A deburring control method for a numerical control machine tool based on an intelligent algorithm

Abstract

The application discloses a kind of numerical control machine tool deburring control methods based on intelligent algorithm.The method uses depth camera and 2D vision sensor to scan the burr area of workpiece, obtains the three-dimensional point cloud data and surface texture information of burr, so as to accurately calculate the position coordinates, height value and distribution density of burr.Through particle swarm optimization algorithm, the optimal motion trajectory of tool is calculated with machining efficiency and path smoothness as target.According to the optimized motion trajectory, numerical control machine tool executes deburring processing, and collects cutting force data.When cutting force exceeds preset threshold, feed speed and spindle speed are dynamically adjusted, and the adjusted parameters are fed back to particle swarm optimization algorithm for re-trajectory planning.The method realizes efficient and accurate deburring processing by combining intelligent sensing and algorithm optimization, effectively improves machining quality and efficiency, and has strong environmental adaptability and real-time adjustment capability, suitable for machining scenes of various complex workpieces.

Description

Technical Field

[0001] This invention relates to the field of CNC machine tool control technology, and in particular to a deburring control method for CNC machine tools based on intelligent algorithms. Background Technology

[0002] In recent years, with the continuous improvement of industrial automation, CNC machine tools have been widely used in the manufacturing field. During workpiece machining, deburring, as a crucial step in ensuring machining quality and finished product performance, has gradually gained importance. Current technologies mostly employ mechanical deburring or automated methods based on fixed trajectories. These methods typically rely on manually setting the machining path and combining it with simple sensor feedback to complete the machining task. Meanwhile, some intelligent machining methods are beginning to explore optimizing machining paths through algorithms to improve machining efficiency and accuracy.

[0003] However, some problems still exist in existing technologies. Mechanical deburring methods are easily limited by the complexity of workpiece shape and material, and have low processing efficiency; fixed-path automated methods lack flexibility and are difficult to adapt to the varied burr distribution on complex workpiece surfaces; while intelligent machining technologies have made some improvements, they are still not perfect in terms of data acquisition accuracy, real-time performance, and path optimization capabilities, and cannot meet the needs of efficient and precise machining. In addition, existing technologies are also insufficient in their ability to adjust machining parameters in real time to cope with dynamic changes.

[0004] Therefore, it is necessary to provide a new technical solution to solve the above problems. Summary of the Invention

[0005] This application provides a deburring control method for CNC machine tools based on intelligent algorithms to improve the deburring quality and efficiency of workpieces.

[0006] This application provides a deburring control method for CNC machine tools based on intelligent algorithms, including:

[0007] A depth camera and a 2D vision sensor are used to scan the burr area of ​​the workpiece to obtain the 3D point cloud data and surface texture information of the burr.

[0008] Using the three-dimensional point cloud data and combined with surface texture information, the location coordinates, height value and distribution density of the burrs are calculated.

[0009] Based on the position coordinates, height value, and distribution density of the burrs, the optimal motion trajectory of the tool is calculated using a particle swarm optimization algorithm, with processing efficiency and path smoothness as optimization objectives.

[0010] Deburring is performed according to the optimal motion trajectory, and cutting force data is collected in real time;

[0011] When the cutting force exceeds the preset threshold, the feed rate and spindle speed are adjusted, and the adjusted parameters are fed back to the particle swarm optimization algorithm to replan the trajectory.

[0012] The beneficial effects of the technical solution provided in this application include:

[0013] (1) This method combines a depth camera and a 2D vision sensor to acquire three-dimensional point cloud data and surface texture information of workpiece burrs, which can accurately locate the specific position and morphological characteristics of burrs, thereby effectively improving the accuracy of deburring and meeting the processing requirements of highly complex workpieces. (2) Using the particle swarm optimization algorithm with processing efficiency and path smoothness as optimization objectives, the optimal motion trajectory of the tool is calculated, which significantly reduces processing time and tool path redundancy and improves overall processing efficiency. (3) Cutting force data is collected in real time, and the feed rate and spindle speed are dynamically adjusted when the cutting force exceeds the preset threshold. At the same time, the trajectory is replanned through the feedback mechanism to effectively cope with dynamic changes in the processing process and ensure the stability and adaptability of the processing process. (4) Through dynamic optimization and adjustment of the processing path and parameters, processing defects caused by tool overload or improper motion trajectory are avoided, and the surface quality of the workpiece and the reliability of the processing process are improved. Attached Figure Description

[0014] Figure 1 This is a flowchart of a deburring control method for CNC machine tools based on intelligent algorithms, provided in the first embodiment of this application. Detailed Implementation

[0015] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0016] The first embodiment of this application provides a deburring control method for CNC machine tools based on intelligent algorithms. Please refer to... Figure 1 This figure is a schematic diagram of the first embodiment of this application. The following is in conjunction with... Figure 1 The first embodiment of this application provides a detailed description of a deburring control method for CNC machine tools based on intelligent algorithms.

[0017] Step S101: Use a depth camera and a 2D vision sensor to scan the burr area of ​​the workpiece to obtain the three-dimensional point cloud data and surface texture information of the burr.

[0018] In step S101, a depth camera and a 2D vision sensor are used to scan the burr area of ​​the workpiece to obtain the three-dimensional point cloud data and surface texture information of the burr. This is the basic step of the whole method and directly determines the accuracy and reliability of subsequent data processing and optimization.

[0019] First, it is necessary to select a depth camera and a 2D vision sensor suitable for the workpiece machining environment. The depth camera can be a device based on laser, structured light, or Time-of-Flight (ToF) technology to capture the three-dimensional geometric information of the workpiece surface; the 2D vision sensor can be a high-resolution industrial camera capable of clearly acquiring the texture features of the workpiece surface. To ensure the accuracy and comprehensiveness of data acquisition, the depth camera and 2D vision sensor need to be mounted in specific locations on the CNC machine tool, such as on the machine tool spindle or a dedicated scanning mechanism, to provide a stable, unobstructed field of view within the machining area.

[0020] Before data acquisition, it is crucial to perform necessary pretreatment on the workpiece surface. For example, for smooth or highly reflective surfaces, matte powder can be sprayed or a reflective-suppressing coating can be applied to reduce optical noise during the acquisition process; for dirty or contaminated surfaces, cleaning methods can be used to improve the quality of the acquired data. Subsequently, the coordinate system consistency between the depth camera and the 2D vision sensor is ensured through joint calibration. This process can be performed using a calibration board or other tools with known geometric features, generating accurate extrinsic and intrinsic parameter matrices.

[0021] In the specific scanning operation, the movement of the CNC machine tool is controlled to gradually cover the burr area of ​​the workpiece according to the set scanning path. The scanning path can be preset according to the geometry of the workpiece, using linear scanning, serpentine scanning, or spiral scanning modes. During scanning, the depth camera generates point cloud data, which contains the three-dimensional coordinate information of each point on the workpiece surface; at the same time, the surface texture image captured by the 2D vision sensor is fused with the point cloud data through pixel-level matching, thereby generating a composite dataset with rich geometric and texture information.

[0022] To further improve data quality, the acquired point cloud data and texture information can be preprocessed. Point cloud data preprocessing includes denoising, filtering, and downsampling operations to reduce data redundancy and improve computational efficiency. For example, statistical filtering or radius filtering methods can be used to remove outliers; voxel downsampling techniques can be used to reduce point cloud resolution while preserving key geometric features. For 2D texture information, image enhancement techniques can be used to improve the visibility of texture details, such as adjusting contrast, sharpening edges, or eliminating uneven lighting.

[0023] After scanning and data fusion are completed, the spur areas need to be initially segmented and labeled. This step can be based on the feature differences between point clouds and textures, using clustering algorithms (such as DBSCAN or region growing algorithms) to identify possible spur areas and separate them from non-spur areas. To improve the segmentation accuracy, edge detection results from 2D images (such as the Canny algorithm) can be combined as auxiliary information.

