Optimal Index of Precision Measurement Configuration for Numerical Control Equipment and Its Kinematic Calibration Method

Abstract

The present invention discloses an optimal index for accuracy measurement configuration of numerical control equipment and its kinematic calibration method, belonging to the technical field of kinematic calibration design of industrial robots, including S1, establishing a geometric error transfer model of the target robot and screening parameters to be identified; S2, establishing a new index based on the ill-condition theory of matrices and the spatial analysis theory; S3, using the new index as the fitness function of the genetic algorithm to obtain the optimal configuration for parameter identification; S4, using the regularization method to obtain the identification result of geometric error parameters; S5, selecting verification configurations in the full working area for error compensation; The present invention first uses the POE method to establish a geometric error transfer model of the robot and screen identification parameters, then discretizes the working space of the robot and uses it as the optimization range of the genetic algorithm, uses the new index as the fitness function to find the optimal configuration, and finally uses the regularization method for parameter identification. By making a comparative analysis with the previous observability index, the superiority of the new index is revealed.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial robot kinematic calibration, and in particular to a preferred index for precision measurement configuration of numerical control equipment and its kinematic calibration method. Background Art

[0002] The absolute accuracy and repeatability accuracy of the end effector of an industrial robot are important accuracy indicators for industrial production and manufacturing. In addition to thermal deformation errors and control system errors, there are still certain geometric errors and structural errors in precision robots, and accuracy compensation can be carried out through kinematic calibration. The process of kinematic calibration mainly includes the following four links: geometric error modeling, end error measurement, error parameter identification, and error compensation. In this step of error parameter identification, the parameter identification results obtained with a well-behaved identification matrix can often improve the accuracy of parameter identification.

[0003] In order for the identification matrix to calculate the estimated value of geometric error parameters through the identification algorithm, it is necessary to construct at least a matrix with the same number of rows as the number of error source parameters, and this matrix needs to be as conducive to parameter identification as possible, which requires the matrix to have characteristics such as low ill-conditioning degree, strong robustness, and full rank.

[0004] In previous studies, multiple observability indicators have been explored as the basis for screening the optimal configuration. Among them, the most commonly used is the condition number. The meaning of the condition number is to find the maximum difference in the stretching degree of the matrix as a whole in different dimensions, trading the accuracy of judging the matrix properties for the maximum fault tolerance range. Its disadvantage is that it cannot reflect the distribution state of the remaining singular values and loses a large amount of information from the remaining singular values.

[0005] Among the five observability indicators O1 to O5, O2, O3, and O4 are all variants of the condition number calculation method. In essence, they have no difference from the geometric meaning symbolized by the condition number and also lose a large amount of information; O1 and O5 retain all the singular value information. O1 represents the square root mean of the product of each singular value, and its geometric meaning is a quantity related to the volume of the ellipsoid, emphasizing the overall characteristics. The larger this value is, the more it indicates that the selected configuration is conducive to parameter identification; while O5 represents the harmonic mean of all singular values, and its geometric meaning is similar to that of O1. The disadvantage of the latter is that it cannot eliminate redundant information, increasing the calculation burden, and points with large difference values may affect the overall identification result.

[0006] Therefore, although the existing observability indicators can screen out the pose combinations that are conducive to parameter identification from a large number of poses through evolutionary algorithms, due to the omission of matrix information, the parameter identification effect cannot be further improved, and this difficulty remains to be solved. Summary of the Invention

[0007] The object of the present invention is to provide an optimal index for the precision measurement configuration of a numerical control equipment and its kinematic calibration method, so as to solve the problems existing in the above-mentioned background technology.

