Machining parameter global optimization method and device based on machining deformation sparse key points

Abstract

The invention belongs to the related technical field of robot machining deformation parameter optimization, and discloses a machining parameter global optimization method and device based on machining deformation sparse key points. A spectral clustering method is adopted to obtain milling force sparse key points and rigidity sparse key points in the machining process; (2) optimizing and partitioning a machining area of the workpiece to be machined based on the obtained milling force sparse key point and rigidity sparse key point; and (3) optimizing parameters of the to-be-processed workpiece based on the obtained optimization partition result, the global optimization deformation index, the single-point milling force prediction model, the rigidity model, the milling force sparse key point and the rigidity sparse key point. The calculation cost is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field related to the optimization of robot processing deformation parameters, and more specifically, relates to a global optimization method and device for processing parameters based on sparse key points of processing deformation. Background Art

[0002] Complex surface parts, such as large-thickness blades like propeller blades, complex aerospace cabin sections, and hydroturbine runners, etc., are widely used in industries such as aerospace, energy, and national defense. Large-thickness blades like propeller blades have the characteristics of large thickness curvature changes, many overlapping regions, and sudden curvature changes at the blade tips. Moreover, due to their large size and complex structural shapes, the tool contact area changes during the cutting process, and the analytical calculation is difficult, the material removal fluctuates greatly, the force-induced deformation distribution during robot processing is complex, and it is difficult to characterize, which directly affects the machining accuracy of the normal vector of the blade surface. Therefore, it is particularly important to control the global machining deformation of complex surfaces such as propeller blades. Most of the existing studies focus on the deformation optimization of a single tool position point, lacking the establishment of global optimization indicators and global optimization strategies for large surfaces, and less optimizing processing parameters from a global perspective to reduce the overall deformation of the machined surface of the part. Moreover, there are many tool position points on large surfaces, the deformation calculation cost is high, the calculation time is long, and the cost required for optimization iteration is even higher.

[0003] Therefore, a global optimization method for processing parameters based on sparse key points of deformation is needed to optimize processing parameters considering the overall deformation of the workpiece to be machined, and provide guidance for the selection of actual processing parameters and the selection of parameters in different regions of large surfaces. Summary of the Invention

[0004] In view of the above defects or improvement requirements of the prior art, the present invention provides a global optimization method and device for processing parameters based on sparse key points of processing deformation, aiming to solve the problem of high cost caused by parameter optimization from the perspective of point positions in the prior art.

[0005] To achieve the above object, according to one aspect of the present invention, a global optimization method for processing parameters based on sparse key points of processing deformation is provided. The optimization method includes the following steps:

[0006] (1) Based on the single-point milling force prediction model and stiffness model of the workpiece to be machined, use the spectral clustering method to obtain the milling force sparse key points and stiffness sparse key points during the processing;

[0007] (2) Optimize and partition the processing area of the workpiece to be machined based on the obtained milling force sparse key points and stiffness sparse key points;

[0008] (3) Optimize the parameters of the workpiece to be machined based on the obtained results of the optimized partition, the global optimization deformation index, the single-point milling force prediction model, the stiffness model, the sparse key points of milling force, and the sparse key points of stiffness; among them, the expression of the global optimization deformation index is:

[0009]

[0010] In the formula: σ 2 is the uniformity index of the overall deformation distribution on the workpiece surface, Δ Dmean represents the overall deformation amount of the workpiece surface, Δ Di is the normal deformation value at the tool position of the sparse key point i; n is the number of sparse key points; Δ D = Φ(K kp (x), F kp (y)), where K kp (x) is the expression of the stiffness sparse key point, and F kp (y) is the expression of the milling force sparse key point.

[0011] Furthermore, the single-point milling force prediction model is a model that combines principal component analysis and Gaussian process regression to deal with the regression prediction problem of multi-dimensional data sets.

