A surface processing method for industrial robots based on region division

Abstract

The present invention discloses an industrial robot surface processing method based on regional division. By constructing a robot stiffness model, a regional stiffness index is proposed; the influence of the robot stiffness performance is added to the division criteria of the clustering algorithm, the robot processing trajectory is divided, and the surface to be processed is divided into different sub-regions; for the different divided sub-regions, the regional stiffness index is used as the optimization target, and the artificial bee colony algorithm is used to optimize the robot redundancy and the workpiece turntable angle to obtain the best processing parameters. The present invention divides the region with stiffness as an index, and the regional division based on the processing strategy can lay a foundation for the optimization of the process parameters of subsequent processing, so that the regional division serves the processing more effectively. The optimization of the robot surface processing trajectory is realized by optimizing the regional parameters, reducing the terminal deformation during the robot processing process, reducing the contour error of the processed parts, and having a significant effect on improving the surface processing quality.

Description

Technical Field

[0001] The invention relates to the technical field of industrial robot surface milling optimization, and in particular to an industrial robot surface processing method based on region division. Background Art

[0002] Industrial robots have advantages such as flexibility, intelligence, and low cost, and are expected to replace traditional CNC machine tools in industries such as automobile manufacturing and aerospace. However, robots also have defects that cannot be ignored: low rigidity and low repeatability. Robot rigidity is a key factor affecting processing accuracy. During the robot milling process, the robot end is prone to deformation, causing path deviation, and may also cause vibration problems, further reducing the robot's processing stability, seriously affecting the service life of the robot and spindle and the surface processing quality of the workpiece. At present, the mainstream is to optimize the processing trajectory of industrial robots and improve the rigidity of the robot during processing. In actual processing, it is difficult to avoid positions with insufficient rigidity for complex surfaces even through posture optimization.

[0003] Currently, the stiffness index of industrial robots is concentrated on the evaluation of a certain end posture, while the processing quality is often evaluated on a processing area. Therefore, there is a lack of stiffness evaluation indicators for the overall posture of the robot processing an area.

[0004] Dividing the workpiece into different sub-areas and processing the sub-areas with different processing parameters can simplify the calculation of path optimization and improve the processing stability and quality of industrial robots. At present, the division is mainly based on the geometric features of the workpiece surface, which does not have practical processing significance. The purpose of regional segmentation is to lay the foundation for subsequent tool path planning and processing parameter optimization. Therefore, establishing the relationship between partition criteria and processing parameters can give the partition a practical processing significance and make the partition serve the processing more effectively.

[0005] By utilizing the robot's redundant degrees of freedom and workpiece turntable, combined with the area-based variable process parameter processing method, the processing method has a higher applicability for workpieces. By optimizing the redundancy and turntable angle of each area, each processing area corresponds to a set of processing parameters, which can significantly improve processing efficiency, improve robot mobility and reduce processing errors. Summary of the invention

[0006] Aiming at the insufficient rigidity of an industrial robot during curved surface machining, the present invention provides an industrial robot curved surface machining method based on area division, so as to improve the machining quality, stability and machining efficiency of the curved surface milling machining performed by the industrial robot.

[0007] To achieve the above object, the present invention provides an industrial robot surface processing method based on region division, comprising the following steps:

[0008] S1: Obtain the trajectory of surface machining and the kinematic parameters of the robot, establish the robot kinematic model, and convert the machining trajectory into the robot end position trajectory;

[0009] S2: establishing a robot regional stiffness index according to the surface machining trajectory, kinematic parameters and milling parameters, and constructing a regional stiffness optimization model;

[0010] S3: according to the surface processing trajectory, using a clustering algorithm, dividing the processing trajectory, and dividing the surface into different processing sub-areas;

[0011] S4: according to the different processing sub-regions divided as described above, a regional stiffness optimization model is established and solved to determine the optimal redundancy and turntable angle of the sub-region processing;

[0012] S5: According to the obtained optimal processing parameters, the processing trajectory is transmitted to the milling robot to complete the surface processing.

