Intelligent numerical control processing control system of strain power hardware manufacturing equipment

Abstract

The application provides an intelligent numerical control machining control system of a tension electric power fitting manufacturing device, and belongs to the field of tension electric power fitting manufacturing. The system comprises the following steps: collecting a target profile through contact touch and generating a discrete initial point set, constructing a composite fitting weight by combining local geometric curvature and normalized arc length distance, and using a weighted least square method to obtain initial control points of a double Bezier curve and fitting residual error; then generating a curvature smoothing adjustment amount based on a curvature difference value at a common connection point, a transition influence factor and a space displacement conversion coefficient, adjusting non-common control points on both sides of the connection point, and obtaining a double Bezier reference track; after machining, residual profile error is extracted, a three-dimensional compensation vector is solved, and a Jacobian matrix is iteratively updated, so that the trajectory compensation capability and forming quality of multi-pass machining are improved.

Description

Technical Field

[0001] This application belongs to the field of tension power fitting manufacturing, and particularly relates to an intelligent CNC machining control system for tension power fitting manufacturing equipment. Background Technology

[0002] Tension power fittings are critical components in overhead transmission lines, used to withstand conductor tension and maintain structural stability. Their mechanical strength and surface profile accuracy directly affect the operational safety of the power grid system. CNC machining technology, with its advantages of high efficiency and controllable precision, has become an important method for manufacturing such fittings. In the process of planning the machining trajectory for the spatial profile of tension power fittings, it is usually necessary to use measuring equipment to obtain discrete data points on the fitting surface and fit these discrete points using Bézier curves. However, traditional curve fitting methods often assign equal weights to each data point when processing discrete data, failing to fully consider the differences in local geometric features of the profile. This is particularly prone to local fitting distortion in areas with significant curvature changes, causing the generated theoretical trajectory to deviate from the target profile. Furthermore, due to the complex profile shape of tension power fittings, trajectory planning typically requires the splicing of multiple Bézier curve segments. Conventional splicing methods focus on positional and tangential continuity at the connection point, but do not fully consider the coordination of curvature on both sides of the connection point. This can easily lead to abrupt curvature changes at the common connection point, which in turn causes fluctuations in machine tool feed speed and acceleration, resulting in surface vibration marks or uneven cutting, thus affecting the quality of the formed surface.

[0003] In the actual CNC machining of tension power fittings, factors such as tool wear, cutting force deformation, and system following errors can easily lead to bow height and spatial contour errors between the actual tool path and the planned theoretical double Bezier reference path. Existing machining error compensation technologies mostly employ offline geometric offsets or static compensation methods based on preset parameters, which are difficult to adapt to the accumulation of nonlinear errors caused by changes in working conditions during the cutting process. When entering the next machining pass, if only the theoretical path from the previous pass is used without targeted correction based on the already generated spatial errors, the generated machining errors will continue to propagate to subsequent cutting processes, leading to decreased machining accuracy and affecting product yield. Traditional compensation methods lack an effective closed-loop feedback mechanism, making it difficult to perform correlation analysis and iterative updates of the error solution model based on the actual measured residual contour errors. This makes it difficult for the compensation model to adaptively adjust during multi-pass machining, hindering the continuous improvement of error prediction and compensation accuracy, and restricting further improvements in the CNC machining accuracy of tension power fittings. Summary of the Invention

[0004] To address the problems of single trajectory fitting weights, abrupt curvature changes at curve joints, and difficulty in compensating for multi-pass machining errors in existing technologies, this invention proposes an intelligent CNC machining control system for tension-resistant power fitting manufacturing equipment, comprising the following modules: The module is used to obtain a discretized initial point set based on the contact touch test of the target profile of the tension power fitting. It constructs a composite fitting weight based on the local geometric curvature of each initial point and the normalized arc length distance from the point to the preset common connection point. Based on the composite fitting weight, the initial control points and fitting residuals of the first and second Bézier curves are obtained by weighted least squares method. The adjustment module is used to obtain the curvature difference between the two curves at the common connection point and the transition influence factor constructed based on the fitting residual. It obtains the curvature smoothing adjustment amount based on the curvature difference, the transition influence factor and the preset spatial displacement conversion coefficient, and generates an adjustment vector based on the curvature smoothing adjustment amount and the tangent direction at the common connection point. It adjusts the positions of the non-common control points on both sides of the connection point based on the adjustment vector to obtain the double Bezier reference trajectory. The update module is used to control the tool to cut along the double Bezier reference trajectory to obtain the spatial error feature vector of the same cutting segment, call the Jacobian matrix to solve for the three-dimensional compensation vector of the next machining pass and update the control point coordinates to obtain the compensatory machining path. After completing the next machining pass along the compensatory machining path, the newly acquired residual contour error is correlated with the three-dimensional compensation vector to obtain the iteratively updated Jacobian matrix.

[0005] Preferably, the contact-based touch testing based on the target contour of the tension-resistant power fitting to obtain a discretized initial point set includes: The control is set at the end of the machine tool spindle. The contact probe touches the contour surface of the tension power fitting in a predetermined step length according to the mutually parallel guide trajectory. The three-dimensional coordinates of the probe trigger point are recorded to obtain a mixed data point set. Extract the boundary surface of the three-dimensional nominal model and offset it bidirectionally along the normal direction by a preset tolerance to obtain the defined threshold envelope surface; Remove all abnormal noise points in the mixed data point set that exceed the defined threshold envelope, and arrange the actual surface coordinates that remain within the envelope in sequence to obtain the discretized initial point set.

[0006] Preferably, the step of constructing composite fitting weights based on the local geometric curvature of each initial point and the normalized arc length distance from the point to the preset common connection point includes: The local geometric curvature corresponding to each initial point is calculated by the adjacent three-point microcircle fitting method, and the local geometric curvature is normalized into a curvature weight term. Calculate the cumulative arc length from each initial point to the common connection point along the actual measured trajectory, divide the cumulative arc length by the total arc length of the trajectory segment, and subtract the quotient from 1 to obtain the corresponding normalized arc length distance. The first and second predefined proportion coefficients are called, the curvature weight term is multiplied by the first proportion coefficient, and the normalized arc length distance is multiplied by the second proportion coefficient. The summation yields the composite fitting weight.