[0012] Furthermore, when constructing the single-point milling force prediction model, the principal component analysis method is used to reduce the dimension of the high-dimensional data of the workpiece to be machined during the machining process to achieve the compression of the high-dimensional data; among them, the compression of the high-dimensional data is carried out according to the cumulative contribution amount of the data; then Gaussian process regression is used to train and learn the dimension-reduced data to obtain the single-point milling force prediction model.

[0013] Furthermore, the Gaussian process is uniquely determined by the mean function m(x) and the covariance function k(x, x ′ ), and its data expression is:

[0014] f(x) ~ GP(m(x), k(x, x ′ ))

[0015] where m(x) is the mean function and k(x, x ′ ) is the covariance function to establish the PCA-GPR milling force prediction model.

[0016] Furthermore, when constructing the stiffness model, let τ be the joint balance force matrix, q be the joint differential deformation matrix, and K q be the joint stiffness matrix of the robot, and we get τ = K q q. According to the principle of virtual work, the virtual work done by the force F at the end of the robot is equal to the sum of the virtual work done by the torques of each joint of the robot:

[0017] F T Δ = τT δq

[0018]

[0019] C(q) is the stiffness model at the end of the robot, which represents the operational relationship between the force F and the deformation Δ in the end effector workspace, i.e.:

[0020]

[0021] Furthermore, taking the cutter location points in the machining area of the milling force or stiffness in the workpiece as clustering parameters, the spectral clustering method is used to cluster and segment the corresponding milling force or stiffness values, so as to obtain the milling force or stiffness sub-region segmentation boundary line in the machining area. The points on the segmentation boundary line are the sparse key points.

[0022] Furthermore, the robot machining deformation distribution is jointly determined by the sparse key points of the end stiffness and the milling force. By determining the sparse key points of the milling force and stiffness, the robot machining deformation distribution information is obtained, i.e.:

[0023] Δ D =Φ(K kp (x),F kp (y))

[0024] K kp (x) is the stiffness key point expression, and F kp (y) is the milling force key point expression.

[0025] Furthermore, first, the workpiece machining area is divided into k regions based on the boundary line composed of sparse key points. The Latin hypercube sampling method is used for population parameter initialization. The initialized population parameters are substituted into the global optimization deformation index for objective function calculation to evaluate the initial overall workpiece deformation. Then, an optimization algorithm is used to find m optimal Pareto solutions, different region machining parameter combinations are selected, the optimization results are calculated and analyzed, and thus the optimal machining parameter combination under the working conditions is obtained.

[0026] The present invention also provides a global machining parameter optimization system based on machining deformation key points. The system includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it executes the global machining parameter optimization method based on machining deformation sparse key points as described above.

[0027] The present invention also provides a computer-readable storage medium. The computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by the processor, the machine-executable instructions cause the processor to implement the global machining parameter optimization method based on machining deformation sparse key points as described above.

[0028] Generally speaking, compared with the prior art by the above technical solution conceived by the present invention, the global optimization method and device for machining parameters based on sparse key points of machining deformation provided by the present invention mainly have the following beneficial effects:

[0029] 1. The optimization partition and global optimization deformation index based on the sparse key points of milling force and stiffness proposed by the present invention replace all tool path points to be measured and calculated on the surface of the workpiece to be machined with sparse key points, and the data volume is compressed to 10%, solving the problems of a large number of tool path points on the surface of large curved surface workpieces, large global deformation calculation amount, high calculation cost, etc., and reducing the calculation complexity and calculation cost of the optimization objective function.

[0030] 2. The optimization objective is replaced from the deformation calculation of a single tool path point of the workpiece to the calculation of the mean and variance of the deformation of sparse key points with global deformation characteristics of the workpiece, and the optimization variables are replaced from the machining parameters of a single region to the machining parameters of a multi-region combination. The overall machining deformation distribution of the workpiece is optimized from the global perspective of the workpiece to be machined, solving the problem that existing research rarely optimizes machining parameters from the perspective of reducing the global deformation of the workpiece, and the situation of uneven deformation distribution on the surface of the workpiece to be machined, so that while the average value of the global deformation of large curved surface workpieces after optimization decreases, the uniformity is greatly improved. The machining parameters obtained by global optimization can provide guidance for actual machining.

