Geometric error parameter detection method of a curved surface processing machine tool and electronic device

Abstract

The present disclosure provides a geometric error parameter detection method of a curved surface machining machine tool and an electronic device, and relates to the technical field of curved surface machining machine tools. The method comprises the following steps: obtaining a machine tool simulation model of a curved surface machining machine tool constructed in advance; obtaining a measurement vector value of an error vector corresponding to a reference curved surface; performing coarse adjustment iterative processing on a vector value of a parameter vector constituted by each target geometric error parameter according to a plurality of discrete values of each target geometric error parameter to obtain a first vector value; determining a fine adjustment interval of each target geometric error parameter according to the first vector value obtained by coarse adjustment iterative processing; performing fine adjustment iterative processing on the vector value of the parameter vector according to the fine adjustment interval to obtain a second vector value; and obtaining the numerical values of a plurality of target geometric error parameters of the curved surface machining machine tool according to the second vector value obtained by fine adjustment iterative processing. In this way, the accuracy of the detection result of the geometric error parameters of the curved surface machining machine tool can be improved.

Description

Technical Field

[0001] This disclosure relates to the field of surface machining machine tool technology, and more specifically, to a method and electronic device for detecting geometric error parameters of a surface machining machine tool. Background Technology

[0002] With the increasing demand for AR glasses and high-end custom glasses in various scenarios such as sports and medical care, the demand for progressive multifocal lenses with continuous vision range adaptation capabilities from far to intermediate to near has increased significantly. These lenses are complex freeform optical surfaces, which have higher requirements for surface accuracy and continuous transition characteristics compared to single-focal lenses.

[0003] These types of curved surfaces are precision-machined on specialized lathes under a wide range of posture changes. The lathe has three axes: a vertical axis for the rotary table's oscillation; a main spindle for mounting the workpiece, horizontally positioned on the rotary table's oscillation axis with its axis orthogonal to it; and a tool feed axis, also known as the fast tool axis, which performs the axial feed turning motion. Machine tool errors typically consist of geometric, dynamic, and thermal errors. Geometric errors, as the fundamental static errors of the machine tool, create systematic spatial orientation deviations during actual machining, resulting in workpiece surface shape errors and severely affecting surface accuracy. Therefore, identifying the static errors of the machine tool is crucial for the machining accuracy of freeform surface lenses.

[0004] Existing methods for identifying geometric errors in machine tools often employ indirect measurement, obtaining characteristic data from machined samples and combining this data with the machine tool's kinematic model to identify error parameters. Commonly used local optimization algorithms include least squares and Newton's method. However, these methods suffer from problems such as strong dependence on initial values, getting trapped in local optima, and unstable convergence, especially in situations involving multi-parameter coupling, significant differences in sensitivity, and multiple extrema on the cost surface. Summary of the Invention

[0005] One object of this disclosure is to provide a new technical solution for detecting errors in surface machining machine tools.

[0006] According to a first aspect of the present disclosure, a method for detecting geometric error parameters of a surface machining machine tool is provided, characterized in that it includes: Obtain a pre-constructed machine tool simulation model of the surface machining machine tool, wherein the machine tool simulation model is applied to characterize the relationship between multiple target geometric error parameters of the surface machining machine tool and the height corresponding to each target sampling point; Obtain the measurement vector value of the reference surface corresponding to the error vector, wherein the components in the error vector correspond one-to-one with the target sampling points, and the measurement vector value represents the measurement error of each target sampling point in the reference surface; Based on multiple discrete values ​​of each target geometric error parameter, a first vector value is obtained by coarsely adjusting the vector value of the parameter vector formed by the target geometric error parameters; wherein, the first vector value minimizes the cost function constructed based on the parameter vector, the machine tool simulation model and the measurement vector value, and each element of the first vector value belongs to multiple discrete values ​​of the corresponding target geometric error parameter; The fine-tuning range for each target geometric error parameter is determined based on the first vector value obtained from the coarse-tuning iterative process. The second vector value is obtained by performing fine-tuning iteration on the vector values ​​of the parameter vector according to the fine-tuning interval; wherein the second vector value minimizes the objective function constructed based on the parameter vector, the machine tool simulation model and the measurement vector value, and each element of the second vector value is within the fine-tuning interval of the corresponding target geometric error parameter; Based on the second vector value obtained from the fine-tuning iterative process, the values ​​of the plurality of target geometric error parameters of the surface machining machine tool are obtained.

[0007] Optionally, the step of performing coarse-tuning iterative processing on the vector values ​​of the parameter vector formed by the target geometric error parameters based on multiple discrete values ​​of each target geometric error parameter to obtain the first vector value includes: The current vector value of the parameter vector in the current coarse adjustment iteration process is obtained based on multiple discrete values ​​of each target geometric error parameter. Then, the current vector value is substituted into the cost function for coarse adjustment iteration processing. Each element of the current vector value belongs to multiple discrete values ​​of the corresponding target geometric error parameter. When the result of the coarse adjustment iteration process meets the first termination condition, the coarse adjustment iteration process is terminated, and the current vector value obtained by the coarse adjustment iteration process is used as the first vector value; otherwise, the coarse adjustment iteration process continues.

[0008] Optionally, obtaining the current vector value of the parameter vector in the current coarse-tuning iteration process based on multiple discrete values ​​of each target geometric error parameter includes: Obtain the probability density of multiple discrete values ​​of each target geometric error parameter; For each target geometric error parameter, a discrete value is obtained from the corresponding multiple discrete values ​​according to the probability density as the current value; The current vector value of the parameter vector is obtained based on the current value of each target geometric error parameter.

[0009] Optionally, obtaining the probability density of multiple discrete values ​​for each target geometric error parameter includes: Based on the cost value of the cost function and the historical cost value, the reward and penalty amount for this round of coarse adjustment iteration is obtained; Construct a kernel function centered on the current vector value; The probability density is updated based on the kernel function and the reward / penalty amount.

[0010] Optionally, determining the fine-tuning range for each target geometric error parameter based on the first vector value obtained from the coarse-tuning iteration includes: Based on the probability density corresponding to multiple discrete values ​​of each target geometric error parameter, the upper and lower quantiles of the cumulative distribution of the corresponding target geometric error parameter are obtained; The center point corresponding to each target geometric error parameter is determined based on the first vector value; Based on the center point, upper quantile, and lower quantile corresponding to each target geometric error parameter, the fine-tuning range of the corresponding target geometric error parameter is obtained.

[0011] Optionally, the method further includes: Based on the maximum and minimum values ​​of multiple discrete values ​​of each target geometric error parameter, determine the maximum half-width constraint and minimum half-width constraint of the corresponding target geometric error parameter; Furthermore, based on the maximum half-width constraint and minimum half-width constraint of each target geometric error parameter, the fine-tuning range of the corresponding target geometric error parameter is obtained.

[0012] Optionally, the step of performing fine-tuning iteration on the vector values ​​of the parameter vector according to the fine-tuning interval to obtain the second vector value includes: Based on the fine-tuning interval and the set dimensionless boundary, construct the transformation relationship between the vector values ​​of the parameter vector and the dimensionless vector values; Based on the dimensionless boundary, obtain the current dimensionless vector value of the parameter vector in this round of fine-tuning iteration, substitute the current dimensionless vector value into the objective function, and use the trust region reflection algorithm for fine-tuning iteration. When the result of the fine-tuning iteration process satisfies the second termination condition, the fine-tuning iteration process is terminated, and the second vector value of the parameter vector is determined according to the transformation relationship and the dimensionless vector value obtained by the fine-tuning iteration process; otherwise, the fine-tuning iteration process continues.

