A process optimization method for complex curved surfaces

Abstract

This invention belongs to the field of complex surface process optimization technology. It discloses a process optimization method for complex surfaces. The method includes: acquiring a three-dimensional structural diagram of the workpiece surface; dividing the workpiece surface into N sub-surfaces based on the three-dimensional model; extracting features from the N sub-surfaces to obtain corresponding feature data; and adaptively optimizing the idle travel path of the N sub-surfaces based on the feature data. This solution overcomes the limitations of traditional CNC machining by employing an intelligent optimization algorithm for idle travel path optimization. It optimizes the machining sequence of the sub-surfaces, automatically selecting the optimal machining sequence to reduce non-cutting idle travel and repeated cutting, thereby improving the machining quality of the workpiece surface.

Description

Technical Field

[0001] This invention relates to the field of process optimization technology for complex curved surfaces, and more specifically, to a process optimization method for complex curved surfaces. Background Technology

[0002] CNC machining technology, with its superior machining accuracy and stability, has been widely applied, and valve surface machining is one of its important applications. With the rapid development of modern industry, valves are undertaking increasingly critical fluid control tasks under various complex operating conditions, prompting valve surface structures to continuously evolve towards higher complexity and multi-surface combinations. Today, valve surface shapes are becoming increasingly complex and diverse, such as irregular streamlined shapes and multi-curvature composite surfaces, bringing unprecedented challenges to valve manufacturing. The machining quality of valve surfaces has become one of the core factors determining the overall performance of a valve.

[0003] A Chinese patent application with authorization publication number CN109901514B proposes a method for optimizing and adjusting the CNC process of complex parts for process reuse, aiming to solve the technical problem of low efficiency in existing CNC process optimization and adjustment methods. The technical solution optimizes and adjusts both the machining tools and the depth of cut. Regarding the machining tools, it optimizes tool combinations for complex features and merges and adjusts tools for machining local structures, thereby reducing tool changes and shortening idle tool travel. Regarding the depth of cut parameter, it optimizes the depth of cut by deciding on a reasonable cutting mode, obtaining an optimized machining area, thus fully utilizing the tool's machining capabilities, ultimately achieving a highly efficient CNC process solution and improving the machining efficiency of complex parts.

[0004] In valve surface machining, existing technologies generally employ pre-set machining paths and surface parameters. This traditional machining method has significant drawbacks. Due to the lack of real-time dynamic monitoring and adaptive adjustment mechanisms for the valve surface machining process, it is difficult to simultaneously consider multiple key performance indicators. This often results in a large number of non-cutting idle strokes and repeated cutting during machining, which prolongs the machining time and affects the machining quality of the valve surface. Ultimately, it fails to meet the high precision and high efficiency requirements of modern industry for valve surface machining.

[0005] Therefore, developing a valve surface CNC machining technology with real-time monitoring and adaptive adjustment capabilities to achieve synergistic optimization of multiple key performance indicators has become an important problem that urgently needs to be solved in the valve manufacturing field. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a process optimization method for complex curved surfaces, comprising:

[0007] Obtain the 3D structural diagram of the machined part's curved surface;

[0008] Based on the 3D model of the workpiece surface, the workpiece surface is divided into N sub-surfaces;

[0009] Feature extraction is performed on N sub-surfaces of the machined parts to obtain the feature data of the machined parts sub-surfaces corresponding to the N sub-surfaces of the machined parts;

[0010] The idle travel path for machining the surface of N machining parts is adaptively optimized based on the feature data of N machining parts sub-surfaces.

[0011] Furthermore, the method for adaptively optimizing the idle travel path of machining the N sub-surfaces of the workpiece based on the feature data of the N sub-surfaces of the workpiece includes:

[0012] Input the feature data of N sub-surfaces of the workpiece into the machining path evaluation model to obtain H candidate machining paths for the workpiece; each candidate machining path contains the machining sequence of the N sub-surfaces of the workpiece; construct a machining path population from the H candidate machining paths;

[0013] The fitness of the H candidate machining paths in the machining path population is evaluated to obtain the corresponding machining path fitness values.

[0014] Sort the fitness values ​​of the H processing paths in descending order, and select the top Q processing paths fitness values ​​to construct an elite population of processing paths;

[0015] The processing path elite population is mutated according to a preset method to obtain R mutated processing candidate paths and added to the processing path elite population; the processing path elite population is then overlaid on the processing path population, and H = Q + R is set.

[0016] Repeat the above operation until the number of iterations reaches the preset maximum number of iterations, then stop execution; select the candidate path for machining the workpiece with the largest fitness value from the machining path population as the optimal solution for the empty travel path of the workpiece surface machining.

[0017] Furthermore, the method for obtaining the R candidate processing paths for the mutated workpieces includes:

[0018] Preset path fitness threshold 1 and path fitness threshold 2, where path fitness threshold 1 is less than path fitness threshold 2; divide the elite population of processing paths according to path fitness threshold 1 and path fitness threshold 2 to obtain a set of high-fitness paths, a set of medium-fitness paths, and a set of low-fitness paths.

[0019] Randomly cross-reference each other in the high-fitness path set to obtain NUM1 candidate processing paths for the workpiece, and add them to the high-fitness path set;

[0020] Randomly cross the processing candidate paths of the workpiece in the medium fitness path set with the processing candidate paths of the workpiece in the high fitness path set to obtain NUM2 processing candidate paths, which are then added to the medium fitness path set; randomly cross the processing candidate paths in the medium fitness path set to obtain NUM3 processing candidate paths, which are then added to the medium fitness path set.

[0021] Randomly cross the processing candidate paths of the workpiece in the low fitness path set with the processing candidate paths of the workpiece in the high fitness path set to obtain NUM4 processing candidate paths, which are then added to the low fitness path set; randomly cross the processing candidate paths of the workpiece in the low fitness path set to obtain NUM5 processing candidate paths, which are then added to the low fitness path set.

[0022] When NUM1+NUM2+NUM3+NUM4+NUM5=R, the above random crossover operation is stopped, and R candidate paths for processing mutated parts are obtained.

[0023] Furthermore, the method for obtaining the high-fitness path set, the medium-fitness path set, and the low-fitness path set includes:

[0024] Candidate processing paths for workpieces from the elite processing path population whose processing path fitness values ​​are greater than or equal to the path fitness threshold two are constructed into a high-fitness path set; candidate processing paths for workpieces from the elite processing path population whose processing path fitness values ​​are greater than or equal to the path fitness threshold one and less than the path fitness threshold two are constructed into a medium-fitness path set; and candidate processing paths for workpieces from the elite processing path population whose processing path fitness values ​​are less than the path fitness threshold one are constructed into a low-fitness path set.