[0024] Through the above steps, high-precision 3D point cloud data and surface texture information of burrs can be obtained. This data not only provides fundamental support for subsequent steps, but also ensures the robustness and applicability of the method through a comprehensive and meticulous scanning and processing strategy, laying a solid technical foundation for deburring complex workpieces.

[0025] Furthermore, the step of using a depth camera and a 2D vision sensor to scan the burr area of ​​the workpiece to acquire 3D point cloud data and surface texture information of the burrs includes:

[0026] During the scanning process, the exposure time of the depth camera and the intensity of the light source are adjusted to dynamically adapt to the material properties and light reflection characteristics of the workpiece surface;

[0027] Based on real-time quality assessment of the scan data, it is determined whether there is missing data or noise interference. When missing data is detected, the scanning angle is adjusted and a supplementary scan is performed.

[0028] When noise interference is detected, a filtering algorithm based on spatial neighborhood and texture consistency is applied to optimize the data, thereby obtaining complete and accurate 3D point cloud data of burrs and surface texture information.

[0029] During the scanning process, the exposure time and light source intensity of the depth camera are dynamically adjusted based on the workpiece surface material characteristics and light reflection properties. This adjustment is achieved through an automatic exposure control mechanism, enabling the depth camera to effectively capture burr features on highly reflective or dark surfaces. For example, to address the strong reflection issues that may arise from smooth metal surfaces, reducing the light source intensity and shortening the exposure time can effectively reduce light interference; while for dark or rough surfaces, increasing the light source intensity and extending the exposure time improves scanning quality. Through this dynamic adjustment, the depth camera can adapt to various complex surface conditions, ensuring scanning clarity and resolution.

[0030] After acquiring the scanned data, its real-time quality is evaluated to determine its completeness and accuracy. Quality assessment is performed by analyzing the distribution characteristics of point cloud density and texture information. For example, if the point cloud density in a certain area is below a set threshold, or if there are obvious breaks in the texture information, this may indicate missing data or noise interference. In this case, the system will use a real-time feedback mechanism to prompt the operator to adjust the scanning angle or position and perform a supplementary scan to cover the missed areas. During supplementary scanning, a multi-angle scanning strategy is prioritized to ensure that each burr feature point is completely captured and to avoid blind spots.

[0031] When noise interference is detected in the scanned data, such as irregular spikes in point cloud data or random spots in texture information, optimization can be performed by applying filtering algorithms based on spatial neighborhood and texture consistency. Specifically, for 3D point cloud data, spatial neighborhood averaging filtering or Gaussian filtering methods can be used to remove isolated noise points by utilizing the distance and positional relationships between adjacent points; for surface texture information, texture consistency detection is used to identify abnormal pixels and perform smoothing. The optimized data will serve as the basis for subsequent calculations of burr features (such as position coordinates, height values, and distribution density), ensuring that the data accuracy meets the requirements of machining trajectory planning.

[0032] The above steps enable high-quality scanning of the burr area of ​​the workpiece, obtaining complete and accurate 3D point cloud data and surface texture information, providing reliable input data for subsequent deburring based on intelligent algorithms. This method is particularly suitable for workpieces with complex surface characteristics and uneven burr distribution, significantly improving the adaptability and efficiency of the deburring process.

[0033] Step S102: Using the three-dimensional point cloud data and combined with surface texture information, calculate the position coordinates, height value and distribution density of the burrs.

[0034] In step S102, the location coordinates, height values, and distribution density of burrs are calculated using 3D point cloud data and surface texture information acquired from a depth camera and a 2D vision sensor. This is one of the core steps in achieving intelligent deburring. This step requires deep fusion and fine processing of multi-source data to ensure a comprehensive and accurate characterization of burr properties.

[0035] First, the 3D point cloud data and surface texture information acquired in step S101 are aligned and synchronized. Point cloud data typically contains a large number of spatial coordinate points, while texture information represents the detailed features of the workpiece surface in image form. By registering the point cloud and texture image, the texture features are mapped to the corresponding positions of each point in the point cloud data, thereby generating point cloud data with texture enhancement properties. The registration process can employ geometric correction and feature matching algorithms, such as using SIFT or ORB feature points to establish a mapping relationship between the point cloud and the image, followed by precise alignment using the Iterative Closest Point (ICP) algorithm.

[0036] Next, the fused data is used for burr region segmentation and feature extraction. The segmentation process first requires preprocessing the point cloud data, such as using normal vector calculation and curvature analysis to distinguish burrs from smooth surfaces. Burrs typically appear as high-curvature regions; therefore, a curvature threshold can be set, and regions exceeding this threshold are marked as candidate burr points. Simultaneously, edge features from the texture information are combined to further confirm the boundary positions of the burrs. Edge features can be extracted from the texture image using the Canny edge detection algorithm, and then cross-validated with high-curvature points in the point cloud data to ensure segmentation accuracy.

[0037] For the segmented burr point cloud region, its three-dimensional position coordinates, height value, and distribution density are further calculated. The position coordinates are obtained by directly extracting three-dimensional spatial points from the point cloud, representing the specific location of the burr. The height value is calculated with reference to the workpiece's reference plane, by subtracting the reference plane's Z-axis coordinate from the Z-axis coordinate of the burr point in the point cloud. This process can use a plane fitting algorithm (such as the least squares method) to determine the reference plane. The distribution density calculation involves statistically analyzing the point density of the burr point cloud per unit volume. This can be accomplished by dividing the three-dimensional space into a voxel grid, counting and normalizing the points within each grid to obtain the density distribution of the burrs in different regions.

[0038] Furthermore, considering that the complex morphology of burrs may affect the processing, additional features of the burr region, such as surface roughness and sharpness, can be extracted. Surface roughness can be calculated using the rate of change of the normal vector of the point cloud, while sharpness can be determined by the angular relationship between a point and its neighboring points. These additional features help to further optimize the deburring strategy in subsequent processing steps.

[0039] Finally, the calculated location coordinates, height values, and distribution density are standardized before being input into the next step of the particle swarm optimization algorithm. The standardization process can employ linear normalization or Z-score standardization to map the data to a uniform range, thereby improving the processing efficiency and robustness of subsequent algorithms.

[0040] By implementing the above steps, step S102 can accurately quantify the characteristics of the burr area, providing high-quality data support for intelligent deburring.

[0041] Furthermore, the step of using the three-dimensional point cloud data and combining it with surface texture information to calculate the location coordinates, height value, and distribution density of the burrs includes:

[0042] The 3D point cloud data is initially processed by denoising filtering and voxelization to simplify the data volume, while retaining the main geometric features of the burrs.

[0043] Based on point cloud normal vector calculation and curvature analysis, burr regions are identified, and the three-dimensional spatial coordinates of each burr point are extracted as the position coordinates of the burr.

[0044] By calculating the height of the burr point, the difference between its Z-axis coordinate and the workpiece reference plane is calculated. The least squares method is used to fit the reference plane and ensure the accuracy of the calculation.

[0045] The burr distribution density is calculated based on the point cloud density per unit volume within the burr region, combined with the feature distribution of texture information.

[0046] First, the 3D point cloud data needs preliminary processing to reduce the data volume and improve computational efficiency. During processing, denoising filtering techniques are applied to remove isolated points and noise introduced by environmental interference or equipment errors. Spatial neighborhood filtering can be used, for example, calculating the weighted average of neighboring points using a Gaussian filter to smooth the data. Simultaneously, to further simplify the data volume, voxelization is employed to sparsify the point cloud, dividing it into a 3D mesh and replacing the original point data with the center point or average point of each mesh. This simplification method significantly reduces the computational burden while preserving the main geometric features of the burrs, providing optimized input data for subsequent steps.