[0013] Optionally, obtaining the current dimensionless vector value of the parameter vector in this round of fine-tuning iterations based on the dimensionless boundary includes: Based on the current dimensionless vector value, determine the first function value of the objective function and the first predicted value of the local quadratic model; Within the set trust region radius, determine the step size for this round of fine-tuning iterations; The target dimensionless vector value is obtained based on the step size and the current dimensionless vector value obtained from the solution. Then, the second function value of the objective function and the second predicted value of the local quadratic model are determined based on the target dimensionless vector value. Determine whether the step size obtained by solving is acceptable based on the first function value, the first predicted value, the second function value, and the second predicted value; If the judgment result indicates that it is unacceptable, update the trust region radius and recalculate the step size of the current fine-tuning iteration process; if the judgment result indicates that it is acceptable, use the target dimensionless vector value as the current dimensionless vector value.

[0014] Optionally, the method further includes: Sensitivity analysis indices corresponding to multiple candidate geometric error parameters of the surface machining machine tool are obtained; Based on the sensitivity analysis index, the target geometric error parameter that has a greater impact on the machining error of the surface machining machine tool is obtained from among the multiple candidate geometric error parameters.

[0015] Optionally, obtaining the sensitivity analysis index corresponding to each candidate geometric error parameter includes: Iterate through the multiple candidate geometric error parameters; Determine the vector value of the sensitivity vector corresponding to the candidate geometric error parameters currently being traversed; Determine the root mean square of the vector value of the sensitivity vector corresponding to the candidate geometric error parameter currently being traversed, and determine the sensitivity intensity index of the candidate geometric error parameter currently being traversed. Based on the sensitivity intensity index of the candidate geometric error parameter being traversed and the search interval width corresponding to the candidate geometric error parameter being traversed, the sensitivity analysis index of the candidate geometric error parameter being traversed is determined. Determine whether the geometric error parameter traversal has ended. If yes, end the traversal; otherwise, continue the traversal.

[0016] Optionally, determining the vector value of the sensitivity vector corresponding to the candidate geometric error parameter currently being traversed includes: Determine the current perturbation step size corresponding to the candidate geometric error parameters of the current traversal; The first and second perturbation values ​​of the candidate geometric error parameters for the current traversal are determined based on the current perturbation step size. The first error vector value of the error vector is determined when other geometric error parameters are zero and the candidate geometric error parameter of the current traversal is a first perturbation value; the second error vector value of the error vector is determined when other geometric error parameters are zero and the candidate geometric error parameter of the current traversal is a second perturbation value. Based on the first error vector value, the second error vector value, and the current perturbation step size, the vector value of the sensitivity vector corresponding to the candidate geometric error parameter currently being traversed is obtained.

[0017] According to a second aspect of this disclosure, an electronic device is provided, including a processor and a memory, the memory being used to store a computer program, and the processor being used to perform the method as described in the first aspect of this disclosure under the control of the computer program.

[0018] According to a third aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect of this disclosure.

[0019] Through the embodiments of this disclosure, the approximate range of the target geometric error parameters is quickly locked within the discrete value range through coarse-tuning iteration, avoiding excessively long detection cycles caused by blind searching; then, fine-tuning iteration precisely optimizes within the boundaries determined by coarse-tuning, ensuring the accuracy of the detection results. Simultaneously, cost functions and objective functions are constructed based on the machine tool simulation model and measurement vector values, achieving a precise correlation between error parameters and machining errors. This eliminates the need for complex on-site disassembly and inspection, reducing detection difficulty and equipment wear, and balancing detection efficiency and accuracy. It provides reliable parameter support for error compensation and accuracy improvement of surface machining machine tools, and is applicable to geometric error detection scenarios for various complex surface machining machine tools.

[0020] In the coarse-tuning stage, the reinforcement learning algorithm is leveraged for its strong global search capability, fast convergence speed, and lack of initial value requirement. A probability density update mechanism enhances the global search and convergence capabilities, enabling rapid locking of the fine-tuning intervals for each target geometric error parameter within a large parameter space. In the fine-tuning stage, a trust-region reflection algorithm is used within these intervals for nonlinear least-squares fine-tuning. Utilizing its high local convergence accuracy and interval constraints, the final detection results for the target geometric error parameters are determined within the fine-tuning intervals. This improves the stability and accuracy of the detection results, providing a basis for error compensation in surface machining machines.

[0021] Other features and advantages of the invention will become clear from the following detailed description of exemplary embodiments of the invention with reference to the accompanying drawings. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments of the invention and, together with their description, serve to explain the principles of the invention.

[0023] Figure 1 This is a block diagram illustrating the hardware configuration of an electronic device that can implement embodiments of the present disclosure; Figure 2This is a flowchart of a method for detecting geometric error parameters of a surface machining machine tool according to an embodiment of the present disclosure; Figure 3 This is a schematic diagram of the topological relationship of each unit body in a surface machining machine tool according to an embodiment of the present disclosure; Figure 4 This is a bar chart of sensitivity analysis indices for candidate geometric error parameters according to an embodiment of the present disclosure; Figure 5 This is a block diagram of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0024] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention.

[0025] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0026] Techniques, methods, and apparatus known to those skilled in the art in the relevant field may not be discussed in detail, but where appropriate, such techniques, methods, and apparatus should be considered part of the specification.

[0027] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0028] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0029] Figure 1 This is a block diagram illustrating the hardware configuration of an electronic device 1000 that can implement embodiments of the present disclosure.

[0030] Electronic device 1000 can be a portable computer, desktop computer, mobile phone, tablet computer, etc. For example... Figure 1As shown, the electronic device 1000 may include a processor 1100, a memory 1200, an interface device 1300, a communication device 1400, a display device 1500, an input device 1600, a speaker 1700, a microphone 1800, etc. The processor 1100 may be a CPU, a microprocessor (MCU), etc. The memory 1200 may include, for example, ROM (Read-Only Memory), RAM (Random Access Memory), or non-volatile memory such as a hard disk. The interface device 1300 may include, for example, a USB interface, a headphone jack, etc. The communication device 1400 may be capable of wired or wireless communication, specifically including Wi-Fi communication, Bluetooth communication, 2G / 3G / 4G / 5G communication, etc. The display device 1500 may be, for example, an LCD screen, a touch screen, etc. The input device 1600 may include, for example, a touch screen, a keyboard, motion input, etc. Users can input / output voice information through the speaker 1700 and the microphone 1800.

[0031] Figure 1 The electronic devices shown are merely illustrative and in no way intended to limit this disclosure, its application, or use. In embodiments applied to this disclosure, the memory 1200 of the electronic device 1000 is used to store instructions for controlling the processor 1100 to operate to perform any of the methods provided in the embodiments of this disclosure. Those skilled in the art will understand that, although... Figure 1 The electronic device 1000 is shown with multiple devices shown; however, this disclosure may relate only to some of these devices. For example, electronic device 1000 may only relate to processor 1100 and memory 1200. Those skilled in the art can design instructions based on the schemes disclosed herein. How the instructions control the processor to operate is well known in the art and will not be described in detail here.