[0025] Furthermore, the method for obtaining the N sub-surfaces of the machined part includes:

[0026] S100: M sampling points are randomly distributed on the three-dimensional structural diagram of the machined part's curved surface;

[0027] S101: Let m be initialized to 1, and the range of m is from 1 to M; let n be initialized to 1, and n is a counting variable;

[0028] S102: Obtain the coordinates of the workpiece surface parameters at the m-th sampling point, and obtain the corresponding workpiece surface parameter equation based on the workpiece surface parameter coordinates; calculate the curvature of the workpiece surface at the m-th sampling point based on the workpiece surface parameter equation;

[0029] S103: Determine the corresponding workpiece surface movement step size and workpiece sub-surface type based on the workpiece surface curvature at the m-th sampling point; perform a diffusion operation based on the workpiece surface movement step size and workpiece sub-surface type to obtain the workpiece surface diffusion region at the m-th sampling point;

[0030] S104: Using the edge point of the diffusion region of the workpiece surface at the m-th sampling point as the new diffusion starting point, repeat S103. When there are no more points of the same type as the workpiece sub-surface at the m-th sampling point during the diffusion process, the diffusion stops, and the workpiece sub-surface at the n-th sampling point is obtained; let n = n + 1, and execute S105;

[0031] S105: Let m = m + 1. If m is less than or equal to M, then execute S106; if m is greater than M, then end the current process, assign n to N, and obtain N sub-surfaces of the processed parts.

[0032] S106: If the m-th sampling point is already included in the constructed sub-surface of the machined part, then execute S105; if the m-th sampling point is not included in the constructed sub-surface of the machined part, then execute S102 to S104.

[0033] Furthermore, the method for obtaining the diffusion region of the workpiece surface at the m-th sampling point includes:

[0034] With the m-th sampling point as the center and the workpiece surface movement step size as the radius, a diffusion processing area for the m-th sampling point is formed; the curvature of the workpiece surface at the point on the edge of the diffusion processing area is calculated and denoted as the edge surface curvature; the corresponding workpiece sub-surface type is obtained based on the edge surface curvature.

[0035] Mark the points on the edge of the diffusion region to be processed that have the same sub-surface type as the sub-surface type of the workpiece at the m-th sampling point as the connection points to be connected at the m-th sampling point; connect all the connection points to be connected at the m-th sampling point to form the diffusion region of the workpiece surface at the m-th sampling point.

[0036] Furthermore, the method for obtaining the curvature of the workpiece surface at the m-th sampling point includes:

[0037] Parametric equations of the surface of the machined part for u m Find the first-order partial derivative to obtain the first first-order partial derivative; the parametric equation of the machined part surface with respect to v mFind the first-order partial derivative to obtain the second-order partial derivative; u m and v m Given the surface parameter coordinates of the workpiece at the m-th sampling point, the unit normal vector of the m-th sampling point is calculated based on the first and second first-order partial derivatives.

[0038] Multiplying the first-order partial derivatives together yields the corresponding first-order formal parameters; multiplying the first-order partial derivatives together yields the corresponding second-order formal parameters; multiplying the second-order partial derivatives together yields the corresponding third-order formal parameters.

[0039] The first-order partial derivative with respect to u m Find the second-order partial derivative to obtain the first-order partial derivative; the first-order partial derivative with respect to v m Find the second partial derivative to obtain the second second partial derivative; the second first partial derivative with respect to v m Find the second-order partial derivative to obtain the third second-order partial derivative;

[0040] Multiplying the first second-order partial derivative with the unit normal vector yields the first second-order formal parameter of the m-th sampling point; multiplying the second second-order partial derivative with the unit normal vector yields the second second-order formal parameter of the m-th sampling point; multiplying the third second-order partial derivative with the unit normal vector yields the third second-order formal parameter of the m-th sampling point.

[0041] The curvature of the workpiece surface at the m-th sampling point is calculated based on the first-order formal parameters, the second-order formal parameters, the third-order formal parameters, the first-second-order formal parameters, the second-second-order formal parameters, and the third-second-order formal parameters.

[0042] Furthermore, the method for determining the corresponding workpiece surface movement step size and workpiece sub-surface type based on the workpiece surface curvature at the m-th sampling point includes:

[0043] Preset a workpiece surface curvature threshold 1 and a workpiece surface curvature threshold 2, wherein the workpiece surface curvature threshold 1 is less than the workpiece surface curvature threshold 2; preset a workpiece surface first movement step, a workpiece surface second movement step, and a workpiece surface third movement step, wherein the workpiece surface first movement step is less than the workpiece surface second movement step, and the workpiece surface second movement step is less than the workpiece surface third movement step;

[0044] If the curvature of the workpiece surface at the m-th sampling point is greater than or equal to the workpiece surface curvature threshold two, then the workpiece sub-surface type is a high curvature region, and the workpiece surface movement step size is the first movement step size of the workpiece surface.

[0045] If the curvature of the workpiece surface at the m-th sampling point is greater than or equal to the workpiece surface curvature threshold one and less than the workpiece surface curvature threshold two, then the workpiece sub-surface type is a medium curvature region, and the workpiece surface movement step size is the workpiece surface second movement step size.

[0046] If the curvature of the workpiece surface at the m-th sampling point is less than the workpiece surface curvature threshold, then the workpiece sub-surface type is a low curvature region, and the workpiece surface movement step size is the third movement step size of the workpiece surface.

[0047] Furthermore, the methods for obtaining the feature data of the N machined part sub-surfaces include:

[0048] S200: Let the initial value of n be 1, and the range of n is from 1 to N;

[0049] S201: Obtain the coordinate range, first-order formal parameters, second-order formal parameters, and third-order formal parameters of the nth workpiece sub-surface; calculate the corresponding workpiece sub-surface area based on the coordinate range, first-order formal parameters, second-order formal parameters, and third-order formal parameters; divide the boundary of the nth workpiece sub-surface into J boundary points, and calculate the boundary length of the nth workpiece sub-surface based on the coordinates of the J boundary points;

[0050] The curvature of the workpiece surface at J boundary points is obtained using the same method as for obtaining the curvature of the workpiece surface; the surface boundary complexity of the nth workpiece sub-surface is calculated based on the curvature of the workpiece surface at J boundary points.

[0051] S202: Construct the corresponding machining part feature data from the area, boundary length, and boundary complexity of the nth machining part sub-surface;

[0052] S203: Let n = n + 1. If n is less than or equal to N, continue executing S201 to S202. If n is greater than N, end the current process.

[0053] Furthermore, the training method for the processing path evaluation model includes:

[0054] A pre-constructed processing path evaluation dataset is provided, comprising Y sets of processing path evaluation data and H candidate processing paths for the corresponding parts, where Y is a positive integer greater than 0. The processing path evaluation data includes N sub-surface feature data of the parts. The processing path evaluation dataset is divided into a processing path evaluation data training set and a processing path evaluation data validation set. The processing path evaluation data training set is used for parameter learning of the processing path evaluation model, and the processing path evaluation data validation set is used for real-time evaluation of the generalization ability of the processing path evaluation model.