[0013] Furthermore, the regional stiffness optimization model includes an objective function and constraints, wherein the objective function is used to minimize the end deformation of the robot during the surface processing process, and the objective function of the regional stiffness optimization model is expressed by the following formula:

[0014] min f = min F(γ t ,γ w ,cls)

[0015] Among them, γ t represents the redundant degrees of freedom of the six-degree-of-freedom robot performing five-axis machining tasks, γ w represents the degree of freedom of the workpiece turntable, cls represents the machining trajectory, and F represents the regional stiffness index of the set of trajectory points in the machining area;

[0016]

[0017] in, Indicates the robot's overall ability to resist deformation when processing an area. Indicates the fluctuation degree of the end deformation during robot processing, j represents the jth trajectory point in the area, k is the custom weight, N is the number of trajectory points in the area, and k a It is the deformation evaluation index, which refers to the ability of the robot's end to resist deformation due to external forces in a certain posture;

[0018]

[0019] Among them, C fd is the robot translational flexibility matrix, which is determined by the robot posture, e f is the unit force on the end of the robot.

[0020] Furthermore, the constraints include:

[0021] The first constraint condition is that the robot's machining posture should satisfy the joint limit constraint to ensure that the joint angle of each axis is in the central part of the joint rotation angle as much as possible;

[0022] The second constraint condition is that the robot processing posture should meet the operability constraint, and the operability constraint is used to avoid singular points in the processing process;

[0023] The third constraint is that when processing the workpiece, the robot should avoid collision with the workpiece and the turntable.

[0024] Furthermore, the first constraint condition is expressed by the following formula:

[0025]

[0026]

[0027] in, is the middle position of joint i, q imax and q imin Respectively represent the maximum and minimum limit positions of each joint, q i represents the rotation angle of joint i, H(q) represents the distance from the current joint angle to the center position of the joint. The smaller the value, the closer the joint is to the center position. H0 is the threshold of joint performance.

[0028] Furthermore, the second constraint condition is expressed by the following formula:

[0029]

[0030] Where μ(q) represents the maneuverability of the robot, q is the robot joint angle, and A = J θ J θ T ,λ max and λ min are the maximum and minimum eigenvalues ​​of matrix A, μ0 is the threshold of operability, J θ is the Jacobian matrix of the robot.

[0031] Furthermore, the third constraint condition is expressed by the following formula:

[0032] f colt (q,γ w )=0

[0033] Among them, f coll () is the collision detection function in the kinematics simulation software

[0034] Furthermore, the processing trajectory is divided into different processing sub-areas including:

[0035] For the trajectory of the surface to be processed, take the trajectory points as vertices V and the vertices as edges E, and construct a weighted undirected graph G = (V, E);

[0036] The robot stiffness and end position corresponding to the vertex are used as similarity measures to calculate the weight w between the vertices ij , construct the similarity matrix W of the trajectory points of the surface to be processed;

[0037] Construct the similarity matrix D and the Laplace matrix L according to the similarity matrix W, perform eigenvalue decomposition on the Laplace matrix, obtain the eigenvectors corresponding to the first k eigenvalues, arrange them into a matrix Q by columns, and standardize the matrix Q;

[0038] Use K-means to cluster the standardized matrix Q to obtain the classification results of the surface trajectory points.

[0039] Furthermore, the robot stiffness and end position corresponding to the vertex are used as the similarity measure to calculate the weight w between the vertices. ij , construct the weight matrix W of the trajectory points of the surface to be processed, including:

[0040] For the trajectory points to be classified, the robot joint angles corresponding to the tool trajectory points are solved by inverse solution of robot kinematics, and the robot stiffness model is used to calculate the terminal deformation ε caused by unit force in the current posture of the robot i ;

[0041] Combine the trajectory point position coordinates and the end deformation into a new six-dimensional vector V i =x i ,y i ,z i ,ε xi ,ε yi ,ε zi ), and normalized, where x i ,yi,z i is the end position of trajectory point i, ε xi ,ε yi ,ε zi Indicates that the current trajectory point corresponds to the deformation of the robot end;

[0042] Use Euclidean distance to measure the six-dimensional vector V of two vertices i The similarity between the two vertices is calculated by using the Gaussian kernel function. ij , construct a symmetric weight matrix W based on the weights between all vertices.

[0043] Furthermore, according to the different sub-areas divided, the regional stiffness optimization model of the robot system is solved to determine the optimal redundancy and turntable angle of regional processing, and the optimal processing parameters are obtained, including:

[0044] Initialize the parameters of the artificial bee colony algorithm, randomly generate a robot processing parameter sequence in the search space, i.e., the initial honey source, and calculate the corresponding fitness value based on the processing parameter sequence;

[0045] According to the nectar source mining formula of employed bees, follower bees and observer bees in the artificial bee colony, new nectar sources are continuously mined in the search space;

[0046] If a nectar source has not been updated after reaching the mining limit, the nectar source will be abandoned and a new nectar source will be randomly generated in the search space to replace it;

[0047] According to the above steps, the nectar source is continuously iterated until the iteration termination condition is met, and the final optimal processing parameters are output.