[0032] This disclosure provides a method for detecting geometric error parameters of a surface machining machine tool, which can be implemented by electronic equipment. Specifically, the method for detecting geometric error parameters of a surface machining machine tool can be implemented by, for example... Figure 1 The electronic device 1000 shown is implemented.

[0033] Figure 2 This is a flowchart of a method for detecting geometric error parameters of a surface machining machine tool according to an embodiment of the present disclosure.

[0034] like Figure 2 As shown, the geometric error parameter detection method for this surface machining machine tool includes the following steps S2100 to S2600: Step S2100: Obtain the pre-built machine tool simulation model of the surface machining machine tool. The machine tool simulation model is used to characterize the relationship between multiple target geometric error parameters of the surface machining machine tool and the height corresponding to each target sampling point.

[0035] In this embodiment, the surface machining machine tool has three motion axes: the oscillating axis of the rotary table, the spindle, and the cutting tool feed axis. The oscillating axis of the rotary table has a vertical axis; the spindle is used to mount the workpiece and is horizontally placed on the oscillating axis of the rotary table, with its axis orthogonal to the oscillating axis; the cutting tool feed axis, also known as the fast tool axis, completes the axial feed turning motion.

[0036] Geometric errors in surface machining machines are categorized into inter-axis errors and intra-axis errors. Intra-axis errors are functional errors that vary with stroke and angle, requiring pre-calibration using specialized metrology instruments to generate error curves. However, given the compact spatial configuration of this surface machining machine, instruments such as laser interferometers and ballbars cannot be installed inside the machine. Furthermore, the three motion axes of the surface machining machine itself have high precision, and their impact on surface accuracy is less than that of inter-axis assembly errors. Therefore, within the inverse framework of using the surface shape of the machined sample as an indirect observation quantity, inter-axis errors, which are parameterizable and have a more significant impact on surface shape, are selected as the identification object.

[0037] In this embodiment, multibody system kinematics theory is used to model the spatial error of the machine tool. First, the unit bodies of the machine tool system and their connections are represented in the form of a topological diagram using the low-order volume array method. The machine tool system consists of five unit bodies: the machine bed, the rotary axis, the workpiece spindle, the workpiece (i.e., the sample), and the high-speed tool axis. The kinematic chain relationships of each unit body of the machine tool system include the workpiece chain and the tool chain. Specifically, the tool chain is: machine bed 0 → rotary axis 1 → workpiece spindle 2 → feature sample 3; the tool chain is: machine bed 0 → high-speed tool axis 4. Their topological relationships are as follows: Figure 3 As shown.

[0038] Specifically, we can first define the machine tool bed as 0, the rotary axis as C, the spindle as B, the workpiece as W, and the fast tool axis as F, and establish the bed coordinate system. Principal coordinate system Rotation axis coordinate system Workpiece coordinate system and tool coordinate system The geometric errors from the bed to the rotary axis, from the rotary axis to the spindle, and from the bed to the fast tool axis are all decomposed into translational and angular errors according to 6 degrees of freedom. Therefore, the geometric error parameters of this surface machining machine tool include the following three groups of 18 items: Geometric error parameters: ; Geometric error parameters: ; Geometric error parameters: .

[0039] The translation error parameters include: The angle error parameters include: .

[0040] The coordinate system transformation of any unit cell can be represented by the following homogeneous matrix:

[0041] Where R is a 3×3 rotation submatrix and p is a 3×1 translation vector.

[0042] Considering error disturbance The effect of the transformation of the actual coordinate system of the unit cell It can be represented as: .

[0043] The geometric error matrix can be expressed as: ; The geometric error matrix can be expressed as: ; The geometric error matrix can be expressed as: .in, include , include , include , include , include , include .

[0044] For any angle error parameter ,have .

[0045] For any translation error parameter The first-order homogeneous error matrix can be obtained. .

[0046] Because the workpiece in the surface machining machine uses a high-precision positioning datum and fixture clamping, the repeatability error of the workpiece's attitude and position relative to the spindle is much smaller than the impact of the machine tool's inter-axis geometric errors on machining accuracy. To avoid introducing non-dominant error sources and reduce the dimensionality of parameter identification, the clamping error of the workpiece relative to the spindle is not considered as an identification error term. .

[0047] Eighteen errors are embedded into the kinematic chain to form an error-containing spatial pose model. A composite transformation of the workpiece chain and the tool chain is then performed. The composite transformation of the workpiece chain can be expressed as:

[0048] The composite transformation of the toolchain can be expressed as:

[0049] The actual coordinates of the tool position point in the workpiece coordinate system can be obtained by multiplying the inverse of the workpiece-side matrix by the tool-side matrix. Therefore, the actual coordinates of the tool position point in the workpiece coordinate system... The actual coordinates below can be represented as: .in, The tool position point is in the tool coordinate system Lower coordinates.

[0050] Therefore, given the geometric error parameters and the sequence of NC command points, the axis command value corresponding to each NC command point is substituted into the formula point by point. The tool position point of the circular arc turning tool in the workpiece coordinate system is obtained point by point. The actual position below, the kth NC instruction point can be represented as .in, It is the actual tool position after taking into account the influence of geometric errors.

[0051] In this embodiment, a regular grid can be established in the (x,y) plane. .in, Define the grid step size and define the grid height matrix. The initial value is +∞.

[0052] Tool position point is determined based on machine tool spatial error model. With unit normal direction vector For each discrete tool position k, with the radius of the circular turning tool... Based on this, in relation to Constructing orthogonal bases in an orthogonal plane ,in, These are two mutually perpendicular unit directions within the plane, used to define local directions within the plane; the intersection circle is the circular profile formed by the intersection of the circular arc turning tool and the plane at the current tool position, used to characterize the spatial geometric boundary of the circular arc turning tool at that tool position. The intersection circle is defined by angle... Discretized into circular points .

[0053] Connect the circular points of adjacent tool positions k and tool position k+1 with the same number, and use two triangles to form a swept surface mesh. Summarize all the triangles of k and m to obtain the discrete triangle set τ.

[0054]

[0055] For any triangle First, calculate its projected bounding box, which is the smallest bounding rectangle formed by the triangle after projection onto the plane. For each grid point... The centroid coordinates are used to determine whether a point falls within the triangle projection. Specifically, if there exists a set of three weight coefficients u, v, w such that the point can be represented as a convex combination of the two-dimensional coordinates of the three vertices of the triangle, and satisfies the following condition: .

[0056] Then u, v, w are the centroid coordinates relative to the three vertices of the triangle. This is true when u, v, w are all not less than the given tolerance. At that time, it is assumed that the point is inside the triangle, and the height of the triangle at that point is obtained by linear interpolation. Where u, v, w are the barycenter coordinates obtained from the above formula. These are the height values ​​of the three vertices of the triangle.

[0057] The machined surface shape is determined by the lower envelope surface of the tool sweeping envelope, thus obtaining the machine tool simulation model: .

[0058] in, For grid points height, For grid points The minimum height of the triangular piece at that point.

[0059] In this embodiment, the target sampling point can be a pre-acquired sampling point.

[0060] Specifically, it can be setting parameters for spiral sampling, such as the starting radius. Termination radius Pitch Number of sampling points Using equal-angle sampling, let The corresponding radius increases linearly with the angle. This allows us to obtain the coordinates of the helix within the workpiece plane. This leads to the formation of a spiral sampling point set. .in, .

