Machine learning based mobile phone frame cnc processing monitoring method and system

Abstract

The application relates to the technical field of numerical control machining and intelligent manufacturing, and discloses a mobile phone frame CNC machining monitoring method and system based on machine learning. The method comprises the following steps: obtaining three-dimensional geometric data of a bending area of a mobile phone frame and establishing a geometric model, extracting curvature radius distribution characteristics to match a tool path template, determining a contact point sequence and adaptively adjusting a tool posture and a vibration frequency. Sensor signals are collected in real time during the machining process, deformation correction vectors are generated and classified, and the tool posture is optimized in combination with force feedback. The optimized algorithm is used to adjust the contact point and the vibration amplitude, the posture parameter group is generated, the numerical control system path is updated, the final tool posture sequence is iteratively corrected, high-precision machining control is realized, and the precision, surface finish and stability of the machining process of the bending area of the mobile phone frame are improved.

Description

Technical Field

[0001] This application relates to the fields of CNC machining and intelligent manufacturing technology, and in particular to a method and system for monitoring CNC machining of mobile phone frames based on machine learning. Background Technology

[0002] Mobile phone frame machining falls under the field of precision manufacturing and is a crucial aspect of the design and structural performance of modern smartphones. As consumers' demands for appearance, feel, and quality continue to rise, mobile phone frames commonly employ complex curved surfaces and structures to enhance the grip and visual appeal. However, existing CNC machining methods primarily rely on fixed tool paths and preset parameters, making it difficult to effectively adapt to rapid changes in the radius of curvature in curved areas. At points of abrupt curvature changes, uneven force distribution between the tool and material can easily lead to machining errors, surface scratches, or insufficient smoothness. Simultaneously, the material often undergoes slight deformation during cutting due to stress and temperature, further complicating precision control. Current technologies, when handling such complex surfaces, lack real-time adaptation to curvature changes and dynamic compensation for material deformation, resulting in insufficient machining stability and consistency, thus limiting the quality and production efficiency of high-end smartphone frames.

[0003] Therefore, providing a machine learning-based CNC machining monitoring method and system for mobile phone frames to achieve adaptive adjustment of curvature changes, real-time compensation for material deformation, and improve machining accuracy, surface finish and overall production efficiency has become an urgent technical problem to be solved. Summary of the Invention

[0004] This application provides a machine learning-based CNC machining monitoring method and system for mobile phone frames, which can improve the machining accuracy, surface finish, and stability of the machining process in the curved areas of mobile phone frames.

[0005] Firstly, this application provides a machine learning-based method for monitoring CNC machining of mobile phone bezels, the machine learning-based method for monitoring CNC machining of mobile phone bezels includes:

[0006] Obtain the three-dimensional geometric data of the curved area of ​​the mobile phone frame, generate the frame geometric model, and determine the distribution characteristics of the frame curvature radius;

[0007] Based on the curvature radius distribution characteristics of the border, the tool path template is matched, the tool contact point sequence is determined, and the tool posture and vibration frequency are adjusted to adapt to the curvature change.

[0008] Acquire sensor signals during the processing, determine the degree of material deformation based on these signals, and generate a deformation correction vector;

[0009] The deformation correction vector is classified to determine the deformation type, and the tool force feedback and attitude are adjusted based on the deformation type to generate a deformation feature set;

[0010] Based on the deformation feature set, the tool attitude parameters and contact point sequence are optimized, and the vibration amplitude of the tool is adjusted to generate an optimized attitude parameter set.

[0011] An attitude adjustment command is generated based on the optimized attitude parameter set, and the machining accuracy is verified in conjunction with sensor feedback. If the accuracy does not meet the requirements, deformation data is generated iteratively.

[0012] The tool posture sequence is refined based on the iterated deformation data to generate the final tool posture sequence, and machining control is executed.

[0013] Secondly, this application provides a machine learning-based CNC machining monitoring system for mobile phone bezels, the machine learning-based CNC machining monitoring system for mobile phone bezels comprising:

[0014] The data acquisition module is used to acquire three-dimensional geometric data of the curved area of ​​the mobile phone frame, generate a geometric model of the frame, and determine the distribution characteristics of the frame curvature radius.

[0015] The path matching module is used to match the tool path template according to the distribution characteristics of the radius of curvature of the border, determine the sequence of tool contact points, and adjust the tool posture and vibration frequency to adapt to the curvature change.

[0016] The correction generation module is used to acquire sensor signals during the processing, determine the degree of material deformation, and generate a deformation correction vector.

[0017] The classification feature module is used to classify the deformation correction vector, determine the deformation type, and adjust the tool force feedback and attitude based on the deformation type to generate a deformation feature set.

[0018] The attitude optimization module is used to optimize the tool attitude parameters and contact point sequence based on the deformation feature set, and adjust the vibration amplitude of the tool to generate an optimized attitude parameter set.

[0019] The accuracy verification module is used to generate attitude adjustment instructions based on the optimized attitude parameter set, and verify the processing accuracy in combination with sensor feedback. If the accuracy does not meet the requirements, deformation data is generated iteratively.

[0020] The control execution module is used to refine the tool posture sequence based on the iterated deformation data, generate the final tool posture sequence, and execute machining control.

[0021] The technical solution provided in this application acquires three-dimensional geometric data of the curved area of ​​the mobile phone frame and establishes a complete geometric model. Combined with the curvature radius distribution characteristics, tool path matching and attitude adjustment are performed, enabling adaptive machining of complex curved surfaces. By acquiring force feedback and temperature signals in real time, deformation correction vectors are generated and classified, which can effectively compensate for the stress and thermal deformation generated during machining. Furthermore, machine learning and optimization algorithms are used to dynamically optimize tool attitude parameters and vibration control, improving machining accuracy and surface finish, extending tool life, and increasing overall production efficiency. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart of the machine learning-based CNC machining monitoring method for mobile phone bezels according to this application;

[0024] Figure 2 This is an analysis diagram of the deformation monitoring and control effect of CNC machining of mobile phone frame based on machine learning in this application;

[0025] Figure 3 This is a flowchart of the adaptive control of tool posture and vibration frequency based on curvature variation in this application.

