according to the size of the defect points to be detected, the neighborhood point cloud within an appropriate radius at the defect point position is extracted. RANSAC plane fitting is performed on the neighborhood point cloud, and the normal vector of the fitted plane is considered as the blade normal direction at the defect point; the distance from each point within the defect point cloud cluster to the fitted plane is calculated, and the longest distance is considered as the depth of the defect point.

[0147] Step S7 includes:

[0148] (1) Contact area analysis:

[0149] According to Hertz contact theory, the contact area and the contact stress distribution within the contact area during the grinding head dressing process are analyzed.

[0150] The contact area can be calculated using the following formula:

[0151] S = πa s b s

[0152] a s and b s are respectively the major semi-axis length and minor semi-axis length of the elliptical contact area, which can be calculated based on parameters such as the contact pressure.

[0153] Generally, in the elliptical contact area, the maximum principal stress occurs at the center of the ellipse, and the maximum principal stress can be calculated using the following formula:

[0154]

[0155] k is the ratio of the major and minor semi-axis lengths of the elliptical contact area, k' is the reciprocal of k, Δ is calculated from the elastic moduli and Poisson's ratios of the two contact surfaces respectively, Z represents the distance from the stress position to the contact plane (along the normal direction of the contact plane), and E is the first kind of elliptic integral of the elliptical contact area.

[0156] Within the entire elliptical contact area, the contact stress distribution can be calculated using the following formula:

[0157]

[0158] (2) Material removal analysis:

[0159] Based on the Hertz contact theory and the Preston equation, a material removal model is established. The material removal model is:

[0160]

[0161] where Μ is the material removal amount on the entire contact area, and k w is a coefficient, affected by the Preston coefficient, surface hardness, and material density. For a specific grinding tool and blade, the coefficient k w is a constant. v is the sliding speed, t is the dwell time, and F is the grinding pressure.

[0162] The coefficient K is a constant in the scenario of determining the grinding tool and blade materials. Therefore, the coefficient K can be calibrated through experiments, and the material removal equation can be established:

[0163]

[0164] (3) Planning the grinding contact force

[0165] A spiral grinding tool path is planned at the defect point. However, when using the constant force control of the spiral trajectory, there is still a problem of uneven curvature transition at the edge of the spiral trajectory after grinding. Therefore, a variable force control of the grinding contact force is added. As the spiral trajectory expands from the inside out, the grinding pressure is controlled to gradually decrease to obtain a smaller and smaller material removal amount, so as to ensure a smooth transition of the surface curvature of the blade after grinding.

[0166] As Figure 7 shown, a robot automatic grinding system for local defects of blades provided by an embodiment of the present invention specifically includes:

[0167] An input module for collecting robot pose data and line laser three-dimensional point cloud data;

[0168] A calibration module for calibrating the hand-eye relationship of the line laser measuring instrument;

[0169] A point cloud processing module for preprocessing the measured point cloud;

[0170] A defect detection module for extracting defect position point clouds;

[0171] A material removal module for establishing a grinding material removal model and calibrating the material removal coefficient;

[0172] A contact force planning module for planning a spiral grinding trajectory and a grinding contact force.

[0173] I. The specific application fields or related products of the present invention.

[0174] Application Example 1: Grinding of Corrosion Pits on the Surface of Aeroengine Blades

[0175] Aeroengine blades are in an environment of strong corrosion, high temperature and pressure, and high dynamic load for a long time. In this case, local defects such as notches, pits, pitting corrosion, scratches, breakages, and tears often appear on the blade surface. These defects have a great impact on the performance and safety of aeroengines. Most of these defects are corrosion pits on the blade surface. These pits are small in size and are not easy to detect and repair using traditional methods. By using the robot measurement and grinding method of the present invention, these corrosion pits can be effectively detected and repaired.

[0176] The specific implementation process is as follows:

[0177] Calibrate the hand-eye relationship of the line laser measuring instrument using a standard ball with a known diameter; fix the blade in the robot's motion space; use the line laser measuring instrument to scan the blade and synchronously collect the robot pose information corresponding to the measured point cloud; convert the measured point cloud coordinate system to the robot base coordinate by combining the hand-eye relationship matrix of the line laser measuring instrument, the line laser measurement data, and the robot pose data; process the data using the method described in this patent to obtain feature information such as the position, depth, and normal of the defect points; plan the grinding tool path and grinding contact force according to the information obtained from the measurement data, and use the robot to hold the end force control actuator to repair the defects. Compared with the traditional processing method, the robot grinding method of the present invention can significantly improve the defect detection and repair accuracy, while improving the efficiency and reducing the production cost.

[0178] Second, evidence related to the technical effects obtained in the embodiments of the present invention.

[0179] Figure 8 The grinding effects of traditional constant force control and the contact force control planned by the present invention are compared. As shown in Figure (a), only using the Archimedean spiral, there is still a problem of uneven curvature transition at the edge of the spiral trajectory after grinding. As shown in Figure (b), as the spiral trajectory expands from the inside to the outside, the grinding pressure is gradually reduced to obtain a smaller and smaller material removal amount to ensure the smooth transition of the curvature of the blade surface after grinding.