[0061] Furthermore, the helical sampling points are interpolated according to the CNC instructions to obtain the interpolated sampling points, and the helical sampling points and the interpolated sampling points are used as the target sampling points.

[0062] Based on this, sampling points are obtained. Target identification data at the location , Sampling points represented by the machine tool simulation model The corresponding height.

[0063] If sampling points Data points outside the measurement coverage area will be invalid, defined as follows: Invalid marker. Filter according to the following validity criteria. Only valid target sampling points are retained.

[0064] In some embodiments, the method further includes: obtaining sensitivity analysis indices corresponding to multiple candidate geometric error parameters of the surface machining machine tool; and obtaining the target geometric error parameter that has a greater impact on the machining error of the surface machining machine tool among the multiple candidate geometric error parameters based on the sensitivity analysis indices.

[0065] The candidate geometric error parameters in this embodiment may include three sets of 18 geometric error parameters for surface machining tools: , , .

[0066] In this embodiment, the sensitivity analysis index can reflect the influence of the corresponding candidate geometric error parameters on the machining error of the surface machining machine tool.

[0067] Specifically, a candidate geometric error parameter that is much larger than other sensitivity analysis parameters can be selected as the target geometric error parameter.

[0068] In one example, the sensitivity analysis indices corresponding to multiple candidate geometric error parameters could be as follows: Figure 4 As shown, the horizontal axis represents candidate geometric error parameters, and the vertical axis represents sensitivity analysis indicators. Figure 4 It can be seen that the sensitivity indices of 5 geometric errors are significantly smaller than those of the remaining 13, and their first-order influence is significantly smaller, exhibiting an approximately zero first-order contribution under the current parameterization. Their influence on the surface error is negligible. Therefore, the set of the 13 selected geometric error parameters can be denoted as […]. The set of the remaining 5 candidate geometric error parameters is denoted as... In the machine tool simulation model, the value is set to 0.

[0069] This embodiment uses sensitivity analysis indicators to screen target geometric error parameters that significantly impact machining errors. This reduces the number of parameters involved in iterative detection, simplifies the detection process, lowers detection difficulty and cost, and shortens the detection cycle. Furthermore, the screened target geometric error parameters are directly related to the core influencing factors of machine tool machining errors. Precise detection of these parameters can more efficiently identify key issues affecting machining accuracy, providing a more targeted basis for machine tool error compensation and accuracy optimization, and further improving the machining quality of surface machining tools.

[0070] In some embodiments, obtaining the sensitivity analysis index corresponding to each candidate geometric error parameter includes: traversing multiple candidate geometric error parameters; determining the vector value of the sensitivity vector corresponding to the currently traversed candidate geometric error parameter; determining the root mean square of the vector value of the sensitivity vector corresponding to the currently traversed candidate geometric error parameter as the sensitivity intensity index of the currently traversed candidate geometric error parameter; determining the sensitivity analysis index of the currently traversed candidate geometric error parameter based on the sensitivity intensity index of the currently traversed candidate geometric error parameter and the search interval width corresponding to the currently traversed candidate geometric error parameter; determining whether the traversal of the geometric error parameter has ended, and if so, ending the traversal, otherwise continuing the traversal.

[0071] In this embodiment, determining the vector value of the sensitivity vector corresponding to the candidate geometric error parameter being traversed includes: determining the current perturbation step size corresponding to the candidate geometric error parameter being traversed; determining the first perturbation value and the second perturbation value of the candidate geometric error parameter being traversed based on the current perturbation step size; determining the first error vector value of the error vector when other geometric error parameters are zero and the candidate geometric error parameter being traversed is the first perturbation value; determining the second error vector value of the error vector when other geometric error parameters are zero and the candidate geometric error parameter being traversed is the second perturbation value; and obtaining the vector value of the sensitivity vector corresponding to the candidate geometric error parameter being traversed based on the first error vector value, the second error vector value, and the current perturbation step size.

[0072] In this embodiment, the perturbation step size corresponding to different candidate geometric error parameters can be the same or different. The perturbation step size corresponding to translation error parameters and rotation error parameters is different.

[0073] The candidate geometric error parameter being traversed so far is the i-th candidate geometric error parameter. In the case of , the central difference can be used to approximate its first-order partial derivative with respect to dP, that is, the vector value of the sensitivity vector corresponding to the i-th candidate geometric error parameter.

[0074]

[0075] in, Let be the unit vector in the direction of the i-th candidate geometric error parameter. For the i-th candidate geometric error parameter The corresponding perturbation step size, It is a vector of length N, where N is the number of target sampling points. For all other geometric error parameters to be zero, the i-th candidate geometric error parameter is the first perturbation value. The first error vector value of the error vector at time ). For all other geometric error parameters to be zero, the i-th candidate geometric error parameter is the second perturbation value. The second error vector value of the time error vector, where each component of the error vector corresponds one-to-one with a target sampling point. The value of any component in the error vector represents the value of the corresponding target sampling point when all other geometric error parameters are zero and the i-th candidate geometric error parameter is the first disturbance value obtained from the machine tool simulation model. The difference between the predicted height and the ideal height of the target sampling point.

[0076] In this embodiment, Let be a vector of length N. The root mean square of the vector values ​​of the sensitivity vector corresponding to the i-th candidate geometric error parameter, i.e., the sensitivity intensity index of the i-th candidate geometric error parameter, can be determined by the following formula:

[0077] Considering the amplification effect of the allowable range of parameter variation on the actual impact, the sensitivity intensity index can be multiplied by the corresponding search interval width to form an impact quantity with uniform dimensions, which can then be used as an analytical indicator. .

[0078] in, This represents the width of the search interval for the corresponding candidate geometric error parameter, i.e., the difference between the maximum and minimum values ​​of the search interval. If... Let be the translation error parameter, then ;like Let be the angle error parameter, then . This corresponds to the maximum value of the translation error parameter. This corresponds to the minimum value of the translation error parameter. This represents the maximum value of the corresponding angle error parameter. This represents the minimum value of the corresponding angle error parameter.

[0079] This embodiment enables precise quantification of the impact of each candidate geometric error parameter on processing errors. Simultaneously, by considering the search interval width and the searchable range of the parameters themselves, it ensures that the selected target geometric error parameters are both significantly influential and detectable. Furthermore, the comprehensive verification mechanism ensures that all candidate parameters are fully analyzed, avoiding the omission of key error parameters. This further enhances the comprehensiveness and reliability of target geometric error parameter selection, providing a guarantee for the efficient implementation of subsequent inspection processes.

[0080] Step S2200: Obtain the measurement vector value of the reference surface corresponding to the error vector. The components in the error vector correspond one-to-one with the target sampling points. The measurement vector value represents the measurement error of each target sampling point in the reference surface.

[0081] In this embodiment, the reference surface can be any surface machined by a surface machining tool. Alternatively, the reference surface can also be a surface simulated when all target geometric error parameters are zero.

[0082] The measurement error of the target sampling point in the reference surface is the difference between the measured height of the corresponding target sampling point in the reference surface and the set ideal height.

[0083] Step S2300: Based on multiple discrete values ​​of each target geometric error parameter, perform coarse-tuning iterative processing on the vector values ​​of the parameter vector formed by the target geometric error parameters to obtain a first vector value; wherein, the first vector value minimizes the cost function constructed based on the parameter vector, the machine tool simulation model and the measurement vector value, and each element of the first vector value belongs to multiple discrete values ​​of the corresponding target geometric error parameter.