[0007] Preferably, the step of obtaining the curvature difference between the two curves at the common connection point and the transition influence factor constructed based on the fitting residual, obtaining the curvature smoothing adjustment amount according to the curvature difference, the transition influence factor and the preset spatial displacement conversion coefficient, and generating an adjustment vector based on the curvature smoothing adjustment amount and the tangent direction at the common connection point includes: Calculate the first and second derivatives of the first and second Bézier curves at the common connection point, respectively, and calculate the curvature difference vector of the two curves at this point. Calculate the sum of the preset smoothness threshold and the maximum value of the fitting residual, and divide the preset smoothness threshold by the sum to obtain the transition influence factor; Calculate the magnitude of the curvature difference vector, and multiply the magnitude of the curvature difference vector by the transition influence factor and the preset spatial displacement conversion coefficient to obtain the curvature smoothing adjustment amount; The unit tangent direction is determined based on the first derivative at the common connection point. A positive candidate adjustment vector and a negative candidate adjustment vector are constructed based on the curvature smoothing adjustment amount and the unit tangent direction. The actual adjustment vector is selected based on the comparison between the curvature difference after the two candidate adjustment vectors are applied and the curvature difference before adjustment.

[0008] Preferably, the step of adjusting the positions of the non-common control points on both sides of the connection point based on the adjustment vector to obtain the double Bezier reference trajectory includes: The signed length of the adjustment vector along the tangent direction of the common connection point is read as the tangential one-dimensional adjustment value; Extract the original three-dimensional coordinates of the first non-common control point and the second non-common control point on both sides of the common connection point; The first and second non-common control points are orthogonally projected onto a tangent constraint line that passes through the common connection point and is oriented in the tangent direction at the common connection point. The one-dimensional tangential adjustment value is then converted into a displacement along the tangent direction according to a preset allocation ratio and superimposed onto the three-dimensional coordinates of the first and second non-common control points respectively. The collinear constraint of the first non-common control points, the common connection point, and the second non-common control points is maintained. After updating the control point positions, the curvature difference vector and the one-dimensional tangential adjustment value are recalculated and iteratively changed until the two curves satisfy a preset continuity condition at the common connection point.

[0009] Preferably, the step of obtaining the initial control points and fitting residuals of the first and second Bézier curves using the weighted least squares method based on the composite fitting weights includes: Substitute the coordinate parameters of the discretized initial point set and the composite fitting weights into the preset polynomial fitting objective function to construct a weighted error sum of squares equation. Calculate the partial derivatives of each control point variable in the weighted error sum of squares equation, and set each partial derivative to 0 to generate a system of linear normal equations. The three-dimensional coordinates of the initial control point are obtained by solving the linear normal equation system through matrix inversion. Substitute the coordinates of the discretized initial point set into the Bézier curve equation determined by the initial control points, calculate the minimum orthogonal distance sequence from the points to the curve, and obtain the fitting residual.

[0010] Preferably, the step of correlating the newly acquired residual contour error with the three-dimensional compensation vector to obtain the iteratively updated Jacobian matrix includes: Extract the difference between the actual coordinates and the theoretical compensation coordinates of each measuring point obtained by the online measurement module in the next machining pass of the same cutting segment, and construct the current residual contour error vector; Calculate the error difference between the current residual contour error vector and the residual contour error vector of the previous processing pass, and the compensation difference between the three-dimensional compensation vector of the current processing pass and the previous processing pass. Using Breudon's rule, a Jacobian matrix adjustment term is constructed based on the error difference and the compensation difference. The Jacobian matrix adjustment term is then updated by matrix addition with the Jacobian matrix used in the previous processing pass, generating the iteratively updated Jacobian matrix.

[0011] According to another aspect of the present invention, an intelligent numerical control machining control method for tension-resistant power fitting manufacturing equipment is proposed, comprising the following steps: Based on the contact-type touch test of the target profile of tension power fitting, a discretized initial point set is obtained. A composite fitting weight is constructed according to the local geometric curvature of each initial point and the normalized arc length distance from the point to the preset common connection point. Based on the composite fitting weight, the initial control points and fitting residuals of the first and second Bézier curves are obtained by weighted least squares method. Obtain the curvature difference between the two curves at the common connection point and the transition influence factor constructed based on the fitting residual. Obtain the curvature smoothing adjustment amount based on the curvature difference, the transition influence factor and the preset spatial displacement conversion coefficient. Generate an adjustment vector based on the curvature smoothing adjustment amount and the tangent direction at the common connection point. Adjust the positions of the non-common control points on both sides of the connection point based on the adjustment vector to obtain the double Bezier reference trajectory. The tool is controlled to cut along the double Bezier reference trajectory to obtain the spatial error feature vector of the same cutting segment. The Jacobian matrix is ​​called to solve for the three-dimensional compensation vector of the next machining pass and update the control point coordinates to obtain the compensatory machining path. After the next machining pass is completed along the compensatory machining path, the newly acquired residual contour error is correlated with the three-dimensional compensation vector to obtain the iteratively updated Jacobian matrix.

[0012] Optionally, the contact-based touch testing based on the target contour of the tension-resistant power fitting to obtain a discretized initial point set includes: The control is set at the end of the machine tool spindle. The contact probe touches the contour surface of the tension power fitting in a predetermined step length according to the mutually parallel guide trajectory. The three-dimensional coordinates of the probe trigger point are recorded to obtain a mixed data point set. Extract the boundary surface of the three-dimensional nominal model and offset it bidirectionally along the normal direction by a preset tolerance to obtain the defined threshold envelope surface; Remove all abnormal noise points in the mixed data point set that exceed the defined threshold envelope, and arrange the actual surface coordinates that remain within the envelope in sequence to obtain the discretized initial point set.

[0013] Optionally, the step of constructing composite fitting weights based on the local geometric curvature of each initial point and the normalized arc length distance from the point to the preset common connection point includes: The local geometric curvature corresponding to each initial point is calculated by the adjacent three-point microcircle fitting method, and the local geometric curvature is normalized into a curvature weight term. Calculate the cumulative arc length from each initial point to the common connection point along the actual measured trajectory, divide the cumulative arc length by the total arc length of the trajectory segment, and subtract the quotient from 1 to obtain the corresponding normalized arc length distance. The first and second predefined proportion coefficients are called, the curvature weight term is multiplied by the first proportion coefficient, and the normalized arc length distance is multiplied by the second proportion coefficient. The summation yields the composite fitting weight.

