Multi-station self-adaptive surface treatment system for metal spectacle frame

Abstract

The invention discloses a multi-station self-adaptive surface treatment system for a metal spectacle frame, and relates to the technical field of spectacle manufacturing and optical detection. The system mainly solves the problems that in the prior art, when surface treatment is conducted on the complex curved surface of the metal spectacle frame, manual experience is relied on, detection and machining are disjointed, and technological parameters cannot be adjusted in a self-adaptive mode, so that the treatment quality is inconsistent, and efficiency is low. The method comprises the following steps: acquiring a three-dimensional digital model with a defect mark through a three-dimensional scanning and defect initial marking module; the multi-station adaptive execution module plans a path according to the model and executes processing; the online rechecking and monitoring module is used for performing in-situ detection and process monitoring; and the intelligent collaborative master control module synthesizes all information to realize multi-station dynamic scheduling and process parameter optimization. According to the system, full-process closed-loop intelligent treatment from defect recognition and self-adaptive machining to online feedback optimization is achieved, and the automation degree, consistency and yield of surface treatment of the metal spectacle frame are remarkably improved.

Description

Technical Field

[0001] This invention discloses a multi-station adaptive surface treatment system for metal eyeglass frames, relating to the fields of eyeglass manufacturing and optical testing technology. Background Technology

[0002] In the field of surface treatment of metal eyeglass frames, especially in the process of high gloss and complex curved surface finishing, the existing technology system still faces several inherent defects.

[0003] In the defect detection stage, conventional methods typically rely on offline, spot-checking manual visual inspection or single-angle optical equipment. This method is difficult to achieve full-area, blind-spot-free quantitative inspection of the complex three-dimensional curved surface of eyeglass frames, which can easily lead to missed detection of minor scratches, dents and other defects. Furthermore, the inspection results cannot be directly and accurately correlated with subsequent processing procedures.

[0004] Existing surface treatments are mostly performed using mechanized equipment with fixed programs or simple contouring. These devices lack the ability to perceive and adapt to individual differences in workpieces. Their processing paths and process parameters are often preset and uniform, making it impossible to dynamically adjust for defects of different locations and types. This can easily lead to under-processing or over-processing, especially in irregular curved surface areas.

[0005] The system suffers from insufficient integration and coordination. Units such as inspection, rough polishing, fine polishing, and cleaning often operate independently or are only physically connected, forming "information silos." There is a lack of efficient collaboration and dynamic scheduling between workstations based on real-time data. The system cannot intelligently adjust production rhythm and allocate tasks according to the actual results of upstream processes and the real-time status of downstream workstations, thus limiting overall processing efficiency and resource utilization.

[0006] The entire process lacks effective closed-loop quality control. A traceable and analyzable real-time feedback link has not been established between the key parameters in the process and the final quality results. Process optimization relies heavily on the experience of operators, making it difficult to achieve stable and consistent high-quality output, and even more difficult to form a process knowledge base for continuous self-optimization. Summary of the Invention

[0007] The purpose of this invention is to provide a multi-station adaptive surface treatment system for metal eyeglass frames, offering a closed-loop processing solution integrating optical inspection and adaptive machining. The system first performs a full-domain scan and modeling of the workpiece using a 3D scanning and defect initial labeling module, generating a 3D digital model with defect markings. Then, a multi-station adaptive execution module plans and executes targeted surface treatments based on this model. An online re-inspection and monitoring module performs in-situ inspection and process monitoring during and intermittent processing, generating real-time quality reports. Finally, an intelligent collaborative control module integrates all information to dynamically schedule and optimize process parameters across the multi-station process, thereby achieving closed-loop control throughout the entire process, from accurate defect identification, adaptive path planning, online quality feedback to intelligent decision-making and scheduling, effectively improving processing consistency and intelligence.

[0008] The objective of this invention can be achieved through the following technical solutions: A multi-station adaptive surface treatment system for metal eyeglass frames includes a 3D scanning and defect initial labeling module, a multi-station adaptive execution module, an online re-inspection and monitoring module, and an intelligent collaborative control module. The three-dimensional scanning and defect labeling module is used to perform three-dimensional reconstruction and defect identification of metal eyeglass frame workpieces based on multi-angle optical scanning, and output a three-dimensional digital model of the eyeglass frame with global defect markings. The multi-station adaptive execution module is used to implement multi-tool head collaborative path planning and surface treatment of metal eyeglass frame workpieces based on the three-dimensional digital model of the eyeglass frame with global defect marking, so as to obtain a semi-finished eyeglass frame after preliminary treatment. The online re-inspection and monitoring module is used to use the force and light signals collected during the processing to complete the in-situ detection and data comparison of the current processing area of ​​the semi-finished eyeglass frame after preliminary processing, and generate a real-time processing status and quality deviation report. The intelligent collaborative control module is used to integrate the three-dimensional digital model of the eyeglass frame with global defect markings with the real-time processing status and quality deviation reports to perform dynamic scheduling and process parameter optimization of the entire system's multi-station process, so as to achieve finished eyeglass frames with surface quality that meet the standards.