[0031] 3. After optimization, the overall deformation average value decreases by 5.7%, and the uniformity increases by 28.9%. That is, after the partition optimization of multiple parameter variables, the overall machining deformation value of the curved surface workpiece decreases and the distribution becomes more uniform, and the comparison before and after optimization of different tool path trajectories shows good optimization effects; at the same time, the calculation complexity decreases to 10% (that is, the number of sparse key points to be calculated is 10% of the original points). Description of the Drawings

[0032] Figure 1 is a flowchart of a global optimization method for machining parameters based on sparse key points of machining deformation provided by the present invention;

[0033] Figure 2 is a distribution diagram of the obtained sparse key points of milling force and sparse key points of stiffness;

[0034] Figure 3 is a schematic diagram of the optimization partition of the workpiece;

[0035] Figure 4 In (a) and (b) of are respectively schematic diagrams of the shape of the curved surface workpiece to be machined and the deformation distribution of its original machining parameters;

[0036] Figure 5Among them, (1) and (2) are respectively the results trained by the NSGA-II multi-objective optimization model and the optimized machining deformation distribution diagrams corresponding to the selected Pareto solutions. Specific implementation manners

[0037] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various implementation manners of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0038] In the field of intelligent manufacturing of robot machining, computational intelligence technology and optimization methods are used to provide guidance for the selection of machining parameters. Computational intelligence technology and optimization methods solve complex optimization problems by simulating intelligent behaviors in nature, such as biological evolution, group cooperation and brain learning, etc., and are particularly suitable for scenarios that are difficult to handle by traditional methods. The core idea is to gradually find the optimal or near-optimal solution to the problem through iteration, evaluation and adjustment, and it has value for solving deformation optimization problems such as multi-objective functions and multi-parameter inputs.

[0039] Please refer to Figure 1 and Figure 2 , the present invention provides a global optimization method for machining parameters based on sparse key points of machining deformation, and the optimization method mainly includes the following steps:

[0040] Step 1, based on the single-point milling force prediction model and stiffness model of the workpiece to be machined, use the spectral clustering method to obtain the milling force sparse key points and stiffness sparse key points during the machining process.

[0041] In one implementation manner, the single-point milling force prediction model constructed based on PCA-GPR is a model that combines principal component analysis (PCA) and Gaussian process regression (GPR) to process the regression prediction problem of multi-dimensional data sets. When constructing the single-point milling force prediction model, first, the principal component analysis method needs to be used to reduce the dimension of the high-dimensional data of the workpiece to be machined during the machining process to achieve the compression of the high-dimensional data; among them, the compression of the high-dimensional data is carried out according to the cumulative contribution amount of the data. Specifically, the cutting force in the Z direction of the robot milling machining is used as the main force factor for force-induced deformation, and according to the spindle speed n at the end of the robot machining to each tool position point, the tool feed speed f, the cutting depth a p , the cutting width a e , each joint angle θ i(i = 1, 2, …, 6) and the amplitude of the Z-direction milling force are used to construct a sample data set (i.e., high-dimensional data). The main factors affecting the Z-direction milling force are analyzed by the PCA method and dimensionality reduction is performed to obtain new variables. The new variables form a K-dimensional data set X * .

[0042] Then, Gaussian process regression is used to train and learn the dimensionality-reduced data to obtain a single-point milling force prediction model. On the basis of ensuring the high prediction ability and high reliability of the GPR model, the computational cost of the model is reduced and the model prediction efficiency is improved. Specifically, for a given data set D = {(x i , y i ) | i = 1, 2, … N} = {x, y} as the model data training set, where x is the input feature matrix and y is the output measurement value. The Gaussian process is mainly uniquely determined by the mean function m(x) and the covariance function k(x, x ′ ). Its data expression is defined as follows:

[0043] f(x) ~ GP(m(x), k(x, x ′ ))

[0044] where m(x) is the mean function, usually set to 0, and k(x, x ′ ) is the covariance function, and a PCA-GPR milling force prediction model is established.