[0014] Optionally, the step of obtaining the curvature difference between the two curves at the common connection point and the transition influence factor constructed based on the fitting residual, obtaining the curvature smoothing adjustment amount according to the curvature difference, the transition influence factor and the preset spatial displacement conversion coefficient, and generating an adjustment vector based on the curvature smoothing adjustment amount and the tangent direction at the common connection point includes: Calculate the first and second derivatives of the first and second Bézier curves at the common connection point, respectively, and calculate the curvature difference vector of the two curves at this point. Calculate the sum of the preset smoothness threshold and the maximum value of the fitting residual, and divide the preset smoothness threshold by the sum to obtain the transition influence factor; Calculate the magnitude of the curvature difference vector, and multiply the magnitude of the curvature difference vector by the transition influence factor and the preset spatial displacement conversion coefficient to obtain the curvature smoothing adjustment amount; The unit tangent direction is determined based on the first derivative at the common connection point. A positive candidate adjustment vector and a negative candidate adjustment vector are constructed based on the curvature smoothing adjustment amount and the unit tangent direction. The actual adjustment vector is selected based on the comparison between the curvature difference after the two candidate adjustment vectors are applied and the curvature difference before adjustment.

[0015] Optionally, adjusting the positions of the non-common control points on both sides of the connection point based on the adjustment vector to obtain the double Bezier reference trajectory includes: The signed length of the adjustment vector along the tangent direction of the common connection point is read as the tangential one-dimensional adjustment value; Extract the original three-dimensional coordinates of the first non-common control point and the second non-common control point on both sides of the common connection point; The first and second non-common control points are orthogonally projected onto a tangent constraint line that passes through the common connection point and is oriented in the tangent direction at the common connection point. The one-dimensional tangential adjustment value is then converted into a displacement along the tangent direction according to a preset allocation ratio and superimposed onto the three-dimensional coordinates of the first and second non-common control points respectively. The collinear constraint of the first non-common control points, the common connection point, and the second non-common control points is maintained. After updating the control point positions, the curvature difference vector and the one-dimensional tangential adjustment value are recalculated and iteratively changed until the two curves satisfy a preset continuity condition at the common connection point.

[0016] Optionally, obtaining the initial control points and fitting residuals of the first and second Bézier curves using weighted least squares based on the composite fitting weights includes: Substitute the coordinate parameters of the discretized initial point set and the composite fitting weights into the preset polynomial fitting objective function to construct a weighted error sum of squares equation. Calculate the partial derivatives of each control point variable in the weighted error sum of squares equation, and set each partial derivative to 0 to generate a system of linear normal equations. The three-dimensional coordinates of the initial control point are obtained by solving the linear normal equation system through matrix inversion. Substitute the coordinates of the discretized initial point set into the Bézier curve equation determined by the initial control points, calculate the minimum orthogonal distance sequence from the points to the curve, and obtain the fitting residual.

[0017] Optionally, the step of correlating the newly acquired residual contour error with the three-dimensional compensation vector to obtain the iteratively updated Jacobian matrix includes: Extract the difference between the actual coordinates and the theoretical compensation coordinates of each measuring point obtained by the online measurement module in the next machining pass of the same cutting segment, and construct the current residual contour error vector; Calculate the error difference between the current residual contour error vector and the residual contour error vector of the previous processing pass, and the compensation difference between the three-dimensional compensation vector of the current processing pass and the previous processing pass. Using Breudon's rule, a Jacobian matrix adjustment term is constructed based on the error difference and the compensation difference. The Jacobian matrix adjustment term is then updated by matrix addition with the Jacobian matrix used in the previous processing pass, generating the iteratively updated Jacobian matrix.

[0018] The present invention has the following beneficial effects: A composite fitting weight is constructed by combining local geometric curvature and normalized arc length, and the initial control points of the Bézier curve are obtained by weighted least squares method. This approach balances the local details of complex contours with overall fitting accuracy, reducing initial modeling deviations. Curvature smoothing adjustment amounts are generated using curvature difference, fitting residuals, and spatial displacement conversion coefficients, and the positions of control points on both sides of the connection point are adjusted to improve the continuity and smoothness at the splicing of the double Bézier reference trajectories, reducing machining fluctuations caused by abrupt trajectory changes. Simultaneously, a three-dimensional compensation vector is solved based on the spatial error feature vector of the cutting segment, and the Jacobian matrix is ​​iteratively updated by combining residual contour error feedback. This improves the error compensation capability in multi-pass machining and enhances the accuracy of CNC machining trajectories and the quality of the formed surface. Attached Figure Description

[0019] Figure 1 This is a flowchart of the closed-loop iterative compensation process. Figure 2 This is the convergence curve for error compensation. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] To facilitate understanding of this invention, the terminology involved in this invention will be explained below: Bézier curve: A smooth parametric curve determined by a start point, an end point, and several control points, used to represent complex contours and plan processing trajectories.

[0022] Bernstein basis functions: polynomial weighting functions that constitute the Bézier curve, determining the degree of influence of each control point at different parameter positions on the curve.

[0023] Breuden method: A quasi-Newton iterative method that gradually improves the error mapping model by updating the approximate Jacobian matrix using the changes in error and compensation between two consecutive iterations.

[0024] In the first embodiment, the present invention proposes an intelligent CNC machining control system for tension-resistant power fitting manufacturing equipment, comprising the following modules: The module is used to obtain a discretized initial point set based on the contact-type touch test of the target profile of the tension power fitting. It constructs a composite fitting weight based on the local geometric curvature of each initial point and the normalized arc length distance from the point to the preset common connection point. Based on the composite fitting weight, the initial control points and fitting residuals of the first and second Bézier curves are obtained by weighted least squares method.

[0025] The blank of the tension power fitting is fixed on the CNC machine tool table. The high-precision contact trigger probe mounted on the CNC machine tool spindle is driven to contact the surface of the fitting according to the preset scanning path. The three-dimensional coordinate data of the machine tool mechanical coordinate system at the moment of probe triggering is recorded. After probe radius compensation and Gaussian filtering for smoothing and noise reduction, a discretized initial point set is generated.