[0084] In this embodiment, the multiple discrete values ​​of each target geometric error parameter can be obtained by dividing the search interval of the corresponding target geometric error into L small intervals on average, and the resulting discrete grid points are the corresponding discrete values.

[0085] For example, the i-th target geometric error parameter The search range is Then, the i-th target geometric error parameter Multiple discrete values ​​can be .

[0086] This embodiment employs a continuous action reinforcement learning automaton to perform a global coarse search for multiple target geometric error parameters. Parameter coarse tuning is based on the continuous action reinforcement learning automaton (CARLA) algorithm, but improvements are made to the reinforcement signal and update kernel. The reinforcement signal uses a bounded bidirectional reward and penalty system based on historical cost quantiles, and in the penalty case, a suppressive kernel update and adaptive grid repartitioning are introduced. Simultaneously, the cost function uses a MAD-scale-normalized Huber robust cost to improve noise resistance, sampling, and identification efficiency.

[0087] Specifically, based on multiple discrete values ​​of each target geometric error parameter, a coarse-tuning iterative process is performed on the vector values ​​of the parameter vector formed by the target geometric error parameters to obtain a first vector value. This includes: obtaining the current vector value of the parameter vector in the current round of coarse-tuning iteration based on multiple discrete values ​​of each target geometric error parameter, and then substituting the current vector value into the cost function for coarse-tuning iterative processing; each element of the current vector value belongs to multiple discrete values ​​of the corresponding target geometric error parameter; when the result of the coarse-tuning iterative processing meets the first termination condition, the coarse-tuning iterative processing is terminated, and the current vector value obtained by the coarse-tuning iterative processing is used as the first vector value; otherwise, the coarse-tuning iterative processing continues.

[0088] In this embodiment, the target geometric error parameters form a parameter vector. The remaining candidate geometric error parameters are fixed at 0 in the machine tool simulation model.

[0089] Specifically, the current vector value of the parameter vector. Substituting the values ​​into the machine tool simulation model, we obtain the predicted vector values ​​of the error vector. .

[0090] The residual vector value is obtained by comparing the predicted vector value of the error vector with the measured vector value. ,in, The measurement vector value of the reference surface corresponding to the error vector obtained in step S2200.

[0091] To improve robustness to measurement noise and interpolation outliers, the cost function adopts the Huber robust cost function based on absolute median difference (MAD) scaling normalization. The robustness scale is calculated first. .

[0092] And order . Huber's loss is , The cost function obtained in the t-th iteration can be expressed as:

[0093] Each round of calculations and the current best Compare; if better, refresh. When the result of the coarse-tuning iteration meets the first termination condition, output... As the result of the coarse adjustment, the initial values ​​of the interval are used for the next step of fine-tuning the parameters.

[0094] In this embodiment, the result of the coarse adjustment iteration process satisfies the first termination condition, including: the number of coarse adjustment iterations t reaches the set maximum number of iterations; and / or, the current optimal... No update was performed for M consecutive iterations, where M is a set positive integer.

[0095] In this embodiment, the first vector value obtained from the coarse adjustment iteration can be used as the initial value of the interval for the fine adjustment iteration.

[0096] This embodiment limits the current vector value of the parameter vector to the discrete value range of each target geometric error parameter. This ensures that each round of coarse adjustment iteration is based on reasonable candidate parameter values, avoiding invalid iterations and improving the targeting and efficiency of the coarse adjustment iteration. Simultaneously, by setting a first termination condition, the termination timing of the coarse adjustment iteration can be flexibly controlled. This prevents excessive deviation in the first vector value due to insufficient iteration, which could affect the subsequent fine adjustment effect, and also avoids wasted time due to excessive iteration. This ensures that the coarse adjustment iteration can efficiently and stably output initial vector values ​​that meet the requirements of subsequent fine adjustment, laying a solid foundation for the accurate implementation of fine adjustment iteration and further optimizing the rationality and efficiency of the overall detection process.

[0097] In some embodiments, obtaining the current vector value of the parameter vector in the current coarse adjustment iteration process based on multiple discrete values ​​of each target geometric error parameter includes: obtaining the probability density of multiple discrete values ​​of each target geometric error parameter; for each target geometric error parameter, obtaining a discrete value from the corresponding multiple discrete values ​​as the current value based on the probability density; and obtaining the current vector value of the parameter vector based on the current value of each target geometric error parameter.

[0098] In the first iteration, the probability density of multiple discrete values ​​of each target geometric error parameter can be initialized to a uniform probability density, meaning that the probability density of multiple discrete values ​​of each target geometric error parameter is equal, and ,in Let be the probability density corresponding to the geometric error parameter of the i-th target.

[0099] In the t-th iteration, for each parameter From its current probability density Action values ​​are obtained by mid-sampling Generate random numbers in the interval [0,1], and obtain the current value of the target geometric error parameter by inverse transformation of the current cumulative distribution function: .

[0100] in, .

[0101] This forms the current vector value of the parameter vector during the t-th iteration. .

[0102] This embodiment guides the selection of discrete values ​​based on probability density, making the selection of current values ​​for the parameter vector during coarse-tuning iterations more scientific and targeted. This avoids problems such as chaotic iteration direction and slow convergence caused by randomly selecting discrete values. For each target geometric error parameter, filtering the current value from discrete values ​​based on probability density prioritizes selecting discrete values ​​closer to the optimal solution, accelerating the convergence speed of coarse-tuning iterations, reducing the number of iterations, and lowering detection time. Simultaneously, this selection method ensures the comprehensiveness of the global search, avoiding overlooking the discrete value range where the optimal solution lies. This ensures that the first vector value obtained from the coarse-tuning iterations is closer to the true error parameters, further improving the efficiency of subsequent fine-tuning iterations and the accuracy of the detection results.

[0103] In some embodiments, obtaining the probability density of multiple discrete values ​​of each target geometric error parameter includes: obtaining the reward / penalty amount of the current coarse adjustment iteration process based on the cost value of the cost function and the historical cost value; constructing a kernel function centered on the current vector value; and updating the probability density based on the kernel function and the reward / penalty amount.

[0104] In this embodiment, the value of the cost function obtained in the t-th iteration process can be used. Compare the value of the cost function obtained in the previous t-1 iterations, and calculate the reward / penalty amount according to the following formula. .

[0105]

[0106] in, These are the 10%, 50%, and 90% quantiles of the cost function values ​​obtained in the first t-1 iterations.

[0107] In this embodiment, the quantile threshold of the cost function obtained from the first t-1 iterations is used to construct a bounded reward / penalty system. Significantly better than historical averages (lower than historical averages) A positive reward will be given at that time. ;when Significantly worse than historical averages (higher than historical averages) When this happens, a negative punishment will be given. The remaining intervals show a linear transition, making .

[0108] Indicates a reward. This design represents a penalty. It aligns rewards and penalties with the relative degree of improvement, avoiding drastic fluctuations in updates caused by relying solely on the magnitude of a single cost.

[0109] For each target geometric error parameter Using the current value of this iteration Constructing a kernel function around the center and updating the probability density, its basic form is a combination of the old distribution and kernel enhancement. Among them, learning rate Follow As the range increases, the Gaussian kernel width increases due to the interval scaling coefficient. Confirmed, indicated as .

