Machine learning-based numerical control machine tool processing path optimization method and system

Abstract

The invention provides a numerical control machine tool machining path optimization method and system based on machine learning, and relates to the technical field of machine tool machining, and the method comprises the steps: carrying out the expansion matching of historical machining data based on the workpiece geometric features and machining demands of a target machining workpiece, and carrying out the machine learning according to a historical machining log set, constructing a processing quality prediction model; candidate cutter location identification is carried out according to geometric features of the workpiece, and candidate cutter location point sets are analyzed based on the machining quality prediction model; screening the candidate cutter location point set based on the candidate cutter location processing quality data set; and according to the candidate cutter location machining quality data set and the key cutter location point set, independent machining parameter optimization is executed, and linkage numerical control machine tool machining constraint conditions are called for linkage machining path optimization. Through the method, the technical problem of poor accuracy of numerical control machine tool machining path optimization in the prior art can be solved, and the technical effect of improving the accuracy of numerical control machine tool machining path optimization is achieved.

Description

Technical Field

[0001] This application relates to the field of machine tool processing technology, and in particular to a method and system for optimizing CNC machine tool processing paths based on machine learning. Background Technology

[0002] With the widespread application of CNC machining in complex curved surfaces, high-precision parts, and multi-axis linkage machining scenarios, machining path planning has become a key factor affecting machining efficiency and quality.

[0003] Currently, existing machining path planning methods generally employ similarity retrieval based on simple geometric features when matching reference machining data. For example, they filter historical workpieces based on rules such as similar dimensions, curvature ranges, or similar machining area distributions, and use these filtering results as input for quality prediction or path planning. However, this matching method often fails to adequately characterize the differences in workpiece geometric features within complex and varying regions. While different workpieces may be similar in overall structure, significant differences may exist in key locations such as local features, edge regions, and transition curvatures. These differences are often the core factors affecting machining quality and path stability. Because existing methods cannot meticulously distinguish these feature differences, key samples in historical machining data are easily overlooked during the matching process, leading to biases in the constructed machining knowledge system and making it difficult for quality prediction models to accurately judge new workpiece geometric features.

[0004] In summary, existing technologies suffer from the problem that matching historical processing data relies solely on simple geometric similarity, which fails to accurately identify local feature differences affecting processing quality. This results in an insufficient data foundation for path planning and quality prediction, further impacting the accuracy of processing path generation, the reliability of processing quality prediction, and the overall stability of complex workpiece processing. Summary of the Invention

[0005] The purpose of this application is to provide a CNC machine tool machining path optimization method and system based on machine learning, in order to solve the technical problem in the prior art that the matching of historical machining data relies only on simple geometric similarity, which cannot accurately identify the local feature differences that affect the machining quality, resulting in insufficient data foundation for path planning and quality prediction, and further affecting the accuracy of machining path generation, the reliability of machining quality prediction and the overall stability of the machining process of complex workpieces.

[0006] In view of the above problems, this application provides a method and system for optimizing CNC machine tool machining paths based on machine learning.

[0007] Firstly, this application provides a machine learning-based CNC machine tool machining path optimization method, implemented through a machine learning-based CNC machine tool machining path optimization system. The method includes: acquiring the geometric model of the target workpiece and extracting its geometric features to obtain workpiece geometric features; expanding and matching historical machining data based on the workpiece geometric features and machining requirements to obtain a historical machining log set including historical machining data, tool path data, process parameters, and machining results; performing machine learning based on the historical machining log set to construct a machining quality prediction model; identifying candidate tool positions based on the workpiece geometric features to obtain a candidate tool position set; analyzing the candidate tool position set based on the machining quality prediction model to obtain a candidate tool position machining quality data set including surface quality prediction values, geometric error prediction values, machining coverage, and estimated machining time; filtering the candidate tool position set based on the candidate tool position machining quality data set to determine a key tool position set; performing independent machining parameter optimization based on the candidate tool position machining quality data set and the key tool position set, and invoking the linkage CNC machine tool machining constraints to perform linkage machining path optimization to obtain the optimal machining path.

[0008] Preferably, the CNC machine tool machining path optimization method based on machine learning further includes: performing initial matching of historical machining data according to the workpiece geometric features and machining requirements to obtain an initial historical machining data set; traversing the initial historical machining data set to perform edge directional expansion to obtain a historical machining log set; dividing the historical machining log set into a base class dataset and a new class dataset; and performing machine learning based on the base class dataset and the new class dataset to construct the machining quality prediction model.

[0009] Preferably, the CNC machine tool machining path optimization method based on machine learning further includes: using the base class dataset to perform machine learning training on the framework built based on the feedforward neural network to obtain an initial machining quality prediction model; and using the new class dataset to perform generalization training on the initial machining quality prediction model to obtain a machining quality prediction model.

[0010] Preferably, the CNC machine tool machining path optimization method based on machine learning further includes: extracting similar representative data from the initial historical machining data set to obtain representative centers; filtering edge data from the initial historical machining data set based on the representative centers to determine edge historical machining data sets; filtering multi-point cluster centers from the edge historical machining data sets to obtain multiple clustered edge historical machining data centers; and expanding the initial historical machining data set with edge directionality based on the multiple clustered edge historical machining data centers to obtain a historical machining log set.