[0026] The central difference algorithm is used to calculate the local geometric curvature of each initial point in the discretized initial point set. The spatial chord lengths between adjacent points are then accumulated to obtain the cumulative arc length from each initial point along the contour to a preset common connection point. This cumulative arc length is divided by the total arc length of the trajectory segment containing that initial point, and the ratio is subtracted from 1 to obtain the normalized arc length distance used to characterize the proximity of the common connection point. A linear combination function with local geometric curvature and the proximity of the common connection point as positively correlated terms is established to calculate the composite fitting weight corresponding to each initial point.

[0027] The design matrix is ​​constructed based on the Bernstein basis function of the third-order Bézier curve. A diagonal weight matrix is ​​constructed using composite fitting weights. The discretized initial point set is used as the observation vector. The weighted squared error function is minimized by matrix differentiation. The three-dimensional coordinates of the four initial control points of the first and second Bézier curves are obtained. The shortest Euclidean distance from each point in the discretized initial point set to the corresponding fitted Bézier curve is calculated as the fitting residual.

[0028] In an optional embodiment, the contact-based touch testing based on the target profile of the tension-resistant power fitting to obtain a discretized initial point set includes: The control is set at the end of the machine tool spindle. The contact probe touches the contour surface of the tension power fitting in a predetermined step length according to the mutually parallel guide trajectory. The three-dimensional coordinates of the probe trigger point are recorded to obtain a mixed data point set. Extract the boundary surface of the three-dimensional nominal model and offset it bidirectionally along the normal direction by a preset tolerance to obtain the defined threshold envelope surface; Remove all abnormal noise points in the mixed data point set that exceed the defined threshold envelope, and arrange the actual surface coordinates that remain within the envelope in sequence to obtain the discretized initial point set.

[0029] The CNC system of the machine tool controls a contact probe installed at the end of the spindle to guide a series of parallel scanning paths with a predetermined step size, such as d=0.5 mm, to perform discrete contact testing on the tension power fitting. Each time the probe is triggered, it records a three-dimensional coordinate P(x,y,z) in the machine tool coordinate system. Assuming that a total of N=1000 points are collected, a mixed data point set is formed.

[0030] Import the 3D CAD nominal model of the tension power fitting, extract the boundary mesh data of the model surface, and offset it bidirectionally in the normal direction by a preset tolerance, such as δ=0.2 mm, to generate a boundary threshold envelope surface with a distance of 0.4 mm between the inner and outer surfaces.

[0031] Calculate the minimum distance D from each point in the mixed data point set to the nominal model surface. If D > δ, for example, the distance to a measurement point is 0.35 mm, it is judged as abnormal noise and removed. Arrange, for example, 985 valid actual surface coordinates retained in the envelope plane after filtering according to the spatial order of the scan trajectory to generate a discretized initial point set.

[0032] In an optional embodiment, the step of constructing composite fitting weights based on the local geometric curvature of each initial point and the normalized arc length distance from the point to a preset common connection point includes: The local geometric curvature corresponding to each initial point is calculated by the adjacent three-point microcircle fitting method, and the local geometric curvature is normalized into a curvature weight term. Calculate the cumulative arc length from each initial point to the common connection point along the actual measured trajectory, divide the cumulative arc length by the total arc length of the trajectory segment, and subtract the quotient from 1 to obtain the corresponding normalized arc length distance. The first and second predefined proportion coefficients are called, the curvature weight term is multiplied by the first proportion coefficient, and the normalized arc length distance is multiplied by the second proportion coefficient. The summation yields the composite fitting weight.

[0033] For any point in the initial discretized point set, extract the two adjacent points before and after it. Calculate the radius R of the circumcircle formed by these three points using a three-point microcircle fitting method. Calculate the reciprocal of R to obtain the local geometric curvature C = 1 / R at that point. For example, if the coordinates of the three adjacent points are (0,0,0), (1,0.1,0), and (2,0,0), the calculated circle radius is approximately 5.05 mm, and the curvature C is approximately 0.198. By dividing the curvature of all points by the maximum curvature value of the point set, assuming the maximum curvature is 0.5, the normalized curvature weight term for that point is calculated to be 0.396.

[0034] Calculate the cumulative Euclidean distance from the initial point to the preset common connection point along the actual measured trajectory as the cumulative arc length L, for example, 15 mm. Divide this arc length by the total arc length of the current trajectory segment, for example, 50 mm, and subtract the quotient from 1 to obtain the normalized arc length distance, i.e., 1-(15 / 50)=0.7.

[0035] The preset first proportion coefficient α=0.6 and second proportion coefficient β=0.4 are retrieved, and a weighted operation is performed on these two items to calculate the final composite fitting weight W=0.6×0.396+0.4×0.7=0.5176 for this point. This weight matrix can improve the fitting accuracy in high curvature regions and give discrete points near common connection points a higher constraint effect during the fitting process.

[0036] In an optional embodiment, obtaining the initial control points and fitting residuals of the first and second Bézier curves using weighted least squares based on the composite fitting weights includes: Substitute the coordinate parameters of the discretized initial point set and the composite fitting weights into the preset polynomial fitting objective function to construct a weighted error sum of squares equation. Calculate the partial derivatives of each control point variable in the weighted error sum of squares equation, and set each partial derivative to 0 to generate a system of linear normal equations. The three-dimensional coordinates of the initial control point are obtained by solving the linear normal equation system through matrix inversion. Substitute the coordinates of the discretized initial point set into the Bézier curve equation determined by the initial control points, calculate the minimum orthogonal distance sequence from the points to the curve, and obtain the fitting residual.

[0037] For a Bézier curve of order n, for example, n=3, the three-dimensional coordinates of the i-th discretized initial point within the current curve segment are obtained. The composite fitting weights corresponding to this point and the parameter nodes obtained from the parameterization calculation based on the chord length. Substitute this into the Bézier curve weighted least squares objective function J; where the parameter nodes The objective function is the ratio of the cumulative spatial distance between adjacent points before the i-th discretization initial point to the sum of the spatial distances between adjacent points in the current curve segment. The parameter node corresponding to the first discretization initial point is 0, and the parameter node corresponding to the last discretization initial point is 1. , ,in, For the control points to be determined, The coefficients are binomial. By calculating the squared deviations between the coordinates of each discretized initial point and the theoretical coordinates of the Bézier curve at the corresponding parameter nodes, and multiplying each squared deviation term by the corresponding composite fitting weight, the weighted error sum-of-squares equation is constructed.