[0055] During the training of the processing path evaluation model, a deep neural network structure based on multilayer perceptrons is adopted. The processing path evaluation data is converted into feature vectors as input, and nonlinear features in the data are extracted through multiple hidden layers. Finally, the softmax activation function is used in the output layer to generate the probability distribution of H processing candidate paths for processing parts. The H processing candidate paths corresponding to the highest probability are output as the final prediction results. The training process aims to minimize the cross-entropy loss function. At the same time, an early stopping strategy is introduced to monitor the performance of the processing path evaluation data validation set. When the prediction accuracy on the processing path evaluation data validation set reaches a preset threshold, the processing path evaluation model is considered to have converged, and the training stops.

[0056] Compared with existing technologies, the technical effects and advantages of the process optimization method for complex curved surfaces proposed in this invention are as follows:

[0057] This solution is based on a 3D structural model of the workpiece surface. It divides and extracts features from the workpiece surface, subdividing complex workpiece surfaces into N sub-surfaces to ensure personalized optimization of the machining strategy for different regions. An intelligent optimization algorithm-based idle travel path tuning is employed to optimize the machining sequence of the workpiece sub-surfaces, automatically selecting the optimal machining sequence and reducing ineffective tool movement between workpiece sub-surfaces, making the machining process more intelligent and adaptive.

[0058] In addition, this solution takes into account the problem that the tool may encounter abrupt trajectory changes or sharp tool turning when switching from one sub-surface to another during the machining of the workpiece surface. The smoothness of the surface machining path when switching between sub-surfaces of the workpiece is further optimized. Based on the smoothness of the surface machining path of different candidate paths, the optimal tool movement path is adaptively selected.

[0059] In summary, this solution breaks through the limitations of traditional CNC machining. It can not only optimize the path within the sub-surface, but also intelligently optimize the switching across sub-surfaces, ensuring continuous trajectory and smooth transition during tool switching. Ultimately, it reduces non-cutting idle travel and repeated cutting, thereby improving the machining quality of the workpiece surface. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of a process optimization system for complex curved surfaces according to Embodiment 1 of the present invention;

[0061] Figure 2 This is a flowchart of a process optimization method for complex curved surfaces according to Embodiment 3 of the present invention;

[0062] Figure 3This is a schematic diagram of a process optimization system for complex curved surfaces according to Embodiment 2 of the present invention;

[0063] Figure 4 A flowchart illustrating a method for adaptively optimizing the idle travel path during surface machining of workpieces. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be described in detail, clearly, and completely below with reference to the accompanying drawings. It should be particularly noted that the specific embodiments described below are only for better illustrating and explaining the technical solutions of the present invention, and are intended to enable those skilled in the art to better understand and implement the present invention, and should not be construed as limiting the scope of protection of the present invention. Without departing from the spirit and substance of the present invention, those skilled in the art can modify, adjust, or make equivalent substitutions based on the content disclosed in the present invention, and these should all be considered within the scope of protection of the present invention.

[0065] Example 1

[0066] Please see Figure 1 As shown, this embodiment discloses a process optimization system for complex curved surfaces, including a first acquisition module, a first processing module, a second processing module, and a first tuning module. Each module is connected by wired and / or wireless means to achieve data transmission.

[0067] The first acquisition module is used to acquire the three-dimensional structural diagram of the curved surface of the workpiece.

[0068] It should be noted that the 3D structural diagram of the machined part surface is obtained from the 3D model database of the machined part manufacturing company. In the early stages of product development, the manufacturing company uses a computer-aided design (CAD) system to create a detailed model of the overall structure and surfaces of the machined part, thereby obtaining the 3D structural diagram of the machined part surface corresponding to different machined parts. The machined part includes valves.

[0069] The first processing module divides the surface of the workpiece into N sub-surfaces based on the 3D model of the workpiece surface.

[0070] The methods for obtaining the N sub-surfaces of the machined parts include:

[0071] S100: M sampling points are randomly distributed on the three-dimensional structural diagram of the machined part's curved surface;

[0072] S101: Let m be initially set to 1, and the value of m ranges from 1 to M; let n be initially set to 1, and n is a counting variable used to record the number of currently identified sub-surfaces of the workpiece.

[0073] S102: Obtain the coordinates of the workpiece surface parameters at the m-th sampling point, and obtain the corresponding workpiece surface parameter equation based on the workpiece surface parameter coordinates; calculate the curvature of the workpiece surface at the m-th sampling point based on the workpiece surface parameter equation;

[0074] S103: Determine the corresponding workpiece surface movement step size and workpiece sub-surface type based on the workpiece surface curvature at the m-th sampling point; perform a diffusion operation based on the workpiece surface movement step size and workpiece sub-surface type to obtain the workpiece surface diffusion region at the m-th sampling point;

[0075] S104: Using the edge point of the diffusion region of the workpiece surface at the m-th sampling point as the new diffusion starting point, repeat S103. When there are no more points of the same type as the workpiece sub-surface at the m-th sampling point during the diffusion process, the diffusion stops, and the workpiece sub-surface at the n-th sampling point is obtained; let n = n + 1, and execute S105;

[0076] S105: Let m = m + 1. If m is less than or equal to M, then execute S106; if m is greater than M, then end the current process, assign n to N, and obtain N sub-surfaces of the processed parts.

[0077] S106: If the m-th sampling point is already included in the constructed sub-surface of the machined part, then execute S105; if the m-th sampling point is not included in the constructed sub-surface of the machined part, then execute S102 to S104.

[0078] The methods for obtaining the diffusion region of the workpiece surface at the m-th sampling point include:

[0079] With the m-th sampling point as the center and the workpiece surface movement step size as the radius, a diffusion processing area for the m-th sampling point is formed; the curvature of the workpiece surface at the point on the edge of the diffusion processing area is calculated and denoted as the edge surface curvature; the corresponding workpiece sub-surface type is obtained based on the edge surface curvature.

[0080] Mark the points on the edge of the diffusion region to be processed that have the same sub-surface type as the sub-surface type of the workpiece at the m-th sampling point as the connection points to be connected at the m-th sampling point; connect all the connection points to be connected at the m-th sampling point to form the diffusion region of the workpiece surface at the m-th sampling point.