[0009] Preferably, the data acquisition unit of the three-dimensional scanning and defect initial labeling module includes: An optical scanning assembly consisting of a projector and multiple industrial cameras, wherein the projector projects a specific coded structured light pattern onto the surface of a metal eyeglass frame workpiece, and the multiple industrial cameras are configured to be synchronously triggered by the projector to acquire pattern-modulated workpiece surface images from different angles to obtain complete contour information for 3D reconstruction. High-resolution spectral confocal sensors emit a probe beam to scan the surface of a workpiece point by point. By analyzing the changes in the wavelength or spectral distribution of the returned beam, they capture the microscopic height fluctuations and material reflection characteristics of the surface, thereby identifying the spectral features of defects such as scratches and pits.

[0010] Preferably, the data processing unit of the 3D scanning and defect initial labeling module includes: The module receives multi-view images and spectral data transmitted by the data acquisition unit, performs multi-view point cloud registration and triangular mesh surface reconstruction, and constructs an accurate three-dimensional model of the eyeglass frame. The module also integrates local material hardness distribution information obtained by laser ranging or stylus detection to form a comprehensive three-dimensional digital model that includes geometric shape and physical properties. The comprehensive three-dimensional digital model is analyzed to quantify the depth and projected area of ​​each identified defect. Combined with the location of the defect in the lens frame, temple, or nose pad functional area, the urgency score for repair is automatically calculated based on a preset rule base. Finally, a defect repair priority map is generated, which is superimposed on the three-dimensional model and marked with different colors or markers to indicate the repair order.

[0011] Preferably, the planning system of the multi-station adaptive execution module includes: The system analyzes the 3D model and defect map to calculate the collision-free optimal toolpath and process sequence for different workstations; and it also has a parameter matching unit that matches tool head model, spindle speed and feed rate from the process library based on the characteristics of different areas of the workpiece.

[0012] Preferably, the execution system of the multi-station adaptive execution module includes: Import the 3D digital model and priority map with defect markings generated by the 3D scanning and defect initial marking module, and calculate the collision-free and time-optimal tool contact path and feed trajectory for each defect area to be processed based on the tool head workspace and kinematic model of each processing station. Generate a global, time-sequential collaborative station operation instruction sequence in a coordinated manner. Based on the characteristics of different areas of the workpiece automatically identified in the 3D model, and combined with the defect type and priority, the optimal combination of processing parameters is dynamically matched and output from the preset process knowledge base. The parameter combination includes at least the recommended tool head model, spindle rotation speed, tool feed rate, and polishing media supply flow rate.

[0013] Preferably, the detection system of the online re-inspection and monitoring module includes: It is fixedly mounted on the end flange of the robotic arm that performs surface treatment, maintaining a defined relative position with the tool head; it adopts a coaxial optical path design, and automatically controls the lens to quickly focus on the current processing area during the interval of tool lifting or moving, and acquires high-resolution digital images for evaluating surface roughness and micromorphology. It consists of a thermocouple and a vibration accelerometer embedded inside the polishing or grinding tool head; the measuring end of the thermocouple is close to the working surface of the tool to directly monitor the instantaneous temperature change in the polishing area; the vibration accelerometer collects multi-dimensional vibration spectrum signals of the tool head in real time during the processing to analyze the tool status and the stability of the processing process.

[0014] Preferably, the analysis system of the online re-inspection and monitoring module includes: The system receives high-definition images of the processing area transmitted by the detection system and performs high-precision registration and pixel-level comparison with the ideal surface morphology of the corresponding area in the original three-dimensional digital model generated by the three-dimensional scanning and defect initial labeling module. The system then identifies and quantifies the size, area, and precise three-dimensional coordinates of the residual defects in the area through an algorithm, and generates a quality deviation report containing the quantitative evaluation results. The system continuously receives and analyzes the polishing temperature and tool vibration spectrum signals collected by the detection system. By comparing the preset process safety threshold with the signal reference model under normal operating conditions, it can determine in real time whether there is an abnormal temperature rise or vibration instability in the processing process. Once an abnormality is detected, it immediately generates and sends an alarm signal or process parameter correction instruction containing a specific workstation identifier to the upper control system according to the predefined rule base.