[0045] When constructing the stiffness model, the six joint axes of the robot are regarded as six torsional spring components. When these six components are subjected to torque, they produce angular deformation. The ability of the six joints in the robot to resist this deformation is called the joint stiffness of the robot. Let τ be the joint balance force matrix, q be the joint differential deformation matrix, and K q be the joint stiffness matrix of the robot. Then τ = K q q. According to the principle of virtual work, the virtual work done by the force F at the end of the robot is equal to the sum of the virtual work done by the torques of each joint of the robot:

[0046] F T Δ = τ T δq

[0047]

[0048] C(q) is the stiffness model at the end of the robot. It represents the operational relationship between the force F and the deformation Δ in the end effector space, that is:

[0049]

[0050] Taking the cutter location points in the machining area of the workpiece with milling force or stiffness as clustering parameters, the spectral clustering method is used to cluster and segment the corresponding milling force or stiffness values, so as to obtain the milling force or stiffness sub-region segmentation boundary line in the machining area. The points on the segmentation boundary line are the sparse key points. The milling force or stiffness distribution trends in each sub-region are similar, and the physical quantity values at different cutter location points in different sub-regions can be restored through fewer characteristic quantities such as the milling force values or stiffness values of the region segmentation boundary line and its sparse key points. Specifically, taking stiffness as an example, clustering is performed on the machining cutter location points and the corresponding robot end stiffness, so as to obtain the segmentation results of different regions. By drawing contour lines, the trend changes of stiffness values in different regions can be observed. It can be seen from the contour map that the boundary line of the segmented region is consistent with the contour line trend in the region. Therefore, the fitted cutter location boundary line is of great value for characterizing the stiffness change trend in this region, can reflect the data distribution trend of the clustering subsets, and the points on the fitted boundary line (segmentation boundary line) are the required key points. After clustering and fitting the milling force and stiffness respectively, the key point distributions are obtained.

[0051] Step 2: Optimize the partition of the machining area of the workpiece to be machined based on the obtained milling force sparse key points and stiffness sparse key points.

[0052] Theoretically, the machining deformation of the robot is determined by the robot end stiffness and the milling force. By analogy, the machining deformation distribution of the robot is jointly determined by the sparse key points of the end stiffness and the milling force. By determining the sparse key points of the milling force and the stiffness, the machining deformation distribution information of the robot is obtained, that is:

[0053] Δ D =Φ(K kp (x),F kp (y))

[0054] K kp (x) is the stiffness key point expression, and F kp (y) is the milling force key point expression.

[0055] Based on the acquisition of the sparse key points of the robot machining stiffness and the milling force, the machining deformation distribution information of the robot is obtained. In the specified workpiece area, the sparse key points K kp ={K b{1,2,3,4,5} (x),K {m1,m2,m3,m4,m5,min,max}} of the robot end stiffness and the corresponding sparse key points F kp ={F {1,2,3,4,5} (x),F {m1,m2,m3,m4,m5,min,max}} of the end milling force are obtained. After the partition of the machining physical quantities of the robot composed of them is simplified, the partition method based on the force-stiffness key points is as Figure 3 shown.

[0056] Step 3: Optimize the parameters of the workpiece to be machined based on the obtained results of optimized partitioning, the global optimization deformation index, the single-point milling force prediction model, the stiffness model, the milling force sparse key points, and the stiffness sparse key points; among them, the expression of the global optimization deformation index is:

[0057]

[0058] In the formula: σ 2 is the uniformity index of the overall deformation distribution on the workpiece surface, Δ Dmean represents the overall deformation amount of the workpiece surface, Δ Di is the normal deformation value at the tool position point of the sparse key point i; n is the number of sparse key points; Δ D = Φ(K kp (x), F kp (y)), where K kp (x) is the expression of the stiffness sparse key point, and F kp (y) is the expression of the milling force sparse key point.