[0081] The method for obtaining the surface parameter equation of the m-th workpiece includes:

[0082] S m (u m v m )=(x m (u m v m ), y m (um v m ), z m (u m v m ));

[0083] Where S m (u m v m Let u be the parametric equation of the workpiece surface at the m-th sampling point. m and v m Let x be the coordinates of the workpiece surface parameters at the m-th sampling point. m (u m v m ) for u m and v m The parametric equation of the x-axis of the m-th sampling point is y m (u m v m ) for u m and v m The ordinate parameter equation for the m-th sampling point is z. m (u m v m ) for u m and v m The vertical axis parameter equation for the m-th sampling point is formed.

[0084] For example, if the surface of the machined part is an ellipsoid of revolution, then the parametric equation of the machined part surface is:

[0085] S m (u m v m )=(a×cosu m ×sinv m ,b×sinu m ×sinv m c×cosv m );

[0086] Where a is the semi-axis length of the ellipsoid along the X-axis, b is the semi-axis length of the ellipsoid along the Y-axis, and c is the semi-axis length of the ellipsoid along the Z-axis.

[0087] The methods for obtaining the curvature of the workpiece surface at the m-th sampling point include:

[0088] Parametric equations of the surface of the machined part for u m Find the first-order partial derivative to obtain the first first-order partial derivative; the parametric equation of the machined part surface with respect to v m Find the first-order partial derivative to obtain the second-order first-order partial derivative;

[0089] The unit normal vector of the m-th sampling point is calculated based on the first and second first-order partial derivatives.

[0090] Multiplying the first-order partial derivatives together yields the corresponding first-order formal parameters; multiplying the first-order partial derivatives together yields the corresponding second-order formal parameters; multiplying the second-order partial derivatives together yields the corresponding third-order formal parameters.

[0091] The first-order partial derivative with respect to u m Find the second-order partial derivative to obtain the first-order partial derivative; the first-order partial derivative with respect to v m Find the second partial derivative to obtain the second second partial derivative; the second first partial derivative with respect to v m Find the second-order partial derivative to obtain the third second-order partial derivative;

[0092] Multiplying the first second-order partial derivative with the unit normal vector yields the first second-order formal parameter of the m-th sampling point; multiplying the second second-order partial derivative with the unit normal vector yields the second second-order formal parameter of the m-th sampling point; multiplying the third second-order partial derivative with the unit normal vector yields the third second-order formal parameter of the m-th sampling point.

[0093] The curvature of the workpiece surface at the m-th sampling point is calculated based on the first-order formal parameters, the second-order formal parameters, the third-order formal parameters, the first-second-order formal parameters, the second-second-order formal parameters, and the third-second-order formal parameters.

[0094] The first-order partial derivative is:

[0095]

[0096] in, The first-order partial derivative, The parametric equation for the x-axis of the m-th sampling point is given by u. m The first-order partial derivative, The parametric equation of the ordinate axis for the m-th sampling point is given by u. m The first-order partial derivative, The vertical axis parametric equation for the m-th sampling point is given by u. m The first-order partial derivative.

[0097] The second first-order partial derivative is:

[0098]

[0099] in, The second-first order partial derivative, The parametric equation for the x-axis of the m-th sampling point is given by v. m The first-order partial derivative, The equation of the ordinate parameter for the m-th sampling point is given by v. m The first-order partial derivative, The vertical axis parametric equation for the m-th sampling point is given by v. m The first-order partial derivative.

[0100] Methods for calculating the unit normal vector of the m-th sampling point based on the first and second first-order partial derivatives include:

[0101]

[0102] in, Let be the unit normal vector of the m-th sampling point. Let m be the normal vector of the m-th sampling point. Let be the magnitude of the normal vector at the m-th sampling point.

[0103] The first first-order formal parameters are:

[0104]

[0105] Among them, DYYJ m Let be the first-order formal parameter of the m-th sampling point.

[0106] The second first-order formal parameters are:

[0107]

[0108] Among them, DEYJ m Let be the second first-order formal parameter of the m-th sampling point.

[0109] The third first-order formal parameter is:

[0110]

[0111] Among them, DSYJ m Let be the third first-order formal parameter of the m-th sampling point.

[0112] The first and second partial derivatives are:

[0113]

[0114] in, These are the first and second order partial derivatives. The first-order partial derivative of the x-axis parametric equation at the m-th sampling point with respect to u m The second-order partial derivative, The first-order partial derivative of the ordinate parametric equation at the m-th sampling point with respect to u m The second-order partial derivative, The first-order partial derivative of the vertical axis parametric equation at the m-th sampling point with respect to um The second-order partial derivatives of .

[0115] The second partial derivative is:

[0116]

[0117] in, It is the second-order partial derivative. The first-order partial derivative of the x-axis parametric equation at the m-th sampling point with respect to v m The second-order partial derivative, The first-order partial derivative of the ordinate parametric equation at the m-th sampling point with respect to v m The second-order partial derivative, The first-order partial derivative of the vertical axis parametric equation at the m-th sampling point with respect to v m The second-order partial derivatives of .

[0118] The third second-order partial derivative is:

[0119]

[0120] in, The third and second order partial derivatives, The second first-order partial derivative of the x-axis parametric equation at the m-th sampling point with respect to v m The second-order partial derivative, The second first-order partial derivative of the ordinate parametric equation at the m-th sampling point with respect to v m The second-order partial derivative, The second first-order partial derivative of the vertical axis parametric equation at the m-th sampling point with respect to v m The second-order partial derivatives of .

[0121] The first and second order formal parameters are:

[0122]

[0123] Among them, DYEJ m Let be the first and second order formal parameters of the m-th sampling point.

[0124] The second-order formal parameters are:

[0125]

[0126] Among them, DEEJ m is the second-order formal parameter of the m-th sampling point.

[0127] The third second-order formal parameter is:

[0128]

[0129] Among them, DSEJ m Let be the third second-order formal parameter of the m-th sampling point.

[0130] The methods for obtaining the curvature of the workpiece surface at the m-th sampling point include:

[0131]

[0132] Among them, FMQL m Let be the curvature of the workpiece surface at the m-th sampling point.

[0133] Methods for determining the corresponding workpiece surface movement step size and workpiece sub-surface type based on the workpiece surface curvature at the m-th sampling point include:

[0134] Preset a workpiece surface curvature threshold 1 and a workpiece surface curvature threshold 2, wherein the workpiece surface curvature threshold 1 is less than the workpiece surface curvature threshold 2; preset a workpiece surface first movement step, a workpiece surface second movement step, and a workpiece surface third movement step, wherein the workpiece surface first movement step is less than the workpiece surface second movement step, and the workpiece surface second movement step is less than the workpiece surface third movement step;

[0135] If the curvature of the workpiece surface at the m-th sampling point is greater than or equal to the workpiece surface curvature threshold two, then the corresponding workpiece sub-surface type is a high curvature region, and the corresponding workpiece surface movement step size is the first movement step size of the workpiece surface.