[0015] Preferably, the scheduling system of the intelligent collaborative central control module includes: Construct and maintain a virtual digital twin model that precisely corresponds to the physical processing system. This model receives real-time data on equipment status, workpiece position, and process from the physical system. It predicts subsequent processing flows and potential bottlenecks through parallel simulation and issues preventative workstation start / stop, cycle time adjustment, or path replanning instructions to the physical system based on the simulation results. The system continuously monitors the real-time equipment load rate, tool health status, job queue length, and quality data fed back by the online re-inspection and monitoring module for each physical workstation in the multi-station adaptive execution module. Based on preset objectives such as maximizing throughput, optimizing energy consumption, or prioritizing quality, the system uses a scheduling algorithm to dynamically calculate and assign the optimal flow sequence of the current and subsequent eyeglass frame workpieces in the system and the timing of their entry into each workstation for processing.

[0016] Preferably, the optimization system of the intelligent collaborative control module includes: The system receives real-time quality deviation reports generated by the online re-inspection and monitoring module, and retrieves a set of successful "workpiece-process-result" cases similar to the current eyeglass frame in terms of 3D features, defect type, and material from the historical process knowledge base for matching analysis. Through regression analysis or machine learning algorithms, the system iteratively adjusts and outputs an optimized set of process parameters for subsequent unprocessed workstations or subsequent similar workpieces of the current workpiece. The parameter set includes tool type, spindle speed, feed rate, and processing time. At the end of the single-piece eyeglass frame processing flow, the system automatically collects and integrates multi-source heterogeneous data from the entire process of the workpiece, from 3D scanning, defect marking, processing parameters of each station, online re-inspection report to the final quality inspection result, to form a structured process archive. Based on the evaluation of the processing results in this archive, the system automatically updates the case data and parameter recommendation rules in the historical process knowledge base, so as to realize the continuous accumulation and optimization of the system's process knowledge.

[0017] The beneficial effects of this invention are: This invention effectively improves the quality consistency of surface treatment for metal eyeglass frames. The system achieves accurate digital modeling of defects across the entire complex curved surface through integrated optical scanning and defect identification, driving a multi-station adaptive actuator for targeted, differentiated processing. Combined with in-situ re-inspection and real-time monitoring during processing, a closed-loop quality control system of "detection-processing-feedback" is formed, ensuring uniform and high-standard smoothness across different parts from the lens rim to the temples. This significantly reduces under-processing or over-processing caused by differences in human experience or fixed equipment programs.

[0018] This invention significantly enhances the overall intelligence and process adaptability of the production system. With the help of an intelligent collaborative control module, the system can dynamically schedule workpiece flow and optimize process parameters at each workstation based on real-time operating conditions. It also continuously learns and optimizes based on historical data, achieving collaboration from personalized processing of individual workpieces to optimal allocation of resources across the entire production line. This not only reduces reliance on skilled operators but also enables the production line to flexibly adapt to the processing of eyeglass frames with different styles and defects, improving production flexibility and efficiency. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the structure of a multi-station adaptive surface treatment system for metal eyeglass frames according to the present invention. Detailed Implementation

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

[0021] Example: Figure 1 As shown, a multi-station adaptive surface treatment system for metal eyeglass frames includes a 3D scanning and defect initial labeling module, a multi-station adaptive execution module, an online re-inspection and monitoring module, and an intelligent collaborative control module. The three-dimensional scanning and defect labeling module is used to perform three-dimensional reconstruction and defect identification of metal eyeglass frame workpieces based on multi-angle optical scanning, and output a three-dimensional digital model of the eyeglass frame with global defect markings. The multi-station adaptive execution module is used to implement multi-tool head collaborative path planning and surface treatment of metal eyeglass frame workpieces based on the three-dimensional digital model of the eyeglass frame with global defect marking, so as to obtain a semi-finished eyeglass frame after preliminary treatment. The online re-inspection and monitoring module is used to use the force and light signals collected during the processing to complete the in-situ detection and data comparison of the current processing area of ​​the semi-finished eyeglass frame after preliminary processing, and generate a real-time processing status and quality deviation report. The intelligent collaborative control module is used to integrate the three-dimensional digital model of the eyeglass frame with global defect markings with the real-time processing status and quality deviation reports to perform dynamic scheduling and process parameter optimization of the entire system's multi-station process, so as to achieve finished eyeglass frames with surface quality that meet the standards.

[0022] In this embodiment, three-dimensional reconstruction and defect identification of the metal eyeglass frame workpiece are performed based on multi-angle optical scanning, and a three-dimensional digital model of the eyeglass frame with global defect markings is output. The specific implementation method is as follows: A multimodal optical scanning platform integrating active structured light and passive spectral sensing was constructed. The core of this platform includes a high-precision digital light processing projector, four spatially orthogonally arranged synchronously triggered industrial cameras, a high-resolution spectral confocal sensor, and a precision rotary stage held by a six-axis robot. During scanning initialization, the eyeglass frame workpiece is fixed to the center of the rotary stage using a dedicated non-destructive flexible fixture. The projector sequentially projects a series of phase-encoded Gray code patterns and sinusoidal fringe patterns onto the workpiece surface. The four cameras achieve microsecond-level synchronous exposure under the projector's hardware trigger signal, capturing a sequence of deformed fringe images modulated by the workpiece surface morphology from four main perspectives: front-view, back-view, left-view, and right-view. This synchronous multi-view acquisition strategy ensures strict spatiotemporal correspondence of all 2D image data, providing a foundation for subsequent high-precision 3D reconstruction. Simultaneously, the six-axis robot drives the spectral confocal sensor to perform point-to-point contact measurements on the workpiece surface along a pre-planned scanning trajectory, synchronously recording the spatial coordinates of each point. and its corresponding full-band reflectance spectral vector This spectral data is an important basis for subsequent material identification and microscopic defect recognition.