[0047] After the initial processing of the point cloud data, burr regions are identified based on the calculation of point cloud normal vectors and curvature analysis. Specifically, normal vector calculation determines the geometric properties of each point by estimating the direction of the surface normal. Curvature analysis identifies locations where abrupt changes occur on the surface by calculating the geometric curvature of local regions of the point cloud. Burrs typically appear as high-curvature regions, so a curvature threshold can be set to distinguish burr regions from other parts of the workpiece surface. After the burr regions are identified, the three-dimensional spatial coordinates of the burr points are extracted from the point cloud and recorded as the burr position coordinates, used to describe the positional distribution of burrs on the workpiece surface.

[0048] The height of burr points is calculated using the difference between the Z-axis component of their three-dimensional coordinates and the workpiece reference plane. The workpiece reference plane is obtained through least-squares fitting, a method that constructs a plane model with minimal overall error based on a set of points in the non-burr region of the point cloud. After fitting, the height of each burr point is calculated as the difference between its Z-coordinate and the height of the reference plane. This method accurately describes the prominence of burrs, providing crucial data for subsequent trajectory planning.

[0049] Finally, the calculation of burr distribution density integrates the characteristic distribution of point cloud density and surface texture information. Specifically, the number of point clouds per unit volume within a defined burr region is counted to reflect the spatial density of the burrs. Simultaneously, by combining surface texture information and analyzing the variation patterns of texture features, such as texture gradient or texture direction consistency, the accuracy of the density calculation is further improved. This combined approach can more comprehensively describe the distribution characteristics of burrs, providing important support for optimizing processing paths.

[0050] By following the detailed steps described above, the location coordinates, height, and distribution density of burrs can be accurately calculated, providing reliable data support for subsequent deburring path planning and parameter adjustment. This method has the advantages of strong adaptability and high precision, and is suitable for processing workpieces with different shapes and surface characteristics.

[0051] Step S103: Based on the position coordinates, height value and distribution density of the burr, calculate the optimal motion trajectory of the tool using a particle swarm optimization algorithm, with processing efficiency and path smoothness as optimization objectives.

[0052] In step S103, calculating the optimal motion trajectory of the tool using the particle swarm optimization algorithm is a key step in achieving efficient and precise deburring. This step takes the geometric feature data of the burr as input and uses processing efficiency and path smoothness as optimization objectives to generate a processing trajectory that adapts to the needs of complex workpieces.

[0053] First, the burr location coordinates, height values, and distribution density calculated in step S102 are input into the initial stage of the optimization algorithm. Particle swarm optimization (PSO) is a global optimization algorithm that simulates swarm intelligence behavior. The initial stage requires defining the basic structure of the particle swarm, where each particle represents a potential tool motion trajectory solution, and its dimension corresponds to trajectory parameters (such as path point coordinates and motion speed). The initial solution of the particle swarm can be randomly generated or defined according to heuristic rules, for example, by uniformly distributing trajectory points around the burr area and setting the initial feed rate and spindle speed.

[0054] The optimization objective function considers both machining efficiency and path smoothness. Machining efficiency is closely related to the total path length and time; the shorter the path and the less machining time, the higher the machining efficiency. Path smoothness measures the rate of curvature change and smoothness of the path. Excessively steep or frequent directional changes increase tool load and machining difficulty, thus requiring optimization by limiting the rate of curvature change. The objective function can be defined as a weighted sum of two parts: machining efficiency is calculated as a function of path length and time, and path smoothness is solved by integrating the second derivative (i.e., curvature) of the path. The specific weights are set according to actual machining requirements.

[0055] The iterative process of particle swarm optimization includes two key steps: velocity update and position update. In each iteration, each particle calculates its new velocity based on its own historical best solution, the swarm's historical best solution, and the position of its current solution, and adjusts its trajectory parameters using a velocity update formula. The velocity update formula consists of three parts: an inertia term (preserving the particle's original motion trend), an individual cognition term (tending towards the particle's own historical best solution), and a social cognition term (tending towards the swarm's historical best solution). The position update formula obtains the new position by adding the updated velocity to the current trajectory parameters. To avoid particles getting trapped in local optima, the inertia weight and learning factor can be dynamically adjusted, for example, by using a linearly decreasing inertia weight or a hybrid strategy.

[0056] The constraints of particle swarm optimization require special explanation. To ensure the physical feasibility of the trajectory, the movement range of the particles should be limited to the area around the burr, while also meeting the kinematic constraints and tool load capacity of the CNC machine tool. For example, the distance between trajectory points should not be too small to avoid tool jitter or excessive friction; the feed rate and spindle speed must be within the rated range of the equipment to ensure machining stability and tool life.

[0057] During the iteration process, the fitness (i.e., objective function value) of each particle is evaluated in real time, and the distribution of particles is dynamically adjusted to gradually approach the global optimum. Typically, the algorithm terminates when the objective function values ​​of all particles converge over several consecutive generations, or when the number of iterations reaches a preset upper limit, and outputs the current global optimum as the tool's motion trajectory.

[0058] Furthermore, the performance of particle swarm optimization algorithms can be improved by introducing various enhancement strategies. For example, introducing random perturbations in the early stages of the search can enhance global search capabilities; combining local optimization algorithms (such as gradient descent) in the later stages of the search can refine trajectory adjustment; or parallel computing techniques can be used to accelerate iterative computation of large-scale particle swarms. These improvements can further enhance optimization efficiency and trajectory quality.

[0059] Through the above optimization process, step S103 realizes dynamic trajectory planning based on burr geometry, ensuring that the tool movement path achieves the best balance between processing efficiency and path smoothness, and providing a reliable technical guarantee for high-quality deburring of complex workpieces.

[0060] Furthermore, based on the position coordinates, height value, and distribution density of the burr, the optimal motion trajectory of the tool is calculated using a particle swarm optimization algorithm, wherein processing efficiency and path smoothness are the optimization objectives, including:

[0061] The geometric feature data of the burr area is partitioned, and priority processing area and secondary processing area are divided according to the difference in burr height value and distribution density.

[0062] In the initial stage of the particle swarm optimization algorithm, the trajectory points of the priority processing region are set as high-priority particles and given higher objective function weights.

[0063] By dynamically updating the fitness value of particles and combining the constraints of tool turning radius and machining path continuity, the movement direction and speed of particles are adjusted in real time.

[0064] During the iteration process, the trajectory points in the priority processing area are finely interpolated to ensure the smoothness of the trajectory and the processing quality.

[0065] First, the geometric feature data of the burr region is partitioned. By analyzing the burr height and distribution density, the entire processing area is divided into priority processing areas and secondary processing areas. Priority processing areas typically include regions with larger burr heights or higher distribution densities; these areas have higher processing complexity and difficulty, and the need for path optimization is more urgent. Secondary processing areas mainly contain regions with simpler burr characteristics, suitable for placement as secondary processing steps in the optimized trajectory path. This partitioning allows for targeted allocation of computational resources in path planning, improving optimization efficiency.

[0066] In the initial stage of the particle swarm optimization algorithm, trajectory points in priority processing regions are designated as high-priority particles and assigned higher objective function weights. The weights of processing efficiency and path smoothness in the objective function are dynamically adjusted to accommodate the higher requirements for trajectory optimization in priority processing regions. In this way, the particle swarm optimization focuses its search on priority processing regions, ensuring that trajectory planning in complex regions is optimized first.

