Optimal selection method for calibration sampling data of articulated arm type coordinate measuring machine
Through the optimization and selection method of calibration sampling data based on D-optimal criterion, the problem that the calibration results of the joint arm coordinate measuring machine are affected by unmodeled errors is solved, the measurement accuracy is improved, and the effective reduction of unmodeled errors is achieved. .
Patent Information
- Application Number
- CN202411004298.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-11-12
AI Technical Summary
The calibration results of the articular arm coordinate measuring machine are affected by unmodeled errors, resulting in low measurement accuracy, and it is difficult for the prior art to effectively reduce the impact of these errors on the calibration results.
The calibration sampling data optimization selection method based on the D-optimal criterion is adopted. By establishing a D-H measurement model and an error calibration model based on spatial distance error, the calibration sampling data is optimized by using a search algorithm to reduce the impact of unmodeled error on the calibration results.
The calibration quality of the joint arm coordinate measuring machine is improved, the impact of unmodeled errors on calibration results is reduced, and the measurement accuracy is enhanced.
Smart Images

Figure CN118936389A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of measuring instruments, and in particular is a method for optimizing the selection of calibration sampling data of an articulated arm coordinate measuring machine based on a D-optimal criterion. Background Art
[0002] According to the structural characteristics of the articulated arm coordinate measuring machine, the DH kinematic model is often used to mathematically model it. The coordinates of the probe relative to the base coordinate system are calculated by modeling to achieve the dimensional measurement of parts. Compared with the traditional orthogonal coordinate measuring machine, the articulated arm coordinate measuring machine has the advantages of small size, simple operation, and large measurement range, but its measurement accuracy is lower than that of the orthogonal coordinate measuring machine. The main reason is that the structural error of the articulated arm coordinate measuring machine will be gradually amplified with the connecting rod. Therefore, the structural parameters of the articulated arm coordinate measuring machine need to be calibrated before actual use.
[0003] It is generally believed that the measurement error of the articulated arm coordinate measuring machine is related not only to its own structural parameter error, but also to other unmodeled errors such as force deformation, thermal deformation, connection shake, etc. Since these errors are usually small, they are usually not specially calibrated, but these errors still exist objectively and have an adverse effect on the measurement or calibration results.
[0004] Since the error calibration model of the articulated arm coordinate measuring machine can be regarded as a linear regression model, the optimal experimental design theory can be introduced to optimize the calibration sampling data. The core idea of the optimal experimental design theory is to achieve optimal regression fitting by reasonably arranging the test points, thereby reducing the impact of random errors on the fitting results. Therefore, the calibration sampling data can be optimized based on the D-optimal criterion in the optimal experimental design to reduce the impact of unmodeled errors on the calibration results.
[0005] Based on the above-mentioned deficiencies in the prior art, the present invention is proposed. Summary of the invention
[0006] Aiming at the problem that the calibration results of an articulated arm coordinate measuring machine are affected by unmodeled errors, the present invention proposes an optimization selection method for calibration sampling data of an articulated arm coordinate measuring machine based on the D-optimal criterion. The method can evaluate the influence of the currently selected calibration sampling data on the calibration results, and search for optimized calibration sampling data through a search algorithm, thereby reducing the influence of unmodeled errors in the calibration model of the articulated arm coordinate measuring machine on the calibration results and improving the calibration quality.