[0023] Based on multiple acquired image sequences, a dense 3D point cloud of the workpiece surface is reconstructed using phase decomposition and multi-view stereo matching techniques. For the temporal fringe images captured by each camera, phase decomposition is performed using the phase-shifting method and the multi-frequency heterodyne principle to calculate the absolute phase value of each pixel. Where the subscript 'c' represents the camera index, and (u,v) are the pixel coordinates. The system parameters—including the intrinsic parameter matrices of each camera—are obtained beforehand through high-precision target calibration. extrinsic rotation matrix relative to the world coordinate system Translation vector And the intrinsic and extrinsic parameters of the projector—establishing a mapping relationship from phase to 3D coordinates. Stereo matching is performed by finding points with consistent phase in different camera views, and their 3D coordinates are calculated based on the forward intersection principle. For a set of matched point pairs, the optimal 3D point P can be obtained by solving the following nonlinear least squares problem, the objective of which is to minimize the reprojection error: in, It is the camera projection function. Let P be the measured pixel coordinates of point P on the image plane of the c-th camera. This calculation is performed on all matching points, and sub-point clouds from different viewpoints are fused. Then, the iterative nearest-point algorithm is used for global optimization registration, ultimately generating a complete, seamless, and high-precision 3D point cloud of the eyeglass frame workpiece. .

[0024] After obtaining the accurate geometric point cloud, macroscopic geometric defect detection and microscopic surface defect identification are performed in parallel. Macroscopic defect detection is achieved by comparing the reconstructed point cloud C with the theoretical CAD model of a standard eyeglass frame. First, non-rigid registration is performed between the two, and then the values ​​of each point in the point cloud are calculated. Signed distance to the nearest point on the CAD model surface Set a geometric tolerance threshold. All satisfied The points were classified as macroscopic geometric anomaly regions. Such areas typically correspond to dents or deformations caused by impacts.

[0025] Microscopic defect identification primarily relies on spectral data acquired by a spectral confocal sensor and texture information from high-resolution camera images. For the spectral data, feature vectors are extracted for each measurement point, along with the reflectance at specific characteristic wavelengths. The slope of the spectral curve and curvature Meanwhile, texture features of the corresponding regions are extracted from the camera images, such as the contrast of the gray-level co-occurrence matrix. Entropy These features are then fused into a joint feature vector F. A support vector machine classifier, pre-trained on a large number of samples, is used to determine whether the point is a defect. The classification decision function is as follows: in, It is a Lagrange multiplier. These are the training sample labels. is the radial basis kernel function, and b is the bias term. If If a point is identified as a micro-defect, then all micro-defect points constitute a set. This includes scratches, pitting, oxidation spots, etc.

[0026] Information fusion and model generation are performed. The point cloud C is triangulated to generate a continuous curved surface mesh model M. Macroscopic defect areas are then... and micro-defect set The spatial coordinates are mapped onto the mesh model M, and attribute labels are embedded on the corresponding vertices or faces. Each label contains the defect type. ,size ,depth And a comprehensive priority score The score is calculated using the following heuristic formula to guide the scheduling of subsequent processing steps: in, It is a severity coefficient set according to the defect type. and As the normalization factor, It is the weight of the functional area where the defect is located. to The weighting factor is used. The final output, a "3D digital model of an eyeglass frame with global defect markings," is a structured data entity that integrates geometric, attribute, and defect information, serving as the core digital twin driving the entire adaptive surface treatment system.

[0027] In this embodiment, multi-tool head collaborative path planning and surface treatment of the metal eyeglass frame workpiece are implemented based on the three-dimensional digital model of the eyeglass frame with global defect marking to obtain a semi-finished eyeglass frame after preliminary treatment. The specific implementation method is as follows: The intelligent collaborative control module reads and parses the "3D digital model of the eyeglass frame with global defect labels" generated in the previous step. This model contains mesh geometry M, vertex attributes, and embedded defect labels. The system first decomposes the entire defect repair task into a series of ordered "processing meta-tasks" based on the defect label's type (Type), priority score, and spatial clustering results. Each metatask Defined as the processing of a continuous defect area or a specific functional part, it is associated with the recommended machining station, the initial tool head type, and a theoretical machining allowance calculated based on the defect severity and the functional weight of the area. This margin is the basis for planning the cutting depth, and its calculation comprehensively considers the defect depth and the target surface quality. And material properties, a simplified estimation formula is: in, This is an empirical coefficient related to the material removal rate. This represents the current estimated baseline quality level for the region. The minimum effective machining depth set for the system.