[0038] For each unknown control point vector in the objective function J Taking the partial derivatives and setting them to zero, we transform the system into a linear normal equation system of the form A·P=b, where A is a weighted symmetric coefficient matrix of size (n+1)×(n+1). Solving this linear normal equation system using Gauss-Jordan elimination or SVD yields the control point vector P, i.e. to The three-dimensional coordinates of these four initial control points are used to substitute the initial control points into the Bézier curve equation to determine the curve equations of the first and second Bézier curves. Then, each initial measurement point is traversed, and the minimum orthogonal projection point corresponding to that point on the curve is calculated using the Newton-Raphson iteration method. By calculating the Euclidean distance between these two points, the minimum orthogonal distance sequence containing the fitting deviation values ​​of all points is obtained, and the fitting residual is obtained.

[0039] The adjustment module is used to obtain the curvature difference between the two curves at the common connection point and the transition influence factor constructed based on the fitting residual. It obtains the curvature smoothing adjustment amount based on the curvature difference, the transition influence factor and the preset spatial displacement conversion coefficient, and generates an adjustment vector based on the curvature smoothing adjustment amount and the tangent direction at the common connection point. It adjusts the positions of the non-common control points on both sides of the connection point based on the adjustment vector to obtain the double Bezier reference trajectory.

[0040] The first curvature of the first Bézier curve at its tail end and the second curvature of the second Bézier curve at its head end are calculated using the curve curvature calculation formula. The difference in curvature between the first and second curvatures is obtained by subtracting the first curvature from the second curvature and taking the absolute value.

[0041] Extract the fitting residuals within a specified neighborhood of the common connection point, and calculate the dimensionless transition influence factor based on the fitting residuals and a preset smoothness threshold.

[0042] Multiply the curvature difference by the transition influence factor, and then multiply by the preset spatial displacement conversion coefficient to obtain the curvature smoothing adjustment amount.

[0043] Based on the curvature smoothing adjustment amount and the unit tangent direction at the common connection point, positive candidate adjustment vectors and negative candidate adjustment vectors are constructed respectively. The two candidate adjustment vectors are applied to the non-common control points on both sides of the connection point according to a preset allocation ratio. The non-common control points on both sides are moved relative to each other along the direction of the candidate adjustment vector and the opposite direction, and the curvature difference after the trial calculation is recalculated. The curvature difference after the trial calculation is compared with the curvature difference before adjustment, and the candidate adjustment vector that can reduce the curvature difference and has a better reduction effect is selected as the actual adjustment vector.

[0044] After determining the actual adjustment vector, the second-to-last initial control point of the first Bézier curve and the second-to-last initial control point of the second Bézier curve are modified by three-dimensional coordinate translation based on the actual adjustment vector. The translation direction is determined by the actual adjustment vector, and the translation distance is determined by the curvature smoothing adjustment amount and a preset allocation ratio. By adjusting the positions of these two non-common control points, the curvature difference between the two curve segments at the common connection point is reduced, making the splicing trajectory smoother and obtaining a double Bézier reference trajectory with reduced curvature abrupt changes. The overall process of double Bézier closed-loop iterative compensation for tension power fitting contours is as follows: Figure 1 As shown.

[0045] In an optional embodiment, the step of obtaining the curvature difference between the two curves at the common connection point and the transition influence factor constructed based on the fitting residual, obtaining a curvature smoothing adjustment amount based on the curvature difference, the transition influence factor, and a preset spatial displacement conversion coefficient, and generating an adjustment vector based on the curvature smoothing adjustment amount and the tangent direction at the common connection point includes: Calculate the first and second derivatives of the first and second Bézier curves at the common connection point, respectively, and calculate the curvature difference vector of the two curves at this point. Calculate the sum of the preset smoothness threshold and the maximum value of the fitting residual, and divide the preset smoothness threshold by the sum to obtain the transition influence factor; Calculate the magnitude of the curvature difference vector, and multiply the magnitude of the curvature difference vector by the transition influence factor and the preset spatial displacement conversion coefficient to obtain the curvature smoothing adjustment amount; The unit tangent direction is determined based on the first derivative at the common connection point. A positive candidate adjustment vector and a negative candidate adjustment vector are constructed based on the curvature smoothing adjustment amount and the unit tangent direction. The actual adjustment vector is selected based on the comparison between the curvature difference after the two candidate adjustment vectors are applied and the curvature difference before adjustment.

[0046] Using calculus algorithms, the first and second derivatives of the first and second Bézier curves at the common connection point, corresponding to parameters u=1 and v=0, are calculated respectively. Substituting these derivatives into the curvature formula and its corresponding normal vector, the three-dimensional curvature difference vector ΔK at the splicing point of the two curves is obtained. Assuming that ΔK=[0.015,-0.008,0.002], the unit is 1 / mm.

[0047] Calculate the magnitude of the three-dimensional curvature difference vector ΔK, assuming |ΔK| is approximately 0.017. Obtain the maximum value of the fitting residual. ,like =0.02 mm, retrieve the preset smoothness threshold. =0.01 mm, through the formula =0.01 / (0.02+0.01)=0.333, thus obtaining the transitional impact factor. .

[0048] Read the preset spatial displacement conversion coefficient λ, for example, λ = 50.0 square millimeters. This coefficient is used to convert the difference in curvature domain into displacement in the position domain. The formula Δs = |ΔK| × ×λ, the curvature smoothing adjustment amount Δs is calculated to be approximately 0.285 mm.

[0049] The unit tangent direction t is determined based on the first derivative at the common junction, for example, t=[1,0,0]. Positive candidate adjustment vectors are then constructed based on the curvature smoothing adjustment amount Δs. =Δs·t and reverse candidate adjustment vector =-Δs·t. Adjust the positive candidate vector. The calculations, based on the preset allocation ratio, apply to the first and second non-common control points, where the first non-common control point is along... Direction movement, second non-common control point along Move in the opposite direction to obtain the first set of candidate control points; adjust the reverse candidate vector. The same trial calculations were performed on the first and second non-common control points to obtain a second set of candidate control points. The curvature difference of the common connection point corresponding to the two sets of trial calculations was then recalculated. The first non-common control point is a control point in the first Bézier curve that is adjacent to but does not coincide with the common connection point, and the second non-common control point is a control point in the second Bézier curve that is adjacent to but does not coincide with the common connection point.