[0007] In order to achieve the purpose of the invention, the technical solution provided by the present invention is:
[0008] A method for optimizing the selection of calibration sampling data for an articulated arm coordinate measuring machine comprises the following steps:
[0009] Step S1: According to the nominal structural parameters of the articulated arm coordinate measuring machine, establish the DH measurement model of the articulated arm coordinate measuring machine and the error calibration model based on the spatial distance error, and record the DH structural parameter vector
[0010] Step S2, obtaining calibration sampling data of the error calibration model based on the spatial distance error in step S1;
[0011] Step S21: Fixing the articulated arm coordinate measuring machine and a custom quartz rod on an optical platform;
[0012] Wherein, the customized quartz rod has cone sockets fixedly connected at both ends of the quartz rod, and the distance between the cone sockets at both ends is known;
[0013] Step S22: placing a probe of the articulated arm coordinate measuring machine in one of the first conical sockets on the customized quartz rod;
[0014] Step S23: Move each joint of the articulated arm coordinate measuring machine to ensure that the probe of the articulated arm coordinate measuring machine is closely fitted with the first cone socket, and continuously collect each joint angle of the articulated arm coordinate measuring machine through the host computer and save it as joint angle data θ C1 ;
[0015] Step S24: Place the probe of the articulated arm coordinate measuring machine in another second cone socket on the customized quartz rod, and repeat the operation of step S23 to obtain the joint angle data θ C2 ;
[0016] The joint angle data θ C2 Should be C1 The amount of data should be equal;
[0017] Step S25: Finally, the calibration sampling data θ is obtained C ;
[0018] Step S3, setting the preferred reserved data amount N and the initial preferred reserved data θ b,0 ;
[0019] The initial preferred retention data θ b,0 , is the C Randomly select N groups of calibration sampling data to form θ b,0 , the remaining calibration sampling data is θ r,0 ;
[0020] Step S4, determine whether the termination condition is met, and if so, output the current calibration sampling data θ b,i ;
[0021] The termination condition is: determine whether the number of iterations reaches the maximum value. If it reaches the maximum value, output θ b,i ;
[0022] The current calibration sampling data θ b,i , the subscript i represents the current iteration number;
[0023] If the termination condition is not met, proceed to step S5;
[0024] Step S5, select θ r,i The jth group of calibration sampling data in θ is put into b,i The composition of θ b,i,j , then θ r,i Composition θ r,i,j , for θ b,i,j Calculate the evaluation d(θ b,i,j );Then r,i Do this for each set of data in the b,i,j ) the largest set of data, and get θ b,i,max and θ r,i,max ;
[0025] The θ b,i,max , which means θ r,i In the middle, d(θ b,i,j )The largest set of data is put into θ b,i The data constituted in
[0026] The θ r,i,max , which means θ r,i Eliminate d(θ b,i,j )The data consisting of the largest set of data;
[0027] Step S6, select θ b,i,max The jth group of calibration sampling data in θ is put into r,i,max The composition of θ r,i,max,j , then θ b,i,max Composition θ b,i,max,j , for θ b,i,max,j Calculate the evaluation d(θ b,i,max,j );Then b,i,max Perform the above operation on each set of data in and find the one that makes d(θ b,i,max,j ) the largest set of data, and get θ b,i+1 and θ r,i+1 ;
[0028] The θ r,i+1 , which means θ b,i,max In the middle, d(θ b,i,max,j )The largest set of data is put into θ r,i,max The data constituted in
[0029] The θ b,i+1 , which means θ b,i,max Eliminate d(θ b,i,max,j )The data consisting of the largest set of data;
[0030] The number of iterations is increased by one, and the process goes to step S4.
[0031] The preferred technical solution provided by the present invention is:
[0032] In step S1, the DH measurement model is characterized by describing the spatial posture transformation relationship between two adjacent coordinate systems fixed on the joint in space through four parameters, which are as follows:
[0033] Connecting rod length l k : The shortest distance between two adjacent joint axes, that is, the length of the common perpendicular line between the two axes;
[0034] Connecting rod torsion angle α k : Move the joint axis toward another adjacent joint axis, and the angle formed when they intersect is the joint torsion angle;
[0035] Joint offset d k :Connecting rod length line l k and l k-1 The distance intercepted on the axis of the kth joint;
[0036] Joint rotation angle θ k : The change in relative position between adjacent links connected by revolute joints;
[0037] The spatial posture transformation matrix A between the two coordinate systems can be obtained through the above four parameters. k as follows:
[0038]
[0039] Considering the structure of the actual articulated arm coordinate measuring machine, the DH measurement model can be obtained as follows:
[0040]
[0041] The above k represents the serial number of the coordinate system fixed on the articulated arm coordinate measuring machine, and k=1,2,…,7.