[0028] Perform global path and scheduling planning for multi-tool head collaboration. The system bases its decisions on the processing requirements of all meta-tasks T and the workstations... Given the capability matrix and its current state, and with the primary optimization objectives of minimizing total processing time and balancing tool wear, solve a multi-constraint scheduling problem. This problem can be formalized as a mixed-integer programming problem, with the core objective function being: in, For the task The completion time, This represents the target average value of the tool wear coefficient for each workstation. and These are the weighting coefficients. The planner outputs a global Gantt chart, which clarifies each meta-task. The work station I was assigned The tool head used Start time and a rough sequence of tool contact points. These contact points are directly sampled from the vertices of the defect region mesh of the 3D model M.

[0029] Guided by overall planning, when a workpiece is transferred to a specific workstation Once the clamping is complete, the local controller at this workstation receives the specific meta-task. Information is gathered, and local fine-grained path planning is initiated. The core of this stage is based on a high-precision local 3D model and the selected toolhead geometry. and processing margin This generates collision-free, smooth toolpaths that meet process constraints. First, the rough contact point sequence is processed. Perform B-spline curve fitting to generate an initial tool center point path. , where s is the path parameter. To ensure the quality of the machined surface and avoid chatter, the path needs to be smoothed and optimized, while ensuring that the tool attitude changes continuously in complex curved areas. This is achieved by solving a constrained optimization problem whose cost function includes a path curvature penalty term and an attitude change penalty term: in, and represents the weighting coefficient. The optimized path ensures that the toolhead moves smoothly along the surface in terms of posture and speed.

[0030] During physical execution, the system implements adaptive processing based on real-time sensor feedback. A six-dimensional force / torque sensor mounted at the end of the spindle measures the normal force at the tool-workpiece contact interface in real time. and tangential force These force signals are compared with the process setpoints, and the error is input into a proportional-integral-derivative controller to dynamically adjust the feed speed of the robotic arm. Or the micro-displacement of the principal axis in the normal direction This achieves constant force control. The control law can be simplified to: in, Let the force error be at the k-th sampling time. This is for controller gain. Simultaneously, based on real-time feedback from the machining area, the local controller can fine-tune the spindle speed S or oscillation frequency within a preset parameter range. Through the aforementioned closed-loop control, the system performs precise processing, from material removal to surface polishing, on complex areas such as the curved surface of the eyeglass frame and the slender temple rods. When a meta-task... After all paths have been executed and the area has been confirmed by rapid in-situ scanning to meet the preset intermediate quality standards, the workpiece is transferred to the next station until all planned meta-tasks are completed, and finally a "preliminary processed eyeglass frame semi-finished product" is obtained with surface defects basically eliminated or significantly improved and overall smoothness tending to be consistent.

[0031] In this embodiment, the force and light signals collected during the processing are used to complete the in-situ detection and data comparison of the current processing area of ​​the pre-processed eyeglass frame semi-finished product, generating a real-time processing status and quality deviation report. The specific implementation method is as follows: The implementation of this module relies on an in-situ sensing system deeply integrated into the execution module. This system consists of two parts: first, a dynamic force signal acquisition unit, the core of which is a high-bandwidth six-dimensional force / torque sensor installed between the spindle flange and the tool head, used to measure the machining contact point in the tool coordinate system in real time. The three force components below and three torque components These raw signals are sampled at high frequency and low-pass filtered to remove high-frequency mechanical vibration noise, yielding a cleanliness force signal vector for process monitoring. Secondly, there is the in-situ optical signal acquisition unit, the core of which is a coaxial macro vision sensor mounted at a fixed offset to the tool head. After the robotic arm completes a processing sub-path, it briefly lifts and moves to align the field of view of the vision sensor with the area just processed. The sensor then automatically focuses to acquire a high-resolution digital image of the current area. It acquires multi-angle spectral illumination bright-field images of the same area using a built-in green LED ring light source to enhance the contrast of surface micro-textures. All force signals and image data are tagged with a uniform timestamp and working coordinates from the robot controller. To achieve time and space synchronization.

[0032] The acquired raw signals need to undergo feature extraction for subsequent analysis. For force signals, within a processing step's time window... Within this, calculate its characteristic statistics, such as the root mean square value. Peak and the distribution of signal energy in a specific frequency band. These features constitute the force eigenvector. For optical images First, compare it with the expected ideal surface rendering of the corresponding area in the original 3D digital model generated by the 3D scanning and defect initial labeling module. Sub-pixel registration based on feature points is performed. After registration, two key optical features are calculated: one is the average gray-level gradient magnitude of the local region. Firstly, it is used to characterize surface roughness; secondly, it is used through comparison. and The SSIM index is a structural similarity index for images generated under the same lighting model. The calculation of this SSIM index can be simplified to a comparison of brightness. Comparison of contrast c and structure s: in, and These represent the mean and standard deviation of the image, respectively. For covariance, This is the stability constant. An SSIM value close to 1 indicates high consistency.