[0050] Compare the two sets of calculated curvature difference values ​​with the curvature difference values ​​before adjustment, and select the candidate adjustment vector that can reduce the curvature difference value by a large amount as the actual adjustment vector V.

[0051] In an optional embodiment, adjusting the positions of the non-common control points on both sides of the connection point based on the adjustment vector to obtain the double Bezier reference trajectory includes: The signed length of the adjustment vector along the tangent direction of the common connection point is read as the tangential one-dimensional adjustment value; Extract the original three-dimensional coordinates of the first non-common control point and the second non-common control point on both sides of the common connection point; The first and second non-common control points are orthogonally projected onto a tangent constraint line that passes through the common connection point and is oriented in the tangent direction at the common connection point. The one-dimensional tangential adjustment value is then converted into a displacement along the tangent direction according to a preset allocation ratio and superimposed onto the three-dimensional coordinates of the first and second non-common control points respectively. The collinear constraint of the first non-common control points, the common connection point, and the second non-common control points is maintained. After updating the control point positions, the curvature difference vector and the one-dimensional tangential adjustment value are recalculated and iteratively changed until the two curves satisfy a preset continuity condition at the common connection point.

[0052] After obtaining the adjustment vector, read the signed length of the adjustment vector along the tangent direction of the common connection point to obtain the one-dimensional tangential adjustment value Δs. For example, when the actual adjustment vector is selected as... At that time, Δs = 0.285 mm; when the actual adjustment vector is selected as At that time, Δs = -0.285 mm.

[0053] Read public connection point The first non-common control point on both sides of (10,5,0) (8,5,0) and the second non-common control point The original three-dimensional coordinates of (13,5,0).

[0054] Will and Orthogonally projected onto the path that passes through And on the tangent constraint line with the unit tangent direction t, the projected control points are still denoted as... and These are the control points to be adjusted to satisfy the collinearity constraint. The preset allocation ratio includes a first allocation ratio. Second allocation ratio ,and + =1, for example =0.4、 =0.6; The preset allocation ratio is the same as the allocation ratio used in the candidate adjustment vector trial calculation. The actual displacement offset is calculated according to the preset allocation ratio and converted into a vector along the tangent direction. These vectors are then superimposed onto the projected control point coordinates to obtain the updated control points. and ,at this time , and The three points are kept collinear. Since Δs is a signed value determined based on the results of forward and reverse candidate calculations, the update formula can adaptively change the direction of control point movement according to the direction of decrease in curvature difference.

[0055] After updating the control point position, recalculate the new curvature difference and new one-dimensional adjustment value of the two curves at this point, and perform iterative changes until the connection point meets the preset continuity conditions. The preset continuity conditions include continuity in the direction of the first derivative and curvature difference less than the preset continuity threshold. At this point, stop the iteration and output the determined double Bezier reference trajectory. The preset continuity threshold is preferably 0.002 / mm.

[0056] The update module is used to control the tool to cut along the double Bezier reference trajectory to obtain the spatial error feature vector of the same cutting segment, call the Jacobian matrix to solve for the three-dimensional compensation vector of the next machining pass and update the control point coordinates to obtain the compensatory machining path. After completing the next machining pass along the compensatory machining path, the newly acquired residual contour error is correlated with the three-dimensional compensation vector to obtain the iteratively updated Jacobian matrix.

[0057] The CNC system's built-in interpolator generates CNC machining instructions based on the double Bezier reference trajectory, driving the servo motor to control the ball end mill to perform the first solid cut at the set feed rate and spindle speed. After the cut is completed, the machine tool's contact probe is called again to perform on-machine measurement of the machined surface of the cut section. The measured contour data is spatially registered and compared with the theoretical target contour, and the positional deviation values ​​of each corresponding sampling point in the three-dimensional orthogonal coordinate system are extracted to form the first-pass residual contour error vector. The first-pass residual contour error vector is used as the spatial error feature vector.

[0058] In the control system, a pre-initialized Jacobian matrix is ​​invoked, which is obtained through offline calibration. Specifically, during trial machining or simulation machining, preset small disturbances are applied to the coordinates of the control points to be compensated in the double Bezier reference trajectory. The changes in the residual contour error vector before and after the disturbance are recorded, and the ratio of the change in the residual contour error vector to the corresponding disturbance of the control point is used as the corresponding element of the Jacobian matrix, thereby constructing the initial Jacobian matrix. This matrix represents the nonlinear sensitivity mapping relationship between the small displacement of the Bezier control point and the contour error of the machining trajectory. The pseudo-inverse matrix of the Jacobian matrix is ​​calculated using the singular value decomposition method. The residual contour error vector of the first pass is inverted according to the preset compensation polarity and multiplied by the pseudo-inverse matrix to calculate the three-dimensional compensation vector of the control point used to reduce the current contour error.

[0059] The three-dimensional compensation vector of the control point is split into corresponding three-dimensional compensation components according to the coordinate order of the control point to be compensated, and then superimposed on the coordinates of the corresponding control point in the double Bezier reference trajectory. The compensation machining path is regenerated based on the new control point after displacement update, and the cutting program for the next machining pass is compiled.

[0060] For the first pass where cutting follows the double Bezier reference path, the corresponding three-dimensional compensation vector is recorded as the zero vector, serving as the initial reference. After the CNC machine tool completes the next machining pass along the compensatory machining path, the probe scans the machined surface again to obtain the residual contour error vector of the current machining pass. The residual contour error vector of the current machining pass is subtracted from the residual contour error vector of the previous machining pass to obtain the error change; the three-dimensional compensation vector used in the current machining pass is subtracted from the three-dimensional compensation vector used in the previous machining pass to obtain the compensation change.

[0061] Using the Broyden quasi-Newton iterative algorithm, the Jacobian matrix update increment is calculated based on the error change, the compensation change, and the Jacobian matrix used in the previous processing pass. The update increment is then superimposed on the previous Jacobian matrix to obtain the iteratively updated Jacobian matrix, enabling the error mapping relationship to be gradually corrected as the processing passes progress.