[0026] Figure 4 This is a schematic diagram of the structure of the machine learning-based mobile phone frame CNC machining monitoring system of this application. Detailed Implementation

[0027] This application provides a machine learning-based CNC machining monitoring method and system for mobile phone bezels. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0028] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the CNC machining monitoring method for mobile phone bezels based on machine learning in this application includes:

[0029] Step S1: Obtain the three-dimensional geometric data of the curved area of ​​the mobile phone frame, generate the frame geometric model, and determine the distribution characteristics of the frame curvature radius.

[0030] Specifically, in the technical solution of this application, data is collected from the curved area of ​​the mobile phone frame using a high-precision 3D scanning device. The 3D geometric data can be expressed in the form of point clouds, and its data density and accuracy directly determine the completeness and accuracy of the subsequent geometric model. The collected point cloud data undergoes surface reconstruction and triangulation processing, which generates a continuous frame surface model from discrete spatial points. In this process, filtering and smoothing algorithms can be combined to eliminate acquisition noise and improve surface smoothness, thereby obtaining a geometric model that meets the accuracy requirements. After obtaining the complete geometric model, the system calculates the local curvature of each point on the model surface. Curvature is an important geometric parameter reflecting the degree of surface curvature, and its reciprocal is the radius of curvature. By continuously calculating and sorting the radius of curvature in different regions, a radius of curvature distribution feature can be formed. This distribution feature can reveal the geometric change law of the frame in straight segments, transition segments, and highly curved segments. Based on this radius of curvature distribution feature, the system can not only identify curvature gradient areas and abrupt change points, but also extract important feature points representing the geometric shape, providing a key basis for subsequent toolpath template matching and attitude adjustment.

[0031] Step S2: Based on the distribution characteristics of the radius of curvature of the border, match the tool path template, determine the sequence of tool contact points, and adjust the tool posture and vibration frequency to adapt to the curvature change.

[0032] Specifically, after establishing the geometric model of the phone frame and extracting the curvature radius distribution features, the system compares these features with a pre-defined tool path template library. By calculating the similarity, the system selects the path template that best matches the actual curvature change, and then generates a preliminary contact point sequence for the tool on the complex curved surface of the frame. This contact point sequence includes not only the spatial coordinates of the tool and the workpiece surface but also geometric vectors of the tangential and normal directions, used to accurately describe the contact state between the tool and the material. By analyzing these vectors, the appropriate angles for the tool at each contact point can be calculated, including tilt and rotation angles, ensuring that the cutting edge of the tool maintains a reasonable contact relationship with the curved surface and avoiding overcutting or undercutting due to angle deviations. Simultaneously, by considering the gradual change in curvature radius distribution, the system can adjust the tool's vibration frequency in real time, maintaining high-frequency vibration in small curvature radius regions to reduce cutting resistance, while using a relatively lower frequency in large curvature radius regions to stabilize the machining process. Through the synergistic effect of path matching, attitude angle correction, and vibration frequency adjustment, the cutting tool can adaptively follow the continuous change of the edge curvature, thereby improving surface finish and enhancing machining accuracy.

[0033] Step S3: Acquire sensor signals during the processing, determine the degree of material deformation based on these signals, and generate a deformation correction vector.

[0034] Specifically, during the machining of the curved area of ​​the mobile phone frame, the system collects machining status information in real time through various sensors installed on the CNC machine tool. Force sensors acquire contact force signals between the tool and the workpiece, while temperature sensors monitor the heat distribution in the cutting area. These signals, after filtering and preprocessing, are converted into continuous force curves and temperature field distributions to reveal the mechanical and thermal stresses experienced by the material during machining. When the material undergoes slight deformation due to uneven stress or localized temperature rise, its geometric accuracy is affected. Therefore, the system needs to construct a thermo-mechanical coupling model combining force and temperature data, calculating the displacement and strain fields to determine the degree of deformation. If the deformation value exceeds a preset threshold, the system activates a compensation mechanism, using a reverse correction algorithm to generate a deformation correction vector. This vector includes not only the correction direction but also the correction magnitude, guiding the dynamic adjustment of the tool path. To ensure data availability and stability, the generated correction vector is also normalized and filtered to convert it into standardized input data.

[0035] Step S4: Classify the deformation correction vector to determine the deformation type, and adjust the tool force feedback and attitude based on the deformation type to generate a deformation feature set.

[0036] Specifically, after generating a standardized deformation correction vector, the system inputs this vector into the classification processing module for identification and analysis. This module preferably employs machine learning-based classification algorithms, such as Support Vector Machines (SVM) or Random Forests. By training on historical machining data and typical deformation samples, it can establish effective decision boundaries in a high-dimensional feature space, distinguishing different types of deformation. The main goal of classification is to determine whether the deformation is dominated by local curvature changes or caused by overall contour offset. The former typically manifests as local deviations in areas of abrupt changes in curvature radius, while the latter manifests as macroscopic displacement of the overall shape of the border. When the deformation is determined to be dominated by curvature changes, the system adjusts the cutting force parameters of the tool accordingly, making the cutting process smoother to reduce stress concentration; when the deformation is due to overall contour offset, the geometric accuracy of the border is restored by correcting the tool path offset. Based on this, the system combines the classification results with the corresponding adjustment strategy to generate a deformation feature set, which includes deformation type labels, adjustment coefficients, correction parameters, and related posture compensation information. This feature set not only serves as input for subsequent attitude optimization, but also provides real-time feedback to the CNC system, enabling dynamic perception and adaptive control of deformation in complex surface machining.

[0037] Step S5: Based on the classified deformation feature set, optimize the tool attitude parameters and contact point sequence, and adjust the vibration amplitude of the tool to generate an optimized attitude parameter set.