[0067] Particle swarm optimization (PSO) dynamically updates the fitness values ​​of particles, continuously adjusting their direction and speed during iteration. The particle fitness value is calculated based on an objective function that comprehensively considers machining efficiency, path smoothness, and machining constraints, such as tool turning radius and path continuity. The turning radius constraint ensures that the trajectory planning conforms to the tool's movement capabilities, avoiding machining errors or mechanical failures caused by excessively small turns. The path continuity constraint, by limiting abrupt changes in the trajectory, guarantees the smoothness of motion and the feasibility of the trajectory during machining. The dynamic update mechanism allows particles to adaptively adjust their positions based on the priority and constraints of the current trajectory point, thereby gradually approaching the globally optimal trajectory.

[0068] During the iterative process of particle swarm optimization, fine interpolation is performed on the trajectory points in the priority machining area to further improve the smoothness of the trajectory and the machining quality. Fine interpolation generates a higher resolution sequence of trajectory points by performing local optimization between the initially planned trajectory points. The interpolated trajectory not only reduces vibration and unnecessary pauses in tool movement, but also accurately covers burrs in complex areas, improving the integrity and efficiency of machining.

[0069] Through the above steps, a particle swarm optimization trajectory planning method based on burr characteristics and processing requirements was realized. This method can balance processing efficiency and path smoothness, adapt to workpieces with complex shapes and varied characteristics, and has extremely high practicality and reliability, providing a new solution for actual processing.

[0070] Furthermore, the calculation of the optimal tool trajectory using a particle swarm optimization algorithm based on the burr's position coordinates, height value, and distribution density also includes:

[0071] During the optimization process, the fitness function of the particle swarm is dynamically adjusted by combining the temperature change data of the processing area. When the temperature is detected to be too high, the processing load in the heat accumulation area is reduced by increasing the path interval and reducing the trajectory curvature priority.

[0072] During particle swarm optimization, the system collects real-time temperature change data of the machining area. This data comes from temperature sensors mounted on the tool or workpiece surface and accurately reflects the temperature rise caused by cutting friction or other factors during machining. The temperature change data is assigned to specific machining areas to assess the heat accumulation state of those areas. If the temperature of a machining area exceeds a preset threshold, for example, if the workpiece material may deform due to excessive temperature or the tool may wear faster due to heat accumulation, the system immediately triggers the dynamic adjustment mechanism of the particle swarm fitness function.

[0073] The fitness function is adjusted by changing the priority weights of path spacing and trajectory curvature. Path spacing refers to the spatial distance between trajectories. When heat accumulation is significant in a certain area, increasing the path spacing can effectively distribute the cutting load of the tool, reduce the machining frequency in that area, and thus reduce the continuous accumulation of heat. In addition, the priority of trajectory curvature is also reduced. Lower curvature means a straighter machining path. This path form can reduce friction and cutting force when the tool turns, thereby further reducing heat generation. By dynamically adjusting these key parameters, the particle swarm optimization algorithm can comprehensively consider machining efficiency, path smoothness, and temperature distribution when optimizing the trajectory, ensuring that the planning results still meet machining requirements under high-temperature conditions.

[0074] The adjusted trajectory planning results significantly mitigate the impact of heat accumulation on the machining area. For example, in high-temperature regions, increasing the trajectory spacing prevents the tool from concentrating on machining in the same area for extended periods, while a straighter trajectory reduces tool load, thus protecting both the tool and the workpiece. Furthermore, these adjustments prevent machining quality issues caused by excessive heat concentration, such as thermal deformation of the workpiece surface or premature tool wear. By optimizing the fitness function and incorporating temperature variation data, the trajectory planning results exhibit higher adaptability, ensuring machining stability and accuracy under complex conditions.

[0075] This embodiment incorporates temperature factors during trajectory optimization, making the machining method more intelligent and practical. Whether in high-intensity machining tasks or in machining scenarios involving heat-sensitive materials, this method significantly improves machining efficiency and workpiece quality, reduces tool wear and material damage risks, and provides reliable technical support for high-performance deburring.

[0076] Furthermore, the implementation steps of the particle swarm optimization algorithm include:

[0077] Based on the location coordinates, height value, and distribution density of burrs, the processing area is dynamically prioritized, where the priority P(A) of the processing area is calculated using the following formula 1:

[0078]

[0079] Where A is the processing area, representing the unit area into which the workpiece surface is divided; α is the average height of burr points within processing area A, used to characterize the overall protrusion of burrs within the area; D(A) is the number of burr points per unit area within processing area A, used to reflect the density of burrs within the area; G(A) represents the shortest distance from processing area A to the workpiece boundary, used to assess the processing risk of the processing area; α, β, γ are priority weighting factors.

[0080] After priority calculation is completed, the geometric center of the highest priority machining area and its nearby random perturbation points are used as the initial positions of the particle swarm. Each particle represents a possible tool movement path. The initial velocity of each particle is set to a random small range value. The particle's velocity is updated in each iteration using the following formula 2:

[0081]

[0082] in, Let be the velocity of particle i in the (t+1)th iteration; Let be the velocity of particle i in the t-th iteration; Let be the position of particle i in the t-th iteration; w(t) is the inertia weight; c1 is the individual learning factor; r1 is a uniformly distributed random number in the range [0,1]; p i c1 is the historical best position of particle i, representing the coordinates of the optimal solution found by the particle in all iterations; c2 is the social learning factor; r2 is a uniformly distributed random number in the range [0,1]; g is the historical best position of the swarm, representing the coordinates of the global optimal solution found by the particle in the iteration process of all particles in the swarm.

[0083] The fitness of a particle is determined by an objective function, which is expressed as Equation 3 below:

[0084] F(x)=λ1·∫|κ(s)| 2 ds+λ2·T(x) (3)

[0085] Where λ1 and λ2 are dynamically adjusted weighting factors; T(x) is the total processing time of the trajectory; κ(s) is the trajectory curvature, calculated according to the following formula 4:

[0086]

[0087] Where s is the arc length parameter of the trajectory; x′(s) is the first derivative of the x-coordinate of the trajectory in the direction of arc length s; x″(s) is the second derivative of the x-coordinate of the trajectory in the direction of arc length s; y′(s) is the first derivative of the y-coordinate of the trajectory in the direction of arc length s; and y″(s) is the second derivative of the y-coordinate of the trajectory in the direction of arc length s.

[0088] This embodiment calculates the optimal motion trajectory of the tool using a particle swarm optimization algorithm. The implementation steps of this method include dynamic priority partitioning, particle swarm initialization and velocity update, and fitness evaluation and optimization based on the objective function.

[0089] First, based on the location coordinates, height, and distribution density of the burrs, the processing areas on the workpiece surface are dynamically prioritized. The priority P(A) of the processing area is calculated using the following formula:

[0090]

[0091] Each term in Formula 1 is defined as follows:

[0092] A: The processing area is the area obtained by dividing the surface of the workpiece into regular grids or irregular unit areas, and each area contains a certain number of burr points.

[0093] The average height of burrs within processing area A is calculated using the following formula:

[0094]

[0095] Where N is the number of burrs in the region; z i Let z be the z-coordinate of the burr point i. ref The height of the reference plane on the workpiece surface, fitted by the least squares method, is used to characterize the overall protrusion of burrs within the region.

[0096] D(A): Burr distribution density in processing area A, defined as the number of burrs per unit area, calculated using the following formula:

[0097]

[0098] Where V is the area of ​​region A, used to reflect the density of burrs within the region.

[0099] G(A): The shortest distance from the processing area A to the workpiece boundary, defined as the minimum Euclidean distance from all points in the area to the boundary, used to assess the risk of the processing area.

[0100] α, β, γ: Priority weighting factors, with recommended values ​​ranging from α = 0.5 to 1.0, β = 0.3 to 0.7, and γ = 0.2 to 0.5. These weighting factors can be dynamically adjusted according to processing requirements; for example, when high-density areas need to be processed first, the β value can be increased.