[0059] In one embodiment, for the global optimization of machining parameters, it is desired to optimize the overall machining deformation distribution of the workpiece from a global perspective of the workpiece to be machined, that is, while reducing the machining deformation at each tool position point of the workpiece, ensuring the uniformity of the overall machining deformation distribution on the workpiece surface. Considering the problems of large calculation amount and high calculation cost of the overall deformation of the workpiece surface, a mathematical model of the average value and variance of the machining deformation of the milling force and stiffness sparse key points on the workpiece surface is selected as the deformation evaluation index, and together they form the optimization objective function, that is, the global optimization deformation index:

[0060]

[0061] In the formula: σ 2 is the uniformity index of the overall deformation distribution on the workpiece surface, Δ Dmean represents the overall deformation amount of the workpiece surface, Δ Di is the normal deformation value at the tool position point of the key point i.

[0062] The NSGA-II algorithm or the particle swarm optimization algorithm can be used to globally optimize the parameters of the workpiece to be machined. In one embodiment, first, divide the workpiece machining area into k regions based on the boundary line formed by the key points, and use the Latin hypercube sampling method to initialize the parameters of n populations. The population is Substitute the initialized population parameters into the global optimization deformation index for the calculation of the objective function, evaluate the initial overall workpiece deformation situation, and then use the optimization algorithm to find m optimal Pareto solutions Select different combinations of machining parameters for different regions, calculate and analyze the optimization results, and then obtain the optimal combination of machining parameters under the given working conditions. Among them, the numerical ranges of the machining parameters are given according to process experience.

[0063] Specifically, the spindle speed and the feed rate are simultaneously used as the optimization variables of the zoning optimization strategy for parameter optimization calculation. The mathematical expression of the optimization model composed of the optimization variable range and the objective function is as follows:

[0064]

[0065] Please refer to Figure 4 and Figure 5 , the NSGA-II algorithm is used to optimize and solve the optimization model. The optimization parameters are set as follows: the population size Populationsize is 100, the Pareto fraction of the optimal individuals is set to 0.4, the maximum number of iterations MaxGenrations is 300, and the size of the decision variable matrix is 2×9; 40 groups of Pareto optimal solutions are obtained through optimization and solution. The parameters can be selected according to different machining requirements. Table 1 shows one of the Pareto optimal solutions and its objective function values, the Pareto optimal boundary diagram, and the optimized machining deformation distribution corresponding to the selected Pareto solution.

[0066] Table 1 Regional variables and their objective function values corresponding to the selected Pareto solution

[0067]

[0068] Analysis of optimization results: Calculate the average value and variance of the optimized machining deformation according to a single tool path trajectory. The initial machining parameters are n = 5000 rpm, f = 500 mm / s, ap = 0.5 mm, and ae = 6 mm. By comparing the calculation results of the objective function of different tool path trajectories before and after the optimization of the deformation situation, it can be seen that the overall average deformation after optimization decreases by 5.7%, and the uniformity increases by 28.9%. That is, after the multi-parameter variable zoning optimization, the overall machining deformation value of the curved surface workpiece decreases and the distribution becomes more uniform, and the comparison before and after the optimization of different tool path trajectories shows good optimization effects.

[0069] The present invention also provides a global optimization system for machining parameters based on key points of machining deformation. The system includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it executes the global optimization method for machining parameters based on sparse key points of machining deformation as described above.

[0070] The present invention also provides a computer-readable storage medium storing machine-executable instructions, which, when called and executed by a processor, cause the processor to implement the above-described global optimization method for machining parameters based on sparse key points of machining deformation.

[0071] Those skilled in the art can easily understand that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A global optimization method for machining parameters based on sparse key points of machining deformation, characterized in that: The optimization method includes the following steps: (1) Based on the single-point milling force prediction model and stiffness model of the workpiece to be machined, the spectral clustering method is used to obtain the sparse key points of milling force and stiffness in the machining process; (2) Optimizing the partitioning of the machining area of ​​the workpiece to be machined based on the obtained milling force sparse key points and stiffness sparse key points; (3) Based on the obtained optimization partition results, the global optimization deformation index, the single-point milling force prediction model, the stiffness model, the milling force sparse key points and the stiffness sparse key points, the parameters of the workpiece to be processed are optimized; wherein, the expression of the global optimization deformation index is: Where: 2 is the uniformity index of overall deformation distribution on the workpiece surface, Δ Dmean Indicates the overall deformation of the workpiece surface, Δ Di is the normal deformation value of the sparse key point i under the knife point; n is the number of sparse key points; Δ D =Φ(K kp (x),F kp (y)), where K kp (x) is the stiffness sparse key point expression, F kp (y) is the sparse key point expression of milling force.