[0180] It can be seen that the constant force control and the variable force control achieve similar material removal depths at the center of the defect points. In the transition region between the defect region and the adjacent region, the material removal depth change obtained by the variable force control is much smoother than that of the constant force control, which can effectively solve the problem of uneven curvature transition.

[0181] Use the line laser to measure the blade body after grinding again, and use the aforementioned coordinate conversion method of the blade body point cloud to convert the ground point cloud to the same coordinate system as the point cloud before grinding. Extract the neighborhood point cloud of the defect area for comparative analysis to verify the effectiveness of the planned contact force and tool path in the local defect grinding of the blade.

[0182] Figure 9 The distance changes of the defect area and the neighborhood point cloud from the neighborhood fitting plane before and after grinding are compared. As shown in Figure (a), there are obvious pit defects in the defect area before grinding. As shown in Figure (b), the defect area is relatively flat after grinding, without obvious pits.

[0183] Figure 10 The normal change rates of the defect area and the neighborhood point cloud before and after grinding are compared. As shown in Figure (a), there are obvious mutations in the normal of the defect area before grinding. As shown in Figure (b), the normal change of the defect area is gentle after grinding, without obvious mutations. Compared with before grinding, the depth change on the blade surface is relatively smooth after grinding, the depth changes regularly with the blade curvature, without obvious irregular depressions; the normal change rate of the blade surface tends to be gentle, and the characteristic of the drastic change in the normal at the defect position before grinding disappears.

[0184] Figure 11 The defect area and the adjacent area before and after grinding are compared. As shown in Figure (a), obvious defect pits can be seen in the defect area before grinding. As shown in Figure (b), there are no obvious pits in the defect area after grinding, and the ground part transitions smoothly with the surrounding area, without obvious curvature mutations.

[0185] It should be noted that the implementation mode of the present invention can be realized by hardware, software, or a combination of software and hardware. The hardware part can be realized by using special logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or special designed hardware. Those of ordinary skill in the art can understand that the above devices and methods can be realized by using computer-executable instructions and / or included in the processor control code, for example, such code is provided on a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be realized by the hardware circuit of programmable hardware devices such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable logic devices such as field programmable gate arrays, or can be realized by software executed by various types of processors, or can be realized by a combination of the above hardware circuits and software such as firmware.

[0186] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be covered by the protection scope of the present invention.

Claims

1. A robot automatic grinding method for local defects of blades, characterized in that: The method specifically includes: S1: Use the robot to hold the line laser measuring instrument to scan the blade to be ground, and simultaneously collect the robot posture data and line laser 3D point cloud data; S2: Use the method of multi-pose scanning of the standard ball by the line laser measuring instrument to calibrate the hand-eye relationship of the line laser measuring instrument, and transform the three-dimensional point cloud coordinate data collected by the line laser into the base coordinate system; S3: Use voxel grid downsampling and statistical filtering methods to preprocess the measured point cloud to obtain a spatially uniformly distributed measured point cloud; S4: The blade body point cloud is divided and analyzed, and adaptive threshold segmentation is achieved through the maximum inter-class variance method to obtain the defective part point cloud; S5: Perform neighborhood search on the blade body point cloud at the defect point to obtain the neighborhood point cloud within a certain radius of the defect point, perform RANSAC plane fitting on the neighborhood point cloud, and calculate the depth of the defect point according to the distance between the point cloud at the defect point position and the fitting plane; S6: Analyze the contact area of ​​the spherical grinding head according to Hertz contact theory and Preston equation, establish a grinding material removal model, and calibrate the material removal coefficient through experiments; S7: Planning the spiral grinding trajectory and grinding contact force based on contact area analysis, material removal model and measured defect point depth; S8: Implementing planned contact forces using robotic force-controlled actuators and force control algorithms; S2, using the method of multi-position scanning standard ball of the line laser measuring instrument to calibrate the hand-eye relationship of the line laser measuring instrument, establish an overdetermined linear equation group, and use the least squares method to solve the hand-eye relationship matrix: in s X, s Y, s Z represents the coordinates of the center of the standard ball in the line laser measurement coordinate system, R and t represent the rotation matrix part and the translation matrix part in the homogeneous coordinate transformation matrix, and the subscripts a and b correspond to the robot end coordinate system and the robot base coordinate system, respectively. The final combination is the hand-eye relationship matrix of the line laser measuring instrument; The least squares method is used to solve the equations to obtain the hand-eye relationship matrix, and finally the tool coordinate system of the robot program is Homogeneous transformation matrix calculated from collected robot pose data And the hand-eye relationship matrix of the line laser instrument obtained through the hand-eye calibration experiment Convert the 3D point cloud coordinate data collected by the line laser into the base coordinate system:

2. The robot automatic grinding method for local defects of blades as claimed in claim 1, characterized in that: In step S3, the initial measured point cloud is evenly divided into a plurality of voxel grids according to the spatial coordinates. Assume that the initial point cloud set is P = {p1, p2, ... p n }, divide the initial point cloud set into m voxel grids according to spatial coordinates, and calculate the weighted average coordinates of the measured point cloud in each voxel grid: The weighted average coordinates are used to replace all the measured point clouds in this voxel grid, and finally the downsampled point cloud Q = {q1, q2, ..., q m }, downsample the initial measured point cloud containing n points to a uniform measured point cloud containing m points; perform a k-neighborhood search for each measured point, and calculate the average distance from the measured point to other points in its k-neighborhood: Assume that the distance set D = {d1, d2, ..., d m } is a Gaussian distribution, its mean and standard deviation can be calculated: As needed, a confidence interval (μ-λσ, μ+λσ) is set, and the measurement points falling into the confidence interval are valid points, and the measurement points outside the confidence interval are removed as outlier noise points.