[0042] The further preferred technical solution provided by the present invention is:
[0043] In step S1, in step S1, the structural parameter vector is the column vector composed of various DH structural parameters in the DH measurement model, specifically:
[0044]
[0045] In the formula, [Δθ 0,1 ,…,Δθ 0,6 ] T represents the initial phase angle of each joint, l prob Indicates the probe length.
[0046] A further preferred technical solution provided by the present invention is:
[0047] In step S1, the spatial distance-based error calibration model is described as the distance measurement error Δy and parameter error vector of the articulated arm coordinate measuring machine. The relationship between them is as follows:
[0048]
[0049] Where J is the Jacobian matrix.
[0050] The specific decomposition formula is as follows:
[0051] The articulated arm coordinate measuring machine is used to measure the first cone socket to obtain a set of joint angle data. A set of joint angle data is obtained by measuring at the second cone socket Then the measured distance between the two cones can be obtained as follows:
[0052]
[0053] Since there are errors between the structural parameters of the articulated arm coordinate measuring machine and the theoretical values, it is assumed that the structural parameter error vector is as follows:
[0054]
[0055] Then the actual distance between the two cones can be expressed as:
[0056]
[0057] Therefore there is a distance measurement error:
[0058]
[0059] Or it can be expressed as:
[0060]
[0061] Another preferred technical solution provided by the present invention is:
[0062] In step S3, the preferred amount of retained data is set to 70% of the amount of data acquired in step S2.
[0063] Another further preferred technical solution provided by the present invention is:
[0064] In the step S5, the b,i,j Calculate the evaluation d(θ b,i,j ), where evaluation d(θ b,i,j ) specifically:
[0065] Set as: Evaluation d(θ), to evaluate the current calibration sampling data by D-optimal criterion, the calculation process is as follows:
[0066]
[0067] d(θ)=|M|
[0068] J is the Jacobian matrix calculated by θ, which can be expressed as:
[0069]
[0070] J i (i=1,2,…,N) represents the Jacobian matrix calculated for the i-th group of calibration sampling data;
[0071] M is defined as an information matrix, which is calculated by the Jacobian matrix J and the preferred retained data amount N; the current calibration sampling data evaluation d(θ) is obtained by calculating the determinant |M| of M.
[0072] Another further preferred technical solution provided by the present invention is:
[0073] The determinant |M| of the information matrix M is:
[0074] Parameter Error Vector The solution of is generally obtained by the least squares method. From multivariate statistical analysis, it can be obtained that for a given confidence level α, the confidence region of (1-α) of all parameter error vectors is:
[0075]
[0076] Where p represents the dimension of the parameter error vector, s 2 is the random error variance σ 2 It is believed that the random error is mainly caused by other unmodeled errors in the articulated arm coordinate measuring machine and exists objectively and cannot be changed. p,n-p-1,α represents the α quantile on the F distribution with p and np-1 degrees of freedom; under the normal assumption, given a positive number C = ps 2 F p,n-p-1,α , so we have:
[0077]
[0078] The image of this equation in space is a The ellipsoid centered on is called the confidence ellipsoid, and the volume of the confidence ellipsoid is:
[0079]
[0080] Where V p represents the volume of the p-dimensional unit hypersphere; therefore, when the confidence level α and the dimension p of the parameter error vector are given, the volume of the confidence ellipsoid is only related to the determinant |M| of the information matrix M, and the value of |M| is only related to the distribution of the calibration sampling data in space under the premise of a given calibration model. The volume of the confidence ellipsoid can be reduced by reasonably arranging the calibration points or collecting a large amount of data in space and searching for the optimal calibration sampling data to increase |M|. The least squares solution of the parameter error vector is It is more likely to be close to the actual parameter error vector, thereby improving the calibration accuracy of the parameter error vector and reducing the impact of the unmodeled error on the calibration result.