[0008] To achieve the above object, the present invention provides an optimal index for the precision measurement configuration of a numerical control equipment and its kinematic calibration method, including the following steps:

[0009] S1. Establish a geometric error transfer model of the target robot and screen the parameters to be identified;

[0010] S2. Establish a new index based on the ill-condition theory of matrices and the spatial analysis theory;

[0011] S3. Use the new index as the fitness function of the genetic algorithm to obtain the optimal configuration for parameter identification;

[0012] S4. Use the regularization method to obtain the identification result of the geometric error parameters;

[0013] S5. Select verification configurations in the entire working area for error compensation.

[0014] Preferably, step S1 specifically includes:

[0015] S11. Establish a geometric error transfer model of the robot based on the POE method. The POE method avoids the singularity problem in the modeling process, and the general formula of the model can be expressed as:

[0016] ;

[0017] Among them, represents the geometric pose error of the moving platform of the present motion platform, , is a 6×6 matrix, which is the driving Jacobian matrix of the robot, represents a diagonal matrix with the product of the unit driving force screw of a certain branch chain and all the adjoint transformation matrices of the branch chain as the diagonal elements, represents the structural error of the robot joint, represents the motion error of the joint; represents the error transfer Jacobian matrix, represents the geometric error source vector;

[0018] S12. Eliminate redundant parameters from all the error parameters of the robot.

[0019] Preferably, step S2 specifically includes:

[0020] S21. Perform matrix ill-conditioning analysis on the error transfer Jacobian matrix of the robot. Assume that the identified Jacobian matrix A obtained from the geometric error model of the robot is an m×n matrix. The row vectors of the matrix represent different geometric error parameters of the same configuration of the robot, and the column vectors represent the same geometric error parameter of the robot under different configurations. In order for the identification matrix to calculate the estimated values of geometric error parameters through the identification algorithm, it is necessary to construct at least a matrix with the number of rows equal to the number of error source parameters, and this matrix needs to be as conducive to parameter identification as possible, which requires the matrix to have characteristics such as small ill-conditioning degree, strong robustness, and full rank. By designing indicators based on the properties of the matrix itself, it helps to quickly construct an identification matrix with excellent performance. In order to construct an indicator that can reflect the relationship between the rows (or columns) of the matrix, it is necessary to establish a theoretical basis for the new indicator based on the row and column vectors of the matrix;

[0021] S22. Expand the error transfer Jacobian matrix of the robot in a high-dimensional space and endow it with the geometric meaning of taking the volume of a hyper-parallel polyhedron as the index value.

[0022] Preferably, in step S12, redundant parameters are removed from the error transfer Jacobian matrix of the established overall machine geometric error transfer model through column correlation analysis to ensure that the error transfer Jacobian matrix of the robot can perform various index operations normally.

[0023] Preferably, the construction process of the new indicator in step S2 includes:

[0024] Let the identified matrix A obtained from robot modeling and point selection. Assume m≥n. When A is column full rank, perform elementary transformation on A to convert it into a trapezoidal matrix M, and then take the first n rows of M to obtain an n-dimensional square matrix N, which is called the information matrix;

[0025] Perform Gram-Schmidt (Gram-Schmidt) orthogonalization (hereinafter referred to as G-S orthogonalization) on the information matrix N by column vectors and normalize it to obtain the matrix G: The specific G-S orthogonalization is as follows: Let be a finite or countable number of linearly independent vectors in the inner product space H, then there must be an orthonormal system in H such that for each natural number , is a linear combination of , is a linear combination of , and this is completely determined by except for a constant factor with an absolute value of 1;

[0026] Represent the column vectors in G as a set of bases in the parameter space , the volume formed by the basis vectors is , and this space is an n-dimensional standard space:

[0027] ;

[0028] Normalize the row vectors of the information matrix N to obtain the matrix N0, and perform Gram - Schmidt orthogonalization on N0 to obtain the matrix N a , N a The row vectors of are represented as , and calculate the volume U of the hyper - parallelepiped that the information matrix G can expand accordingly n , then the metric index of the identification matrix is expressed as :

[0029] .