[0011] Preferably, the CNC machine tool machining path optimization method based on machine learning further includes: the representative center is the historical machining data corresponding to the maximum sum of similarity with other historical machining data in the initial historical machining data set.

[0012] Preferably, the CNC machine tool machining path optimization method based on machine learning further includes: constructing a representative center neighborhood of the representative center according to a preset edge iteration filtering bandwidth; performing edge diffusion on the neighborhood edges of the representative center neighborhood according to the preset edge iteration filtering bandwidth to obtain a diffused representative center neighborhood; comparing the data volume difference between the diffused representative center neighborhood and the representative center neighborhood, and if the data volume difference is greater than or equal to a preset threshold, stopping the diffusion and adding the data other than the representative center neighborhood from the initial historical machining data set to the edge historical machining data set; if the data volume difference is less than the preset threshold, continuing to perform edge diffusion on the diffused representative center neighborhood according to the preset edge iteration filtering bandwidth until the data volume difference between two adjacent neighborhoods is less than the preset threshold, thereby obtaining the edge historical machining data set.

[0013] Preferably, the CNC machine tool machining path optimization method based on machine learning further includes: randomly selecting multiple edge historical machining data from the edge historical machining data set, identifying the distance between the multiple edge historical machining data, and obtaining multiple clustered edge historical machining data centers when the distance between any two edge historical machining data is greater than or equal to a preset distance constraint.

[0014] Preferably, the machine learning-based CNC machine tool machining path optimization method further includes: performing independent machining parameter optimization on the key tool position set based on the candidate tool position machining quality data set to obtain a key tool position machining parameter set; treating each key tool position in the key tool position set as a node in a graph structure, and constructing a linkage optimization graph structure based on the spatial distance between key tool positions, the tool axis direction angle, and the adjacency relationship of machining areas; and calling the linkage CNC machine tool machining constraints to perform linkage machining path optimization on the key tool position machining parameter set according to the linkage optimization graph structure to obtain the optimal machining path.

[0015] Preferably, the CNC machine tool machining path optimization method based on machine learning further includes: the machining constraints of the linkage CNC machine tool are to minimize the machining time and minimize the geometric error.

[0016] Secondly, this application also provides a machine learning-based CNC machine tool machining path optimization system for executing the machine learning-based CNC machine tool machining path optimization method as described in the first aspect, including: a workpiece geometric feature acquisition module, used to acquire the geometric model of the target workpiece and extract geometric features to obtain workpiece geometric features; a machining quality prediction model construction module, used to expand and match historical machining data based on the workpiece geometric features and machining requirements to obtain a historical machining log set including historical machining data, tool path data, process parameters and machining results, and to perform machine learning based on the historical machining log set to construct a machining quality prediction model; and a candidate tool position machining quality data set acquisition module, used to... Candidate tool positions are identified based on the workpiece's geometric features to obtain a set of candidate tool position points. The set of candidate tool position points is then analyzed based on the machining quality prediction model to obtain a set of candidate tool position machining quality data, including predicted surface quality values, predicted geometric error values, machining coverage, and estimated machining time. A key tool position point set determination module is used to filter the candidate tool position point set based on the candidate tool position machining quality data set to determine a set of key tool position points. An optimal machining path acquisition module is used to perform independent machining parameter optimization based on the candidate tool position machining quality data set and the set of key tool position points, and to perform linked machining path optimization by invoking the machining constraints of the linked CNC machine tool to obtain the optimal machining path.

[0017] The technical solution provided in this application has at least the following technical effects or advantages: by achieving accurate matching and in-depth utilization of historical processing data based on multi-dimensional local features, a more reliable processing knowledge base is constructed and the technical goal of supporting adaptive path planning and predictive modeling is achieved, thereby realizing more accurate path planning, more reliable processing quality prediction, and improved processing stability in the processing of complex workpieces.

[0018] The above description is merely an overview of the technical solution of this application. To enable a clearer understanding of the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the CNC machine tool machining path optimization method based on machine learning proposed in this application.

[0021] Figure 2 This is a schematic diagram of the CNC machine tool machining path optimization system based on machine learning, as described in this application.

[0022] The attached diagrams are labeled as follows: Module 1 for obtaining workpiece geometric features, Module 2 for constructing a machining quality prediction model, Module 3 for obtaining a set of machining quality data for candidate tool positions, Module 4 for determining a set of key tool positions, and Module 5 for obtaining the optimal machining path. Detailed Implementation

[0023] This application provides a machine learning-based method and system for optimizing CNC machine tool machining paths. It addresses the technical problem in existing technologies where matching historical machining data relies solely on simple geometric similarity, failing to accurately identify local feature differences affecting machining quality. This results in insufficient data for path planning and quality prediction, further impacting the accuracy of machining path generation, the reliability of machining quality prediction, and the overall stability of complex workpiece machining. The method achieves accurate matching and in-depth utilization of historical machining data based on multi-dimensional local features, thereby constructing a more reliable machining knowledge base and supporting adaptive path planning and predictive modeling. Ultimately, this leads to more accurate path planning, more reliable machining quality prediction, and improved machining process stability in complex workpiece machining.