2. The global optimization method of machining parameters based on sparse key points of machining deformation according to claim 1, characterized in that: The single-point milling force prediction model is a model that combines principal component analysis and Gaussian process regression to deal with multidimensional data set regression prediction problems.

3. The global optimization method of machining parameters based on sparse key points of machining deformation according to claim 2, characterized in that: When constructing the single-point milling force prediction model, the principal component analysis method is used to reduce the dimensionality of the high-dimensional data of the workpiece to be processed during the processing to achieve compression of the high-dimensional data; the compression of the high-dimensional data is carried out based on the cumulative contribution of the data; and then Gaussian process regression is used to train and learn the reduced-dimensional data to obtain the single-point milling force prediction model.

4. The global optimization method of machining parameters based on machining deformation sparse key points according to claim 3, characterized in that: The Gaussian process consists of a mean function m(x) and a covariance function k(x,x ′ ) is uniquely determined, and its data expression is: f(x)~GP(m(x),k(x,x ′ )) Where m(x) is the mean function, k(x,x ′ ) is the covariance function, and the PCA-GPR milling force prediction model is established.

5. The global optimization method of machining parameters based on machining deformation sparse key points according to claim 1, characterized in that: When constructing the stiffness model, let τ be the joint equilibrium force matrix, q be the joint differential deformation matrix, and K q is the joint stiffness matrix of the robot, and τ=K q q, according to the principle of virtual work, the virtual work done by the end force F of the robot is equal to the sum of the virtual work done by the torques of each joint of the robot: F T Δ=τ T δq C(q) is the stiffness model of the robot end, which represents the operational relationship between the force F and the deformation Δ in the end operation space, namely:

6. The global optimization method of machining parameters based on machining deformation sparse key points according to claim 1, characterized in that: The tool position of the milling force or stiffness in the machining area of ​​the workpiece is taken as the clustering parameter, and the spectral clustering method is used to cluster and segment the corresponding milling force or stiffness values, so as to obtain the milling force or stiffness sub-area segmentation boundary line under the machining area. The points on the segmentation boundary line are the sparse key points.

7. The global optimization method for machining parameters based on machining deformation sparse key points according to any one of claims 1 to 6, characterized in that: The robot machining deformation distribution is jointly determined by the sparse key points of the end stiffness and the milling force. By determining the sparse key points of the milling force and stiffness, the robot machining deformation distribution information is obtained, that is: D D =Φ(K kp (x),F kp (y)) K kp (x) is the key point expression of stiffness, F kp (y) is the key point expression of milling force.

8. The global optimization method for machining parameters based on machining deformation sparse key points according to any one of claims 1 to 6, characterized in that: Firstly, the workpiece processing area is divided into k areas based on the boundary lines formed by sparse key points, and the Latin hypercube sampling method is used to initialize the population parameters. The initial population parameters are substituted into the global optimization deformation index to calculate the objective function and evaluate the initial overall workpiece deformation. Then, the optimization algorithm is used to find m optimal Pareto solutions, select the processing parameter combinations of different areas, calculate and analyze the optimization results, and then obtain the optimal processing parameter combination under the working conditions.

9. A global optimization system for processing parameters based on key points of processing deformation, characterized in that: The system includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the global optimization method for machining parameters based on machining deformation sparse key points according to any one of claims 1 to 8 is executed.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by the processor, the machine-executable instructions prompt the processor to implement the global optimization method for machining parameters based on machining deformation sparse key points according to any one of claims 1 to 8.

Images