[0038] Specifically, after obtaining the classified deformation feature set, the system uses this feature set as input to the optimization algorithm to dynamically optimize the tool attitude parameters and contact point sequence. Specifically, the deformation feature set includes deformation type, correction coefficients, path offset, and attitude compensation information. This data accurately reflects the distribution characteristics of geometric errors and material deformation during machining. Based on these inputs, the system constructs an objective function for attitude optimization, which typically uses minimizing tool attitude deviation, reducing cutting force unevenness, and improving surface finish as core evaluation indicators. To achieve global optimization, the system employs a genetic algorithm or other swarm intelligence algorithm, iteratively converging to obtain the optimal solution through population initialization, crossover, and mutation operations. During the iteration process, the contact point sequence is corrected according to fine-tuning parameters, making the tool path more closely conform to the curvature variation of the frame, thereby reducing local stress concentration and machining errors. Simultaneously, the system combines real-time feedback information from the deformation feature set to adapt and adjust the tool vibration amplitude, maintaining small-amplitude, high-frequency vibration in high-curvature regions to improve surface finish, while using larger amplitude vibrations in low-curvature regions to increase material removal rate. The system outputs the optimization results as a complete set of attitude parameters, which includes attitude angle and position correction values ​​as well as vibration control strategies, providing stable data support for subsequent accuracy verification and dynamic control.

[0039] Step S6: Generate attitude adjustment instructions based on the optimized attitude parameter set, and verify the machining accuracy in conjunction with sensor feedback. If the accuracy does not meet the requirements, generate deformation data iteratively.

[0040] Specifically, after completing attitude optimization, the system encapsulates the optimized attitude angle, contact point correction, and vibration control parameters into attitude adjustment commands, which are then sent to the CNC controller via the bus to trigger dynamic path updates. During execution, the machine tool simultaneously collects multimodal feedback such as force, displacement, temperature, and vibration, calculates accuracy indicators such as dimensional deviation, contour error, and surface roughness, and compares them with preset thresholds (such as linear dimensions, tolerance band, and Ra target). If any indicator fails to meet the requirements, the system immediately increases the sampling frequency and introduces supplementary features (such as transient peak values, frequency domain energy, and temperature gradient). Based on this, it re-estimates the thermo-mechanical coupled displacement field, iteratively generates a new deformation correction vector, updates the deformation feature set and objective function, re-solves the attitude parameter set, and sends out new commands until the accuracy meets the target.

[0041] Step S7: Refine the tool posture sequence based on the iterated deformation data, generate the final tool posture sequence, and execute machining control.

[0042] Specifically, after obtaining a new round of deformation data, the system first fuses and weights the data with historical attitude parameter sets, generates a continuous, jitter-free tool attitude trajectory through Kalman / spline smoothing, and applies feed, acceleration, and jerk limits under machine tool dynamics constraints. Subsequently, based on the radius of curvature and residual error distribution, the system performs intensive sampling and fine-tuning of key contact points, enabling the attitude to maintain small-angle, high-frequency fine-tuning in the high-curvature region and steady-state cutting in the low-curvature region. The constraint-optimized attitude sequence then undergoes post-processing (including coordinate system mapping, tool length / radius compensation, and vibration parameter embedding) to generate an executable NC command stream, while configuring safety interlocks and abnormal rollback strategies. The CNC system implements closed-loop machining control based on this, and collects force-temperature-vibration feedback to check trajectory deviation and surface quality in real time. It also writes the residual deviation back to the posture sequence cache, forming an online self-learning and iterative convergence mechanism. When each accuracy index reaches the threshold, the sequence is frozen as the "final tool posture sequence" for the current batch execution and subsequent reuse of the same type of parts. This ensures machining stability while achieving high precision, high surface finish and high consistency machining of curved areas.

[0043] It is understood that the executing entity of this application can be a machine learning-based mobile phone frame CNC machining monitoring system, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiments use a server as an example for illustration.

[0044] In one specific embodiment, the process of performing step S1 may specifically include the following steps:

[0045] Three-dimensional geometric data of a specific curved area on the phone's frame are collected using a scanning device;

[0046] The surface reconstruction method is used to process the three-dimensional geometric data to generate the bounding box geometric model;

[0047] Calculate the radius of curvature of each region of the border based on the border geometry model, and generate a gradual distribution feature of the radius of curvature.

[0048] Analyze the gradual distribution characteristics of the radius of curvature to extract key feature points of the frame geometry;

[0049] Based on key feature points, the principal curvature is calculated and converted into curvature radius. A curvature radius sequence is generated according to the processing path order. An interpolation method is used to form a continuously changing curve to determine the curvature radius distribution characteristics of the border.

[0050] Specifically, in one embodiment, the implementation process of this step can be illustrated using the curved area of ​​an aluminum alloy mobile phone frame as an example. For instance, a laser scanner is used to collect point clouds at the curved area of ​​the frame, achieving a point cloud density of 120 sampling points per square millimeter, thus ensuring the accuracy of capturing minute curvature changes. A triangular mesh reconstruction method is used to convert the point cloud data into a continuous surface, and the surface noise is iteratively corrected using a Laplacian smoothing algorithm, ensuring that the surface deviation of the generated geometric model is controlled within 0.03 millimeters. Next, the principal curvature values ​​are calculated in each region of the geometric model and converted into curvature radii using a formula. For example, a principal curvature of 0.2 is detected at the transition section, corresponding to a curvature radius of 5 millimeters, while in a smoother region with a principal curvature of 0.05, the curvature radius is 20 millimeters. Arranging these curvature radii according to the processing path sequence yields a curvature sequence gradually decreasing from 20 millimeters to 5 millimeters, which is then fitted into a continuous curvature change curve using cubic spline interpolation. By analyzing the curve, locations of abrupt curvature changes can be identified and marked as key feature points. For example, at a point 2 mm from the corner of the edge, the radius of curvature drops sharply from 12 mm to 6 mm. This point serves as a crucial basis for subsequent toolpath matching and attitude adjustment. Ultimately, the system stores the resulting radius of curvature distribution features in a database to support subsequent path planning and machining control.