[0030] Preferably, in step S3, the calculation method of the new index is used as the fitness function, and the genetic algorithm is used as the evolutionary algorithm for configuration optimization. This genetic algorithm has dynamically updated crossover probability and genetic probability for faster convergence.

[0031] Preferably, in step S4, the generalized cross - validation criterion is used to optimize the regularization parameter, and the Tikhonov regularization method is used to estimate the parameters of the identification equation system The essence of generalized cross - validation (GCV) can be summarized as: arbitrarily divide the detection data set into two subsets, use the data of one subset to identify the error parameters and use them to predict the other subset. The optimal regularization parameter should make the aforementioned prediction accuracy optimal in a statistical sense. The valuation criterion of the regularization method applied to the present invention can be expressed as:

[0032] ;

[0033] Among them, is the regularization parameter; L is a symmetric positive - definite matrix, and usually the identity matrix I is taken in engineering; is the true value of the geometric error source; is the estimate of the geometric error source parameter, specifically the variable to be optimized and solved. Under the condition of satisfying the parameter estimation criterion, the solution of the error identification model is:

[0034] ;

[0035] Compared with the least - squares solution, the Tikhonov regularized solution improves the stability of the parameter identification result by adding a damping term to the diagonal elements of the normal matrix .

[0036] Preferably, in step S5, a new index configuration optimization combination is used as a parameter identification matrix, and a regularization method is used as the identification method for the geometric error parameters of the robot. They are respectively substituted into the optimal configuration and the verification configuration to calculate the estimated value of the end pose error. The equivalent motion error of the driving joint is obtained from the end attitude error through the mapping matrix. The equivalent motion error is the rod length error of each driving joint, and its opposite number is the error compensation amount. :

[0037] ;

[0038] Among them, is the driving Jacobian matrix of the 6-UPS six-degree-of-freedom parallel robot, is the identification matrix, is the geometric error source combination.

[0039] Preferably, the specific process of the genetic algorithm is as follows:

[0040] Individual coding: After the workspace is discretized according to the specified spatial coordinate interval and attitude angle interval, the pose combination of the robot is encoded using binary, and it is used as a biological individual. In each pose combination, a points in the workspace are selected as internal genes;

[0041] Set the relevant parameters of the genetic algorithm;

[0042] Set the fitness function: The fitness function of the genetic algorithm respectively selects the new evaluation index ;

[0043] Generate the initial population, calculate the identification matrix corresponding to each organism in the population to calculate the fitness function;

[0044] In the population, select the parent generation for crossover according to the corresponding probability generated by the fitness function to generate offspring, and then perform gene mutation on the offspring according to the mutation function to form a new generation of population as a whole;

[0045] Iteration times + 1 until the upper limit of the iteration times is exceeded;

[0046] Obtain the population or individual with the best fitness, and use the pose combination corresponding to this individual as the basis for parameter identification.

[0047] Combine the driving joint variables with the compensation amount and add them to the control system, and then use a laser tracker to measure the end pose information of the compensated pose. Finally, analyze and compare the errors of the six degrees of freedom before and after compensation and the overall error of the optimal configuration obtained based on the new index and the optimal configuration obtained based on the condition number.

[0048] Therefore, the present invention adopts the above-mentioned preferred index for the precision measurement configuration of the numerical control equipment and its kinematic calibration method, and has the following beneficial effects:

[0049] (1) Compared with the original index, the proposed new index improves the effective information carrying capacity of the error transfer Jacobian matrix in the step of configuration optimization, enabling more accurate results to be obtained for the parameter identification of geometric error sources;

[0050] (2) Compared with the traditional method, the genetic algorithm with dynamically updated parameters can converge faster, enabling the algorithm to traverse more configuration combinations in the working space of the robot to find the optimal configuration;

[0051] (3) The Tikhonov regularization method based on the generalized cross-validation criterion used will obtain more stable parameter identification results compared with the least squares method.