[0110] when At that time, adopt the following Gaussian kernel centered on :

[0111] This embodiment allows for nucleus enhancement and increases the probability of being near the rewarded action.

[0112] when At that time, Gaussian kernel It can be determined using the following formula:

[0113] In this embodiment, a suppressive update is performed near the sampling point, which reduces the neighborhood density of the sampling point and transfers the probability quality to a region far away from the point, thereby reducing the probability density of the neighborhood of poor actions.

[0114] After the update, normalization is performed on the probability density to ensure... Subsequently, an equal-probability quality re-partition is performed, redistributing the discrete grid based on the cumulative area. This results in denser grids in high-probability regions and sparser grids in low-probability regions, thereby improving subsequent sampling efficiency and convergence stability.

[0115] This embodiment determines the reward / penalty amount based on the cost value of the cost function in the current coarse-tuning iteration and the historical cost value, and updates the probability density using a kernel function, achieving dynamic adaptive adjustment of the probability density. This allows for continuous optimization of the search direction in the coarse-tuning iteration. When the cost value corresponding to a certain discrete value is better (closer to the minimum of the cost function), its probability density is increased through the reward / penalty amount and kernel function, and this type of discrete value is prioritized in subsequent iterations; conversely, its probability density is decreased to reduce invalid selections, thereby guiding the coarse-tuning iteration to gradually converge towards the optimal solution, further improving the efficiency and accuracy of the coarse-tuning iteration. This dynamic update mechanism eliminates the need for manual intervention to adjust the probability density, improving the automation level of the detection method, reducing human error, and ensuring the stability and reliability of the coarse-tuning iteration process.

[0116] Step S2400: Determine the fine-tuning range of each target geometric error parameter based on the first vector value obtained from the coarse-tuning iterative process.

[0117] In this embodiment, each element in the first vector value can be used as the local fine-tuning center point of the corresponding target geometric error parameter.

[0118] In some embodiments, determining the fine-tuning interval of each target geometric error parameter based on the first vector value obtained from the coarse-tuning iterative process may include: obtaining the upper and lower quantiles of the cumulative distribution of the corresponding target geometric error parameter based on the probability density corresponding to multiple discrete values ​​of each target geometric error parameter; determining the center point corresponding to each target geometric error parameter based on the first vector value; and obtaining the fine-tuning interval of the corresponding target geometric error parameter based on the center point, upper quantile, and lower quantile of each target geometric error parameter.

[0119] In this embodiment, the first boundary of each target geometric error parameter is obtained based on the first vector value and multiple discrete values ​​of each target geometric error parameter. This can be achieved by using each element in the first vector value as the center point of the local fine-tuning of the corresponding target geometric error parameter, and then obtaining the fine-tuning interval of the corresponding target geometric error parameter based on the center point of each target geometric error parameter and multiple discrete values.

[0120] For the A target geometric error parameter, and multiple discrete values ​​obtained from the coarse-tuning iterative process for this target geometric error parameter. Corresponding probability density Calculation of cumulative distribution quantiles and lower quantile :

[0121] in, and Let represent the upper and lower boundaries of the i-th target geometric error parameter during the coarse adjustment iteration process, respectively; The cumulative distribution is constructed from the probability density obtained by coarse-tuning iteration.

[0122] For the i-th target geometric error parameter, it can be based on the center point Based on this, construct the corresponding half-width. .

[0123] The boundary value of the fine-tuning interval for the i-th target geometric error parameter can be expressed as: .

[0124] Then, the fine-tuning interval of the i-th target geometric error parameter can be expressed as: .

[0125] In this embodiment, based on the probability density of the first vector value and discrete values ​​obtained from the coarse-tuning iteration, the fine-tuning boundary is determined by the upper quantile, lower quantile, and center point. This makes the setting of the fine-tuning boundary more scientific and targeted, accurately defining the range of the true error parameters. It avoids the problems of low search efficiency due to an excessively large fine-tuning boundary or missing the true optimal solution due to an excessively small boundary. Simultaneously, by combining the cumulative distribution of the probability density to determine the quantiles, it fully utilizes the parameter distribution information accumulated during the coarse-tuning iteration. This ensures that the fine-tuning boundary not only conforms to the actual distribution of parameters but also focuses on the optimal solution region, providing a precise search range for the fine-tuning iteration. This effectively improves the convergence speed and accuracy of the detection results, achieving efficient integration between coarse and fine-tuning.

[0126] In some embodiments, the method may further include: determining the maximum half-width constraint and the minimum half-width constraint of the corresponding target geometric error parameter based on the maximum and minimum values ​​of multiple discrete values ​​of each target geometric error parameter; and obtaining the fine-tuning range of the corresponding target geometric error parameter based on the maximum half-width constraint and the minimum half-width constraint of each target geometric error parameter.

[0127] In this embodiment, the maximum half-width constraint of the i-th target geometric error parameter can be expressed as: The minimum half-width constraint of the i-th target geometric error parameter can be expressed as: .

[0128] Furthermore, it can be based on the center point Apply minimum and maximum half-width constraints to the half-width constructed based on the baseline, such that... This allows us to obtain the boundary values ​​of the i-th target geometric error parameter. .

[0129] This embodiment adds maximum and minimum half-width constraints to further optimize and limit the fine-tuning boundary, avoiding unreasonable fine-tuning boundaries caused by quantile calculation deviations or abnormal parameter distributions. The maximum half-width constraint prevents the fine-tuning boundary from being too wide, ensuring the search range of the fine-tuning iteration is focused and improving search efficiency; the minimum half-width constraint prevents the fine-tuning boundary from being too narrow, avoiding the omission of true error parameters and ensuring the accuracy of the detection results. This dual constraint mechanism makes the setting of the fine-tuning boundary more robust, adaptable to different types of surface machining tools and the distribution characteristics of different error parameters, further improving the versatility and reliability of the detection method, and ensuring that the fine-tuning iteration can stably and accurately output the optimal solution.

[0130] Step S2500: The vector values ​​of the parameter vector are fine-tuned iteratively processed according to the fine-tuning interval to obtain the second vector value; wherein, the second vector value minimizes the objective function constructed based on the parameter vector, the machine tool simulation model and the measurement vector value, and each element of the second vector value is within the fine-tuning interval of the corresponding target geometric error parameter.

[0131] In this embodiment, a nonlinear least squares problem with boundaries is solved within the fine-tuning interval of each target geometric error parameter to obtain the value of each geometric error parameter.

[0132] In some embodiments, the second vector value is obtained by performing fine-tuning iteration on the vector value of the parameter vector according to the fine-tuning interval, including: constructing a transformation relationship between the vector value of the parameter vector and the dimensionless vector value according to the fine-tuning interval and the set dimensionless boundary; obtaining the current dimensionless vector value of the parameter vector in the current round of fine-tuning iteration according to the dimensionless boundary, substituting the current dimensionless vector value into the objective function, and performing fine-tuning iteration using the Trust-Region Reflective (TRR) algorithm; terminating the fine-tuning iteration when the result of the fine-tuning iteration meets the second termination condition, and determining the second vector value of the parameter vector according to the transformation relationship and the dimensionless vector value obtained by the fine-tuning iteration; otherwise, continuing the fine-tuning iteration.