[0136] If the curvature of the workpiece surface at the m-th sampling point is greater than or equal to the workpiece surface curvature threshold one and less than the workpiece surface curvature threshold two, then the corresponding workpiece sub-surface type is a medium curvature region, and the corresponding workpiece surface movement step size is the workpiece surface second movement step size.

[0137] If the curvature of the workpiece surface at the m-th sampling point is less than the workpiece surface curvature threshold, then the corresponding workpiece sub-surface type is a low curvature region, and the corresponding workpiece surface movement step size is the third movement step size of the workpiece surface.

[0138] It should be noted that during the machining of curved surfaces, since these surfaces are typically composed of multiple regions with varying curvatures, directly optimizing the entire surface uniformly can lead to insufficiently refined path optimization in certain areas, or even over-optimization, negatively impacting machining quality and efficiency. Therefore, by rationally dividing the surface into regions, we can ensure that each sub-surface employs the most suitable machining strategy for its characteristics, thereby improving the accuracy and adaptability of machining path optimization.

[0139] If the workpiece surface is not subdivided, the tool will frequently jump between high-curvature and low-curvature regions during machining, leading to increased tool lift-offs and excessive idle travel, thus reducing overall machining efficiency. However, by dividing the workpiece surface into sub-surfaces, the tool can move continuously within similar areas, reducing toolpath switching and unnecessary movement during machining, thereby improving machining efficiency.

[0140] Furthermore, the division of sub-surfaces in the machining process also considers the smooth transition path between adjacent sub-surfaces, making the tool transitions between different areas more natural, thereby avoiding abrupt path changes and improving the smoothness and stability of the overall machining path. This approach not only optimizes machining accuracy but also effectively reduces errors during the machining process, improves the machining quality of the workpiece surfaces, and ensures the uniformity and consistency of the final machining effect.

[0141] The second processing module is used to extract features from N sub-surfaces of the machined parts to obtain the feature data of the sub-surfaces of the machined parts corresponding to the N sub-surfaces of the machined parts.

[0142] Methods for obtaining the feature data of N machined part sub-surfaces include:

[0143] S200: Let the initial value of n be 1, and the range of n is from 1 to N;

[0144] S201: Obtain the coordinate range, first-order formal parameters, second-order formal parameters, and third-order formal parameters of the nth workpiece sub-surface; calculate the corresponding workpiece sub-surface area based on the coordinate range, first-order formal parameters, second-order formal parameters, and third-order formal parameters; divide the boundary of the nth workpiece sub-surface into J boundary points, and calculate the boundary length of the nth workpiece sub-surface based on the coordinates of the J boundary points;

[0145] The curvature of the workpiece surface at J boundary points is obtained using the same method as for obtaining the curvature of the workpiece surface; the surface boundary complexity of the nth workpiece sub-surface is calculated based on the curvature of the workpiece surface at J boundary points.

[0146] S202: Construct the corresponding machining part feature data from the area, boundary length, and boundary complexity of the nth machining part sub-surface;

[0147] S203: Let n = n + 1. If n is less than or equal to N, continue executing S201 to S202. If n is greater than N, end the current process.

[0148] The method for calculating the area of ​​the sub-surface of the machined part includes:

[0149]

[0150] Among them, QMMJ n Let u be the area of ​​the subsurface of the part processed at the nth boundary point. n，min and v n，min u represents the lower bound of the workpiece surface parameter coordinates for the nth workpiece sub-surface. n，max and v n，max This represents the upper limit of the coordinates of the surface parameters of the nth machined part sub-surface; DYYJ n DEYJ is the first-order formal parameter of the nth boundary point. n DSYJ is the second first-order formal parameter of the nth boundary point. n is the third first-order formal parameter of the nth boundary point.

[0151] The method for calculating the boundary length of the sub-surface of the machined part includes:

[0152]

[0153] Among them, BJCD n Let x be the boundary length of the nth processed part's subsurface. j+1 Let x be the x-axis coordinate of the (j+1)th boundary point. j Let y be the x-axis coordinate of the j-th boundary point. j+1 Let y be the ordinate of the (j+1)th boundary point. j Let z be the ordinate of the j-th boundary point. j+1 Let y be the vertical coordinate of the (j+1)th boundary point. j Let be the vertical coordinate of the j-th boundary point. It calculates the distance between the last point and the first point on the boundary of the sub-surface of the machined part.

[0154] The method for calculating the surface boundary complexity includes:

[0155]

[0156] Among them, FZD n Let λ be the surface boundary complexity of the nth processed part sub-surface. j For the coefficient of the j-th boundary point, FMQL j Let J be the curvature of the workpiece surface at the j-th boundary point, and J be the number of boundary points.

[0157] The first optimization module adaptively optimizes the idle travel path of machining the N sub-surfaces of the workpiece based on the feature data of the N workpiece sub-surfaces.

[0158] It should be noted that optimizing the idle travel path of the machining of curved surfaces, that is, reducing the non-cutting motion of the tool between various sub-surfaces of the machining workpiece, is the key to improving machining efficiency.

[0159] like Figure 4 As shown, the method for adaptively optimizing the idle travel path of machining the surface of N machined parts based on the feature data of N machined part sub-surfaces includes:

[0160] Input the feature data of N sub-surfaces of the workpiece into the machining path evaluation model to obtain H candidate machining paths for the workpiece; each candidate machining path contains the machining sequence of the N sub-surfaces of the workpiece; construct a machining path population from the H candidate machining paths;

[0161] For example, there are 5 sub-surfaces for machining parts, and the corresponding numbers of the 5 sub-surfaces are 1, 2, 3, 4 and 5. There are 3 candidate paths for door machining, which are 1→2→4→3→5, 2→1→5→4→3 and 5→1→4→3→2.

[0162] The fitness of the H candidate machining paths in the machining path population is evaluated to obtain the corresponding machining path fitness values.

[0163] Sort the fitness values ​​of the H processing paths in descending order, and select the top Q processing paths fitness values ​​to construct an elite population of processing paths;

[0164] The processing path elite population is mutated according to a preset method to obtain R mutated processing candidate paths and added to the processing path elite population; the processing path elite population is then overlaid on the processing path population, and H = Q + R is set.

[0165] Repeat the above operation until the number of iterations reaches the preset maximum number of iterations, then stop execution; select the candidate path for machining the workpiece with the largest fitness value from the machining path population as the optimal solution for the empty travel path of the workpiece surface machining.

[0166] It should be noted that the preset method refers to the specific method for performing mutation operations on the elite population of processing paths. The subsequent "Method for Obtaining R Candidate Processing Paths for Mutated Processed Parts" details the specific implementation steps of the preset method.