[0101] After priority calculation, the geometric center of the highest priority machining area and its nearby random perturbation points are set as the initial positions of the particle swarm. Each particle represents a possible tool movement path, with an initial velocity of a small random value. During particle swarm optimization, the velocity of each particle is updated using the following formula:

[0102]

[0103] Each term in the formula is defined as follows:

[0104] The velocity of particle i in the (t+1)th iteration.

[0105] The velocity of particle i in the t-th iteration.

[0106] The position of particle i in the t-th iteration.

[0107] w(t): Inertia weight, used to balance global search and local optimization. The inertia weight typically decreases linearly, and an initial value of w is recommended. max =0.9, the final value is w min =0.4.

[0108] c1 and c2: Individual learning factors and social learning factors, respectively controlling the particle's trajectory towards its historical optimal position p. i The degree of closeness to the group's historical best position g. The recommended value is c1 = c2 = 2.0.

[0109] r1, r2: Uniformly distributed random numbers in the range [0,1], used to introduce randomness.

[0110] p i : The historical best position of particle i.

[0111] g: The global optimal solution in the iterative process of all particles in the particle swarm.

[0112] The fitness of a particle is evaluated using an objective function, which is expressed as follows:

[0113] F(x)=λ1·∫|κ(s)| 2 ds+λ2·T(x) (3)

[0114] Each term in the formula is defined as follows:

[0115] λ1 and λ2: Dynamically adjusted weighting factors that reflect the importance of trajectory smoothness and processing efficiency, respectively. Recommended values ​​are λ1 = 0.6–1.0 and λ2 = 0.4–0.8.

[0116] T(x): The total processing time of the trajectory, calculated by dividing the trajectory length by the average processing speed.

[0117] κ(s): The curvature of the trajectory, calculated using the following formula:

[0118]

[0119] Where s is the arc length parameter of the trajectory, x′(s) and y′(s) are the first derivatives of the trajectory in the s direction, representing the rates of change of the x and y coordinates, respectively; x″(s) and y″(s) are the second derivatives, representing the changes in the acceleration of the trajectory. The derivatives can be calculated by the difference between the trajectory points.

[0120] Through the above steps, intelligent trajectory planning based on particle swarm optimization algorithm is realized, which effectively balances processing efficiency and trajectory smoothness, and ensures the feasibility and efficiency of trajectory planning results in complex processing scenarios.

[0121] Step S104: Perform deburring according to the optimal motion trajectory and collect cutting force data in real time.

[0122] In step S104, deburring is performed based on the optimal motion trajectory calculated using the particle swarm optimization algorithm, while cutting force data is collected in real time. This is the core step in translating intelligent planning into actual machining behavior. This step requires coordination between the CNC machine tool's actuators and the real-time monitoring system to ensure machining accuracy and stability.

[0123] First, the optimal motion trajectory output in step S103 is converted into an instruction set executable by the CNC machine tool. This requires formatting the trajectory data into G-code or other machine tool control languages ​​and matching it with the parameters of the machine tool controller. For example, the spatial coordinates of the trajectory points need to be converted into the machine tool's working coordinate system, and specific dynamic parameters such as feed rate and spindle speed are added to each point to ensure the continuity and stability of the machining process. During this process, the trajectory data also needs to be interpolated to generate a smooth motion path. Spline interpolation or linear interpolation can be used to ensure the consistency of the machine tool motion trajectory in both space and time.

[0124] In actual machining, the CNC machine tool's cutting tool begins its movement and performs cutting operations according to the aforementioned trajectory. The selection of cutting tools for deburring should be determined based on the workpiece material and burr characteristics. For example, carbide tools are used to handle metal burrs, or flexible grinding wheels are selected to handle burrs on plastics or composite materials. Furthermore, the cutting parameters of the tool need to be consistent with the dynamic parameters in the trajectory, including feed rate, spindle speed, and depth of cut. To ensure machining accuracy, the tool can be calibrated before machining, for example, using a laser tool setter or contact sensors to accurately measure the tool length and position.

[0125] Real-time acquisition of cutting force data is crucial for ensuring machining stability as the tool moves along its trajectory. This data acquisition can be achieved using force sensors mounted on the machine tool spindle or tool holder. These sensors capture three-dimensional force changes during the cutting process, including radial, tangential, and axial forces. The sensor sampling frequency should be high enough to capture rapid changes during machining. The cutting force data is then transmitted in real-time to the monitoring module via the machine tool's control system for processing and analysis.

[0126] To ensure the reliability and usability of cutting force data, preprocessing of the acquired data is necessary. For example, high-frequency noise can be removed using filtering algorithms; common filtering methods include low-pass filtering and Kalman filtering. Furthermore, real-time analysis of the cutting force signal can extract useful feature values, such as average cutting force, maximum cutting force, and fluctuation range. These feature values ​​reflect changes in tool load during machining, providing data support for subsequent adjustments.

[0127] During machining, real-time monitoring of cutting force changes can effectively reflect dynamic changes in machining conditions. For example, a sudden increase in cutting force may indicate that the tool has entered an area of ​​abnormal hardness or encountered an unidentified workpiece feature; a decrease in cutting force may indicate tool wear or loosening. The monitoring system must respond promptly to these changes, adjusting machining parameters through the controller to maintain machining stability.

[0128] Furthermore, to improve the adaptability of deburring, the machining path can be dynamically adjusted based on real-time cutting force data. For example, if the cutting force indicates that burrs in certain areas are difficult to machine, the tool feed rate can be appropriately reduced or the depth of cut increased to ensure machining quality. Simultaneously, historical data on cutting force changes can be recorded as a reference for optimizing the next machining path.

[0129] By combining the execution of the optimal trajectory with real-time monitoring of the cutting force, step S104 ensures the precise implementation of the machining behavior, which can achieve efficient deburring and ensure machining quality and equipment safety under dynamic conditions, providing reliable real-time feedback information for subsequent steps.

[0130] Furthermore, the step of performing deburring according to the optimal motion trajectory and collecting cutting force data in real time includes:

[0131] During deburring, a multi-axis force sensor mounted on the tool or spindle collects three-dimensional component data of the cutting force in real time, including tangential force, radial force and axial force.

[0132] By combining real-time acquired cutting force data with the machining characteristic curve of the workpiece material, it is determined whether the machining state is stable; when one or more components of the cutting force fluctuate beyond the preset threshold, the machining parameters, including feed rate, spindle speed and tool depth of cut, are dynamically adjusted, and the adjusted parameters are associated with and stored with the cutting force fluctuation data in real time.

[0133] By performing time-series analysis on the collected cutting force data, a historical curve of the machining state is generated, providing a reference for subsequent machining optimization and tool wear prediction.

[0134] During machining, multi-axis force sensors mounted on the tool or spindle collect real-time data on the three-dimensional components of the cutting force, including tangential force, radial force, and axial force. Tangential force primarily reflects the force on the tool along the cutting direction of the machining path and is an important indicator for evaluating cutting efficiency. Radial force represents the lateral force on the tool perpendicular to the cutting path plane and mainly affects the surface quality of the machined material. Axial force reflects the force on the tool in the depth direction and is directly related to the cutting depth. Real-time monitoring of these force components comprehensively reflects the interaction between the tool and the workpiece during deburring.

[0135] By combining real-time acquired cutting force data with the machining characteristic curve of the workpiece material, it is possible to determine whether the machining state is stable. The machining characteristic curve is predetermined based on the physical and mechanical properties of the workpiece material and is used to describe the theoretical cutting force distribution under different cutting parameters. When the real-time cutting force data deviates from the normal range of the characteristic curve, for example, when the fluctuation of one or more components exceeds a preset safety threshold, this usually indicates an abnormal machining state. Abnormalities may be caused by factors such as uneven burr distribution, tool wear, and mismatched machining parameters.