[0033] Based on feature extraction, the system performs online evaluation by fusing multi-source information. The evaluation is divided into two parallel threads: process stability monitoring and surface quality consistency evaluation.

[0034] For process stability, the system maintains a library of benchmark force signal models for different combinations of "material-tool-process". The force feature vectors extracted in real time are then used... The benchmark model corresponding to the current processing operation A comparison is performed. A process stability index based on Mahalanobis distance is defined. : in, This is the covariance matrix of the baseline model's eigenvectors, used to account for the correlation between the features. If... Exceeding the threshold If the judgment process is abnormal, such as increased tool wear, the appearance of unexpected hard spots, or loosening of the clamps.

[0035] For surface quality consistency assessment, the core is to compare the measured optical characteristics with the preset quality acceptance threshold range based on the current processing stage and area type. and A comparison will be made. Simultaneously, the system will retrieve the initial defect severity score for that region in the digital model. And the theoretical material removal amount of this step. A comprehensive, real-time quality achievement metric. It is calculated to quantify the gap between the current processing effect and the expected goal: in, These are weighted coefficients, and their sum is 1. This is the attenuation coefficient. The formula integrates roughness improvement, topography fidelity, and defect repair depth information.

[0036] Real-time processing status and quality deviation reports are generated in a formatted manner. This report is a structured data object containing at least the following fields: timestamp, workstation ID, toolhead ID, global coordinates of the processing area, and process stability index. Its state and key optical characteristic values Real-time quality achievement And a deviation flag. If If any optical feature exceeds the threshold range, the deviation flag is set, and a preliminary diagnostic suggestion is attached to the report. This report is immediately pushed to the intelligent collaborative control module via the real-time data bus, serving as the core basis for its dynamic scheduling and process parameter optimization.

[0037] In this embodiment, the dynamic scheduling and process parameter optimization of the entire system's multi-station process are performed by combining the three-dimensional digital model of the eyeglass frame with the real-time processing status and quality deviation report, thereby achieving finished eyeglass frames with acceptable surface quality. The specific implementation method is as follows: Driven by the core engine of the intelligent collaborative control module, this engine first builds and maintains a dynamic digital twin model that is precisely synchronized with the physical production line. This digital twin model (DT) includes not only all workstations... The geometric and kinematic models of robots and transmission lines, and more importantly, their real-time state mapping, including: each workpiece in the manufacturing process. Instantiated 3D digital model And its associated defect marking and handling progress, equipment at each workstation Real-time availability status The currently loaded tool head ID and its wear estimate. and all material buffer queues The length. This twin model continuously receives two key data streams from the physical world through an Industrial Internet of Things (IIoT) interface: one is from the 3D scanning and defect initialization module, and the other is from each newly loaded workpiece. The initial "3D digital model with global defect markings" provides First, a unique and authoritative digital mirror was created in the twin world; second, a continuous and timestamp-aligned stream of "real-time processing status and quality deviation reports" from the online review and monitoring module. These reports are instantly linked to the corresponding artifacts in the twin model. and workstation Dynamically update the post-processing quality assessment value of the corresponding area of ​​the workpiece. and workstation process stability index .

[0038] Based on this real-time synchronized digital twin, the dynamic scheduler is activated. It faces a multi-objective, multi-constraint real-time optimization problem: how to schedule the continuously arriving new workpieces... For workpieces already online, under random disturbances, the system dynamically assigns and sorts their entry order into each processing station to optimize overall production efficiency, ensure delivery time, and prioritize high-priority defects. The scheduler periodically runs a rolling time-domain optimization algorithm. The core objective function J of this algorithm aims to optimize the process within a finite look-ahead time window. Achieving a balance among multiple objectives: in, It is a workpiece The estimated completion time, It is the overall priority weight calculated based on the initial defect model. This is its current cumulative waiting time. It is the target value for tool wear equalization. It is an indicator function that indicates whether workstation j is in an abnormal state. This is its expected recovery time. These are configurable weighting coefficients. After solving this problem, the scheduler immediately issues the latest workpiece flow instructions to the physical conveying system and each workstation controller, achieving an adaptive production flow that can cope with disturbances.

[0039] In parallel with dynamic scheduling, the process parameter optimizer is specifically responsible for fine-tuning the parameters of machining tasks that are about to be executed or are currently being executed at a micro level. When the scheduler decides to schedule the workpiece... A certain meta-task Assigned to workstation At this time, the optimizer is triggered. Its input includes: task Corresponding detailed region model Material properties, defect types, and initial quality score of the area. ,for Recommended initial process parameter set And most importantly—historical processing reports from this region or similar regions and the latest real-time reports. .