[0052] Next, through the drawings and embodiments, the technical solution of the present invention will be further described in detail. Brief Description of the Drawings

[0053] Figure 1 is a flowchart of the preferred index for the precision measurement configuration of the numerical control equipment and its kinematic calibration method of the present invention;

[0054] Figure 2 is a schematic diagram of the geometric meaning of the new index proposed in the embodiment of the present invention; among them, (a) is non-ill-conditioned, (b) is mildly ill-conditioned, and (c) is severely ill-conditioned;

[0055] Figure 3 is a flowchart of the genetic algorithm in the embodiment of the present invention;

[0056] Figure 4 is an iteration diagram of the genetic algorithm with the condition number as the fitness in the embodiment of the present invention;

[0057] Figure 5 is an iteration diagram of the genetic algorithm with the new index as the fitness in the embodiment of the present invention;

[0058] Figure 6 is a comparison diagram of the root mean square error of multiple parameter identification results in the embodiment of the present invention; among them, (a) is the root mean square error of the parameter identification result based on the condition number, and (b) is the root mean square error of the parameter identification result based on the new index;

[0059] Figure 7 is a comparison diagram before and after the spatial position compensation of the verification configuration in the embodiment of the present invention; among them, (a) is before compensation, (b) is after compensation based on the condition number, and (c) is after compensation based on the new index;

[0060] Figure 8This is a comparison diagram of the attitude angle compensation of the verified configuration in the embodiments of the present invention; among them, (a) is before compensation, (b) is after compensation based on the condition number, and (c) is after compensation based on the new index. Detailed implementation manners

[0061] The following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0062] Please refer to Figure 1 , the preferred index for the accuracy measurement configuration of the numerical control equipment and its kinematic calibration method, including the following steps:

[0063] S1. Establish a geometric error transfer model of a six-degree-of-freedom parallel robot by the exponential product method. The specific process is as follows:

[0064] From the part of the geometric error model of the serial kinematic chain, we can obtain:

[0065] (1)

[0066] Among them, represents the end pose error of the platform, represents the motion error of joint , represents the joint structure error, represents the unit allowable motion infinitesimal displacement screw of joint , represents the adjoint transformation matrix of system relative to system .

[0067] The six-degree-of-freedom platform is connected by six serial kinematic chains between the static platform and the moving platform. Therefore, add " " at the lower right corner of all variables on the right side of Equation (1) to describe the number of the branch chain, and we can obtain:

[0068] (2)

[0069] The formula (2) is further described by the following formula:

[0070] , ;

[0071] ;

[0072] ;

[0073] ;

[0074] Among them, represents the geometric pose error of the moving platform of this motion platform, represents the branch chain 's degree of freedom. Construct the unit allowable motion infinitesimal displacement screw of each joint of each branch chain:

[0075] (3)

[0076] (4)

[0077] Among them, - represents the angular velocity of each joint of each branch chain, represents the coordinate vector from the origin of the system to the center of each hinge point of the moving platform expressed in the system , represents the coordinate vector from the origin of the system to the center of each hinge point of the moving platform expressed in the system , represents the direction vector of the driving rod i in the early system , is the vector from the origin of the system to the origin of the system , and the th branch chain of the robot satisfies , represents the attitude matrix.

[0078] Using the property that the driving (constraint) joint force screw only does work on the unit allowable motion infinitesimal displacement screw it causes, take the inner product of both sides of Equation (2) with respect to the unit driving force screw of the branch chain , and we get:

[0079] (5)

[0080] Write the above formula in matrix form:

[0081] (6)

[0082] is a 6×6 matrix, which is called the driving Jacobian matrix of the 6-UPS six-degree-of-freedom parallel robot. The columns of the transpose of respectively represent a unit driving force screw acting on the moving platform. Note that, in the non-singular configuration, is invertible, so let

[0083] (7)

[0084] Wherein, represents the geometric pose error of the moving platform of this motion platform, , is a 6×6 matrix, which is the driving Jacobian matrix of the robot, represents a diagonal matrix with the product of the unit driving force screw of a certain branch chain and all the adjoint transformation matrices of this branch chain as the diagonal elements, represents the structural error of the robot joint, represents the motion error of the joint. represents the error transfer Jacobian matrix, represents the geometric error source vector. Thus, the overall geometric error model of this 6-UPS six-degree-of-freedom parallel robot is established. The error transfer Jacobian matrix is 6 rows and 216 columns. The 6 rows represent the six-dimensional pose of the robot, and the 216 columns represent the 216 error sources of the robot.