[0024] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0025] Example 1, please refer to the appendix. Figure 1 This application provides a machine learning-based CNC machine tool machining path optimization method, applied to a machine learning-based CNC machine tool machining path optimization system, specifically including: S1: Obtain the geometric model of the target workpiece and extract its geometric features to obtain the workpiece's geometric features.

[0026] Specifically, acquiring the geometric model of the target workpiece involves reading complete data describing the workpiece's shape from a 3D design file. For example, a 3D model format can include the workpiece's outer contour, internal cavities, chamfers, and holes. The 3D geometric model, composed of points, lines, and surfaces, accurately represents the shape features of the workpiece that need to be cut or preserved during actual machining, used for subsequent machining planning. Then, geometric feature extraction is performed, and the geometric model is analyzed to identify key attributes relevant to machining. Mathematical algorithms automatically identify different regions and structural characteristics within the geometric model to provide a basis for machining strategies. Geometric features include surface type, curvature, feature regions, precision-sensitive regions, and dimensions. Surface type refers to the basic shape of the workpiece's geometric surface, such as a plane, cylinder, sphere, or freeform surface. Different surface types require different tool orientations and machining methods, thus needing accurate identification when planning the machining path. Curvature describes the degree of surface bending, corresponding to the rate of change of local geometry on the workpiece surface. A larger curvature indicates a steeper surface, requiring a smaller tool path to ensure surface quality. Feature regions refer to local areas in a geometric model that are significant for machining, such as steps, cavities, bosses, or sharp corners, requiring specific tool orientations or machining strategies. Precision-sensitive regions are areas in the geometry that are highly sensitive to machining errors, such as transition surfaces, mating surfaces, or functional surfaces, requiring higher tool density and stricter control of machining parameters. Dimensions refer to the size information constituting the geometric features of a workpiece, such as length, width, depth, or diameter. Dimensions not only directly affect machining strategies but also determine the range of tool selection and process planning.

[0027] S2: Based on the workpiece's geometric features and processing requirements, historical processing data is expanded and matched to obtain a set of historical processing logs including historical processing data, tool path data, process parameters, and processing results. Machine learning is then performed on the set of historical processing logs to construct a processing quality prediction model.

[0028] Furthermore, this application also includes: performing initial matching of historical processing data based on the workpiece geometric features and processing requirements to obtain an initial historical processing data set; traversing the initial historical processing data set to perform edge direction expansion to obtain a historical processing log set; dividing the historical processing log set into a base class dataset and a new class dataset; and performing machine learning based on the base class dataset and the new class dataset to construct the processing quality prediction model.

[0029] Furthermore, this application also includes: extracting similar representative data from the initial historical processing data set to obtain representative centers; filtering edge data from the initial historical processing data set based on the representative centers to determine an edge historical processing data set; filtering multi-point cluster centers from the edge historical processing data set to obtain multiple clustered edge historical processing data centers; and expanding the initial historical processing data set with edge directionality based on the multiple clustered edge historical processing data centers to obtain a historical processing log set.

[0030] Furthermore, this application also includes: the representative center is the historical processed data corresponding to the maximum sum of similarities with other historical processed data in the initial historical processed data set.

[0031] Furthermore, this application also includes: constructing a representative center neighborhood of the representative center according to a preset edge iteration filtering bandwidth; performing edge diffusion on the neighborhood edges of the representative center neighborhood according to the preset edge iteration filtering bandwidth to obtain a diffused representative center neighborhood; comparing the data volume difference between the diffused representative center neighborhood and the representative center neighborhood, and if the data volume difference is greater than or equal to a preset threshold, stopping the diffusion and adding the data other than the representative center neighborhood from the initial historical processing data set to the edge historical processing data set; if the data volume difference is less than the preset threshold, continuing to perform edge diffusion on the diffused representative center neighborhood according to the preset edge iteration filtering bandwidth until the data volume difference between two adjacent neighborhoods is less than the preset threshold, thereby obtaining the edge historical processing data set.

[0032] Furthermore, this application also includes: randomly selecting multiple edge historical processing data from the edge historical processing data set, identifying the distance between the multiple edge historical processing data, and obtaining multiple clustered edge historical processing data centers when the distance between any two edge historical processing data is greater than or equal to a preset distance constraint.

[0033] Furthermore, this application also includes: using the base class dataset to perform machine learning training on the framework built based on the feedforward neural network to obtain an initial processing quality prediction model; and using the new class dataset to perform generalization training on the initial processing quality prediction model to obtain a processing quality prediction model.

[0034] Specifically, initial matching of historical machining data based on workpiece geometry and machining requirements refers to finding similar machining cases from existing historical machining records based on the current workpiece's shape characteristics and machining requirements. Machining requirements include surface roughness requirements, allowable geometric error ranges, machining time limits, and tool type, among other conditions. Initial matching involves finding the machining instance closest to the target workpiece, thus forming an initial set of historical machining data for subsequent modeling.