[0040] The optimizer first searches the historical process knowledge base to find a set E of "successful cases" that match the current scenario. Each case... Includes the parameters it uses and achieved results indicators The optimizer, based on these cases, establishes a model from process parameters using either Gaussian process regression or a neural network proxy. To the prediction results and the stability of the prediction process mapping relationship The current optimization problem is transformed into finding a new set of parameters. Under the condition of satisfying the physical constraints of the equipment Under the premise of maximizing a comprehensive utility function U: here, It is a weight vector of each dimension of result quality. It is a coefficient that penalizes instability in the prediction process. This is a regularization coefficient to prevent parameters from deviating too much from the initial recommended value. It is obtained after optimization. This will be sent to the workstation as a suggested parameter. The local controller executes.

[0041] The system implements closed-loop self-learning. Each time a meta-task... Complete, its complete "input-output" data pair, that is, the final parameters used. Initial state description of the area And authoritative results from subsequent final quality inspections. It will be automatically packaged into a new case. After consistency verification, the data is incrementally stored in the historical process knowledge base. Updates to the knowledge base not only add data but also trigger changes to the internal predictive model. The system undergoes periodic retraining, allowing its predictive capabilities to continuously evolve over time. Through the synergistic effect of the aforementioned dynamic scheduling and parameter optimization, the system ensures that each eyeglass frame component completes all processing along an efficient, high-quality, and personalized path, ultimately yielding finished eyeglass frames with satisfactory surface quality at the export point, where all key quality indicators meet the preset shipping standards. .

[0042] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0043] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-station adaptive surface treatment system for metal eyeglass frames, characterized in that, Includes a 3D scanning and defect initialization module, a multi-station adaptive execution module, an online re-inspection and monitoring module, and an intelligent collaborative control module: The three-dimensional scanning and defect labeling module is used to perform three-dimensional reconstruction and defect identification of metal eyeglass frame workpieces based on multi-angle optical scanning, and output a three-dimensional digital model of the eyeglass frame with global defect markings. The multi-station adaptive execution module is used to implement multi-tool head collaborative path planning and surface treatment of metal eyeglass frame workpieces based on the three-dimensional digital model of the eyeglass frame with global defect marking, so as to obtain a semi-finished eyeglass frame after preliminary treatment. The online re-inspection and monitoring module is used to use the force and light signals collected during the processing to complete the in-situ detection and data comparison of the current processing area of ​​the semi-finished eyeglass frame after preliminary processing, and generate a real-time processing status and quality deviation report. The intelligent collaborative control module is used to integrate the three-dimensional digital model of the eyeglass frame with global defect markings with the real-time processing status and quality deviation reports to perform dynamic scheduling and process parameter optimization of the entire system's multi-station process, so as to achieve finished eyeglass frames with surface quality that meet the standards.

2. The multi-station adaptive surface treatment system for metal eyeglass frames according to claim 1, characterized in that, The data acquisition unit of the 3D scanning and defect preliminary labeling module includes: An optical scanning assembly consisting of a projector and multiple industrial cameras, wherein the projector projects a specific coded structured light pattern onto the surface of a metal eyeglass frame workpiece, and the multiple industrial cameras are configured to be synchronously triggered by the projector to acquire pattern-modulated workpiece surface images from different angles to obtain complete contour information for 3D reconstruction. High-resolution spectral confocal sensors emit a probe beam to scan the surface of a workpiece point by point. By analyzing the changes in the wavelength or spectral distribution of the returned beam, they capture the microscopic height fluctuations and material reflection characteristics of the surface, thereby identifying the spectral features of defects such as scratches and pits.

3. The multi-station adaptive surface treatment system for metal eyeglass frames according to claim 2, characterized in that, The data processing unit of the three-dimensional scanning and defect preliminary labeling module includes: The module receives multi-view images and spectral data transmitted by the data acquisition unit, performs multi-view point cloud registration and triangular mesh surface reconstruction, and constructs an accurate three-dimensional model of the eyeglass frame. The module also integrates local material hardness distribution information obtained by laser ranging or stylus detection to form a comprehensive three-dimensional digital model that includes geometric shape and physical properties. The comprehensive three-dimensional digital model is analyzed to quantify the depth and projected area of ​​each identified defect. Combined with the location of the defect in the lens frame, temple, or nose pad functional area, the urgency score for repair is automatically calculated based on a preset rule base. Finally, a defect repair priority map is generated, which is superimposed on the three-dimensional model and marked with different colors or markers to indicate the repair order.