[0085] Select individual configurations, and geometric error relationships can be obtained , written in matrix format, which is the identification equation set for error parameter identification:

[0086] (8)

[0087] Wherein, is the identification matrix.

[0088] S2. Construct a new index according to the matrix ill-conditioning theory and the space analysis theory. The construction method is as follows:

[0089] (1) Let the identification matrix A of obtained from robot modeling and point selection. Assume m≥n. When A is column full rank, A can be subjected to elementary transformation to be transformed into a trapezoidal matrix M, and then the first n rows of M are taken to obtain an n-dimensional square matrix N, which is called the information matrix.

[0090] (2) Perform Gram-Schmidt orthogonalization (hereinafter referred to as G-S orthogonalization) on the column vectors of the information matrix N and normalize them to obtain matrix G.

[0091] Gram-Schmidt: Let be a finite or countable number of linearly independent vectors in the inner product space H. Then there must be an orthonormal system in H such that for each natural number , is a linear combination of is a linear combination of. Such except for a constant factor with an absolute value of 1, is completely determined by completely determined.

[0092] Express the column vectors in G as a set of bases for the parameter space , and the volume formed by the basis vectors is , and this space is an n-dimensional standard space;

[0093] (9)

[0094] (3) Normalize the row vectors of the information matrix N to obtain the matrix N0, and perform the G-S orthogonalization on N0 to obtain the matrix N a , and the row vectors of N a are expressed as , and calculate the volume U of the hyper-parallel polyhedron that the information matrix G can expand n , then the metric index of the identification matrix can be expressed as:

[0095] (10)

[0096] That is the new index value.

[0097] Combined with Figure 2 , taking a certain information matrix as an example, assume that its three edge lengths are all 0.6. In the left diagram, the edges of the green polyhedron are perpendicular to each other, so its volume reaches the maximum value of 0.216. In the middle diagram, the included angle between the two edges of the bottom surface of the blue polyhedron is an acute angle, resulting in a slightly smaller volume compared to the green polyhedron, which is 0.187. Finally, in the right diagram, the included angles between all the edges of the red polyhedron are acute angles, which makes the volume reach the minimum value of 0.162.

[0098] S3. As Figure 3 shown, the process of the genetic algorithm for configuration optimization using the new index as the fitness function is as follows:

[0099] (1) Individual coding: After discretizing the workspace according to the specified spatial coordinate interval and attitude angle interval, use binary coding for the pose combination of the robot, and use it as a biological individual. Select n points in the workspace in each pose combination as internal genes.

[0100] (2) Set the relevant parameters of the genetic algorithm: such as the initial population size, dynamic crossover probability, dynamic mutation probability, crossover point, elimination method, number of iterations or algorithm termination conditions, etc.

[0101] (3) Set the fitness function: The fitness functions of the genetic algorithm respectively select the new evaluation index and the condition number , and it is considered that the robot pose combination with a larger new evaluation index or a smaller condition number is more conducive to the parameter identification of kinematic calibration.

[0102] (4) Generate the initial population, and calculate the identification matrix corresponding to each organism in the population to calculate the fitness function.

[0103] (5) In the population, generate corresponding probabilities according to the fitness function to select the parent generation for crossover, generate the offspring, and then perform gene mutation on the offspring according to the mutation function to form the overall new generation population.