[0081] Compared with the prior art, the present invention has the following beneficial effects:
[0082] Aiming at the problem that the calibration results of articulated arm coordinate measuring machines are affected by unmodeled errors, the present invention proposes an optimization selection method for calibration sampling data of articulated arm coordinate measuring machines based on the D-optimal criterion. The method can evaluate the influence of the currently selected calibration sampling data on the calibration results, and search for optimized calibration sampling data through a search algorithm, thereby reducing the influence of unmodeled errors in the calibration model of the articulated arm coordinate measuring machine on the calibration results and improving the calibration quality. The present invention applies the optimal experimental design theory to the calibration process of the articulated arm coordinate measuring machine, optimizes the calibration sampling data based on the spatial distance error calibration model, realizes the extended application of the theory, and provides a reference for the practice of other calibration models. BRIEF DESCRIPTION OF THE DRAWINGS
[0083] Figure 1 It is a schematic diagram of the structure of the articulated arm coordinate measuring machine calibration system applied in the present invention;
[0084] Figure 2 is a flow chart of the steps of the present invention;
[0085] Figure 3 Schematic diagram of the spatial posture transformation relationship between two adjacent coordinate systems fixed on the joint. DETAILED DESCRIPTION
[0086] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0087] Figure 1 It is a schematic diagram of the structure of the articulated arm coordinate measuring machine calibration system applied in the present invention; Figure 2 FIG. 1 is a flow chart of the steps of the present invention. Figure 1 as well as Figure 2 The method for optimizing the selection of calibration sampling data of an articulated arm coordinate measuring machine based on the D-optimal criterion of the present invention is described as follows:
[0088] like Figure 1 As shown, the calibration system of the articulated arm coordinate measuring machine 1 applied in the present invention comprises an optical platform (not shown in the figure), an articulated arm coordinate measuring machine 1 and a special quartz rod 2 on the optical platform, wherein two conical sockets 3 are fixedly connected at both ends of the customized quartz rod 2, and the spatial distance between the two conical sockets is known.
[0089] Combination Figure 2 The specific steps of the present invention are described as follows:
[0090] Step S1: According to the nominal structural parameters of the articulated arm coordinate measuring machine, establish the DH measurement model of the articulated arm coordinate measuring machine and the error calibration model based on the spatial distance error, and record the DH structural parameter vector
[0091] The DH measurement model describes the spatial posture transformation relationship between two adjacent coordinate systems fixed on the joint in space through four parameters (such as Figure 3 ), as follows:
[0092] Connecting rod length l k : The shortest distance between two adjacent joint axes, that is, the length of the common perpendicular line between the two axes;
[0093] Connecting rod torsion angle α k : Move the joint axis toward another adjacent joint axis, and the angle formed when they intersect is the joint torsion angle;
[0094] Joint offset d k :Connecting rod length line l k and l k-1 The distance intercepted on the axis of the kth joint;
[0095] Joint rotation angle θ k : The change in relative position between adjacent links connected by revolute joints.
[0096] Through the above four parameters, the spatial posture transformation matrix between the two coordinate systems can be obtained as follows:
[0097]
[0098] Considering the structure of the actual articulated arm coordinate measuring machine, its measurement model can be obtained as follows:
[0099]
[0100] The structural parameter vector The characteristics are as follows:
[0101]
[0102] In the formula, [Δθ 0,1 ,…,Δθ 0,6 ] T represents the initial phase angle of each joint, l prob Indicates the probe length.
[0103] The above k represents the serial number of the coordinate system fixed on the articulated arm coordinate measuring machine, and k = 1, 2, ..., 7;
[0104] The error calibration model based on spatial distance error is described as the relationship between the distance measurement error and parameter error of the articulated arm coordinate measuring machine as follows:
[0105] Suppose the articulated arm coordinate measuring machine is used to measure the first cone socket to obtain a set of joint angle data. A set of joint angle data is obtained by measuring in another cone socket Then the measured distance between the two cones can be obtained as follows:
[0106]
[0107] Since there are errors between the structural parameters of the articulated arm coordinate measuring machine and the theoretical values, it is assumed that the structural parameter error vector is as follows:
[0108]
[0109] Then the actual distance between the two cones can be expressed as:
[0110]
[0111] Therefore there is a distance measurement error:
[0112]
[0113] Or it can be expressed as:
[0114]
[0115] Step S2: Obtain calibration sampling data of the error calibration model based on spatial distance error.