[0051] In one specific embodiment, the process of performing step S2 may specifically include the following steps:

[0052] Extract an initial path template from the preset toolpath library that matches the distribution features of the curvature radius of the border;

[0053] Based on the initial path template, generate the initial contact point sequence of the tool on the complex geometry of the border;

[0054] For the initial contact point sequence, calculate the tangential and normal vectors of the tool at each contact point;

[0055] The tangential and normal vectors are optimized using a particle swarm optimization algorithm to generate a contact point control sequence for attitude angle adjustment.

[0056] A genetic algorithm is used to optimize the distribution characteristics of the radius of curvature of the border, obtain the vibration frequency control parameters of the tool, and determine the initial vibration mode of the tool based on the vibration frequency control parameters.

[0057] The contact point control sequence is simulated and verified to determine whether it adapts to curvature changes. If it does not adapt, the path template is adjusted and the contact point sequence is regenerated.

[0058] Specifically, a preliminary contact point sequence is generated on the complex geometry of the border based on the initial path template. Then, the tangential and normal vectors of the tool at each contact point are calculated based on this contact point sequence. On this basis, the particle swarm optimization algorithm is used to optimize the tangential and normal vectors to obtain the contact point control sequence after attitude angle adjustment. Furthermore, a genetic algorithm is used to optimize the curvature radius distribution characteristics of the border to obtain the vibration frequency control parameters of the tool, and the initial vibration mode of the tool is determined based on these parameters. Finally, the contact point control sequence is simulated and verified to determine whether it can adapt to curvature changes. If the adaptability requirements are not met, the path template is adjusted and the contact point sequence is regenerated to ensure that the tool path matches the geometric features of the border.

[0059] Taking the processing of titanium alloy mobile phone frames as an example, high-precision 3D scanning equipment is used to acquire point cloud data of the curved area of ​​the frame. The point cloud density is controlled at 100 points per square millimeter. A complete geometric model is generated using surface reconstruction. The radius of curvature of each region is calculated and a gradual distribution feature is formed. The initial path template with the highest similarity is selected from the template library and mapped onto the geometric model to generate approximately 500 contact points. For each contact point, the tangential vector and normal vector are calculated and input into a particle swarm optimization algorithm for attitude angle correction, resulting in an optimized contact point control sequence. This sequence is then combined with a genetic algorithm to analyze the radius of curvature distribution feature. The process involves optimization to output vibration frequency control parameters. The frequency is set to 2000Hz for the small curvature radius region, 1500Hz for the medium region, and 1000Hz for the large curvature radius region. Simultaneously, the initial vibration mode is determined, and the simulation system is imported to verify the adaptability of the contact point control sequence. It is determined whether the deviation in the transition zone where the curvature changes from 12 mm to 4 mm is less than 0.1 mm. If the deviation exceeds the threshold, the path template is adjusted and the contact point sequence is regenerated. Through iterative optimization, a contact point control sequence and vibration parameters that meet the machining accuracy requirements are finally obtained, providing stable input for subsequent CNC execution.

[0060] In one specific embodiment, the process of performing step S3 may specifically include the following steps:

[0061] Force sensors installed on CNC machine tools are used to collect force feedback signals between the cutting tool and the workpiece in real time. At the same time, temperature data of the workpiece area is obtained through infrared temperature sensors.

[0062] Based on the collected force feedback signals, a finite element analysis model is established, the force feedback signals are mapped onto the mesh of the processing area, the stress distribution of each element is calculated, and the material stress distribution is determined.

[0063] By using temperature data and interpolation algorithms, the thermal distribution characteristics of the processing area are calculated to identify high-temperature areas and temperature gradients, thereby predicting thermal stress and possible material deformation.

[0064] Combining the stress and heat distribution characteristics of the material, a thermo-mechanical coupling model is used to calculate the degree of deformation of the material during processing, quantify the degree of deformation into deformation amount, and compare it with a preset threshold.

[0065] If the deformation exceeds a preset threshold, the deformation compensation mechanism is activated, and a deformation correction vector is calculated and generated based on the deformation.

[0066] Specifically, a force sensor installed on the CNC machine tool acquires real-time force feedback signals between the cutting tool and the material during the cutting process. This signal can be represented as the force exerted by the tool on a unit, denoted by F(t), with the unit being Newtons (N), and t representing time. Simultaneously, an infrared temperature sensor acquires temperature field data of the machining area, denoted by T(x,y,z), with the unit being degrees Celsius. ), where (x,y,z) represent spatial coordinates. The force feedback signal is input into the finite element analysis model, and the processing area is meshed. The stress of each element can be expressed as:

[0067]

[0068] in, Indicates the first The average stress of each element, expressed in Pascals (Pa); The force exerted by the cutting tool on the unit; The area of ​​the unit subjected to force is given. Based on temperature data obtained from an infrared temperature sensor, a continuous temperature field is obtained using an interpolation algorithm, and the thermal stress is further calculated.

[0069]

[0070] in, Thermal stress; This is the elastic modulus of the material, expressed in Pascals (Pa). This is the coefficient of linear expansion of the material, in units of... ; ,in The reference temperature is used. Combining mechanical and thermal stresses, a thermo-mechanical coupling model is employed to calculate the total deformation of the processed area.

[0071]

[0072] in, The deformation is expressed in millimeters (mm). The feature length represents the geometric dimensions of the machining area. If the calculated... Exceeding the preset threshold The system will automatically activate the compensation mechanism to generate a correction vector opposite to the deformation direction. Its size is equal to Its direction is opposite to that of the displacement field, and it is used to adjust the tool path and attitude in real time.