[0136] When abnormal fluctuations in cutting force are detected, the system dynamically adjusts machining parameters to restore stability. These adjustments include changing the feed rate, spindle speed, or depth of cut. For example, when the cutting force is too high, the feed rate or depth of cut can be reduced to decrease the load; when the cutting force is too low, the spindle speed can be increased to improve cutting efficiency. This dynamic adjustment not only adapts to changes in machining conditions in real time but also prevents tool damage or workpiece deformation due to excessive force. Simultaneously, all adjusted machining parameters and cutting force fluctuation data are recorded and stored in real time, providing reliable data support for subsequent machining optimization and historical state analysis.

[0137] Furthermore, by performing time-series analysis on the real-time acquired cutting force data, historical curves of the machining process can be generated. These historical curves reflect the dynamic trends of cutting forces during machining, not only helping operators intuitively understand the machining status but also serving as important evidence for optimizing machining parameters and predicting tool wear. For example, a gradual increase in cutting force may indicate accumulated tool wear, thus suggesting the need to replace the tool or adjust the machining strategy. Through this data-driven analysis method, the stability and quality of machining can be further improved.

[0138] By monitoring and dynamically adjusting cutting forces in real time, not only can the working stability of the tool be maintained during machining, but machining efficiency and workpiece quality can also be significantly improved. Its comprehensive historical data storage and analysis capabilities also provide a scientific basis for optimizing future machining tasks and equipment maintenance.

[0139] The following is a reference implementation of the particle swarm optimization algorithm:

[0140]

[0141]

[0142]

[0143]

[0144]

[0145] Step S105: When the cutting force exceeds the preset threshold, adjust the feed rate and spindle speed, and feed the adjusted parameters back to the particle swarm optimization algorithm to replan the trajectory.

[0146] In step S105, when the real-time collected cutting force data exceeds a preset threshold, the stability and efficiency of deburring are ensured by dynamically adjusting the machining parameters and re-optimizing the tool path. This process requires combining real-time monitoring, control algorithms, and optimization mechanisms to quickly respond to abnormal situations during machining, while improving the system's intelligence and adaptability.

[0147] First, for detecting cutting forces exceeding a preset threshold, a reasonable threshold range should be determined based on real-time acquired cutting force signals, combined with the force characteristics during processing and equipment capabilities. The cutting force threshold can be experimentally calibrated and dynamically set according to the processing requirements of different materials, burr heights, and distribution densities. For example, for harder metal workpieces, the threshold can be set higher, while for flexible or brittle materials, the threshold needs to be appropriately lowered. Furthermore, considering that the cutting force signal may be subject to instantaneous interference, the possibility of false alarms can be reduced through sliding window averaging or signal smoothing algorithms, ensuring that the threshold-exceeding signal truly reflects processing abnormalities.

[0148] Once the cutting force exceeds a preset threshold, the system needs to quickly adjust machining parameters to handle the abnormal situation. These adjusted parameters include feed rate, spindle speed, and depth of cut. The adjustment logic for these parameters needs to consider the workpiece's material properties and burr distribution. For example, when the cutting force is too high, the tool load can be reduced by decreasing the feed rate, while the spindle speed can be appropriately reduced to slow the increase in cutting force. If the burrs in a specific area are too numerous or have a high density, the depth of cut can be appropriately increased to remove the burrs in one pass, while avoiding repeated machining that would waste time. Parameter adjustments need to be sent to the CNC machine tool's control system in real time, ensuring the accuracy and stability of its execution.

[0149] While adjusting the parameters, the current cutting force data and the adjusted parameters need to be fed back to the particle swarm optimization algorithm to replan the tool's motion trajectory. The particle swarm optimization algorithm recalculates the optimal trajectory based on the new machining conditions, focusing on the balance between machining efficiency and path smoothness, while avoiding force overload or machining quality issues caused by the new trajectory. For example, for areas with high burr density, the optimization algorithm can adjust the trajectory coverage, increase the number of repeated machining operations, or change the direction of the tool contact path to ensure that burrs are completely removed. The fed-back cutting force data can serve as additional constraints for the optimization algorithm, preventing similar force overload problems from occurring in subsequent trajectory planning.

[0150] To further improve the system's response speed and stability, a multi-threaded parallel processing approach can be adopted, executing the processes of acquiring cutting force data, adjusting machining parameters, and re-optimizing the trajectory simultaneously, ensuring minimal machine tool downtime. Furthermore, to avoid system instability caused by frequent adjustments, an adaptive threshold range can be set. When the cutting force slightly exceeds the preset value but remains within a safe range, only minor adjustments are made without immediate trajectory re-optimization.

[0151] Step S105 enables intelligent handling of abnormal situations during processing, ensuring both high efficiency and quality while improving the system's adaptability to complex workpieces and dynamic environments. This real-time feedback and adjustment mechanism provides a reliable guarantee for the intelligence and stability of deburring processing.

[0152] Furthermore, the step of adjusting the feed rate and spindle speed when the cutting force exceeds a preset threshold, and feeding the adjusted parameters back to the particle swarm optimization algorithm to replan the trajectory, includes:

[0153] The specific workpiece areas where the cutting force exceeds the preset threshold are marked, and the objective function weight of the particle swarm optimization algorithm is dynamically adjusted based on the burr distribution density and processing difficulty of the marked areas, so that the priority of processing efficiency and path smoothness is redistributed according to the actual situation.

[0154] While adjusting the feed rate and spindle speed, regional characteristic parameters are introduced as optimization constraints. These regional characteristic parameters include the local hardness and thermal conductivity of the workpiece material to improve the accuracy of trajectory planning.

[0155] The replanned trajectory is optimized point by point through interpolation to ensure continuity and stability within the adjusted processing parameters.

[0156] When the cutting force in a certain workpiece area exceeds a preset threshold, the system first marks that area. The marking is based on the burr distribution density and processing difficulty. Burr distribution density is determined by the number of burrs per unit area, while processing difficulty depends on the burr height and material properties. After marking, this information is used to dynamically adjust the weights of the objective function of the particle swarm optimization algorithm. The objective function redistributes weights between processing efficiency and path smoothness. For example, when the burr distribution density in a certain area is high, the system prioritizes processing efficiency to complete the processing of complex areas first; while for areas with lower processing difficulty but requiring higher surface quality, the weight of path smoothness is increased. Through this dynamic weight adjustment, the particle swarm optimization algorithm can achieve customized trajectory planning for different processing areas.

[0157] During the adjustment of feed rate and spindle speed, the system further introduces regional characteristic parameters as optimization constraints. These parameters include the local hardness and thermal conductivity of the workpiece material. Hardness determines the mechanical requirements of cutting, while thermal conductivity affects the temperature rise of the workpiece and tool during cutting. By introducing these parameters, the impact of adjusted machining parameters on trajectory planning can be accurately assessed. For example, in high-hardness regions, the system may reduce the feed rate or increase the spindle speed to reduce cutting force fluctuations; in low-thermal-conductivity regions, the system may optimize the path to reduce the tool's dwell time in the same area, thereby reducing the impact of heat accumulation on machining quality. The introduction of regional characteristic parameters makes trajectory planning more consistent with actual machining conditions, improving the scientific validity and feasibility of the planning.

[0158] During the trajectory replanning process using the particle swarm optimization algorithm, the system performs point-by-point optimization interpolation on the planning results to ensure the smoothness and continuity of the trajectory. The purpose of interpolation is to increase the resolution and accuracy of the trajectory points, enabling the trajectory to remain stable within the adjusted machining parameter range. For example, for certain complex areas, interpolation can eliminate jumps between trajectory points, optimize curvature changes, and prevent sudden vibrations of the tool during movement. Furthermore, the interpolation results can effectively improve the accuracy of tool movement and the surface finish of the machined surface, ensuring the final machining quality.