[0116] Step S21: Fix the articulated arm coordinate measuring machine and a custom quartz rod on the optical platform.
[0117] The customized quartz rod has two cone sockets fixedly connected at both ends of the quartz rod, and the distance between the two cone sockets is known.
[0118] Step S22: placing the probe of the articulated arm coordinate measuring machine in the first conical socket on the quartz rod.
[0119] Step S23: Fully move each joint of the articulated arm coordinate measuring machine to ensure that the probe of the articulated arm coordinate measuring machine fits tightly with the cone socket, and continuously collect each joint angle through the host computer and save it as joint angle data θ C1 .
[0120] The joint angle data θ C1 , specifically expressed as:
[0121]
[0122] in It refers to the column vector consisting of the angle data of the mth joint of the articulated arm coordinate measuring machine.
[0123] If the host computer saves n sets of joint angles, then θ C1 It is an n*6 matrix.
[0124] Step S24: Similarly, place the probe of the articulated arm coordinate measuring machine in the second cone socket on the quartz rod, repeat the operation of step S23, and obtain the joint angle data θ C2 Finally, we get the calibration data θ C :
[0125] θ C =[θ C1 ,θ C2 ]
[0126] The joint angle data θ C2 Should be C1 It is also an n*6 matrix, that is, the amount of data collected by the articulated arm coordinate measuring machine on the two cone sockets should be equal.
[0127] Step S3: Set the preferred reserved data volume N and the initial preferred reserved data θ b,0 .
[0128] The preferred reserved data volume is generally set to 70% of the data volume acquired in step S2 or the data volume is set based on experience.
[0129] The initial preferred retention data, at θ C Randomly select N groups of calibration sampling data to form θ b,0 , the remaining calibration sampling data is θ r,0 . Where θ b,0 is an N*6 matrix, θ r,0 It is a (nN)*6 matrix.
[0130] Step S4: Determine whether the termination condition is met. If the termination condition is met, output the current calibration sampling data θ b,i .
[0131] The termination condition is to determine whether the number of iterations reaches the maximum value. If it reaches the maximum value, the output θ b,i .
[0132] The current calibration sampling data θ b,i , where i in the subscript represents the current iteration number.
[0133] Step S5:
[0134] Select θ r,i The jth group of calibration sampling data in θ is put into b,i The composition of θ b,i,j , then θ r,i Composition θ r,i,j , for θ b,i,j Calculate the evaluation d(θ b,i,j );
[0135] Sequentially r,i Do this for each set of data in the b,i,j ) the largest set of data, and get θ b,i,max and θ r,i,max .
[0136] The evaluation d(θ) is to evaluate the current calibration sampling data by the D-optimal criterion, and the calculation process is as follows:
[0137]
[0138] d(θ)=|M|
[0139] J is the Jacobian matrix calculated by θ, which can be expressed as:
[0140]
[0141] J i(i=1,2,…,N) represents the Jacobian matrix calculated for the i-th group of calibration sampling data.
[0142] M is defined as an information matrix, which is calculated by the Jacobian matrix J and the preferred retained data amount N; the current calibration sampling data evaluation d(θ) is obtained by calculating the determinant |M| of M.