[0073] For example, during the machining of the curved area of ​​an aluminum alloy mobile phone frame, the force sensor collected a peak cutting force signal of 120N between the tool and the workpiece in real time, and the infrared temperature sensor monitored a local machining area temperature of approximately 85°C. After inputting these data into the finite element analysis model, the system obtained a maximum local stress distribution of 180 MPa. Simultaneously, temperature field analysis showed that the high-temperature gradient region was concentrated at the corner of the frame, with a predicted thermal stress of approximately 60 MPa. Combined with the thermo-mechanical coupling model, the deformation in this region was calculated to be 0.026 mm, while the preset threshold was 0.02 mm. Therefore, the system automatically triggered a compensation mechanism, generating a correction vector of size 0.006 mm, with its direction opposite to the displacement field. This vector was used to dynamically adjust the tool attitude and path, thereby ensuring the surface machining accuracy and finish of the frame.

[0074] In one specific embodiment, the process of performing step S4 may specifically include the following steps:

[0075] Support vector machines are used to learn the deformed correction vectors on the training dataset and then perform classification.

[0076] Based on the classification results of the support vector machine, determine whether the current deformation type is dominated by curvature change or overall contour shift;

[0077] If the deformation type is dominated by curvature change, then the local curvature change can be adapted by reducing the cutting force parameter;

[0078] If the deformation type is overall contour offset, then the overall geometric change of the material is corrected by optimizing the toolpath, and the toolpath offset is adjusted.

[0079] Based on the classification results, a set of classified deformation features is generated, key deformation features are extracted, and the adjustment coefficient of tool force feedback is calculated based on the key deformation features. The tool force feedback control parameters are updated to optimize the real-time attitude angle of the tool. Among them, the deformation features include the curvature change amplitude and the path offset.

[0080] Specifically, the system inputs standardized deformation correction vectors collected in real time during machining into a support vector machine classifier. This classifier, trained based on historical machining data and labeled samples, can distinguish between deformation types dominated by curvature changes and overall contour shifts. When the classification result indicates that curvature changes are dominant, the system identifies a large change in local curvature radius, for example, a sudden drop in curvature radius from 12mm to 6mm. In this case, the cutting force parameters are reduced to achieve flexible cutting, avoiding surface damage caused by concentrated stress. When the classification result indicates that the overall contour shift occurs, the system identifies an overall geometric shift in the boundary, for example, an overall displacement of 0.02mm. In this case, machining accuracy is restored by correcting the tool path offset. After classification, the system generates a classified deformation feature set, which includes key indicators such as the curvature change magnitude and path offset, for example, a curvature change magnitude of 50% and a path offset of 0.02mm. Based on this feature set, the system further calculates the adjustment coefficient of the tool force feedback, such as an adjustment coefficient of 1.15, and updates the tool force feedback control parameters accordingly, thereby dynamically optimizing the real-time attitude angle of the tool, so that the tool can maintain stable cutting and high-precision machining under complex curvature and overall deformation conditions.

[0081] Taking the machining of aluminum alloy mobile phone frames as an example, by arranging force sensors and infrared temperature sensors on a CNC machine tool, the system acquires cutting force signals and temperature field data of the machining area in real time. The collected signals are then input into a support vector machine (SVM) classification model for analysis. The SVM learns on a pre-built training dataset and can accurately distinguish between deformation types dominated by curvature changes or overall contour shifts. When the classification result indicates that curvature changes are dominant, for example, when the curvature radius at the corner of the frame drops sharply from 12mm to 6mm, the system determines that there is significant local deformation. Therefore, it reduces the cutting force parameters (e.g., from 12N to 9N) to reduce the force between the tool and the material, thus avoiding local damage. When the classification result indicates that the overall contour shift is dominant, for example, when the overall frame shift reaches 0.02mm, the system restores the overall contour accuracy by correcting the tool path shift (e.g., compensating 0.02mm in the X direction). After classification, the system generates a deformation feature set containing key indicators such as curvature change amplitude and path offset, for example, a curvature change amplitude of 50% and a path offset of 0.02mm. Based on this feature set, the system calculates an adjustment coefficient for the tool force feedback, for example, 1.2, and applies it to the force feedback control module to update the tool attitude angle in real time. Through this process, the tool can achieve adaptive control under different material deformation scenarios, thereby improving the machining accuracy and surface finish of the curved areas of the mobile phone frame. (Reference) Figure 2 The figures illustrate the analysis process of deformation monitoring and control in CNC machining of mobile phone frames based on machine learning. Figure (a) shows a deformation classification model built based on the support vector machine algorithm. The model clearly distinguishes between two deformation modes: "curvature change-dominated" and "overall contour offset," achieving a classification accuracy of 96.2%. Figure (b) highlights the detection effect of local curvature changes. By comparing the standard contour with the deformed contour, an abnormal change in the curvature radius, dropping abruptly from 12mm to 6mm, was identified at the corner of the frame. The system dynamically adjusted the cutting force parameters, improving the roughness of this area by 43.4%. Figure (c) demonstrates the monitoring capability of overall contour offset. A high-precision laser displacement sensor detected a systematic offset of 0.02mm. A path compensation algorithm achieved a compensation accuracy of ±0.001mm, improving the contour dimension accuracy by 25%. Figure (d) verifies the overall effect of the intelligent control system through comparative experiments. After optimization, the surface roughness of the machined part decreased from 2.1 μm to 0.8 μm, which was particularly obvious in the curvature variation region. At the same time, the machining efficiency was improved by 17.8% and the tool life was extended by 26.4%.

[0082] In one specific embodiment, the process of performing step S5 may specifically include the following steps:

[0083] The tool posture angle is fine-tuned by iteratively calculating the parameters of the tool posture angle using a genetic algorithm, and the tool posture is continuously adjusted by a multi-generation evolutionary optimization algorithm.

[0084] Adjust the contact point sequence according to the fine-tuning parameters to optimize the contact between the tool and the material surface;

[0085] By combining the deformation feature set and analyzing the effects of material stress and temperature changes, the vibration amplitude parameters are optimized.

[0086] A resonance mode adjustment strategy is generated based on the optimized vibration amplitude parameters to avoid resonance phenomena.

[0087] An optimized set of attitude parameters was created based on the resonant mode adjustment strategy, and simulation was used to confirm whether it could improve surface finish and extend tool life.