3. The robot automatic grinding method for blade local defects according to claim 1 is characterized in that: The S4 comprises: (1) Based on the spatial arrangement information contained in the rows and columns of the arrival point cloud measured by the line laser, the blade body measurement point cloud is grid-segmented. For each segmented area, the diagonal center distance is calculated and used as an evaluation index to identify the defective area. For the four corner points of each grid area: A(x a ,y a ,z a ),B(x b ,y b ,z b ),C(x c ,y c ,z c ),D(x d ,y d ,z d ) Due to the spatial characteristics of the grid area, it is considered that: x a =x d ,x b =x c ,and a =and b ,and c =and d , Therefore, the regional diagonal center distance is used as the evaluation index, The center points of the diagonals are approximately coincident, V≈0, while in the defect area, the corner points of the area falling into the defect pit will experience a dramatic change in coordinates, and the center points of the diagonals cannot coincide, V≠0, so the defect location can be determined based on this; (2) Use the maximum inter-class variance method to perform adaptive threshold segmentation to divide the defect part and the non-defect part. Set the threshold t to traverse all possible values. Use the threshold t to divide the point cloud into the foreground point cloud family (V>t) and the background point cloud family (V≤t). Calculate the V value mean μ1, μ2 and weight w1, w2 of the foreground point cloud family and the background point cloud family respectively. The weight is the proportion of the number of point clouds in the point cloud family to the total number of point clouds. Calculate the inter-class variance of the point cloud family: s 2 (t)=w1w2(μ1-μ2) 2 The threshold t traverses all possible values ​​and obtains the threshold t corresponding to the maximum inter-class variance. max , which is the final threshold. The area whose diagonal center distance is higher than the threshold is considered to be the defect area. The defect position point cloud is extracted through the point cloud index of the defect area. The center of the defect point cloud family is the position of the defect point.

4. The robot automatic grinding method for blade local defects according to claim 1 is characterized in that: The S5 extracts the neighborhood point cloud within a suitable radius at the defect point location according to the size of the defect point to be detected; performs RANSAC plane fitting on the neighborhood point cloud, and the normal vector of the fitting plane is considered to be the normal direction of the blade at the defect point; calculates the distance from each point in the defect point cloud family to the fitting plane, and the longest distance is considered to be the depth of the defect point.

5. The robot automatic grinding method for blade local defects as claimed in claim 1, characterized in that: The S7 comprises: (1) Contact area analysis: According to the Hertz contact theory, the contact area during the grinding process and the contact stress distribution in the contact area are analyzed. The contact area can be calculated using the following formula: S=πab a and b are the major and minor semi-axis lengths of the elliptical contact area, respectively, which are calculated based on the contact pressure. Usually, in the elliptical contact area, the maximum principal stress occurs at the center of the ellipse. The maximum principal stress can be calculated using the following formula: k is the ratio of the length of the major and minor semi-axes of the elliptical contact area, k' is the reciprocal of k, Δ is calculated by the elastic modulus and Poisson's ratio of the two contact surfaces, Z represents the distance between the stress position and the contact plane (along the normal direction of the contact plane), and E is the first kind of elliptical integral of the elliptical contact area. In the entire elliptical contact area, the contact stress distribution can be calculated by the following formula: (2) Material removal analysis: A material removal model is established based on Hertz contact theory and Preston equation. The material removal model is: Where M is the material removal amount on the entire contact area, k w is a coefficient, which is affected by Preston coefficient, surface hardness and material density. For specific grinding tools and blades, the coefficient k w is a constant, v is the sliding speed, t is the dwell time, and F is the grinding pressure. The coefficient K is calibrated through experiments to establish the material removal equation: (3) Planning the grinding contact force: A spiral grinding tool path is planned at the defect point, and variable force control of the grinding contact force is added. As the spiral trajectory expands from the inside to the outside, the grinding pressure is controlled to gradually decrease to obtain a smaller and smaller amount of material removal to ensure a smooth transition of the blade surface curvature after grinding.

6. A robot automatic grinding system for blade local defects according to claims 1-5, characterized in that: The system specifically includes: Input module, used to collect robot posture data and line laser 3D point cloud data; Calibration module, used to calibrate the hand-eye relationship of the line laser measuring instrument; Point cloud processing module, used for preprocessing the measured point cloud; Defect detection module, used to extract defect location point cloud; Material removal module, used to establish grinding material removal model and calibrate material removal coefficient; The contact force planning module is used to plan the spiral grinding trajectory and grinding contact force.

Images