[0035] Furthermore, extracting representative data from the initial historical machining dataset involves identifying the record that best represents the overall data distribution from all matched historical machining data, serving as the representative center. This extraction is based on similarity calculations, which can be comprehensively measured by factors such as geometric features, machining parameters, workpiece surface complexity, or machining results. For example, a machining record with the smallest sum of differences in curvature, tool feed rate, or surface roughness compared to other records in the dataset is more likely to become a representative center.

[0036] The representative center is the historical machining data point with the highest sum of similarities to other historical machining data in the initial historical machining data set. In other words, it is the machining data point selected from all initial machining records that is most similar to all other records as a whole. The sum of similarities refers to the total similarity obtained by comparing a given data point with all other data in the set one by one. Similarity can be calculated based on multiple features such as curvature, surface type, toolpath type, machining speed, machining time, or surface roughness. For example, the difference in curvature, dimension, or machining temperature between two machining records can be comprehensively scored; the smaller the value, the smaller the difference and the higher the similarity. By calculating and summing the similarities of each historical machining data point with several other data points in the set, a series of similarity sums can be obtained. The data point with the highest sum is then selected, indicating that it maintains a high similarity to the most data points overall, and therefore can serve as the central reference for the entire dataset.

[0037] Furthermore, constructing a representative center neighborhood according to a preset edge iteration filtering bandwidth means selecting historical processed data that are close to the representative center's features based on a pre-defined bandwidth range to form an initial neighborhood. The preset edge iteration filtering bandwidth represents an allowed range of feature differences.

[0038] Edge diffusion, which iteratively filters the neighborhood edges of a representative center using a preset bandwidth, refers to expanding the data outwards from the initial neighborhood. The neighborhood edge refers to the outermost ring of data points in the initial neighborhood. By applying the same bandwidth again, more records that might belong to the neighborhood can be found in the outer ring. The diffusion operation involves progressively expanding a multidimensional sphere centered on the representative center in the data space, gradually increasing the neighborhood range to determine the central distribution range of the data.

[0039] Comparing the difference in data volume between the neighborhood of the diffusion representative center and the original neighborhood of the diffusion center is a measure of the diffusion effect; that is, comparing the change in the amount of data contained in the neighborhood before and after diffusion. If the amount of data in the neighborhood after diffusion increases significantly compared to the original neighborhood, and reaches a preset threshold, it indicates that the data distribution has significantly expanded in that direction. At this point, diffusion is terminated, and all data in the initial historical processed data set that do not belong to the neighborhood are included in the marginal historical processed data set. The preset threshold is used to control the scale of diffusion, effectively identifying data that is far from the center and difficult to classify into the core area, making it marginal data, which helps to supplement the boundary parts of the data distribution in the future.

[0040] If the difference in data volume is less than the preset threshold, it means that the neighborhood after diffusion has not increased significantly and the data distribution has not widened significantly. At this time, continue to carry out the next round of diffusion on the basis of the diffusion neighborhood with the same bandwidth until the difference in data volume generated by two consecutive diffusions is lower than the preset threshold. That is, detect whether the neighborhood has reached the natural boundary of data density. When the neighborhood after diffusion gradually stabilizes and no longer increases significantly, it means that the neighborhood has fully covered the core data area. At this time, the final edge historical processing data set can be generated.

[0041] Furthermore, multi-point cluster center screening of the edge historical processing data set refers to further searching for multiple center points in the identified edge data that are not very close to each other but jointly represent the edge distribution. These are called clustered edge historical processing data centers. Cluster screening measures the distance between each pair of edge data. When the distance is greater than a set minimum constraint, it can be used as an independent cluster center.

[0042] Off-center directional augmentation of the initial historical processing data set, based on multiple clustered edge historical processing data centers, refers to starting from the cluster center and searching for or generating more relevant processing records in different directions within the data space, ultimately forming a more complete and richer set of historical processing logs. Off-center directional augmentation is used to supplement sparse or underrepresented areas in the data, making the data coverage more comprehensive under various processing conditions.

[0043] Furthermore, dividing the historical machining log set into base class datasets and new class datasets means organizing historical data hierarchically according to their typicality and representativeness. The base class dataset contains a large amount of machining data that is stable and highly representative, such as cases of planar milling or conventional surface machining; the new class dataset contains relatively sparse data that is frequently updated or has special characteristics, such as new cutting tools, extreme curvature regions, or high-precision feature surfaces. This facilitates the phased learning of machine learning models: first acquiring basic capabilities from stable data, and then expanding and generalizing through new class data.