[0159] Through the above steps, a dynamic combination of machining parameter adjustment and trajectory optimization is achieved under abnormal cutting force conditions. The system can not only respond quickly to changes in machining status, but also comprehensively consider burr characteristics, material properties, and trajectory planning requirements, enabling the machining process to achieve an optimal balance between efficiency, quality, and stability.

[0160] Furthermore, the step of adjusting the feed rate and spindle speed when the cutting force exceeds a preset threshold, and feeding the adjusted parameters back to the particle swarm optimization algorithm to replan the trajectory, also includes:

[0161] During the cutting force adjustment process, the vibration spectrum data of the contact area between the tool and the workpiece is collected in real time to determine whether there is a mechanical resonance phenomenon.

[0162] If resonance is detected, the feed rate is reduced first and the tool depth of cut is adjusted appropriately. At the same time, a vibration constraint factor is introduced into the particle swarm optimization algorithm to avoid repeated machining of the resonance area caused by subsequent trajectory planning.

[0163] When the cutting force exceeds a preset threshold, the system not only adjusts the feed rate and spindle speed, but also collects vibration spectrum data of the contact area in real time through vibration sensors mounted on the tool or workpiece surface. This data describes the vibration characteristics between the tool and workpiece in the frequency domain, including amplitude and frequency information. By analyzing the vibration spectrum, the system can determine whether mechanical resonance exists. Mechanical resonance typically manifests as a sharp increase in vibration amplitude within a specific frequency range. It may be caused by the natural frequency of the tool or workpiece structure matching the frequency of the external force applied during machining, leading to machining instability, accelerated tool wear, or a decrease in workpiece surface quality.

[0164] If mechanical resonance is detected, the system prioritizes reducing the excitation frequency generated during machining by decreasing the feed rate and adjusting the tool depth of cut, thereby avoiding coupling with the structure's natural frequency. Simultaneously, the spindle speed can be fine-tuned according to the resonance frequency to further mitigate vibration. This adjustment not only effectively reduces vibration amplitude but also reduces the potential harm of resonance to tool life and workpiece quality.

[0165] Furthermore, to prevent repeated processing of resonant regions during subsequent trajectory planning, the system introduces a vibration constraint factor into the particle swarm optimization algorithm. This factor, calculated based on vibration spectrum data, directly affects the fitness evaluation of particles. For example, for regions where resonance is detected, the vibration constraint factor lowers the priority of that region, prompting the particle swarm optimization algorithm to avoid that region or reduce the number of times it processes it when planning the trajectory. In this way, the trajectory optimization results can proactively avoid the risk of resonance, further improving the stability of the processing.

[0166] This embodiment achieves real-time control of mechanical resonance by combining vibration spectrum monitoring with machining parameter adjustment. Simultaneously, by introducing vibration constraints into the trajectory optimization process, repetitive machining in the resonance region is fundamentally reduced, ensuring the rationality and stability of the machining trajectory. This method is particularly suitable for deburring complex workpieces, significantly improving machining efficiency and quality, extending tool life, and reducing the risk of machining errors caused by vibration.

[0167] Furthermore, the aforementioned CNC machine tool deburring control method based on intelligent algorithms also includes:

[0168] During the machining process, the tool wear status is determined based on the real-time changes in cutting force signal and spindle current; when the tool wear exceeds the preset threshold, the process switches to the backup optimized trajectory to continue machining.

[0169] The surface temperature and vibration data of the workpiece are collected in real time, and the processing parameters are dynamically adjusted based on the temperature and vibration data to ensure processing quality and efficiency until the deburring process is completed.

[0170] During machining, real-time monitoring of cutting force signals and spindle current changes is fundamental to determining tool wear status. Cutting force signals directly reflect the physical force between the tool and the workpiece, while spindle current changes indirectly reflect the changing trend of tool load. By combining these two signals, the degree of tool wear can be captured more accurately. In practice, cutting force signals can be acquired using high-precision force sensors, while spindle current is obtained through the current detection module of the machine tool controller. Both data must be acquired synchronously to ensure real-time analysis and accuracy. When determining tool wear status, various algorithms can be used, such as analyzing the amplitude changes, frequency characteristics, and trends of cutting force and spindle current signals, combined with preset wear models or thresholds, to determine the current tool usage status.

[0171] When tool wear exceeds a preset threshold, the system automatically switches to a backup optimized trajectory to continue machining. This process aims to maximize machining continuity and avoid production interruptions caused by tool changes or downtime adjustments. When generating the backup trajectory, the characteristics of the remaining burr area on the current workpiece are considered, and the trajectory is recalculated using particle swarm optimization or other intelligent planning methods. This ensures that the new trajectory bypasses the area affected by the worn tool while still meeting machining efficiency and quality requirements. Backup trajectories can be pre-calculated and stored in the system at the start of machining, or generated in real-time upon wear detection. To improve machining flexibility, multiple backup trajectories can be designed and dynamically switched for different wear conditions.

[0172] Furthermore, to further improve machining quality and adaptability, real-time acquisition of workpiece surface temperature and vibration data is a crucial step. Changes in workpiece surface temperature reflect the accumulation of cutting heat, directly affecting the material properties of the machining area and tool life. Vibration data is a key indicator of dynamic stability during machining; excessive vibration can lead to increased machining errors and rapid tool wear. In practice, temperature data can be acquired using non-contact infrared sensors, and vibration data can be monitored in real time using accelerometers. These data need to undergo filtering and feature extraction processing to remove environmental noise and extract key change information.

[0173] Based on the collected temperature and vibration data, the system optimizes the machining process by adaptively adjusting machining parameters. For example, when the temperature is too high, the feed rate or spindle speed can be reduced to decrease heat accumulation; when the vibration amplitude exceeds a threshold, the toolpath, depth of cut, or a spare tool can be adjusted to restore machining stability. Furthermore, predictive control strategies can be employed, using machine learning models or empirical rules to anticipate potential anomalies and make preventative adjustments.

[0174] This method achieves intelligent deburring through comprehensive monitoring and dynamic adjustment of tool wear, workpiece temperature, and vibration. Its advantages include improved machining quality, enhanced system adaptability and robustness, extended tool life, and reduced downtime losses.

[0175] A second embodiment of this application provides an electronic device, the electronic device comprising:

[0176] processor;

[0177] The memory is used to store a program, which, when read and executed by the processor, executes a deburring control method for CNC machine tools based on intelligent algorithms provided in the first embodiment of this application.

[0178] The third embodiment of this application provides a computer-readable storage medium storing a computer program thereon. When the program is executed by a processor, it executes a deburring control method for CNC machine tools based on intelligent algorithms provided in the first embodiment of this application.

[0179] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A deburring control method for CNC machine tools based on intelligent algorithms, characterized in that, include: A depth camera and a 2D vision sensor are used to scan the burr area of ​​the workpiece to obtain the 3D point cloud data and surface texture information of the burr. Using the three-dimensional point cloud data and combined with surface texture information, the location coordinates, height value, and distribution density of the burrs are calculated; Based on the position coordinates, height value, and distribution density of the burrs, the optimal motion trajectory of the tool is calculated using a particle swarm optimization algorithm, with processing efficiency and path smoothness as optimization objectives. Deburring is performed according to the optimal motion trajectory, and cutting force data is collected in real time; When the cutting force exceeds the preset threshold, the feed rate and spindle speed are adjusted, and the adjusted parameters are fed back to the particle swarm optimization algorithm to replan the trajectory.