[0104] (6) Iteration count +1 until the upper limit of the iteration count is exceeded.

[0105] (7) Obtain the population or individual with the best fitness, and use the pose combination corresponding to this individual as the basis for parameter identification.

[0106] Combined with Figure 4 and Figure 5 , they are respectively the iteration diagrams of the genetic algorithm obtained with the condition number and the new index as the fitness function. The biological individuals in the genetic algorithm are a set of configuration combinations containing 40 different points. The number of the biological population is 200. The initial crossover probability of the genetic algorithm is 0.75, and the number of iterations is 100 times. When the change degree of the fitness function is less than 1.0×10 -6 for 80 consecutive times during the algorithm process or after reaching the maximum number of iterations, it is considered that the optimal configuration is found, and the algorithm stops after the number of iterations reaches 200 times. The optimal condition number obtained by the genetic algorithm is 2.3914×10 7 ; the optimal new index value obtained by the genetic algorithm is 0.7276.

[0107] As the number of iterations increases, the mean value of the condition number of the population changes violently, indicating that using the condition number as the fitness function of the genetic algorithm cannot effectively ensure the stable evolution of the population. Since the mutation function in the genetic algorithm maintains the diversity of the species, the mean value of the last generation is not necessarily the minimum mean value during the iteration process.

[0108] Combined with Figure 6 , the root mean square error between the estimated value of the robot end error and the true value obtained by verifying the optimal configuration combination obtained by the genetic algorithm with the new index as the fitness function through randomly generated verification configurations is smaller than the result obtained by the optimal configuration combination obtained by the genetic algorithm with the condition number as the fitness.

[0109] S4. The error compensation method for the verified configuration of the robot is as follows: Using the new index configuration optimization combination as the parameter identification matrix and the least squares method as the identification method for the geometric error parameters of the robot, substitute them into the optimal configuration and the verified configuration respectively, and calculate the estimated value of the end pose error. The equivalent motion error of the driving joints is obtained through the mapping matrix for the end attitude error. The equivalent motion error is actually the rod length error of each driving joint, and its opposite number is the error compensation amount. :

[0110] (11)

[0111] Among them, is the driving Jacobian matrix of the 6-UPS six-degree-of-freedom parallel robot, is the identification matrix, is the geometric error source combination.

[0112] Combine the driving joint variables with the compensation amount and add them to the control system, and then use a laser tracker to measure the end pose information of the compensated pose. Finally, analyze and compare the errors of the six degrees of freedom and the overall error before and after compensation for the optimal configuration obtained based on the new index and the optimal configuration obtained based on the condition number.

[0113] Combined with Figure 7 and Figure 8 , through comparison, the compensation result of the spatial coordinates based on the new index is significantly better than that based on the condition number, and good optimization effects can be obtained for a large number of points; while the compensation result of the attitude angle based on the new index does not have an obvious advantage compared with that based on the condition number, and there are differences in the optimization effects between different points.