[0044] Furthermore, multiple edge historical processing data are randomly selected from the edge historical processing data set. This means that several data samples are arbitrarily extracted from the historical processing data after edge screening. These samples are used to further analyze the distribution of data in the high-dimensional feature space. Edge historical processing data refers to historical processing records that are far from the representative center and can reflect the overall change boundary of the data. The distance between multiple edge historical processing data is identified. This means that the feature difference between any two data points is calculated using metrics such as Euclidean distance, Manhattan distance, or cosine distance to determine their relative position in the feature space. When the distance between any two edge historical processing data is greater than or equal to a preset distance constraint, it means that as long as the distance between any two samples is not lower than the threshold set by the system, for example, if the distance threshold is set to 5 and the distance between two points in the distance measurement is 6 mm, it is considered to meet the constraint, thus indicating that these data have sufficient dispersion and representativeness in space. Multiple clustered edge historical processing data centers are obtained. This means that several data points that are far apart from each other in the feature space are selected and used as multiple starting points for subsequent edge direction expansion. These serve as cluster centers of the data edge region to describe the data distribution boundary in different directions.

[0045] Furthermore, a stable, abundant, and clearly distributed base class dataset is used to train the framework built on a feedforward neural network for machine learning, obtaining an initial machining quality prediction model. A feedforward neural network refers to a structure where data propagates sequentially from the input layer through several hidden layers to the output layer, like a computational chain flowing from left to right. For example, several features such as curvature, surface type, cutting speed, and feed rate are input into the input layer. After multiple linear transformations and nonlinear activations, the output is a predicted value for surface roughness or machining error.

[0046] Subsequently, the initial machining quality prediction model was generalized and trained using a new dataset to obtain a new machining quality prediction model. This means adding a small amount of new data with unique features and a distribution that deviates from mainstream machining scenarios to the training, so that the initial machining quality prediction model is not only applicable to common situations, but can also handle some uncommon but real machining conditions such as special curved surfaces, extreme machining parameters, and offset tool axis directions. The new dataset covers edge features, such as conditions with steeper curvature, more complex surface areas, or more extreme machining speeds. Generalization training refers to continuing training or fine-tuning to ensure that the initial machining quality prediction model can maintain its predictive ability in the new feature space without significant shift.

[0047] S3: Based on the workpiece's geometric features, candidate tool positions are identified to obtain a set of candidate tool position locations. Based on the machining quality prediction model, the set of candidate tool position locations is analyzed to obtain a set of candidate tool position machining quality data, including surface quality prediction value, geometric error prediction value, machining coverage, and machining estimated time.

[0048] Specifically, candidate tool positions are identified based on the workpiece's geometric features. This involves filtering potential tool entry or machining locations based on the workpiece's geometric characteristics, such as surface types in different regions, local curvature variations, dimensional features, and precision-sensitive areas. Candidate tool positions refer to spatial points that can be used by the tool to contact or cut the workpiece. Candidate tool position identification selects a set of spatial points from all possible machining spaces that satisfy machining accessibility, tool posture safety, and machining coverage. For example, in areas with high curvature, a denser distribution of tool positions may be needed to ensure machining accuracy, while in planar areas, sparser tool positions can be selected. Areas with large surface dimensions may generate candidate tool positions in multiple sub-regions to ensure machining integrity. The selected tool positions are stored as a set to obtain the candidate tool position set.

[0049] Then, the candidate tool position sets are analyzed based on the machining quality prediction model. This means using the machining quality prediction model to predictively evaluate each candidate tool position under possible machining conditions. The machining quality prediction model can take into account the geometric attributes, local curvature, and tool parameters of the tool position and output corresponding quality indicators. A set of machining quality data for candidate tool positions is obtained, including predicted surface quality values, predicted geometric error values, machining coverage, and estimated machining time. In other words, each candidate tool position generates a set of data containing multiple evaluation indicators. The predicted surface quality value represents the possible surface finish of the workpiece after machining at that tool position, such as a predicted value based on roughness parameters; the predicted geometric error value represents the deviation between the actual contour and the ideal geometry after machining, such as predicted shape or dimensional errors; machining coverage refers to the coverage of the local area by the tool position in the machining path planning, such as the percentage of an area that a tool position can cover; the estimated machining time represents the time required to complete the local machining task under the conditions of that tool position, such as the ratio of tool feed distance to feed rate.

[0050] S4: Based on the set of candidate tool positions machining quality data, the set of candidate tool positions is filtered to determine the set of key tool positions.

[0051] Specifically, the candidate tool position set is screened based on the candidate tool position machining quality data set. This means using the candidate tool position machining quality data as a judgment criterion to evaluate all candidate tool positions and eliminate those that do not meet machining requirements or have insufficient quality. From the screened tool positions, the set of tool positions that are most representative of machining, have the best quality, or have the strongest coverage is selected to determine the set of key tool positions, which serves as a reference for subsequent machining path planning.

[0052] S5: Optimize independent machining parameters based on the candidate tool position machining quality data set and the key tool position set, and optimize the linkage machining path by calling the linkage CNC machine tool machining constraints to obtain the optimal machining path.

[0053] Furthermore, this application also includes: performing independent machining parameter optimization on the set of key tool positions based on the candidate tool position machining quality data set to obtain a set of key tool position machining parameters; taking each key tool position in the set of key tool positions as a node in a graph structure, and constructing a linkage optimization graph structure based on the spatial distance between key tool positions, the included angle of the tool axis direction, and the adjacency relationship of the machining area; calling the linkage CNC machine tool machining constraints to perform linkage machining path optimization on the set of key tool position machining parameters according to the linkage optimization graph structure to obtain the optimal machining path.