[0088] Specifically, a genetic algorithm, a global optimization method based on natural selection and genetic mechanisms, is used to iteratively calculate the fine-tuning parameters of the tool's attitude angle. Through multiple generations of evolutionary operations such as population initialization, crossover, and mutation, the tool's attitude on complex curved surfaces is continuously optimized, enabling it to better match the curvature changes of the boundary geometry. Based on the optimized fine-tuning parameters, the contact point sequence is adjusted accordingly, improving the cutting contact state between the tool and the workpiece surface and reducing local stress concentration and surface defects caused by attitude deviations. Furthermore, the system analyzes the deformation feature set, generated by the previous classification module, which covers the stress distribution of the material under stress and thermal deformation information caused by temperature rise. These features are used to further optimize the tool's vibration amplitude, enabling differentiated vibration control capabilities in different curvature regions. Subsequently, the system generates a resonance mode adjustment strategy based on the optimized vibration amplitude parameters to avoid harmful resonance between the tool and the workpiece at specific frequencies, ensuring the stability of the machining process. Based on this resonance mode adjustment strategy, an optimized attitude parameter set is created, and its adaptability in actual machining is verified through simulation testing.

[0089] In one specific embodiment, the process of performing step S6 may specifically include the following steps:

[0090] Extract the attitude adjustment command for the current machining segment from the optimized attitude parameter set, and determine the required attitude adjustment angle and contact point position based on the current machining state and path of the tool.

[0091] The attitude adjustment command, attitude adjustment angle, and contact position are transmitted to the CNC system to ensure that the CNC system adjusts the tool path and attitude in real time and performs dynamic path updates.

[0092] Obtain the updated path data and compare it with the preset machining path to ensure that the toolpath is consistent with the updated path data;

[0093] The processing accuracy data of the updated path is collected by sensors, the processing accuracy is monitored in real time, and it is determined whether the processing result meets the preset accuracy requirements. If it does not meet the requirements, additional sensor signals are collected, and based on the additional sensor signals, the previous attitude parameter set is fused to iteratively generate deformation data.

[0094] Specifically, the system extracts a subset of data corresponding to the current machining segment from the optimized attitude parameter set, and combines this with the actual machining state of the tool and the preset path to determine the required attitude adjustment angle and contact point position for this segment, thus generating an attitude adjustment command. This command, along with the attitude angle and contact point position information, is transmitted to the CNC system. The CNC controller parses and applies this information in real time during machining, dynamically updating the tool path and attitude to ensure continuous matching with changes in the complex curvature of the frame. After completing the dynamic path update, the CNC system outputs the updated path data, which is then compared and verified with the preset machining path to ensure that the actual tool movement trajectory remains consistent with the new path data, avoiding the accumulation of deviations. Simultaneously, sensors collect real-time machining accuracy data of the updated path, including indicators such as surface roughness, dimensional deviation, and stress distribution, and compare this data with preset accuracy thresholds. If the monitoring results show that the machining accuracy does not meet the requirements, the system will automatically activate the redundant sensor module to collect additional signals. Based on this, the newly acquired data will be fused with the previous attitude parameter set to iteratively generate new deformation data, thereby providing a basis for subsequent optimization and compensation, and ensuring the stability and high-precision output of the entire machining process.

[0095] In one specific embodiment, the process of performing step S7 may specifically include the following steps:

[0096] A genetic algorithm is used to process the iterative deformation data to generate a refined tool posture sequence;

[0097] A reinforcement learning algorithm is used to train a refined tool posture sequence. The tool path and posture are dynamically adjusted through a reward mechanism to generate the final tool posture sequence. This sequence is then converted into CNC program code and transmitted to the CNC system for dynamic path updates and tool adjustments.

[0098] Specifically, in one embodiment, the system first inputs the iteratively modified deformation data into the genetic algorithm model. The genetic algorithm performs a global search and optimization of the tool posture angle and contact point through operations such as population initialization, crossover, and mutation, ultimately generating a more refined tool posture sequence. For example, in the region where the bending radius of the frame transitions from 15mm to 5mm, the refined sequence obtained by the genetic algorithm can make the posture angle adjustment smoother, avoiding over-cutting or under-cutting.

[0099] The system inputs the refined posture sequence into the reinforcement learning model. The reinforcement learning algorithm comprehensively evaluates the machining accuracy, surface finish, and tool stress stability by setting a reward mechanism, and dynamically adjusts the tool path and posture during the training process to obtain better results.

[0100] For example, when sensor feedback indicates a decrease in surface roughness, the model receives a positive reward, thereby further optimizing the attitude adjustment strategy. After multiple training and updates, the system finally generates an optimized tool attitude sequence and automatically converts this sequence into CNC program code, which is then transmitted to the CNC system for dynamic path updates and tool attitude adjustments, achieving intelligent and high-precision control of the machining process.

[0101] In one specific embodiment, the process of adjusting the tool posture and vibration frequency to adapt to the curvature change in S2 can specifically include the following steps:

[0102] Based on the distribution characteristics of the curvature radius of the border, the attitude adjustment angle of the tool in each machining area is calculated, and a dynamic attitude control sequence is generated.

[0103] By combining dynamic attitude control sequence, the vibration frequency parameters of the tool are adjusted and simulation test is conducted to verify whether the adaptability of the vibration frequency parameters to the change of frame curvature meets the requirements.

[0104] If the vibration frequency parameters do not meet the requirements, recalculate the attitude adjustment angle and update the dynamic attitude control sequence.

[0105] A vibration pattern matching the updated dynamic attitude control sequence is generated and stored for real-time control of tool attitude adjustment angle and vibration frequency parameters during machining.

[0106] Specifically, based on the curvature radius distribution characteristics of the phone's frame, the system calculates the tool's attitude adjustment angles in various machining areas. These angles are derived from the geometric characteristics and machining requirements of different curvature radius regions. For example, in sharp bends in the frame where the curvature radius is less than 8mm, the calculated tool attitude angle may need to be increased to 15 degrees to ensure more precise contact between the tool and the material. Subsequently, the system generates a dynamic attitude control sequence based on these attitude adjustment angles to ensure that the tool's motion trajectory adapts to changes in complex curved surfaces. (Reference) Figure 3 The figure illustrates the adaptive control process of tool posture and vibration frequency based on curvature variation.