2. The deburring control method for CNC machine tools based on intelligent algorithms according to claim 1, characterized in that, Also includes: During the machining process, the tool wear condition is determined based on the real-time changes in cutting force signals and spindle current. When tool wear exceeds a preset threshold, switch to a backup optimized trajectory to continue machining. The surface temperature and vibration data of the workpiece are collected in real time, and the processing parameters are dynamically adjusted based on the temperature and vibration data to ensure processing quality and efficiency until the deburring process is completed.

3. The deburring control method for CNC machine tools based on intelligent algorithms according to claim 1, characterized in that, The step of scanning the burr area of ​​the workpiece using a depth camera and a 2D vision sensor to acquire the 3D point cloud data and surface texture information of the burr includes: During the scanning process, the exposure time of the depth camera and the intensity of the light source are adjusted to dynamically adapt to the material properties and light reflection characteristics of the workpiece surface; Based on real-time quality assessment of the scan data, it is determined whether there is missing data or noise interference. When missing data is detected, the scanning angle is adjusted and a supplementary scan is performed. When noise interference is detected, a filtering algorithm based on spatial neighborhood and texture consistency is applied to optimize the data, thereby obtaining complete and accurate 3D point cloud data of burrs and surface texture information.

4. The deburring control method for CNC machine tools based on intelligent algorithms according to claim 1, characterized in that, The step of calculating the location coordinates, height value, and distribution density of burrs using the three-dimensional point cloud data and combined with surface texture information includes: The 3D point cloud data is initially processed by denoising filtering and voxelization to simplify the data volume, while retaining the main geometric features of the burrs. Based on point cloud normal vector calculation and curvature analysis, burr regions are identified, and the three-dimensional spatial coordinates of each burr point are extracted as the position coordinates of the burr. By calculating the height of the burr point, the difference between its Z-axis coordinate and the workpiece reference plane is calculated. The least squares method is used to fit the reference plane and ensure the accuracy of the calculation. The burr distribution density is calculated based on the point cloud density per unit volume within the burr region, combined with the feature distribution of texture information.

5. The deburring control method for CNC machine tools based on intelligent algorithms according to claim 1, characterized in that, The optimal tool trajectory is calculated using a particle swarm optimization algorithm based on the position coordinates, height, and distribution density of the burrs, with machining efficiency and path smoothness as optimization objectives, including: The geometric feature data of the burr area is partitioned, and priority processing area and secondary processing area are divided according to the difference in burr height value and distribution density. In the initial stage of the particle swarm optimization algorithm, the trajectory points of the priority processing region are set as high-priority particles and given higher objective function weights. By dynamically updating the fitness value of particles and combining the constraints of tool turning radius and machining path continuity, the movement direction and speed of particles are adjusted in real time. During the iteration process, the trajectory points in the priority processing area are finely interpolated to ensure the smoothness of the trajectory and the processing quality.

6. The deburring control method for CNC machine tools based on intelligent algorithms according to claim 1, characterized in that, The process of performing deburring according to the optimal motion trajectory and collecting cutting force data in real time includes: During deburring, a multi-axis force sensor mounted on the tool or spindle collects three-dimensional component data of the cutting force in real time, including tangential force, radial force and axial force. By combining real-time acquired cutting force data with the machining characteristic curve of the workpiece material, it is determined whether the machining state is stable; when one or more components of the cutting force fluctuate beyond the preset threshold, the machining parameters, including feed rate, spindle speed and tool depth of cut, are dynamically adjusted, and the adjusted parameters are associated with and stored with the cutting force fluctuation data in real time. By performing time-series analysis on the collected cutting force data, a historical curve of the machining state is generated, providing a reference for subsequent machining optimization and tool wear prediction.

7. The deburring control method for CNC machine tools based on intelligent algorithms according to claim 1, characterized in that, When the cutting force exceeds a preset threshold, the feed rate and spindle speed are adjusted, and the adjusted parameters are fed back to the particle swarm optimization algorithm to replan the trajectory, including: The specific workpiece areas where the cutting force exceeds the preset threshold are marked, and the objective function weight of the particle swarm optimization algorithm is dynamically adjusted based on the burr distribution density and processing difficulty of the marked areas, so that the priority of processing efficiency and path smoothness is redistributed according to the actual situation. While adjusting the feed rate and spindle speed, regional characteristic parameters are introduced as optimization constraints. These regional characteristic parameters include the local hardness and thermal conductivity of the workpiece material to improve the accuracy of trajectory planning. The replanned trajectory is optimized point by point through interpolation to ensure continuity and stability within the adjusted processing parameters.

8. The deburring control method for CNC machine tools based on intelligent algorithms according to claim 5, characterized in that, The calculation of the optimal tool trajectory using a particle swarm optimization algorithm based on the burr's position coordinates, height value, and distribution density also includes: During the optimization process, the fitness function of the particle swarm is dynamically adjusted by combining the temperature change data of the processing area. When the temperature is detected to be too high, the processing load in the heat accumulation area is reduced by increasing the path interval and reducing the trajectory curvature priority.

9. The deburring control method for CNC machine tools based on intelligent algorithms according to claim 7, characterized in that, The step of adjusting the feed rate and spindle speed when the cutting force exceeds a preset threshold, and feeding the adjusted parameters back to the particle swarm optimization algorithm to replan the trajectory, also includes: During the cutting force adjustment process, the vibration spectrum data of the contact area between the tool and the workpiece is collected in real time to determine whether there is a mechanical resonance phenomenon. If resonance is detected, the feed rate is reduced first and the tool depth of cut is adjusted appropriately. At the same time, a vibration constraint factor is introduced into the particle swarm optimization algorithm to avoid repeated machining of the resonance area caused by subsequent trajectory planning.

10. The deburring control method for CNC machine tools based on intelligent algorithms according to claim 8, characterized in that, The implementation steps of the particle swarm optimization algorithm include: Based on the location coordinates, height value, and distribution density of burrs, the processing area is dynamically prioritized, where the priority P(A) of the processing area is calculated using the following formula 1: Where A is the processing area, representing the unit area into which the workpiece surface is divided; α is the average height of burr points within processing area A, used to characterize the overall protrusion of burrs within the area; D(A) is the number of burr points per unit area within processing area A, used to reflect the density of burrs within the area; G(A) represents the shortest distance from processing area A to the workpiece boundary, used to assess the processing risk of the processing area; α, β, and γ are priority weighting factors. After priority calculation is completed, the geometric center of the highest priority machining area and its nearby random perturbation points are used as the initial positions of the particle swarm. Each particle represents a possible tool movement path. The initial velocity of each particle is set to a random small range value. The particle's velocity is updated in each iteration using the following formula 2: in, Let be the velocity of particle i in the (t+1)th iteration; Let i be the velocity of particle i in the t-th iteration; Let be the position of particle i in the t-th iteration; w(t) is the inertia weight; c1 is the individual learning factor; r1 is a uniformly distributed random number in the range [0,1]; p i c1 is the historical best position of particle i, representing the coordinates of the optimal solution found by the particle in all iterations; c2 is the social learning factor; r2 is a uniformly distributed random number in the range [0,1]; g is the historical best position of the swarm, representing the coordinates of the global optimal solution found by the particle in the iteration process of all particles in the swarm. The fitness of a particle is determined by an objective function, which is expressed as Equation 3 below: F(x)=λ1·∫|κ(s)| 2 ds+λ2·T(x) (3) Where λ1 and λ2 are weighting factors; T(x) is the total processing time of the trajectory; and к(s) is the trajectory curvature, calculated according to the following formula 4: Where s is the arc length parameter of the trajectory; x′(s) is the first derivative of the x-coordinate of the trajectory in the direction of arc length s; x″(s) is the second derivative of the x-coordinate of the trajectory in the direction of arc length s; y′(s) is the first derivative of the y-coordinate of the trajectory in the direction of arc length s; and y″(s) is the second derivative of the y-coordinate of the trajectory in the direction of arc length s.

Images