[0054] Furthermore, this application also includes: the machining constraints of the linkage CNC machine tool are to minimize the machining time and minimize the geometric error.

[0055] Specifically, based on the machining quality data set of candidate tool positions, independent machining parameter optimization is performed on the set of key tool position points. This means that machining parameters are individually optimized for each of the selected key tool position points using machining quality prediction data associated with each candidate tool position point. Independent machining parameter optimization refers to adjusting parameters such as tool posture, feed rate, and spindle speed for each key tool position point to achieve the optimal machining quality for that tool position point.

[0056] Each critical tool point in the set of critical tool points is treated as a node in the graph structure, meaning each critical tool point is abstracted as an independent node in the graph. A linked optimization graph structure is constructed based on the spatial distance between critical tool points, the tool axis angle, and the adjacency relationship of machining areas. This represents a weighted graph built based on several types of geometric and machining relationships between nodes. Spatial distance refers to the positional difference of tool points in three-dimensional space, used to measure the time consumption and path rationality of tool position switching; the tool axis angle represents the difficulty of tool posture change when transitioning from one tool position to the next; the machining area adjacency relationship indicates whether two tool positions belong to adjacent machining areas, thus affecting path continuity and machining efficiency. The constructed linked optimization graph structure allows subsequent optimization to find the optimal machining sequence among nodes.

[0057] The linkage CNC machine tool machining constraint conditions are invoked, and based on the linkage optimization graph structure, the machining path optimization is performed on the set of machining parameters for key tool positions. This means that overall path planning is performed on the constructed graph structure according to the physical and control constraints of the CNC machine tool, ultimately obtaining the optimal machining path. The linkage CNC machine tool machining constraints may include maximum acceleration limits, tool rotation limits, joint range of motion, and upper limits of machining speed, used to ensure that path planning is both executable and efficient. Linkage machining path optimization refers to comprehensively considering machining continuity, tool posture change costs, machining time, and quality to find the optimal order and corresponding parameters passing through all key tool positions. For example, if a path segment violates the joint limits of the machine tool, it will be eliminated by the path optimization algorithm, thus ensuring that the final generated path is truly machinable.

[0058] Furthermore, the machining constraints for linked CNC machine tools are minimizing machining time and minimizing geometric errors. This means that during the optimization of machining paths and parameters, two constraint objectives are set: to minimize the overall machining time while ensuring the executability of multi-axis linkage on the machine tool, and to minimize the geometric deviations generated during machining. Linked CNC machine tools refer to CNC machining equipment capable of achieving synchronous multi-axis motion. When planning the path, they must meet the displacement range, speed limits, and attitude change capabilities of each axis. Machining constraints refer to the target or restriction terms used to guide the optimization algorithm. Minimizing machining time means aiming to minimize the overall machining time by rationally planning the tool position sequence, reducing position switching distances, and avoiding unnecessary idle travel. Minimizing geometric errors means minimizing the dimensional and shape deviations of the final machined surface by controlling tool attitude changes, maintaining a reasonable feed rate, and reducing the risk of overcutting.

[0059] In summary, the machine learning-based CNC machine tool machining path optimization method provided in this application has the following technical effects: by achieving accurate matching and in-depth utilization of historical machining data based on multi-dimensional local features, a more reliable machining knowledge base is constructed, supporting the technical goal of adaptive path planning and predictive modeling. This achieves the technical effects of more accurate path planning, more reliable machining quality prediction, and improved machining process stability in the machining of complex workpieces.

[0060] Example 2: Based on the same inventive concept as the machine learning-based CNC machine tool machining path optimization method in the foregoing examples, this application also provides a machine learning-based CNC machine tool machining path optimization system. Please refer to the appendix. Figure 2The system includes: a workpiece geometry feature acquisition module 1, used to acquire the geometric model of the target workpiece and extract its geometric features; a machining quality prediction model construction module 2, used to expand and match historical machining data based on the workpiece geometry features and machining requirements to obtain a historical machining log set including historical machining data, tool path data, process parameters, and machining results, and to perform machine learning based on the historical machining log set to construct a machining quality prediction model; a candidate tool position machining quality data set acquisition module 3, used to identify candidate tool positions based on the workpiece geometry features to obtain a candidate tool position set, and to analyze the candidate tool position set based on the machining quality prediction model to obtain a candidate tool position machining quality data set including surface quality prediction value, geometric error prediction value, machining coverage, and machining estimation time; a key tool position set determination module 4, used to filter the candidate tool position set based on the candidate tool position machining quality data set to determine the key tool position set; and an optimal machining path acquisition module 5, used to perform independent machining parameter optimization based on the candidate tool position machining quality data set and the key tool position set, and to call the linkage CNC machine tool machining constraints to perform linkage machining path optimization to obtain the optimal machining path.