[0062] In an optional embodiment, the step of correlating the newly acquired residual contour error with the three-dimensional compensation vector to obtain the iteratively updated Jacobian matrix includes: Extract the difference between the actual coordinates and the theoretical compensation coordinates of each measuring point obtained by the online measurement module in the next machining pass of the same cutting segment, and construct the current residual contour error vector; Calculate the error difference between the current residual contour error vector and the residual contour error vector of the previous processing pass, and the compensation difference between the three-dimensional compensation vector of the current processing pass and the previous processing pass. Using Breudon's rule, a Jacobian matrix adjustment term is constructed based on the error difference and the compensation difference. The Jacobian matrix adjustment term is then updated by matrix addition with the Jacobian matrix used in the previous processing pass, generating the iteratively updated Jacobian matrix.

[0063] After completing the m-th machining pass, for example, m=2, the actual coordinates of each inspection point on the cutting section are captured by the servo online measurement module. By subtracting the theoretical compensation coordinates of that pass, a current residual contour error vector in the form of E(m)=[0.015,-0.005,0.002] mm is constructed.

[0064] Retrieve the residual contour error vector E(m-1) from the previous processing pass (m-1), the 3D compensation vector ΔC(m) used in the current processing pass, and the 3D compensation vector ΔC(m-1) from the previous processing pass. When the previous processing pass is the first pass and the first pass does not use a compensatory processing path, ΔC(m-1) is recorded as the zero vector. Perform a subtraction operation to obtain the error difference ΔE = E(m) - E(m-1) and the corresponding compensation difference variable ΔU = ΔC(m) - ΔC(m-1). Assume the calculated compensation variation is [0.02, 0, 0.01]. Using the Breudend method, denote the Jacobian matrix of the previous processing pass as J(m-1), and use the formula... A Jacobian matrix adjustment term ΔJ is constructed, and this adjustment term is added to J(m-1) using a matrix addition operation, i.e., J(m) = J(m-1) + ΔJ, to generate the iteratively updated Jacobian matrix. The convergence characteristics of contour residual error compensation under multi-pass machining are as follows: Figure 2 As shown.

[0065] In the second embodiment, the present invention also proposes an intelligent CNC machining control method for tension-resistant power fitting manufacturing equipment, comprising the following steps: Based on the contact-type touch test of the target profile of tension power fitting, a discretized initial point set is obtained. A composite fitting weight is constructed according to the local geometric curvature of each initial point and the normalized arc length distance from the point to the preset common connection point. Based on the composite fitting weight, the initial control points and fitting residuals of the first and second Bézier curves are obtained by weighted least squares method. Obtain the curvature difference between the two curves at the common connection point and the transition influence factor constructed based on the fitting residual. Obtain the curvature smoothing adjustment amount based on the curvature difference, the transition influence factor and the preset spatial displacement conversion coefficient. Generate an adjustment vector based on the curvature smoothing adjustment amount and the tangent direction at the common connection point. Adjust the positions of the non-common control points on both sides of the connection point based on the adjustment vector to obtain the double Bezier reference trajectory. The tool is controlled to cut along the double Bezier reference trajectory to obtain the spatial error feature vector of the same cutting segment. The Jacobian matrix is ​​called to solve for the three-dimensional compensation vector of the next machining pass and update the control point coordinates to obtain the compensatory machining path. After the next machining pass is completed along the compensatory machining path, the newly acquired residual contour error is correlated with the three-dimensional compensation vector to obtain the iteratively updated Jacobian matrix.

[0066] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. An intelligent CNC machining control system for a tension-resistant power fitting manufacturing equipment, characterized in that, Includes the following modules: The module is used to obtain a discretized initial point set based on the contact touch test of the target profile of the tension power fitting. It constructs a composite fitting weight based on the local geometric curvature of each initial point and the normalized arc length distance from the point to the preset common connection point. Based on the composite fitting weight, the initial control points and fitting residuals of the first and second Bézier curves are obtained by weighted least squares method. The adjustment module is used to obtain the curvature difference between the two curves at the common connection point and the transition influence factor constructed based on the fitting residual. It obtains the curvature smoothing adjustment amount based on the curvature difference, the transition influence factor and the preset spatial displacement conversion coefficient, and generates an adjustment vector based on the curvature smoothing adjustment amount and the tangent direction at the common connection point. It adjusts the positions of the non-common control points on both sides of the connection point based on the adjustment vector to obtain the double Bezier reference trajectory. The update module is used to control the tool to cut along the double Bezier reference trajectory to obtain the spatial error feature vector of the same cutting segment, call the Jacobian matrix to solve for the three-dimensional compensation vector of the next machining pass and update the control point coordinates to obtain the compensatory machining path. After completing the next machining pass along the compensatory machining path, the newly acquired residual contour error is correlated with the three-dimensional compensation vector to obtain the iteratively updated Jacobian matrix.

2. The system according to claim 1, characterized in that, The contact-based touch testing method based on the target contour of tension-resistant power fittings to obtain a discretized initial point set includes: The control is set at the end of the machine tool spindle. The contact probe touches the contour surface of the tension power fitting in a predetermined step length according to the mutually parallel guide trajectory. The three-dimensional coordinates of the probe trigger point are recorded to obtain a mixed data point set. Extract the boundary surface of the three-dimensional nominal model and offset it bidirectionally along the normal direction by a preset tolerance to obtain the defined threshold envelope surface; Remove all abnormal noise points in the mixed data point set that exceed the defined threshold envelope, and arrange the actual surface coordinates that remain within the envelope in sequence to obtain the discretized initial point set.

3. The system according to claim 1, characterized in that, The step of constructing composite fitting weights based on the local geometric curvature of each initial point and the normalized arc length distance from the point to the preset common connection point includes: The local geometric curvature corresponding to each initial point is calculated by the adjacent three-point microcircle fitting method, and the local geometric curvature is normalized into a curvature weight term. Calculate the cumulative arc length from each initial point to the common connection point along the actual measured trajectory, divide the cumulative arc length by the total arc length of the trajectory segment, and subtract the quotient from 1 to obtain the corresponding normalized arc length distance. The first and second predefined proportion coefficients are called, the curvature weight term is multiplied by the first proportion coefficient, and the normalized arc length distance is multiplied by the second proportion coefficient. The summation yields the composite fitting weight.