[0143] The determinant of the information matrix M is |M|, characterized in that the parameter error vector The solution of is generally obtained by the least squares method. From multivariate statistical analysis, it can be obtained that for a given confidence level α, the confidence region of (1-α) of all parameter error vectors is:
[0144]
[0145] Where p represents the dimension of the parameter error vector, s 2 is the random error variance σ 2 It is believed that the random error is mainly caused by other unmodeled errors in the articulated arm coordinate measuring machine and exists objectively and cannot be changed. p,n-p-1,α represents the α quantile on the F distribution with p and np-1 degrees of freedom. Under the normal assumption, given a positive number C = ps 2 F p,n-p-1,α , so we have:
[0146]
[0147] The image of this equation in space is a The ellipsoid centered on is called the confidence ellipsoid, and the volume of the confidence ellipsoid is:
[0148]
[0149] Where V p represents the volume of the p-dimensional unit hypersphere. Therefore, when the confidence level α and the dimension p of the parameter error vector are given, the volume of the confidence ellipsoid is only related to the determinant |M| of the information matrix M, and the value of |M| is only related to the distribution of the calibration sampling data in space under the premise of a given calibration model. By reasonably arranging the calibration points or collecting a large amount of data in space and searching for the optimal calibration sampling data to increase |M|, the volume of the confidence ellipsoid can be reduced, and the least squares solution of the parameter error vector is It is more likely to be close to the actual parameter error vector, thereby improving the calibration accuracy of the parameter error vector and reducing the impact of the unmodeled error on the calibration result.
[0150] The θ b,i,max , which means θ r,i In the middle, d(θ b,i,j)The largest set of data is put into θ b,i The data constructed in .
[0151] The θ r,i,max , which means θ r,i Eliminate d(θ b,i,j ) is the data consisting of the largest group of data.
[0152] Step S6:
[0153] Select θ b,i,max The jth group of calibration sampling data in θ is put into r,i,max The composition of θ r,i,max,j , then θ b,i,max Composition θ b,i,max,j , for θ b,i,max,j Calculate the evaluation d(θ b,i,max,j );
[0154] Sequentially b,i,max Do this for each set of data in the b,i,max,j ) the largest set of data, and get θ b,i+1 and θ r,i+1 .
[0155] The θ r,i+1 , which means θ b,i,max In the middle, d(θ b,i,max,j )The largest set of data is put into θ r,i,max The data constructed in .
[0156] The θ b,i+1 , which means θ b,i,max Eliminate d(θ b,i,max,j ) is the data consisting of the largest group of data.
[0157] The number of iterations is increased by one, and the process goes to step S4.
[0158] Aiming at the problem that the calibration result of an articulated arm coordinate measuring machine is affected by the unmodeled error, the present invention proposes an optimization selection method for calibration sampling data of an articulated arm coordinate measuring machine based on the D-optimal criterion.
[0159] This method can evaluate the impact of the currently selected calibration sampling data on the calibration results.
[0160] The optimized calibration sampling data is searched through the search algorithm.
[0161] This reduces the influence of unmodeled errors in the calibration model of the articulated arm coordinate measuring machine on the calibration results and improves the calibration quality.
[0162] The present invention applies the optimal experimental design theory to the calibration process of an articulated arm coordinate measuring machine, optimizes the calibration sampling data based on a spatial distance error calibration model, realizes the extended application of the theory, and provides a reference for the practice of other calibration models.
[0163] Since the error calibration model of the articulated arm coordinate measuring machine can be regarded as a linear regression model, the optimal experimental design theory can be introduced to optimize the calibration sampling data. The optimal experimental design theory is used to find the optimal data based on the D-optimal criterion in a large amount of calibration sampling data to achieve optimal regression fitting, thereby reducing the impact of unmodeled errors on the fitting results.
[0164] Although the functions and working processes of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the above-mentioned specific functions and working processes. The above-mentioned specific implementation methods are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which are all within the protection of the present invention.