[0107] Based on the generated dynamic attitude control sequence, the system adjusts the tool's vibration frequency parameters. For example, in regions with low curvature (curvature radius < 6 mm), the tool's vibration frequency may need to be increased to 2000 Hz to maintain cutting stability and reduce thermal deformation of the material. The system verifies the adaptability of the vibration frequency parameters to changes in the frame curvature through simulation experiments, evaluating whether it can maintain stable cutting force and high-precision machining results in different curvature regions. If the simulation results show that the vibration frequency parameters do not meet the machining accuracy requirements, the system will recalculate the attitude adjustment angle and update the dynamic attitude control sequence to ensure that the tool attitude and vibration frequency match the actual machining requirements.

[0108] The system generates vibration patterns that match the updated dynamic attitude control sequence and stores them in the system. These patterns serve as the basis for real-time control of tool attitude adjustment angles and vibration frequency parameters during machining. The system can achieve adaptive control throughout the machining process, ensuring tool accuracy and surface finish in the curved areas of the edge.

[0109] The above describes the machine learning-based CNC machining monitoring method for mobile phone bezels in the embodiments of this application. The following describes the machine learning-based CNC machining monitoring system for mobile phone bezels in the embodiments of this application. Please refer to [link to relevant documentation]. Figure 4 One embodiment of the CNC machining monitoring system for mobile phone bezels based on machine learning in this application includes:

[0110] The data acquisition module 201 is used to acquire three-dimensional geometric data of the curved area of ​​the mobile phone frame, generate a geometric model of the frame, and determine the distribution characteristics of the frame curvature radius.

[0111] The path matching module 202 is used to match the tool path template according to the distribution characteristics of the radius of curvature of the border, determine the sequence of tool contact points, and adjust the tool posture and vibration frequency to adapt to the curvature change.

[0112] The correction generation module 203 is used to acquire sensor signals during the processing, determine the degree of material deformation based on these signals, and generate a deformation correction vector.

[0113] The classification feature module 204 is used to classify the deformation correction vector, determine the deformation type, and adjust the tool force feedback and attitude based on the deformation type to generate a deformation feature set.

[0114] The attitude optimization module 205 is used to optimize the tool attitude parameters and contact point sequence based on the deformation feature set, and adjust the vibration amplitude of the tool to generate an optimized attitude parameter set.

[0115] The accuracy verification module 206 is used to generate attitude adjustment instructions based on the optimized attitude parameter set and verify the machining accuracy in combination with sensor feedback. If the accuracy does not meet the requirements, deformation data is generated iteratively.

[0116] The control execution module 207 is used to refine the tool posture sequence based on the iterated deformation data, generate the final tool posture sequence, and execute machining control.

[0117] Through the collaborative efforts of the aforementioned components, intelligent monitoring and dynamic control of the entire process for machining the curved areas of mobile phone frames are achieved. This system can adapt to geometric changes in real time under complex curvature environments and effectively compensate for stress and thermal deformation generated during machining. Simultaneously, by combining classification features and optimization algorithms, it dynamically adjusts tool posture and vibration parameters, thereby improving machining accuracy and surface finish, extending tool life, and significantly enhancing overall production efficiency and machining stability.