4. The system according to claim 1, characterized in that, The process of obtaining the curvature difference between the two curves at the common connection point and the transition influence factor constructed based on the fitting residual, obtaining the curvature smoothing adjustment amount based on the curvature difference, the transition influence factor, and a preset spatial displacement conversion coefficient, and generating an adjustment vector based on the curvature smoothing adjustment amount and the tangent direction at the common connection point includes: Calculate the first and second derivatives of the first and second Bézier curves at the common connection point, respectively, and calculate the curvature difference vector of the two curves at this point. Calculate the sum of the preset smoothness threshold and the maximum value of the fitting residual, and divide the preset smoothness threshold by the sum to obtain the transition influence factor; Calculate the magnitude of the curvature difference vector, and multiply the magnitude of the curvature difference vector by the transition influence factor and the preset spatial displacement conversion coefficient to obtain the curvature smoothing adjustment amount; The unit tangent direction is determined based on the first derivative at the common connection point. A positive candidate adjustment vector and a negative candidate adjustment vector are constructed based on the curvature smoothing adjustment amount and the unit tangent direction. The actual adjustment vector is selected based on the comparison between the curvature difference after the two candidate adjustment vectors are applied and the curvature difference before adjustment.

5. The system according to claim 4, characterized in that, The process of adjusting the positions of the non-common control points on both sides of the connection point based on the adjustment vector to obtain the double Bezier reference trajectory includes: The signed length of the adjustment vector along the tangent direction of the common connection point is read as the tangential one-dimensional adjustment value; Extract the original three-dimensional coordinates of the first non-common control point and the second non-common control point on both sides of the common connection point; The first and second non-common control points are orthogonally projected onto a tangent constraint line that passes through the common connection point and is oriented in the tangent direction at the common connection point. The one-dimensional tangential adjustment value is then converted into a displacement along the tangent direction according to a preset allocation ratio and superimposed onto the three-dimensional coordinates of the first and second non-common control points respectively. The collinear constraint of the first non-common control points, the common connection point, and the second non-common control points is maintained. After updating the control point positions, the curvature difference vector and the one-dimensional tangential adjustment value are recalculated and iteratively changed until the two curves satisfy a preset continuity condition at the common connection point.

6. The system according to claim 1, characterized in that, The process of obtaining the initial control points and fitting residuals of the first and second Bézier curves using the weighted least squares method based on the composite fitting weights includes: Substitute the coordinate parameters of the discretized initial point set and the composite fitting weights into the preset polynomial fitting objective function to construct a weighted error sum of squares equation. Calculate the partial derivatives of each control point variable in the weighted error sum of squares equation, and set each partial derivative to 0 to generate a system of linear normal equations. The three-dimensional coordinates of the initial control point are obtained by solving the linear normal equation system through matrix inversion. Substitute the coordinates of the discretized initial point set into the Bézier curve equation determined by the initial control points, calculate the minimum orthogonal distance sequence from the points to the curve, and obtain the fitting residual.

7. The system according to claim 1, characterized in that, The Jacobian matrix is ​​obtained by correlating the newly acquired residual contour error with the three-dimensional compensation vector to obtain the iteratively updated matrix, including: Extract the difference between the actual coordinates and the theoretical compensation coordinates of each measuring point obtained by the online measurement module in the next machining pass of the same cutting segment, and construct the current residual contour error vector; Calculate the error difference between the current residual contour error vector and the residual contour error vector of the previous processing pass, and the compensation difference between the three-dimensional compensation vector of the current processing pass and the previous processing pass. Using Breudon's rule, a Jacobian matrix adjustment term is constructed based on the error difference and the compensation difference. The Jacobian matrix adjustment term is then updated by matrix addition with the Jacobian matrix used in the previous processing pass, generating the iteratively updated Jacobian matrix.

8. A method for intelligent numerical control machining of tension-resistant power fitting manufacturing equipment, characterized in that, Includes the following steps: Based on the contact-type touch test of the target profile of tension power fitting, a discretized initial point set is obtained. A composite fitting weight is constructed according to the local geometric curvature of each initial point and the normalized arc length distance from the point to the preset common connection point. Based on the composite fitting weight, the initial control points and fitting residuals of the first and second Bézier curves are obtained by weighted least squares method. Obtain the curvature difference between the two curves at the common connection point and the transition influence factor constructed based on the fitting residual. Obtain the curvature smoothing adjustment amount based on the curvature difference, the transition influence factor and the preset spatial displacement conversion coefficient. Generate an adjustment vector based on the curvature smoothing adjustment amount and the tangent direction at the common connection point. Adjust the positions of the non-common control points on both sides of the connection point based on the adjustment vector to obtain the double Bezier reference trajectory. The tool is controlled to cut along the double Bezier reference trajectory to obtain the spatial error feature vector of the same cutting segment. The Jacobian matrix is ​​called to solve for the three-dimensional compensation vector of the next machining pass and update the control point coordinates to obtain the compensatory machining path. After the next machining pass is completed along the compensatory machining path, the newly acquired residual contour error is correlated with the three-dimensional compensation vector to obtain the iteratively updated Jacobian matrix.

9. The method according to claim 8, characterized in that, The contact-based touch testing method based on the target contour of tension-resistant power fittings to obtain a discretized initial point set includes: The control is set at the end of the machine tool spindle. The contact probe touches the contour surface of the tension power fitting in a predetermined step length according to the mutually parallel guide trajectory. The three-dimensional coordinates of the probe trigger point are recorded to obtain a mixed data point set. Extract the boundary surface of the three-dimensional nominal model and offset it bidirectionally along the normal direction by a preset tolerance to obtain the defined threshold envelope surface; Remove all abnormal noise points in the mixed data point set that exceed the defined threshold envelope, and arrange the actual surface coordinates that remain within the envelope in sequence to obtain the discretized initial point set.

10. The method according to claim 8, characterized in that, The step of constructing composite fitting weights based on the local geometric curvature of each initial point and the normalized arc length distance from the point to the preset common connection point includes: The local geometric curvature corresponding to each initial point is calculated by the adjacent three-point microcircle fitting method, and the local geometric curvature is normalized into a curvature weight term. Calculate the cumulative arc length from each initial point to the common connection point along the actual measured trajectory, divide the cumulative arc length by the total arc length of the trajectory segment, and subtract the quotient from 1 to obtain the corresponding normalized arc length distance. The first and second predefined proportion coefficients are called, the curvature weight term is multiplied by the first proportion coefficient, and the normalized arc length distance is multiplied by the second proportion coefficient. The summation yields the composite fitting weight.

Images