[0167] For example, consider a ball valve. A ball valve is a machined component used to control fluid flow. Its core component includes a rotatable ball with a channel. Rotating the ball 90° controls the flow or blockage of fluid. The machined surfaces of a ball valve include multiple complex free-form surfaces such as the outer surface of the ball, the valve seat sealing surface, and the flow channel connection area. Due to the structural characteristics of the ball valve ball, during CNC machining, the tool needs to move between multiple sub-surfaces with different curvatures. Especially when machining the contact area between the ball and the valve seat, the tool switching path needs to be optimized to reduce idle travel and improve machining quality and efficiency.

[0168] The training method for the processing path evaluation model includes:

[0169] A pre-constructed processing path evaluation dataset is provided, comprising Y sets of processing path evaluation data and H candidate processing paths for the corresponding parts, where Y is a positive integer greater than 0. The processing path evaluation data includes N sub-surface feature data of the parts. The processing path evaluation dataset is divided into a processing path evaluation data training set and a processing path evaluation data validation set. The processing path evaluation data training set is used for parameter learning of the processing path evaluation model, and the processing path evaluation data validation set is used for real-time evaluation of the generalization ability of the processing path evaluation model.

[0170] During the training of the processing path evaluation model, a deep neural network structure based on multilayer perceptrons is adopted. The processing path evaluation data is converted into feature vectors as input, and nonlinear features in the data are extracted through multiple hidden layers. Finally, the softmax activation function is used in the output layer to generate the probability distribution of H processing candidate paths for processing parts. The H processing candidate paths corresponding to the highest probability are output as the final prediction results. The training process aims to minimize the cross-entropy loss function. At the same time, an early stopping strategy is introduced to monitor the performance of the processing path evaluation data validation set. When the prediction accuracy on the processing path evaluation data validation set reaches a preset threshold, the processing path evaluation model is considered to have converged, and the training stops.

[0171] The methods for obtaining R candidate processing paths for mutated workpieces include:

[0172] Preset path fitness threshold 1 and path fitness threshold 2, where path fitness threshold 1 is less than path fitness threshold 2; divide the elite population of processing paths according to path fitness threshold 1 and path fitness threshold 2 to obtain a set of high-fitness paths, a set of medium-fitness paths, and a set of low-fitness paths.

[0173] Randomly cross-reference each other in the high-fitness path set to obtain NUM1 candidate processing paths for the workpiece, and add them to the high-fitness path set;

[0174] Randomly cross the processing candidate paths of the workpiece in the medium fitness path set with the processing candidate paths of the workpiece in the high fitness path set to obtain NUM2 processing candidate paths, which are then added to the medium fitness path set; randomly cross the processing candidate paths in the medium fitness path set to obtain NUM3 processing candidate paths, which are then added to the medium fitness path set.

[0175] Randomly cross the processing candidate paths of the workpiece in the low fitness path set with the processing candidate paths of the workpiece in the high fitness path set to obtain NUM4 processing candidate paths, which are then added to the low fitness path set; randomly cross the processing candidate paths of the workpiece in the low fitness path set to obtain NUM5 processing candidate paths, which are then added to the low fitness path set.

[0176] When NUM1+NUM2+NUM3+NUM4+NUM5=R, the above random crossover operation is stopped, and R candidate paths for processing mutated parts are obtained.

[0177] The methods for obtaining the high-fitness path set, the medium-fitness path set, and the low-fitness path set include:

[0178] Candidate processing paths for workpieces from the elite processing path population whose processing path fitness values ​​are greater than or equal to the path fitness threshold two are constructed into a high-fitness path set; candidate processing paths for workpieces from the elite processing path population whose processing path fitness values ​​are greater than or equal to the path fitness threshold one and less than the path fitness threshold two are constructed into a medium-fitness path set; and candidate processing paths for workpieces from the elite processing path population whose processing path fitness values ​​are less than the path fitness threshold one are constructed into a low-fitness path set.

[0179] The method for obtaining the fitness value of the processing path includes:

[0180] Input the H candidate machining paths and the corresponding subsurface feature data of the machining parts into the fitness evaluation model to obtain the fitness value of the corresponding machining path.

[0181] The training method for the fitness evaluation model includes:

[0182] A fitness evaluation dataset is pre-collected, which includes P sets of fitness evaluation data and the corresponding processing path fitness values ​​for the P sets of fitness evaluation data, where P is a positive integer greater than 0. The fitness evaluation data includes candidate processing paths for the workpiece and subsurface feature data of the workpiece. The fitness evaluation dataset is divided into a training set and a validation set, where the training set is used to train the fitness evaluation model and the validation set is used to evaluate the generalization performance of the fitness evaluation model.

[0183] During the training of the fitness evaluation model, minimizing the cross-entropy loss function is used as the optimization objective. An early stopping strategy is used to monitor the performance on the validation set. The model performance is optimized by continuously adjusting the network parameters. When the prediction accuracy on the validation set reaches the expected accuracy, the fitness evaluation model is considered to have converged and training is stopped. The fitness evaluation model is trained using a deep neural network based on a multilayer perceptron.

[0184] The fitness evaluation data is converted into feature vectors. The input layer of the fitness evaluation model receives the feature vectors and extracts the nonlinear relationships in the data through several hidden layers. Finally, the output layer of the fitness evaluation model calculates the probability distribution of the fitness values ​​of the processing paths through the softmax activation function and outputs the fitness value of the processing path with the highest probability as the final prediction result.

[0185] Example 2

[0186] Please see Figure 3 As shown, this embodiment provides a process optimization system for complex curved surfaces, and also includes:

[0187] The second optimization module adaptively optimizes the smoothness of the surface machining path for N machined parts sub-surfaces based on the adaptively optimized empty travel path.

[0188] Methods for adaptively optimizing the smoothness of surface machining paths for N machined parts sub-surfaces based on adaptively optimized idle travel paths include:

[0189] S300: Obtain the machining sequence of the sub-surface of the workpiece corresponding to the adaptively optimized empty travel path; set the initial value of sx to 1, and the value range of sx is from 1 to N;

[0190] S301: If sx is less than or equal to N-1, then obtain the boundary region between the sxth and sx+1th sub-surfaces of the workpiece from the machining sequence; if sx is equal to N, then obtain the boundary region between the sxth and 1st sub-surfaces of the workpiece from the machining sequence.

[0191] Set the initial and final positions of the tool for the boundary region; let T be the number of moves from the initial to the final position; initialize DDYZ full tool movement paths, calculate the smoothness of the surface machining path corresponding to each full tool movement path, and record the single-move smoothness corresponding to each tool movement; add the tool single movement paths with single-move smoothness greater than the single-move smoothness threshold to the single movement path set; iterate and optimize the single movement paths in the single movement path set and those not in the single movement path set according to a preset number of iterations to obtain a preset number of tool movement paths, and calculate the corresponding surface machining path smoothness;

[0192] The tool movement path with the greatest smoothness in the surface machining path is selected as the corresponding adaptively optimized tool movement path.