[0118] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for monitoring the CNC machining of a mobile phone frame based on machine learning, characterized in that, The method includes: S1. Obtain the three-dimensional geometric data of the curved area of ​​the mobile phone frame, generate the frame geometric model, and determine the distribution characteristics of the frame curvature radius. S2. Based on the distribution characteristics of the radius of curvature of the border, match the tool path template, determine the sequence of tool contact points, and adjust the tool posture and vibration frequency to adapt to the curvature change. S3. Acquire sensor signals during the processing, determine the degree of material deformation based on these signals, and generate a deformation correction vector. S4. Classify the deformation correction vector to determine the deformation type, and adjust the tool force feedback and attitude based on the deformation type to generate a deformation feature set; S5. Based on the deformation feature set, optimize the tool attitude parameters and contact point sequence, and adjust the vibration amplitude of the tool to generate an optimized attitude parameter set. S6. Generate attitude adjustment instructions based on the optimized attitude parameter set, and verify the processing accuracy in conjunction with sensor feedback. If the accuracy does not meet the requirements, generate deformation data iteratively. S7. Refine the tool posture sequence based on the iterated deformation data, generate the final tool posture sequence, and execute machining control; S4 includes: using a support vector machine to learn the deformation correction vector on the training dataset and completing classification processing; determining whether the current deformation type is dominated by curvature change or overall contour offset based on the classification result of the support vector machine; if the deformation type is dominated by curvature change, then reducing the cutting force parameter to adapt to the local curvature change; if the deformation type is the overall contour offset, then optimizing the tool path to correct the overall geometric change of the material and adjusting the tool path offset; generating a classified deformation feature set based on the classification result, extracting key deformation features, and calculating the tool force feedback adjustment coefficient based on the key deformation features, updating the tool force feedback control parameters, thereby optimizing the real-time attitude angle of the tool, wherein the deformation features include the curvature change amplitude and the path offset; S5 includes: iteratively calculating the fine-tuning parameters of the tool posture angle using a genetic algorithm, and continuously adjusting the tool posture using a multi-generation evolutionary optimization algorithm; adjusting the contact point sequence according to the fine-tuning parameters to optimize the contact between the tool and the material surface; optimizing the vibration amplitude parameters by analyzing the influence of material stress and temperature changes in combination with the deformation feature set; generating a resonance mode adjustment strategy based on the optimized vibration amplitude parameters to avoid resonance; creating the optimized posture parameter set according to the resonance mode adjustment strategy, and confirming whether it can improve surface finish and extend tool life through simulation detection; S6 includes: extracting the attitude adjustment command for the current machining segment from the optimized attitude parameter set; determining the required attitude adjustment angle and contact point position based on the current machining state and path of the tool; transmitting the attitude adjustment command, the attitude adjustment angle, and the contact point position to the CNC system to ensure that the CNC system adjusts the tool path and attitude in real time and performs dynamic path updates; acquiring the updated path data and comparing it with the preset machining path to ensure that the tool path is consistent with the updated path data; collecting machining accuracy data of the updated path through sensors, monitoring machining accuracy in real time, and determining whether the machining result meets the preset accuracy requirements. If it does not meet the requirements, additional sensor signals are collected, and the deformation data is iteratively generated by fusing the previous attitude parameter set based on the additional sensor signals. 2.The machine learning based mobile phone bezel CNC machining monitoring method according to claim 1, characterized in that, S1 includes: The three-dimensional geometric data of a specific curved area of ​​the mobile phone frame is acquired using a scanning device; The three-dimensional geometric data is processed using a surface reconstruction method to generate the bounding box geometric model; Based on the geometric model of the border, the radius of curvature of each region of the border is calculated, and a gradual distribution feature of the radius of curvature is generated. The curvature radius gradient distribution characteristics are analyzed to extract key feature points of the frame geometry; Based on the key feature points, the principal curvature is calculated and converted into curvature radius. A curvature radius sequence is generated according to the processing path order. An interpolation method is used to form a continuously changing curve to determine the curvature radius distribution characteristics of the border. 3.The machine learning based mobile phone bezel CNC machining monitoring method of claim 1, wherein, S2 includes: Extract an initial path template from the preset toolpath library that matches the curvature radius distribution characteristics of the border; Based on the initial path template, a sequence of initial contact points of the tool on the complex geometry of the border is generated; For the initial contact point sequence, calculate the tangential vector and normal vector of the tool at each contact point; The tangential and normal vectors are optimized using a particle swarm optimization algorithm to generate a contact point control sequence for attitude angle adjustment. A genetic algorithm is used to optimize the distribution characteristics of the radius of curvature of the border, obtain the vibration frequency control parameters of the tool, and determine the initial vibration mode of the tool based on the vibration frequency control parameters. The contact point control sequence is simulated and verified to determine whether it adapts to curvature changes. If it does not adapt, the path template is adjusted and the contact point sequence is regenerated. 4.The method of claim 1, wherein, S3 includes: Force sensors installed on CNC machine tools are used to collect force feedback signals between the cutting tool and the workpiece in real time. At the same time, temperature data of the workpiece area is obtained through infrared temperature sensors. Based on the collected force feedback signal, a finite element analysis model is established, the force feedback signal is mapped onto the mesh of the processing area, the stress distribution of each element is calculated, and the material stress distribution is determined. Using the temperature data, the heat distribution characteristics of the processing area are calculated through an interpolation algorithm to identify high-temperature areas and temperature gradients, thereby predicting thermal stress and possible material deformation. Combining the stress distribution and heat distribution characteristics of the material, a thermo-mechanical coupling model is used to calculate the degree of deformation of the material during processing, the degree of deformation is quantified into deformation amount, and compared with a preset threshold. If the deformation exceeds a preset threshold, the deformation compensation mechanism is activated, and a deformation correction vector is calculated and generated based on the deformation. 5.The machine learning based mobile phone bezel CNC machining monitoring method of claim 1, wherein, S7 includes: A genetic algorithm is used to process the iterative deformation data to generate a refined tool posture sequence; The refined tool posture sequence is trained using a reinforcement learning algorithm. The tool path and posture are dynamically adjusted through a reward mechanism to generate the final tool posture sequence. This sequence is then converted into CNC program code and transmitted to the CNC system for dynamic path updates and tool adjustments. 6.The machine learning based mobile phone bezel CNC machining monitoring method of claim 1, wherein, In step S2, adjusting the tool posture and vibration frequency to adapt to curvature changes includes: Based on the curvature radius distribution characteristics of the border, the attitude adjustment angle of the tool in each processing area is calculated, and a dynamic attitude control sequence is generated. Based on the dynamic attitude control sequence, the vibration frequency parameters of the tool are adjusted, and simulation experiments are conducted to verify whether the adaptability of the vibration frequency parameters to the change of the frame curvature meets the requirements. If the vibration frequency parameter does not meet the requirements, recalculate the attitude adjustment angle and update the dynamic attitude control sequence; A vibration pattern matching the updated dynamic attitude control sequence is generated and stored for real-time control of the tool's attitude adjustment angle and vibration frequency parameters during machining. 7.A machine learning based mobile phone frame CNC machining monitoring system, configured to implement the machine learning based mobile phone frame CNC machining monitoring method according to any one of claims 1-6, characterized in that, The machine learning-based mobile phone frame CNC machining monitoring system includes: The data acquisition module is used to acquire three-dimensional geometric data of the curved area of ​​the mobile phone frame, generate a geometric model of the frame, and determine the distribution characteristics of the frame curvature radius. The path matching module is used to match the tool path template according to the distribution characteristics of the radius of curvature of the border, determine the sequence of tool contact points, and adjust the tool posture and vibration frequency to adapt to the curvature change. The correction generation module is used to acquire sensor signals during the processing, determine the degree of material deformation, and generate a deformation correction vector. The classification feature module is used to classify the deformation correction vector, determine the deformation type, and adjust the tool force feedback and attitude based on the deformation type to generate a deformation feature set. The attitude optimization module is used to optimize the tool attitude parameters and contact point sequence based on the deformation feature set, and adjust the vibration amplitude of the tool to generate an optimized attitude parameter set. The accuracy verification module is used to generate attitude adjustment instructions based on the optimized attitude parameter set, and verify the processing accuracy in combination with sensor feedback. If the accuracy does not meet the requirements, deformation data is generated iteratively. The control execution module is used to refine the tool posture sequence based on the iterated deformation data, generate the final tool posture sequence, and execute machining control.

Images