[0133] By employing dimensionless variables, the influence of dimensional and sensitivity differences in the target geometric error parameters on the solver is reduced. Based on the fine-tuning interval and the defined dimensionless boundaries, a transformation relationship is constructed between the vector values ​​of the parameter vector and the dimensionless vector values. This transformation relationship can be expressed as follows: .

[0134] Where p represents the vector value of the parameter vector to be fine-tuned; represents the center vector value of the parameter vector, where each element represents the center point of the fine-tuning interval of the corresponding target geometric error parameter; h is the dimension-wise half-width vector corresponding to the fine-tuning interval of each target geometric error parameter; q is the dimensionless variable corresponding to the target geometric error parameter, which is the variable actually solved by the TRR algorithm in the fine-tuning iteration process; ⊙ represents element-wise multiplication, indicating that corresponding elements are multiplied separately. After normalization, all dimensionless variables are restricted to the constraint interval [-1,1], enabling the TRR algorithm to perform local fine-tuning in a unified scale space.

[0135] In this embodiment, TRR uses a constrained nonlinear least squares optimization method to solve for the second vector value:

[0136] In this embodiment, the current dimensionless vector value can be substituted into the machine tool simulation model to obtain the predicted vector value of the error vector. Then, the predicted vector value of the error vector is compared with the measured vector value to obtain the vector value of the residual vector. ,in, The measurement vector value of the reference surface corresponding to the error vector obtained in step S2200.

[0137] Based on this, first calculate the center vector value of the residual vector at the center of the fine-tuning interval. Then use MAD to obtain the scale. .

[0138] The residual vector values ​​are then normalized using the following formula:

[0139] The vector value of the residual vector after normalization.

[0140] In the residual vector and boundary constraints Based on this, the trust region reflection algorithm is used to solve the following objective function:

[0141] The current dimensionless vector value can be 0 during the first round of fine-tuning iteration.

[0142] The process involves obtaining the current dimensionless vector value of the parameter vector in the current fine-tuning iteration based on the dimensionless boundary, including: determining the first function value of the objective function and the first predicted value of the local quadratic model based on the current dimensionless vector value; solving for the step size of the current fine-tuning iteration within the set trust region radius; obtaining the target dimensionless vector value based on the solved step size and the current dimensionless vector value, and then determining the second function value of the objective function and the second predicted value of the local quadratic model based on the target dimensionless vector value; judging whether the solved step size is acceptable based on the first function value, the first predicted value, the second function value, and the second predicted value; updating the trust region radius and resolving the step size of the current fine-tuning iteration if the judgment result indicates that it is unacceptable; and using the target dimensionless vector value as the current dimensionless vector value if the judgment result indicates that it is acceptable.

[0143] At the k-th (k is an integer greater than 1) fine-tuning iteration point At this point, the residual vector is linearized to first order using the following formula:

[0144] in, The Jacobian matrix is ​​obtained by differencing the residual vectors. This leads to the local quadratic model. .

[0145] Based on this, the following formula is used to define the trust region radius. Internally solve for the step size of the k-th fine-tuning iteration :

[0146] When the step size causes some components of the dimensionless vector value to go out of bounds, a reflection strategy is used to map them back into the feasible region boundary. The out-of-bounds components are corrected by reflection along the boundary, thereby maintaining the descent direction as much as possible while maintaining feasibility.

[0147] Based on this, the ratio of the actual decline to the predicted decline is calculated using the following formula:

[0148] in, This is the current dimensionless vector value. Let the target be a dimensionless vector value. The first function value, The second function value, This represents the actual decrease. The first predicted value, This is the second predicted value. To predict the amount of decline.

[0149] like If the prediction is unreliable, reject the step size, reduce the trust region radius, and solve for the step size again. Until the acceptance conditions are met. If If the current step size is acceptable, the trust region radius remains unchanged; if... When the actual decrease is consistent with the prediction, the update is accepted and the confidence region radius is appropriately increased. Here, a is less than b, and both a and b are positive numbers.

[0150] The step size is obtained in the k-th fine-tuning iteration. After completing the acceptance criterion, if the objective function converges, the descent amount is below the threshold. ,Right now If the result of the fine-tuning iteration process satisfies the second termination condition, the iteration is terminated and the current dimensionless vector value is output. .

[0151] In this embodiment, it can be that... The transformation relationship between the vector values ​​of the pre-constructed parameter vector and the dimensionless vector values ​​is introduced. In this process, the second vector value of the parameter vector is obtained.

[0152] This embodiment constructs the transformation relationship between the vector values ​​of the parameter vector and the dimensionless vector values. Fine-tuning iterations are performed using dimensionless boundaries, and a trust-region reflection algorithm is employed. This addresses issues such as inconsistent parameter scales, cumbersome boundary handling, and unstable convergence during fine-tuning iterations. Dimensionlessness eliminates the dimensional differences between different target geometric error parameters, making the search step size of each parameter more reasonable during iteration and improving iteration stability. The trust-region reflection algorithm strictly ensures that the iteration point is within the fine-tuning boundary, avoiding iteration failure caused by parameter out-of-bounds errors. It also possesses good convergence and robustness, quickly converging to the minimum value of the objective function. Furthermore, a second termination condition controls the iteration termination, ensuring that the accuracy of the fine-tuning iteration meets the detection requirements, ultimately obtaining high-precision error parameter values, providing a more reliable basis for machine tool accuracy compensation.

[0153] Step S2600: Based on the second vector value obtained from the fine-tuning iteration process, the values ​​of multiple target geometric error parameters of the surface machining machine tool are obtained.

[0154] In this embodiment, the element in the second vector value corresponding to each target geometric error parameter represents the value of the corresponding target geometric error parameter.

[0155] Through the embodiments of this disclosure, the approximate range of the target geometric error parameters is quickly locked within the discrete value range through coarse-tuning iteration, avoiding excessively long detection cycles caused by blind searching; then, fine-tuning iteration precisely optimizes within the boundaries determined by coarse-tuning, ensuring the accuracy of the detection results. Simultaneously, cost functions and objective functions are constructed based on the machine tool simulation model and measurement vector values, achieving a precise correlation between error parameters and machining errors. This eliminates the need for complex on-site disassembly and inspection, reducing detection difficulty and equipment wear, and balancing detection efficiency and accuracy. It provides reliable parameter support for error compensation and accuracy improvement of surface machining machine tools, and is applicable to geometric error detection scenarios for various complex surface machining machine tools.

[0156] In the coarse-tuning stage, the reinforcement learning algorithm is leveraged for its strong global search capability, fast convergence speed, and lack of initial value requirement. A probability density update mechanism enhances the global search and convergence capabilities, enabling rapid locking of the fine-tuning intervals for each target geometric error parameter within a large parameter space. In the fine-tuning stage, a trust-region reflection algorithm is used within these intervals for nonlinear least-squares fine-tuning. Utilizing its high local convergence accuracy and interval constraints, the final detection results for the target geometric error parameters are determined within the fine-tuning intervals. This improves the stability and accuracy of the detection results, providing a basis for error compensation in surface machining machines.

[0157] This embodiment provides an electronic device, such as... Figure 5As shown, the electronic device 1000 may include a processor 1100 and a memory 1200. The memory 1200 is used to store computer programs, and the processor 1100 is used to control the electronic device to execute the methods of any embodiment of this disclosure under the control of the computer programs.

[0158] This embodiment provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the methods described in any of the method embodiments of this disclosure.