[0048] Furthermore, according to the nectar source mining formula of employed bees, follower bees and observer bees in the artificial bee colony algorithm, continuously mining new nectar sources in the search space includes:

[0049] Select SN bees from the population as employed bees and SN bees as follower bees. The employed bees correspond to the nectar sources one by one.

[0050] Employed bees adopt directional learning strategy and elite learning strategy to mine new nectar sources near nectar sources through formulas and share nectar source information with other bees in the population:

[0051] After the employed bees complete the nectar mining, they calculate the probability value of each nectar source based on the fitness value. The follower bees use the roulette wheel method to select the nectar source to follow, and mine the nectar source in the same way as the employed bees.

[0052] For the newly mined nectar source, it is necessary to determine whether the nectar source meets the constraints. If not, the initial nectar source is still selected; if it is, a greedy selection is made between the new and old nectar sources, and the one with a higher fitness value is selected to update the nectar source.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] First, the overall stiffness index of the processing area of ​​the industrial robot is established; then, according to the stiffness performance index of each path point of the robot milling, the processing area is segmented, and the industrial robot regional variable process parameter processing optimization model is established; finally, the industrial robot regional variable process parameter processing optimization model is used to search and solve, and effective optimization is performed to determine the optimal processing path, and the parts are milled according to the optimal processing parameters. In summary, the present invention establishes a processing area stiffness performance index, which can more intuitively reflect the stiffness performance of the robot milling process, and at the same time, the processing trajectory is segmented, which improves the efficiency of robot milling and reduces the calculation amount of the optimization model. The regional variable process parameter optimization model is used to solve the overall stiffness of the robot during the processing process, which effectively improves the stability and processing quality of the industrial robot milling process. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A schematic diagram of a robot processing system of the present invention;

[0056] Figure 2 A schematic flow chart of an embodiment of an industrial robot surface processing method based on area division provided by the present invention;

[0057] Figure 3 The present invention provides Figure 2 Schematic diagram of the process of the embodiment of step S3;

[0058] Figure 4 The present invention provides Figure 2 Schematic diagram of the process of the embodiment of step S4;

[0059] Figure 5 The present invention provides Figure 4 Schematic diagram of the process of the embodiment of step S42;

[0060] Figure 6 It is a schematic diagram of the flow of the artificial bee colony algorithm of the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0063] The embodiment of the present invention provides an industrial robot surface processing method based on region division. In conjunction with FIG. 2 , Figure 2 The schematic diagram of the process of an embodiment of the industrial robot surface processing method provided by the present invention includes steps S1 to S5, wherein:

[0064] S1: Obtain the trajectory of surface machining and the kinematic parameters of the robot, establish the robot kinematic model, and convert the machining trajectory into the robot end position trajectory;

[0065] S2: establishing a robot regional stiffness index according to the surface machining trajectory, kinematic parameters and milling parameters, and constructing a regional stiffness optimization model;

[0066] S3: according to the surface processing trajectory, using a clustering algorithm, dividing the processing trajectory, and dividing the surface into different processing sub-areas;

[0067] S4: according to the different processing sub-regions divided as described above, a regional stiffness optimization model is established and solved to determine the optimal redundancy and turntable angle of the sub-region processing;

[0068] S5: According to the obtained optimal processing parameters, the processing trajectory is transmitted to the milling robot to complete the surface processing.

[0069] In step S1, trajectory parameters of surface machining, kinematic parameters of the robot and milling parameters are obtained;

[0070] In this embodiment, first, the trajectory parameters of the surface to be processed are obtained, which are composed of a set of CLS points with uniform discrete distribution, CLS = {cls1, cls2, ..., cls j}, where cls j =(x j ,y j ,z j ,i j ,j j ,k j ), x j ,y j ,z j Indicates the end position of the tool axis, i j ,j j ,k j Represents the direction vector of the tool axis, which is converted into the robot end coordinate system O t The pose matrix T tool ;

[0071] Furthermore, at the robot milling cutter, a tool coordinate system O is established. t According to the kinematic parameters, the POE method is used to establish the kinematic model of the six-degree-of-freedom robot Among them, ξ i is the rotation axis of each joint, [ξ i ] is ξ i The antisymmetric matrix of θi is the robot joint angle, and M is the robot terminal posture matrix under the initial posture. Establish the relationship between joint velocity and terminal velocity, and construct the Jacobian matrix J of the robot θ . Where T is the robot end pose matrix, [ξ i ] is the robot's i-th axis ξ i The antisymmetric matrix, θ i is the joint angle of the robot's i-th axis.