[0193] S302: Let sx = sx + 1. If sx is less than or equal to SX, then continue to execute S301. If sx is greater than SX, then end the current process.

[0194] The method for obtaining the smoothness of the surface processing path includes:

[0195]

[0196] Among them, PHX ((sx，sx+1)，ddyz) Let ddyz be the smoothness of the machining path of the ddyz-th surface corresponding to the boundary region between the sx-th and sx+1-th machined sub-surfaces, where ddyz ranges from 1 to DDYZ, and PHX is the value of the surface. ((sx，sx+1)，ddyz) The larger the value, the smoother the surface machining path, and the more suitable the corresponding tool movement path; FMQL (ddyz，t+1) Let FMQL be the curvature of the workpiece surface corresponding to the (t+1)th tool movement position of the ddyz-th tool movement path. (ddyz，t) YDJL represents the curvature of the workpiece surface corresponding to the t-th tool movement position of the ddyz-th tool movement path. (ddyz，t) Let DJJD be the tool movement distance for the t-th time in the ddyz-th tool movement path. (ddyz，t) Let be the tool movement angle of the t-th time in the ddyz-th tool movement path.

[0197] It should be noted that when the tool for machining the surface of a workpiece switches from one sub-surface to another, it may encounter issues such as abrupt changes in the trajectory or sharp tool turns. Therefore, it is necessary to adaptively optimize the smoothness of the surface machining path to ensure smooth transitions, avoid sudden acceleration or deceleration, reduce idle travel, and improve the machining quality of the workpiece surface.

[0198] Example 3

[0199] Please see Figure 2 As shown, this embodiment provides a process optimization method for complex curved surfaces, implemented based on a process optimization system for complex curved surfaces, including:

[0200] Obtain the 3D structural diagram of the machined part's curved surface;

[0201] Based on the 3D model of the workpiece surface, the workpiece surface is divided into N sub-surfaces;

[0202] Feature extraction is performed on N sub-surfaces of the machined parts to obtain the feature data of the machined parts sub-surfaces corresponding to the N sub-surfaces of the machined parts;

[0203] The idle travel path for machining the surface of N machining parts is adaptively optimized based on the feature data of N machining parts sub-surfaces.

Claims

1. A process optimization method for complex curved surfaces, characterized in that, include: Obtain the 3D structural diagram of the machined part's curved surface; Based on the 3D model of the workpiece surface, the workpiece surface is divided into N sub-surfaces; Feature extraction is performed on N sub-surfaces of the machined parts to obtain the feature data of the machined parts sub-surfaces corresponding to the N sub-surfaces of the machined parts; Adaptive optimization of the idle travel path for machining the surface of N machining parts based on the feature data of N machining parts sub-surfaces; The method for adaptively optimizing the idle travel path of machining the surface of N machined parts based on the feature data of N machined parts sub-surfaces includes: Input the feature data of N sub-surfaces of the workpiece into the machining path evaluation model to obtain H candidate machining paths for the workpiece; each candidate machining path contains the machining sequence of the N sub-surfaces of the workpiece; construct a machining path population from the H candidate machining paths; The fitness of the H candidate machining paths in the machining path population is evaluated to obtain the corresponding machining path fitness values. Sort the fitness values ​​of the H processing paths in descending order, and select the top Q processing paths fitness values ​​to construct an elite population of processing paths; The processing path elite population is mutated according to a preset method to obtain R mutated processing candidate paths for processing parts, which are then added to the processing path elite population; the processing path elite population is then overlaid onto the processing path population, and H = Q + R is set. Repeat the above operation until the number of iterations reaches the preset maximum number of iterations, then stop execution; select the candidate path for machining the workpiece with the largest fitness value from the machining path population as the optimal solution for the empty travel path of the workpiece surface machining.

2. The process optimization method for complex curved surfaces according to claim 1, characterized in that, The methods for obtaining R candidate processing paths for mutated workpieces include: Preset path fitness threshold 1 and path fitness threshold 2, where path fitness threshold 1 is less than path fitness threshold 2; divide the elite population of processing paths according to path fitness threshold 1 and path fitness threshold 2 to obtain a set of high-fitness paths, a set of medium-fitness paths, and a set of low-fitness paths. Randomly cross over the high-fitness path set to obtain Each workpiece has a candidate processing path, which is then added to the set of highly fit paths. Randomly cross-reference the candidate machining paths for the workpiece in the medium-fitness path set with those in the high-fitness path set to obtain... Each candidate processing path for a workpiece is selected and added to the medium-fitness path set; random cross-paths are then performed within the medium-fitness path set to obtain... Each workpiece is selected as a candidate processing path and added to the set of medium-fitness paths. Randomly cross-reference the candidate machining paths for the workpiece in the low-fitness path set with those in the high-fitness path set to obtain... Each candidate processing path for a workpiece is selected and added to the low-fitness path set; random cross-paths are then performed within the low-fitness path set to obtain... Each workpiece is selected as a candidate processing path and added to the low-fitness path set. when When the above random crossover operation is stopped, R candidate processing paths for the mutated workpiece are obtained.

3. The process optimization method for complex curved surfaces according to claim 2, characterized in that, The methods for obtaining the high-fitness path set, the medium-fitness path set, and the low-fitness path set include: Candidate processing paths for workpieces from the elite processing path population whose processing path fitness values ​​are greater than or equal to the path fitness threshold two are constructed into a high-fitness path set; candidate processing paths for workpieces from the elite processing path population whose processing path fitness values ​​are greater than or equal to the path fitness threshold one and less than the path fitness threshold two are constructed into a medium-fitness path set; and candidate processing paths for workpieces from the elite processing path population whose processing path fitness values ​​are less than the path fitness threshold one are constructed into a low-fitness path set.

4. The process optimization method for complex curved surfaces according to claim 3, characterized in that, The methods for obtaining the N sub-surfaces of the machined parts include: S100: M sampling points are randomly distributed on the three-dimensional structural diagram of the machined part's curved surface; S101: Let m be initialized to 1, and the range of m is from 1 to M; let n be initialized to 1, and n is a counting variable; S102: Obtain the coordinates of the workpiece surface parameters at the m-th sampling point, and obtain the corresponding workpiece surface parameter equation based on the workpiece surface parameter coordinates; calculate the curvature of the workpiece surface at the m-th sampling point based on the workpiece surface parameter equation; S103: Determine the corresponding workpiece surface movement step size and workpiece sub-surface type based on the workpiece surface curvature at the m-th sampling point; perform a diffusion operation based on the workpiece surface movement step size and workpiece sub-surface type to obtain the workpiece surface diffusion region at the m-th sampling point; S104: Using the edge point of the diffusion region of the workpiece surface at the m-th sampling point as the new diffusion starting point, repeat S103. When there are no more points of the same type as the workpiece sub-surface at the m-th sampling point during the diffusion process, the diffusion stops, and the workpiece sub-surface at the n-th sampling point is obtained; let n=n+1, and execute S105. S105: Let m = m + 1. If m is less than or equal to M, then execute S106; if m is greater than M, then end the current process, assign n to N, and obtain N sub-surfaces of the processed parts. S106: If the m-th sampling point is already included in the constructed sub-surface of the machined part, then execute S105; if the m-th sampling point is not included in the constructed sub-surface of the machined part, then execute S102 to S104.