4. The multi-station adaptive surface treatment system for metal eyeglass frames according to claim 1, characterized in that, The planning system of the multi-station adaptive execution module includes: The system analyzes the 3D model and defect map to calculate the collision-free optimal toolpath and process sequence for different workstations; and it also has a parameter matching unit that matches tool head model, spindle speed and feed rate from the process library based on the characteristics of different areas of the workpiece.

5. The multi-station adaptive surface treatment system for metal eyeglass frames according to claim 4, characterized in that, The execution system of the multi-station adaptive execution module includes: Import the 3D digital model and priority map with defect markings generated by the 3D scanning and defect initial marking module, and calculate the collision-free and time-optimal tool contact path and feed trajectory for each defect area to be processed based on the tool head workspace and kinematic model of each processing station. Generate a global, time-sequential collaborative station operation instruction sequence in a coordinated manner. Based on the characteristics of different areas of the workpiece automatically identified in the 3D model, and combined with the defect type and priority, the optimal combination of processing parameters is dynamically matched and output from the preset process knowledge base. The parameter combination includes at least the recommended tool head model, spindle rotation speed, tool feed rate, and polishing media supply flow rate.

6. The multi-station adaptive surface treatment system for metal eyeglass frames according to claim 1, characterized in that, The detection system of the online re-inspection and monitoring module includes: It is fixedly mounted on the end flange of the robotic arm that performs surface treatment, maintaining a defined relative position with the tool head; it adopts a coaxial optical path design, and automatically controls the lens to quickly focus on the current processing area during the interval of tool lifting or moving, and acquires high-resolution digital images for evaluating surface roughness and micromorphology. It consists of a thermocouple and a vibration accelerometer embedded inside the polishing or grinding tool head; the measuring end of the thermocouple is close to the working surface of the tool to directly monitor the instantaneous temperature change in the polishing area; the vibration accelerometer collects multi-dimensional vibration spectrum signals of the tool head in real time during the processing to analyze the tool status and the stability of the processing process.

7. The multi-station adaptive surface treatment system for metal eyeglass frames according to claim 6, characterized in that, The analysis system of the online re-inspection and monitoring module includes: The system receives high-definition images of the processing area transmitted by the detection system and performs high-precision registration and pixel-level comparison with the ideal surface morphology of the corresponding area in the original three-dimensional digital model generated by the three-dimensional scanning and defect initial labeling module. The system then identifies and quantifies the size, area, and precise three-dimensional coordinates of the residual defects in the area through an algorithm, and generates a quality deviation report containing the quantitative evaluation results. The system continuously receives and analyzes the polishing temperature and tool vibration spectrum signals collected by the detection system. By comparing the preset process safety threshold with the signal reference model under normal operating conditions, it can determine in real time whether there is an abnormal temperature rise or vibration instability in the processing process. Once an abnormality is detected, it immediately generates and sends an alarm signal or process parameter correction instruction containing a specific workstation identifier to the upper control system according to the predefined rule base.

8. The multi-station adaptive surface treatment system for metal eyeglass frames according to claim 1, characterized in that, The scheduling system of the intelligent collaborative central control module includes: Construct and maintain a virtual digital twin model that precisely corresponds to the physical processing system. This model receives real-time data on equipment status, workpiece position, and process from the physical system. It predicts subsequent processing flows and potential bottlenecks through parallel simulation and issues preventative workstation start / stop, cycle time adjustment, or path replanning instructions to the physical system based on the simulation results. The system continuously monitors the real-time equipment load rate, tool health status, job queue length, and quality data fed back by the online re-inspection and monitoring module for each physical workstation in the multi-station adaptive execution module. Based on preset objectives such as maximizing throughput, optimizing energy consumption, or prioritizing quality, the system uses a scheduling algorithm to dynamically calculate and assign the optimal flow sequence of the current and subsequent eyeglass frame workpieces in the system and the timing of their entry into each workstation for processing.

9. The multi-station adaptive surface treatment system for metal eyeglass frames according to claim 8, characterized in that, The optimization system of the intelligent collaborative central control module includes: The system receives real-time quality deviation reports generated by the online re-inspection and monitoring module, and retrieves a set of successful "workpiece-process-result" cases similar to the current eyeglass frame in terms of 3D features, defect type, and material from the historical process knowledge base for matching analysis. Through regression analysis or machine learning algorithms, the system iteratively adjusts and outputs an optimized set of process parameters for subsequent unprocessed workstations or subsequent similar workpieces of the current workpiece. The parameter set includes tool type, spindle speed, feed rate, and processing time. At the end of the single-piece eyeglass frame processing flow, the system automatically collects and integrates multi-source heterogeneous data from the entire process of the workpiece, from 3D scanning, defect marking, processing parameters of each station, online re-inspection report to the final quality inspection result, to form a structured process archive. Based on the evaluation of the processing results in this archive, the system automatically updates the case data and parameter recommendation rules in the historical process knowledge base, so as to realize the continuous accumulation and optimization of the system's process knowledge.

Images