[0072] In step S2, a robot regional stiffness index is established according to the robot kinematic parameters, and a regional stiffness optimization model is constructed;

[0073] In this embodiment, first, for a six-degree-of-freedom robot, when performing a five-axis machining task, there is a redundant degree of freedom γ t , that is, the direction of rotation of the tool axis. Changing the redundant degrees of freedom will not affect the processing task. Therefore, the robot end coordinate system O t Posture can be defined as For a given processing task It can be seen that the robot's terminal posture is a redundant degree of freedom γ t and workpiece turntable angle γ w function.

[0074] in, is the attitude matrix of the turntable coordinate system relative to the base coordinate system, is the posture matrix of the workpiece coordinate system relative to the turntable, is the posture matrix corresponding to the robot processing task, roz ( t ) is the coordinate system around the end o t Tool axis rotationγ t The transformation matrix, T roz (γ w ) is the workpiece coordinate system o w Rotate around the Z axisγ w The transformation matrix of .

[0075] According to the robot kinematic parameters, the kinematic inverse solution of the processing task is obtained to obtain the joint angle of each processing posture Construct the robot end flexibility matrix. The stiffness performance of the robot in the joint space can be expressed by the diagonal matrix K θ It is expressed as follows: the relationship between the flexibility matrix of the robot end in Cartesian space and the joint stiffness coefficient in joint space is established, which is expressed as follows:

[0076]

[0077] K θ =diag(k θ1 , k θ2 , k θ3 , k θ4 , k θ5 , k θ6 )

[0078] Among them, C fd , C fδ , C md , C mδ They represent the force-translation flexibility matrix, force-rotation flexibility matrix, torque-translation flexibility matrix and torque-rotation flexibility matrix respectively, K θ is the robot joint stiffness matrix.

[0079] Only the translation deformation caused by the end force is considered, and the end translation deformation ε caused by the unit force is used as the stiffness performance evaluation index of the robot. However, the deformation ε caused by the unit force is a vector, which is converted into a scalar index ka;

[0080]

[0081]

[0082] Among them, ε is the deformation of the end under unit force, f is the magnitude of the external force on the end, and e f is the unit force on the end of the robot, that is, the direction of the external force on the end, ε x , ε y , ε z Indicates the deformation of the robot end in the next direction under unit force;

[0083] In actual processing, the evaluation of processing effect is often aimed at the entire surface or a certain area, rather than a certain processing point. The evaluation of a certain point is not enough to explain the processing of the entire surface. It is necessary to consider the overall stiffness of the entire processed surface so that the overall stiffness of the surface meets the processing requirements, and consider that the deformation fluctuation at the end of the overall processing is within a certain range. The regional stiffness index F is defined as:

[0084]

[0085] Where j represents the jth trajectory point in the region, k is a custom weight that affects the overall stiffness and stiffness fluctuation, N is the number of trajectory points in the region, and k a It is a deformation evaluation index, which refers to the deformation of the robot's end subjected to unit force in a certain posture.

[0086] Furthermore, the regional stiffness optimization model includes an objective function and constraints. The objective function is used to minimize the end deformation of the robot during milling. When a given processing task is given, the regional stiffness index of the robot is affected by redundant degrees of freedom and the workpiece turntable. The objective function is expressed by the following formula:

[0087] min f = min F(γ t , γ w )

[0088] Among them, γ t is the robot's redundant degrees of freedom, γ w It is the rotation angle of the workpiece coordinate system around the Z axis.

[0089] The constraints include:

[0090] The first constraint condition is to make the joint angle of the robot's machining posture meet the joint limit and ensure that the joint angle of each axis is as close to the center of the joint angle as possible, which can be expressed by the following formula:

[0091]

[0092]

[0093] in, is the middle position of joint i, q imax and q imin Respectively represent the maximum and minimum limit positions of each joint, q i represents the rotation angle of joint i, H(q) represents the distance from the current joint angle to the center position of the joint. The smaller the value, the closer the joint is to the center position. H0 is the threshold of joint performance.