Claims
1. A method for optimizing the selection of calibration sampling data for an articulated arm coordinate measuring machine, characterized in that: The following steps are involved: Step S1: According to the nominal structural parameters of the articulated arm coordinate measuring machine, establish the DH measurement model of the articulated arm coordinate measuring machine and the error calibration model based on the spatial distance error, and record the DH structural parameter vector Step S2, obtaining calibration sampling data of the error calibration model based on the spatial distance error in step S1; Step S21: Fixing the articulated arm coordinate measuring machine and a custom quartz rod on an optical platform; Wherein, the customized quartz rod has cone sockets fixedly connected at both ends of the quartz rod, and the distance between the cone sockets at both ends is known; Step S22: placing a probe of the articulated arm coordinate measuring machine in one of the first conical sockets on the customized quartz rod; Step S23: Move each joint of the articulated arm coordinate measuring machine to ensure that the probe of the articulated arm coordinate measuring machine is closely fitted with the first cone socket, and continuously collect the angles of each joint of the articulated arm coordinate measuring machine through the host computer and save them as joint angle data θ C1 ; Step S24: Place the probe of the articulated arm coordinate measuring machine in another second cone socket on the customized quartz rod, and repeat the operation of step S23 to obtain the joint angle data θ C2 ; The joint angle data θ C2 Should be C1 The amount of data should be equal; Step S25: Finally, the calibration sampling data θ is obtained C ; Step S3, setting the preferred reserved data amount N and the initial preferred reserved data θ b,0 ; The initial preferred retention data θ b,0 , is the C Randomly select N groups of calibration sampling data to form θ b,0 , the remaining calibration sampling data is θ r,0 ; Step S4, determine whether the termination condition is met, and if so, output the current calibration sampling data θ b,i ; The termination condition is: determine whether the number of iterations reaches the maximum value. If it reaches the maximum value, output θ b,i ; The current calibration sampling data θ b,i , the subscript i represents the current iteration number; If the termination condition is not met, proceed to step S5; Step S5, select θ r,i The jth group of calibration sampling data in θ is put into b,i The composition of θ b,i,j , then θ r,i Composition θ r,i,j , for θ b,i,j Calculate the evaluation d(θ b,i,j );Then r,i Do this for each set of data in the b,i,j ) the largest set of data, and get θ b,i,max and θ r,i,max ; The θ b,i,max , which means θ r,i In the middle, d(θ b,i,j )The largest set of data is put into θ b,i The data constituted in The θ r,i,max , which means θ r,i Eliminate d(θ b,i,j )The data consisting of the largest set of data; Step S6, select θ b,i,max The jth group of calibration sampling data in θ is put into r,i,max The composition of θ r,i,max,j , then θ b,i,max Composition θ b,i,max,j , for θ b,i,max,j Calculate the evaluation d(θ b,i,max,j );Then b,i,max Perform the above operation on each set of data in and find the one that makes d(θ b,i,max,j ) the largest set of data, and get θ b,i+1 and θ r,i+1 ; The θ r,i+1 , which means θ b,i,max In the middle, d(θ b,i,max,j )The largest set of data is put into θ r,i,max The data constituted in The θ b,i+1 , which means θ b,i,max Eliminate d(θ b,i,max,j )The data consisting of the largest set of data; The number of iterations is increased by one, and the process goes to step S4.
2. The method for optimizing the selection of calibration sampling data for an articulated arm coordinate measuring machine according to claim 1, characterized in that: In step S1, the DH measurement model is characterized by describing the spatial posture transformation relationship between two adjacent coordinate systems fixed on the joint in space through four parameters, which are as follows: Connecting rod length l k : The shortest distance between two adjacent joint axes, that is, the length of the common perpendicular line between the two axes; Connecting rod torsion angle α k : Move the joint axis toward another adjacent joint axis, and the angle formed when they intersect is the joint torsion angle; Joint offset d k :Connecting rod length line l k and l k-1 The distance intercepted on the axis of the kth joint; Joint rotation angle θ k : The change in relative position between adjacent links connected by revolute joints; The spatial posture transformation matrix A between the two coordinate systems can be obtained through the above four parameters. k as follows: Considering the structure of the actual articulated arm coordinate measuring machine, the DH measurement model can be obtained as follows: The above k represents the serial number of the coordinate system fixed on the articulated arm coordinate measuring machine, and k=1,2,…,7.