[0114] Therefore, the present invention adopts the above-mentioned optimal index for precision measurement configuration of the numerical control equipment and its kinematic calibration method. First, establish a geometric error transfer model of the target robot based on the POE method and eliminate redundant parameters, then use the proposed new index as the fitness function of the genetic algorithm to find the optimal configuration, and finally use the regularization method for parameter estimation. Compare and analyze the estimation results with the observability index represented by the condition number to reveal and verify the regularity and superiority of the new index.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and such modifications or equivalent replacements do not make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A kinematic calibration method for numerical control equipment, characterized in that, It includes the following steps: S1. Establish the geometric error transfer model of the target robot and screen the parameters to be identified; S2. Establish a new index based on the matrix ill-condition theory and the spatial analysis theory; S3. Use the new index as the fitness function of the genetic algorithm to obtain the optimal configuration for parameter identification; S4. Use the regularization method to obtain the identification results of the geometric error parameters; S5. Select the verification configuration in the full working area for error compensation; The construction process of the new index in step S2 includes: Let the identification matrix A of obtained by robot modeling and point selection. Assume m≥n. When A has full column rank, perform elementary row operations on A to transform it into a trapezoidal matrix M. Then take the first n rows of M to obtain an n×n square matrix N, which is called the information matrix. The information matrix N is orthonormalized column by column using the Gram - Schmidt method and then normalized to obtain the matrix G: The Gram - Schmidt orthonormalization is as follows: Let be a finite or countable set of linearly independent vectors in the inner product space H. Then there must exist an orthonormal system in H such that for each natural number , , is a linear combination of , and is a linear combination of . Represent the column vectors in G as a set of bases for the parameter space , and the volume formed by the basis vectors is , and this space is an n-dimensional standard space: ； Normalize the row vectors of the information matrix N to obtain the matrix N0, and perform Gram - Schmidt orthogonalization on N0 to obtain the matrix N a , N a The row vectors of are expressed as , and calculate the volume U of the hyper - parallelepiped that the information matrix G can expand n , then the metric index of the identification matrix is expressed as :[[]]END]] 。 2. The kinematic calibration method of the numerical control equipment according to claim 1, characterized in that Step S1 specifically includes: S11. Establish the geometric error transfer model of the robot based on the POE method; S12. Eliminate the redundant parameters from all the error parameters of the robot.

3. The kinematic calibration method of the numerical control equipment according to claim 2, characterized in that, Step S2 specifically includes: S21. Conduct matrix ill-condition analysis on the error transfer Jacobian matrix of the robot; S22. Expand the error transfer Jacobian matrix of the robot in the high-dimensional space and endow it with the geometric meaning of taking the volume of the hyper-parallel polyhedron as the index value.

4. The kinematic calibration method of the numerical control equipment according to claim 3, characterized in that: In step S12, the redundant parameters are eliminated from the error transfer Jacobian matrix of the established overall geometric error transfer model of the whole machine through column correlation analysis.

5. The kinematic calibration method of the numerical control equipment according to claim 1, characterized in that: In step S3, the calculation method of the new index is used as the fitness function, and the genetic algorithm is used as the evolutionary algorithm for configuration optimization. The genetic algorithm has a dynamically updated crossover probability and genetic probability.

6. The kinematic calibration method of the numerical control equipment according to claim 1, characterized in that: In step S4, the generalized cross-validation criterion is adopted to optimize the regularization parameter, and the Tikhonov regularization method is used to identify the geometric error parameters.

7. The kinematic calibration method of the numerical control equipment according to claim 1, characterized in that: Step S5 uses the new index configuration optimal combination as the parameter identification matrix and the regularization method as the identification method for the geometric error parameters of the robot. They are respectively substituted into the optimal configuration and the verification configuration to calculate the estimated value of the end pose error. , and the end attitude error is used to obtain the equivalent motion error of the driving joints through the mapping matrix. , the equivalent motion error is the rod length error amount of each driving joint, and its opposite number is the error compensation amount. .

8. The kinematic calibration method of the numerical control equipment according to claim 5, characterized in that The specific process of the genetic algorithm is: Individual coding: After discretizing the working space according to the specified spatial coordinate interval and attitude angle interval, use binary coding for the pose combination of the robot, and use it as a biological individual. Select a points in the working space in each pose combination as internal genes; Set the relevant parameters of the genetic algorithm; Set the fitness function: The fitness functions of the genetic algorithm respectively select new evaluation indicators ; Generate the initial population, calculate the identification matrix corresponding to each organism in the population to calculate the fitness function; In the population, generate the corresponding probability according to the fitness function to select the parent generation for crossover, generate the offspring, and then perform gene mutation on the offspring according to the mutation function to form a new generation of population as a whole; Iteration times +1 until the upper limit of the iteration times is exceeded; Obtain the population or individual with the best fitness, and use the pose combination corresponding to this individual as the basis for parameter identification.

Images