[0094] The second constraint condition is to make the robot's processing posture maneuverability meet the requirements. Maneuverability is used to evaluate the robot's current posture's ability to move in all directions. Maneuverability constraints are used to avoid singular points in the processing process. It is expressed by the following formula:

[0095]

[0096] Where q is the robot joint angle, A = J θ J θ T ,λ max and λ min are the maximum eigenvalue and minimum eigenvalue of matrix A respectively, and μ0 is the threshold of operability.

[0097] The third constraint condition is that when processing the workpiece, the robot should avoid collision with the workpiece and the turntable. The collision detection in the kinematic simulation is used to determine whether the current posture meets the requirements, which is expressed by the following formula:

[0098] f coll (q, γ w )=0

[0099] As a preferred embodiment, Figure 3 Come and see, Figure 3 The present invention provides Figure 2 The flowchart of step S3 of an embodiment of the present invention is as follows, step S3 includes steps S31 to S35. In step S3, according to the surface processing trajectory, a clustering algorithm is used to divide the processing trajectory, and the surface is divided into different processing areas, including:

[0100] In step S31, for the trajectory points to be segmented, the robot joint angles corresponding to the tool trajectory points are solved by inverse solution of robot kinematics, and the end deformation ε caused by unit force in the current posture of the robot is calculated using the robot stiffness model. i , merge the trajectory point position coordinates and the terminal deformation into a new six-dimensional vector V i =(x i ,y i , z i , ε xi , ε yi , ε zi ) and normalize it;

[0101] V i =[P i , ε i ]=(x i ,y i , z i , ε xi , ε yi , ε zi )

[0102] Among them, P i is the robot trajectory point position vector, ε i is the terminal deformation of the trajectory point corresponding to the robot posture.

[0103] In step S32, a weighted undirected graph G=(V, E) is constructed with the processing trajectory point as the vertex V, where E is a set of edges;

[0104] In step S33, the Euclidean distance is used to measure the distance between two vertices V i The similarity between the two vertices is calculated by using the Gaussian kernel function. ij :

[0105]

[0106] Among them, dist ijrepresents the Euclidean distance between two vertices i and j, σ is a scale parameter related to the distance function of all vertices;

[0107] In step S34, a symmetric n×n similarity matrix W is constructed, where N represents the number of trajectory points and represents the similarity between the robot postures corresponding to all trajectory points;

[0108]

[0109] In step S35, the similarity matrix W is used to construct the degree matrix D and the Laplace matrix L, and the Laplace matrix is ​​decomposed to obtain the eigenvalues ​​corresponding to its first k eigenvalues, and the eigenvectors are arranged into a matrix Q by columns. All rows of the matrix Q are standardized, and K-means is used for clustering to obtain the final classification result.

[0110] As a specific embodiment, the embodiment of the present invention establishes multiple constraints to ensure the effective constraints of multiple conditions, wherein the first constraint ensures that the joint angles of the robot's processing posture meet the joint limits, the second constraint ensures that the robot's processing posture meets the operability requirements, and the third constraint ensures that the robot does not collide with the turntable, workpiece, etc. during the processing.

[0111] As a preferred embodiment, Figure 4 Come and see, Figure 4 The present invention provides Figure 2 The flowchart of step S4 is a schematic diagram of an embodiment. Step S4 includes steps S41 to S44. In step S4, according to the different sub-areas classified above, the regional stiffness optimization model of the robot system is solved for the redundancy of the robot system and the workpiece parameters to obtain the optimal processing parameters, wherein:

[0112] In step S41, the artificial bee colony algorithm parameters are initialized, and the redundant angles γ of SN groups of robots are randomly generated in the search space. t and workpiece turntable angle γ w , that is, the initial nectar source x i,j , according to the initial honey source and processing task CLS, determine the robot stiffness value Fit corresponding to each set of redundant angles and turntable angles i , assign tasks to the bee population, select SN bees as employed bees and SN bees as follower bees;

[0113] In step S42, the employed bees correspond to the nectar sources one by one, and adopt the direction learning strategy and the elite learning strategy to mine new nectar sources near the nectar sources through the formula, and judge whether the newly mined nectar sources meet the constraint conditions. If not, the initial nectar source is retained; if so, a greedy selection is made between the new and old nectar sources, and the one with the higher fitness value is selected to update the nectar source; after the employed bees have mined the nectar sources, the follower bees select the nectar source to follow according to the probability value by gambling on roulette, and mine the nectar sources in the same way as the employed bees;