3. The method for optimizing the selection of calibration sampling data for an articulated arm coordinate measuring machine according to claim 1, characterized in that: In step S1, in step S1, the structural parameter vector is the column vector composed of various DH structural parameters in the DH measurement model, specifically: In the formula, [Δθ 0,1 ,…,Δθ 0,6 ] T represents the initial phase angle of each joint, l prob Indicates the probe length.
4. The method for optimizing the selection of calibration sampling data for an articulated arm coordinate measuring machine according to claim 1, characterized in that: In step S1, the spatial distance-based error calibration model is described as the distance measurement error Δy and parameter error vector of the articulated arm coordinate measuring machine. The relationship between them is as follows: Where J is the Jacobian matrix.
5. The method for optimizing the selection of calibration sampling data for an articulated arm coordinate measuring machine according to claim 4, characterized in that: The distance measurement error Δy and parameter error vector of the articulated arm coordinate measuring machine The relationship between them is specifically decomposed into: The articulated arm coordinate measuring machine is used to measure the first cone socket to obtain a set of joint angle data. A set of joint angle data is obtained by measuring at the second cone socket Then the measured distance between the two cones can be obtained as follows: Since there are errors between the structural parameters of the articulated arm coordinate measuring machine and the theoretical values, it is assumed that the structural parameter error vector is as follows: Then the actual distance between the two cones can be expressed as: Therefore there is a distance measurement error: Or it can be expressed as:
6. The method for optimizing the selection of calibration sampling data for an articulated arm coordinate measuring machine according to claim 1, characterized in that: In step S3, the preferred amount of retained data is set to 70% of the amount of data acquired in step S2.
7. The method for optimizing the selection of calibration sampling data for an articulated arm coordinate measuring machine according to claim 1, characterized in that: In the step S5, the b,i,j Calculate the evaluation d(θ b,i,j ), where evaluation d(θ b,i,j ) specifically: Set as: Evaluation d(θ), to evaluate the current calibration sampling data by D-optimal criterion, the calculation process is as follows: d(θ)=|M| J is the Jacobian matrix calculated by θ, which can be expressed as: J i (i=1,2,…,N) represents the Jacobian matrix calculated for the i-th group of calibration sampling data; M is defined as an information matrix, which is calculated by the Jacobian matrix J and the preferred retained data amount N; the current calibration sampling data evaluation d(θ) is obtained by calculating the determinant |M| of M.
8. The method for optimizing the selection of calibration sampling data for an articulated arm coordinate measuring machine according to claim 7, characterized in that: The determinant |M| of the information matrix M is: Parameter Error Vector The solution of is generally obtained by the least squares method. From multivariate statistical analysis, it can be obtained that for a given confidence level α, the confidence region of (1-α) of all parameter error vectors is: Where p represents the dimension of the parameter error vector, s 2 is the random error variance σ 2 It is believed that the random error is mainly caused by other unmodeled errors in the articulated arm coordinate measuring machine and exists objectively and cannot be changed. p,n-p-1,α represents the α quantile on the F distribution with p and np-1 degrees of freedom; under the normal assumption, given a positive number C = ps 2 F p,n-p-1,α , so we have: The image of this equation in space is a The ellipsoid centered on is called the confidence ellipsoid, and the volume of the confidence ellipsoid is: Where V p represents the volume of the p-dimensional unit hypersphere; therefore, when the confidence level α and the dimension p of the parameter error vector are given, the volume of the confidence ellipsoid is only related to the determinant |M| of the information matrix M, and the value of |M| is only related to the distribution of the calibration sampling data in space under the premise of a given calibration model. The volume of the confidence ellipsoid can be reduced by reasonably arranging the calibration points or collecting a large amount of data in space and searching for the optimal calibration sampling data to increase |M|. The least squares solution of the parameter error vector is It is more likely to be close to the actual parameter error vector, thereby improving the calibration accuracy of the parameter error vector and reducing the impact of the unmodeled error on the calibration result.