[0061] Furthermore, the CNC machine tool machining path optimization system based on machine learning is also used for: performing initial matching of historical machining data according to the workpiece geometric features and machining requirements to obtain an initial historical machining data set; traversing the initial historical machining data set to perform edge directional expansion to obtain a historical machining log set; dividing the historical machining log set into a base class dataset and a new class dataset; and performing machine learning based on the base class dataset and the new class dataset to construct the machining quality prediction model.

[0062] Furthermore, the CNC machine tool machining path optimization system based on machine learning is also used to: train the framework based on the feedforward neural network using the base class dataset to obtain an initial machining quality prediction model; and generalize the initial machining quality prediction model using the new class dataset to obtain a machining quality prediction model.

[0063] Furthermore, the CNC machine tool machining path optimization system based on machine learning is also used for: extracting similar representative data from the initial historical machining data set to obtain representative centers; filtering edge data from the initial historical machining data set based on the representative centers to determine edge historical machining data sets; filtering multi-point cluster centers from the edge historical machining data sets to obtain multiple clustered edge historical machining data centers; and expanding the initial historical machining data set with edge directionality based on the multiple clustered edge historical machining data centers to obtain a historical machining log set.

[0064] Furthermore, the CNC machine tool machining path optimization system based on machine learning is also used to: the representative center is the historical machining data corresponding to the maximum sum of similarity with other historical machining data in the initial historical machining data set.

[0065] Furthermore, the CNC machine tool machining path optimization system based on machine learning is also used for: constructing a representative center neighborhood of the representative center according to a preset edge iteration filtering bandwidth; performing edge diffusion on the neighborhood edges of the representative center neighborhood according to the preset edge iteration filtering bandwidth to obtain a diffused representative center neighborhood; comparing the difference in data volume between the diffused representative center neighborhood and the representative center neighborhood, and if the difference in data volume is greater than or equal to a preset threshold, stopping the diffusion and adding the data other than the representative center neighborhood from the initial historical machining data set to the edge historical machining data set; if the difference in data volume is less than the preset threshold, continuing to perform edge diffusion on the diffused representative center neighborhood according to the preset edge iteration filtering bandwidth until the difference in data volume between two adjacent neighborhoods is less than the preset threshold, thereby obtaining the edge historical machining data set.

[0066] Furthermore, the CNC machine tool machining path optimization system based on machine learning is also used to: randomly select multiple edge historical machining data from the edge historical machining data set, identify the distance between the multiple edge historical machining data, and obtain multiple clustered edge historical machining data centers when the distance between any two edge historical machining data is greater than or equal to a preset distance constraint.

[0067] Furthermore, the machine learning-based CNC machine tool machining path optimization system is also used for: performing independent machining parameter optimization on the set of key tool positions based on the candidate tool position machining quality data set to obtain a set of key tool position machining parameters; taking each key tool position in the set of key tool positions as a node in a graph structure, and constructing a linkage optimization graph structure based on the spatial distance between key tool positions, the included angle of the tool axis direction, and the adjacency relationship of the machining area; calling the linkage CNC machine tool machining constraints to perform linkage machining path optimization on the set of key tool position machining parameters according to the linkage optimization graph structure to obtain the optimal machining path.

[0068] Furthermore, the machine learning-based CNC machine tool machining path optimization system is also used to: link the CNC machine tool machining constraints to minimize machining time and minimize geometric error.

[0069] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The machine learning-based CNC machine tool machining path optimization method and specific examples in the aforementioned Embodiment 1 are also applicable to the machine learning-based CNC machine tool machining path optimization system of this embodiment. Through the foregoing detailed description of the machine learning-based CNC machine tool machining path optimization method, those skilled in the art can clearly understand the machine learning-based CNC machine tool machining path optimization system of this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0070] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0071] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A CNC machine tool machining path optimization method based on machine learning, characterized in that, The method includes: Obtain the geometric model of the target workpiece and extract its geometric features to obtain the workpiece's geometric features; Based on the workpiece's geometric features and processing requirements, historical processing data is expanded and matched to obtain a set of historical processing logs, including historical processing data, tool path data, process parameters, and processing results. Machine learning is then performed on the set of historical processing logs to construct a processing quality prediction model. Candidate tool positions are identified based on the workpiece's geometric features to obtain a set of candidate tool position locations. The set of candidate tool position locations is then analyzed based on the machining quality prediction model to obtain a set of candidate tool position machining quality data, which includes surface quality prediction value, geometric error prediction value, machining coverage, and machining estimated time. The candidate tool position set is filtered based on the candidate tool position machining quality data set to determine the key tool position set; Based on the candidate tool position machining quality data set and the key tool position set, perform independent machining parameter optimization, and call the linkage CNC machine tool machining constraints to optimize the linkage machining path to obtain the optimal machining path.

2. The CNC machine tool machining path optimization method based on machine learning as described in claim 1, characterized in that, Based on the workpiece's geometric features and machining requirements, historical machining data is expanded and matched to obtain a historical machining log set including historical machining data, toolpath data, process parameters, and machining results. Machine learning is then performed on this historical machining log set to construct a machining quality prediction model, including: Based on the workpiece's geometric features and processing requirements, initial matching of historical processing data is performed to obtain an initial set of historical processing data. The initial historical processing data set is traversed and edge-directional expansion is performed to obtain a historical processing log set; The historical processing log set is divided into a base class dataset and a new class dataset; Machine learning is performed on the base class dataset and the new class dataset to construct the processing quality prediction model.