[0114] In step S43, if no new nectar source is found after the upper limit of the exploitation of a certain nectar source is reached, the nectar source is abandoned, and the bees corresponding to the nectar source are turned into observation bees, and a new nectar source is randomly generated in the search space to replace it;

[0115] In step S44, according to the mining formula of employed bees, follower bees and observer bees in the bee population, the processing parameters corresponding to all nectar sources are updated multiple times until the iteration termination condition is met, and the final optimal processing parameters are output;

[0116] As a preferred embodiment, Figure 5 Come and see, Figure 5 The present invention provides Figure 4 In the flowchart of step S42 of an embodiment, step S42 includes steps S421 to S424, wherein:

[0117] In step S421, SN bees are selected from the population as employed bees, and SN bees are follower bees, and the employed bees correspond to the nectar sources one by one;

[0118] In step S422, the employed bees adopt the directional learning strategy and the elite learning strategy to mine new nectar sources near the nectar sources through formulas, and share the nectar source information with other bees;

[0119] In step S423, the follower bee calculates the following probability of each nectar source based on the nectar source information shared by the employed bee, selects a nectar source to follow by gambling on a roulette wheel, and mines the nectar source in the same way as the employed bee;

[0120] In step S424, according to the mining formula of employed bees, follower bees and observer bees in the bee population, the processing parameters corresponding to all nectar sources are updated multiple times until the iteration termination condition is met, and the optimal processing parameters of the processing area are output.

[0121] In step S5, according to the obtained optimal processing parameters of the sub-areas, the processing trajectories of different sub-areas are spliced ​​into a complete processing surface trajectory in the order of the segmented labels, and the complete processing trajectory is transmitted to the robot system through the Socket protocol, and the surface processing task is completed by offline programming.

[0122] In a specific embodiment of the present invention, the search process of the artificial bee colony algorithm is as follows:

[0123] The first step is to set the initial number of nectar sources SN, the number of employed bees SN, the number of follower bees SN, the maximum number of nectar source mining times limit, and the maximum number of iterations trial max ;

[0124] The second step is to randomly generate a sequence of robot processing parameters in the search space, namely the initial population;

[0125] x i，j =x min，j +rand(0,1)(x max，j -x minj，j )=(γ t , γ w )

[0126] Among them, x ij represents the i-th nectar source x i The value of the jth dimension, x max，j and x min，j Respectively represent the maximum and minimum values ​​of the j-th dimension;

[0127] The third step is to perform task assignment, calculate the fitness value Fit of each nectar source, and select and record the top M nectar sources with the current fitness value;

[0128] Fit=F(x i，j )=F(γ t , γ w )

[0129] The fourth step is to select SN bees as hired bees, correspond one to one with the initial nectar source, adopt the directional learning strategy and elite learning strategy, mine near the initial nectar source according to the hired honey source mining formula, update the nectar source, and share the nectar source information:

[0130] Among them, j is the dimension of the honey source, δ ij is a random number in {-1, 1}, which is the direction factor in the jth dimension of the i-th honey source and is used to control the search direction; is a random number in {0, 1}, used to control the search step length. is a randomly selected nectar source from the top M nectar sources in terms of fitness value; in the directional learning strategy, the directional factor δ ij Used to record the search directions of different individuals. If the generated v ij The fitness value is better than x ij , it indicates that the current search direction can improve the fitness value of the nectar source, δ ij Remain unchanged, otherwise, the current search direction is not suitable for searching, δij The opposite number should be taken; in the elite learning strategy, the search equation is introduced only in the nectar sources with the top M fitness values, which improves the convergence speed of the algorithm. However, if M is too small, it will lead to premature algorithm. Therefore, different emphases should be placed in different search stages, which should not only speed up the convergence speed but also avoid local optimality. M is defined as:

[0131]

[0132] Among them, trial is the current iteration number, trial max is the upper limit of the number of algorithm iterations, and μ is a custom coefficient used to prevent the number of elite populations from being too small, causing the algorithm to converge prematurely. For newly mined nectar sources, it is necessary to determine whether the nectar source meets the constraints. If it does not meet the constraints, the initial nectar source is still selected; if it does, a greedy selection is made between the new and old nectar sources, and the one with the highest fitness value is selected to update the nectar source; in the fifth step, the remaining SN bees are selected as follower bees, and the follow probability value p of each nectar source is calculated based on the nectar source information shared by the employed bees. i , choose the nectar source to follow by gambling on roulette, and use the same strategy as hired bees to mine near the nectar source:

[0133] Among them, Fit i is the fitness value of the current nectar source, N p is the number of nectar sources.