5. The process optimization method for complex curved surfaces according to claim 4, characterized in that, The methods for obtaining the diffusion region of the workpiece surface at the m-th sampling point include: With the m-th sampling point as the center and the workpiece surface movement step size as the radius, a diffusion processing area for the m-th sampling point is formed; the curvature of the workpiece surface at the point on the edge of the diffusion processing area is calculated and denoted as the edge surface curvature; the corresponding workpiece sub-surface type is obtained based on the edge surface curvature. Mark the points on the edge of the diffusion region to be processed that have the same sub-surface type as the sub-surface type of the workpiece at the m-th sampling point as the connection points to be connected at the m-th sampling point; connect all the connection points to be connected at the m-th sampling point to form the diffusion region of the workpiece surface at the m-th sampling point.

6. The process optimization method for complex curved surfaces according to claim 5, characterized in that, The methods for obtaining the curvature of the workpiece surface at the m-th sampling point include: Parametric equations of machined part surfaces Find the first-order partial derivative to obtain the first first-order partial derivative; the parametric equations of the machined part surface are... Find the first-order partial derivative to obtain the second-order first-order partial derivative; and Given the surface parameter coordinates of the workpiece at the m-th sampling point, the unit normal vector of the m-th sampling point is calculated based on the first and second first-order partial derivatives. Multiplying the first-order partial derivatives together yields the corresponding first-order formal parameters; multiplying the first-order partial derivatives together yields the corresponding second-order formal parameters; multiplying the second-order partial derivatives together yields the corresponding third-order formal parameters. The first-order partial derivative pairs Finding the second-order partial derivative yields the first and second-order partial derivatives; the first-order partial derivative is related to... Find the second partial derivative to obtain the second second partial derivative; the second first partial derivative is related to... Find the second-order partial derivative to obtain the third second-order partial derivative; Multiplying the first second-order partial derivative with the unit normal vector yields the first second-order formal parameter of the m-th sampling point; multiplying the second second-order partial derivative with the unit normal vector yields the second second-order formal parameter of the m-th sampling point; multiplying the third second-order partial derivative with the unit normal vector yields the third second-order formal parameter of the m-th sampling point. The curvature of the workpiece surface at the m-th sampling point is calculated based on the first-order formal parameters, the second-order formal parameters, the third-order formal parameters, the first-second-order formal parameters, the second-second-order formal parameters, and the third-second-order formal parameters.

7. The process optimization method for complex curved surfaces according to claim 6, characterized in that, Methods for determining the corresponding workpiece surface movement step size and workpiece sub-surface type based on the workpiece surface curvature at the m-th sampling point include: Preset a workpiece surface curvature threshold 1 and a workpiece surface curvature threshold 2, wherein the workpiece surface curvature threshold 1 is less than the workpiece surface curvature threshold 2; preset a workpiece surface first movement step, a workpiece surface second movement step, and a workpiece surface third movement step, wherein the workpiece surface first movement step is less than the workpiece surface second movement step, and the workpiece surface second movement step is less than the workpiece surface third movement step; If the curvature of the workpiece surface at the m-th sampling point is greater than or equal to the workpiece surface curvature threshold two, then the workpiece sub-surface type is a high curvature region, and the workpiece surface movement step size is the first movement step size of the workpiece surface. If the curvature of the workpiece surface at the m-th sampling point is greater than or equal to the workpiece surface curvature threshold one and less than the workpiece surface curvature threshold two, then the workpiece sub-surface type is a medium curvature region, and the workpiece surface movement step size is the workpiece surface second movement step size. If the curvature of the workpiece surface at the m-th sampling point is less than the workpiece surface curvature threshold, then the workpiece sub-surface type is a low curvature region, and the workpiece surface movement step size is the third movement step size of the workpiece surface.

8. The process optimization method for complex curved surfaces according to claim 7, characterized in that, Methods for obtaining the feature data of N machined part sub-surfaces include: S200: Let the initial value of n be 1, and the range of n is from 1 to N; S201: Obtain the coordinate range, first-order formal parameters, second-order formal parameters, and third-order formal parameters of the nth workpiece sub-surface; calculate the corresponding workpiece sub-surface area based on the coordinate range, first-order formal parameters, second-order formal parameters, and third-order formal parameters; divide the boundary of the nth workpiece sub-surface into J boundary points, and calculate the boundary length of the nth workpiece sub-surface based on the coordinates of the J boundary points; The curvature of the workpiece surface at J boundary points is obtained using the same method as for obtaining the curvature of the workpiece surface; the surface boundary complexity of the nth workpiece sub-surface is calculated based on the curvature of the workpiece surface at J boundary points. S202: Construct the corresponding machining part feature data from the area, boundary length, and boundary complexity of the nth machining part sub-surface; S203: Let n = n + 1. If n is less than or equal to N, continue executing S201 to S202. If n is greater than N, end the current process.

9. The process optimization method for complex curved surfaces according to claim 8, characterized in that, The training method for the processing path evaluation model includes: A pre-constructed processing path evaluation dataset is provided, comprising Y sets of processing path evaluation data and H candidate processing paths for the corresponding parts, where Y is a positive integer greater than 0. The processing path evaluation data includes N sub-surface feature data of the parts. The processing path evaluation dataset is divided into a processing path evaluation data training set and a processing path evaluation data validation set. The processing path evaluation data training set is used for parameter learning of the processing path evaluation model, and the processing path evaluation data validation set is used for real-time evaluation of the generalization ability of the processing path evaluation model. During the training of the processing path evaluation model, a deep neural network structure based on multilayer perceptrons is adopted. The processing path evaluation data is converted into feature vectors as input, and nonlinear features in the data are extracted through multiple hidden layers. Finally, the softmax activation function is used in the output layer to generate the probability distribution of H processing candidate paths for processing parts. The H processing candidate paths corresponding to the highest probability are output as the final prediction results. The training process aims to minimize the cross-entropy loss function. At the same time, an early stopping strategy is introduced to monitor the performance of the processing path evaluation data validation set. When the prediction accuracy on the processing path evaluation data validation set reaches a preset threshold, the processing path evaluation model is considered to have converged, and the training stops.

Images