[0159] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0160] This disclosure may be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement any of the methods in the foregoing embodiments of this disclosure.

[0161] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media may include, for example, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), compact disc-read-only memory (CD-ROM), digital versatile disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any combination thereof. The computer-readable storage medium used herein is not to be interpreted as a transient signal itself, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0162] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include one or more of copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to computer-readable storage media in the respective computing / processing device.

[0163] The computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source or object programs written in any combination of one or more programming languages, including object-oriented programming languages ​​(such as Smalltalk, C++, etc.) and conventional procedural programming languages ​​(such as the "C" language or similar programming languages). The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network (e.g., a local area network or a wide area network), or it may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays, or programmable logic arrays, can execute computer-readable program instructions to implement various aspects of the embodiments of this disclosure by utilizing state information from the computer-readable program instructions.

[0164] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0165] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0166] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions that execute on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0167] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. It should be noted that implementation in hardware, implementation in software, and implementation using a combination of software and hardware are all equivalent.

[0168] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, and are not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein. The scope of this disclosure is defined by the appended claims.

Claims

1. A method for detecting geometric error parameters of a surface machining machine tool, characterized in that, include: Obtain a pre-constructed machine tool simulation model of the surface machining machine tool, wherein the machine tool simulation model is used to characterize the relationship between multiple target geometric error parameters of the surface machining machine tool and the height corresponding to each target sampling point; Obtain the measurement vector value of the reference surface corresponding to the error vector, wherein the components in the error vector correspond one-to-one with the target sampling points, and the measurement vector value represents the measurement error of each target sampling point in the reference surface; Based on multiple discrete values ​​of each target geometric error parameter, a first vector value is obtained by coarsely adjusting the vector value of the parameter vector formed by the target geometric error parameters; wherein, the first vector value minimizes the cost function constructed based on the parameter vector, the machine tool simulation model and the measurement vector value, and each element of the first vector value belongs to multiple discrete values ​​of the corresponding target geometric error parameter; The fine-tuning range for each target geometric error parameter is determined based on the first vector value obtained from the coarse-tuning iterative process. The second vector value is obtained by performing fine-tuning iteration on the vector values ​​of the parameter vector according to the fine-tuning interval; wherein the second vector value minimizes the objective function constructed based on the parameter vector, the machine tool simulation model and the measurement vector value, and each element of the second vector value is within the fine-tuning interval of the corresponding target geometric error parameter; Based on the second vector value obtained from the fine-tuning iterative process, the values ​​of the plurality of target geometric error parameters of the surface machining machine tool are obtained.

2. The method according to claim 1, characterized in that, The step of performing coarse-tuning iterative processing on the vector values ​​of the parameter vector formed by the target geometric error parameters based on multiple discrete values ​​of each target geometric error parameter to obtain the first vector value includes: The current vector value of the parameter vector in the current coarse adjustment iteration is obtained based on multiple discrete values ​​of each target geometric error parameter, and then the current vector value is substituted into the cost function for coarse adjustment iteration processing; each element of the current vector value belongs to multiple discrete values ​​of the corresponding target geometric error parameter; When the result of the coarse adjustment iteration process meets the first termination condition, the coarse adjustment iteration process is terminated, and the current vector value obtained by the coarse adjustment iteration process is used as the first vector value; otherwise, the coarse adjustment iteration process continues.

3. The method according to claim 2, characterized in that, The step of obtaining the current vector value of the parameter vector in this round of coarse adjustment iterations based on multiple discrete values ​​of each target geometric error parameter includes: Obtain the probability density of multiple discrete values ​​of each target geometric error parameter; For each target geometric error parameter, a discrete value is obtained from the corresponding multiple discrete values ​​according to the probability density as the current value; The current vector value of the parameter vector is obtained based on the current value of each target geometric error parameter.

4. The method according to claim 3, characterized in that, The process of obtaining the probability density of multiple discrete values ​​for each target geometric error parameter includes: Based on the cost value of the cost function and the historical cost value, the reward and penalty amount for this round of coarse adjustment iteration is obtained; Construct a kernel function centered on the current vector value; The probability density is updated based on the kernel function and the reward / penalty amount.

5. The method according to claim 1, characterized in that, Determining the fine-tuning range for each target geometric error parameter based on the first vector value obtained from the coarse-tuning iteration includes: Based on the probability density corresponding to multiple discrete values ​​of each target geometric error parameter, the upper and lower quantiles of the cumulative distribution of the corresponding target geometric error parameter are obtained. The center point corresponding to each target geometric error parameter is determined based on the first vector value; Based on the center point, upper quantile, and lower quantile corresponding to each target geometric error parameter, the fine-tuning range of the corresponding target geometric error parameter is obtained.

6. The method according to claim 5, characterized in that, The method further includes: Based on the maximum and minimum values ​​of multiple discrete values ​​of each target geometric error parameter, determine the maximum half-width constraint and minimum half-width constraint of the corresponding target geometric error parameter; Furthermore, based on the maximum half-width constraint and minimum half-width constraint of each target geometric error parameter, the fine-tuning range of the corresponding target geometric error parameter is obtained.

7. The method according to claim 1, characterized in that, The step of performing fine-tuning iteration on the vector values ​​of the parameter vector according to the fine-tuning interval to obtain the second vector value includes: Based on the fine-tuning interval and the set dimensionless boundary, construct the transformation relationship between the vector values ​​of the parameter vector and the dimensionless vector values; Based on the dimensionless boundary, obtain the current dimensionless vector value of the parameter vector in this round of fine-tuning iteration, substitute the current dimensionless vector value into the objective function, and use the trust region reflection algorithm for fine-tuning iteration. When the result of the fine-tuning iteration process satisfies the second termination condition, the fine-tuning iteration process is terminated, and the second vector value of the parameter vector is determined according to the transformation relationship and the dimensionless vector value obtained by the fine-tuning iteration process; otherwise, the fine-tuning iteration process continues.

8. The method according to claim 7, characterized in that, The step of obtaining the current dimensionless vector value of the parameter vector in this round of fine-tuning iterations based on the dimensionless boundary includes: Based on the current dimensionless vector value, determine the first function value of the objective function and the first predicted value of the local quadratic model; Within the set trust region radius, determine the step size for this round of fine-tuning iterations; The target dimensionless vector value is obtained based on the step size and the current dimensionless vector value obtained from the solution. Then, the second function value of the objective function and the second predicted value of the local quadratic model are determined based on the target dimensionless vector value. Determine whether the step size obtained by solving is acceptable based on the first function value, the first predicted value, the second function value, and the second predicted value; If the judgment result indicates that it is unacceptable, update the trust region radius and recalculate the step size of the current fine-tuning iteration process; if the judgment result indicates that it is acceptable, use the target dimensionless vector value as the current dimensionless vector value.

9. The method according to claim 1, characterized in that, The method further includes: Sensitivity analysis indices corresponding to multiple candidate geometric error parameters of the surface machining machine tool are obtained; Based on the sensitivity analysis index, the target geometric error parameter that has a greater impact on the machining error of the surface machining machine tool is obtained from among the multiple candidate geometric error parameters.

10. An electronic device, characterized in that, It includes a processor and a memory, the memory being used to store a computer program, and the processor being used, under the control of the computer program, to execute the method as described in any one of claims 1 to 9.

Images