[0134] In the sixth step, the nectar sources that have reached the upper limit of the exploitation limit but have not been updated are abandoned, and the corresponding bees are converted into observer bees, which randomly exploit nectar sources in the search space as replacements. The way to generate nectar sources is the same as in the second step;

[0135] Step 7, repeating steps 3 to 6, updating the nectar source through the employed bee, follower bee and observer bee formulas until the iteration limit is reached, and outputting the final optimal processing parameters;

[0136] The invention discloses an industrial robot surface processing method based on area division. First, the overall stiffness index of the processing area of ​​the industrial robot is established; then, the processing area is divided according to the stiffness performance of each path point of the robot, and an industrial robot regional variable process parameter processing optimization model is established; finally, the industrial robot regional variable process parameter optimization model is used to search and solve, and effective optimization is performed to determine the optimal processing parameters, and the parts are milled according to the optimal processing parameters.

[0137] The technical solution of the present invention establishes a processing area stiffness performance index, which more intuitively reflects the stiffness performance of the robot in the milling process, and segments the processing trajectory according to the stiffness performance of the robot posture corresponding to the processing trajectory, laying the foundation for subsequent process parameter optimization, so that the regional segmentation serves the processing more effectively, improves the efficiency of robot milling and reduces the calculation amount of the optimization model, and utilizes the solution of the regional variable process parameter optimization model to make the overall stiffness of the robot meet the requirements during the processing, effectively improving the stability and processing quality of the industrial robot in the milling process.

[0138] The above-described embodiments are merely descriptions of preferred implementations of the present invention, and are not intended to limit the concept and scope of the present invention. Without departing from the design concept of the present invention, various modifications and improvements made by ordinary persons in the art to the technical solution of the present invention should fall within the protection scope of the present invention, and the technical contents for which protection is sought in the present invention have been fully recorded in the claims.

Claims

1. An industrial robot surface processing method based on region division, characterized in that: The steps include: S1: Obtain the trajectory of surface machining and the kinematic parameters of the robot, establish the robot kinematic model, and convert the machining trajectory into the robot end position trajectory; S2: Establishing the robot regional stiffness index according to the surface machining trajectory, kinematic parameters and milling parameters, and constructing a regional stiffness optimization model; the regional stiffness optimization model includes an objective function and constraint conditions, the objective function is used to minimize the terminal deformation of the robot during the surface machining process, and the objective function of the regional stiffness optimization model is expressed by the following formula: minf=minF(γ t ,c w ,cls) Among them, γ t represents the redundant degrees of freedom of the six-degree-of-freedom robot performing five-axis machining tasks, γ w represents the degree of freedom of the workpiece turntable, cls represents the machining trajectory, and F represents the regional stiffness index of the set of trajectory points in the machining area; in, Indicates the robot's overall ability to resist deformation when processing an area. Indicates the fluctuation degree of the end deformation during robot processing, j represents the jth trajectory point in the area, k is the custom weight, N is the number of trajectory points in the area, and k a It is the deformation evaluation index, which refers to the ability of the robot's end to resist deformation due to external forces in a certain posture; Among them, C fd is the robot translational flexibility matrix, which is determined by the robot posture, e f is the unit force on the end of the robot; S3: according to the surface processing trajectory, using a clustering algorithm, the processing trajectory is divided into different processing sub-areas; including: for the surface trajectory to be processed, with the trajectory points as vertices V and the vertices as edges E, constructing a weighted undirected graph G = (V, E); The robot stiffness and end position corresponding to the vertex are used as similarity measures to calculate the weight w between the vertices ij , construct the similarity matrix W of the trajectory points of the surface to be processed; Construct the similarity matrix D and the Laplace matrix L according to the similarity matrix W, perform eigenvalue decomposition on the Laplace matrix, obtain the eigenvectors corresponding to the first k eigenvalues, arrange them into a matrix Q by columns, and standardize the matrix Q; Use K-means to cluster the standardized matrix Q to obtain the classification results of the surface trajectory points; S4: according to the different processing sub-regions divided as described above, a regional stiffness optimization model is established and solved to determine the optimal redundancy and turntable angle of the sub-region processing; S5: According to the obtained optimal processing parameters, the processing trajectory is transmitted to the milling robot to complete the surface processing.