3. The CNC machine tool machining path optimization method based on machine learning as described in claim 2, characterized in that, Machine learning is performed based on the base class dataset and the new class dataset to construct the processing quality prediction model, including: The framework built on a feedforward neural network is trained using the base class dataset to obtain an initial processing quality prediction model. The initial processing quality prediction model is generalized and trained using a new dataset to obtain a new processing quality prediction model.

4. The CNC machine tool machining path optimization method based on machine learning as described in claim 2, characterized in that, The initial historical processing data set is traversed and expanded directionally at the edges to obtain a historical processing log set, including: Similar representative data are extracted from the initial historical processing data set to obtain representative centers; Based on the representative center, the initial historical processing data set is filtered for edge data to determine the edge historical processing data set; The edge historical processing data set is filtered by multiple points to obtain multiple clustered edge historical processing data centers; The initial historical processing data set is expanded directionally at the edge based on the multiple clustered edge historical processing data centers to obtain a historical processing log set.

5. The CNC machine tool machining path optimization method based on machine learning as described in claim 4, characterized in that, The representative center is the historical processed data corresponding to the maximum sum of similarities with other historical processed data in the initial historical processed data set.

6. The CNC machine tool machining path optimization method based on machine learning as described in claim 4, characterized in that, Based on the representative centers, the initial historical processing data set is filtered for edge data to determine the edge historical processing data set, including: Based on the preset edge iteration bandwidth selection, construct the representative center neighborhood of the representative center; The neighborhood edges of the representative central neighborhood are diffused according to a preset edge iteration filtering bandwidth to obtain a diffused representative central neighborhood; Compare the difference in data volume between the representative center neighborhood and the representative center neighborhood. If the difference in data volume is greater than or equal to a preset threshold, stop the diffusion and add the data other than the representative center neighborhood from the initial historical processing data set to the edge historical processing data set. If the difference in data volume is less than a preset threshold, the edge diffusion of the representative center neighborhood continues according to the preset edge iteration filtering bandwidth until the difference in data volume between two adjacent neighborhoods is less than the preset threshold, and the edge historical processing data set is obtained.

7. The CNC machine tool machining path optimization method based on machine learning as described in claim 4, characterized in that, Multiple edge historical processing data are randomly selected from the edge historical processing data set, and the distance between the multiple edge historical processing data is identified. When the distance between any two edge historical processing data is greater than or equal to a preset distance constraint, multiple clustered edge historical processing data centers are obtained.

8. The CNC machine tool machining path optimization method based on machine learning as described in claim 1, characterized in that, Based on the candidate tool position machining quality data set and the key tool position set, independent machining parameter optimization is performed, and the linkage machining path is optimized by invoking the linkage CNC machine tool machining constraints to obtain the optimal machining path, including: Based on the candidate tool position machining quality data set, independent machining parameter optimization is performed on the key tool position set to obtain the key tool position machining parameter set. Each critical tool point in the set of critical tool points is taken as a node in the graph structure, and a linkage optimization graph structure is constructed based on the spatial distance between critical tool points, the included angle of the tool axis direction, and the adjacency relationship of the machining area. By invoking the machining constraints of the linkage CNC machine tool and optimizing the machining path of the key tool position point machining parameter set according to the linkage optimization diagram structure, the optimal machining path is obtained.

9. The CNC machine tool machining path optimization method based on machine learning as described in claim 8, characterized in that, The machining constraints of a linkage CNC machine tool are to minimize machining time and minimize geometric error.

10. A CNC machine tool machining path optimization system based on machine learning, characterized in that, The steps for implementing the machine learning-based CNC machine tool machining path optimization method according to any one of claims 1 to 9 include: The workpiece geometry feature acquisition module is used to acquire the geometric model of the target workpiece and extract its geometric features to obtain the workpiece geometry features. The machining quality prediction model construction module is used to expand and match historical machining data based on the workpiece's geometric features and machining requirements to obtain a set of historical machining logs including historical machining data, tool path data, process parameters, and machining results. Machine learning is then performed on the set of historical machining logs to construct a machining quality prediction model. The candidate tool position machining quality data set acquisition module is used to identify candidate tool positions based on the workpiece geometric features, obtain a candidate tool position set, and analyze the candidate tool position set based on the machining quality prediction model to obtain a candidate tool position machining quality data set that includes surface quality prediction value, geometric error prediction value, machining coverage and machining estimated time. The critical tool position set determination module is used to filter the candidate tool position set based on the candidate tool position machining quality data set to determine the critical tool position set; The optimal machining path acquisition module is used to perform independent machining parameter optimization based on the candidate tool position machining quality data set and the key tool position set, and to call the linkage CNC machine tool machining constraints to perform linkage machining path optimization to obtain the optimal machining path.

Images