2. The method for surface processing of an industrial robot based on area division according to claim 1, characterized in that: The constraints include: The first constraint condition is that the robot's machining posture should satisfy the joint limit constraint to ensure that the joint angle of each axis is in the central part of the joint rotation angle as much as possible; The second constraint condition is that the robot processing posture should meet the operability constraint, and the operability constraint is used to avoid singular points in the processing process; The third constraint is that when processing the workpiece, the robot should avoid collision with the workpiece and the turntable.

3. The method for surface processing of an industrial robot based on area division according to claim 2, characterized in that: The first constraint condition is expressed by the following formula: in, is the middle position of joint i, q imax and q imin Respectively represent the maximum and minimum limit positions of each joint, q i represents the rotation angle of joint i, H(q) represents the distance from the current joint angle to the center position of the joint. The smaller the value, the closer the joint is to the center position. H0 is the threshold of joint performance.

4. The method for surface processing of an industrial robot based on area division according to claim 2, characterized in that: The second constraint is expressed by the following formula: Where μ(q) represents the maneuverability of the robot, q is the robot joint angle, and A = J θ J θ T ,λ max and λ min are the maximum and minimum eigenvalues ​​of matrix A, μ0 is the threshold of operability, J θ is the Jacobian matrix of the robot.

5. The method for surface processing of an industrial robot based on area division according to claim 2, characterized in that: The third constraint condition is expressed by the following formula: f coll (q,γ w )=0 Among them, f coll () is the collision detection function in the kinematic simulation software.

6. The method for surface processing of an industrial robot based on area division according to claim 1, characterized in that: The robot stiffness and end position corresponding to the vertex are used as similarity measures to calculate the weight w between the vertices ij , construct the weight matrix W of the trajectory points of the surface to be processed, including: For the trajectory points to be classified, the robot joint angles corresponding to the tool trajectory points are solved by inverse solution of robot kinematics, and the robot stiffness model is used to calculate the terminal deformation ε caused by unit force in the current posture of the robot i ; Combine the trajectory point position coordinates and the terminal deformation into a new six-dimensional vector V i =(x i ,y i ,z i ,ε xi ,ε yi ,ε zi ), and normalized, where x j ,y j ,z j is the end position of the trajectory point, ε xi ,ε yi ,ε zi Indicates that the current trajectory point corresponds to the deformation of the robot end; Use the Euclidean distance to measure the distance between two vertices V i The similarity between the two vertices is calculated by using the Gaussian kernel function. ij , construct a symmetric weight matrix W based on the weights between all vertices.

7. The method for surface processing of an industrial robot based on area division according to claim 1, characterized in that: According to the different sub-areas, the regional stiffness optimization model of the robot system is solved to determine the optimal redundancy and turntable angle of regional processing, and the optimal processing parameters are obtained, including: Initialize the parameters of the artificial bee colony algorithm, randomly generate a robot processing parameter sequence in the search space, i.e., the initial honey source, and calculate the corresponding fitness value based on the processing parameter sequence; According to the nectar source mining formula of employed bees, follower bees and observer bees in the artificial bee colony, new nectar sources are continuously mined in the search space; If a nectar source has not been updated after reaching the mining limit, the nectar source will be abandoned and a new nectar source will be randomly generated in the search space to replace it; According to the above steps, the nectar source is continuously iterated until the iteration termination condition is met, and the final optimal processing parameters are output.

8. The method for surface processing of an industrial robot based on area division according to claim 7, characterized in that: According to the nectar source mining formula of employed bees, follower bees and observer bees in the artificial bee colony algorithm, new nectar sources are continuously mined in the search space, including: Select SN bees from the population as employed bees and SN bees as follower bees. The employed bees correspond to the nectar sources one by one. Employed bees adopt directional learning strategy and elite learning strategy to mine new nectar sources near nectar sources through formulas and share nectar source information with other bees in the population: After the employed bees complete the nectar mining, they calculate the probability value of each nectar source based on the fitness value. The follower bees use the roulette wheel method to select the nectar source to follow, and mine the nectar source in the same way as the employed bees. For the newly mined nectar source, it is necessary to determine whether the nectar source meets the constraints. If not, the initial nectar source is still selected; if it is, a greedy selection is made between the new and old nectar sources, and the one with a higher fitness value is selected to update the nectar source.

Images